BMJ Open Is Committed to Open Peer Review. As Part of This Commitment We Make the Peer Review History of Every Article We Publish Publicly Available
BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
BMJ Open is committed to open peer review. As part of this commitment we make the peer review history of every article we publish publicly available.
When an article is published we post the peer reviewers’ comments and the authors’ responses online. We also post the versions of the paper that were used during peer review. These are the versions that the peer review comments apply to.
The versions of the paper that follow are the versions that were submitted during the peer review process. They are not the versions of record or the final published versions. They should not be cited or distributed as the published version of this manuscript.
BMJ Open is an open access journal and the full, final, typeset and author-corrected version of record of the manuscript is available on our site with no access controls, subscription charges or pay-per-view fees (http://bmjopen.bmj.com).
If you have any questions on BMJ Open’s open peer review process please email [email protected] http://bmjopen.bmj.com/ on September 26, 2021 by guest. Protected copyright. BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Development and internal validation of prognostic models to predict negative health outcomes in older patients with multimorbidity and polypharmacy in general practice ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2020-039747
Article Type: Original research
Date Submitted by the 24-Apr-2020 Author:
Complete List of Authors: Müller, Beate; Goethe University Frankfurt, Institute of General Practice Uhlmann, Lorenz ; University of Heidelberg, Institute of Medical Biometry and Informatics Ihle, Peter; University of Cologne, PMV Research Group, Faculty of Medicine and University Hospital Cologne Stock, Christian; University of Heidelberg, Institute of Medical Biometry and Informatics von Buedingen, Fiona; Goethe University Frankfurt, Institute of General Practice Beyer, Martin; Goethe University Frankfurt, Institute of General Practice Gerlach, Ferdinand; Goethe University Frankfurt, Institute of General Practice
Perera, Rafael; University of Oxford, Nuffield Department of Primary http://bmjopen.bmj.com/ Care Health Sciences Valderas, Jose; University of Exeter Medical School, APEx Collaboration for Academic Primary Care Glasziou, Paul; Bond University, Centre for Research in Evidence-Based Practice, Faculty of Health Sciences and Medicine van den Akker, Marjan; Goethe University Frankfurt, Institute of General Practice; Maastricht University, Department of Family Medicine, School CAPHRI
Muth, Christiane; Goethe University Frankfurt, Institute of General on September 26, 2021 by guest. Protected copyright. Practice
PRIMARY CARE, THERAPEUTICS, GERIATRIC MEDICINE, HEALTH Keywords: SERVICES ADMINISTRATION & MANAGEMENT
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 48 BMJ Open
1 2
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 7 8 9 I, the Submitting Author has the right to grant and does grant on behalf of all authors of the Work (as defined 10 in the below author licence), an exclusive licence and/or a non-exclusive licence for contributions from authors 11 who are: i) UK Crown employees; ii) where BMJ has agreed a CC-BY licence shall apply, and/or iii) in accordance 12 with the terms applicable for US Federal Government officers or employees acting as part of their official 13 duties; on a worldwide, perpetual, irrevocable, royalty-free basis to BMJ Publishing Group Ltd (“BMJ”) its 14 licensees and where the relevant Journal is co-owned by BMJ to the co-owners of the Journal, to publish the 15 Work in this journal and any other BMJ products and to exploit all rights, as set out in our licence. 16 17 The Submitting Author accepts and understands that any supply made under these terms is made by BMJ to 18 the Submitting Author Forunless you peer are acting as review an employee on behalf only of your employer or a postgraduate 19 student of an affiliated institution which is paying any applicable article publishing charge (“APC”) for Open 20 Access articles. Where the Submitting Author wishes to make the Work available on an Open Access basis (and 21 intends to pay the relevant APC), the terms of reuse of such Open Access shall be governed by a Creative 22 Commons licence – details of these licences and which Creative Commons licence will apply to this Work are set 23 out in our licence referred to above. 24 25 Other than as permitted in any relevant BMJ Author’s Self Archiving Policies, I confirm this Work has not been 26 accepted for publication elsewhere, is not being considered for publication elsewhere and does not duplicate 27 material already published. I confirm all authors consent to publication of this Work and authorise the granting 28 of this licence. 29 30 31 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 48
1 2
3 1 TITLE BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 2 Development and internal validation of prognostic models to predict negative health 6 3 outcomes in older patients with multimorbidity and polypharmacy in general practice 7 4 8 9 5 Corresponding author 10 11 6 Beate S. Müller, Institute of General Practice, Goethe University, Frankfurt/Main, 12 7 Theodor-Stern-Kai 7, 60590 Frankfurt, Germany, [email protected] 13 8 frankfurt.de 14 15 9 16 10 Authors 17 11 1. Beate S. Müller, Institute of General Practice, Goethe University, Frankfurt/Main, 18 For peer review only 19 12 Germany 20 13 2. Lorenz Uhlmann, Institute of Medical Biometry and Informatics, University of 21 14 Heidelberg, Heidelberg, Germany 22 23 15 3. Peter Ihle, PMV Research Group, Faculty of Medicine and University Hospital 24 16 Cologne, University of Cologne, Köln, Germany 25 17 4. Christian Stock, Institute of Medical Biometry and Informatics, University of 26 27 18 Heidelberg, Heidelberg, Germany 28 19 5. Fiona v. Büdingen, Institute of General Practice, Goethe University, 29 20 Frankfurt/Main, Germany 30 31 21 6. Martin Beyer, Institute of General Practice, Goethe University, Frankfurt/Main, 32 22 Germany 33 23 7. Ferdinand M. Gerlach, Institute of General Practice, Goethe University, 34 35 24 Frankfurt/Main, Germany 36 25 8. Rafael Perera, Nuffield Department of Primary Care Health Sciences, University 37 26 of Oxford, Oxford, UK http://bmjopen.bmj.com/ 38 39 27 9. Jose M. Valderas, APEx Collaboration for Academic Primary Care, University of 40 28 Exeter Medical School, Exeter, UK 41 29 10.Paul P. Glasziou, Centre for Research in Evidence-Based Practice (CREBP), 42 43 30 Faculty of Health Sciences and Medicine, Bond University, Robina, Australia 44 31 11.Marjan van den Akker, (1) Institute of General Practice, Goethe University, 45 32 Frankfurt/Main, Germany, (2) Department of Family Medicine, School CAPHRI, on September 26, 2021 by guest. Protected copyright. 46 47 33 Maastricht University, Maastricht, Netherlands 48 34 12.Christiane Muth, Institute of General Practice, Goethe University, Frankfurt/Main, 49 35 Germany 50 51 36 52 37 Word count: 4,400 53 54 38 55 56 39 57 58 59 60
1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 48 BMJ Open
1 2
3 40 ABSTRACT BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 41 Background: Polypharmacy interventions are resource-intensive and should ideally 7 8 42 target those at risk of negative health outcomes. Our aim was to predict health- 9 10 43 related quality of life (HRQoL) and the combined outcome of falls, hospitalisation, 11 12 44 institutionalisation, and nursing care needs, in older patients with multimorbidity and 13 14 15 45 polypharmacy in general practices. 16 17 46 Methods: Design: two independent datasets, one comprising health insurance claims 18 For peer review only 19 47 data (N=592,456) and the other data from the PRIoritising MUltimedication in 20 21 22 48 Multimorbidity (PRIMUM) cluster-RCT (N=502). Population: ≥60 years, ≥5 drugs, ≥3 23 24 49 chronic diseases, excluding dementia. Outcomes: the combined outcome of falls, 25 26 50 hospitalisation, institutionalisation, and nursing care needs (after 6, 9 and 24 months) 27 28 (claims data); and HRQoL (after 6 and 9 months) (trial data). Predictor variables in 29 51 30 31 52 both datasets: age, sex, morbidity-related variables (e.g., disease count), medication- 32 33 53 related variables (e.g., EU-PIM), health service utilisation (e.g., number of involved 34 35 54 physicians). Predictor variables exclusively in trial data: additional socio- 36
37 http://bmjopen.bmj.com/ 38 55 demographics, morbidity-related variables (e.g., Cumulative Illness Rating Scale, 39 40 56 depression), Medication Appropriateness Index (MAI), lifestyle (e.g., smoking status), 41 42 57 functional status and HRQoL (EuroQol EQ5D-3L). Analysis: mixed regression 43 44
45 58 models, combined with stepwise variable selection, 10-fold cross validation and on September 26, 2021 by guest. Protected copyright. 46 47 59 sensitivity analyses. 48 49 60 Results: The most important predictors of EQ5D-3L at 6 months in the best model 50 51 (Nagelkerke’s R² 0.507) were depressive symptoms (-2.73 [95%CI: -3.56 to -1.91]), 52 61 53 54 62 MAI (-0.39 [-0.7 to -0.08]) and baseline EQ5D-3L (0.55 [95%CI: 0.47 to 0.64]). 55 56 63 Models based on claims data and those predicting long-term outcomes based on 57 58 64 both datasets produced low R² values. 59 60
2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 48
1 2
3 65 Conclusions: The best trial data-based model performed well to predict HRQoL after BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 66 6 months and included parameters of well-being not found in claims data. The 7 8 67 performance of claims data-based models and models predicting long-term outcomes 9 10 68 was relatively weak. To demonstrate generalisability, future studies should refit the 11 12 69 model by considering a variety of parameters representing well-being and functional 13 14 15 70 status. 16 17 71 18 For peer review only 19 72 STRENGHTS AND LIMITATIONS OF THIS STUDY 20 21 22 73 We developed our predictive models using two completely different datasets – 23 24 74 claims data and data primarily collected in a cluster-randomised trial. 25 26 75 The claims data contained a large number of cases, enabling our models to 27 28 29 76 include many possible predictors without any convergence issues. 30 31 77 The trial data provided a rich set of potential predictor variables of high data 32 33 78 quality and included data on patient-reported outcome measures, such as 34 35 36 79 well-being and functional status.
37 http://bmjopen.bmj.com/ 38 80 Both datasets have their own methodological limitations, such as imprecise 39 40 81 claims data (collected for reimbursement purposes) und the trial’s small 41 42 43 82 sample size. 44
45 83 The nature of the data meant neither dataset could be used to validate a on September 26, 2021 by guest. Protected copyright. 46 47 84 predictive model based on the other. 48 49 50 85 51 52 86 Keywords: Polypharmacy [MeSH], multimorbidity [MeSH], aged [MeSH], general 53 54 87 practice [MeSH], primary care [MeSH], prognostic model, clinical prediction [MeSH], 55 56 88 drug therapy [MeSH], chronic disease [MeSH] 57 58 59 60
3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 48 BMJ Open
1 2
3 89 BACKGROUND BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 90 Currently, up to 80% of primary care consultations involve patients with multiple 7 8 91 chronic conditions (multimorbidity).[1] A multiplicity of disorders in patients is 9 10 92 associated with polypharmacy. Both multimorbidity and polypharmacy are recognised 11 12 93 as a major challenge facing health care systems.[2–5] Polypharmacy can increase 13 14 15 94 the risk of mortality, hospitalisation,[6, 7] and falls and fall-related injuries with 16 17 95 resulting disability and loss of autonomy.[8, 9] It can also reduce cognitive and 18 For peer review only 19 96 physical function, as well as health-related quality of life (HRQoL).[2, 10] 20 21 22 97 The number of drugs increases the probability of adverse drug reactions (ADRs), but 23 24 98 the relationship is inconsistent, suggesting that the number of medications alone may 25 26 99 not adequately indicate the quality of an individual’s medication regimen.[11, 12] The 27 28 kind of drugs prescribed plays an important role in the type of reaction, with certain 29 100 30 31 101 medication classes, such as benzodiazepines, demonstrating a significant 32 33 102 association with falls, and medications with anti-cholinergic properties being 34 35 103 associated with impaired cognitive and physical function in elderly individuals.[13–16] 36
37 http://bmjopen.bmj.com/ 38 104 At a physician level, the cause of these negative health outcomes of polypharmacy 39 40 105 may be inappropriate prescribing, including undertreatment.[17–22] At a patient level, 41 42 106 a high number of drugs and the complexity of a drug regimen is often associated with 43 44
45 107 poor adherence,[23] which may be exacerbated by the presence of depression on September 26, 2021 by guest. Protected copyright. 46 47 108 and/or cognitive impairment.[24] Moreover, polypharmacy may also be the result of 48 49 109 an accumulation of potentially inappropriate medications (PIMs). 50 51 In the older primary care population, these issues, along with high inter-individual 52 110 53 54 111 variability in the pharmacologic response to drugs, the heterogeneity of underlying 55 56 112 disease patterns, and the severity of multimorbidity, make it difficult to identify 57 58 113 patients at risk of negative health outcomes from polypharmacy.[25–28] This is 59 60 114 particularly relevant when targeting patients that are at risk of negative health 4 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 48
1 2
3 115 outcomes and that would benefit most from resource-intensive interventions.[29–36] BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 116 The course of multimorbidity (and associated polypharmacy) has been characterised 7 8 117 by a decline in well-being (e.g., due to the progression of one or more chronic 9 10 118 diseases) interrupted by adverse events (e.g., exacerbations of chronic diseases or 11 12 119 adverse drug reactions).[37, 38] In order to identify the target population, it is 13 14 15 120 therefore necessary to predict a wide array of possible negative health outcomes. 16 17 121 Several prognostic models have been developed to predict mortality or unplanned 18 For peer review only 19 122 hospital (re-)admission, but only two studies exist that attempted to predict future 20 21 22 123 declines in quality of life, and no studies involved polypharmacy-related 23 24 124 predictors.[39] 25 26 125 The aim of this exploratory study was to develop and internally validate models to 27 28 predict hospitalisation, falls, level of required nursing care, institutionalisation and 29 126 30 31 127 health-related quality of life (HRQoL) in general practice patients with multimorbidity 32 33 128 and polypharmacy. The models were based on morbidity and medication-related 34 35 129 variables, as well as sociodemographic characteristics and parameters of health care 36
37 http://bmjopen.bmj.com/ 38 130 utilisation. 39 40 131 41 42 132 METHODS 43 44
45 133 We developed and internally validated two prognostic models to identify key health on September 26, 2021 by guest. Protected copyright. 46 47 134 problems linked with multimorbidity and associated polypharmacy (decline in well- 48 49 135 being and adverse events: Figure 1). (1) Based on claims data, we predicted the 50 51 combined endpoint of hospitalisation, falls / fall-related injuries, need for nursing care, 52 136 53 54 137 deterioration in the required level of care (nursing level), or institutionalisation, after 55 56 138 six, nine and 24 months. (2) We predicted health-related quality of life (HRQoL) after 57 58 139 six and nine months based on data from a cluster-randomised trial.[33] 59 60 140 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 48 BMJ Open
1 2
3 141 [About here: Figure 1: Predicted outcomes with regard to general trajectories of well- BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 142 being and quality of life in patients with multimorbidity and polypharmacy] 7 8 143 9 10 144 Design and Setting / Study Samples 11 12 145 Two datasets were used in modelling: 13 14 15 146 Claims data obtained from the Techniker Krankenkasse (TK) statutory health 16 17 147 insurance company between 01/2012 and 12/2014. TK is the largest statutory health 18 For peer review only 19 148 insurer in Germany and provided health insurance to 8.1 million persons in 2012.[40] 20 21 22 149 Trial data from the cluster-randomised PRIMUM (PRIoritising MUltimedication in 23 24 150 Multimorbidity) trial [33] conducted in general practices in Hesse, Germany from 25 26 151 08/2010 to 02/2012. 27 28 29 152 30 31 153 Population 32 33 154 Claims-based models: We aimed to use the same inclusion criteria for both datasets 34 35 155 as far as possible. We therefore included health insurance claims data of older 36
37 http://bmjopen.bmj.com/ 38 156 patients (≥60 years) with multimorbidity (at least three documented chronic diseases, 39 40 157 from a list of 46 diagnoses and conditions, from 01/01/2012 to 31/12/2012) [41] and 41 42 158 polypharmacy (at least five documented and concurrent prescriptions from 43 44
45 159 01/07/2012 to 31/12/2012). Included patients had to have been continuously insured on September 26, 2021 by guest. Protected copyright. 46 47 160 by TK from 01/01/2012 to 31/12/2014 (except in case of death at any time after 48 49 161 31/12/2012) and had to have contacted a primary care provider at least once in 2012. 50 51 Patients were excluded if they were diagnosed with dementia (ICD-10: F00-03, 52 162 53 54 163 F05.1, G30-31, R54) or under guardianship from 01/01/2012 to 31/12/2012. 55 56 164 57 58 165 Trial data-based models: We included data from patients that participated in the 59 60 166 cluster-randomised PRIMUM trial (N=503, intervention group: n=252, control group 6 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 48
1 2
3 167 n=251).[33] Patients with multimorbidity and polypharmacy were included in the study BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 168 if they were at least 60 years old, had at least three chronic diseases from two or 7 8 169 more chapters of ICD-10, and at least five prescriptions. Patients were excluded if 9 10 170 they were cognitively impaired (defined as a score lower or equal to 26 on the 11 12 171 MiniMental Status Exam [42]), had an alcohol or drug addiction, or were not able to 13 14 15 172 participate in telephone interviews, fill in questionnaires or express their own free will. 16 17 173 18 For peer review only 19 174 Outcomes 20 21 22 175 Models based on claims data: we predicted the combined endpoint of hospitalisation, 23 24 176 falls / fall-related injuries, or institutionalisation in a long-term care facility, or if the 25 26 177 need for nursing care was recognised, or the level of care (“Pflegestufe”) had 27 28 worsened at 6-, 9-, 24-month follow-up. Outcomes were operationalised as follows: 29 178 30 31 179 Hospitalisation: We included all-cause hospitalisations, as our data did not 32 33 180 permit us to differentiate between unplanned and elective hospitalisations. 34 35 36 181 Falls and fall-related injuries: We included all fractures and injuries coded in
37 http://bmjopen.bmj.com/ 38 182 ICD-10 chapters “S” and “T”. We excluded ICD codes for severe body injuries 39 40 183 such as S31 (“open wound of abdomen, lower back and pelvis”), which we 41 42 184 assessed as related to severe bodily impact, rather than drug-related falls (see 43 44
45 185 Additional file 1 for all excluded ICD codes). We also excluded osteoporosis- on September 26, 2021 by guest. Protected copyright. 46 47 186 related fractures (ICD-10 M80). 48 49 187 Institutionalisation was defined as the admission of a patient to a long-term 50 51 52 188 care facility for at least 28 days (in Germany, this is the maximum length of 53 54 189 time considered as ‘short-term care’ in such facilities). 55 56 190 Level of (nursing) care (“Pflegestufe”) referred to dependency on care. In the 57 58 59 191 period under review, the German nursing care insurance system recognised 60 192 four levels of care (“1” – lowest level to “3” – highest level, and “H”, which was 7 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 48 BMJ Open
1 2
3 193 mainly used for people with mental illnesses who are in need for support). The BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 194 onset of care and any increase in care level were taken into consideration. 7 8 195 Models based on trial data: we predicted HRQoL six and nine months after baseline. 9 10 196 HRQoL was measured using the EQ5D-3L index score.[43–46] The EQ5D-3L index 11 12 197 score [45] is a weighted summary score of five different dimensions of health 13 14 15 198 (mobility, self-care, usual activities, pain/discomfort, and anxiety/depression). Each 16 17 199 dimension has three levels. The index score is calculated based on TTO norm values 18 For peer review only 19 200 and ranges from 0 to 1, with “0” signifying death and “1” in full health.[46] 20 21 22 201 23 24 202 Potential predictors 25 26 203 Potential predictors in both modelling approaches are based on claims and trial data 27 28 (core predictors): To compare the two models, we first used a set of shared predictor 29 204 30 31 205 variables. 32 33 206 Socio-demographics: age, sex 34 35 36 207 Morbidity-related (excluding dementia): number of chronic diseases (based on
37 http://bmjopen.bmj.com/ 38 208 a modified list of 46 diagnoses and conditions), [41] Charlson comorbidity 39 40 209 index, [47] number of specific chronic conditions according to Diederichs’ list, 41 42 210 [48] consisting of 17 chronic diseases identified in a systematic review of 43 44
45 211 existing comorbidity indices. As dementia was excluded, the final list contained on September 26, 2021 by guest. Protected copyright. 46 47 212 16 diagnoses. 48 49 213 Medication: number of prescriptions (defined as Anatomical Therapeutic 50 51 52 214 Chemical (ATC) agents), excluding drugs for topical applications and drug 53 54 215 groups that were irrelevant to our research question, e.g. contrast agents 55 56 216 (ATC V-08). 57 58 59 217 Potentially Inappropriate Medication (PIM): we constructed two patient co- 60 218 variables: (1) exposure to any PIM (yes/no) and (2) number of PIMs between 8 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 48
1 2
3 219 01/07/2012 and 31/12/2012 (claims-based models) and at baseline (trial data- BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 220 based models). We used the following two lists to identify PIMs: 7 8 221 o Modified EU-PIM list [49]: The list of PIMs for the elderly contains 282 9 10 222 chemical substances or drug classes divided into 34 therapeutic 11 12 223 groups. 13 14 15 224 o Modified PRISCUS list [50]: The German list of PIMs for the elderly 16 17 225 includes 83 chemical substances from a total of 18 drug classes. 18 For peer review only 19 226 We excluded from the lists PIMs that referred to specific doses, treatment 20 21 22 227 duration, and disease severity, as valid information on these could not be 23 24 228 obtained from the claims data. 25 26 229 Anticholinergic drug burden: scores were calculated based on all prescribed 27 28 29 230 drugs with anticholinergic properties per patient. Despite substantial 30 31 231 differences between existing scales, associations with adverse clinical 32 33 232 outcomes, such as hospital admissions, fall-related hospitalisations, length of 34 35 stays in hospital, and GP visits, have been found for all of them.[16] 36 233
37 http://bmjopen.bmj.com/ 38 234 o Anticholinergic Drug Scale (ADS) [51, 52]: The ADS weights 39 40 235 anticholinergic properties per drug from ‘0’ – no anticholinergic activity, 41 42 236 ‘1’ – mild, ‘2’ – moderate to ‘3’ – strong anticholinergic activity. The 43 44
45 237 overall anticholinergic burden per patient was calculated as a sum on September 26, 2021 by guest. Protected copyright. 46 47 238 score for the entire medication regimen. 48 49 239 o Modified Anticholinergic Drug Burden Index (DBI) [13, 14, 53]: The DBI 50 51 52 240 comprises drugs with sedative effects (which form the sedative burden 53 54 241 (BS)), and drugs with anticholinergic or both sedative and 55 56 242 anticholinergic effects (which form the anticholinergic burden (BAC)). As 57 58 243 claims data do not provide dosages, the cumulative number of sedative 59 60 244 and anticholinergic drugs was calculated (modified DBI score). 9 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 48 BMJ Open
1 2
3 245 o Anticholinergic drug burden according to Duran et al.[54]: A systematic BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 246 review of assessment instruments and additional searches in the 7 8 247 Martindale® drug information database revealed N=100 drugs with 9 10 248 clinically relevant anticholinergic properties (n=47 high potency and 11 12 249 n=53 low potency) from which a 3-point (0 to 2) classification of 13 14 15 250 anticholinergic potency was derived. 16 17 251 Healthcare utilisation: For each patient, we obtained information on all-cause 18 For peer review only 19 252 hospitalisations (yes/no), falls and fall-related injuries (yes/no) and the 20 21 22 253 number of physicians involved in ambulatory health care, between 01/01/2012 23 24 254 and 31/12/2012 for models based on claims data, and in the 6 months 25 26 255 previous to baseline for models based on trial data. 27 28 29 256 30 31 257 Additional potential predictors used exclusively to re-fit models based on trial 32 33 258 data: 34 35 36 259 Socio-demographics: Education (Comparative Analysis of Social Mobility in
37 http://bmjopen.bmj.com/ 38 260 Industrial Nations, CASMIN [55]) and number of persons living in the 39 40 261 household 41 42 43 262 Lifestyle: Alcohol consumption (audit-C), [56] smoking status and body mass 44
45 263 index on September 26, 2021 by guest. Protected copyright. 46 47 264 Inappropriateness of medication: Medication Appropriateness Index (MAI) [57] 48 49 50 265 Morbidity-related: Severity of multimorbidity, as measured using the 51 52 266 Cumulative Illness Rating Scale (CIRS), [58] with scores calculated as the 53 54 267 total sum score, the number of affected organ systems, and the Health- 55 56 268 Related Quality of Life comorbidity index (HRQoL-CI) [59, 60] 57 58 59 269 Depressive symptoms, as measured using the Geriatric Depression Scale with 60 270 15 items (GDS) [61, 62] 10 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 48
1 2
3 271 HRQoL at baseline, as measured using the EQ5D-3L index score [43, 44] BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 272 7 8 273 Missing values and imputation 9 10 274 There were no missing values in the claims data, so no imputation was carried out in 11 12 models that were based on them. In models based on trial data, imputation of 13 275 14 15 276 missing values in predictors and outcomes was conducted using multiple imputation 16 17 277 via chained equations.[63, 64] We created 50 datasets and fitted the model for each 18 For peer review only 19 278 of them. We then combined estimates using 'Rubin’s rules'.[63] 20 21 22 279 23 24 280 Statistical analyses 25 26 281 In both models, we first investigated the core predictors that were available in both 27 28 29 282 datasets, including socio-demographics, morbidity- and medication-related variables, 30 31 283 and variables for healthcare utilisation. We then refitted the trial data-based models 32 33 284 using the additional predictors that were exclusively available for trial data, such as 34 35 variables for lifestyle and well-being. 36 285
37 http://bmjopen.bmj.com/ 38 286 39 40 287 Models based on claims data: In order to develop a prediction model for the binary 41 42 288 combined outcome (containing all-cause hospitalisation, falls / fall-related injuries, 43 44
45 289 institutionalisation or level of (nursing) care required) at 6-, 9- and 24-month follow- on September 26, 2021 by guest. Protected copyright. 46 47 290 up, we performed multiple logistic regression analyses with the occurrence of at least 48 49 291 one of the components at 6-, 9-, and 24-month follow-up as the dependent variable. 50 51 52 292 As patients were not always assigned a single general practice, we did not consider 53 54 293 cluster structures in the claims data. 55 56 294 Models based on trial data: In order to develop a prediction model for the continuous 57 58 295 outcome HRQoL at 6- and 9-month follow-up, we performed multiple linear 59 60 296 regression analyses using the EQ5D-3L index score at 6- and 9-month follow-up as 11 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 48 BMJ Open
1 2
3 297 the dependent variable. The cluster structure of the data was taken into account by BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 298 including a random intercept to produce a mixed regression model. We assumed a 7 8 299 compound symmetry structure when estimating the covariance matrix. 9 10 300 Univariate analyses in both claims and trial data: Prior to conducting regression 11 12 301 analyses, we performed univariate analyses to identify any associations between our 13 14 15 302 potential predictors (at baseline) and the outcomes (at 6-, 9-, and 24-month follow- 16 17 303 up). 18 For peer review only 19 304 Regression analyses and variable selection: To find out which predictor variables 20 21 22 305 influence the outcome variables, we used a stepwise variable selection procedure 23 24 306 (combining forward and backward steps). We started with the full model and all 25 26 307 potential predictor variables. After this, we used a selection procedure based on p- 27 28 values.[65] In the backward selection step, we deleted the variable with the highest 29 308 30 31 309 p-value from the model if its p value was greater than 0.157. In the forward selection 32 33 310 step, the variable with the lowest p-value was included in the model if its p-value was 34 35 311 less than 0.156. As long as each covariate had only one degree of freedom, the use 36
37 http://bmjopen.bmj.com/ 38 312 of these boundaries led to the same results as variable selection using the Akaike 39 40 313 Information Criterion (AIC).[66] The resulting models are presented by providing the 41 42 314 estimated regression coefficients (models based on trial data) or odds ratios (OR, 43 44
45 315 models based on claims) with 95% confidence intervals and corresponding p-values. on September 26, 2021 by guest. Protected copyright. 46 47 316 As we expected the large sample size of claims-based models to result in low p- 48 49 317 values, we calculated additional z values and continuous net reclassification indices 50 51 (NRI) to gain information on the predictive power of each variable.[67] 52 318 53 54 319 55 56 320 Performance of the models 57 58 321 We calculated R2 for linear models based on trial data (according to Nakagawa and 59 60 322 Schielzeth [68]), and Nagelkerke’s R2 for logistic models (according to Steyerberg 12 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 48
1 2
3 323 and Nagelkerke [69, 70]) based on claims data. Furthermore, in order to assess BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 324 performance more realistically and to internally validate the models, we used the 7 8 325 AUC (area under the receiver operator curve, equivalent to the concordance index) to 9 10 326 validate the logistic regression model based on claims data, and R2 to validate the 11 12 327 linear regression model based on RCT data, in combination with 10-fold cross- 13 14 2 2 15 328 validation. R and Nagelkerke’s R are measures of the overall model’s ability to 16 17 329 assess explained variance. The AUC provides a measure of the model’s 18 For peer review only 19 330 discriminatory ability to distinguish patients at risk from those that are not. 20 21 22 331 23 24 332 Sensitivity analyses 25 26 333 Using sensitivity analysis, we applied two further modelling approaches (at first 27 28 separately and then in combination): 1) modelling without multiple imputation and 2) 29 334 30 31 335 modelling without variable selection. 32 33 336 34 35 337 Software: We made use of different statistical packages to analyse the data in R.[71– 36
37 http://bmjopen.bmj.com/ 38 338 78] 39 40 339 We used TRIPOD reporting guidelines (transparent reporting of a multivariable 41 42 340 prediction model for individual prognosis or diagnosis) in the preparation of this 43 44
45 341 manuscript.[79] on September 26, 2021 by guest. Protected copyright. 46 47 342 48 49 343 Patient and Public Involvement statement 50 51 Neither patients nor the public were involved in this study. 52 344 53 54 345 55 56 346 RESULTS 57 58 347 Participants 59 60 348 Claims data 13 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 48 BMJ Open
1 2
3 349 The total sample of those ≥60 years that were continuously insured by TK from BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 350 01/01/2012 to 31/12/2014, and had at least one primary care contact during 2012, 7 8 351 amounted to 1,377,917 persons. Overall, 592,456 patients met the pre-specified 9 10 352 criteria and were included in the analyses (see study flow-chart, Figure 2). 11 12 353 Trial data 13 14 15 354 Of the 505 patients that participated in the PRIMUM trial,[32] all but 3 were 60 years 16 17 355 or older. The final analyses therefore included 502 patients. 18 For peer review only 19 356 Key characteristics of study participants are shown in table 1. 20 21 22 357 23 24 358 [About here: Figure 2: Flow chart of the study population in both datasets] 25 26 359 27 28 Table 1: Characteristics of study participants 29 360 30 31 Characteristic Claims data* CRT data* 32 N=592,456 N=502 33 Data collection period January 2012 to August 2010 to 34 December 2014# February 2012 35 36 Study design Cohort study Cluster-RCT
37 Setting Claims data from the TK 72 General Practices http://bmjopen.bmj.com/ 38 health insurance fund. TK in Hesse, Germany 39 serves about 10 Mio. 40 people in Germany 41 Inclusion criteria ≥60 years ≥60 years 42 43 ≥3 chronic diseases ≥3 chronic diseases 44 ≥5 prescriptions ≥5 prescriptions
45 ≥1 GP visit ≥1 GP visit on September 26, 2021 by guest. Protected copyright. 46 Continuously insured 47 (except in case of death in 48 follow-up period) 49 Exclusion criteria Person under legal Person under legal 50 51 guardianship guardianship 52 Diagnosed dementia Cognitive dysfunction 53 including dementia 54 (MMSE ≤ 26) 55 Outcomes to be predicted Combined† binary HRQoL (continuous 56 outcome after 6-, 9-, 24- outcome) after 6- and 57 58 month follow-up 9-month follow-up 59 Potential predictors in both samples at baseline** 60 Age (years) 71.3 (7.06) 72.2 (6.86)
14 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 48
1 2
3 Male sex (n, %) 319,453 (54) 240 (48) BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Morbidity: 5 6 . Disease count 9.7 (3.75) 9.6 (3.25) 7 . No. of specific chronic 4.3 (1.97) 4.1 (1.60) 8 diseases (Diederichs) 9 . CCI 3.0 (2.54) 2.6 (1.92) 10 . HRQoL-CI, mental 2.8 (2.12) 2.1 (1.81) 11 . HRQoL-CI, physical 8.0 (3.57) 7.6 (3.12) 12 13 Medication: 14 . No. of drugs 8.6 (3.80) 8.1 (2.57) 15 . No. of PIM (EU-PIM) 1.1 (1.15) 0.9 (0.96) 16 . ACh burden (ADS) 1.0 (1.45) 0.8 (1.21) 17 . Mod. Drug Burden Index 0.8 (1.03) 0.5 (0.77) 18 No. of involved physiciansFor peer9.95 review (5.26) only2.6 (1.77) 19 20 Previous Hospitalisation: ‡ 21 . Patients that have 194,984 (33) 81 (16) 22 undergone hospital 23 treatment(n, %) 24 . No. of hospitalisations 1.67 (1.25) 1.5 (0.86) ‡ 25 . No. of days in hospital 14.5 (18.20) 17 (12.66) ‡ 26 Patients with previous 163,387 (28) 83 (17) ‡ 27 28 falls/fall-related injuries (n, %) 29 Patients requiring nursing 30 care: 31 . Any nursing level (n, %) 28,310 (5) - 32 . Nursing level 1 (n, %) 19,030 (3) - 33 . Nursing level 2 (n, %) 7,968 (1) - 34 35 . Nursing level 3 (n, %) 1,273 (0.2) - 36 . Nursing level H (n, %) 39 (0.007) -
37 Additional predictor variables in CRT data at baseline** http://bmjopen.bmj.com/ 38 Socio-demographics 39 . Educational level - 1.4 (0.66) 40 (CASMIN) 41 42 . No. of persons living in - 1.8 (0.70) 43 household 44 Lifestyle
45 . Alcohol intake (AUDIT C) - 1.9 (1.96) [mv: 39] on September 26, 2021 by guest. Protected copyright. 46 . Smoker (n, %) - 46 (10) [mv: 25] 47 . Body Mass Index - 30.1 (6.58) 48 49 Morbidity: 50 . CIRS sum score 7.7 (4.56) 51 . CIRS, no. of organ 4.5 (2.35) 52 systems 53 . Depressive Symptoms 2.4 (2.29) [mv: 8] 54 (GDS) 55 Medication: 56 57 . MAI - 4.7 (5.56) 58 HRQoL: 59 . EQ5D-3L Index Score - 74.3 (23.72) [mv: 24] 60 361
15 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 48 BMJ Open
1 2
3 362 Legend: *Values are arithmetic means and standard deviations unless otherwise BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 # 6 363 indicated. The anamnestic period for baseline data ran from 01/01/2012 to 7 8 364 31/12/2012, except for medication data, for which it ran from 01/07/2012 to 9 10 365 31/12/2012. The follow-up period started on 01/01/2013. †Combined outcome 11 12 366 included hospitalization, fall/fall-related injuries, institutionalization and care level. ‡6 13 14 15 367 months before study entry. **Number of patients with missing values (mv) is zero 16 17 368 unless indicated in square parentheses. 18 For peer review only 19 369 Abbreviations: ACh burden – Anticholinergic drug burden, ADS – Anticholinergic 20 21 22 370 Drug Scale, AUDIT - Alcohol Use Disorders Identification Test (WHO), CASMIN - 23 24 371 Comparative Analysis of Social Mobility in Industrial Nations, CCI - Charlson 25 26 372 Comorbidity Index, CIRS – Cumulative Illness Rating Scale, GDS – Geriatric 27 28 Depression Scale, GP – General Practitioner, HRQoL – Health-Related Quality of 29 373 30 31 374 Life, HRQoL-CI – Health-Related Quality of Life Comorbidity Index, MAI – Medication 32 33 375 Appropriateness Index, MMSE – Mini Mental Status Exam, PIM – Potentially 34 35 376 Inappropriate Medication, CRT – cluster-randomised controlled trial. 36
37 http://bmjopen.bmj.com/ 38 39 377 40 41 378 Univariate Analyses 42 43 379 In the claims data, univariate analyses revealed significant associations between the 44
45 on September 26, 2021 by guest. Protected copyright. 380 combined outcome and the following predictors: age, sex, disease count, Charlson 46 47 48 381 comorbidity index, EU-PIMs, ADS, DBI, previous hospitalisations, previous falls and 49 50 382 number of physicians involved in the patient’s care at all follow-ups (after six, nine 51 52 383 and 24 months) (Additional file 2). In the trial data, HRQoL was significantly 53 54 55 384 correlated with the shared predictor variables disease count, number of chronic 56 57 385 prescriptions, previous falls and sex, and the additional predictors depression and 58 59 386 HRQoL at baseline (Additional file 3). 60
16 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 48
1 2
3 387 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 388 Prediction models 7 8 389 Claims data 9 10 390 The model predicting the combined endpoint at 6 months performed best (AUC with 11 12 391 10-fold cross validation: 0.71, see table 2), and had a low explanation of variance 13 14 2 15 392 (Nagelkerke’s R without cross validation: 0.16). Variables in the model with the 16 17 393 highest predictive power were previous falls/fall-related injuries and previous 18 For peer review only 19 394 hospitalisations, as well as age, number of involved physicians, and number of 20 21 22 395 chronic diseases (“disease count”) (see Additional file 4). The models predicting the 23 24 396 combined outcome at nine and 24 months had AUCs calculated with 10-fold cross 25 26 397 validation of 0.68 (R² without cross validation: 0.15) and 0.69 (R² without cross 27 28 validation: 0.13) respectively. 29 398 30 31 399 32 33 400 Trial data 34 35 401 All results presented in this section are based on the modelling approach and involve 36
37 http://bmjopen.bmj.com/ 38 402 multiple imputation of missing values and the variable selection procedure. Models 39 40 403 predicting the HRQoL endpoint at 6 months that were based on predictors available 41 42 404 in both claims and trial data showed low predictive accuracy (R2 with 10-fold cross 43 44
45 405 validation: 0.111) (see Additional file 5, model 2.4). HRQoL at 6 months was best on September 26, 2021 by guest. Protected copyright. 46 47 406 predicted when predictors that were exclusively available in the trial data were also 48 49 407 included (R2 with 10-fold cross validation: 0.507). The variables with the highest 50 51 predictive power were depressive symptoms (GDS) and EQ5D-3L Index Score 52 408 53 54 409 (Baseline). MAI was also predictive (see Additional file 5, model 3.4). 55 56 410 57 58 411 Comparison of model quality and sensitivity analyses 59 60
17 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 48 BMJ Open
1 2
3 412 The shorter the predicted time span, the better model quality was. However, model BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 413 quality stayed low when it only included predictor variables that were available for 7 8 414 both claims and trial data. Sensitivity analyses confirmed the results (see table 2). 9 10 415 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
18 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
BMJ Open Page 20 of 48
1 2 3 Table 2: Comparison of models 4 5 2 6 Models based on claims data AUC* R 7 1.1 Combined outcome after 6 months 0.71 {0.70} 0.16 {0.16} 8 1.2 Combined outcome after 9 months 0.69 {0.69} 0.15 {0.14} 9 10 1.3 Combined outcome after 24 months 0.68 {0.68} 0.13 {0.12} 11 Models based on CRT data: core predictors# AIC R2 R2 (10x) 12 EQ5D-3L after 6 months For peer review only 13 14 2.1 No imputation, no variable selection 4,138.86 {4,069.41} 0.155 {0.159} 0.112 {0.103} 15 2.2 No imputation, with variable selection 4,138.81 {4,068.69} 0.150 {0.155} 0.129 {0.122} 16 2.3 With imputation, no variable selection 4,582.30 {4,507.71} 0.159http://bmjopen.bmj.com/ {0.163} 0.094 {0.108} 17 2.4 With imputation, with variable selection 4,583.15 {4,507.47} 0.919 {0.925} 0.111 {0.128} 18 19 EQ5D-3L after 9 months 20 2.5 No imputation, no variable selection 3,917.75 {3,917.75} 0.150 {0.150} 0.030 {0.030} 21 2.6 No imputation, with variable selection 3,921.95 {3,921.95} 0.146 {0.146} 0.053 {0.053} 22 23 2.7 With imputation, no variable selection 4,540.58 {4,505.52} 0.156 {0.152} 0.090 {0.093} 24 2.8 With imputation, with variable selection 4,546.42 {4,511.10} 0.221 {0.218} 0.107 {0.106} on September 26, 2021 by guest. Protected copyright. 25 Models based on CRT data: additional predictors# 26 EQ-5D, after 6 months 27 28 3.1 No imputation, no variable selection 3,205.13 {3,205.13} 0.034 {0.034} 0.442 {0.442} 29 3.2 With imputation, no variable selection 4,308.94 {4,308.94} 0.538 {0.538} 0.481 {0.481} 30 3.3 No imputation, with variable selection 3,197.37 {3,197.37} 0.526 {0.526} 0.483 {0.483} 31 3.4 With imputation, with variable selection† 4,307.47 {4,307.47} 0.677 {0.677} 0.507 {0.507} 32 33 Models with "fixed variables" 34 3.5 No imputation, no variable selection 3,208.58 {3,208.58} 0.514 {0.514} 0.468 {0.468} 35 3.6 With imputation, with variable selection 4,308.90 {4,308.90} 0.665 {0.665} 0.499 {0.499} 36 EQ-5D, after 9 months 37 38 3.7 No imputation, no variable selection 3,061.06 {3,113.53} 0.042 {0.028} 0.411 {0.409} 39 3.8 With imputation, no variable selection 4,307.28 {4,361.36} 0.498 {0.477} 0.433 {0.404} 40 41 42 19 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Page 21 of 48 BMJ Open
1 2 3 3.9 No imputation, with variable selection 3,062.03 {3,108.61} 0.490 {0.485} 0.448 {0.443} 4 3.10 With imputation, with variable selection 4,309.88 {4,360.32} 0.453 {0.346} 0.455 {0.431} 5 6 Models with fixed variables 7 3.11 No imputation, no variable selection 3,064.76 {3,113.08} 0.490 {0.485} 0.439 {0.434} 8 3.12 With imputation, with variable selection 4,310.92 {4,363.62} 0.113 {0.071} 0.447 {0.423} 9 10 11 # 12 Legend: {sensitivity analyses}, ModelsFor based peeron RCT data: fixedreview effects, †Best Model only; Abbreviations: AUC* - Area under the curve 13 14 after 10-fold cross validation, AIC – Aikaike’s information criterium, R² - Nagelkerke’s R², R2 (10x) – Nagelkerke’s R² with 10-fold cross 15 16 validation http://bmjopen.bmj.com/ 17 18 19 20 21 22 23 24 on September 26, 2021 by guest. Protected copyright. 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 20 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 22 of 48
1 2
3 DISCUSSION BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 Main results 7 8 On the basis of claims data, the occurrence of an adverse health-related event (all- 9 10 cause hospitalisation, or falls / fall-related injuries, or institutionalisation, or receipt of 11 12 nursing care / worsening of the required level of care (nursing level)) could not be 13 14 15 predicted satisfactorily using socio-demographic, disease- or medication-related 16 17 variables. As the number of cases was large, all predictor variables reached a high 18 For peer review only 19 level of significance, and AUC-values after 10-fold validation were also moderately 20 21 22 high (max. 0.71). Nevertheless, the explanation of variance remained low (max. 23 24 Pseudo-Nagelkerke R²: 0.16). The variables with the highest predictive power were 25 26 previous falls/fall-related injuries and previous hospitalisations, as well as age, 27 28 number of involved physicians, and number of chronic diseases. 29 30 31 Based on trial data, health-related quality of life was poorly predicted by variables 32 33 available for both datasets (R² max. 0.13). However, adding variables that were only 34 35 available for trial data improved the explanation of variance (best model: R2 with 10- 36
37 http://bmjopen.bmj.com/ 38 fold cross validation 0.51). Depressive symptoms and the initial level of health-related 39 40 quality of life had relatively high predictive power, while medication appropriateness 41 42 (MAI) also had an impact. 43 44
45 Regardless of modelling approach, quality improved with shorter forecast periods. on September 26, 2021 by guest. Protected copyright. 46 47 This is unsurprising, as adverse health-related events often occur shortly after receipt 48 49 of a prescription. 50 51 When analysing both data sources, endpoint components had a relatively high 52 53 54 predictive power (claims: previous falls and hospitalisations; trial data: initial level of 55 56 health-related quality of life). 57 58 59 60 Comparison with the literature 21 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 23 of 48 BMJ Open
1 2
3 The presented results are consistent with results from other studies. The AUC values BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 in our claims-based models (AUC 0.68-0.71) are comparable to those of 23 7 8 prognostic models for Case Finding conducted in elderly patients in primary care. 9 10 These models predicted (re)hospitalisation, functional impairment, institutionalisation 11 12 and death.[80] The quality of models with a low risk of bias was AUC 0.60-0.78, but 13 14 15 no explanation of variance was provided. The best model for predicting death within 16 17 four years (AUC: 0.82) included 12 predictors comprising age, sex, BMI, chronic 18 For peer review only 19 diseases, smoking status and functional parameters.[80] Models that included 20 21 22 additional trial data (e.g. clinical data) predicted endpoints better than models based 23 24 only on claims data.[80–82] In many models described in other studies, health care 25 26 utilisation parameters, and especially previous hospitalisations, were predictive of 27 28 (re)hospitalisations, emergency admissions and functional impairment.[81, 83, 84] 29 30 31 The predictive power of sex is inconsistent: in 18/27 risk models, sex was included in 32 33 the final model [81]; in 7/23 risk models, male sex was predictive [80], while a further 34 35 25 studies found sex to have no influence.[83, 84] Model quality also improved in 36
37 http://bmjopen.bmj.com/ 38 studies that included multimorbidity and polypharmacy parameters.[81, 83, 85] 39 40 However, the parameters and instruments used in modelling (e.g. CIRS, Charlson 41 42 Comorbidity Index and disease count, as reported here) varied considerably among 43 44
45 studies. They were neither consistently predictive, nor were certain parameters or on September 26, 2021 by guest. Protected copyright. 46 47 instruments better than others.[81, 84, 85] 48 49 Most published models were developed to predict the risk of hospitalisation.[81, 83– 50 51 89] Other models predicted functional outcomes,[85] while four models predicted 52 53 54 adverse drug reactions.[89] So far, little is known about the predictive power of 55 56 polypharmacy parameters and the appropriateness of prescriptions, especially the MAI 57 58 has never been used in prognostic models. Furthermore, no models have yet been 59 60 developed to predict health-related quality of life in patients with multimorbidity.[39, 85] 22 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 24 of 48
1 2
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 Relevance for primary care 7 8 This exploratory study identified predictors that could be used as warning parameters 9 10 ('red flags') to support general practitioners in recognising older patients with 11 12 multimorbidity and polypharmacy that are at risk of adverse health-related events and 13 14 15 that may therefore benefit from a medication review: 16 17 . Patients with reduced health-related quality of life and/or depressive 18 For peer review only 19 symptoms are at risk of a decline in their health-related quality of life in the 20 21 22 subsequent six months. Depending on the options available for collecting such 23 24 information from patients, clinicians may want to consider the feasibility of 25 26 single item HRQoL assessment instruments and two item screening tests for 27 28 depression, as compared with the more robust psychometric properties of EQ- 29 30 31 5D and GDS.[90–93] 32 33 . Patients with previous falls and hospitalisation are at risk of further falls, 34 35 (re)hospitalisation and institutionalisation in the subsequent six months. This 36
37 http://bmjopen.bmj.com/ 38 information is available in discharge letters and electronic health records. 39 40 41 42 Strengths and Limitations 43 44
45 One strength of our study is that we could use two data sources with differing on September 26, 2021 by guest. Protected copyright. 46 47 advantages in our exploratory analysis: claims data contained a large number of 48 49 cases with a high event rate of "hard" endpoints, and trial data provided additional 50 51 high-quality patient data. Both datasets also have their limitations, since claims are 52 53 54 documented for billing purposes and are therefore imprecise, whereas our trial 55 56 dataset consisted of only a limited number of observations. Thus, each dataset 57 58 allows its own endpoints to be modelled. Risk modelling is especially complex in 59 60 multimorbid patients with polypharmacy, as predictor variables in this patient 23 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 25 of 48 BMJ Open
1 2
3 collective are often associated with one another (e.g. diagnoses and prescriptions). In BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 addition, comparable risk situations can lead to different endpoints, as risk often 7 8 depends on context. For example, a drug-induced fall may have no health-related 9 10 consequences or may lead to impairment and institutionalisation. We plan to conduct 11 12 an individual patient-data meta-analysis to further develop and externally validate the 13 14 15 models presented here (PROSPERO ID: CRD42018088129). 16 17 18 For peer review only 19 Conclusions 20 21 22 This study provided models of risk that explore the impact of multimorbidity and 23 24 polypharmacy in an older population. Previous falls, hospital stays, reduced health- 25 26 related quality of life, and depression, were strong predictors of negative health 27 28 outcomes in our models. Future studies should independently test their predictive 29 30 31 power as 'red flags', as this could help family practitioners identify the high-risk 32 33 general practice patients that would most benefit from comprehensive medication 34 35 reviews. 36
37 http://bmjopen.bmj.com/ 38 39 40 List of abbreviations 41 42 ACh burden: Anticholinergic drug burden, ADS: Anticholinergic Drug Scale, AIC: 43 44
45 Aikaike‘s information criterium, AUC*: Area under the curve after 10-fold cross on September 26, 2021 by guest. Protected copyright. 46 47 validation, AUDIT: Alcohol Use Disorders Identification Test (WHO), CASMIN: 48 49 Comparative Analysis of Social Mobility in Industrial Nations, CCI: Charlson 50 51 Comorbidity Index, CIRS: Cumulative Illness Rating Scale, GDS: Geriatric 52 53 54 Depression Scale, GP: General Practitioner, HRQoL: Health-Related Quality of Life, 55 56 HRQoL-CI: Health-Related Quality of Life Comorbidity Index, MAI: Medication 57 58 Appropriateness Index, MMSE: Mini Mental Status Exam, NRI: Continous Net 59 60 Reclassification Index, OR: Odds Ratio, PIM: Potential Inappropriate Medication, 24 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 48
1 2 2 (10x) 3 PRIMUM: Prioritising Multimedication in Multimorbidity, R²: Nagelkerke‘s R², R : BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 Nagelkerke’s R² with 10-fold cross validation, CRT: cluster-randomised controlled 7 8 trial, SD: Standard Deviation, TK: Techniker Krankenkasse, TRIPOD: Transparent 9 10 reporting of a multivariable prediction model for individual prognosis or diagnosis 11 12 13 14 15 16 17 Declarations 18 For peer review only 19 Ethics approval and consent to participate 20 21 Claims may be analysed in accordance with § 284 of Social Code Book V. When 22 23 24 claims are anonymously analysed in accordance with Good Practice in Claims Data 25 26 Analysis, [94] no further ethics vote is required. 27 28 Regarding our trial data, the ethics commission of the medical faculty of the Johann 29 30 31 Wolfgang Goethe University, Frankfurt / Main approved the PRIMUM trial (resolution 32 33 number E 46/10, file number 123/10, date: 20/05/2010) and all of the participants 34 35 gave their written informed consent before taking part. 36
37 http://bmjopen.bmj.com/ 38 39 40 Consent for publication 41 42 Not applicable. 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 Availability of data and material 48 49 The datasets generated and analysed in the current study are not publicly available, 50 51 as further analyses are ongoing. 52 53 54 55 56 Competing interests 57 58 FG, BM, MB and CM received grants from the German Statutory Healthcare 59 60 Insurance Company Techniker Krankenkasse during the course of the study. CS has 25 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 48 BMJ Open
1 2
3 been employed by Boehringer Ingelheim Pharma GmbH & Co. KG since October BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 2019. The company had no role in the design, analysis or interpretation of the current 7 8 study. MvA, LU, PI, FvB, RP, PG and JV declare that they have no competing 9 10 interests. 11 12 13 14 15 Funding 16 17 This study was supported by the German Statutory Healthcare Insurance Company 18 For peer review only 19 Techniker Krankenkasse. 20 21 22 23 24 Authors' contributions 25 26 MB, FG and CM designed the study. PI, LU and CS analysed the data. All authors 27 28 contributed to the interpretation of the data. CM and BM drafted the manuscript and 29 30 31 all authors revised it critically for important intellectual content. All authors approved 32 33 the version to be submitted for publication. LU, CS and CM had full access to all data 34 35 and are responsible for the integrity and the accuracy of the data analysis. 36
37 http://bmjopen.bmj.com/ 38 39 40 Acknowledgements 41 42 The authors would like to thank Phillip Elliott for the language review of the paper. 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
26 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 48
1 2
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 REFERENCES 5 1 Salisbury C, Johnson L, Purdy S, et al. Epidemiology and impact of multimorbidity in primary 6 care: A retrospective cohort study. Br J Gen Pract 2011;61(582):e12-21. 7 8 2 Moßhammer D, Haumann H, Mörike K, et al. Polypharmacy-an Upward Trend with 9 Unpredictable Effects. Dtsch Arztebl Int 2016;113(38):627–33. 10 3 Muth C, Blom JW, Smith SM, et al. Evidence supporting the best clinical management of 11 patients with multimorbidity and polypharmacy: A systematic guideline review and expert 12 13 consensus. J Intern Med 2019;285(3):272–88. 14 4 Palmer K, Marengoni A, Forjaz MJ, et al. Multimorbidity care model: Recommendations from 15 the consensus meeting of the Joint Action on Chronic Diseases and Promoting Healthy Ageing 16 17 across the Life Cycle (JA-CHRODIS). Health Policy 2018;122(1):4–11. 18 5 Nobili A, MarengoniFor A, Tettamanti peer M, etreview al. Association between only clusters of diseases and 19 polypharmacy in hospitalized elderly patients: results from the REPOSI study. Eur J Intern Med 20 2011;22(6):597–602. 21 22 6 Oscanoa TJ, Lizaraso F, Carvajal A. Hospital admissions due to adverse drug reactions in the 23 elderly. A meta-analysis. Eur J Clin Pharmacol 2017;73(6):759–70. 24 7 Angamo MT, Chalmers L, Curtain CM, et al. Adverse-Drug-Reaction-Related Hospitalisations in 25 26 Developed and Developing Countries: A Review of Prevalence and Contributing Factors. Drug 27 Saf 2016;39(9):847–57. 28 8 Deandrea S, Lucenteforte E, Bravi F, et al. Risk factors for falls in community-dwelling older 29 people: a systematic review and meta-analysis. Epidemiology 2010;21(5):658–68. 30 31 9 Heinrich S, Rapp K, Rissmann U, et al. Cost of falls in old age: a systematic review. Osteoporos 32 Int 2010;21(6):891–902. 33 10 Fried TR, O'Leary J, Towle V, et al. Health outcomes associated with polypharmacy in 34 35 community-dwelling older adults: a systematic review. J Am Geriatr Soc 2014;62(12):2261–72. 36 11 Payne RA, Abel GA, Avery AJ, et al. Is polypharmacy always hazardous? A retrospective cohort 37 analysis using linked electronic health records from primary and secondary care. Br J Clin http://bmjopen.bmj.com/ 38 Pharmacol 2014;77(6):1073–82. 39 40 12 Gnjidic D, Hilmer SN, Blyth FM, et al. Polypharmacy cutoff and outcomes: five or more 41 medicines were used to identify community-dwelling older men at risk of different adverse 42 outcomes. J Clin Epidemiol 2012;65(9):989–95. 43 44 13 Hilmer SN, Mager DE, Simonsick EM, et al. A drug burden index to define the functional burden
45 of medications in older people. Arch Intern Med 2007;167(8):781–87. on September 26, 2021 by guest. Protected copyright. 46 14 Hilmer SN, Mager DE, Simonsick EM, et al. Drug burden index score and functional decline in 47 older people. Am J Med 2009;122(12):1142-1149.e1-2. 48 49 15 Salahudeen MS, Duffull SB, Nishtala PS. Anticholinergic burden quantified by anticholinergic 50 risk scales and adverse outcomes in older people: a systematic review. BMC Geriatr 2015;15:31. 51 16 Salahudeen MS, Hilmer SN, Nishtala PS. Comparison of anticholinergic risk scales and 52 53 associations with adverse health outcomes in older people. J Am Geriatr Soc 2015;63(1):85–90. 54 17 Gurwitz JH, Field TS, Harrold LR, et al. Incidence and preventability of adverse drug events 55 among older persons in the ambulatory setting. JAMA 2003;289(9):1107–16. 56 18 Kuijpers MAJ, van Marum RJ, Egberts ACG, et al. Relationship between polypharmacy and 57 58 underprescribing. Br J Clin Pharmacol 2008;65(1):130–33. 59 19 Steinman MA, Landefeld CS, Rosenthal GE, et al. Polypharmacy and prescribing quality in older 60 people. J Am Geriatr Soc 2006;54(10):1516–23.
27 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 48 BMJ Open
1 2
3 20 Meid AD, Quinzler R, Freigofas J, et al. Medication Underuse in Aging Outpatients with BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Cardiovascular Disease: Prevalence, Determinants, and Outcomes in a Prospective Cohort 5 6 Study. PLoS One 2015;10(8):e0136339. 7 21 Meid AD, Quinzler R, Groll A, et al. Longitudinal evaluation of medication underuse in older 8 outpatients and its association with quality of life. Eur J Clin Pharmacol 2016;72(7):877–85. 9 10 22 Meid AD, Haefeli WE. Age-Dependent Impact of Medication Underuse and Strategies for 11 Improvement. Gerontology 2016;62(5):491–99. 12 23 Müller BS, Uhl MC, Nguyen Truc Sophia, et al. Patienten mit Multimedikation: Ambulante 13 Herausforderungen und Lösungswege: Eine qualitative Studie. Zeitschrift für Allgemeinmedizin 14 15 2018;94(10):396–400. 16 24 Horne R, Weinman J, Barber N, Elliott R, Morgan M, Cribb A, Kellar I. Concordance, adherence 17 and compliance in medicine taking: Report for the National Co-ordinating Centre for NHS 18 For peer review only 19 Service Delivery and Organisation R & D (NCCSDO) 2005. 20 25 Motter FR, Fritzen JS, Hilmer SN, et al. Potentially inappropriate medication in the elderly: A 21 systematic review of validated explicit criteria. Eur J Clin Pharmacol 2018;74(6):679–700. 22 26 Corsonello A, Pedone C, Incalzi RA. Age-related pharmacokinetic and pharmacodynamic 23 24 changes and related risk of adverse drug reactions. Curr Med Chem 2010;17(6):571–84. 25 27 Chalk D, Manzi S, Britten N, et al. Can agent-based simulation be used as a tool to support 26 polypharmacy prescribing practice? BMJ STEL 2017;3(3):94–98. 27 28 28 Ble A, Masoli JAH, Barry HE, et al. Any versus long-term prescribing of high risk medications in 29 older people using 2012 Beers Criteria: Results from three cross-sectional samples of primary 30 care records for 2003/4, 2007/8 and 2011/12. BMC Geriatr 2015;15(1):1–10. 31 29 Muth C, Harder S, Uhlmann L, et al. Pilot study to test the feasibility of a trial design and 32 33 complex intervention on PRI oritising MU ltimedication in M ultimorbidity in general practices 34 (PRIMUM pilot). BMJ Open 2016;6(7):e011613. 35 30 Johansson T, Abuzahra ME, Keller S, et al. Impact of strategies to reduce polypharmacy on 36
37 clinically relevant endpoints: A systematic review and meta-analysis. Br J Clin Pharmacol http://bmjopen.bmj.com/ 38 2016;82(2):532–48. 39 31 Rankin A, Cadogan CA, Patterson SM, et al. Interventions to improve the appropriate use of 40 polypharmacy for older people. Cochrane Database Syst Rev 2018;9:CD008165. 41 42 32 Willeboordse F, Schellevis FG, Chau SH, et al. The effectiveness of optimised clinical medication 43 reviews for geriatric patients: Opti-Med a cluster randomised controlled trial. Fam Pract 44 2017;34(4):437–45. 45 on September 26, 2021 by guest. Protected copyright. 46 33 Muth C, Uhlmann L, Haefeli WE, et al. Effectiveness of a complex intervention on Prioritising 47 Multimedication in Multimorbidity (PRIMUM) in primary care: Results of a pragmatic cluster 48 randomised controlled trial. BMJ Open 2018;8(2):e017740. 49 34 Blom J, den Elzen W, van Houwelingen AH, et al. Effectiveness and cost-effectiveness of a 50 51 proactive, goal-oriented, integrated care model in general practice for older people. A cluster 52 randomised controlled trial: Integrated Systematic Care for older People--the ISCOPE study. 53 Age Ageing 2016;45(1):30–41. 54 55 35 Bosch-Lenders D, Maessen DWHA, Stoffers HEJHJ, et al. Factors associated with appropriate 56 knowledge of the indications for prescribed drugs among community-dwelling older patients 57 with polypharmacy. Age Ageing 2016;45(3):402–08. 58 59 36 Müller CA, Wilm S, Thürmann PA, et al. Die RIME Studie – eine clusterrandomisierte 60 kontrollierte Studie zur Reduktion von potentiell inadäquater Medikation in der Hausarztpraxis – Studienprotokoll. German Medical Science GMS Publishing House. Salzburg 2011. 28 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 48
1 2
3 37 Murray SA, Kendall M, Mitchell G, et al. Palliative care from diagnosis to death. BMJ BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 2017;356:j878. 5 6 38 Lynn J. Living long in fragile health: the new demographics shape end of life care. Hastings Cent 7 Rep 2005;Spec No:S14-8. 8 39 National Institute for Health and Care Excellence. Multimorbidity: clinical assessment and 9 10 management: NICE guideline NG 56 2016. 11 40 Bundesministerium für Gesundheit. KM 6-Statistik 2013. Available at: 12 https://www.bundesgesundheitsministerium.de/themen/krankenversicherung/zahlen-und- 13 fakten-zur-krankenversicherung/mitglieder-und-versicherte.html Accessed April 23, 2020. 14 15 41 Schäfer I, Leitner E-C von, Schön G, et al. Multimorbidity patterns in the elderly: A new 16 approach of disease clustering identifies complex interrelations between chronic conditions. 17 PLoS One 2010;5(12):e15941. 18 For peer review only 19 42 Folstein MF, Folstein SE, McHugh PR. "Mini-mental state". A practical method for grading the 20 cognitive state of patients for the clinician. J Psychiatr Res 1975;12(3):189–98. 21 43 EuroQol--a new facility for the measurement of health-related quality of life. Health Policy 22 1990;16(3):199–208. 23 24 44 Graf Schulenburg JM, Claes C, Greiner W, et al. Die deutsche Version des EuroQol-Fragebogens. 25 J Public Health (Germany) 2009;6(1):3–20. 26 45 Janssen B, Szende A. Self-Reported Population Health: An International Perspective based on 27 28 EQ-5D: Population Norms for the EQ-5D. Dordrecht 2014. 29 46 Agborsangaya CB, Lahtinen M, Cooke T, et al. Comparing the EQ-5D 3L and 5L: Measurement 30 properties and association with chronic conditions and multimorbidity in the general 31 population. Health Qual Life Outcomes 2014;12:74. 32 33 47 Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in 34 longitudinal studies: Development and validation. J Chronic Dis 1987;40(5):373–83. 35 48 Diederichs C, Berger K, Bartels DB. The measurement of multiple chronic diseases--a systematic 36
37 review on existing multimorbidity indices. J Gerontol A Biol Sci Med Sci 2011;66(3):301–11. http://bmjopen.bmj.com/ 38 49 Renom-Guiteras A, Meyer G, Thürmann Pa. The EU(7)-PIM list: A list of potentially 39 inappropriate medications for older people consented by experts from seven European 40 countries. European Journal of Clinical Pharmacology 2015;71(7):861–75. 41 42 50 Holt S, Schmiedl S, Thürmann PA. Potentially Inappropiate medications in the Elderly: The 43 PRISCUS List. Dtsch Arztebl 2010;107(1):31–32. 44 51 Carnahan RM, Lund BC, Perry PJ, et al. The relationship of an anticholinergic rating scale with 45 on September 26, 2021 by guest. Protected copyright. 46 serum anticholinergic activity in elderly nursing home residents. Psychopharmacol Bull 47 2002;36(4):14–19. 48 52 Carnahan RM, Lund BC, Perry PJ, et al. The Anticholinergic Drug Scale as a measure of drug- 49 related anticholinergic burden: Associations with serum anticholinergic activity. J Clin 50 51 Pharmacol 2006;46(12):1481–86. 52 53 Cao Y-J, Mager DE, Simonsick EM, et al. Physical and cognitive performance and burden of 53 anticholinergics, sedatives, and ACE inhibitors in older women. Clin Pharmacol Ther 54 55 2008;83(3):422–29. 56 54 Durán CE, Azermai M, Vander Stichele RH. Systematic review of anticholinergic risk scales in 57 older adults. Eur J Clin Pharmacol 2013;69(7):1485–96. 58 59 55 Brauns H, Steinmann S (1999). Educational Reform in France, West-Germany, the United 60 Kingdom and Hungary.: Updating the CASMIN Educational Classification. ZUMA-Nachrichten, 1999:7–44. Available at: 29 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 48 BMJ Open
1 2
3 www.gesis.org/fileadmin/upload/forschung/publikationen/zeitschriften/zuma_nachrichten/zn BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 _44.pdf Accessed April 23, 2020. 5 6 56 Saunders JB, Aasland OG, Babor TF, et al. Development of the Alcohol Use Disorders 7 Identification Test (AUDIT): WHO Collaborative Project on Early Detection of Persons with 8 Harmful Alcohol Consumption--II. Addiction 1993;88(6):791–804. 9 10 57 Hanlon JT, Schmader KE, Samsa GP, et al. A method for assessing drug therapy appropriateness. 11 J Clin Epidemiol 1992;45(10):1045–51. 12 58 Hudon C, Fortin M, Soubhi H. Abbreviated guidelines for scoring the Cumulative Illness Rating 13 Scale (CIRS) in family practice. J Clin Epidemiol 2007;60(2):212. 14 15 59 Mukherjee B, Ou H-T, Wang F, et al. A new comorbidity index: The health-related quality of life 16 comorbidity index. J Clin Epidemiol 2011;64(3):309–19. 17 60 Ou H-T, Mukherjee B, Erickson SR, et al. Comparative performance of comorbidity indices in 18 For peer review only 19 predicting health care-related behaviors and outcomes among Medicaid enrollees with type 2 20 diabetes. Popul Health Manag 2012;15(4):220–29. 21 61 Sheikh JI, Yesavage JA, Brooks JO, et al. Proposed factor structure of the Geriatric Depression 22 Scale. Int Psychogeriatr 1991;3(1):23–28. 23 24 62 Yesavage JA, Brink TL, Rose TL, et al. Development and validation of a geriatric depression 25 screening scale: A preliminary report. J Psychiatr Res 1982;17(1):37–49. 26 63 Rubin DB. Multiple imputation for nonresponse in surveys. Hoboken, N.J.: Wiley-Interscience 27 28 2004. 29 64 van Buuren S, Groothuis-Oudshoorn K. mice: Multivariate Imputation by Chained Equations in 30 R. J Stat Softw 2011;45(3). 31 65 Wood AM, White IR, Royston P. How should variable selection be performed with multiply 32 33 imputed data? Stat Med 2008;27(17):3227–46. 34 66 Sauerbrei W. The Use of Resampling Methods to Simplify Regression Models in Medical 35 Statistics. J R Statist Soc C 1999;48(3):313–29. 36
37 67 Kerr KF, Wang Z, Janes H, et al. Net reclassification indices for evaluating risk prediction http://bmjopen.bmj.com/ 38 instruments: A critical review. Epidemiology 2014;25(1):114–21. 39 68 Nakagawa S, Schielzeth H, O'Hara RB. A general and simple method for obtaining R2 from 40 generalized linear mixed-effects models. Methods Ecol Evol 2013;4(2):133–42. 41 42 69 Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: A 43 framework for traditional and novel measures. Epidemiology 2010;21(1):128–38. 44 70 Nagelkerke NJD. A note on a general definition of the coefficient of determination. Biometrika 45 on September 26, 2021 by guest. Protected copyright. 46 1991;78(3):691–92. 47 71 R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R 48 Foundation for Statistical Computing 2016. 49 72 Pinhero J, Bates D, DebRoy S, et al. nlme: Linear and Nonlinear Mixed Effects Models. r package 50 51 version 3.1-128 2016. 52 73 van Buuren S, Groothuis-Oudshoorn K. mice: Multivariate Imputation by Chained Equations in 53 R. 54 55 74 Kundu S, Aulchenko YS, Janssens, A. Cecile J. W. PredictABEL: Assessment of Risk Prediction 56 Models. r package version 1.2-2 2014. 57 75 Lumley T. mitools: Tools for multiple imputation of missing data. r package version 2.3 2014. 58 59 76 Nakazawa M. fmsb: Functions for Medical Statistics Book with some Demographic Data. r 60 package version 0.5.2 2015. 77 Dahl DB. xtable: Export Tables to LaTeX or HTML. r package version 1.8-2 2016. 30 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 48
1 2
3 78 Xie Y. knitr: A General-Purpose Package for Dynamic Report Generation in R. r package version BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 1.13 2016. 5 6 79 Collins GS, Reitsma JB, Altman DG, et al. Transparent Reporting of a multivariable prediction 7 model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med 8 2015;162(1):55–63. 9 10 80 O'Caoimh R, Cornally N, Weathers E, et al. Risk prediction in the community: A systematic 11 review of case-finding instruments that predict adverse healthcare outcomes in community- 12 dwelling older adults. Maturitas 2015;82(1):3–21. 13 81 Wallace E, Stuart E, Vaughan N, et al. Risk prediction models to predict emergency hospital 14 15 admission in community-dwelling adults: A systematic review. Med Care 2014;52(8):751–65. 16 82 Coleman EA, Min S-j, Chomiak A, et al. Posthospital care transitions: Patterns, complications, 17 and risk identification. Health Serv Res 2004;39(5):1449–65. 18 For peer review only 19 83 Campbell SE, Seymour DG, Primrose WR. A systematic literature review of factors affecting 20 outcome in older medical patients admitted to hospital. Age Ageing 2004;33(2):110–15. 21 84 García-Pérez L, Linertová R, Lorenzo-Riera A, et al. Risk factors for hospital readmissions in 22 elderly patients: A systematic review. QJM 2011;104(8):639–51. 23 24 85 Alonso-Morán E, Nuño-Solinis R, Onder G, et al. Multimorbidity in risk stratification tools to 25 predict negative outcomes in adult population. Eur J Intern Med 2015;26(3):182–89. 26 86 Wallace E, Hinchey T, Dimitrov BD, et al. A systematic review of the probability of repeated 27 28 admission score in community-dwelling adults. J Am Geriatr Soc 2013;61(3):357–64. 29 87 Kansagara D, Englander H, Salanitro A, et al. Risk prediction models for hospital readmission: A 30 systematic review. JAMA 2011;306(15):1688–98. 31 88 Vest JR, Gamm LD, Oxford BA, et al. Determinants of preventable readmissions in the United 32 33 States: A systematic review. Implement Sci 2010;5:88. 34 89 Stevenson JM, Williams JL, Burnham TG, et al. Predicting adverse drug reactions in older adults; 35 a systematic review of the risk prediction models. Clin Interv Aging 2014;9:1581–93. 36
37 90 Cunny KA, Perri M. Single-item vs multiple-item measures of health-related quality of life. http://bmjopen.bmj.com/ 38 Psychol Rep 1991;69(1):127–30. 39 91 Krantz E, Wide U, Trimpou P, et al. Comparison between different instruments for measuring 40 health-related quality of life in a population sample, the WHO MONICA Project, Gothenburg, 41 42 Sweden: An observational, cross-sectional study. BMJ Open 2019;9(4):e024454. 43 92 Bosanquet K, Della Bailey, Gilbody S, et al. Diagnostic accuracy of the Whooley questions for 44 the identification of depression: A diagnostic meta-analysis. BMJ Open 2015;5(12):e008913. 45 on September 26, 2021 by guest. Protected copyright. 46 93 Whooley MA, Avins AL, Miranda J, et al. Case-finding instruments for depression. Two 47 questions are as good as many. J Gen Intern Med 1997;12(7):439–45. 48 94 Swart E, Gothe H, Geyer S, et al. Gute Praxis Sekundärdatenanalyse (GPS): Leitlinien und 49 Empfehlungen. Gesundheitswesen 2015;77(2):120–26. 50 51 52 53 54 55 56 57 58 59 60
31 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 48 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Figure 1: Predicted outcomes with regard to general trajectories of well-being and quality of life in patients 32 with multimorbidity and polypharmacy 33 http://bmjopen.bmj.com/ 254x190mm (96 x 96 DPI) 34 35 36 37 38 39 40 41 on September 26, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 48 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 For peer review only 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Figure 2: Flow chart of the study population in both datasets 32 33 254x190mm (96 x 96 DPI) http://bmjopen.bmj.com/ 34 35 36 37 38 39 40 41 on September 26, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 48 BMJ Open
1 2
3 Additional file 1: Falls and fall-related injuries: list of excluded ICD-10-codes BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 Conditions ICD-10-Code 7 8 (ICD-10-GM Version 2014) 9 10 . Osteoporosis with pathological fracture M80 11 12 . Open wound of neck S11 13 14 15 . Dislocation, sprain and strain of joints and S13 16 17 ligaments at neck level 18 For peer review only 19 . Injury of nerves and spinal cord at neck level S14 20 21 22 . Injury of blood vessels at neck level S15 23 24 . Injury of muscle and tendon at neck level S16 25 26 . Crushing injury of neck S17 27 28 29 . Traumatic amputation at neck level S18 30 31 . Other and unspecified injuries of neck S19 32 33 . Injury of blood vessels of thorax S25 34 35 . Injury of heart S26 36
37 http://bmjopen.bmj.com/ 38 . Injury of other and unspecified intrathoracic S27 39 40 organs 41 42 . Crushing injury of thorax and traumatic amputation S28 43 44
45 of part of thorax on September 26, 2021 by guest. Protected copyright. 46 47 . Other and unspecified injuries of thorax S29 48 49 . Open wound of abdomen, lower back and pelvis S31 50 51 52 . Injury of blood vessels at abdomen, lower back S35 53 54 and pelvis level 55 56 . Injury of pancreas S36.2 57 58 . Injury of stomach S36.3 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 48
1 2
3 . Injury of small intestine S36.4 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 . 6 Injury of colon S36.5 7 8 . Injury of rectum S36.6 9 10 . Injury of multiple intra-abdominal organs S36.7 11 12 . Injury of other intra-abdominal organs S36.8 13 14 15 . Injury of ureter S37.1 16 17 . Injury of bladder S37.2 18 For peer review only 19 . Injury of urethra S37.3 20 21 22 . Injury of ovary S37.4 23 24 . Injury of fallopian tube S37.5 25 26 . Injury of uterus S37.6 27 28 . Injury of multiple pelvic organs S37.7 29 30 31 . Injury of other pelvic organs S37.8 32 33 . Crushing injury and traumatic amputation of part S38 34 35 of abdomen, lower back and pelvis 36
37 http://bmjopen.bmj.com/ 38 . Injury of intra-abdominal organ(s) with pelvic S39.6 39 40 organ(s) 41 42 . Other multiple injuries of abdomen, lower back S39.7 43 44
45 and pelvis on September 26, 2021 by guest. Protected copyright. 46 47 . Other specified injuries of abdomen, lower back S39.8 48 49 and pelvis 50 51 . Crushing injury of shoulder and upper arm S47 52 53 54 . Traumatic amputation of shoulder and upper arm S48 55 56 . Traumatic amputation of forearm S58 57 58 . Traumatic amputation of wrist and hand S68 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 37 of 48 BMJ Open
1 2
3 . Traumatic amputation of hip and thigh S78 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 . 6 Traumatic amputation of lower leg S88 7 8 . Traumatic amputation of ankle and foot S98 9 10 . Crushing injuries involving multiple body regions T04 11 12 . Traumatic amputations involving multiple body T05 13 14 15 regions 16 17 . Effects of foreign body entering through natural T15-19 18 For peer review only 19 orifice 20 21 22 . Burns and corrosions T20-32 23 24 . Frostbite T33-35 25 26 . Poisoning by specified narcotics and T40.0-T40.1, T40.3, T40.5- 27 28 psychodysleptic agents (hallucinogenic drugs) T40.9 29 30 31 . Toxic effects of substances chiefly nonmedicinal T51-65 32 33 as to source 34 35 . Other and unspecified effects of external causes T66-77 36
37 http://bmjopen.bmj.com/ 38 . Complications of surgical and medical care, not T80-87 39 40 elsewhere classified 41 42 . Sequelae of injuries, of poisoning and of other T90-98 43 44
45 consequences of external causes on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
BMJ Open Page 38 of 48
Additional file 2: Univariate Analyses: Association between predictor variables and combined outcome 1 2 combined outcome reached combined outcome reached combined outcome reached 3 4 after 6 months after 9 months after 24 months 5 6 Yes No Total Yes No Total Yes No Total 7 8 (n = (n = (n = (n = (n = (n = (n = (n = (n = 9 10 192357) 400099) 592456) 244190) 348266) 592456) 387951) 204505) 592456) 11 12 Mean MeanFor Mean peer p-value Mean review Mean Mean only p-value Mean Mean Mean p-value 13 14 (SD) (SD) (SD) (SD) (SD) (SD) (SD) (SD) (SD) 15 16
Age 72.2 (7.3) 70.8 (6.9) 71.3 (7.1) <0.001 72.1 (7.3) 70.7 (6.9) 71.3 (7.1) <0.001 http://bmjopen.bmj.com/ 71.8 (7.2) 70.2 (6.7) 71.3 (7.1) <0.001 17 18 Disease count 10.5 (4.0) 9.3 (3.6) 9.7 (3.8) <0.001 10.4 (3.9) 9.2 (3.5) 9.7 (3.8) <0.001 10.2 (3.9) 8.8 (3.4) 9.7 (3.8) <0.001 19 20 CCI 3.4 (2.8) 2.8 (2.4) 3.0 (2.5) <0.001 3.4 (2.7) 2.8 (2.4) 3.0 (2.5) <0.001 3.2 (2.7) 2.6 (2.3) 3.0 (2.5) <0.001 21 22 No. of specific chronic 4.7 (2.1) 4.1 (1.9) 4.3 (2.0) <0.001 4.7 (2.1) 4.1 (1.9) 4.3 (2.0) <0.001 4.6 (2.0) 3.9 (1.8) 4.3 (2.0) <0.001 23 24 diseases (Diederichs) on September 26, 2021 by guest. Protected copyright. 25 26 No. of PIM (EU-PIM) 1.3 (1.2) 1.1 (1.1) 1.1 (1.2) <0.001 1.3 (1.2) 1.1 (1.1) 1.1 (1.2) <0.001 1.2 (1.2) 1.0 (1.0) 1.1 (1.1) <0.001 27 28 ADS 1.2 (1.6) 0.9 (1.4) 1.0 (1.5) <0.001 1.2 (1.6) 0.9 (1.3) 1.0 (1.5) <0.001 1.1 (1.5) 0.8 (1.3) 1.0 (1.5) <0.001 29 30 DBI 0.9 (1.2) 0.7 (1.0) 0.8 (1.0) <0.001 0.9 (1.1) 0.7 (0.9) 0.8 (1.0) <0.001 0.8 (1.1) 0.6 (0.9) 0.8 (1.0) <0.001 31 32 Number of involved 11.0 (5.6) 9.4 (5.0) 10.0 (5.3) <0.001 10.9 (5.6) 9.3 (4.9) 10.0 (5.3) <0.001 10.6 (5.4) 8.8 (4.7) 10.0 (5.3) <0.001 33 34 physicians 35 36 Abbreviations: ACh burden – Anticholinergic drug burden, ADS – Anticholinergic Drug Scale, CCI - Charlson Comorbidity Index, DBI – Drug 37 38 Burden Index, PIM – Potentially Inappropriate Medication. 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Page 39 of 48 BMJ Open
combined outcome reached combined outcome reached combined outcome reached 1 2 after 6 months after 9 months after 24 months 3 4 Yes No Total Yes No Total Yes No Total 5 6 (n = (n = (n = (n = (n = (n = (n = (n = (n = 7 8 192357) 400099) 592456) 244190) 348266) 592456) 387951) 204505) 592456) 9 10 n (%) n (%) n (%) p-value n (%) n (%) n (%) p-value n (%) n (%) n (%) p-value 11 12 Sex Female 90924 182079For 273003 peer 114859 review 158144 273003 only 180222 92781 273003 13 14 (47.3%) (45.5%) (46.1%) (47.0%) (45.4%) (46.1%) (46.5%) (45.4%) (46.1%) 15 16 Male 101433 218020 319453 <0.001 129331 190122 319453 <0.001 207729 111724 319453 <0.001 http://bmjopen.bmj.com/ 17 18 (52.7%) (54.5%) (53.9%) (53.0%) (54.6%) (53.9%) (53.5%) (54.6%) (53.9%) 19 20 Previous No 107627 289845 397472 141326 256146 397472 240676 156796 397472 21 22 hospitalis (56%) (72.4%) (67.1%) (57.9%) (73.5%) (67.1%) (62%) (76.7%) (67.1%) 23 ation 24 yes 84730 110254 194984 <0.001 102864 92120 194984 <0.001 147275 47709 194984 <0.001 on September 26, 2021 by guest. Protected copyright. 25 (44%) (27.6%) (32.9%) (42.1%) (26.5%) (32.9%) (38%) (23.3%) (32.9%) 26 27 Previous No 103734 325335 429069 142477 286592 429069 255691 173378 429069 28 29 falls (53.9%) (81.3%) (72.4%) (58.3%) (82.3%) (72.4%) (65.9%) (84.8%) (72.4%) 30 31 yes 88623 74764 163387 <0.001 101713 61674 163387 <0.001 132260 31127 163387 <0.001 32 33 (46.1%) (18.7%) (27.6%) (41.7%) (17.7%) (27.6%) (34.1%) (15.2%) (27.6%) 34 35 36 37 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 40 of 48
1 2
3 Additional file 3: Univariate analyses BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 Table A3.1: Association between predictor variables and EQ5D-3L (Pearson 7 8 Correlation) 9 10 Predictor variable EQ5D-3L (T1) 11 12 13 Predictors available in both study samples 14 15 . Age -0.09 16 17 . Disease count -0.24 18 For peer review only 19 . Charlson Comorbidity Index (CCI) -0.07 20 21 22 . No. of specific chronic diseases (Diederichs) -0.19 23 24 . No. of drugs -0.26 25 26 . No. of PIM (EU-PIM) -0.18 27 28 29 . ACh burden (ADS) -0.16 30 31 . Mod. Drug Burden Index -0.20 32 33 . No. of involved physicians -0.06 34 35 36 Additional predictors available only in trial data
37 http://bmjopen.bmj.com/ 38 . No. of persons living in household 0.04 39 40 . CASMIN 0.09 41 42 . Alcohol intake (AUDIT C) 0.13 43 44
45 . Body Mass Index -0.15 on September 26, 2021 by guest. Protected copyright. 46 47 . MAI -0.24 48 49 . CIRS sum score -0.27 50 51 52 . CIRS, no. of organ systems -0.22 53 54 . HRQoL-CI, mental -0.24 55 56 . HRQoL-CI, physical -0.20 57 58 59 . Depressive Symptoms (GDS) -0.52 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 48 BMJ Open
1 2
3 . EQ5D-3L (Baseline) 0.68 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 7 8 Abbreviations: ACh burden – Anticholinergic drug burden, ADS – Anticholinergic 9 10 Drug Scale, AUDIT - Alcohol Use Disorders Identification Test (WHO), CASMIN - 11 12 Comparative Analysis of Social Mobility in Industrial Nations, CCI - Charlson 13 14 15 Comorbidity Index, CIRS – Cumulative Illness Rating Scale, GDS – Geriatric 16 17 Depression Scale, HRQoL – Health-Related Quality of Life, HRQoL-CI – Health- 18 For peer review only 19 Related Quality of Life Comorbidity Index, MAI – Medication Appropriateness Index, 20 21 22 PIM – Potentially Inappropriate Medication. 23 24 25 26 27 28 Table A3.2: Association between predictor variables and EQ5D-3L (T-Test) 29 30 31 Predictor variable Mean (SD) Mean (SD) p-value 32 33 34 Predictors available in both study samples 35 36 . Sex (female / male) 67.7 (25.77) 78.5 (22.97) <0.001
37 http://bmjopen.bmj.com/ 38 . Previous hospitalisation (yes / no) 71.2 (26.76) 73.3 (24.67) 0.531 39 40 41 . Previous falls (yes/no) 65.9 (25.28) 74.3 (24.76) 0.011 42 43 Additional predictors available only in trial data 44 45 . Smoker (yes / no) 74.1 (25.81) 73.1 (25.07) 0.807 on September 26, 2021 by guest. Protected copyright. 46 47 48 Abbreviations: SD – Standard Deviation 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 42 of 48
1 2 Additional file 4: Best model based on claims data (table 2, model 1.1)
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 OR [95%-CI] p-value z-value NRI 5 6 7 Age (years) 1.02 [1.02; 1.02] <0.001 45.4 0.11 8 9 Sex (female) 0.99 [0.97; 1.00] 0.025 -2.2 -0.03 10 11 Disease count 1.02 [1.02; 1.03] <0.001 19.2 0.05 12 13 14 CCI 1.03 [1.03; 1.04] <0.001 22.7 0.02 15 16 No. of specific chronic 1.01 [1.00; 1.01] <0.001 3.9 0.01 17 18 diseases (Diederichs)For peer review only 19 20 No. of PIM (EU-PIM) 1.03 [1.02; 1.03] <0.001 8.9 0.02 21 22 23 ACh burden (ADS) 1.04 [1.03; 1.05] <0.001 14.9 0.04 24 25 Mod. Drug Burden Index 1.08 [1.07; 1.08] <0.001 20.1 0.08 26 27 Previous hospitalisations 1.67 [1.65; 1.70] <0.001 82.3 0.34 28 29 30 Previous falls/fall-related 3.29 [3.25; 3.34] <0.001 188.6 0.55 31 32 injuries 33 34 No. of involved physicians 1.02 [1.02; 1.02] <0.001 29.0 0.08 35 36
37 http://bmjopen.bmj.com/ 38 39 Legend: ACh burden – Anticholinergic drug burden, ADS - Anticholinergic Drug 40 41 Scale, CCI - Charlson Comorbidity Index, NRI - Continuous Net Reclassification 42 43 44 Index, OR – Odds Ratio - PIM – Potentially Inappropriate Medication
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 43 of 48 BMJ Open
1 2 Additional file 5: Predictive models based on CRT data
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Estimates [95%-CI] p-value 5 6 7 Table 2, model 2.4 based on predictors 8 9 available in both study samples 10 11 Intercept 101.18 [93.11; 109.25] <0.001 12 13 14 Sex (female) -11.26 [-15.59; -6.94] <0.001 15 16 No. of specific chronic diseases (Diederichs) -2.18 [-3.64; -0.73] 0.004 17 18 No. of drugs For peer review-1.28 only[-2.23; -0.32] 0.010 19 20 21 Mod. Drug Burden Index -5.19 [-8.26; -2.12] 0.001 22 23 Previous falls -6.11 [-12.07; -0.15] 0.045 24 25 Table 2, model 3.4: best model, with 26 27 additional predictors available only in 28 29 30 CRT data 31 32 Intercept 51.74 [38.91; 64.57] <0.001 33 34 Sex (female) -3.61 [-6.96; -0.27] 0.036 35 36
37 No. of specific chronic diseases (Diederichs) -1.03 [-2.08; 0.01] 0.055 http://bmjopen.bmj.com/ 38 39 No. of involved physicians 0.80 [-0.13; 1.74] 0.093 40 41 Body Mass Index (BMI) -0.28 [-0.53; -0.03] 0.031 42 43 44 Medication Appropriateness Index (MAI) -0.39 [-0.70; -0.08] 0.015
45 on September 26, 2021 by guest. Protected copyright. 46 Depressive symptoms (GDS) -2.73 [-3.56; -1.91] <0.001 47 48 EQ-5D Index Score (Baseline) 0.55 [0.47; 0.64] <0.001 49 50 51 52 53 Legend: GDS – Geriatric Depression Scale 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 44 of 48 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 3 4 Reporting checklist for prediction model 5 6 7 development and validation study. 8 9 10 11 Based on the TRIPOD guidelines. 12 13 14 15 Instructions to authors 16 For peer review only 17 Complete this checklist by entering the page numbers from your manuscript where readers will find 18 19 20 each of the items listed below. 21 22 23 Your article may not currently address all the items on the checklist. Please modify your text to 24 25 include the missing information. If you are certain that an item does not apply, please write "n/a" and 26 27 provide a short explanation. 28 29 30 Upload your completed checklist as an extra file when you submit to a journal. 31 32 http://bmjopen.bmj.com/ 33 In your methods section, say that you used the TRIPODreporting guidelines, and cite them as: 34 35 36 37 Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction 38 39 model for individual prognosis or diagnosis (TRIPOD): The TRIPOD statement. 40 41 on September 26, 2021 by guest. Protected copyright. 42 Page 43 44 Reporting Item Number 45 46 47 Title 48 49 50 #1 Identify the study as developing and / or validating a 1 51 52 53 multivariable prediction model, the target population, and 54 55 the outcome to be predicted. 56 57 58 Abstract 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 45 of 48 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 #2 Provide a summary of objectives, study design, setting, 2 3 4 participants, sample size, predictors, outcome, statistical 5 6 analysis, results, and conclusions. 7 8 9 Introduction 10 11 12 #3a Explain the medical context (including whether diagnostic 4 13 14 15 or prognostic) and rationale for developing or validating 16 For peer review only 17 the multivariable prediction model, including references 18 19 to existing models. 20 21 22 #3b Specify the objectives, including whether the study 5 23 24 describes the development or validation of the model or 25 26 27 both. 28 29 30 Methods 31 32 33 Source of data #4a Describe the study design or source of data (e.g., 5/6 http://bmjopen.bmj.com/ 34 35 randomized trial, cohort, or registry data), separately for 36 37 the development and validation data sets, if applicable. 38 39 40 41 Source of data #4b Specify the key study dates, including start of accrual; 6 on September 26, 2021 by guest. Protected copyright. 42 43 end of accrual; and, if applicable, end of follow-up. 44 45 46 Participants #5a Specify key elements of the study setting (e.g., primary 6 47 48 care, secondary care, general population) including 49 50 number and location of centres. 51 52 53 54 Participants #5b Describe eligibility criteria for participants. 6 55 56 57 Participants #5c Give details of treatments received, if relevant n/a 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 46 of 48 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Outcome #6a Clearly define the outcome that is predicted by the 7 3 4 prediction model, including how and when assessed. 5 6 7 Outcome #6b Report any actions to blind assessment of the outcome n/a 8 9 to be predicted. 10 11 12 Predictors #7a Clearly define all predictors used in developing or 8 13 14 15 validating the multivariable prediction model, including 16 For peer review only 17 how and when they were measured 18 19 20 Predictors #7b Report any actions to blind assessment of predictors for n/a 21 22 the outcome and other predictors. 23 24 25 Sample size #8 Explain how the study size was arrived at. 6 26 27 28 Missing data #9 Describe how missing data were handled (e.g., 11 29 30 31 complete-case analysis, single imputation, multiple 32 33 imputation) with details of any imputation method. http://bmjopen.bmj.com/ 34 35 36 Statistical #10a If you are developing a prediction model describe how 11 37 38 analysis methods predictors were handled in the analyses. 39 40 41 Statistical #10b If you are developing a prediction model, specify type of 11-13 on September 26, 2021 by guest. Protected copyright. 42 43 44 analysis methods model, all model-building procedures (including any 45 46 predictor selection), and method for internal validation. 47 48 49 Statistical #10c If you are validating a prediction model, describe how the 11-13 50 51 analysis methods predictions were calculated. 52 53 54 Statistical #10d Specify all measures used to assess model performance 11-13 55 56 57 analysis methods and, if relevant, to compare multiple models. 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 47 of 48 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Statistical #10e If you are validating a prediction model, describe any 11-13 3 4 analysis methods model updating (e.g., recalibration) arising from the 5 6 validation, if done 7 8 9 Risk groups #11 Provide details on how risk groups were created, if done. n/a 10 11 12 Development vs. #12 For validation, identify any differences from the 11-13 13 14 15 validation development data in setting, eligibility criteria, outcome, 16 For peer review only 17 and predictors. 18 19 20 Results 21 22 23 Participants #13a Describe the flow of participants through the study, Figure 2 24 25 including the number of participants with and without the 26 27 28 outcome and, if applicable, a summary of the follow-up 29 30 time. A diagram may be helpful. 31 32 33 Participants #13b Describe the characteristics of the participants (basic Table 1, http://bmjopen.bmj.com/ 34 35 demographics, clinical features, available predictors), Additional 36 37 including the number of participants with missing data for file 2 38 39 40 predictors and outcome. 41 on September 26, 2021 by guest. Protected copyright. 42 43 Participants #13c For validation, show a comparison with the development Additional 44 45 data of the distribution of important variables file 2 46 47 (demographics, predictors and outcome). 48 49 50 Model #14a If developing a model, specify the number of participants Additional 51 52 53 development and outcome events in each analysis. files 3-5 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 48 of 48 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Model #14b If developing a model, report the unadjusted association, Additional 3 4 development if calculated between each candidate predictor and files 3-5 5 6 outcome. 7 8 9 Model #15a If developing a model, present the full prediction model to Additional 10 11 specification allow predictions for individuals (i.e., all regression files 3-5 12 13 14 coefficients, and model intercept or baseline survival at a 15 16 Forgiven peer time point). review only 17 18 19 Model #15b If developing a prediction model, explain how to the use 17 20 21 specification it. 22 23 24 Model #16 Report performance measures (with CIs) for the Additional 25 26 27 performance prediction model. files 3-5 28 29 30 Model-updating #17 If validating a model, report the results from any model 17 31 32 updating, if done (i.e., model specification, model 33 http://bmjopen.bmj.com/ 34 performance). 35 36 37 Discussion 38 39 40 41 Limitations #18 Discuss any limitations of the study (such as 23 on September 26, 2021 by guest. Protected copyright. 42 43 nonrepresentative sample, few events per predictor, 44 45 missing data). 46 47 48 Interpretation #19a For validation, discuss the results with reference to 22 49 50 performance in the development data, and any other 51 52 53 validation data 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 49 of 48 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Interpretation #19b Give an overall interpretation of the results, considering 23 3 4 objectives, limitations, results from similar studies, and 5 6 other relevant evidence. 7 8 9 Implications #20 Discuss the potential clinical use of the model and 23 10 11 implications for future research 12 13 14 15 Other information 16 For peer review only 17 18 Supplementary #21 Provide information about the availability of 25 19 20 information supplementary resources, such as study protocol, Web 21 22 calculator, and data sets. 23 24 25 Funding #22 Give the source of funding and the role of the funders for 26 26 27 28 the present study. 29 30 31 None The TRIPOD checklist is distributed under the terms of the Creative Commons Attribution 32 33 License CC-BY. This checklist can be completed online using https://www.goodreports.org/, a tool http://bmjopen.bmj.com/ 34 35 made by the EQUATOR Network in collaboration with Penelope.ai 36 37 38 39 40 41 on September 26, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Development and internal validation of prognostic models to predict negative health outcomes in older patients with multimorbidity and polypharmacy in general practice ForJournal: peerBMJ Open review only Manuscript ID bmjopen-2020-039747.R1
Article Type: Original research
Date Submitted by the 26-Aug-2020 Author:
Complete List of Authors: Müller, Beate; Goethe University Frankfurt, Institute of General Practice Uhlmann, Lorenz ; University of Heidelberg, Institute of Medical Biometry and Informatics Ihle, Peter; University of Cologne, PMV Research Group, Faculty of Medicine and University Hospital Cologne Stock, Christian; University of Heidelberg, Institute of Medical Biometry and Informatics von Buedingen, Fiona; Goethe University Frankfurt, Institute of General Practice Beyer, Martin; Goethe University Frankfurt, Institute of General Practice Gerlach, Ferdinand; Goethe University Frankfurt, Institute of General Practice
Perera, Rafael; University of Oxford, Nuffield Department of Primary http://bmjopen.bmj.com/ Care Health Sciences Valderas, Jose; University of Exeter Medical School, APEx Collaboration for Academic Primary Care Glasziou, Paul; Bond University, Centre for Research in Evidence-Based Practice, Faculty of Health Sciences and Medicine van den Akker, Marjan; Goethe University Frankfurt, Institute of General Practice; Maastricht University, Department of Family Medicine, School CAPHRI
Muth, Christiane; Goethe University Frankfurt, Institute of General on September 26, 2021 by guest. Protected copyright. Practice
Primary Subject General practice / Family practice Heading:
Secondary Subject Heading: Geriatric medicine, Health services research
PRIMARY CARE, THERAPEUTICS, GERIATRIC MEDICINE, HEALTH Keywords: SERVICES ADMINISTRATION & MANAGEMENT
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 1 of 90 BMJ Open
1 2
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 7 8 9 I, the Submitting Author has the right to grant and does grant on behalf of all authors of the Work (as defined 10 in the below author licence), an exclusive licence and/or a non-exclusive licence for contributions from authors 11 who are: i) UK Crown employees; ii) where BMJ has agreed a CC-BY licence shall apply, and/or iii) in accordance 12 with the terms applicable for US Federal Government officers or employees acting as part of their official 13 duties; on a worldwide, perpetual, irrevocable, royalty-free basis to BMJ Publishing Group Ltd (“BMJ”) its 14 licensees and where the relevant Journal is co-owned by BMJ to the co-owners of the Journal, to publish the 15 Work in this journal and any other BMJ products and to exploit all rights, as set out in our licence. 16 17 The Submitting Author accepts and understands that any supply made under these terms is made by BMJ to 18 the Submitting Author Forunless you peer are acting as review an employee on behalf only of your employer or a postgraduate 19 student of an affiliated institution which is paying any applicable article publishing charge (“APC”) for Open 20 Access articles. Where the Submitting Author wishes to make the Work available on an Open Access basis (and 21 intends to pay the relevant APC), the terms of reuse of such Open Access shall be governed by a Creative 22 Commons licence – details of these licences and which Creative Commons licence will apply to this Work are set 23 out in our licence referred to above. 24 25 Other than as permitted in any relevant BMJ Author’s Self Archiving Policies, I confirm this Work has not been 26 accepted for publication elsewhere, is not being considered for publication elsewhere and does not duplicate 27 material already published. I confirm all authors consent to publication of this Work and authorise the granting 28 of this licence. 29 30 31 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 2 of 90
1 2
3 1 TITLE BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 2 Development and internal validation of prognostic models to predict negative health 6 3 outcomes in older patients with multimorbidity and polypharmacy in general practice 7 4 8 9 5 Corresponding author 10 11 6 Beate S. Müller, Institute of General Practice, Goethe University, Frankfurt/Main, 12 7 Theodor-Stern-Kai 7, 60590 Frankfurt, Germany, [email protected] 13 8 frankfurt.de 14 15 9 16 10 Authors 17 11 1. Beate S. Müller, Institute of General Practice, Goethe University, Frankfurt/Main, 18 For peer review only 19 12 Germany 20 13 2. Lorenz Uhlmann, Institute of Medical Biometry and Informatics, University of 21 14 Heidelberg, Heidelberg, Germany 22 23 15 3. Peter Ihle, PMV Research Group, Faculty of Medicine and University Hospital 24 16 Cologne, University of Cologne, Köln, Germany 25 17 4. Christian Stock, Institute of Medical Biometry and Informatics, University of 26 27 18 Heidelberg, Heidelberg, Germany 28 19 5. Fiona von Buedingen, Institute of General Practice, Goethe University, 29 20 Frankfurt/Main, Germany 30 31 21 6. Martin Beyer, Institute of General Practice, Goethe University, Frankfurt/Main, 32 22 Germany 33 23 7. Ferdinand M. Gerlach, Institute of General Practice, Goethe University, 34 35 24 Frankfurt/Main, Germany 36 25 8. Rafael Perera, Nuffield Department of Primary Care Health Sciences, University 37 26 of Oxford, Oxford, UK http://bmjopen.bmj.com/ 38 39 27 9. Jose M. Valderas, APEx Collaboration for Academic Primary Care, University of 40 28 Exeter Medical School, Exeter, UK 41 29 10.Paul P. Glasziou, Centre for Research in Evidence-Based Practice (CREBP), 42 43 30 Faculty of Health Sciences and Medicine, Bond University, Robina, Australia 44 31 11.Marjan van den Akker, (1) Institute of General Practice, Goethe University, 45 32 Frankfurt/Main, Germany, (2) Department of Family Medicine, School CAPHRI, on September 26, 2021 by guest. Protected copyright. 46 47 33 Maastricht University, Maastricht, Netherlands 48 34 12.Christiane Muth, Institute of General Practice, Goethe University, Frankfurt/Main, 49 35 Germany 50 51 36 52 37 Word count: 4,707 53 54 38 55 56 39 57 58 59 60
1 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 3 of 90 BMJ Open
1 2
3 40 ABSTRACT BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 41 Background: Polypharmacy interventions are resource-intensive and should be 7 8 42 targeted to those at risk of negative health outcomes. Our aim was to develop and 9 10 43 internally validate prognostic models to predict health-related quality of life (HRQoL) 11 12 44 and the combined outcome of falls, hospitalization, institutionalization, and nursing 13 14 15 45 care needs, in older patients with multimorbidity and polypharmacy in general 16 17 46 practices. 18 For peer review only 19 47 Methods: Design: two independent datasets, one comprising health insurance claims 20 21 22 48 data (N=592,456), the other data from the PRIoritising MUltimedication in 23 24 49 Multimorbidity (PRIMUM) cluster-RCT (N=502). Population: ≥60 years, ≥5 drugs, ≥3 25 26 50 chronic diseases, excluding dementia. Outcomes: combined outcome of falls, 27 28 hospitalization, institutionalization, and nursing care needs (after 6, 9, 24 months) 29 51 30 31 52 (claims data); and HRQoL (after 6, 9 months) (trial data). Predictor variables in both 32 33 53 datasets: age, sex, morbidity-related variables (disease count), medication-related 34 35 54 variables (EU-PIM), health service utilization. Predictor variables exclusively in trial 36
37 http://bmjopen.bmj.com/ 38 55 data: additional socio-demographics, morbidity-related variables (Cumulative Illness 39 40 56 Rating Scale, depression), Medication Appropriateness Index, lifestyle, functional 41 42 57 status, HRQoL (EuroQol EQ5D-3L). Analysis: mixed regression models, combined 43 44
45 58 with stepwise variable selection, 10-fold cross validation, sensitivity analyses. on September 26, 2021 by guest. Protected copyright. 46 47 59 Results: Most important predictors of EQ5D-3L at 6 months in best model 48 49 60 (Nagelkerke’s R² 0.507) were depressive symptoms (-2.73 [95%CI: -3.56 to -1.91]), 50 51 MAI (-0.39 [-0.7 to -0.08]), baseline EQ5D-3L (0.55 [95%CI: 0.47 to 0.64]). Models 52 61 53 54 62 based on claims data and those predicting long-term outcomes based on both 55 56 63 datasets produced low R² values. In claims data-based model with highest 57 58 64 explanatory power (R²=0.16), previous falls/fall-related injuries, previous 59 60
2 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 4 of 90
1 2
3 65 hospitalizations, age, number of involved physicians, and disease count were most BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 66 important predictor variables. 7 8 67 Conclusions: Best trial data-based model predicted HRQoL after 6 months well and 9 10 68 included parameters of well-being not found in claims. Performance of claims data- 11 12 69 based models and models predicting long-term outcomes was relatively weak. For 13 14 15 70 generalizability, future studies should refit models by considering parameters 16 17 71 representing well-being and functional status. 18 For peer review only 19 72 20 21 22 73 STRENGTHS AND LIMITATIONS OF THIS STUDY 23 24 74 We developed our predictive models using two completely different datasets – 25 26 75 claims data and data primarily collected in a cluster-randomized trial. 27 28 29 76 The claims data contained a large number of cases, enabling our models to 30 31 77 include many possible predictors without any convergence issues. 32 33 78 The trial data provided a rich set of potential predictor variables of high data 34 35 36 79 quality and included data on patient-reported outcome measures, such as
37 http://bmjopen.bmj.com/ 38 80 well-being and functional status. 39 40 81 Both datasets have their own methodological limitations, such as imprecise 41 42 43 82 claims data (collected for reimbursement purposes) und the trial’s small 44
45 83 sample size. on September 26, 2021 by guest. Protected copyright. 46 47 84 The nature of the data meant neither dataset could be used to validate a 48 49 50 85 predictive model based on the other. 51 52 86 53 54 87 Keywords: Polypharmacy [MeSH], multimorbidity [MeSH], aged [MeSH], general 55 56 88 practice [MeSH], primary care [MeSH], prognostic model, clinical prediction [MeSH], 57 58 59 89 drug therapy [MeSH], chronic disease [MeSH] 60
3 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 5 of 90 BMJ Open
1 2
3 90 BACKGROUND BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 91 Currently, up to 80% of primary care consultations involve patients with multiple 7 8 92 chronic conditions (multimorbidity).[1] A multiplicity of disorders in patients is 9 10 93 associated with polypharmacy. Both multimorbidity and polypharmacy are recognized 11 12 94 as a major challenge facing health care systems.[2–5] Polypharmacy can increase 13 14 15 95 the risk of mortality, hospitalization,[6, 7] and falls and fall-related injuries with 16 17 96 resulting disability and loss of autonomy.[8, 9] It can also reduce cognitive and 18 For peer review only 19 97 physical function, as well as health-related quality of life (HRQoL).[2, 10] 20 21 22 98 The number of drugs increases the probability of adverse drug reactions (ADRs), but 23 24 99 the relationship is inconsistent, suggesting that the number of medications alone may 25 26 100 not adequately indicate the quality of an individual’s medication regimen.[11, 12] The 27 28 kind of drugs prescribed plays an important role in the type of reaction, with certain 29 101 30 31 102 medication classes, such as benzodiazepines, demonstrating a significant 32 33 103 association with falls, and medications with anti-cholinergic properties being 34 35 104 associated with impaired cognitive and physical function in elderly individuals.[13, 14] 36
37 http://bmjopen.bmj.com/ 38 105 At a physician level, the cause of these negative health outcomes of polypharmacy 39 40 106 may be inappropriate prescribing, including undertreatment.[15–18] At a patient level, 41 42 107 a high number of drugs and the complexity of a drug regimen is often associated with 43 44
45 108 poor adherence,[19] which may be exacerbated by the presence of depression on September 26, 2021 by guest. Protected copyright. 46 47 109 and/or cognitive impairment.[20] Moreover, polypharmacy may also result in an 48 49 110 accumulation of potentially inappropriate medications (PIMs). 50 51 Several complex interventions have been developed to optimize (inappropriate) 52 111 53 54 112 polypharmacy. However, despite their evidence-based rationale, they have led to 55 56 113 inconsistent improvements in process parameters of care and failed to impact 57 58 114 patient-relevant outcomes.[21, 22] One possible reason for this is that the included 59 60 115 populations are too heterogeneous in terms of their baseline risk and potentially 4 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 6 of 90
1 2
3 116 achievable intervention effects. For example, the majority of the study population BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 117 included in the PRIMUM trial showed very good quality of life and functional status at 7 8 118 baseline, even though participants had at least three chronic conditions affecting 9 10 119 more than two organ systems, five or more chronic drug prescriptions and were 60 11 12 120 years of age or older. The authors therefore concluded that there was not enough 13 14 15 121 room for improvement.[23] This highlights current difficulties in defining inclusion 16 17 122 criteria in polypharmacy trials in such a way that selected populations have a 18 For peer review only 19 123 considerable baseline risk and can be expected to benefit from the intervention. 20 21 22 124 Moreover, as polypharmacy interventions tend to address inappropriate prescribing, 23 24 125 health care coordination etc., they are generally complex.[21, 22] As the complex 25 26 126 interventions are also resource-intensive, it would be preferable for a stratified 27 28 approach to address patients that are at high risk of negative health outcomes and 29 127 30 31 128 most likely to benefit from them.[24] 32 33 129 The course of multimorbidity (and associated polypharmacy) has been characterised 34 35 130 by a decline in well-being (e.g., functional decline or worsening of quality of life due to 36
37 http://bmjopen.bmj.com/ 38 131 inappropriate prescriptions and/or deterioration in one or more chronic diseases), 39 40 132 interrupted by adverse events (e.g., exacerbations of chronic diseases or adverse 41 42 133 drug reactions).[25, 26] In order to identify a population at high risk, it is therefore 43 44
45 134 necessary to predict a wide array of possible negative health outcomes. Several on September 26, 2021 by guest. Protected copyright. 46 47 135 prognostic models have predicted single outcomes, mainly mortality or unplanned 48 49 136 hospital (re-)admission, and to a lesser extend a future decline in quality of life, but 50 51 no studies have investigated the risk for the above-mentioned combined endpoints, 52 137 53 54 138 or involved polypharmacy-related predictors.[27] 55 56 139 The aim of this exploratory study was to develop and internally validate prognostic 57 58 140 models to predict the risk of adverse events or a decline in well-being in general 59 60 141 practice patients with multimorbidity and polypharmacy, and to operationalize these 5 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 7 of 90 BMJ Open
1 2
3 142 negative health outcomes in terms of hospitalization, falls, level of required nursing BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 143 care, institutionalization and health-related quality of life (HRQoL). The models were 7 8 144 based on morbidity and medication-related variables, as well as sociodemographic 9 10 145 characteristics and parameters of health care utilization. 11 12 146 13 14 15 147 METHODS 16 17 148 We developed and internally validated prognostic models to identify key health 18 For peer review only 19 149 problems linked with multimorbidity and associated polypharmacy (decline in well- 20 21 22 150 being and adverse events: Figure 1). (1) Based on claims data, we predicted the 23 24 151 combined endpoint of hospitalization, falls / fall-related injuries, need for nursing care, 25 26 152 deterioration in the required level of care (nursing level), or institutionalization, after 27 28 six, nine and 24 months. (2) We predicted health-related quality of life (HRQoL) after 29 153 30 31 154 six and nine months based on data from a cluster-randomized trial.[23] 32 33 155 34 35 156 [About here: Figure 1: Predicted outcomes with regard to general trajectories of well- 36
37 http://bmjopen.bmj.com/ 38 157 being and quality of life in patients with multimorbidity and polypharmacy over time 39 40 158 and Figure 2: Models and sensitivity analyses with regard to data source and 41 42 159 predictor set] 43 44
45 160 on September 26, 2021 by guest. Protected copyright. 46 47 161 Design and Setting / Study Samples 48 49 162 Two datasets were used in modelling: 50 51 Claims data obtained from the Techniker Krankenkasse (TK) statutory health 52 163 53 54 164 insurance company between 01/2012 and 12/2014. TK is the largest statutory health 55 56 165 insurer in Germany and provided health insurance to 8.1 million persons in 2012.[28] 57 58 166 In accordance with Social Code book V, all statutory health insurance companies in 59 60 167 Germany collect basic data on socio-demographics, details of pharmacological and 6 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 8 of 90
1 2
3 168 non-pharmacological prescriptions and information on other health services utilization BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 169 and data on morbidity. 7 8 170 Trial data from the cluster-randomized PRIMUM (PRIoritising MUltimedication in 9 10 171 Multimorbidity) trial [23] conducted in general practices in Hesse, Germany from 11 12 172 08/2010 to 02/2012. 13 14 15 173 16 17 174 Population 18 For peer review only 19 175 Claims-based models: We aimed to use the same inclusion criteria for both datasets 20 21 22 176 as far as possible. We therefore included health insurance claims data of older 23 24 177 patients (≥60 years) with multimorbidity (at least three documented chronic diseases, 25 26 178 from a list of 46 diagnoses and conditions, from 01/01/2012 to 31/12/2012) [29] and 27 28 polypharmacy (at least five documented and concurrent prescriptions from 29 179 30 31 180 01/07/2012 to 31/12/2012). Included patients had to have been continuously insured 32 33 181 by TK from 01/01/2012 to 31/12/2014 (except in case of death at any time after 34 35 182 31/12/2012) and had to have contacted a primary care provider at least once in 2012. 36
37 http://bmjopen.bmj.com/ 38 183 Patients were excluded if they were diagnosed with dementia (ICD-10: F00-03, 39 40 184 F05.1, G30-31, R54) or under guardianship from 01/01/2012 to 31/12/2012. 41 42 185 43 44
45 186 Trial data-based models: We included data from patients that participated in the on September 26, 2021 by guest. Protected copyright. 46 47 187 cluster-randomized PRIMUM trial (N=5032, intervention group: n=252, control group 48 49 188 n=2501).[23] Patients with multimorbidity and polypharmacy were included in the 50 51 study if they were at least 60 years old, had at least three chronic diseases from two 52 189 53 54 190 or more chapters of ICD-10, and at least five prescriptions. Patients were excluded if 55 56 191 they were cognitively impaired (defined as a score lower or equal to 26 on the 57 58 192 MiniMental Status Exam [30]), had an alcohol or drug addiction, or were not able to 59 60
7 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 9 of 90 BMJ Open
1 2
3 193 participate in telephone interviews, fill in questionnaires or express their own free will. BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 194 Four out of the 502 patients (0.79%) died during the nine-month follow-up period. 7 8 195 9 10 196 Outcomes 11 12 197 Models based on claims data: we predicted the combined endpoint of hospitalization, 13 14 15 198 falls / fall-related injuries, or institutionalization in a long-term care facility, or if the 16 17 199 need for nursing care was recognised, or the level of care (“Pflegestufe”) had 18 For peer review only 19 200 worsened at 6-, 9-, 24-month follow-up. We treated the parameters of health service 20 21 22 201 use (hospitalization, level of nursing care and institutionalization) as surrogate 23 24 202 parameters for a decline in functional status and well-being, as details of these are 25 26 203 not included in German claims data. Outcomes were operationalized as follows: 27 28 29 204 Hospitalization: We included all-cause hospitalizations, as our data did not 30 31 205 permit us to differentiate between unplanned and elective hospitalizations. 32 33 206 Falls and fall-related injuries: We included all fractures and injuries coded in 34 35 36 207 ICD-10 chapters “S” and “T”. We excluded ICD codes for severe body injuries
37 http://bmjopen.bmj.com/ 38 208 such as S31 (“open wound of abdomen, lower back and pelvis”), which we 39 40 209 assessed as related to severe bodily impact, rather than drug-related falls (see 41 42 210 Additional file 1 for all excluded ICD codes). We also excluded osteoporosis- 43 44
45 211 related fractures (ICD-10 M80). on September 26, 2021 by guest. Protected copyright. 46 47 212 Institutionalization was defined as the admission of a patient to a long-term 48 49 213 care facility for at least 28 days (in Germany, this is the maximum length of 50 51 52 214 time considered as ‘short-term care’ in such facilities). 53 54 215 Level of (nursing) care (“Pflegestufe”) referred to dependency on care. In the 55 56 216 period under review, the German nursing care insurance system recognized 57 58 59 217 four levels of care (“1” – lowest level to “3” – highest level, and “H”, which was 60
8 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 10 of 90
1 2
3 218 mainly used for people with mental illnesses who are in need for support). The BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 219 onset of care and any increase in care level were taken into consideration. 7 8 220 Models based on trial data: we predicted HRQoL six and nine months after baseline. 9 10 221 HRQoL was measured using the EQ5D-3L index score.[31–33] The EQ5D-3L index 11 12 222 score is a weighted summary score of five different dimensions of health (mobility, 13 14 15 223 self-care, usual activities, pain/discomfort, and anxiety/depression). Each dimension 16 17 224 has three levels. The index score is calculated based on TTO norm values and 18 For peer review only 19 225 ranges from 0 to 1, with “0” signifying death and “1” in full health. Patients who died 20 21 22 226 during follow-up were assigned the value “0”. 23 24 227 25 26 228 Potential predictors 27 28 The potential predictors that were initially used in the two modelling approaches were 29 229 30 31 230 on available in both claims and trial data (‘core predictors’, see Figure 2): To compare 32 33 231 the two models, we first used these ’core predictors’ (all variables were continuous 34 35 232 variables, if not stated otherwise). 36
37 http://bmjopen.bmj.com/ 38 233 Socio-demographics: age (in years), sex (m/f, binary) 39 40 234 Morbidity-related (excluding dementia): number of chronic diseases (based on 41 42 235 a modified list of 46 diagnoses and conditions), [29] Charlson comorbidity 43 44
45 236 index, [34] number of specific chronic conditions according to Diederichs’ list on September 26, 2021 by guest. Protected copyright. 46 47 237 [35] consisting of 17 chronic diseases identified in a systematic review of 48 49 238 existing comorbidity indices. As dementia was excluded, the final list contained 50 51 52 239 16 diagnoses. (All instruments including ICD-10 codes are provided in 53 54 240 Additional file 2.) 55 56 241 Medication: number of prescriptions (defined as Anatomical Therapeutic 57 58 59 242 Chemical (ATC) agents using 5th-level coding, ATC version 2011 to 2014), 60
9 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 11 of 90 BMJ Open
1 2
3 243 excluding drugs for topical applications and drug groups that were irrelevant to BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 244 our research question, e.g. contrast agents (ATC V-08, 3-digit level). 7 8 245 Potentially Inappropriate Medication (PIM): we constructed two patient co- 9 10 246 variables: (1) exposure to any PIM (yes/no) and (2) number of PIMs between 11 12 01/07/2012 and 31/12/2012 (claims-based models) and at baseline (trial data- 13 247 14 15 248 based models). We used the following two lists to identify PIMs: 16 17 249 o Modified EU-PIM list [36]: The list of PIMs for the elderly contains 282 18 For peer review only 19 250 chemical substances or drug classes divided into 34 therapeutic 20 21 22 251 groups. 23 24 252 o Modified PRISCUS list [37]: The German list of PIMs for the elderly 25 26 253 includes 83 chemical substances from a total of 18 drug classes. 27 28 29 254 We excluded from the lists PIMs that referred to specific doses, treatment 30 31 255 duration, and disease severity, as valid information on these could not be 32 33 256 obtained from the claims data. (All instruments including ATC codes are 34 35 provided in Additional file 3.) 36 257
37 http://bmjopen.bmj.com/ 38 258 Anticholinergic drug burden: scores were calculated based on all prescribed 39 40 259 drugs with anticholinergic properties per patient. Despite substantial 41 42 260 differences between existing scales, associations with adverse clinical 43 44
45 261 outcomes, such as hospital admissions, fall-related hospitalizations, length of on September 26, 2021 by guest. Protected copyright. 46 47 262 stays in hospital, and GP visits, have been found for all of them.[38] As the 48 49 263 evidence does not support the preferred use of any particular scale, we tested 50 51 52 264 the following (all instruments including ATC codes are provided in Additional 53 54 265 file 3): 55 56 266 o Anticholinergic Drug Scale (ADS) [39]: The ADS weights anticholinergic 57 58 59 267 properties per drug from ‘0’ – no anticholinergic activity, ‘1’ – mild, ‘2’ – 60 268 moderate to ‘3’ – strong anticholinergic activity. The overall 10 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 12 of 90
1 2
3 269 anticholinergic burden per patient was calculated as a sum score for the BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 270 entire medication regimen. 7 8 271 o Modified Anticholinergic Drug Burden Index (DBI) [13]: The DBI 9 10 272 comprises drugs with sedative effects (which form the sedative burden 11 12 273 (B )), and drugs with anticholinergic or both sedative and 13 S 14 15 274 anticholinergic effects (which form the anticholinergic burden (BAC)). As 16 17 275 claims data do not provide dosages, the cumulative number of sedative 18 For peer review only 19 276 and anticholinergic drugs was calculated (modified DBI score). 20 21 22 277 Healthcare utilization: For each patient, we obtained information on all-cause 23 24 278 hospitalizations (yes/no), falls and fall-related injuries (yes/no) and the number 25 26 279 of physicians involved in ambulatory health care, between 01/01/2012 and 27 28 29 280 31/12/2012 for models based on claims data, and in the 6 months previous to 30 31 281 baseline for models based on trial data. 32 33 282 34 35 Additional potential predictor variables were used exclusively to re-fit models 36 283
37 http://bmjopen.bmj.com/ 38 284 based on trial data, as they were only available in these data (‘additional 39 40 285 predictors’, see Figure 2; all variables were continuous variables unless stated 41 42 286 otherwise): 43 44
45 287 Socio-demographics: Education (Comparative Analysis of Social Mobility in on September 26, 2021 by guest. Protected copyright. 46 47 288 Industrial Nations, CASMIN [40]) and number of persons living in the 48 49 289 household 50 51 52 290 Lifestyle: Alcohol consumption (audit-C, categorical variables on number of 53 54 291 drinking occasions and amount of alcohol consumed), [41] smoking status 55 56 292 (smoker/non-smoker, binary) and body mass index 57 58 59 293 Inappropriateness of medication: Medication Appropriateness Index (MAI) 60 294 consists of 10 items (indication, effectiveness, correctness of dosage, 11 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 13 of 90 BMJ Open
1 2
3 295 correctness of direction, practicality of direction, drug–drug interactions, drug– BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 296 disease interactions, unnecessary drug duplications, correctness of treatment 7 8 297 duration and costs). [42] The MAI item on cost was omitted because variable 9 10 298 discount contracts between pharmaceutical companies and statutory health 11 12 299 insurers preclude cost comparisons in Germany. The medication reviews were 13 14 15 300 conducted by a trained clinical pharmacologist (SH), who rated nine items for 16 17 301 each prescription. Values ranged from ‘0’ (appropriate) to ‘2’ (inappropriate) 18 For peer review only 19 302 whereby ‘1’ represented a middle rating of uncertain appropriateness. The 20 21 22 303 assigned values were summed to give an MAI score between 0 and 18 for 23 24 304 each prescription and across the entire medication regimen of the patient.[23]] 25 26 305 Morbidity-related: Severity of multimorbidity, as measured using the 27 28 29 306 Cumulative Illness Rating Scale (the CIRS differentiates between 14organ 30 31 307 systems, which are assessed on a 5-point Likert scale according to severity of 32 33 308 impairment, with the ratings ranging from no impairment to extreme 34 35 impairment), [43] with scores calculated as the total sum score, the number of 36 309
37 http://bmjopen.bmj.com/ 38 310 affected organ systems, and the Health-Related Quality of Life comorbidity 39 40 311 index (HRQoL-CI consists of a mental and a physical subscale, whereby the 41 42 312 presence of certain diseases are assigned weights from ‘1’ to ‘3’, see 43 44
45 313 Additional file 2) [44] on September 26, 2021 by guest. Protected copyright. 46 47 314 Depressive symptoms, as measured using the Geriatric Depression Scale with 48 49 315 15 items (GDS) [45] 50 51 52 316 HRQoL at baseline, as measured using the EQ5D-3L index score [31–33] 53 54 317 55 56 318 Missing values and imputation 57 58 59 319 There were no missing values in the claims data, so no imputation was carried out in 60 320 models that were based on them. In models based on trial data, imputation of 12 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 14 of 90
1 2
3 321 missing values in predictors and outcomes was conducted using multiple imputation BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 322 via chained equations (MICE).[46, 47] We used a fully conditional specification 7 8 323 approach by setting up an appropriate conditional density for each variable. In the 9 10 324 imputation process, we included all variables that were used in each model. We 11 12 325 imputed m=50 datasets and combined the results using ‘Rubin’s rules’.[46] 13 14 15 326 16 17 327 Statistical analyses 18 For peer review only 19 328 In both models, we first investigated the core predictors that were available in both 20 21 22 329 datasets, including socio-demographics, morbidity- and medication-related variables, 23 24 330 and variables for healthcare utilization. We then refitted the trial data-based models 25 26 331 using the additional predictors that were exclusively available for trial data, such as 27 28 variables for lifestyle and well-being (see Figure 2). 29 332 30 31 333 32 33 334 Models based on claims data: In order to develop a prediction model for the binary 34 35 335 combined outcome (containing all-cause hospitalization, falls / fall-related injuries, 36
37 http://bmjopen.bmj.com/ 38 336 institutionalization or level of (nursing) care required) at 6-, 9- and 24-month follow- 39 40 337 up, we performed multiple logistic regression analyses with the occurrence of at least 41 42 338 one of the components at 6-, 9-, and 24-month follow-up as the dependent variable. 43 44
45 339 As patients were not always assigned a single general practice,[48] we did not on September 26, 2021 by guest. Protected copyright. 46 47 340 perform cluster analysis on the claims data. 48 49 341 50 51 Models based on trial data: In order to develop a prediction model for the continuous 52 342 53 54 343 outcome HRQoL at 6- and 9-month follow-up, we performed multiple linear 55 56 344 regression analyses using the EQ5D-3L index score at 6- and 9-month follow-up as 57 58 345 the dependent variable. The cluster structure of the data was taken into account by 59 60
13 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 15 of 90 BMJ Open
1 2
3 346 including a random intercept to produce a mixed regression model. We assumed a BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 347 compound symmetry structure when estimating the covariance matrix. 7 8 348 9 10 349 Univariate analyses in both claims and trial data: Prior to conducting regression 11 12 350 analyses, we performed univariate analyses to identify any associations between our 13 14 15 351 potential predictors (at baseline) and the outcomes (at 6-, 9-, and 24-month follow- 16 17 352 up). 18 For peer review only 19 353 20 21 22 354 Regression analyses and variable selection: To find out which predictor variables 23 24 355 influence the outcome variables, we used a stepwise variable selection procedure 25 26 356 (combining forward and backward steps). We started with the full model and all 27 28 potential predictor variables. After this, we used a selection procedure based on p- 29 357 30 31 358 values.[49] In the backward selection step, we deleted the variable with the highest 32 33 359 p-value from the model if its p value was greater than 0.157. In the forward selection 34 35 360 step, the variable with the lowest p-value was included in the model if its p-value was 36
37 http://bmjopen.bmj.com/ 38 361 less than 0.156. As long as each covariate had only one degree of freedom, the use 39 40 362 of these boundaries led to the same results as variable selection using the Akaike 41 42 363 Information Criterion (AIC).[50] The resulting models are presented by providing the 43 44
45 364 estimated regression coefficients (models based on trial data) or odds ratios (OR, on September 26, 2021 by guest. Protected copyright. 46 47 365 models based on claims) with 95% confidence intervals and corresponding p-values. 48 49 366 As we expected the large sample size of claims-based models to result in low p- 50 51 values, we calculated additional z values and continuous net reclassification indices 52 367 53 54 368 (NRI) to gain information on the predictive power of each variable.[51] Multi- 55 56 369 collinearity was assessed using the variance inflation factor (VIF).[52] In the models 57 58 370 based on trial data, we did not account for the clustering structure when we 59 60 371 calculated the VIF. 14 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 16 of 90
1 2
3 372 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 373 Performance of the models 7 8 374 We calculated R2 for linear models based on trial data (according to Nakagawa and 9 10 375 Schielzeth [53]), and Nagelkerke’s R2 for logistic models (according to Steyerberg 11 12 376 and Nagelkerke [54, 55]) based on claims data. Furthermore, in order to assess 13 14 15 377 performance more realistically and to internally validate the models, we used the 16 17 378 AUC (area under the receiver operator curve, equivalent to the concordance index) to 18 For peer review only 19 379 validate the logistic regression model based on claims data, and R2 to validate the 20 21 22 380 linear regression model based on RCT data, in combination with 10-fold cross- 23 24 381 validation.[56] R2 and Nagelkerke’s R2 are measures of the overall model’s ability to 25 26 382 assess explained variance. The AUC provides a measure of the model’s 27 28 discriminatory ability to distinguish patients at risk from those that are not. 29 383 30 31 384 32 33 385 Sensitivity analyses 34 35 386 Using sensitivity analysis, we applied two further modelling approaches (at first 36
37 http://bmjopen.bmj.com/ 38 387 separately and then in combination): 1) modelling without multiple imputation and 2) 39 40 388 modelling without variable selection. 41 42 389 43 44
45 390 Software: We made use of different statistical packages to analyse the data in R.[47, on September 26, 2021 by guest. Protected copyright. 46 47 391 57–63] 48 49 392 We used TRIPOD reporting guidelines (transparent reporting of a multivariable 50 51 prediction model for individual prognosis or diagnosis) in the preparation of this 52 393 53 54 394 manuscript.[64] 55 56 395 57 58 396 Patient and Public Involvement statement 59 60 397 Neither patients nor the public were involved in this study. 15 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 17 of 90 BMJ Open
1 2
3 398 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 399 RESULTS 7 8 400 Participants 9 10 401 Claims data 11 12 402 The total sample of those ≥60 years that were continuously insured by TK from 13 14 15 403 01/01/2012 to 31/12/2014, and had at least one primary care contact during 2012, 16 17 404 amounted to 1,377,917 persons. Overall, 592,456 patients met the pre-specified 18 For peer review only 19 405 criteria and were included in the analyses (see study flow-chart, Additional file 4). 20 21 22 406 Trial data 23 24 407 Of the 505 patients that participated in the PRIMUM trial, all but 3 were 60 years or 25 26 408 older. The final analyses therefore included 502 patients. 27 28 Key characteristics of study participants are shown in table 1. 29 409 30 31 410 32 33 411 Table 1: Characteristics of study participants 34 35 Characteristic Claims data* CRT data* 36
37 N=592,456 N=502 http://bmjopen.bmj.com/ 38 Data collection period January 2012 to August 2010 to 39 December 2014# February 2012 40 Study design Cohort study Cluster-RCT 41 Setting Claims data from the TK 72 General Practices 42 health insurance fund. TK in Hesse, Germany 43 44 serves about 10 Mio.
45 people in Germany on September 26, 2021 by guest. Protected copyright. 46 Inclusion criteria ≥60 years ≥60 years 47 ≥3 chronic diseases ≥3 chronic diseases 48 ≥5 prescriptions ≥5 prescriptions 49 ≥1 GP visit ≥1 GP visit 50 51 Continuously insured 52 (except in case of death in 53 follow-up period) 54 Exclusion criteria Person under legal Person under legal 55 guardianship guardianship 56 Diagnosed dementia Cognitive dysfunction 57 including dementia 58 59 (MMSE ≤ 26) 60
16 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 18 of 90
1 2 † 3 Outcomes to be predicted Combined binary HRQoL (continuous BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 outcome after 6-, 9-, 24- outcome) after 6- and 5 6 month follow-up 9-month follow-up 7 Potential predictors in both samples at baseline** 8 Age (years) 71.3 (7.06) 72.2 (6.86) 9 Male sex (n, %) 319,453 (54) 240 (48) 10 Morbidity: 11 . Disease count 9.7 (3.75) 9.6 (3.25) 12 13 . No. of specific chronic 4.3 (1.97) 4.1 (1.60) 14 diseases (Diederichs) 15 . CCI 3.0 (2.54) 2.6 (1.92) 16 . HRQoL-CI, mental 2.8 (2.12) 2.1 (1.81) 17 . HRQoL-CI, physical 8.0 (3.57) 7.6 (3.12) 18 Medication: For peer review only 19 20 . No. of drugs 8.6 (3.80) 8.1 (2.57) 21 . No. of PIM (EU-PIM) 1.1 (1.15) 0.9 (0.96) 22 . ACh burden (ADS) 1.0 (1.45) 0.8 (1.21) 23 . Mod. Drug Burden Index 0.8 (1.03) 0.5 (0.77) 24 No. of involved physicians 9.95 (5.26) 2.6 (1.77) 25 Previous Hospitalization: 26 ‡ 27 . Patients that have 194,984 (33) 81 (16) 28 undergone hospital 29 treatment (n, %) 30 . No. of hospitalizations 1.67 (1.25) 1.5 (0.86) ‡ 31 . No. of days in hospital 14.5 (18.20) 17 (12.66) ‡ 32 Patients with previous 163,387 (28) 83 (17) ‡ 33 falls/fall-related injuries (n, %) 34 35 Patients requiring nursing 36 care:
37 . Any nursing level (n, %) 28,310 (5) - http://bmjopen.bmj.com/ 38 . Nursing level 1 (n, %) 19,030 (3) - 39 . Nursing level 2 (n, %) 7,968 (1) - 40 . Nursing level 3 (n, %) 1,273 (0.2) - 41 42 . Nursing level H (n, %) 39 (0.007) - 43 Additional predictor variables in CRT data at baseline** 44 Socio-demographics 45 . Educational level - 1.4 (0.66) on September 26, 2021 by guest. Protected copyright. 46 (CASMIN) 47 . No. of persons living in - 1.8 (0.70) 48 49 household 50 Lifestyle 51 . Alcohol intake (AUDIT C) - 1.9 (1.96) [mv: 39] 52 . Smoker (n, %) - 46 (10) [mv: 25] 53 . Body Mass Index - 30.1 (6.58) 54 Morbidity: 55 56 . CIRS sum score 7.7 (4.56) 57 . CIRS, no. of organ 4.5 (2.35) 58 systems 59 . Depressive Symptoms 2.4 (2.29) [mv: 8] 60 (GDS)
17 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 19 of 90 BMJ Open
1 2
3 Medication: BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 . MAI - 4.7 (5.56) 5 6 HRQoL: 7 . EQ5D-3L Index Score - 74.3 (23.72) [mv: 24] 8 412 9 10 11 413 Legend: *Values are arithmetic means and standard deviations unless otherwise 12 13 414 indicated. #The anamnestic period for baseline data ran from 01/01/2012 to 14 15 16 415 31/12/2012, except for medication data, for which it ran from 01/07/2012 to 17 18 416 31/12/2012. The follow-upFor peer period started review on 01/01/2013. only †Combined outcome 19 20 417 included hospitalization, fall/fall-related injuries, institutionalization and care level. ‡6 21 22 months before study entry. **Number of patients with missing values (mv) is zero 23 418 24 25 419 unless indicated in square parentheses. 26 27 420 Abbreviations: ACh burden – Anticholinergic drug burden, ADS – Anticholinergic 28 29 421 Drug Scale, AUDIT - Alcohol Use Disorders Identification Test (WHO), CASMIN - 30 31 32 422 Comparative Analysis of Social Mobility in Industrial Nations, CCI - Charlson 33 34 423 Comorbidity Index, CIRS – Cumulative Illness Rating Scale, GDS – Geriatric 35 36 424 Depression Scale, GP – General Practitioner, HRQoL – Health-Related Quality of 37 http://bmjopen.bmj.com/ 38 39 425 Life, HRQoL-CI – Health-Related Quality of Life Comorbidity Index, MAI – Medication 40 41 426 Appropriateness Index, MMSE – Mini Mental Status Exam, PIM – Potentially 42 43 427 Inappropriate Medication, CRT – cluster-randomized controlled trial. 44
45 on September 26, 2021 by guest. Protected copyright. 46 428 47 48 49 429 Univariate Analyses 50 51 430 In the claims data, univariate analyses revealed significant associations between the 52 53 431 combined outcome and the following predictors: age, sex, disease count, Charlson 54 55 56 432 comorbidity index, EU-PIMs, ADS, DBI, previous hospitalizations, previous falls and 57 58 433 number of physicians involved in the patient’s care at all follow-ups (after six, nine 59 60 434 and 24 months) (Additional file 5). In the trial data, HRQoL was significantly
18 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 20 of 90
1 2
3 435 correlated with the shared predictor variables disease count, number of chronic BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 436 prescriptions, previous falls and sex, and the additional predictors depression and 7 8 437 HRQoL at baseline (Additional file 6). 9 10 438 11 12 439 Prognostic models 13 14 15 440 Claims data 16 17 441 The model predicting the combined endpoint at 6 months had the highest C-statistic 18 For peer review only 19 442 (AUC with 10-fold cross validation: 0.71, see table 2), but a low explanation of 20 21 2 22 443 variance (Nagelkerke’s R without cross validation: 0.16). Variables in the model with 23 24 444 the highest predictive power were previous falls/fall-related injuries and previous 25 26 445 hospitalizations, as well as age, number of involved physicians, and number of 27 28 chronic diseases (“disease count”) (see Table 3). The models predicting the 29 446 30 31 447 combined outcome at nine and 24 months had AUCs calculated with 10-fold cross 32 33 448 validation of 0.68 (R² without cross validation: 0.15) and 0.69 (R² without cross 34 35 449 validation: 0.13) respectively. The VIF (to assess any multi-collinearity) showed 36
37 http://bmjopen.bmj.com/ 38 450 moderate values (max. 7.5). 39 40 451 41 42 452 Trial data 43 44
45 453 All results presented in this section are based on the modelling approach and involve on September 26, 2021 by guest. Protected copyright. 46 47 454 multiple imputation of missing values and the variable selection procedure. Models 48 49 455 predicting the HRQoL endpoint at 6 months that were based on core predictors 50 51 available in both claims and trial data showed low predictive accuracy (R2 with 10- 52 456 53 54 457 fold cross validation: 0.111) (see Table 3, model 2.4). HRQoL at 6 months was best 55 56 458 predicted when additional predictors that were exclusively available in the trial data 57 58 459 were also included (R2 with 10-fold cross validation: 0.507). The variables with the 59 60 460 highest predictive power were depressive symptoms (GDS) and EQ5D-3L Index 19 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 21 of 90 BMJ Open
1 2
3 461 Score (Baseline). MAI was also predictive (see Table 3, model 3.4). The VIF showed BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 462 small values (max. 2.2). 7 8 463 9 10 464 Comparison of model quality and sensitivity analyses 11 12 465 The shorter the time span of the prediction, the better the explained variance and 13 14 15 466 hence, the performance of the model. However, model performance remained fair to 16 17 467 poor when it only included predictor variables that were available for both claims and 18 For peer review only 19 468 trial data. Sensitivity analyses confirmed these results (see Table 2). 20 21 22 469 23 24 25 26 27 28 29 30 31 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
20 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
BMJ Open Page 22 of 90
1 2 3 Table 2: Comparison of models 4 5 2 6 Models based on claims data: core predictors AUC* R 7 1.1 Combined outcome after 6 months 0.71 {0.70} 0.16 {0.16} 8 1.2 Combined outcome after 9 months 0.69 {0.69} 0.15 {0.14} 9 10 1.3 Combined outcome after 24 months 0.68 {0.68} 0.13 {0.12} 11 Models based on CRT data: core predictors# AIC R2 R2 (10x) 12 EQ5D-3L after 6 months For peer review only 13 14 2.1 No imputation, no variable selection 4,138.86 {4,069.41} 0.155 {0.159} 0.112 {0.103} 15 2.2 No imputation, with variable selection 4,138.81 {4,068.69} 0.150 {0.155} 0.129 {0.122} 16 2.3 With imputation, no variable selection 4,582.30 {4,507.71} 0.159http://bmjopen.bmj.com/ {0.163} 0.094 {0.108} 17 2.4 With imputation, with variable selection 4,583.15 {4,507.47} 0.919 {0.925} 0.111 {0.128} 18 19 EQ5D-3L after 9 months 20 2.5 No imputation, no variable selection 3,917.75 {3,917.75} 0.150 {0.150} 0.030 {0.030} 21 2.6 No imputation, with variable selection 3,921.95 {3,921.95} 0.146 {0.146} 0.053 {0.053} 22 23 2.7 With imputation, no variable selection 4,540.58 {4,505.52} 0.156 {0.152} 0.090 {0.093} 24 2.8 With imputation, with variable selection 4,546.42 {4,511.10} 0.221 {0.218} 0.107 {0.106} on September 26, 2021 by guest. Protected copyright. 25 Models based on CRT data: core predictors and additional 26 predictors# 27 EQ-5D, after 6 months 28 29 3.1 No imputation, no variable selection 3,205.13 {3,205.13} 0.034 {0.034} 0.442 {0.442} 30 3.2 With imputation, no variable selection 4,308.94 {4,308.94} 0.538 {0.538} 0.481 {0.481} 31 3.3 No imputation, with variable selection 3,197.37 {3,197.37} 0.526 {0.526} 0.483 {0.483} 32 3.4 With imputation, with variable selection† 4,307.47 {4,307.47} 0.677 {0.677} 0.507 {0.507} 33 34 Models with "fixed variables" 35 3.5 No imputation, no variable selection 3,208.58 {3,208.58} 0.514 {0.514} 0.468 {0.468} 36 3.6 With imputation, with variable selection 4,308.90 {4,308.90} 0.665 {0.665} 0.499 {0.499} 37 EQ-5D, after 9 months 38 39 3.7 No imputation, no variable selection 3,061.06 {3,113.53} 0.042 {0.028} 0.411 {0.409} 40 41 42 21 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Page 23 of 90 BMJ Open
1 2 3 3.8 With imputation, no variable selection 4,307.28 {4,361.36} 0.498 {0.477} 0.433 {0.404} 4 3.9 No imputation, with variable selection 3,062.03 {3,108.61} 0.490 {0.485} 0.448 {0.443} 5 6 3.10 With imputation, with variable selection 4,309.88 {4,360.32} 0.453 {0.346} 0.455 {0.431} 7 Models with fixed variables 8 3.11 No imputation, no variable selection 3,064.76 {3,113.08} 0.490 {0.485} 0.439 {0.434} 9 3.12 With imputation, with variable selection 4,310.92 {4,363.62} 0.113 {0.071} 0.447 {0.423} 10 11 12 # For peer review only 13 Legend: {sensitivity analyses}, Models based on RCT data: fixed effects, †Best Overall Model; Abbreviations: AUC* - Area under 14 15 the curve after 10-fold cross validation, AIC – Aikaike’s information criterium, R² - Nagelkerke’s R², R2 (10x) – Nagelkerke’s R² with 10- 16 http://bmjopen.bmj.com/ 17 18 fold cross validation 19 20 21 22 Table 3: Best performing models per data set and set of predictors 23 24 on September 26, 2021 by guest. Protected copyright. 25 Best model based on claims data: core OR [95%-CI] p-value z-value NRI 26 predictors (model 1.1) 27 Age (years) 1.02 [1.02; 1.02] <0.001 45.4 0.11 28 29 Sex (female) 0.99 [0.97; 1.00] 0.025 -2.2 -0.03 30 Disease count 1.02 [1.02; 1.03] <0.001 19.2 0.05 31 CCI 1.03 [1.03; 1.04] <0.001 22.7 0.02 32 33 No. of specific chronic diseases (Diederichs) 1.01 [1.00; 1.01] <0.001 3.9 0.01 34 No. of PIM (EU-PIM) 1.03 [1.02; 1.03] <0.001 8.9 0.02 35 ACh burden (ADS) 1.04 [1.03; 1.05] <0.001 14.9 0.04 36 37 Mod. Drug Burden Index 1.08 [1.07; 1.08] <0.001 20.1 0.08 38 Previous hospitalizations 1.67 [1.65; 1.70] <0.001 82.3 0.34 39 Previous falls/fall-related injuries 3.29 [3.25; 3.34] <0.001 188.6 0.55 40 41 42 22 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
BMJ Open Page 24 of 90
1 2 3 No. of involved physicians 1.02 [1.02; 1.02] <0.001 29.0 0.08 4 5 Best model based on CRT data: Coefficient [95% CI] p-value 6 core predictors (model 2.4) 7 Intercept 101.18 [93.11; 109.25] <0.001 8 9 Sex (female) -11.26 [-15.59; -6.94] <0.001 10 No. of specific chronic diseases (Diederichs) -2.18 [-3.64; -0.73] 0.004 11 No. of drugs -1.28 [-2.23; -0.32] 0.010 12 Mod. Drug Burden Index For peer-5.19 review[-8.26; -2.12] only0.001 13 14 Previous falls -6.11 [-12.07; -0.15] 0.045 15 Best model based on CRT data: core Coefficient [95% CI] p-value 16 http://bmjopen.bmj.com/ 17 predictors and additional predictors (model 18 3.4, best overall model) 19 Intercept 51.74 [38.91; 64.57] <0.001 20 21 Sex (female) -3.61 [-6.96; -0.27] 0.036 22 No. of specific chronic diseases (Diederichs) -1.03 [-2.08; 0.01] 0.055 23 No. of involved physicians 0.80 [-0.13; 1.74] 0.093 24 on September 26, 2021 by guest. Protected copyright. 25 Body Mass Index (BMI) -0.28 [-0.53; -0.03] 0.031 26 Medication Appropriateness Index (MAI) -0.39 [-0.70; -0.08] 0.015 27 Depressive symptoms (GDS) -2.73 [-3.56; -1.91] <0.001 28 EQ-5D Index Score (Baseline) 0.55 [0.47; 0.64] <0.001 29 30 31 Legend: ACh burden – Anticholinergic drug burden, ADS - Anticholinergic Drug Scale, CCI - Charlson Comorbidity Index, GDS – 32 33 Geriatric Depression Scale, NRI - Continuous Net Reclassification Index, OR – Odds Ratio - PIM – Potentially Inappropriate 34 Medication 35 36 37 38 39 40 41 42 23 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 25 of 90 BMJ Open
1 2
3 470 DISCUSSION BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 471 Main results 7 8 472 Our best overall prognostic model predicted HRQoL after six months in older general 9 10 473 practice patients with multimorbidity and polypharmacy. It performed well, was based 11 12 474 on trial data and explained more than half of the variance. The most important 13 14 15 475 predictors were depressive symptoms, the initial level of health-related quality of life 16 17 476 and medication appropriateness (MAI) – all of which were only available as 18 For peer review only 19 477 ‘additional predictors’ in trial data. Prognostic models in trial data, which were 20 21 22 478 exclusively developed from ‘core predictors’ (available in both data sets) performed 23 24 479 worse, as well as claims based models and models based on both data sets that had 25 26 480 longer forecast periods (nine months or more). In both trial data- and claims-based 27 28 models, outcome components at baseline had a relatively high impact (i.e., HRQoL at 29 481 30 31 482 baseline in the trial data-based model and previous hospitalization and previous 32 33 483 falls/fall-related injuries in claims-based models). Although this is unsurprising and is 34 35 484 often the case in prognostic models,[65] it nonetheless seems reasonable to retain 36
37 http://bmjopen.bmj.com/ 38 485 the variables in the model. Furthermore, we identified further predictors, such as 39 40 486 depressive symptoms and medication appropriateness, which had a relatively high 41 42 487 predictive power. 43 44
45 488 on September 26, 2021 by guest. Protected copyright. 46 47 489 Comparison with the literature 48 49 490 The presented results are consistent with results from other studies. The AUC values 50 51 in our claims-based models (AUC 0.68-0.71) are comparable to those of 23 52 491 53 54 492 prognostic models for Case Finding conducted in elderly patients in primary care. 55 56 493 These models predicted (re)hospitalization, functional impairment, institutionalization 57 58 494 and death.[65] The quality of models with a low risk of bias was AUC 0.60-0.78, but 59 60 495 no explanation of variance was provided. The best model for predicting death within 24 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 26 of 90
1 2
3 496 four years (AUC: 0.82) included 12 predictors comprising age, sex, BMI, chronic BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 497 diseases, smoking status and functional parameters.[65] Models that included 7 8 498 additional trial data (e.g. clinical data) predicted endpoints better than models based 9 10 499 only on claims data.[65–67] In many models described in other studies, health care 11 12 500 utilization parameters, and especially previous hospitalizations, were predictive of 13 14 15 501 (re)hospitalizations, emergency admissions and functional impairment.[66, 68, 69] 16 17 502 The predictive power of sex is inconsistent: in 18/27 risk models, sex was included in 18 For peer review only 19 503 the final model [66]; in 7/23 risk models, male sex was predictive [65], while a further 20 21 22 504 25 studies found sex to have no influence.[68, 69] Model quality also improved in 23 24 505 studies that included multimorbidity and polypharmacy parameters.[66, 68, 70] 25 26 506 However, the parameters and instruments used in modelling (e.g. CIRS, Charlson 27 28 Comorbidity Index and disease count, as reported here) varied considerably among 29 507 30 31 508 studies. They were neither consistently predictive, nor were certain parameters or 32 33 509 instruments better than others.[66, 69, 70] 34 35 510 Most published models were developed to predict the risk of hospitalization.[66, 68– 36
37 http://bmjopen.bmj.com/ 38 511 74] Other models predicted functional outcomes,[70] while four models predicted 39 40 512 adverse drug reactions.[74] So far, little is known about the predictive power of 41 42 513 polypharmacy parameters and the appropriateness of prescriptions, especially the MAI 43 44
45 514 has never been used in prognostic models. Furthermore, no models have yet been on September 26, 2021 by guest. Protected copyright. 46 47 515 developed to predict health-related quality of life in patients with multimorbidity and 48 49 516 polypharmacy in general practice.[27, 70] 50 51 52 517 53 54 518 Strengths and Limitations 55 56 519 One strength of our study is that we could use two data sources with differing 57 58 520 advantages in our exploratory analysis: claims data contained a large number of 59 60 521 cases, and trial data provided additional high-quality patient data including functional 25 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 27 of 90 BMJ Open
1 2
3 522 status and HRQoL. Both datasets also have their limitations, since claims are BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 523 documented for billing purposes and are therefore imprecise, whereas our trial 7 8 524 dataset consisted of only a limited number of observations. Thus, each dataset 9 10 525 allows its own endpoints to be modelled. Risk modelling is especially complex in 11 12 526 multimorbid patients with polypharmacy, as predictor variables in this patient 13 14 15 527 collective are often associated with one another (e.g. diagnoses and prescriptions). In 16 17 528 addition, comparable risk situations can lead to different endpoints, as risk often 18 For peer review only 19 529 depends on context. For example, a drug-induced fall may have no health-related 20 21 22 530 consequences or may lead to impairment and institutionalization. 23 24 531 Further to these key limitations, our results need careful interpretation for several 25 26 532 reasons: Firstly, the combined endpoint in the claims-based models yielded a high 27 28 event rate, which may have resulted in overoptimistic results in our logistic 29 533 30 31 534 regression. However, other approaches would not have resolved this problem to suit 32 33 535 our purposes either. Additionally, we still have enough cases in both categories of the 34 35 536 dependent variable to conduct a valid model estimation. Nonetheless, the low 36
37 http://bmjopen.bmj.com/ 38 537 performance of the claims model may have been because predictors acted in 39 40 538 different ways on the different elements of the combined outcome, thus resulting in 41 42 539 greater heterogeneity.[75] Secondly, the small sample size of the trial population may 43 44
45 540 have led to some overfitting of the model. At the same time, the VIF (to assess any on September 26, 2021 by guest. Protected copyright. 46 47 541 multi-collinearity) showed only up to moderate values. The application of shrinkage 48 49 542 methods would have been a possible alternative to address this limitation.[76] 50 51 However, there is an ongoing debate whether it solves such problems, and a recent 52 543 53 54 544 study has suggested that although shrinkage can result in improved calibration, it 55 56 545 may not be superior in terms of reducing overfitting.[77] Furthermore, shrinkage 57 58 546 models lead to biased estimates of the regression coefficients, thus making results 59 60 547 more difficult to interpret. Thirdly, in our modelling approach we tested disease-based 26 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 28 of 90
1 2
3 548 indicators such as the CCI and CIRS that were developed and validated for other BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 549 purposes. However, we chose indicators that showed a strong association with 7 8 550 negative health outcomes.[35] 9 10 551 11 12 552 Relevance for primary care and research implications 13 14 15 553 As the models derived in our study have not been externally validated and our 16 17 554 methods have some limitations, we do not claim to have developed comprehensive 18 For peer review only 19 555 prognostic models to identify older general practice patients with multimorbidity and 20 21 22 556 polypharmacy at risk of negative health outcomes. For this reason, we plan to 23 24 557 conduct an individual patient data-based meta-analysis to further develop and 25 26 558 externally validate the models presented here (PROSPERO ID: CRD42018088129). 27 28 It is, however, very likely that baseline components of our predicted endpoints are 29 559 30 31 560 important predictors, especially considering these results are unsurprising and 32 33 561 entirely plausible. A decline in HRQoL, a previous hospitalization and a previous 34 35 562 falls/fall-related injury can therefore be seen as a warning parameter ('red flag') that 36
37 http://bmjopen.bmj.com/ 38 563 may help general practitioners in recognising older patients with multimorbidity and 39 40 564 polypharmacy at high risk of adverse health outcomes. These patients are therefore 41 42 565 more likely to benefit from an intervention than others with low or no risk.[24] Hence, 43 44
45 566 researchers evaluating polypharmacy interventions, such as medication reviews, may on September 26, 2021 by guest. Protected copyright. 46 47 567 like to consider our models when deciding upon selection and inclusion criteria for a 48 49 568 study population. 50 51 52 569 53 54 570 Conclusions 55 56 571 This study provides prognostic models to identify older general practice patients with 57 58 572 multimorbidity and polypharmacy at high risk of deterioration in health-related quality 59 60 573 of life, hospitalization, falls/fall-related injuries, institutionalization, and a need of 27 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 29 of 90 BMJ Open
1 2
3 574 nursing care. Outcome components, such as previous falls, hospital stays, reduced BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 575 health-related quality of life, and depression, were important predictors of these 7 8 576 negative health outcomes in our models. They can be seen as warning signs of 9 10 577 future worsening and an indication that these patients are likely to benefit from 11 12 578 interventions to optimize their medication. Future studies should externally validate 13 14 15 579 the models and evaluate the effectiveness of polypharmacy interventions in high risk 16 17 580 patients. 18 For peer review only 19 581 20 21 22 582 List of abbreviations 23 24 583 ACh burden: Anticholinergic drug burden, ADS: Anticholinergic Drug Scale, AIC: 25 26 584 Aikaike‘s information criterium, AUC*: Area under the curve after 10-fold cross 27 28 validation, AUDIT: Alcohol Use Disorders Identification Test (WHO), CASMIN: 29 585 30 31 586 Comparative Analysis of Social Mobility in Industrial Nations, CCI: Charlson 32 33 587 Comorbidity Index, CIRS: Cumulative Illness Rating Scale, GDS: Geriatric 34 35 588 Depression Scale, GP: General Practitioner, HRQoL: Health-Related Quality of Life, 36
37 http://bmjopen.bmj.com/ 38 589 HRQoL-CI: Health-Related Quality of Life Comorbidity Index, MAI: Medication 39 40 590 Appropriateness Index, MMSE: Mini Mental Status Exam, NRI: Continous Net 41 42 591 Reclassification Index, OR: Odds Ratio, PIM: Potential Inappropriate Medication, 43 44 2 (10x) 45 592 PRIMUM: Prioritising Multimedication in Multimorbidity, R²: Nagelkerke‘s R², R : on September 26, 2021 by guest. Protected copyright. 46 47 593 Nagelkerke’s R² with 10-fold cross validation, CRT: cluster-randomized controlled 48 49 594 trial, SD: Standard Deviation, TK: Techniker Krankenkasse, TRIPOD: Transparent 50 51 reporting of a multivariable prediction model for individual prognosis or diagnosis 52 595 53 54 596 55 56 597 57 58 598 Declarations 59 60 599 Ethics approval and consent to participate 28 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 30 of 90
1 2
3 600 Claims may be analysed by statutory health insurance companies in accordance with BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 601 § 284 of Social Code Book V. For the research questions of this project, Cologne 7 8 602 University, Goethe-University Frankfurt and Heidelberg University were analysed 9 10 603 collaboratively with TK. When claims are anonymously analysed in accordance with 11 12 604 Good Practice in Claims Data Analysis, [78] no further ethics vote is required. 13 14 15 605 Regarding our trial data, the ethics commission of the medical faculty of the Johann 16 17 606 Wolfgang Goethe University, Frankfurt / Main approved the PRIMUM trial (resolution 18 For peer review only 19 607 number E 46/10, file number 123/10, date: 20/05/2010) and all of the participants 20 21 22 608 gave their written informed consent before taking part. 23 24 25 609 26 27 610 Consent for publication 28 29 611 Not applicable. 30 31 32 612 33 34 613 Availability of data and material 35 36 614 The datasets generated and analysed in the current study are not publicly available, 37 http://bmjopen.bmj.com/ 38 39 615 as further analyses are ongoing. 40 41 616 42 43 617 Competing interests 44
45 on September 26, 2021 by guest. Protected copyright. 618 FG, BM, MB and CM received grants from the German Statutory Healthcare 46 47 48 619 Insurance Company Techniker Krankenkasse during the course of the study. CS has 49 50 620 been employed by Boehringer Ingelheim GmbH & Co. KG since October 2019. The 51 52 621 company had no role in the design, analysis or interpretation of the current study. 53 54 55 622 MvA, LU, PI, FvB, RP, PG and JV declare that they have no competing interests. 56 57 623 58 59 624 Funding 60
29 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 31 of 90 BMJ Open
1 2
3 625 This study was supported by the German Statutory Healthcare Insurance Company BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 626 Techniker Krankenkasse. 7 8 627 9 10 628 Authors' contributions 11 12 629 MB, FG and CM designed the study. PI, LU and CS analysed the data. BM, LU, PI, 13 14 15 630 CS, FvB, MB, FG, RP, JV, PG, MvdA and CM contributed to the interpretation of the 16 17 631 data. CM and BM drafted the manuscript and all authors revised it and subsequent 18 For peer review only 19 632 versions of the manuscript critically for important intellectual content. All authors 20 21 22 633 approved the version to be submitted for publication. LU, CS and CM had full access 23 24 634 to all data and are responsible for the integrity and the accuracy of the data analysis. 25 26 635 27 28 Acknowledgements 29 636 30 31 637 The authors would like to thank Phillip Elliott for the language review of the paper. 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
30 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 32 of 90
1 2
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 638 REFERENCES 5 639 1 Salisbury C, Johnson L, Purdy S, et al. Epidemiology and impact of multimorbidity in primary 6 640 care: A retrospective cohort study. Br J Gen Pract 2011;61(582):e12-21. 7 8 641 2 Moßhammer D, Haumann H, Mörike K, et al. Polypharmacy-an Upward Trend with 9 642 Unpredictable Effects. Dtsch Arztebl Int 2016;113(38):627–33. 10 643 3 Muth C, Blom JW, Smith SM, et al. Evidence supporting the best clinical management of 11 644 patients with multimorbidity and polypharmacy: A systematic guideline review and expert 12 13 645 consensus. J Intern Med 2019;285(3):272–88. 14 646 4 Palmer K, Marengoni A, Forjaz MJ, et al. Multimorbidity care model: Recommendations from 15 647 the consensus meeting of the Joint Action on Chronic Diseases and Promoting Healthy Ageing 16 17 648 across the Life Cycle (JA-CHRODIS). Health Policy 2018;122(1):4–11. 18 649 5 Nobili A, MarengoniFor A, Tettamanti peer M, etreview al. Association between only clusters of diseases and 19 650 polypharmacy in hospitalized elderly patients: results from the REPOSI study. Eur J Intern Med 20 651 2011;22(6):597–602. 21 22 652 6 Oscanoa TJ, Lizaraso F, Carvajal A. Hospital admissions due to adverse drug reactions in the 23 653 elderly. A meta-analysis. Eur J Clin Pharmacol 2017;73(6):759–70. 24 654 7 Angamo MT, Chalmers L, Curtain CM, et al. Adverse-Drug-Reaction-Related Hospitalisations in 25 26 655 Developed and Developing Countries: A Review of Prevalence and Contributing Factors. Drug 27 656 Saf 2016;39(9):847–57. 28 657 8 Deandrea S, Lucenteforte E, Bravi F, et al. Risk factors for falls in community-dwelling older 29 658 people: a systematic review and meta-analysis. Epidemiology 2010;21(5):658–68. 30 31 659 9 Heinrich S, Rapp K, Rissmann U, et al. Cost of falls in old age: a systematic review. Osteoporos 32 660 Int 2010;21(6):891–902. 33 661 10 Fried TR, O'Leary J, Towle V, et al. Health outcomes associated with polypharmacy in 34 35 662 community-dwelling older adults: a systematic review. J Am Geriatr Soc 2014;62(12):2261–72. 36 663 11 Payne RA, Abel GA, Avery AJ, et al. Is polypharmacy always hazardous? A retrospective cohort 37 664 analysis using linked electronic health records from primary and secondary care. Br J Clin http://bmjopen.bmj.com/ 38 665 Pharmacol 2014;77(6):1073–82. 39 40 666 12 Gnjidic D, Hilmer SN, Blyth FM, et al. Polypharmacy cutoff and outcomes: five or more 41 667 medicines were used to identify community-dwelling older men at risk of different adverse 42 668 outcomes. J Clin Epidemiol 2012;65(9):989–95. 43 44 669 13 Hilmer SN, Mager DE, Simonsick EM, et al. A drug burden index to define the functional burden
45 670 of medications in older people. Arch Intern Med 2007;167(8):781–87. on September 26, 2021 by guest. Protected copyright. 46 671 14 Salahudeen MS, Duffull SB, Nishtala PS. Anticholinergic burden quantified by anticholinergic 47 672 risk scales and adverse outcomes in older people: a systematic review. BMC Geriatr 2015;15:31. 48 49 673 15 Gurwitz JH, Field TS, Harrold LR, et al. Incidence and preventability of adverse drug events 50 674 among older persons in the ambulatory setting. JAMA 2003;289(9):1107–16. 51 675 16 Kuijpers MAJ, van Marum RJ, Egberts ACG, et al. Relationship between polypharmacy and 52 53 676 underprescribing. Br J Clin Pharmacol 2008;65(1):130–33. 54 677 17 Steinman MA, Landefeld CS, Rosenthal GE, et al. Polypharmacy and prescribing quality in older 55 678 people. J Am Geriatr Soc 2006;54(10):1516–23. 56 679 18 Meid AD, Quinzler R, Freigofas J, et al. Medication Underuse in Aging Outpatients with 57 58 680 Cardiovascular Disease: Prevalence, Determinants, and Outcomes in a Prospective Cohort 59 681 Study. PLoS One 2015;10(8):e0136339. 60
31 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 33 of 90 BMJ Open
1 2
3 682 19 Müller BS, Uhl MC, Nguyen Truc Sophia, et al. Patienten mit Multimedikation: Ambulante BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 683 Herausforderungen und Lösungswege: Eine qualitative Studie. Zeitschrift für Allgemeinmedizin 5 6 684 2018;94(10):396–400. 7 685 20 Horne R, Weinman J, Barber N, Elliott R, Morgan M, Cribb A, Kellar I. Concordance, adherence 8 686 and compliance in medicine taking: Report for the National Co-ordinating Centre for NHS 9 10 687 Service Delivery and Organisation R & D (NCCSDO) 2005. 11 688 21 Rankin A, Cadogan CA, Patterson SM, et al. Interventions to improve the appropriate use of 12 689 polypharmacy for older people. Cochrane Database Syst Rev 2018;9:CD008165. 13 690 22 Johansson T, Abuzahra ME, Keller S, et al. Impact of strategies to reduce polypharmacy on 14 15 691 clinically relevant endpoints: A systematic review and meta-analysis. Br J Clin Pharmacol 16 692 2016;82(2):532–48. 17 693 23 Muth C, Uhlmann L, Haefeli WE, et al. Effectiveness of a complex intervention on Prioritising 18 For peer review only 19 694 Multimedication in Multimorbidity (PRIMUM) in primary care: Results of a pragmatic cluster 20 695 randomised controlled trial. BMJ Open 2018;8(2):e017740. 21 696 24 Glasziou PP, Irwig LM. An evidence based approach to individualising treatment. BMJ 22 697 1995;311(7016):1356–59. 23 24 698 25 Murray SA, Kendall M, Mitchell G, et al. Palliative care from diagnosis to death. BMJ 25 699 2017;356:j878. 26 700 26 Lynn J. Living long in fragile health: the new demographics shape end of life care. Hastings Cent 27 28 701 Rep 2005;Spec No:S14-8. 29 702 27 National Institute for Health and Care Excellence. Multimorbidity: clinical assessment and 30 703 management: NICE guideline NG 56 2016. 31 704 28 Bundesministerium für Gesundheit. KM 6-Statistik 2013. Available at: 32 33 705 https://www.bundesgesundheitsministerium.de/themen/krankenversicherung/zahlen-und- 34 706 fakten-zur-krankenversicherung/mitglieder-und-versicherte.html Accessed April 23, 2020. 35 707 29 Schäfer I, Leitner E-C von, Schön G, et al. Multimorbidity patterns in the elderly: A new 36
37 708 approach of disease clustering identifies complex interrelations between chronic conditions. http://bmjopen.bmj.com/ 38 709 PLoS One 2010;5(12):e15941. 39 710 30 Folstein MF, Folstein SE, McHugh PR. "Mini-mental state". A practical method for grading the 40 711 cognitive state of patients for the clinician. J Psychiatr Res 1975;12(3):189–98. 41 42 712 31 Graf Schulenburg JM, Claes C, Greiner W, et al. Die deutsche Version des EuroQol-Fragebogens. 43 713 J Public Health (Germany) 2009;6(1):3–20. 44 714 32 Agborsangaya CB, Lahtinen M, Cooke T, et al. Comparing the EQ-5D 3L and 5L: Measurement 45 on September 26, 2021 by guest. Protected copyright. 46 715 properties and association with chronic conditions and multimorbidity in the general 47 716 population. Health Qual Life Outcomes 2014;12:74. 48 717 33 EuroQol--a new facility for the measurement of health-related quality of life. Health Policy 49 718 1990;16(3):199–208. 50 51 719 34 Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in 52 720 longitudinal studies: Development and validation. J Chronic Dis 1987;40(5):373–83. 53 721 35 Diederichs C, Berger K, Bartels DB. The measurement of multiple chronic diseases--a systematic 54 55 722 review on existing multimorbidity indices. J Gerontol A Biol Sci Med Sci 2011;66(3):301–11. 56 723 36 Renom-Guiteras A, Meyer G, Thürmann Pa. The EU(7)-PIM list: A list of potentially 57 724 inappropriate medications for older people consented by experts from seven European 58 59 725 countries. European Journal of Clinical Pharmacology 2015;71(7):861–75. 60 726 37 Holt S, Schmiedl S, Thürmann PA. Potentially Inappropiate medications in the Elderly: The 727 PRISCUS List. Dtsch Arztebl 2010;107(1):31–32. 32 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 34 of 90
1 2
3 728 38 Salahudeen MS, Hilmer SN, Nishtala PS. Comparison of anticholinergic risk scales and BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 729 associations with adverse health outcomes in older people. J Am Geriatr Soc 2015;63(1):85–90. 5 6 730 39 Carnahan RM, Lund BC, Perry PJ, et al. The Anticholinergic Drug Scale as a measure of drug- 7 731 related anticholinergic burden: Associations with serum anticholinergic activity. J Clin 8 732 Pharmacol 2006;46(12):1481–86. 9 10 733 40 Brauns H, Steinmann S (1999). Educational Reform in France, West-Germany, the United 11 734 Kingdom and Hungary.: Updating the CASMIN Educational Classification. ZUMA-Nachrichten, 12 735 1999:7–44. Available at: 13 736 www.gesis.org/fileadmin/upload/forschung/publikationen/zeitschriften/zuma_nachrichten/zn 14 15 737 _44.pdf Accessed April 23, 2020. 16 738 41 Saunders JB, Aasland OG, Babor TF, et al. Development of the Alcohol Use Disorders 17 739 Identification Test (AUDIT): WHO Collaborative Project on Early Detection of Persons with 18 For peer review only 19 740 Harmful Alcohol Consumption--II. Addiction 1993;88(6):791–804. 20 741 42 Hanlon JT, Schmader KE, Samsa GP, et al. A method for assessing drug therapy appropriateness. 21 742 J Clin Epidemiol 1992;45(10):1045–51. 22 743 43 Hudon C, Fortin M, Soubhi H. Abbreviated guidelines for scoring the Cumulative Illness Rating 23 24 744 Scale (CIRS) in family practice. J Clin Epidemiol 2007;60(2):212. 25 745 44 Mukherjee B, Ou H-T, Wang F, et al. A new comorbidity index: The health-related quality of life 26 746 comorbidity index. J Clin Epidemiol 2011;64(3):309–19. 27 28 747 45 Yesavage JA, Brink TL, Rose TL, et al. Development and validation of a geriatric depression 29 748 screening scale: A preliminary report. J Psychiatr Res 1982;17(1):37–49. 30 749 46 Rubin DB. Multiple imputation for nonresponse in surveys. Hoboken, N.J.: Wiley-Interscience 31 750 2004. 32 33 751 47 van Buuren S, Groothuis-Oudshoorn K. mice: Multivariate Imputation by Chained Equations in 34 752 R. J Stat Softw 2011;45(3). 35 753 48 Grandt D, Lappe V, Schubert I. Barmer Arzneimittelreport 2018. Wuppertal 2018. 36
37 754 49 Wood AM, White IR, Royston P. How should variable selection be performed with multiply http://bmjopen.bmj.com/ 38 755 imputed data? Stat Med 2008;27(17):3227–46. 39 756 50 Sauerbrei W. The Use of Resampling Methods to Simplify Regression Models in Medical 40 757 Statistics. J R Statist Soc C 1999;48(3):313–29. 41 42 758 51 Kerr KF, Wang Z, Janes H, et al. Net reclassification indices for evaluating risk prediction 43 759 instruments: A critical review. Epidemiology 2014;25(1):114–21. 44 760 52 James G, Witten D, Hastie T, et al. An introduction to statistical learning: With applications in R, 45 on September 26, 2021 by guest. Protected copyright. 46 761 8th edn. New York, Heidelberg, Dordrecht, London: Springer 2017. 47 762 53 Nakagawa S, Schielzeth H, O'Hara RB. A general and simple method for obtaining R2 from 48 763 generalized linear mixed-effects models. Methods Ecol Evol 2013;4(2):133–42. 49 764 54 Steyerberg EW, Vickers AJ, Cook NR, et al. Assessing the performance of prediction models: A 50 51 765 framework for traditional and novel measures. Epidemiology 2010;21(1):128–38. 52 766 55 Nagelkerke NJD. A note on a general definition of the coefficient of determination. Biometrika 53 767 1991;78(3):691–92. 54 55 768 56 Hastie T, Tibshirani R, Friedman JH. The elements of statistical learning: Data mining, inference, 56 769 and prediction, 12th edn. New York, NY: Springer 2017. 57 770 57 R Core Team. R: A Language and Environment for Statistical Computing. Vienna, Austria: R 58 59 771 Foundation for Statistical Computing 2016. 60 772 58 Pinhero J, Bates D, DebRoy S, et al. nlme: Linear and Nonlinear Mixed Effects Models. r package 773 version 3.1-128 2016. 33 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 35 of 90 BMJ Open
1 2
3 774 59 Kundu S, Aulchenko YS, Janssens, A. Cecile J. W. PredictABEL: Assessment of Risk Prediction BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 775 Models. r package version 1.2-2 2014. 5 6 776 60 Lumley T. mitools: Tools for multiple imputation of missing data. r package version 2.3 2014. 7 777 61 Nakazawa M. fmsb: Functions for Medical Statistics Book with some Demographic Data. r 8 778 package version 0.5.2 2015. 9 10 779 62 Dahl DB. xtable: Export Tables to LaTeX or HTML. r package version 1.8-2 2016. 11 780 63 Xie Y. knitr: A General-Purpose Package for Dynamic Report Generation in R. r package version 12 781 1.13 2016. 13 782 64 Collins GS, Reitsma JB, Altman DG, et al. Transparent Reporting of a multivariable prediction 14 15 783 model for Individual Prognosis or Diagnosis (TRIPOD): the TRIPOD statement. Ann Intern Med 16 784 2015;162(1):55–63. 17 785 65 O'Caoimh R, Cornally N, Weathers E, et al. Risk prediction in the community: A systematic 18 For peer review only 19 786 review of case-finding instruments that predict adverse healthcare outcomes in community- 20 787 dwelling older adults. Maturitas 2015;82(1):3–21. 21 788 66 Wallace E, Stuart E, Vaughan N, et al. Risk prediction models to predict emergency hospital 22 789 admission in community-dwelling adults: A systematic review. Med Care 2014;52(8):751–65. 23 24 790 67 Coleman EA, Min S-j, Chomiak A, et al. Posthospital care transitions: Patterns, complications, 25 791 and risk identification. Health Serv Res 2004;39(5):1449–65. 26 792 68 Campbell SE, Seymour DG, Primrose WR. A systematic literature review of factors affecting 27 28 793 outcome in older medical patients admitted to hospital. Age Ageing 2004;33(2):110–15. 29 794 69 García-Pérez L, Linertová R, Lorenzo-Riera A, et al. Risk factors for hospital readmissions in 30 795 elderly patients: A systematic review. QJM 2011;104(8):639–51. 31 796 70 Alonso-Morán E, Nuño-Solinis R, Onder G, et al. Multimorbidity in risk stratification tools to 32 33 797 predict negative outcomes in adult population. Eur J Intern Med 2015;26(3):182–89. 34 798 71 Wallace E, Hinchey T, Dimitrov BD, et al. A systematic review of the probability of repeated 35 799 admission score in community-dwelling adults. J Am Geriatr Soc 2013;61(3):357–64. 36
37 800 72 Kansagara D, Englander H, Salanitro A, et al. Risk prediction models for hospital readmission: A http://bmjopen.bmj.com/ 38 801 systematic review. JAMA 2011;306(15):1688–98. 39 802 73 Vest JR, Gamm LD, Oxford BA, et al. Determinants of preventable readmissions in the United 40 803 States: A systematic review. Implement Sci 2010;5:88. 41 42 804 74 Stevenson JM, Williams JL, Burnham TG, et al. Predicting adverse drug reactions in older adults; 43 805 a systematic review of the risk prediction models. Clin Interv Aging 2014;9:1581–93. 44 806 75 Glynn RJ, Rosner B. Methods to evaluate risks for composite end points and their individual 45 on September 26, 2021 by guest. Protected copyright. 46 807 components. Journal of Clinical Epidemiology 2004;57(2):113–22. 47 808 76 Steyerberg EW, Eijkemans MJ, Harrell FE, et al. Prognostic modeling with logistic regression 48 809 analysis: in search of a sensible strategy in small data sets. Med Decis Making 2001;21(1):45– 49 810 56. 50 51 811 77 van Calster B, van Smeden M, Cock B de, et al. Regression shrinkage methods for clinical 52 812 prediction models do not guarantee improved performance: Simulation study. Stat Methods 53 813 Med Res 2020:962280220921415. 54 55 814 78 Swart E, Gothe H, Geyer S, et al. Gute Praxis Sekundärdatenanalyse (GPS): Leitlinien und 56 815 Empfehlungen. Gesundheitswesen 2015;77(2):120–26. 57 816 58 59 817 60 818 819 34 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 36 of 90
1 2
3 820 Figures: BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 821 Figure 1: Predicted outcomes with regard to general trajectories of well-being and 6 7 822 quality of life over time 8 823 Figure 2: Models and sensitivity analyses with regard to data source and predictor set 9 10 824 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
35 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Page 37 of 90 BMJ Open Events of the combined outcome 1 2 3 High 4 5 6 7 8 Decline 9 10 For peer review only of 11 12 http://bmjopen.bmj.com/ HRQoL
13 HRQoL 14 15 16 17 18 19 20 on September 26, 2021 by guest. Protected copyright.
21 being/ 22 - 23 24 25
26 Well 27 28 29 Low 30 Death 31 32 33 34 Baseline Follow-up 35 36 Time 37 38 Sudden death Intermittent decline (e.g., multimorbidity) 39 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 40 Rapid decline Gradual decline (e.g., frailty) 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
BMJ Open Page 38 of 90
1 2 CRT data Claims data 3 4 5 6 7 8 Models 2.1, 2.2, 2.3, 9 Models 1.1, 1.2, 1.3 10 For peer review only 11 2.4, 2.5, 2.6, 2.7, 2.8 http://bmjopen.bmj.com/ 12 Core 13 Sensitivity analyses
14 predictors Sensitivity analyses 15 16 17 18 19 20 on September 26, 2021 by guest. Protected copyright. 21 22 23 Models 3.1, 3.2, 3.3, 24 25 3.4†, 3.5, 3.6, 3.7, 3.8, 26 27 28 3.9, 3.10, 3.11, 3.12 29 30 31 Additional predictors 32 Sensitivity analyses 33 34 35 36 37 38 39 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 39 of 90 BMJ Open
1 2
3 Additional file 1: Falls and fall-related injuries: list of excluded ICD-10-codes BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 Conditions ICD-10-Code 7 8 (ICD-10-GM Version 2014) 9 10 . Osteoporosis with pathological fracture M80 11 12 . Open wound of neck S11 13 14 15 . Dislocation, sprain and strain of joints and S13 16 17 ligaments at neck level 18 For peer review only 19 . Injury of nerves and spinal cord at neck level S14 20 21 22 . Injury of blood vessels at neck level S15 23 24 . Injury of muscle and tendon at neck level S16 25 26 . Crushing injury of neck S17 27 28 29 . Traumatic amputation at neck level S18 30 31 . Other and unspecified injuries of neck S19 32 33 . Injury of blood vessels of thorax S25 34 35 . Injury of heart S26 36
37 http://bmjopen.bmj.com/ 38 . Injury of other and unspecified intrathoracic S27 39 40 organs 41 42 . Crushing injury of thorax and traumatic amputation S28 43 44
45 of part of thorax on September 26, 2021 by guest. Protected copyright. 46 47 . Other and unspecified injuries of thorax S29 48 49 . Open wound of abdomen, lower back and pelvis S31 50 51 52 . Injury of blood vessels at abdomen, lower back S35 53 54 and pelvis level 55 56 . Injury of pancreas S36.2 57 58 . Injury of stomach S36.3 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 40 of 90
1 2
3 . Injury of small intestine S36.4 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 . Injury of colon S36.5 6 7 8 . Injury of rectum S36.6 9 10 . Injury of multiple intra-abdominal organs S36.7 11 12 . Injury of other intra-abdominal organs S36.8 13 14 15 . Injury of ureter S37.1 16 17 . Injury of bladder S37.2 18 For peer review only 19 . Injury of urethra S37.3 20 21 22 . Injury of ovary S37.4 23 24 . Injury of fallopian tube S37.5 25 26 . Injury of uterus S37.6 27 28 . Injury of multiple pelvic organs S37.7 29 30 31 . Injury of other pelvic organs S37.8 32 33 . Crushing injury and traumatic amputation of part S38 34 35 of abdomen, lower back and pelvis 36
37 http://bmjopen.bmj.com/ 38 . Injury of intra-abdominal organ(s) with pelvic S39.6 39 40 organ(s) 41 42 . Other multiple injuries of abdomen, lower back S39.7 43 44
45 and pelvis on September 26, 2021 by guest. Protected copyright. 46 47 . Other specified injuries of abdomen, lower back S39.8 48 49 and pelvis 50 51 . Crushing injury of shoulder and upper arm S47 52 53 54 . Traumatic amputation of shoulder and upper arm S48 55 56 . Traumatic amputation of forearm S58 57 58 . Traumatic amputation of wrist and hand S68 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 41 of 90 BMJ Open
1 2
3 . Traumatic amputation of hip and thigh S78 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 . Traumatic amputation of lower leg S88 6 7 8 . Traumatic amputation of ankle and foot S98 9 10 . Crushing injuries involving multiple body regions T04 11 12 . Traumatic amputations involving multiple body T05 13 14 15 regions 16 17 . Effects of foreign body entering through natural T15-19 18 For peer review only 19 orifice 20 21 22 . Burns and corrosions T20-32 23 24 . Frostbite T33-35 25 26 . Poisoning by specified narcotics and T40.0-T40.1, T40.3, T40.5- 27 28 psychodysleptic agents (hallucinogenic drugs) T40.9 29 30 31 . Toxic effects of substances chiefly nonmedicinal T51-65 32 33 as to source 34 35 . Other and unspecified effects of external causes T66-77 36
37 http://bmjopen.bmj.com/ 38 . Complications of surgical and medical care, not T80-87 39 40 elsewhere classified 41 42 . Sequelae of injuries, of poisoning and of other T90-98 43 44
45 consequences of external causes on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 42 of 90
1 2 Potentially inappropriate drugs EU-PIM ATC-Code ATC-Code 2013 ATC-Code 2012 ATC-Code 2014 2011
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Magnesium hydroxide 1 A02AA04 A02AA04 A02AA04 A02AA04 5 Aluminium-containing antacids 1 A02AB01 A02AB01 A02AB01 A02AB01 6 Aluminium-containing antacids 2 A02AB02 A02AB02 A02AB02 A02AB02 7 Aluminium-containing antacids 3 A02AB03 A02AB03 A02AB03 A02AB03 8 Aluminium-containing antacids 4 A02AB04 A02AB04 A02AB04 A02AB04 9 Aluminium-containing antacids 5 A02AB05 A02AB05 A02AB05 A02AB05 10 Aluminium-containing antacids 6 A02AB06 A02AB06 A02AB06 A02AB06 11 Aluminium-containing antacids 7 A02AB07 A02AB07 A02AB07 A02AB07 12 Aluminium-containing antacids 8 A02AB10 A02AB10 A02AB10 A02AB10 13 Aluminium-containing antacids 15 A02AD02 A02AD02 A02AD02 A02AD02 14 Aluminium-containing antacids 16 A02AD03 A02AD03 A02AD03 A02AD03 15 Aluminium-containing antacids 17 A02AD04 A02AD04 A02AD04 A02AD04 16 Aluminium-containing antacids 18 A02AD05 A02AD05 A02AD05 A02AD05 17 Aluminium-containing antacids 9 A02AD06 A02AD06 A02AD06 A02AD06 18 Aluminium-containing antacidsFor 10 (Magnesium peer reviewA02AD10 A02AD10 only A02AD10 A02AD10 19 hydroxide 2) 20 Aluminium-containing antacids 11 A02AF01 A02AF01 A02AF01 A02AF01 21 Aluminium-containing antacids 12 A02AF03 A02AF03 A02AF03 A02AF03 22 Aluminium-containing antacids 13 A02AF04 A02AF04 A02AF04 A02AF04 23 A02AF05 A02AF05 A02AF05 A02AF05 24 Aluminium-containing antacids 14 25 Cimetidine 1 A02BA01 A02BA01 A02BA01 A02BA01 26 Ranitidine 1 A02BA02 A02BA02 A02BA02 A02BA02 27 Famotidine 1 A02BA03 A02BA03 A02BA03 A02BA03 28 Ranitidine 2 A02BA07 A02BA07 A02BA07 A02BA07 29 Cimetidine 2 A02BA51 A02BA51 A02BA51 A02BA51 30 Famotidine 2 A02BA53 A02BA53 A02BA53 A02BA53 31 Mebeverine A03AA04 A03AA04 A03AA04 A03AA04 32 Trimebutine A03AA05 A03AA05 A03AA05 A03AA05 33 Dihexyverine A03AA08 A03AA08 A03AA08 A03AA08 34 Otilonium bromide 1 A03AB06 A03AB06 A03AB06 A03AB06 35 Tiemonium (iodide) 1 A03AB17 A03AB17 A03AB17 A03AB17 36 Trospium 2 A03AB20 A03AB20 A03AB20 A03AB20
37 Pinaverium A03AX04 A03AX04 A03AX04 A03AX04 http://bmjopen.bmj.com/ 38 Belladonna alkaloids 2 A03BA01 A03BA01 A03BA01 A03BA01 39 Belladonna alkaloids 3 (Hyoscyamine 1) A03BA03 A03BA03 A03BA03 A03BA03 40 Belladonna alkaloids 4 A03BA04 A03BA04 A03BA04 A03BA04 41 Belladonna alkaloids 5 A03BA20 A03BA20 A03BA20 A03BA20 42 Belladonna alkaloids 6 A03BB01 A03BB01 A03BB01 A03BB01 43 Belladonna alkaloids 7 A03BB02 A03BB02 A03BB02 A03BB02 44 Belladonna alkaloids 8 A03BB03 A03BB03 A03BB03 A03BB03 45 Belladonna alkaloids 9 A03BB04 A03BB04 A03BB04 A03BB04 on September 26, 2021 by guest. Protected copyright. 46 Belladonna alkaloids 10 A03BB05 A03BB05 A03BB05 A03BB05 47 Clidinium A03CA02 A03CA02 A03CA02 A03CA02 48 Otilonium bromide 2 A03CA04 A03CA04 A03CA04 A03CA04 49 Belladonna alkaloids 11 A03CB01 A03CB01 A03CB01 A03CB01 50 Belladonna alkaloids 12 A03CB02 A03CB02 A03CB02 A03CB02 51 Belladonna alkaloids 13 A03CB03 A03CB03 A03CB03 A03CB03 52 Belladonna alkaloids 14 A03CB04 A03CB04 A03CB04 A03CB04 53 Belladonna alkaloids 15 + Hyoscyamine 2 A03CB31 A03CB31 A03CB31 A03CB31 54 Belladonna alkaloids 16 A03CB37 A03CB37 A03CB37 A03CB37 55 Belladonna alkaloids 17 A03CB38 A03CB38 A03CB38 A03CB38 56 Pitofenone A03DA02 A03DA02 A03DA02 A03DA02 57 A03DA06 A03DA06 A03DA06 A03DA06 58 Trospium 3 59 Tiemonium (iodide) 2 A03DA07 A03DA07 A03DA07 A03DA07 60 Belladonna alkaloids 18 A03DB04 A03DB04 A03DB04 A03DB04 Trospium 4 A03EA04 A03EA04 A03EA04 A03EA04
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 43 of 90 BMJ Open
1 2 Metoclopramide 1 A03FA01 A03FA01 A03FA01 A03FA01 Alizapride A03FA05 A03FA05 A03FA05 A03FA05
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Metoclopramide 2 A03FA51 A03FA51 A03FA51 A03FA51 5 Dimenhydrinate 1 A04AB02 A04AB02 A04AB02 A04AB02 6 Meclozine 3 A04AB04 A04AB04 A04AB04 A04AB04 7 Diphenhydramine 3 A04AB05 A04AB05 A04AB05 A04AB05 8 Dimenhydrinate 2 A04AB52 A04AB52 A04AB52 A04AB52 9 Meclozine 4 A04AB54 A04AB54 A04AB54 A04AB54 10 Diphenhydramine 6 A04AB55 A04AB55 A04AB55 A04AB55 11 Doxylamine 4 A04AB56 A04AB56 A04AB56 A04AB56 12 Promethazine 3 A04AB58 A04AB58 A04AB58 A04AB58 13 Scopolamine 1 (=Hyoscin) A04AD01 A04AD01 A04AD01 A04AD01 14 Metopimazine A04AD05 A04AD05 A04AD05 A04AD05 15 Scopolamine 2 (=Hyoscin) A04AD51 A04AD51 A04AD51 A04AD51 16 Viscous paraffin (=Liquid paraffin) 1 A06AA01 A06AA01 A06AA01 A06AA01 17 Docusate sodium (oral) A06AA02 A06AA02 A06AA02 A06AA02 18 Viscous paraffin (=Liquid paraffin)For 2 peer reviewA06AA51 A06AA51 only A06AA51 A06AA51 19 Castor oil (=Ricinus communis, =Neoloid) A06AB05 A06AB05 A06AB05 A06AB05 20 Senna glycosides 1 A06AB06 A06AB06 A06AB06 A06AB06 21 Cascara sagrada 1 A06AB07 A06AB07 A06AB07 A06AB07 22 Sodium picosulfate 1 A06AB08 A06AB08 A06AB08 A06AB08 23 24 Aloe 1 A06AB13 A06AB13 A06AB13 A06AB13 25 Belladonna alkaloids 1 A06AB30 A06AB30 A06AB30 A06AB30 26 Senna glycosides 2 A06AB56 A06AB56 A06AB56 A06AB56 27 Cascara sagrada 2 A06AB57 A06AB57 A06AB57 A06AB57 28 Sodium picosulfate 2 A06AB58 A06AB58 A06AB58 A06AB58 29 Aloe 2 A06AB63 A06AB63 A06AB63 A06AB63 30 Prucalopride A06AX05 A03AE04 A03AE04 A03AE04 31 Diphenoxylate-Atropin A07DA01 A07DA01 A07DA01 A07DA01 32 Racecadotril A07XA04 A07XA04 A07XA04 A07XA04 33 Norephedrine (=Phenylpropanolamine) 3 A08AA13 A08AA13 A08AA13 A08AA13 34 Norephedrine (=Phenylpropanolamine) 4 A08AA63 A08AA63 A08AA63 A08AA63 35 Glibenclamide 1 A10BB01 A10BB01 A10BB01 A10BB01 36 Chlorpropamide A10BB02 A10BB02 A10BB02 A10BB02
37 Carbutamide A10BB06 A10BB06 A10BB06 A10BB06 http://bmjopen.bmj.com/ 38 Glipizide A10BB07 A10BB07 A10BB07 A10BB07 39 Glimepiride 1 A10BB12 A10BB12 A10BB12 A10BB12 40 Glimepiride 2 A10BD04 A10BD04 A10BD04 A10BD04 41 Pioglitazone 3 A10BD05 A10BD05 A10BD05 A10BD05 42 Glimepiride 3 + Pioglitazone 2 A10BD06 A10BD06 A10BD06 A10BD06 43 Sitagliptine 3 A10BD07 A10BD07 A10BD07 A10BD07 44 Vildagliptine 2 A10BD08 A10BD08 A10BD08 A10BD08 45 Pioglitazone 4 A10BD09 A10BD09 A10BD09 A10BD09 on September 26, 2021 by guest. Protected copyright. 46 Pioglitazone 5 A10BD12 A10BD12 A10BD12 n/a 47 Sitagliptine 4 A10BD12 A10BD12 A10BD12 n/a 48 Glibenclamide 2 A10BD15 n/a n/a n/a 49 Acarbose A10BF01 A10BF01 A10BF01 A10BF01 50 Pioglitazone 1 A10BG03 A10BG03 A10BG03 A10BG03 51 Sitagliptine 1 A10BH01 A10BH01 A10BH01 A10BH01 52 Vildagliptine 1 A10BH02 A10BH02 A10BH02 A10BH02 53 Sitagliptine 2 A10BH51 n/a n/a n/a 54 Cyproheptadine 2 A15AA01 A15AA01 A15AA01 A15AA01 55 Cyproheptadine 3 A15AA51 A15AA51 A15AA51 A15AA51 56 Acenocoumarol B01AA07 B01AA07 B01AA07 B01AA07 57 58 Ticlopidine B01AC05 B01AC05 B01AC05 B01AC05 59 Dipyridamole 1 B01AC07 B01AC07 B01AC07 B01AC07 60 Prasugrel B01AC22 B01AC22 B01AC22 B01AC22 Dipyridamole 2 B01AC36 n/a n/a n/a
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 44 of 90
1 2 Dabigatran B01AE07 B01AE07 B01AE07 B01AE07 Rivaroxaban B01AF01 B01AF01 B01AX06 B01AX06
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Apixaban B01AF02 B01AF02 B01AX08 n/a 5 Acetyldigitoxin C01AA01 C01AA01 C01AA01 C01AA01 6 Acetyldigoxin 1 C01AA02 C01AA02 C01AA02 C01AA02 7 Digitoxin 1 C01AA04 C01AA04 C01AA04 C01AA04 8 Digoxin 1 C01AA05 C01AA05 C01AA05 C01AA05 9 Metildigoxin 1 C01AA08 C01AA08 C01AA08 C01AA08 10 Acetyldigoxin 2 C01AA52 C01AA52 C01AA52 C01AA52 11 Digitoxin 2 C01AA54 C01AA54 C01AA54 C01AA54 12 Digoxin 2 C01AA55 C01AA55 C01AA55 C01AA55 13 Metildigoxin 2 C01AA58 C01AA58 C01AA58 C01AA58 14 Quinidine and combinations 1 C01BA01 C01BA01 C01BA01 C01BA01 15 Procainamide C01BA02 C01BA02 C01BA02 C01BA02 16 Disopyramide C01BA03 C01BA03 C01BA03 C01BA03 17 Quinidine and combinations 2 C01BA51 C01BA51 C01BA51 C01BA51 18 Quinidine and combinations 3For peer reviewC01BA71 C01BA71 only C01BA71 C01BA71 19 Propafenone C01BC03 C01BC03 C01BC03 C01BC03 20 Flecainide C01BC04 C01BC04 C01BC04 C01BC04 21 Amiodarone C01BD01 C01BD01 C01BD01 C01BD01 22 Dronedarone C01BD07 C01BD07 C01BD07 C01BD07 23 C01DX21 C01DX21 C01DX21 C01DX21 24 Dipyridamole 4 25 Dipyridamole 3 C01DX71 C01DX71 C01DX71 C01DX71 26 Indometacin 2 C01EB03 C01EB03 C01EB03 C01EB03 27 Trimetazidine C01EB15 C01EB15 C01EB15 C01EB15 28 Ivabradine C01EB17 C01EB17 C01EB17 C01EB17 29 Theophylline 2 C01EB22 C01EB22 C01EB22 C01EB22 30 Theophylline 3 C01EX66 C01EX66 C01EX66 C01EX66 31 Reserpine 1 C02AA02 C02AA02 C02AA02 C02AA02 32 Reserpine 2 C02AA52 C02AA52 C02AA52 C02AA52 33 Methyldopa 1 C02AB01 C02AB01 C02AB01 C02AB01 34 Methyldopa 2 C02AB02 C02AB02 C02AB02 C02AB02 35 Clonidine 1 C02AC01 C02AC01 C02AC01 C02AC01 36 Guanfacine C02AC02 C02AC02 C02AC02 C02AC02
37 Moxonidine 1 C02AC05 C02AC05 C02AC05 C02AC05 http://bmjopen.bmj.com/ 38 Rilmenidine C02AC06 C02AC06 C02AC06 C02AC06 39 Prazosin 1 C02CA01 C02CA01 C02CA01 C02CA01 40 Doxazosin 1 C02CA04 C02CA04 C02CA04 C02CA04 41 Urapidil C02CA06 C02CA06 C02CA06 C02CA06 42 Terazosin 2 C02CA08 C02CA08 C02CA08 C02CA08 43 Guanethidine 1 C02CC02 C02CC02 C02CC02 C02CC02 44 Hydralazine 1 C02DB02 C02DB02 C02DB02 C02DB02 45 Reserpine 3 C02LA01 C02LA01 C02LA01 C02LA01 on September 26, 2021 by guest. Protected copyright. 46 Reserpine 4 C02LA51 C02LA51 C02LA51 C02LA51 47 Reserpine 5 C02LA71 C02LA71 C02LA71 C02LA71 48 Methyldopa 3 C02LB01 C02LB01 C02LB01 C02LB01 49 Clonidine 2 C02LC01 C02LC01 C02LC01 C02LC01 50 Moxonidine 2 C02LC05 C02LC05 C02LC05 C02LC05 51 Clonidine 3 C02LC51 C02LC51 C02LC51 C02LC51 52 Prazosin 2 C02LE01 C02LE01 C02LE01 C02LE01 53 Guanethidine 2 C02LF01 C02LF01 C02LF01 C02LF01 54 Hydralazine 2 C02LG02 C02LG02 C02LG02 C02LG02 55 Niacin (=Nicotinic acid) 2 C04AC51 C04AC51 C04AC51 C04AC51 56 Pentoxifylline C04AD03 C04AD03 C04AD03 C04AD03 57 58 Ergoloid mesylate (= dihydroergotoxine) 1 C04AE01 C04AE01 C04AE01 C04AE01 59 Nicergoline 1 C04AE02 C04AE02 C04AE02 C04AE02 60 Dihydroergocristine 1 C04AE04 C04AE04 C04AE04 C04AE04 Ergoloid mesylate (= dihydroergotoxine) 2 C04AE51 C04AE51 C04AE51 C04AE51
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 45 of 90 BMJ Open
1 2 Dihydroergocristine 2 C04AE54 C04AE54 C04AE54 C04AE54 Cyclandelate (=Cyclospasmol) 1 C04AX01 C04AX01 C04AX01 C04AX01
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Vincamine 1 C04AX07 C04AX07 C04AX07 C04AX07 5 Moxisylyte 1 C04AX10 C04AX10 C04AX10 C04AX10 6 Piribedil 2 C04AX13 C04AX13 C04AX13 C04AX13 7 Vinburnine C04AX17 C04AX17 C04AX17 C04AX17 8 Buflomedil C04AX20 C04AX20 C04AX20 C04AX20 9 Naftidrofuryl C04AX21 C04AX21 C04AX21 C04AX21 10 Quinine and derivatives 5 C05AF01 C05AF01 C05AF01 C05AF01 11 Quinine and derivatives 6 C05AF51 C05AF51 C05AF51 C05AF51 12 Digitoxin 3 C05BZ05 C05BZ05 C05BZ05 C05BZ05 13 Escin (=Aescin) 3 C05BZ09 C05BZ09 C05BZ09 C05BZ09 14 Escin (=Aescin) 4 C05BZ59 C05BZ59 C05BZ59 C05BZ59 15 Hidrosmin C05CA05 C05CA05 C05CA05 C05CA05 16 Escin (=Aescin) 1 C05CA07 C05CA07 C05CA07 C05CA07 17 Vincamine-Rutoside C05CA51 C05CA51 C05CA51 C05CA51 18 Troxerutin-Vincamine For peer reviewC05CA54 C05CA54 only C05CA54 C05CA54 19 Escin (=Aescin) 2 C05CA57 C05CA57 C05CA57 C05CA57 20 Dihydroergotaminmesilat 1 C06AA02 C06AA02 C06AA02 C06AA02 21 Dihydroergotaminmesilat 2 C06AA50 C06AA50 C06AA50 C06AA50 22 Oxprenolol 1 C07AA02 C07AA02 C07AA02 C07AA02 23 24 Pindolol 1 C07AA03 C07AA03 C07AA03 C07AA03 25 Propranolol 1 C07AA05 C07AA05 C07AA05 C07AA05 26 Sotalol 1 C07AA07 C07AA07 C07AA07 C07AA07 27 Nadolol 1 C07AA12 C07AA12 C07AA12 C07AA12 28 Sotalol 2 C07AA57 C07AA57 C07AA57 C07AA57 29 Labetalol 1 C07AG01 C07AG01 C07AG01 C07AG01 30 Oxprenolol 2 C07BA02 C07BA02 C07BA02 C07BA02 31 Propranolol 5 C07BA05 C07BA05 C07BA05 C07BA05 32 Sotalol 3 C07BA07 C07BA07 C07BA07 C07BA07 33 Nadolol 2 C07BA12 C07BA12 C07BA12 C07BA12 34 Labetalol 3 C07BG01 C07BG01 C07BG01 C07BG01 35 Oxprenolol 3 C07CA02 C07CA02 C07CA02 C07CA02 36 Pindolol 2 C07CA03 C07CA03 C07CA03 C07CA03
37 Propranolol 4 C07CA05 C07CA05 C07CA05 C07CA05 http://bmjopen.bmj.com/ 38 Labetalol 2 C07CG01 C07CG01 C07CG01 C07CG01 39 Propranolol 2 C07DA05 C07DA05 C07DA05 C07DA05 40 Pindolol 3 C07EA03 C07EA03 C07EA03 C07EA03 41 Propranolol 6 C07EA05 C07EA05 C07EA05 C07EA05 42 Oxprenolol 4 C07FA02 C07FA02 C07FA02 C07FA02 43 Propranolol 3 C07FA05 C07FA05 C07FA05 C07FA05 44 Nifedipine 4 C07FB22 C07FB22 C07FB22 C07FB22 45 Nifedipine 5 C07FB23 C07FB23 C07FB23 C07FB23 on September 26, 2021 by guest. Protected copyright. 46 Nicardipine C08CA04 C08CA04 C08CA04 C08CA04 47 Nifedipine 1 C08CA05 C08CA05 C08CA05 C08CA05 48 Nifedipine 2 C08CA55 C08CA55 C08CA55 C08CA55 49 Verapamil 1 C08DA01 C08DA01 C08DA01 C08DA01 50 Verapamil 5 C08DA51 C08DA51 C08DA51 C08DA51 51 Verapamil 2 C08DA81 C08DA81 C08DA81 C08DA81 52 Diltiazem C08DB01 C08DB01 C08DB01 C08DB01 53 Nifedipine 3 C08GA01 C08GA01 C08GA01 C08GA01 54 Verapamil 3 C08GA02 C08GA02 C08GA02 C08GA02 55 Verapamil 4 C09BB10 C09BB10 C09BB10 C09BB10 56 Niacin (=Nicotinic acid) 1 C10AD02 C10AD02 C10AD02 C10AD02 57 C10AD52 C10AD52 C10AD52 C10AD52 58 Niacin (=Nicotinic acid) 3 59 Niacin (=Nicotinic acid) 4 C10BA01 C10BA01 C10BA01 C10BA01 60 Niacin (=Nicotinic acid) 5 C10BB01 C10BB01 C10BB01 C10BB01 Dimetindene 2 D04AA13 D04AA13 D04AA13 D04AA13
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 46 of 90
1 2 Bromocriptine 2 G02CB01 G02CB01 G02CB01 G02CB01 Cabergoline 2 G02CB03 G02CB03 G02CB03 G02CB03
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Oestrogen (oral) 1 G03CA01 G03CA01 G03CA01 G03CA01 5 Oestrogen (oral) 2 G03CA03 G03CA03 G03CA03 G03CA03 6 Oestrogen (oral) 3 G03CA04 G03CA04 G03CA04 G03CA04 7 Oestrogen (oral) 4 G03CA06 G03CA06 G03CA06 G03CA06 8 Oestrogen (oral) 5 G03CA07 G03CA07 G03CA07 G03CA07 9 Oestrogen (oral) 6 G03CA09 G03CA09 G03CA09 G03CA09 10 Oestrogen (oral) 7 G03CA10 G03CA10 G03CA10 G03CA10 11 Oestrogen (oral) 8 G03CA53 G03CA53 G03CA53 G03CA53 12 Oestrogen (oral) 9 G03CA57 G03CA57 G03CA57 G03CA57 13 Oestrogen (oral) 10 G03CB01 G03CB01 G03CB01 G03CB01 14 Oestrogen (oral) 11 G03CB02 G03CB02 G03CB02 G03CB02 15 Oestrogen (oral) 12 G03CB03 G03CB03 G03CB03 G03CB03 16 Oestrogen (oral) 13 G03CB04 G03CB04 G03CB04 G03CB04 17 Oestrogen (oral) 14 G03CC02 G03CC02 G03CC02 G03CC02 18 Oestrogen (oral) 15 For peer reviewG03CC03 G03CC03 only G03CC03 G03CC03 19 Oestrogen (oral) 16 G03CC04 G03CC04 G03CC04 G03CC04 20 Oestrogen (oral) 17 G03CC05 G03CC05 G03CC05 G03CC05 21 Oestrogen (oral) 18 G03CC06 G03CC06 G03CC06 G03CC06 22 Oestrogen (oral) 19 G03CC07 G03CC07 G03CC07 G03CC07 23 G03CC08 G03CC08 G03CC08 G03CC08 24 Oestrogen (oral) 20 25 Oestrogen (oral) 21 G03CX01 G03CX01 G03CX01 G03CX01 26 Flavoxat G04BD02 G04BD02 G04BD02 G04BD02 27 Oxybutynine G04BD04 G04BD04 G04BD04 G04BD04 28 Tolterodine G04BD07 G04BD07 G04BD07 G04BD07 29 Solifenacin 1 G04BD08 G04BD08 G04BD08 G04BD08 30 Trospium 1 G04BD09 G04BD09 G04BD09 G04BD09 31 Darifenacin G04BD10 G04BD10 G04BD10 G04BD10 32 Fesoterodin G04BD11 G04BD11 G04BD11 G04BD11 33 Trospium 5 G04BD59 G04BD59 G04BD59 G04BD59 34 Moxisylyte 2 G04BE06 G04BE06 G04BE06 G04BE06 35 Magnesium hydroxide 3 G04BX01 G04BX01 G04BX01 G04BX01 36 Terazosin 1 G04CA03 G04CA03 G04CA03 G04CA03
37 Doxazosin 2 G04CA05 G04CA05 G04CA05 G04CA05 http://bmjopen.bmj.com/ 38 Solifenacin 2 G04CA53 n/a n/a n/a 39 Ofloxacin J01MA01 J01MA01 J01MA01 J01MA01 40 Amantadine 2 J05AC04 J05AC04 J05AC04 J05AC04 41 Demecolcin L01CC01 L01CC01 L01CC01 L01CC01 42 Celecoxib 2 L01XX33 L01XX33 L01XX33 L01XX33 43 Phenylbutazone 1 M01AA01 M01AA01 M01AA01 M01AA01 44 Phenylbutazone 2 M01AA51 M01AA51 M01AA51 M01AA51 45 Indometacin 1 M01AB01 M01AB01 M01AB01 M01AB01 on September 26, 2021 by guest. Protected copyright. 46 Diclofenac 1 M01AB05 M01AB05 M01AB05 M01AB05 47 Acemetacin M01AB11 M01AB11 M01AB11 M01AB11 48 Ketorolac M01AB15 M01AB15 M01AB15 M01AB15 49 Aceclofenac 1 M01AB16 M01AB16 M01AB16 M01AB16 50 Indometacin 3 M01AB51 M01AB51 M01AB51 M01AB51 51 Diclofenac 2 M01AB55 M01AB55 M01AB55 M01AB55 52 Piroxicam 1 M01AC01 M01AC01 M01AC01 M01AC01 53 Lornoxicam M01AC05 M01AC05 M01AC05 M01AC05 54 Meloxicam 1 M01AC06 M01AC06 M01AC06 M01AC06 55 Meloxicam 2 M01AC56 M01AC56 n/a n/a 56 Ketoprofen 1 M01AE03 M01AE03 M01AE03 M01AE03 57 58 Flurbiprofen 1 M01AE09 M01AE09 M01AE09 M01AE09 59 Dexketoprofen M01AE17 M01AE17 M01AE17 M01AE17 60 Ketoprofen 2 M01AE53 M01AE53 M01AE53 M01AE53 Mefenamic acid M01AG01 M01AG01 M01AG01 M01AG01
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 47 of 90 BMJ Open
1 2 Celecoxib 1 M01AH01 M01AH01 M01AH01 M01AH01 Etoricoxib M01AH05 M01AH05 M01AH05 M01AH05
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Nabumetone M01AX01 M01AX01 M01AX01 M01AX01 5 Phenylbutazone 3 M01BA01 M01BA01 M01BA01 M01BA01 6 Phenylbutazone 5 M02AA01 M02AA01 M02AA01 M02AA01 7 Piroxicam 2 M02AA07 M02AA07 M02AA07 M02AA07 8 Ketoprofen 3 M02AA10 M02AA10 M02AA10 M02AA10 9 Diclofenac 3 M02AA15 M02AA15 M02AA15 M02AA15 10 Flurbiprofen 2 M02AA19 M02AA19 M02AA19 M02AA19 11 Indometacin 4 M02AA23 M02AA23 M02AA23 M02AA23 12 Aceclofenac 2 M02AA25 M02AA25 M02AA25 M02AA25 13 Indometacin 5 M02AA73 M02AA73 M02AA73 M02AA73 14 Niacin (=Nicotinic acid) 6 M02AD50 M02AD50 M02AD50 M02AD50 15 Carisoprodol 1 M03BA02 M03BA02 M03BA02 M03BA02 16 Methocarbamol 1 M03BA03 M03BA03 M03BA03 M03BA03 17 Carisoprodol 2 M03BA52 M03BA52 M03BA52 M03BA52 18 Methocarbamol 2 For peer reviewM03BA53 M03BA53 only M03BA53 M03BA53 19 Meprobamate 4 M03BA57 M03BA57 M03BA57 M03BA57 20 Carisoprodol 3 M03BA72 M03BA72 M03BA72 M03BA72 21 Methocarbamol 3 M03BA73 M03BA73 M03BA73 M03BA73 22 Orphenadrine 1 M03BC01 M03BC01 M03BC01 M03BC01 23 M03BC51 M03BC51 M03BC51 M03BC51 24 Orphenadrine 2 25 Baclofen M03BX01 M03BX01 M03BX01 M03BX01 26 Tizanidine M03BX02 M03BX02 M03BX02 M03BX02 27 Tetrazepam M03BX07 M03BX07 M03BX07 M03BX07 28 Cyclobenzaprine M03BX08 M03BX08 M03BX08 M03BX08 29 Colchicin M04AC01 M04AC01 M04AC01 M04AC01 30 Strontium ranelate M05BX03 M05BX03 M05BX03 M05BX03 31 Quinine and derivatives 1 M09AA01 M09AA01 M09AA01 M09AA01 32 Quinine and derivatives 2 M09AA02 M09AA02 M09AA02 M09AA02 33 Quinine and derivatives 3 M09AA52 M09AA52 M09AA52 M09AA52 34 Quinine and derivatives 4 M09AA72 M09AA72 M09AA72 M09AA72 35 Diphenhydramine 4 N01BX06 N01BX06 N01BX06 N01BX06 36 Ethylmorphine 2 N02AA57 N02AA57 N02AA57 N02AA57
37 Diclofenac 4 N02AA65 N02AA65 N02AA65 N02AA65 http://bmjopen.bmj.com/ 38 Pethidine (=Meperidine) 1 N02AB02 N02AB02 N02AB02 N02AB02 39 Pethidine (=Meperidine) 2 N02AB52 N02AB52 N02AB52 N02AB52 40 Pethidine (=Meperidine) 3 N02AB72 N02AB72 N02AB72 N02AB72 41 Methadone 3 (Levomethadone) N02AC06 N02AC06 N02AC06 N02AC06 42 Methadone 2 N02AC52 N02AC52 N02AC52 N02AC52 43 Pentazocine N02AD01 N02AD01 N02AD01 N02AD01 44 Pethidine (=Meperidine) 4 N02AG03 N02AG03 N02AG03 N02AG03 45 Tramadol 1 N02AX02 N02AX02 N02AX02 N02AX02 on September 26, 2021 by guest. Protected copyright. 46 Tramadol 2 N02AX52 N02AX52 N02AX52 N02AX52 47 Dihydroergotamine 1 N02CA01 N02CA01 N02CA01 N02CA01 48 Ergotamine 1 N02CA02 N02CA02 N02CA02 N02CA02 49 Dihydroergotamine 2 N02CA51 N02CA51 N02CA51 N02CA51 50 Ergotamine 2 N02CA52 N02CA52 N02CA52 N02CA52 51 Dihydroergotamine 3 N02CA71 N02CA71 N02CA71 N02CA71 52 Ergotamine 3 N02CA72 N02CA72 N02CA72 N02CA72 53 Triptanes / Selektive Serotonin-5HT1-Agonisten 1 N02CC01 N02CC01 N02CC01 N02CC01 54 Triptanes / Selektive Serotonin-5HT1-Agonisten 2 N02CC02 N02CC02 N02CC02 N02CC02 55 Triptanes / Selektive Serotonin-5HT1-Agonisten 3 N02CC03 N02CC03 N02CC03 N02CC03 56 Triptanes / Selektive Serotonin-5HT1-Agonisten 4 N02CC04 N02CC04 N02CC04 N02CC04 57 N02CC05 N02CC05 N02CC05 N02CC05 58 Triptanes / Selektive Serotonin-5HT1-Agonisten 5 59 Triptanes / Selektive Serotonin-5HT1-Agonisten 6 N02CC06 N02CC06 N02CC06 N02CC06 60 Triptanes / Selektive Serotonin-5HT1-Agonisten 7 N02CC07 N02CC07 N02CC07 N02CC07 Clonidine 4 N02CX02 N02CX02 N02CX02 N02CX02
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 48 of 90
1 2 Topiramate 2 N02CX12 N02CX12 N02CX12 N02CX12 Metoclopramide 3 N02CX59 N02CX59 N02CX59 N02CX59
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Phenobarbital 1 N03AA02 N03AA02 N03AA02 N03AA02 5 Phenytoin 1 N03AB02 N03AB02 N03AB02 N03AB02 6 Phenytoin 3 (Fosphenytoin) N03AB05 N03AB05 N03AB05 N03AB05 7 Phenytoin 2 N03AB52 N03AB52 N03AB52 N03AB52 8 Clonazepam N03AE01 N03AE01 N03AE01 N03AE01 9 Midazolam 2 N03AE02 N03AE02 n/a n/a 10 Carbamazepine N03AF01 N03AF01 N03AF01 N03AF01 11 Topiramate 1 N03AX11 N03AX11 N03AX11 N03AX11 12 Trihexyphenidyl N04AA01 N04AA01 N04AA01 N04AA01 13 Biperiden N04AA02 N04AA02 N04AA02 N04AA02 14 Tropatepin N04AA12 N04AA12 N04AA12 N04AA12 15 Orphenadrine 3 N04AB02 N04AB02 N04AB02 N04AB02 16 Benzatropine 1 N04AC01 N04AC01 N04AC01 N04AC01 17 Benzatropine 2 (Etybenzatropin) N04AC30 N04AC30 N04AC30 N04AC30 18 Amantadine 1 For peer reviewN04BB01 N04BB01 only N04BB01 N04BB01 19 Bromocriptine 1 N04BC01 N04BC01 N04BC01 N04BC01 20 Pergolide N04BC02 N04BC02 N04BC02 N04BC02 21 Dihydroergocryptine N04BC03 N04BC03 N04BC03 N04BC03 22 Ropinirole N04BC04 N04BC04 N04BC04 N04BC04 23 24 Pramipexole N04BC05 N04BC05 N04BC05 N04BC05 25 Cabergoline 1 N04BC06 N04BC06 N04BC06 N04BC06 26 Piribedil 1 N04BC08 N04BC08 N04BC08 N04BC08 27 Rotigotine N04BC09 N04BC09 N04BC09 N04BC09 28 Selegiline N04BD01 N04BD01 N04BD01 N04BD01 29 Chlorpromazine N05AA01 N05AA01 N05AA01 N05AA01 30 Levomepromazine N05AA02 N05AA02 N05AA02 N05AA02 31 Clorazepate-Acepromazine 1 N05AA04 N05AA04 N05AA04 N05AA04 32 Cyamemazine N05AA06 N05AA06 N05AA06 N05AA06 33 Fluphenazine N05AB02 N05AB02 N05AB02 N05AB02 34 Perphenazine N05AB03 N05AB03 N05AB03 N05AB03 35 Prochlorperazine N05AB04 N05AB04 N05AB04 N05AB04 36 Trifluoperazine N05AB06 N05AB06 N05AB06 N05AB06
37 Propericiazine (=Periciazine) N05AC01 N05AC01 N05AC01 N05AC01 http://bmjopen.bmj.com/ 38 Thioridazine N05AC02 N05AC02 N05AC02 N05AC02 39 Pipotiazine N05AC04 N05AC04 N05AC04 N05AC04 40 Droperidol N05AD08 N05AD08 N05AD08 N05AD08 41 Sertindole N05AE03 N05AE03 N05AE03 N05AE03 42 Ziprasidone N05AE04 N05AE04 N05AE04 N05AE04 43 Flupentixole N05AF01 N05AF01 N05AF01 N05AF01 44 Chlorprothixen N05AF03 N05AF03 N05AF03 N05AF03 45 Zuclopenthixol N05AF05 N05AF05 N05AF05 N05AF05 on September 26, 2021 by guest. Protected copyright. 46 Pimozide N05AG02 N05AG02 N05AG02 N05AG02 47 Clozapine N05AH02 N05AH02 N05AH02 N05AH02 48 Lithium N05AN01 N05AN01 N05AN01 N05AN01 49 Aripiprazole N05AX12 N05AX12 N05AX12 N05AX12 50 Reserpine 6 N05AX15 N05AX15 N05AX15 N05AX15 51 Diazepam N05BA01 N05BA01 N05BA01 N05BA01 52 Chlordiazepoxide N05BA02 N05BA02 N05BA02 N05BA02 53 Medazepam N05BA03 N05BA03 N05BA03 N05BA03 54 Clorazepate-Acepromazine 2 (Dipotassium clorazepate) N05BA05 N05BA05 N05BA05 N05BA05 55 56 Bromazepam N05BA08 N05BA08 N05BA08 N05BA08 57 58 Clobazam N05BA09 N05BA09 N05BA09 N05BA09 59 Prazepam N05BA11 N05BA11 N05BA11 N05BA11 60 Alprazolam N05BA12 N05BA12 N05BA12 N05BA12 Halazepam N05BA13 N05BA13 N05BA13 N05BA13
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 49 of 90 BMJ Open
1 2 Nordazepam N05BA16 N05BA16 N05BA16 N05BA16 (Ethyl-) Loflazepate N05BA18 N05BA18 N05BA18 N05BA18
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Hydroxyzine 1 N05BB01 N05BB01 N05BB01 N05BB01 5 Hydroxyzine 3 N05BB51 N05BB51 N05BB51 N05BB51 6 Meprobamate 1 N05BC01 N05BC01 N05BC01 N05BC01 7 Meprobamate 2 N05BC51 N05BC51 N05BC51 N05BC51 8 Phenobarbital 2 N05CA24 N05CA24 N05CA24 N05CA24 9 Chloralhydrate 1 N05CC01 N05CC01 N05CC01 N05CC01 10 Flurazepam N05CD01 N05CD01 N05CD01 N05CD01 11 Nitrazepam N05CD02 N05CD02 N05CD02 N05CD02 12 Flunitrazepam N05CD03 N05CD03 N05CD03 N05CD03 13 Estazolam N05CD04 N05CD04 N05CD04 N05CD04 14 Triazolam N05CD05 N05CD05 N05CD05 N05CD05 15 Temazepam N05CD07 N05CD07 N05CD07 N05CD07 16 Midazolam 1 N05CD08 N05CD08 N05CD08 N05CD08 17 Quazepam N05CD10 N05CD10 N05CD10 N05CD10 18 Clomethiazole 1 For peer reviewN05CM02 N05CM02 only N05CM02 N05CM02 19 Scopolamine 3 (=Hyoscin) N05CM05 N05CM05 N05CM05 N05CM05 20 Propiomazine N05CM06 N05CM06 N05CM06 N05CM06 21 Diphenhydramine 5 N05CM20 N05CM20 N05CM20 N05CM20 22 Doxylamine 3 N05CM21 N05CM21 N05CM21 N05CM21 23 N05CM22 N05CM22 N05CM22 N05CM22 24 Promethazine 4 25 Meprobamate 3 N05CX01 N05CX01 N05CX01 N05CX01 26 Clomethiazole 2 N05CX04 N05CX04 N05CX04 N05CX04 27 Diphenhydramine 2 N05CX07 N05CX07 N05CX07 N05CX07 28 Chloralhydrate 2 N05CX11 N05CX11 N05CX11 N05CX11 29 Promethazine 5 N05CX13 N05CX13 N05CX13 N05CX13 30 Desipramine N06AA01 N06AA01 N06AA01 N06AA01 31 Imipramine 1 N06AA02 N06AA02 N06AA02 N06AA02 32 Imipramine 2 N06AA03 N06AA03 N06AA03 N06AA03 33 Clomipramine N06AA04 N06AA04 N06AA04 N06AA04 34 Trimipramine N06AA06 N06AA06 N06AA06 N06AA06 35 Amitriptyline 1 N06AA09 N06AA09 N06AA09 N06AA09 36 Nortriptyline 1 N06AA10 N06AA10 N06AA10 N06AA10
37 Doxepin N06AA12 N06AA12 N06AA12 N06AA12 http://bmjopen.bmj.com/ 38 Dosulepin 1 N06AA16 N06AA16 N06AA16 N06AA16 39 Amoxapine N06AA17 N06AA17 N06AA17 N06AA17 40 Maprotiline N06AA21 N06AA21 N06AA21 N06AA21 41 Amitriptyline 3 N06AA25 N06AA25 N06AA25 N06AA25 42 Fluoxetine 1 N06AB03 N06AB03 N06AB03 N06AB03 43 Paroxetine N06AB05 N06AB05 N06AB05 N06AB05 44 Fluvoxamine N06AB08 N06AB08 N06AB08 N06AB08 45 Tranylcypromine 1 N06AF04 N06AF04 N06AF04 N06AF04 on September 26, 2021 by guest. Protected copyright. 46 Bupropion 1 N06AX12 N06AX12 N06AX12 N06AX12 47 Venlafaxine 1 N06AX16 N06AX16 N06AX16 N06AX16 48 Reboxetine N06AX18 N06AX18 N06AX18 N06AX18 49 Venlafaxine 2 (Desvenlafaxine) N06AX23 N06AX23 N06AX23 N06AX23 50 Methylphenidat N06BA04 N06BA04 N06BA04 N06BA04 51 Piracetam N06BX03 N06BX03 N06BX03 N06BX03 52 Amitriptyline 2 N06CA01 N06CA01 N06CA01 N06CA01 53 Fluoxetine 2 N06CA03 n/a n/a n/a 54 Nortriptyline 2 N06CA06 N06CA06 N06CA06 N06CA06 55 Tranylcypromine 2 N06CA07 N06CA03 N06CA03 N06CA03 56 Dosulepin 2 N06CA10 N06CA10 N06CA10 N06CA10 57 N06DP01 N06DP01 N06DP01 N06DH01 58 Ginkgo biloba 59 Ergoloid mesylate (= dihydroergotoxine) 4 N06DX07 N06DX07 N06DX07 N06DX07 60 Vincamine 2 N06DX09 N06DX09 N06DX09 N06DX09 Nicergoline 2 N06DX13 N06DX13 N06DX13 N06DX13
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 50 of 90
1 2 Cyclandelate (=Cyclospasmol) 2 N06DX14 N06DX14 N06DX14 N06DX14 Dihydroergocristine 3 N06DX19 N06DX19 N06DX19 N06DX19
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Ergoloid mesylate (= dihydroergotoxine) 3 N06DX57 N06DX57 N06DX57 N06DX57 5 Bethanechol N07AB02 N07AB02 N07AB02 N07AB02 6 Bupropion 2 N07BA02 N07BA02 N07BA02 N07BA02 7 Clonidine 5 N07BB06 N07BB06 N07BB06 N07BB06 8 Methadone 1 N07BC02 N07BC02 N07BC02 N07BC02 9 Dextrometorphan 2 N07XX59 n/a n/a n/a 10 Quinine and derivatives 7 P01BC01 P01BC01 P01BC01 P01BC01 11 Norephedrine (=Phenylpropanolamine) 1 R01BA01 R01BA01 R01BA01 R01BA01 12 Pseudoephedrine 1 R01BA02 R01BA02 R01BA02 R01BA02 13 Norephedrine (=Phenylpropanolamine) 2 R01BA51 R01BA51 R01BA51 R01BA51 14 Pseudoephedrine 2 R01BA52 R01BA52 R01BA52 R01BA52 15 Flurbiprofen 3 R02AX01 R02AX01 R02AX01 R02AX01 16 Terbutaline (oral) 2 R03AC03 R03AC03 R03AC03 R03AC03 17 Terbutaline (oral) 1 R03CC03 R03CC03 R03CC03 R03CC03 18 Terbutaline (oral) 3 For peer reviewR03CC53 R03CC53 only R03CC53 R03CC53 19 Theophylline 1 R03DA04 R03DA04 R03DA04 R03DA04 20 Theophylline 4 R03DA54 R03DA54 R03DA54 R03DA54 21 Theophylline 5 R03DA74 R03DA74 R03DA74 R03DA74 22 Theophylline 6 R03DB04 R03DB04 R03DB04 R03DB04 23 24 Ethylmorphine 1 R05DA01 R05DA01 R05DA01 R05DA01 25 Dextrometorphan 1 R05DA09 R05DA09 R05DA09 R05DA09 26 Dextrometorphan 3 R05DA59 R05DA59 R05DA59 R05DA59 27 Phenylbutazone 4 R05XA10 R05XA10 R05XA10 R05XA10 28 Diphenhydramine 1 R06AA02 R06AA02 R06AA02 R06AA02 29 Clemastine 1 R06AA04 R06AA04 R06AA04 R06AA04 30 Carbinoxamine R06AA08 R06AA08 R06AA08 R06AA08 31 Doxylamine 1 R06AA09 R06AA09 R06AA09 R06AA09 32 Diphenhydramine 7 R06AA52 R06AA52 R06AA52 R06AA52 33 Clemastine 2 R06AA54 R06AA54 R06AA54 R06AA54 34 Doxylamine 2 R06AA59 R06AA59 n/a n/a 35 Brompheniramine 1 R06AB01 R06AB01 R06AB01 R06AB01 36 Dexchlorpheniramine 1 R06AB02 R06AB02 R06AB02 R06AB02
37 Dimetindene 1 R06AB03 R06AB03 R06AB03 R06AB03 http://bmjopen.bmj.com/ 38 Chlorpheniramine (=Chlorphenamine) 1 R06AB04 R06AB04 R06AB04 R06AB04 39 Pheniramine R06AB05 R06AB05 R06AB05 R06AB05 40 Dexbrompheniramin 1 R06AB06 R06AB06 R06AB06 R06AB06 41 Brompheniramine 2 R06AB51 R06AB51 R06AB51 R06AB51 42 Dexchlorpheniramine 2 R06AB52 R06AB52 R06AB52 R06AB52 43 Chlorpheniramine (=Chlorphenamine) 2 R06AB54 R06AB54 R06AB54 R06AB54 44 Dexbrompheniramin 2 R06AB56 R06AB56 R06AB56 R06AB56 45 Tripelennamine R06AC04 R06AC04 R06AC04 R06AC04 on September 26, 2021 by guest. Protected copyright. 46 Alimemazine R06AD01 R06AD01 R06AD01 R06AD01 47 Promethazine 1 R06AD02 R06AD02 R06AD02 R06AD02 48 Promethazine 8 (Hydroxyethylpromethazin 1) R06AD05 R06AD05 R06AD05 R06AD05 49 Mequitazine R06AD07 R06AD07 R06AD07 R06AD07 50 Oxomemazine R06AD08 R06AD08 R06AD08 R06AD08 51 Promethazine 7 (Dioxopromethazin) R06AD10 R06AD10 R06AD10 R06AD10 52 Promethazine 2 R06AD52 R06AD52 R06AD52 R06AD52 53 Promethazine 9 (Hydroxyethylpromethazin 2) R06AD55 R06AD55 R06AD55 R06AD55 54 Buclizine 1 R06AE01 R06AE01 R06AE01 R06AE01 55 Cyclizine 1 R06AE03 R06AE03 R06AE03 R06AE03 56 Meclozine 1 R06AE05 R06AE05 R06AE05 R06AE05 57 R06AE51 R06AE51 R06AE51 R06AE51 58 Buclizine 2 59 Cyclizine 2 R06AE53 R06AE53 R06AE53 R06AE53 60 Meclozine 2 R06AE55 R06AE55 R06AE55 R06AE55 Cyproheptadine 1 R06AX02 R06AX02 R06AX02 R06AX02
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 51 of 90 BMJ Open
1 2 Triprolidine 1 R06AX07 R06AX07 R06AX07 R06AX07 Terfenadine R06AX12 R06AX12 R06AX12 R06AX12
3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Ebastine R06AX22 R06AX22 R06AX22 R06AX22 5 Pimethixene R06AX23 R06AX23 R06AX23 R06AX23 6 Hydroxyzine 2 R06AX32 R06AX32 R06AX32 R06AX32 7 Triprolidine 2 R06AX57 R06AX57 R06AX57 R06AX57 8 Promethazine 6 V03AB05 V03AB05 V03AB05 V03AB05 9 10 11 12 13 14 15 16 17 18 For peer review only 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 52 of 90
1 2 ATC-Code ATC-Code ATC-Code ATC-Code PRICSCUS List
3 2014 2013 2012 2011 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Acemetacin M01AB11 M01AB11 M01AB11 M01AB11 5 6 Acetyldigoxin 1 C01AA02 C01AA02 C01AA02 C01AA02 7 Acetyldigoxin 2 C01AA52 C01AA52 C01AA52 C01AA52 8 Alprazolam N05BA12 N05BA12 N05BA12 N05BA12 9 Amitriptyline 1 N06AA09 N06AA09 N06AA09 N06AA09 10 Amitriptyline 2 N06CA01 N06CA01 N06CA01 N06CA01 11 12 Amitriptyline 3 N06AA25 N06AA25 N06AA25 N06AA25 13 Baclofen M03BX01 M03BX01 M03BX01 M03BX01 14 Bromazepam N05BA08 N05BA08 N05BA08 N05BA08 15 Chinidin = Quinidine 1 C01BA01 C01BA01 C01BA01 C01BA01 16 17 Chinidin = Quinidine 2 C01BA51 C01BA51 C01BA51 C01BA51 18 Chinidin = Quinidine 3 For peer C01reviewBA71 C01BA71 only C01BA71 C01BA71 19 Chloralhydrat 1 N05CC01 N05CC01 N05CC01 N05CC01 20 Chloralhydrat 2 N05CX11 N05CX11 N05CX11 N05CX11 21 Chlordiazepoxid N05BA02 N05BA02 N05BA02 N05BA02 22 23 Chlorpheniramine (=Chlorphenamine) 1 R06AB04 R06AB04 R06AB04 R06AB04 24 Chlorpheniramine (=Chlorphenamine) 2 R06AB54 R06AB54 R06AB54 R06AB54 25 Clemastin 1 R06AA04 R06AA04 R06AA04 R06AA04 26 Clemastin 2 R06AA54 R06AA54 R06AA54 R06AA54 27 Clobazam N05BA09 N05BA09 N05BA09 N05BA09 28 29 Clomipramin N06AA04 N06AA04 N06AA04 N06AA04 30 Clonidin 1 C02AC01 C02AC01 C02AC01 C02AC01 31 Clonidin 2 C02LC01 C02LC01 C02LC01 C02LC01 32 Clonidin 3 C02LC51 C02LC51 C02LC51 C02LC51 33 Clonidin 4 N02CX02 N02CX02 N02CX02 N02CX02 34 35 Clonidin 5 N07BB06 N07BB06 N07BB06 N07BB06 36 Clozapin N05AH02 N05AH02 N05AH02 N05AH02
37 Diazepam N05BA01 N05BA01 N05BA01 N05BA01 http://bmjopen.bmj.com/ 38 Dickflüssiges Paraffin 1 A06AA01 A06AA01 A06AA01 A06AA01 39 Dickflüssiges Paraffin 2 40 A06AA51 A06AA51 A06AA51 A06AA51 41 Digoxin 1 C01AA05 C01AA05 C01AA05 C01AA05 42 Digoxin 2 C01AA55 C01AA55 C01AA55 C01AA55 43 Dihydroergocryptinmesilat N04BC03 N04BC03 N04BC03 N04BC03 44 Dihydroergotamin 1 N02CA01 N02CA01 N02CA01 N02CA01 45 on September 26, 2021 by guest. Protected copyright. 46 Dihydroergotamin 2 N02CA51 N02CA51 N02CA51 N02CA51 47 Dihydroergotamin 3 N02CA71 N02CA71 N02CA71 N02CA71 48 Dihydroergotaminmesilat 1 C06AA02 C06AA02 C06AA02 C06AA02 49 Dihydroergotaminmesilat 2 C06AA50 C06AA50 C06AA50 C06AA50 50 Dihydroergotoxin 1 N06DX07 N06DX07 N06DX07 N06DX07 51 52 Dihydroergotoxin 2 N06DX57 N06DX57 N06DX57 N06DX57 53 Dikaliumclorazepat N05BA05 N05BA05 N05BA05 N05BA05 54 Dimenhydrinate 1 A04AB02 A04AB02 A04AB02 A04AB02 55 Dimenhydrinate 2 A04AB52 A04AB52 A04AB52 A04AB52 56 Dimetindene 1 R06AB03 R06AB03 R06AB03 R06AB03 57 58 Dimetindene 2 D04AA13 D04AA13 D04AA13 D04AA13 59 Diphenhydramine 1 R06AA02 R06AA02 R06AA02 R06AA02 60 Diphenhydramine 2 N05CX07 N05CX07 N05CX07 N05CX07 Diphenhydramine 3 A04AB05 A04AB05 A04AB05 A04AB05
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 53 of 90 BMJ Open
1 2 Diphenhydramine 4 N01BX06 N01BX06 N01BX06 N01BX06
3 Diphenhydramine 5 N05CM20 N05CM20 N05CM20 N05CM20 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Diphenhydramine 6 A04AB55 A04AB55 A04AB55 A04AB55 5 6 Diphenhydramine 7 R06AA52 R06AA52 R06AA52 R06AA52 7 Doxazosin 1 C02CA04 C02CA04 C02CA04 C02CA04 8 Doxazosin 2 G04CA05 G04CA05 G04CA05 G04CA05 9 Doxepin N06AA12 N06AA12 N06AA12 N06AA12 10 Doxylamine 1 R06AA09 R06AA09 R06AA09 R06AA09 11 12 Doxylamine 2 R06AA59 R06AA59 n/a n/a 13 Doxylamine 3 N05CM21 N05CM21 N05CM21 N05CM21 14 Doxylamine 4 A04AB56 A04AB56 A04AB56 A04AB56 15 Ergotamine 1 N02CA02 N02CA02 N02CA02 N02CA02 16 17 Ergotamine 2 N02CA52 N02CA52 N02CA52 N02CA52 18 Ergotamine 3 For peer N0review2CA72 N02CA72 only N02CA72 N02CA72 19 Etoricoxib M01AH05 M01AH05 M01AH05 M01AH05 20 Flecainid C01BC04 C01BC04 C01BC04 C01BC04 21 Flunitrazepam N05CD03 N05CD03 N05CD03 N05CD03 22 23 Fluoxetine 1 N06AB03 N06AB03 N06AB03 N06AB03 24 Fluoxetine 2 N06CA03 n/a n/a n/a 25 Fluphenazin N05AB02 N05AB02 N05AB02 N05AB02 26 Flurazepam N05CD01 N05CD01 N05CD01 N05CD01 27 Hydroxyzine 1 N05BB01 N05BB01 N05BB01 N05BB01 28 29 Hydroxyzine 2 R06AX32 R06AX32 R06AX32 R06AX32 30 Hydroxyzine 3 N05BB51 N05BB51 N05BB51 N05BB51 31 Imipramin N06AA02 N06AA02 N06AA02 N06AA02 32 Imipraminoxid N06AA03 N06AA03 N06AA03 N06AA03 33 Indometacin 1 M01AB01 M01AB01 M01AB01 M01AB01 34 35 Indometacin 2 C01EB03 C01EB03 C01EB03 C01EB03 36 Indometacin 3 M01AB51 M01AB51 M01AB51 M01AB51
37 Indometacin 4 M02AA23 M02AA23 M02AA23 M02AA23 http://bmjopen.bmj.com/ 38 Indometacin 5 M02AA73 M02AA73 M02AA73 M02AA73 39 40 Ketoprofen 1 M01AE03 M01AE03 M01AE03 M01AE03 41 Ketoprofen 2 M01AE53 M01AE53 M01AE53 M01AE53 42 Ketoprofen 3 M02AA10 M02AA10 M02AA10 M02AA10 43 Levomepromazin N05AA02 N05AA02 N05AA02 N05AA02 44 Maprotilin N06AA21 N06AA21 N06AA21 N06AA21 45 on September 26, 2021 by guest. Protected copyright. 46 Medazepam N05BA03 N05BA03 N05BA03 N05BA03 47 Meloxicam 1 M01AC06 M01AC06 M01AC06 M01AC06 48 Meloxicam 2 M01AC56 M01AC56 n/a n/a 49 Methyldopa 1 C02AB01 C02AB01 C02AB01 C02AB01 50 Methyldopa 2 C02AB02 C02AB02 C02AB02 C02AB02 51 52 Methyldopa 3 C02LB01 C02LB01 C02LB01 C02LB01 53 Metildigoxin 1 C01AA08 C01AA08 C01AA08 C01AA08 54 Metildigoxin 2 C01AA58 C01AA58 C01AA58 C01AA58 55 Naftidrofuryl C04AX21 C04AX21 C04AX21 C04AX21 56 Nicergoline 1 C04AE02 C04AE02 C04AE02 C04AE02 57 58 Nicergoline 2 N06DX13 N06DX13 N06DX13 N06DX13 59 Nifedipine 1 C08CA05 C08CA05 C08CA05 C08CA05 60 Nifedipine 2 C08CA55 C08CA55 C08CA55 C08CA55 Nifedipine 3 C08GA01 C08GA01 C08GA01 C08GA01
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 54 of 90
1 2 Nifedipine 4 C07FB22 C07FB22 C07FB22 C07FB22
3 Nifedipine 5 C07FB23 C07FB23 C07FB23 C07FB23 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Nitrazepam N05CD02 N05CD02 N05CD02 N05CD02 5 6 Nitrofurantoin 1 J01XE01 J01XE01 J01XE01 J01XE01 7 Nitrofurantoin 2 J01XE51 J01XE51 J01XE51 J01XE51 8 Oxybutynin G04BD04 G04BD04 G04BD04 G04BD04 9 Pentoxifyllin C04AD03 C04AD03 C04AD03 C04AD03 10 Perphenazin N05AB03 N05AB03 N05AB03 N05AB03 11 12 Pethidine (=Meperidine) 1 N02AB02 N02AB02 N02AB02 N02AB02 13 Pethidine (=Meperidine) 2 N02AB52 N02AB52 N02AB52 N02AB52 14 Pethidine (=Meperidine) 3 N02AB72 N02AB72 N02AB72 N02AB72 15 Pethidine (=Meperidine) 4 N02AG03 N02AG03 N02AG03 N02AG03 16 17 Phenobarbital 1 N03AA02 N03AA02 N03AA02 N03AA02 18 Phenobarbital 2 For peer N05CA24review N05CA24 only N05CA24 N05CA24 19 Phenylbutazone 1 M01AA01 M01AA01 M01AA01 M01AA01 20 Phenylbutazone 2 M01AA51 M01AA51 M01AA51 M01AA51 21 Phenylbutazone 3 M01BA01 M01BA01 M01BA01 M01BA01 22 23 Phenylbutazone 4 R05XA10 R05XA10 R05XA10 R05XA10 24 Phenylbutazone 5 M02AA01 M02AA01 M02AA01 M02AA01 25 Piracetam N06BX03 N06BX03 N06BX03 N06BX03 26 Piroxicam 1 M01AC01 M01AC01 M01AC01 M01AC01 27 Piroxicam 2 M02AA07 M02AA07 M02AA07 M02AA07 28 29 Prasugrel B01AC22 B01AC22 B01AC22 B01AC22 30 Prazepam N05BA11 N05BA11 N05BA11 N05BA11 31 Prazosin 1 C02CA01 C02CA01 C02CA01 C02CA01 32 Prazosin 2 C02LE01 C02LE01 C02LE01 C02LE01 33 Reserpine 1 C02AA02 C02AA02 C02AA02 C02AA02 34 35 Reserpine 2 C02AA52 C02AA52 C02AA52 C02AA52 36 Reserpine 3 C02LA01 C02LA01 C02LA01 C02LA01
37 Reserpine 4 C02LA51 C02LA51 C02LA51 C02LA51 http://bmjopen.bmj.com/ 38 Reserpine 5 C02LA71 C02LA71 C02LA71 C02LA71 39 40 Reserpine 6 N05AX15 N05AX15 N05AX15 N05AX15 41 Solifenacin 1 G04BD08 G04BD08 G04BD08 G04BD08 42 Solifenacin 2 G04CA53 n/a n/a n/a 43 Sotalol 1 C07AA07 C07AA07 C07AA07 C07AA07 44 Sotalol 2 C07AA57 C07AA57 C07AA57 C07AA57 45 on September 26, 2021 by guest. Protected copyright. 46 Sotalol 3 C07BA07 C07BA07 C07BA07 C07BA07 47 Temazepam N05CD07 N05CD07 N05CD07 N05CD07 48 Terazosin 1 G04CA03 G04CA03 G04CA03 G04CA03 49 Terazosin 2 C02CA08 C02CA08 C02CA08 C02CA08 50 Tetrazepam M03BX07 M03BX07 M03BX07 M03BX07 51 52 Thioridazin N05AC02 N05AC02 N05AC02 N05AC02 53 Ticlopidin B01AC05 B01AC05 B01AC05 B01AC05 54 Tolterodin G04BD07 G04BD07 G04BD07 G04BD07 55 Tranylcypromine 1 N06AF04 N06AF04 N06AF04 N06AF04 56 Tranylcypromine 2 N06CA07 N06CA03 N06CA03 N06CA03 57 58 Triazolam N05CD05 N05CD05 N05CD05 N05CD05 59 Trimipramin N06AA06 N06AA06 N06AA06 N06AA06 60 Verapamil in Kombination mit Chinidin C08DA81 C08DA81 C08DA81 C08DA81
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 55 of 90 BMJ Open
1 2 Anticholinergic Drug Scale ATC-Code ATC-Code ATC-Code ATC-Code Score
3 2014 2013 2012 2011 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Dexamethasone 2 A01AC02 A01AC02 A01AC02 A01AC02 1 5 6 Hydrocortisone 2 A01AC03 A01AC03 A01AC03 A01AC03 1 7 Cimetidine A02BA01 A02BA01 A02BA01 A02BA01 2 8 Ranitidine 1 A02BA02 A02BA02 A02BA02 A02BA02 2 9 Famotidine 1 A02BA03 A02BA03 A02BA03 A02BA03 1 10 Nizatidine A02BA04 A02BA04 A02BA04 A02BA04 1 11 12 Ranitidine 2 A02BA07 A02BA07 A02BA07 A02BA07 2 13 Famotidine 2 A02BA53 A02BA53 A02BA53 A02BA53 1 14 Omeprazole A02BC01 A02BC01 A02BC01 A02BC01 0 15 Pantoprazole A02BC02 A02BC02 A02BC02 A02BC02 0 16 17 Lansoprazole A02BC03 A02BC03 A02BC03 A02BC03 0 18 Rabeprazol ForA02BC04 peer A02BC04 review A02BC04 only A02BC04 0 19 Dicyclomine (=Dicycloverin) 1 A03AA07 A03AA07 A03AA07 A03AA07 3 20 21 Propantheline 1 A03AB05 A03AB05 A03AB05 A03AB05 3 22 23 Atropine 1 A03BA01 A03BA01 A03BA01 A03BA01 3 24 Hyoscyamine 1 A03BA03 A03BA03 A03BA03 A03BA03 3 25 Propantheline 2 A03CA34 A03CA34 A03CA34 A03CA34 3 26 Hyoscyamine 2 A03CB31 A03CB31 A03CB31 A03CB31 3 27 Metoclopramide 1 A03FA01 A03FA01 A03FA01 A03FA01 0 28 29 Metoclopramide 2 A03FA51 A03FA51 A03FA51 A03FA51 0 30 Dimenhydrinate 1 A04AB02 A04AB02 A04AB02 A04AB02 3 31 Meclozine 3 A04AB04 A04AB04 A04AB04 A04AB04 3 32 Diphenhydramine 3 A04AB05 A04AB05 A04AB05 A04AB05 3 33 Dimenhydrinate 2 A04AB52 A04AB52 A04AB52 A04AB52 3 34 35 Meclozine 4 A04AB54 A04AB54 A04AB54 A04AB54 3 36 Diphenhydramine 6 A04AB55 A04AB55 A04AB55 A04AB55 3
37 Promethazine 3 A04AB58 A04AB58 A04AB58 A04AB58 3 http://bmjopen.bmj.com/ 38 Scopolamine 1 (=Hyoscin) A04AD01 A04AD01 A04AD01 A04AD01 3 39 40 Scopolamine 2 (=Hyoscin) A04AD51 A04AD51 A04AD51 A04AD51 3 41 Bisacodyl A06AB02 A06AB02 A06AB02 A06AB02 0 42 Senna glycosides 1 A06AB06 A06AB06 A06AB06 A06AB06 0 43 Senna glycosides 2 A06AB56 A06AB56 A06AB56 A06AB56 0 44 Vancomycin 2 A07AA09 A07AA09 A07AA09 A07AA09 1 45 on September 26, 2021 by guest. Protected copyright. 46 Diphenoxylate A07DA01 A07DA01 A07DA01 A07DA01 0 47 Loperamide 1 A07DA03 A07DA03 A07DA03 A07DA03 1 48 Morphine 2 A07DA52 A07DA52 A07DA52 A07DA52 1 49 Loperamide 2 A07DA53 A07DA53 A07DA53 A07DA53 1 50 Insulin 1 A10AB01 A10AB01 A10AB01 A10AB01 0 51 52 Insulin 2 A10AB02 A10AB02 A10AB02 A10AB02 0 53 Insulin 3 A10AB03 A10AB03 A10AB03 A10AB03 0 54 Insulin 4 A10AB04 A10AB04 A10AB04 A10AB04 0 55 Insulin 5 A10AB05 A10AB05 A10AB05 A10AB05 0 56 Insulin 6 A10AB06 A10AB06 A10AB06 A10AB06 0 57 58 Insulin 7 A10AB30 A10AB30 A10AB30 A10AB30 0 59 Insulin 8 A10AC01 A10AC01 A10AC01 A10AC01 0 60 Insulin 9 A10AC02 A10AC02 A10AC02 A10AC02 0 Insulin 10 A10AC03 A10AC03 A10AC03 A10AC03 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 56 of 90
1 2 Insulin 11 A10AC04 A10AC04 A10AC04 A10AC04 0
3 Insulin 12 A10AC30 A10AC30 A10AC30 A10AC30 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Insulin 13 A10AD01 A10AD01 A10AD01 A10AD01 0 5 6 Insulin 14 A10AD02 A10AD02 A10AD02 A10AD02 0 7 Insulin 15 A10AD03 A10AD03 A10AD03 A10AD03 0 8 Insulin 16 A10AD04 A10AD04 A10AD04 A10AD04 0 9 Insulin 17 A10AD05 A10AD05 A10AD05 A10AD05 0 10 Insulin 18 A10AD30 A10AD30 A10AD30 A10AD30 0 11 12 Insulin 19 A10AE01 A10AE01 A10AE01 A10AE01 0 13 Insulin 20 A10AE02 A10AE02 A10AE02 A10AE02 0 14 Insulin 21 A10AE03 A10AE03 A10AE03 A10AE03 0 15 Insulin 22 A10AE04 A10AE04 A10AE04 A10AE04 0 16 17 Insulin 23 A10AE05 A10AE05 A10AE05 A10AE05 0 18 Insulin 24 ForA10AE30 peer A10AE30 review A10AE30 only A10AE30 0 19 Insulin 25 A10AF01 A10AF01 A10AF01 A10AF01 0 20 Metformin A10BA02 A10BA02 A10BA02 A10BA02 0 21 Glipizide A10BB07 A10BB07 A10BB07 A10BB07 0 22 23 Pioglitazone 3 A10BD05 A10BD05 A10BD05 A10BD05 0 24 Pioglitazone 2 A10BD06 A10BD06 A10BD06 A10BD06 0 25 Pioglitazone 4 A10BD09 A10BD09 A10BD09 A10BD09 0 26 Pioglitazone 5 A10BD12 n/a n/a n/a 0 27 Rosiglitazone A10BG02 A10BG02 A10BG02 A10BG02 0 28 29 Pioglitazone 1 A10BG03 A10BG03 A10BG03 A10BG03 0 30 Cyproheptadine 2 A15AA01 A15AA01 A15AA01 A15AA01 2 31 Cyproheptadine 3 A15AA51 A15AA51 A15AA51 A15AA51 2 32 Clopidogrel B01AC04 B01AC04 B01AC04 B01AC04 0 33 Dipyridamole 1 B01AC07 B01AC07 B01AC07 B01AC07 1 34 35 Dipyridamole 2 B01AC36 n/a n/a n/a 1 36 Digitoxin 1 C01AA04 C01AA04 C01AA04 C01AA04 1
37 Digoxin 1 C01AA05 C01AA05 C01AA05 C01AA05 1 http://bmjopen.bmj.com/ 38 Digitoxin 2 C01AA54 C01AA54 C01AA54 C01AA54 1 39 40 Digoxin 2 C01AA55 C01AA55 C01AA55 C01AA55 1 41 Disopyramide C01BA03 C01BA03 C01BA03 C01BA03 2 42 Nitroglycerin C01DA02 C01DA02 C01DA02 C01DA02 0 43 Isosorbidedinitrate 1 C01DA08 C01DA08 C01DA08 C01DA08 1 44 Isosorbidemononitrate C01DA14 C01DA14 C01DA14 C01DA14 1 45 on September 26, 2021 by guest. Protected copyright. 46 Isosorbidedinitrate 2 C01DA58 C01DA58 C01DA58 C01DA58 1 47 Dipyridamole 4 C01DX21 C01DX21 C01DX21 C01DX21 1 48 Dipyridamole 3 C01DX71 C01DX71 C01DX71 C01DX71 1 49 Theophylline 2 C01EB22 C01EB22 C01EB22 C01EB22 1 50 Theophylline 3 C01EX66 C01EX66 C01EX66 C01EX66 1 51 52 Hydralazine 1 C02DB02 C02DB02 C02DB02 C02DB02 1 53 Hydralazine 2 C02LG02 C02LG02 C02LG02 C02LG02 1 54 Hydrochlorothiazide C03AA03 C03AA03 C03AA03 C03AA03 0 55 Chlortalidone 1 C03BA04 C03BA04 C03BA04 C03BA04 1 56 Chlortalidone 2 C03BB04 C03BB04 C03BB04 C03BB04 1 57 58 Furosemide 1 C03CA01 C03CA01 C03CA01 C03CA01 1 59 Furosemide 2 C03CB01 C03CB01 C03CB01 C03CB01 1 60 Spironolactone C03DA01 C03DA01 C03DA01 C03DA01 0 Triamterene 1 C03DB02 C03DB02 C03DB02 C03DB02 1
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 57 of 90 BMJ Open
1 2 Chlortalidone 3 C03EA06 C03EA06 C03EA06 C03EA06 1
3 Triamterene 2 C03EA21 C03EA21 C03EA21 C03EA21 1 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Furosemide 3 C03EB01 C03EB01 C03EB01 C03EB01 1 5 6 Furosemide 4 + Triamterene C03EB21 C03EB21 C03EB21 C03EB21 1 7 3 8 Isosorbidedinitrate 3 C05AE02 C05AE02 C05AE02 C05AE02 1 9 Digitoxin 3 C05BZ05 C05BZ05 C05BZ05 C05BZ05 1 10 Propranolol 1 C07AA05 C07AA05 C07AA05 C07AA05 0 11 12 Timolol C07AA06 C07AA06 C07AA06 C07AA06 0 13 Metoprolol 1 C07AB02 C07AB02 C07AB02 C07AB02 0 14 Atenolol 1 C07AB03 C07AB03 C07AB03 C07AB03 0 15 Atenolol 8 (Enantiomer S- C07AB11 C07AB11 C07AB11 C07AB11 0 16 17 Atenolol) 18 Metoprolol 2 ForC07AB52 peer C07AB52 review C07AB52 only C07AB52 0 19 Propranolol 5 C07BA05 C07BA05 C07BA05 C07BA05 0 20 Metoprolol 3 C07BB02 C07BB02 C07BB02 C07BB02 0 21 Atenolol 2 C07BB03 C07BB03 C07BB03 C07BB03 0 22 23 Metoprolol 4 C07BB52 C07BB52 C07BB52 C07BB52 0 24 Propranolol 4 C07CA05 C07CA05 C07CA05 C07CA05 0 25 Metoprolol 5 C07CB02 C07CB02 C07CB02 C07CB02 0 26 Atenolol 3 C07CB03 C07CB03 C07CB03 C07CB03 0 27 Atenolol 4 C07CB53 C07CB53 C07CB53 C07CB53 0 28 29 Propranolol 2 C07DA05 C07DA05 C07DA05 C07DA05 0 30 Atenolol 5 C07DB01 C07DB01 C07DB01 C07DB01 0 31 Propranolol 6 C07EA05 C07EA05 C07EA05 C07EA05 0 32 Propranolol 3 C07FA05 C07FA05 C07FA05 C07FA05 0 33 Metoprolol 6 C07FB02 C07FB02 C07FB02 C07FB02 0 34 35 Atenolol 7 C07FB03 C07FB03 C07FB03 C07FB03 0 36 Metoprolol 7 C07FB22 C07FB22 C07FB22 C07FB22 0
37 Atenolol 6 C07FB23 C07FB23 C07FB23 C07FB23 0 http://bmjopen.bmj.com/ 38 Metoprolol 8 C07FB24 C07FB24 C07FB24 C07FB24 0 39 40 Amlodipine C08CA01 C08CA01 C08CA01 C08CA01 0 41 Verapamil 1 C08DA01 C08DA01 C08DA01 C08DA01 0 42 Verapamil 5 C08DA51 C08DA51 C08DA51 C08DA51 0 43 Verapamil 2 C08DA81 C08DA81 C08DA81 C08DA81 0 44 Diltiazem C08DB01 C08DB01 C08DB01 C08DB01 1 45 on September 26, 2021 by guest. Protected copyright. 46 Verapamil 3 C08GA02 C08GA02 C08GA02 C08GA02 0 47 Captopril 1 C09AA01 C09AA01 C09AA01 C09AA01 1 48 Enalapril C09AA02 C09AA02 C09AA02 C09AA02 0 49 Lisinopril C09AA03 C09AA03 C09AA03 C09AA03 0 50 Benazepril C09AA07 C09AA07 C09AA07 C09AA07 0 51 52 Trandolapril C09AA10 C09AA10 C09AA10 C09AA10 0 53 Captopril 2 C09BA01 C09BA01 C09BA01 C09BA01 1 54 Verapamil 4 C09BB10 C09BB10 C09BB10 C09BB10 0 55 Losartan C09CA01 C09CA01 C09CA01 C09CA01 0 56 Simvastatin C10AA01 C10AA01 C10AA01 C10AA01 0 57 58 Atorvastatin C10AA05 C10AA05 C10AA05 C10AA05 0 59 Gemfibrozil C10AB04 C10AB04 C10AB04 C10AB04 0 60 Bromocriptine 2 G02CB01 G02CB01 G02CB01 G02CB01 1 Flavoxate G04BD02 G04BD02 G04BD02 G04BD02 3
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 58 of 90
1 2 Oxybutynin G04BD04 G04BD04 G04BD04 G04BD04 3
3 Tolterodine G04BD07 G04BD07 G04BD07 G04BD07 3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Darifenacin G04BD10 G04BD10 G04BD10 G04BD10 3 5 6 Dicyclomine (=Dicycloverin) 2 G04BD13 G04BD13 G04BD13 G04BD13 3 7 8 Atropine 2 G04BD15 G04BD15 G04BD15 G04BD15 3 9 Dicyclomine (=Dicycloverin) 3 G04BD63 G04BD63 G04BD63 G04BD63 3 10 11 12 Atropine 3 G04BD65 G04BD65 G04BD65 G04BD65 3 13 Dexamethasone 1 H02AB02 H02AB02 H02AB02 H02AB02 1 14 Methylprednisolone 1 H02AB04 H02AB04 H02AB04 H02AB04 1 15 Prednisolone 1 H02AB06 H02AB06 H02AB06 H02AB06 1 16 17 Prednisolone 4 (Prednisone) H02AB07 H02AB07 H02AB07 H02AB07 1 18 For peer review only 19 Triamcinolone 1 H02AB08 H02AB08 H02AB08 H02AB08 1 20 Hydrocortisone 1 H02AB09 H02AB09 H02AB09 H02AB09 1 21 Cortisone H02AB10 H02AB10 H02AB10 H02AB10 1 22 23 Methylprednisolone 2 H02AB54 H02AB54 H02AB54 H02AB54 1 24 Prednisolone 2 H02AB56 H02AB56 H02AB56 H02AB56 1 25 Methylprednisolone 3 H02BX01 H02BX01 H02BX01 H02BX01 1 26 Dexamethasone 3 H02BX02 H02BX02 H02BX02 H02BX02 1 27 Prednisolone 3 H02BX06 H02BX06 H02BX06 H02BX06 1 28 29 Triamcinolone 2 H02BX08 H02BX08 H02BX08 H02BX08 1 30 Ampicillin 1 J01CA01 J01CA01 J01CA01 J01CA01 1 31 Ampicillin 5 (Prodrug J01CA02 J01CA02 J01CA02 J01CA02 1 32 Pivampicillin) 33 Amoxicillin J01CA04 J01CA04 J01CA04 J01CA04 0 34 35 Ampicillin 6 (Prodrug J01CA06 J01CA06 J01CA06 J01CA06 1 36 PiperacillinBacampicilin) 1 J01CA12 J01CA12 J01CA12 J01CA12 1 37 http://bmjopen.bmj.com/ Ampicillin 7 (Prodrug J01CA14 J01CA14 J01CA14 J01CA14 1 38 39 Metampicillin) 40 Ampicillin 8 (Prodrug J01CA15 J01CA15 J01CA15 J01CA15 1 41 Talampicillin) 42 Ampicillin 2 J01CA51 J01CA51 J01CA51 J01CA51 1 43 44 Ampicillin 3 J01CR01 J01CR01 J01CR01 J01CR01 1
45 Piperacillin 2 J01CR05 J01CR05 J01CR05 J01CR05 1 on September 26, 2021 by guest. Protected copyright. 46 Cefalexin J01DB01 J01DB01 J01DB01 J01DB01 0 47 Cefalotin J01DB03 J01DB03 J01DB03 J01DB03 1 48 Cefoxitin J01DC01 J01DC01 J01DC01 J01DC01 1 49 50 Cefamandole J01DC03 J01DC03 J01DC03 J01DC03 1 51 Trimethoprim J01EA01 J01EA01 J01EA01 J01EA01 0 52 Clindamycin J01FF01 J01FF01 J01FF01 J01FF01 1 53 Gentamicin 1 J01GB03 J01GB03 J01GB03 J01GB03 1 54 Gentamicin 2 J01GB53 J01GB53 J01GB53 J01GB53 1 55 56 Levofloxacin J01MA12 J01MA12 J01MA12 J01MA12 0 57 Vancomycin 1 J01XA01 J01XA01 J01XA01 J01XA01 1 58 Cycloserine J04AB01 J04AB01 J04AB01 J04AB01 1 59 Amantadine 2 J05AC04 J05AC04 J05AC04 J05AC04 1 60 Celecoxib 2 L01XX33 L01XX33 L01XX33 L01XX33 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 59 of 90 BMJ Open
1 2 Tamoxifen L02BA01 L02BA01 L02BA01 L02BA01 0
3 Ciclosporine L04AD01 L04AD01 L04AD01 L04AD01 1 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Azathioprine L04AX01 L04AX01 L04AX01 L04AX01 1 5 6 Methotrexate L04AX03 L04AX03 L04AX03 L04AX03 0 7 Piroxicam 1 M01AC01 M01AC01 M01AC01 M01AC01 0 8 Ibuprofen M01AE01 M01AE01 M01AE01 M01AE01 0 9 Ketoprofen 1 M01AE03 M01AE03 M01AE03 M01AE03 0 10 11 Ketoprofen 2 M01AE53 M01AE53 M01AE53 M01AE53 0 12 Celecoxib 1 M01AH01 M01AH01 M01AH01 M01AH01 0 13 Piroxicam 2 M02AA07 M02AA07 M02AA07 M02AA07 0 14 Ketoprofen 3 M02AA10 M02AA10 M02AA10 M02AA10 0 15 Pancuronium M03AC01 M03AC01 M03AC01 M03AC01 1 16 17 Carisoprodol 1 M03BA02 M03BA02 M03BA02 M03BA02 0 18 Carisoprodol 2 ForM03BA52 peer M03BA52 review M03BA52 only M03BA52 0 19 Carisoprodol 3 M03BA72 M03BA72 M03BA72 M03BA72 0 20 Orphenadrine 1 M03BC01 M03BC01 M03BC01 M03BC01 3 21 Orphenadrine 2 M03BC51 M03BC51 M03BC51 M03BC51 3 22 23 Baclofen M03BX01 M03BX01 M03BX01 M03BX01 0 24 Cyclobenzaprine M03BX08 M03BX08 M03BX08 M03BX08 2 25 Allopurinol M04AA01 M04AA01 M04AA01 M04AA01 0 26 Colchicine M04AC01 M04AC01 M04AC01 M04AC01 0 27 Fentanyl 2 N01AH01 N01AH01 N01AH01 N01AH01 1 28 29 Fentanyl 3 N01AH51 N01AH51 N01AH51 N01AH51 1 30 Diphenhydramine 4 N01BX06 N01BX06 N01BX06 N01BX06 3 31 Morphine 1 N02AA01 N02AA01 N02AA01 N02AA01 1 32 Oxycodone 1 N02AA05 N02AA05 N02AA05 N02AA05 1 33 34 Codeine 10 (Dihydrocodeine N02AA08 N02AA08 N02AA08 N02AA08 1 35 1) 36 Morphine 3 N02AA51 N02AA51 N02AA51 N02AA51 1 37 Oxycodone 2 N02AA55 N02AA55 N02AA55 N02AA55 1 http://bmjopen.bmj.com/ 38 Codeine 11 (Dihydrocodeine N02AA58 N02AA58 N02AA58 N02AA58 1 39 40 2) 41 Codeine 2 N02AA59 N02AA59 N02AA59 N02AA59 1 42 Codeine 3 N02AA64 N02AA64 N02AA64 N02AA64 1 43 Codeine 4 N02AA65 N02AA65 N02AA65 N02AA65 1 44 Codeine 5 N02AA66 N02AA66 N02AA66 N02AA66 1 45 on September 26, 2021 by guest. Protected copyright. 46 Codeine 6 N02AA69 N02AA69 N02AA69 N02AA69 1 47 Codeine 7 N02AA79 N02AA79 N02AA79 N02AA79 1 48 Pethidine (=Meperidine) 1 N02AB02 N02AB02 N02AB02 N02AB02 2 49 50 51 Fentanyl 1 N02AB03 N02AB03 N02AB03 N02AB03 1 52 Pethidine (=Meperidine) 2 2 53 N02AB52 N02AB52 N02AB52 N02AB52 54 Pethidine (=Meperidine) 3 2 55 N02AB72 N02AB72 N02AB72 N02AB72 56 57 Morphine 4 N02AG01 N02AG01 N02AG01 N02AG01 1 58 Pethidine (=Meperidine) 4 2 59 N02AG03 N02AG03 N02AG03 N02AG03 60 Tramadol 1 N02AX02 N02AX02 N02AX02 N02AX02 1 Tramadol 2 N02AX52 N02AX52 N02AX52 N02AX52 1
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 60 of 90
1 2 Acetylsalicylicacid N02BA01 N02BA01 N02BA01 N02BA01 0
3 Paracetamol N02BE01 N02BE01 N02BE01 N02BE01 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Topiramate 2 N02CX12 N02CX12 N02CX12 N02CX12 0 5 6 Codeine 8 N02CX58 N02CX58 N02CX58 N02CX58 1 7 Metoclopramide 3 N02CX59 N02CX59 N02CX59 N02CX59 0 8 Phenobarbital 1 N03AA02 N03AA02 N03AA02 N03AA02 0 9 Phenytoin 1 N03AB02 N03AB02 N03AB02 N03AB02 0 10 Phenytoin 3 (Fosphenytoin) N03AB05 N03AB05 N03AB05 N03AB05 0 11 12 13 Phenytoin 2 N03AB52 N03AB52 N03AB52 N03AB52 0 14 Clonazepam N03AE01 N03AE01 N03AE01 N03AE01 1 15 Midazolam 2 N03AE02 N03AE02 n/a n/a 1 16 17 Carbamazepine N03AF01 N03AF01 N03AF01 N03AF01 2 18 Oxcarbazepine ForN03AF02 peer N03AF02 review N03AF02 only N03AF02 2 19 Valproic acid / Sodium N03AG01 N03AG01 N03AG01 N03AG01 1 20 Valproate 21 Topiramate 1 N03AX11 N03AX11 N03AX11 N03AX11 0 22 23 Trihexyphenidyl N04AA01 N04AA01 N04AA01 N04AA01 3 24 Procyclidine N04AA04 N04AA04 N04AA04 N04AA04 3 25 Orphenadrine 3 N04AB02 N04AB02 N04AB02 N04AB02 3 26 Benzatropine 1 N04AC01 N04AC01 N04AC01 N04AC01 3 27 Benzatropine 2 N04AC30 N04AC30 N04AC30 N04AC30 3 28 29 (Etybenzatropin) 30 Levodopa 1 N04BA01 N04BA01 N04BA01 N04BA01 0 31 Entacapone 2 + Levodopa 3 N04BA03 N04BA03 N04BA03 N04BA03 0 32 33 Levodopa 6 (Melevodopa 1) N04BA04 N04BA04 N04BA04 N04BA04 0 34 35 36 Levodopa 7 (Melevodopa 2) N04BA05 N04BA05 N04BA05 N04BA05 0
37 http://bmjopen.bmj.com/ 38 Levodopa 5 (Etilevodopa) N04BA06 N04BA06 N04BA06 N04BA06 0 39 40 Levodopa 2 N04BA10 N04BA10 N04BA10 N04BA10 0 41 Levodopa 4 N04BA11 N04BA11 N04BA11 N04BA11 0 42 Amantadine 1 N04BB01 N04BB01 N04BB01 N04BB01 1 43 Bromocriptine 1 N04BC01 N04BC01 N04BC01 N04BC01 1 44 Ropinirole N04BC04 N04BC04 N04BC04 N04BC04 0 45 on September 26, 2021 by guest. Protected copyright. 46 Pramipexol N04BC05 N04BC05 N04BC05 N04BC05 0 47 Selegiline N04BD01 N04BD01 N04BD01 N04BD01 0 48 Entacapone 1 N04BX02 N04BX02 N04BX02 N04BX02 0 49 Chlorpromazine N05AA01 N05AA01 N05AA01 N05AA01 3 50 Levomepromazine N05AA02 N05AA02 N05AA02 N05AA02 2 51 52 Fluphenazine N05AB02 N05AB02 N05AB02 N05AB02 1 53 Perphenazine N05AB03 N05AB03 N05AB03 N05AB03 1 54 Prochlorperazine N05AB04 N05AB04 N05AB04 N05AB04 1 55 Trifluoperazine N05AB06 N05AB06 N05AB06 N05AB06 1 56 Thioridazine N05AC02 N05AC02 N05AC02 N05AC02 3 57 58 Haloperidol N05AD01 N05AD01 N05AD01 N05AD01 0 59 Molindone N05AE02 N05AE02 N05AE02 N05AE02 2 60 Tiotixene N05AF04 N05AF04 N05AF04 N05AF04 1 Pimozide N05AG02 N05AG02 N05AG02 N05AG02 2
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 61 of 90 BMJ Open
1 2 Loxapine N05AH01 N05AH01 N05AH01 N05AH01 2
3 Clozapine N05AH02 N05AH02 N05AH02 N05AH02 3 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Olanzapine N05AH03 N05AH03 N05AH03 N05AH03 1 5 6 Quetiapine (fumarate) N05AH04 N05AH04 N05AH04 N05AH04 0 7 Lithium N05AN01 N05AN01 N05AN01 N05AN01 0 8 Diazepam N05BA01 N05BA01 N05BA01 N05BA01 1 9 Chlordiazepoxide N05BA02 N05BA02 N05BA02 N05BA02 1 10 Oxazepam N05BA04 N05BA04 N05BA04 N05BA04 1 11 12 Clorazepate N05BA05 N05BA05 N05BA05 N05BA05 1 13 Lorazepam 1 N05BA06 N05BA06 N05BA06 N05BA06 1 14 Alprazolam N05BA12 N05BA12 N05BA12 N05BA12 1 15 Lorazepam 2 N05BA56 N05BA56 N05BA56 N05BA56 1 16 17 Hydroxyzine 1 N05BB01 N05BB01 N05BB01 N05BB01 3 18 Hydroxyzine 3 ForN05BB51 peer N05BB51 review N05BB51 only N05BB51 3 19 Phenobarbital 2 N05CA24 N05CA24 N05CA24 N05CA24 0 20 Flurazepam N05CD01 N05CD01 N05CD01 N05CD01 1 21 Estazolam N05CD04 N05CD04 N05CD04 N05CD04 1 22 23 Triazolam N05CD05 N05CD05 N05CD05 N05CD05 1 24 Temazepam N05CD07 N05CD07 N05CD07 N05CD07 1 25 Midazolam 1 N05CD08 N05CD08 N05CD08 N05CD08 1 26 Zopiclone N05CF01 N05CF01 N05CF01 N05CF01 0 27 Zolpidem N05CF02 N05CF02 N05CF02 N05CF02 0 28 29 Scopolamine 3 (=Hyoscin) N05CM05 N05CM05 N05CM05 N05CM05 3 30 Diphenhydramine 5 N05CM20 N05CM20 N05CM20 N05CM20 3 31 Promethazine 4 N05CM22 N05CM22 N05CM22 N05CM22 3 32 Diphenhydramine 2 N05CX07 N05CX07 N05CX07 N05CX07 3 33 Promethazine 5 N05CX13 N05CX13 N05CX13 N05CX13 3 34 35 Desipramine N06AA01 N06AA01 N06AA01 N06AA01 3 36 Imipramine 1 N06AA02 N06AA02 N06AA02 N06AA02 3
37 Imipramine 2 N06AA03 N06AA03 N06AA03 N06AA03 3 http://bmjopen.bmj.com/ 38 Clomipramine N06AA04 N06AA04 N06AA04 N06AA04 3 39 40 Trimipramine N06AA06 N06AA06 N06AA06 N06AA06 3 41 Amitriptyline 1 N06AA09 N06AA09 N06AA09 N06AA09 3 42 Nortriptyline 1 N06AA10 N06AA10 N06AA10 N06AA10 3 43 Protriptyline N06AA11 N06AA11 N06AA11 N06AA11 3 44 Doxepin N06AA12 N06AA12 N06AA12 N06AA12 3 45 on September 26, 2021 by guest. Protected copyright. 46 Amitriptyline 3 N06AA25 N06AA25 N06AA25 N06AA25 3 47 Fluoxetine 1 N06AB03 N06AB03 N06AB03 N06AB03 1 48 Citalopram N06AB04 N06AB04 N06AB04 N06AB04 0 49 Paroxetine N06AB05 N06AB05 N06AB05 N06AB05 1 50 Fluvoxamine N06AB08 N06AB08 N06AB08 N06AB08 1 51 52 Escitalopram N06AB10 N06AB10 N06AB10 N06AB10 0 53 Phenelzine N06AF03 N06AF03 N06AF03 N06AF03 1 54 Trazodone N06AX05 N06AX05 N06AX05 N06AX05 0 55 Nefazodone N06AX06 N06AX06 N06AX06 N06AX06 0 56 Mirtazapine N06AX11 N06AX11 N06AX11 N06AX11 0 57 58 Bupropion 1 N06AX12 N06AX12 N06AX12 N06AX12 0 59 Venlafaxine 1 N06AX16 N06AX16 N06AX16 N06AX16 0 60 Duloxetine N06AX21 N06AX21 N06AX21 N06AX21 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 62 of 90
1 2 Venlafaxine 2 N06AX23 N06AX23 N06AX23 N06AX23 0
3 (Desvenlafaxine) BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Amitriptyline 2 N06CA01 N06CA01 N06CA01 N06CA01 3 5 6 Fluoxetine 2 N06CA03 n/a n/a n/a 1 7 Nortriptyline 2 N06CA06 N06CA06 N06CA06 N06CA06 3 8 Donepezil N06DA02 N06DA02 N06DA02 N06DA02 0 9 Galantamine N06DA04 N06DA04 N06DA04 N06DA04 0 10 Bupropion 2 N07BA02 N07BA02 N07BA02 N07BA02 0 11 12 Pseudoephedrine 1 R01BA02 R01BA02 R01BA02 R01BA02 0 13 Pseudoephedrine 2 R01BA52 R01BA52 R01BA52 R01BA52 0 14 Terbutaline (oral) 2 R03AC03 R03AC03 R03AC03 R03AC03 0 15 Fluticasone-salmeterol R03AK06 R03AK61 R03AK61 R03AK61 1 16 17 Ipratropium R03BB01 R03BB01 R03BB01 R03BB01 0 18 Terbutaline (oral) 1For R03CC03 peer R03CC03 review R03CC03 only R03CC03 0 19 Terbutaline (oral) 3 R03CC53 R03CC53 R03CC53 R03CC53 0 20 Theophylline 1 R03DA04 R03DA04 R03DA04 R03DA04 1 21 Theophylline 4 R03DA54 R03DA54 R03DA54 R03DA54 1 22 23 Theophylline 5 R03DA74 R03DA74 R03DA74 R03DA74 1 24 Theophylline 6 R03DB04 R03DB04 R03DB04 R03DB04 1 25 Guaifenesin R05CA03 R05CA03 R05CA03 R05CA03 0 26 Hydrocodone R05DA03 R05DA03 R05DA03 R05DA03 0 27 Codeine 1 R05DA04 R05DA04 R05DA04 R05DA04 1 28 29 Morphine 5 R05DA05 R05DA05 R05DA05 R05DA05 1 30 Codeine 12 (Dihydrocodeine R05DA14 R05DA14 R05DA14 R05DA14 1 31 3) 32 Codeine 9 R05DA54 R05DA54 R05DA54 R05DA54 1 33 34 Codeine 13 (Dihydrocodeine R05DA64 R05DA64 R05DA64 R05DA64 1 35 4) 36 Ampicillin 4 R05GB05 R05GB05 R05GB05 R05GB05 1 37 Diphenhydramine 1 R06AA02 R06AA02 R06AA02 R06AA02 3 http://bmjopen.bmj.com/ 38 Clemastine 1 R06AA04 R06AA04 R06AA04 R06AA04 3 39 40 Carbinoxamine R06AA08 R06AA08 R06AA08 R06AA08 3 41 Diphenhydramine 7 R06AA52 R06AA52 R06AA52 R06AA52 3 42 Clemastine 2 R06AA54 R06AA54 R06AA54 R06AA54 3 43 Brompheniramine 1 R06AB01 R06AB01 R06AB01 R06AB01 3 44 Chlorpheniramine R06AB04 R06AB04 R06AB04 R06AB04 3 45 on September 26, 2021 by guest. Protected copyright. 46 (=Chlorphenamine) 1 47 Brompheniramine 2 R06AB51 R06AB51 R06AB51 R06AB51 3 48 Chlorpheniramine 3 49 (=Chlorphenamine) 2 R06AB54 R06AB54 R06AB54 R06AB54 50 Pyrilamine R06AC01 R06AC01 R06AC01 R06AC01 3 51 52 Promethazine 1 R06AD02 R06AD02 R06AD02 R06AD02 3 53 Promethazine 8 3 54 (Hydroxyethylpromethazin 1) 55 R06AD05 R06AD05 R06AD05 R06AD05 56 57 Promethazine 7 R06AD10 R06AD10 R06AD10 R06AD10 3 58 (Dioxopromethazin) 59 Promethazine 2 R06AD52 R06AD52 R06AD52 R06AD52 3 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 63 of 90 BMJ Open
1 2 Promethazine 9 R06AD55 R06AD55 R06AD55 R06AD55 3
3 (Hydroxyethylpromethazin 2) BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 Meclozine 1 R06AE05 R06AE05 R06AE05 R06AE05 3 7 Cetirizin 1 R06AE07 R06AE07 R06AE07 R06AE07 0 8 Cetirizin 3 (Enantiomer R06AE09 R06AE09 R06AE09 R06AE09 0 9 Levocetirizin ) 10 Meclozine 2 R06AE55 R06AE55 R06AE55 R06AE55 3 11 12 Cetirizin 2 R06AE57 R06AE57 R06AE57 R06AE57 0 13 Cyproheptadine 1 R06AX02 R06AX02 R06AX02 R06AX02 2 14 Loratadine 1 R06AX13 R06AX13 R06AX13 R06AX13 0 15 Ketotifen R06AX17 R06AX17 R06AX17 R06AX17 1 16 17 Fexofenadine R06AX26 R06AX26 R06AX26 R06AX26 0 18 Loratadine 2 (EnantiomerFor R06AX27 peer R06AX27 review R06AX27 only R06AX27 0 19 Desloratadin ) 20 Hydroxyzine 2 R06AX32 R06AX32 R06AX32 R06AX32 3 21 Promethazine 6 V03AB05 V03AB05 V03AB05 V03AB05 3 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 64 of 90
1 2 Drug Burden Index
3 Medication IDIS ingredient Medication ATC-Code ATC-Code ATC-Code ATC-Code BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 code (IDIS=Iowa Drug 2014 2013 2012 2011 5 6 Information System) 7 Dicyclomine 12080005 Dicycloverine A03AA07 A03AA07 A03AA07 A03AA07 8 Propantheline 12080008 Propantheline 1 A03AB05 A03AB05 A03AB05 A03AB05 9 Papaverine 86000007 Papaverine 1 A03AD01 A03AD01 A03AD01 A03AD01 10 Belladonna alkaloids 2 A03BA01 A03BA01 A03BA01 A03BA01 11 12 Hyoscyamine 12080079 Hyoscyamine 1 + Belladonna A03BA03 A03BA03 A03BA03 A03BA03 13 alkaloids 3 14 Belladonna alkaloids 4 A03BA04 A03BA04 A03BA04 A03BA04 15 Belladonna alkaloids 5 A03BA20 A03BA20 A03BA20 A03BA20 16 17 Belladonna alkaloids 6 A03BB01 A03BB01 A03BB01 A03BB01 18 ForBelladonna peer alkaloids review 7 A03BB02 only A03BB02 A03BB02 A03BB02 19 Methscopolamine Methscopolamine 1 A03BB03 A03BB03 A03BB03 A03BB03 20 12080007 (Belladonna alkaloids 8) 21 Belladonna alkaloids 9 A03BB04 A03BB04 A03BB04 A03BB04 22 23 Belladonna alkaloids 10 A03BB05 A03BB05 A03BB05 A03BB05 24 Clidinium 12080047 Clidinium A03CA02 A03CA02 A03CA02 A03CA02 25 Propantheline 2 A03CA34 A03CA34 A03CA34 A03CA34 26 Methscopolamine 2 A03CB01 A03CB01 A03CB01 A03CB01 27 (Belladonna alkaloids 11) 28 29 Belladonna alkaloids 12 A03CB02 A03CB02 A03CB02 A03CB02 30 Belladonna alkaloids 13 A03CB03 A03CB03 A03CB03 A03CB03 31 Belladonna alkaloids 14 A03CB04 A03CB04 A03CB04 A03CB04 32 Hyoscyamine 2 (Belladonna 33 alkaloids 15) A03CB31 A03CB31 A03CB31 A03CB31 34 35 Belladonna alkaloids 16 A03CB37 A03CB37 A03CB37 A03CB37 36 Belladonna alkaloids 17 A03CB38 A03CB38 A03CB38 A03CB38
37 Belladonna alkaloids 18 A03DB04 A03DB04 A03DB04 A03DB04 http://bmjopen.bmj.com/ 38 Metoclopramide 56220098 Metoclopramide 1 A03FA01 A03FA01 A03FA01 A03FA01 39 A03FA51 A03FA51 A03FA51 A03FA51 40 Metoclopramide 2 41 Dimenhydrinate 56220003 Dimenhydrinate 1 A04AB02 A04AB02 A04AB02 A04AB02 42 Meclozine 3 A04AB04 A04AB04 A04AB04 A04AB04 43 Diphenhydramine 3 A04AB05 A04AB05 A04AB05 A04AB05 44 Dimenhydrinate 2 A04AB52 A04AB52 A04AB52 A04AB52 45 on September 26, 2021 by guest. Protected copyright. 46 Meclozine 4 A04AB54 A04AB54 A04AB54 A04AB54 47 Diphenhydramine 6 A04AB55 A04AB55 A04AB55 A04AB55 48 Doxylamine 4 A04AB56 A04AB56 A04AB56 A04AB56 49 Promethazine 3 A04AB58 A04AB58 A04AB58 A04AB58 50 Triflupromazine 2 A04AD06 A04AD06 A04AD06 A04AD06 51 52 Belladonna 12080002 Belladonna alkaloids 1 A06AB30 A06AB30 A06AB30 A06AB30 53 Diphenoxylate 56080005 Diphenoxylate A07DA01 A07DA01 A07DA01 A07DA01 54 Opium 28080881 Opium 1 A07DA02 A07DA02 A07DA02 A07DA02 55 Loperamide 56080009 Loperamide 1 A07DA03 A07DA03 A07DA03 A07DA03 56 Morphine 2 A07DA52 A07DA52 A07DA52 A07DA52 57 58 Loperamide 2 A07DA53 A07DA53 A07DA53 A07DA53 59 Cyproheptadine 2 A15AA01 A15AA01 A15AA01 A15AA01 60 Cyproheptadine 3 A15AA51 A15AA51 A15AA51 A15AA51 Disopyramide 24040024 Disopyramide C01BA03 C01BA03 C01BA03 C01BA03
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 65 of 90 BMJ Open
1 2 Reserpine 24080010 Reserpine 1 C02AA02 C02AA02 C02AA02 C02AA02
3 Reserpine 2 C02AA52 C02AA52 C02AA52 C02AA52 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Methyldopa 24080006 Methyldopa 1 C02AB01 C02AB01 C02AB01 C02AB01 5 6 Methyldopa 2 C02AB02 C02AB02 C02AB02 C02AB02 7 Clonidine 24080064 Clonidine 1 C02AC01 C02AC01 C02AC01 C02AC01 8 Guanfacine 24080063 Guanfacine C02AC02 C02AC02 C02AC02 C02AC02 9 Prazosin 12160404 Prazosin 1 C02CA01 C02CA01 C02CA01 C02CA01 10 Doxazosin 12160419 Doxazosin 1 C02CA04 C02CA04 C02CA04 C02CA04 11 12 Terazosin 2 C02CA08 C02CA08 C02CA08 C02CA08 13 Guanethidine 24080003 Guanethidine 1 C02CC02 C02CC02 C02CC02 C02CC02 14 Guanabenz 24080084 Guanoxabenz C02CC07 C02CC07 C02CC07 C02CC07 15 Reserpine 3 C02LA01 C02LA01 C02LA01 C02LA01 16 17 Reserpine 4 C02LA51 C02LA51 C02LA51 C02LA51 18 ForReserpine peer 5 reviewC02LA71 only C02LA71 C02LA71 C02LA71 19 Methyldopa 3 C02LB01 C02LB01 C02LB01 C02LB01 20 Clonidine 2 C02LC01 C02LC01 C02LC01 C02LC01 21 Clonidine 3 C02LC51 C02LC51 C02LC51 C02LC51 22 23 Prazosin 2 C02LE01 C02LE01 C02LE01 C02LE01 24 Guanethidine 2 C02LF01 C02LF01 C02LF01 C02LF01 25 Flavoxate 12080039 Flavoxate G04BD02 G04BD02 G04BD02 G04BD02 26 Oxybutynin 86000004 Oxybutynin G04BD04 G04BD04 G04BD04 G04BD04 27 Tolterodine 86000047 Tolterodine G04BD07 G04BD07 G04BD07 G04BD07 28 29 Papaverine 2 G04BE02 G04BE02 G04BE02 G04BE02 30 Tamulosin 12160411 Tamsulosin 1 G04CA02 G04CA02 G04CA02 G04CA02 31 Terazosin 12160401 Terazosin 1 G04CA03 G04CA03 G04CA03 G04CA03 32 Doxazosin 2 G04CA05 G04CA05 G04CA05 G04CA05 33 Tamsulosin 2 G04CA52 G04CA52 G04CA52 G04CA52 34 35 Tamsulosin 3 G04CA53 G04CA53 G04CA53 G04CA53 36 Carisoprodol 12200001 Carisoprodol 1 M03BA02 M03BA02 M03BA02 M03BA02
37 Methocarbamol 12200005 Methocarbamol 1 M03BA03 M03BA03 M03BA03 M03BA03 http://bmjopen.bmj.com/ 38 Carisoprodol 2 M03BA52 M03BA52 M03BA52 M03BA52 39 M03BA53 M03BA53 M03BA53 M03BA53 40 Methocarbamol 2 41 Meprobamate 4 M03BA57 M03BA57 M03BA57 M03BA57 42 Carisoprodol 3 M03BA72 M03BA72 M03BA72 M03BA72 43 Methocarbamol 3 M03BA73 M03BA73 M03BA73 M03BA73 44 Chlorzoxazone 12200091 Chlorzoxazone 1 M03BB03 M03BB03 M03BB03 M03BB03 45 on September 26, 2021 by guest. Protected copyright. 46 Chlorzoxazone 3 M03BB53 M03BB53 M03BB53 M03BB53 47 Chlorzoxazone 2 M03BB73 M03BB73 M03BB73 M03BB73 48 Orphenadrine 12080804 Orphenadrine 1 M03BC01 M03BC01 M03BC01 M03BC01 49 Orphenadrine 2 M03BC51 M03BC51 M03BC51 M03BC51 50 Cyclobenzaprine 12200009 Cyclobenzaprine M03BX08 M03BX08 M03BX08 M03BX08 51 52 Hexobarbital 28240405 Hexobarbital 1 N01AF02 N01AF02 N01AF02 N01AF02 53 Fentanyl 2 N01AH01 N01AH01 N01AH01 N01AH01 54 Fentanyl 3 N01AH51 N01AH51 N01AH51 N01AH51 55 Diphenhydramine 4 N01BX06 N01BX06 N01BX06 N01BX06 56 Morphine 28080819 Morphine 1 N02AA01 N02AA01 N02AA01 N02AA01 57 58 Opium 2 N02AA02 N02AA02 N02AA02 N02AA02 59 Oxycodone 28080883 Oxycodone 1 N02AA05 N02AA05 N02AA05 N02AA05 60 Codeine 10 (Dihydrocodeine N02AA08 N02AA08 N02AA08 N02AA08 1)
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 66 of 90
1 2 Morphine 3 N02AA51 N02AA51 N02AA51 N02AA51
3 Oxycodone 2 N02AA55 N02AA55 N02AA55 N02AA55 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Codeine 11 (Dihydrocodeine N02AA58 N02AA58 N02AA58 N02AA58 5 6 2) 7 Codeine 2 N02AA59 N02AA59 N02AA59 N02AA59 8 Codeine 3 N02AA64 N02AA64 N02AA64 N02AA64 9 Codeine 4 N02AA65 N02AA65 N02AA65 N02AA65 10 Codeine 5 N02AA66 N02AA66 N02AA66 N02AA66 11 12 Codeine 6 N02AA69 N02AA69 N02AA69 N02AA69 13 Codeine 7 N02AA79 N02AA79 N02AA79 N02AA79 14 Fentanyl 28080810 Fentanyl 1 N02AB03 N02AB03 N02AB03 N02AB03 15 Propoxyphene 28080840 Propoxyphene N02AC04 N02AC04 N02AC04 N02AC04 16 17 Methadone 3 N02AC06 N02AC06 N02AC06 N02AC06 18 For(Levomethadone) peer review only 19 Methadone 2 N02AC52 N02AC52 N02AC52 N02AC52 20 Pentazocine 28080892 Pentazocine N02AD01 N02AD01 N02AD01 N02AD01 21 Morphine 4 N02AG01 N02AG01 N02AG01 N02AG01 22 23 Tramadol 28080854 Tramadol 1 N02AX02 N02AX02 N02AX02 N02AX02 24 Tramadol 2 N02AX52 N02AX52 N02AX52 N02AX52 25 Clonidine 4 N02CX02 N02CX02 N02CX02 N02CX02 26 Codeine 8 N02CX58 N02CX58 N02CX58 N02CX58 27 Metoclopramide 3 N02CX59 N02CX59 N02CX59 N02CX59 28 29 Phenobarbital 28120405 Phenobarbital 1 N03AA02 N03AA02 N03AA02 N03AA02 30 Primidone 28120407 Primidone N03AA03 N03AA03 N03AA03 N03AA03 31 Phenytoin 28120805 Phenytoin 1 N03AB02 N03AB02 N03AB02 N03AB02 32 Phenytoin 3 (Fosphenytoin) N03AB05 N03AB05 N03AB05 N03AB05 33 Phenytoin 2 N03AB52 N03AB52 N03AB52 N03AB52 34 35 Clonazepam 28240212 Clonazepam N03AE01 N03AE01 N03AE01 N03AE01 36 Carbamazepine 28122007 Carbamazepine N03AF01 N03AF01 N03AF01 N03AF01
37 Oxcarbazepine 28122011 Oxcarbazepine N03AF02 N03AF02 N03AF02 N03AF02 http://bmjopen.bmj.com/ 38 Valproic acid 28122015 Valproic acid / Sodium N03AG01 N03AG01 N03AG01 N03AG01 39 40 Valproate 41 Tiagabine 28122034 Tiagabine N03AG06 N03AG06 N03AG06 N03AG06 42 Lamotrigine 28122024 Lamotrigine N03AX09 N03AX09 N03AX09 N03AX09 43 Gabapentin 28122020 Gabapentin N03AX12 N03AX12 N03AX12 N03AX12 44 Levetiracetam 28122040 Levetiracetam N03AX14 N03AX14 N03AX14 N03AX14 45 on September 26, 2021 by guest. Protected copyright. 46 Trihexyphenidyl 12080802 Trihexyphenidyl N04AA01 N04AA01 N04AA01 N04AA01 47 Orphenadrine 3 N04AB02 N04AB02 N04AB02 N04AB02 48 Benztropine 12080806 Benzatropine 1 N04AC01 N04AC01 N04AC01 N04AC01 49 Benzatropine 2 N04AC30 N04AC30 N04AC30 N04AC30 50 (Etybenzatropin) 51 52 Ropinerole 28280011 Ropinerole N04BC04 N04BC04 N04BC04 N04BC04 53 Pramipexole 28280013 Pramipexole N04BC05 N04BC05 N04BC05 N04BC05 54 Selegiline 28160520 Selegiline N04BD01 N04BD01 N04BD01 N04BD01 55 Chlorpromazine 56220089 Chlorpromazine N05AA01 N05AA01 N05AA01 N05AA01 56 Triflupromazine 28160996 Triflupromazine 1 N05AA05 N05AA05 N05AA05 N05AA05 57 58 Fluphenazine 28160906 Fluphenazine N05AB02 N05AB02 N05AB02 N05AB02 59 Perphenazine 28160909 Perphenazine N05AB03 N05AB03 N05AB03 N05AB03 60 Prochlorperazine 56220096 Prochlorperazine N05AB04 N05AB04 N05AB04 N05AB04 Trifluoperazine 28160913 Trifluoperazine N05AB06 N05AB06 N05AB06 N05AB06
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 67 of 90 BMJ Open
1 2 Thioridazine 28160912 Thioridazine N05AC02 N05AC02 N05AC02 N05AC02
3 Haloperidol 28161014 Haloperidol N05AD01 N05AD01 N05AD01 N05AD01 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Ziprasidone 28160844 Ziprasidone N05AE04 N05AE04 N05AE04 N05AE04 5 6 Chlorprothixine 28160804 Chlorprothixene N05AF03 N05AF03 N05AF03 N05AF03 7 Loxapine 28160858 Loxapine N05AH01 N05AH01 N05AH01 N05AH01 8 Olanzapine 28160836 Olanzapine N05AH03 N05AH03 N05AH03 N05AH03 9 Quetiapine 28160834 Quetiapine N05AH04 N05AH04 N05AH04 N05AH04 10 Risperidone 28160822 Risperidone N05AX08 N05AX08 N05AX08 N05AX08 11 12 Reserpine 6 N05AX15 N05AX15 N05AX15 N05AX15 13 Diazepam 28240205 Diazepam N05BA01 N05BA01 N05BA01 N05BA01 14 Chlordiazepoxide 28240202 Chlordiazepoxide N05BA02 N05BA02 N05BA02 N05BA02 15 16 17 Oxazepam 28240215 Oxazepam N05BA04 N05BA04 N05BA04 N05BA04 18 Clorazepate 28240228For Clorazepate peer reviewN05BA05 only N05BA05 N05BA05 N05BA05 19 Lorazepam 28240276 Lorazepam 1 N05BA06 N05BA06 N05BA06 N05BA06 20 Alprazolam 28240232 Alprazolam N05BA12 N05BA12 N05BA12 N05BA12 21 Lorazepam 2 N05BA56 N05BA56 N05BA56 N05BA56 22 23 Hydroxyzine 28160807 Hydroxyzine 1 N05BB01 N05BB01 N05BB01 N05BB01 24 Hydroxyzine 3 N05BB51 N05BB51 N05BB51 N05BB51 25 Meprobamate 28240820 Meprobamate 1 N05BC01 N05BC01 N05BC01 N05BC01 26 Meprobamate 2 N05BC51 N05BC51 N05BC51 N05BC51 27 Buspirone 28240837 Buspirone N05BE01 N05BE01 N05BE01 N05BE01 28 29 Hexobarbital 2 N05CA16 N05CA16 N05CA16 N05CA16 30 Phenobarbital 2 N05CA24 N05CA24 N05CA24 N05CA24 31 Dichloralphenazone Dichloralphenazone N05CC04 N05CC04 N05CC04 N05CC04 32 28240828 33 Flurazepam 28240206 Flurazepam N05CD01 N05CD01 N05CD01 N05CD01 34 35 Estazolam 28240216 Estazolam N05CD04 N05CD04 N05CD04 N05CD04 36 Triazolam 28240222 Triazolam N05CD05 N05CD05 N05CD05 N05CD05
37 Temazepam 28240231 Temazepam N05CD07 N05CD07 N05CD07 N05CD07 http://bmjopen.bmj.com/ 38 Zolpidem 28240834 Zolpidem N05CF02 N05CF02 N05CF02 N05CF02 39 40 Zaleplon 28240856 Zaleplon N05CF03 N05CF03 N05CF03 N05CF03 41 Diphenhydramine 5 N05CM20 N05CM20 N05CM20 N05CM20 42 Doxylamine 3 N05CM21 N05CM21 N05CM21 N05CM21 43 Promethazine 4 N05CM22 N05CM22 N05CM22 N05CM22 44 Meprobamate 3 N05CX01 N05CX01 N05CX01 N05CX01 45 on September 26, 2021 by guest. Protected copyright. 46 Diphenhydramine 2 N05CX07 N05CX07 N05CX07 N05CX07 47 Promethazine 5 N05CX13 N05CX13 N05CX13 N05CX13 48 Desipramine 28160689 Desipramine N06AA01 N06AA01 N06AA01 N06AA01 49 Imipramine 28160602 Imipramine 1 N06AA02 N06AA02 N06AA02 N06AA02 50 Imipramine 2 N06AA03 N06AA03 N06AA03 N06AA03 51 52 Clomipramine 28160688 Clomipramine N06AA04 N06AA04 N06AA04 N06AA04 53 Trimipramine 28160650 Trimipramine N06AA06 N06AA06 N06AA06 N06AA06 54 Amitriptyline 28160601 Amitriptyline 1 N06AA09 N06AA09 N06AA09 N06AA09 55 Nortryptyline 28160695 Nortriptyline 1 N06AA10 N06AA10 N06AA10 N06AA10 56 Doxepin 28160681 Doxepin N06AA12 N06AA12 N06AA12 N06AA12 57 58 Amitriptyline 3 N06AA25 N06AA25 N06AA25 N06AA25 59 Fluoxetine 28160701 Fluoxetine 1 N06AB03 N06AB03 N06AB03 N06AB03 60 Citalopram 28160705 Citalopram N06AB04 N06AB04 N06AB04 N06AB04 Paroxetine 28160702 Paroxetine N06AB05 N06AB05 N06AB05 N06AB05
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 68 of 90
1 2 Sertraline 28160703 Sertraline N06AB06 N06AB06 N06AB06 N06AB06
3 Ecitalopram 28160711 Escitalopram N06AB10 N06AB10 N06AB10 N06AB10 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Phenelzine 28160505 Phenelzine N06AF03 N06AF03 N06AF03 N06AF03 5 6 Tranylcypromine 28160601 Tranylcypromine 1 N06AF04 N06AF04 N06AF04 N06AF04 7 Trazodone 28160415 Trazodone N06AX05 N06AX05 N06AX05 N06AX05 8 Nefazodone 28160486 Nefazodone N06AX06 N06AX06 N06AX06 N06AX06 9 Mirtazepine 28160617 Mirtazapine N06AX11 N06AX11 N06AX11 N06AX11 10 Venlafaxine 28160458 Venlafaxine 1 N06AX16 N06AX16 N06AX16 N06AX16 11 12 Venlafaxine 2 N06AX23 N06AX23 N06AX23 N06AX23 13 (Desvenlafaxine) 14 Amitriptyline 2 N06CA01 N06CA01 N06CA01 N06CA01 15 Fluoxetine 2 N06CA03 n/a n/a n/a 16 17 Nortriptyline 2 N06CA06 N06CA06 N06CA06 N06CA06 18 ForTranylcypromine peer 2 reviewN06CA07 onlyN06CA03 N06CA03 N06CA03 19 Clonidine 5 N07BB06 N07BB06 N07BB06 N07BB06 20 Methadone 28080818 Methadone 1 N07BC02 N07BC02 N07BC02 N07BC02 21 Dextrometorphan 2 N07XX59 n/a n/a n/a 22 23 Hydrocodone 48000072 Hydrocodone R05DA03 R05DA03 R05DA03 R05DA03 24 Codeine 48000063 Codeine 1 R05DA04 R05DA04 R05DA04 R05DA04 25 Morphine 5 R05DA05 R05DA05 R05DA05 R05DA05 26 Dextromethorphan Dextrometorphan 1 R05DA09 R05DA09 R05DA09 R05DA09 27 48000069 28 29 Codeine 12 (Dihydrocodeine R05DA14 R05DA14 R05DA14 R05DA14 30 3) 31 Codeine 9 R05DA54 R05DA54 R05DA54 R05DA54 32 Dextrometorphan 3 R05DA59 R05DA59 R05DA59 R05DA59 33 Codeine 13 (Dihydrocodeine R05DA64 R05DA64 R05DA64 R05DA64 34 35 4) 36 Benzonatate 48000054 Benzonatate R05DB01 R05DB01 R05DB01 R05DB01
37 Diphenhydramine 4000006 Diphenhydramine 1 R06AA02 R06AA02 R06AA02 R06AA02 http://bmjopen.bmj.com/ 38 Clemastine 4000054 Clemastine 1 R06AA04 R06AA04 R06AA04 R06AA04 39 40 Doxylamine 4000068 Doxylamine 1 R06AA09 R06AA09 R06AA09 R06AA09 41 Trimethobenzamide Trimethobenzamide R06AA10 R06AA10 R06AA10 R06AA10 42 56220006 43 Diphenhydramine 7 R06AA52 R06AA52 R06AA52 R06AA52 44 Clemastine 2 R06AA54 R06AA54 R06AA54 R06AA54 45 on September 26, 2021 by guest. Protected copyright. 46 Doxylamine 2 R06AA59 R06AA59 n/a n/a 47 Brompheniramine 4000078 Brompheniramine 1 R06AB01 R06AB01 R06AB01 R06AB01 48 Dexchlorpheniramine Dexchlorpheniramine 1 R06AB02 R06AB02 R06AB02 R06AB02 49 4000084 50 Chlorpheniramine Chlorpheniramine R06AB04 R06AB04 R06AB04 R06AB04 51 52 (=Chlorphenamine) (=Chlorphenamine) 1 53 4000003 54 Pheniramine 4000092 Pheniramine R06AB05 R06AB05 R06AB05 R06AB05 55 Dexbrompheniramine Dexbrompheniramine 1 R06AB06 R06AB06 R06AB06 R06AB06 56 4000083 57 58 Brompheniramine 2 R06AB51 R06AB51 R06AB51 R06AB51 59 Dexchlorpheniramine 2 R06AB52 R06AB52 R06AB52 R06AB52 60 Chlorpheniramine (=Chlorphenamine) 2 R06AB54 R06AB54 R06AB54 R06AB54
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 69 of 90 BMJ Open
1 2 Dexbrompheniramine 2 R06AB56 R06AB56 R06AB56 R06AB56
3 Tripelennamine 4000013 Tripelennamine R06AC04 R06AC04 R06AC04 R06AC04 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 Promethazine 4000010 Promethazine 1 R06AD02 R06AD02 R06AD02 R06AD02 5 6 Promethazine 8 7 (Hydroxyethylpromethazin 8 1) R06AD05 R06AD05 R06AD05 R06AD05 9 Promethazine 7 R06AD10 R06AD10 R06AD10 R06AD10 10 (Dioxopromethazin) 11 12 Promethazine 2 R06AD52 R06AD52 R06AD52 R06AD52 13 Promethazine 9 R06AD55 R06AD55 R06AD55 R06AD55 14 (Hydroxyethylpromethazin 15 2) 16 17 Meclizine 56220005 Meclozine 1 R06AE05 R06AE05 R06AE05 R06AE05 18 Cetirizine 4000031For Cetirizin peer 1 reviewR06AE07 only R06AE07 R06AE07 R06AE07 19 Cetirizin 3 (Enantiomer R06AE09 R06AE09 R06AE09 R06AE09 20 Levocetirizin ) 21 Meclozine 2 R06AE55 R06AE55 R06AE55 R06AE55 22 23 Cetirizin 2 R06AE57 R06AE57 R06AE57 R06AE57 24 Cyproheptadine 4000012 Cyproheptadine 1 R06AX02 R06AX02 R06AX02 R06AX02 25 Triprolidine 4000099 Triprolidine 1 R06AX07 R06AX07 R06AX07 R06AX07 26 Azatadine 4000018 Azatadine R06AX09 R06AX09 R06AX09 R06AX09 27 Astemizole 4000022 Astemizole R06AX11 R06AX11 R06AX11 R06AX11 28 29 Loratadine 4000029 Loratadine 1 R06AX13 R06AX13 R06AX13 R06AX13 30 Loratadine 2 (Enantiomer R06AX27 R06AX27 R06AX27 R06AX27 31 Desloratadin ) 32 Hydroxyzine 2 R06AX32 R06AX32 R06AX32 R06AX32 33 Triprolidine 2 R06AX57 R06AX57 R06AX57 R06AX57 34 35 Promethazine 6 V03AB05 V03AB05 V03AB05 V03AB05 36
37 http://bmjopen.bmj.com/ 38 39 40 41 42 43 44
45 on September 26, 2021 by guest. Protected copyright. 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 70 of 90
1 2 ICD-10-Code (Source: ICD-10-GM ). 1 = yes, 0 = no
3 Schaefer et al. Diederich Charlson HRQoL- Excluded BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 (mod.) Index 5 2014 2013 2012 2011 diagnoses 6 B15 B15 B15 B15 0 0 0 1 0 7 B16 B16 B16 B16 0 0 0 1 0 8 B17 B17 B17 B17 0 0 0 1 0 9 10 B18 B18 B18 B18 0 0 0 1 0 11 B19 B19 B19 B19 0 0 0 1 0 12 B20 B20 B20 B20 0 0 1 1 0 13 14 B21 B21 B21 B21 0 0 1 1 0 15 B22 B22 B22 B22 0 0 1 1 0 16 B23 B23 B23 B23 0 0 1 1 0 17 B24 B24 B24 B24 0 0 1 1 0 18 C00 C00 C00 ForC00 peer1 review1 only1 0 0 19 20 C01 C01 C01 C01 1 1 1 0 0 21 C02 C02 C02 C02 1 1 1 0 0 22 C03 C03 C03 C03 1 1 1 0 0 23 C04 C04 C04 C04 1 1 1 0 0 24 C05 C05 C05 C05 1 1 1 0 0 25 26 C06 C06 C06 C06 1 1 1 0 0 27 C07 C07 C07 C07 1 1 1 0 0 28 C08 C08 C08 C08 1 1 1 0 0 29 C09 C09 C09 C09 1 1 1 0 0 30 C10 C10 C10 C10 1 31 1 1 0 0 32 C11 C11 C11 C11 1 1 1 0 0 33 C12 C12 C12 C12 1 1 1 0 0 34 C13 C13 C13 C13 1 1 1 0 0 35 C14 C14 C14 C14 1 1 1 0 0 36
37 C15 C15 C15 C15 1 1 1 0 0 http://bmjopen.bmj.com/ 38 C16 C16 C16 C16 1 1 1 0 0 39 C17 C17 C17 C17 1 1 1 0 0 40 C18 C18 C18 C18 1 1 1 0 0 41 C19 C19 C19 C19 1 1 1 0 0 42 43 C20 C20 C20 C20 1 1 1 0 0 44 C21 C21 C21 C21 1 1 1 0 0
45 C22 C22 C22 C22 1 1 1 0 0 on September 26, 2021 by guest. Protected copyright. 46 C23 C23 C23 C23 1 1 1 0 0 47 C24 C24 C24 C24 1 1 1 0 0 48 49 C25 C25 C25 C25 1 1 1 0 0 50 C26 C26 C26 C26 1 1 1 0 0 51 C30 C30 C30 C30 1 1 1 0 0 52 C31 C31 C31 C31 1 1 1 0 0 53 C32 C32 C32 C32 1 1 1 0 0 54 55 C33 C33 C33 C33 1 1 1 0 0 56 C34 C34 C34 C34 1 1 1 0 0 57 C37 C37 C37 C37 1 1 1 0 0 58 C38 C38 C38 C38 1 1 1 0 0 59 60 C39 C39 C39 C39 1 1 1 0 0 C40 C40 C40 C40 1 1 1 0 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 71 of 90 BMJ Open
1 2 C41 C41 C41 C41 1 1 1 0 0
3 C43 C43 C43 C43 1 1 1 0 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 C44 C44 C44 C44 1 1 1 0 0 5 6 C45 C45 C45 C45 1 1 1 0 0 7 C46 C46 C46 C46 1 1 1 0 0 8 C47 C47 C47 C47 1 1 1 0 0 9 C48 C48 C48 C48 1 1 1 0 0 10 11 C49 C49 C49 C49 1 1 1 0 0 12 C50 C50 C50 C50 1 1 1 0 0 13 C51 C51 C51 C51 1 1 1 0 0 14 C52 C52 C52 C52 1 1 1 0 0 15 C53 C53 C53 C53 1 1 1 0 0 16 17 C54 C54 C54 C54 1 1 1 0 0 18 C55 C55 C55 ForC55 peer1 review1 only1 0 0 19 C56 C56 C56 C56 1 1 1 0 0 20 C57 C57 C57 C57 1 1 1 0 0 21 C58 C58 C58 C58 1 1 1 0 0 22 23 C60 C60 C60 C60 1 1 1 0 0 24 C61 C61 C61 C61 1 1 1 0 0 25 C62 C62 C62 C62 1 1 1 0 0 26 C63 C63 C63 C63 1 1 1 0 0 27 C64 C64 C64 C64 1 1 1 0 0 28 29 C65 C65 C65 C65 1 1 1 0 0 30 C66 C66 C66 C66 1 1 1 0 0 31 C67 C67 C67 C67 1 1 1 0 0 32 C68 C68 C68 C68 1 1 1 0 0 33 34 C69 C69 C69 C69 1 1 1 0 0 35 C70 C70 C70 C70 1 1 1 0 0 36 C71 C71 C71 C71 1 1 1 0 0 37 C72 C72 C72 C72 1 1 1 0 0 http://bmjopen.bmj.com/ 38 C73 C73 C73 C73 1 1 1 0 0 39 40 C74 C74 C74 C74 1 1 1 0 0 41 C75 C75 C75 C75 1 1 1 0 0 42 C76 C76 C76 C76 1 1 1 0 0 43 C77 C77 C77 C77 1 1 1 0 0 44 C78 C78 C78 C78 1 1 1 0 0 45 on September 26, 2021 by guest. Protected copyright. 46 C79 C79 C79 C79 1 1 1 0 0 47 C80 C80 C80 C80 1 1 1 0 0 48 C81 C81 C81 C81 1 1 1 0 0 49 C82 C82 C82 C82 1 1 1 0 0 50 C83 C83 C83 C83 1 1 1 0 0 51 52 C84 C84 C84 C84 1 1 1 0 0 53 C85 C85 C85 C85 1 1 1 0 0 54 C86 C86 C86 C86 1 1 1 0 0 55 C88 C88 C88 C88 1 1 1 0 0 56 57 C90 C90 C90 C90 1 1 1 0 0 58 C91 C91 C91 C91 1 1 1 0 0 59 C92 C92 C92 C92 1 1 1 0 0 60 C93 C93 C93 C93 1 1 1 0 0 C94 C94 C94 C94 1 1 1 0 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 72 of 90
1 2 C95 C95 C95 C95 1 1 1 0 0
3 C96 C96 C96 C96 1 1 1 0 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 C97 C97 C97 C97 1 1 1 0 0 5 6 D00 D00 D00 D00 1 1 1 0 0 7 D01 D01 D01 D01 1 1 1 0 0 8 D02 D02 D02 D02 1 1 1 0 0 9 D03 D03 D03 D03 1 1 1 0 0 10 11 D04 D04 D04 D04 1 1 1 0 0 12 D05 D05 D05 D05 1 1 1 0 0 13 D06 D06 D06 D06 1 1 1 0 0 14 D07 D07 D07 D07 1 1 1 0 0 15 D09 D09 D09 D09 1 1 1 0 0 16 17 D37 D37 D37 D37 1 1 1 0 0 18 D38 D38 D38 ForD38 peer1 review1 only1 0 0 19 D39 D39 D39 D39 1 1 1 0 0 20 D40 D40 D40 D40 1 1 1 0 0 21 D41 D41 D41 D41 1 1 1 0 0 22 23 D42 D42 D42 D42 1 1 1 0 0 24 D43 D43 D43 D43 1 1 1 0 0 25 D44 D44 D44 D44 1 1 1 0 0 26 D45 D45 D45 D45 1 1 1 0 0 27 D46 D46 D46 D46 1 1 1 0 0 28 29 D47 D47 D47 D47 1 1 1 0 0 30 D48 D48 D48 D48 1 1 1 0 0 31 D50 D50 D50 D50 1 0 0 1 0 32 D51 D51 D51 D51 1 0 0 1 0 33 34 D52 D52 D52 D52 1 0 0 1 0 35 D53 D53 D53 D53 1 0 0 1 0 36 D55 D55 D55 D55 1 0 0 1 0 37 D56 D56 D56 D56 1 0 0 1 0 http://bmjopen.bmj.com/ 38 D57 D57 D57 D57 1 0 0 1 0 39 40 D58 D58 D58 D58 1 0 0 1 0 41 D59.0 D59.0 D59.0 D59.0 1 0 0 1 0 42 D59.1 D59.1 D59.1 D59.1 1 0 0 1 0 43 D59.2 D59.2 D59.2 D59.2 1 0 0 1 0 44 D59.4 D59.4 D59.4 D59.4 1 0 0 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 D59.5 D59.5 D59.5 D59.5 1 0 0 1 0 47 D59.6 D59.6 D59.6 D59.6 1 0 0 1 0 48 D59.8 D59.8 D59.8 D59.8 1 0 0 1 0 49 D59.9 D59.9 D59.9 D59.9 1 0 0 1 0 50 D60.0 D60.0 D60.0 D60.0 1 0 0 1 0 51 52 D60.8 D60.8 D60.8 D60.8 1 0 0 1 0 53 D60.9 D60.9 D60.9 D60.9 1 0 0 1 0 54 D61 D61 D61 D61 1 0 0 1 0 55 D63 D63 D63 D63 1 0 0 1 0 56 57 D64 D64 D64 D64 1 0 0 1 0 58 D66 D66 D66 D66 1 0 0 0 0 59 D67 D67 D67 D67 1 0 0 0 0 60 D68 D68 D68 D68 1 0 0 0 0 D69 D69 D69 D69 1 0 0 0 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 73 of 90 BMJ Open
1 2 E01 E01 E01 E01 1 0 0 1 0
3 E02 E02 E02 E02 1 0 0 1 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 E03 E03 E03 E03 1 0 0 1 0 5 6 E04 E04 E04 E04 1 0 0 1 0 7 E05 E05 E05 E05 1 0 0 1 0 8 E06.1 E06.1 E06.1 E06.1 1 0 0 1 0 9 E06.2 E06.2 E06.2 E06.2 1 0 0 1 0 10 11 E06.3 E06.3 E06.3 E06.3 1 0 0 1 0 12 E06.5 E06.5 E06.5 E06.5 1 0 0 1 0 13 E06.9 E06.9 E06.9 E06.9 1 0 0 1 0 14 E07 E07 E07 E07 1 0 0 1 0 15 E10.0 E10.0 E10.0 E10.0 1 1 1 1 0 16 17 E10.1 E10.1 E10.1 E10.1 1 1 1 1 0 18 E10.2 E10.2 E10.2 ForE10.2 peer1 review1 only1 1 0 19 E10.3 E10.3 E10.3 E10.3 1 1 1 1 0 20 E10.4 E10.4 E10.4 E10.4 1 1 1 1 0 21 E10.5 E10.5 E10.5 E10.5 1 1 1 1 0 22 23 E10.6 E10.6 E10.6 E10.6 1 1 1 1 0 24 E10.7 E10.7 E10.7 E10.7 1 1 1 1 0 25 E10.8 E10.8 E10.8 E10.8 1 1 1 1 0 26 E10.9 E10.9 E10.9 E10.9 1 1 1 1 0 27 E11.0 E11.0 E11.0 E11.0 1 1 1 1 0 28 29 E11.1 E11.1 E11.1 E11.1 1 1 1 1 0 30 E11.2 E11.2 E11.2 E11.2 1 1 1 1 0 31 E11.3 E11.3 E11.3 E11.3 1 1 1 1 0 32 E11.4 E11.4 E11.4 E11.4 1 1 1 1 0 33 34 E11.5 E11.5 E11.5 E11.5 1 1 1 1 0 35 E11.6 E11.6 E11.6 E11.6 1 1 1 1 0 36 E11.7 E11.7 E11.7 E11.7 1 1 1 1 0 37 E11.8 E11.8 E11.8 E11.8 1 1 1 1 0 http://bmjopen.bmj.com/ 38 E11.9 E11.9 E11.9 E11.9 1 1 1 1 0 39 40 E12.0 E12.0 E12.0 E12.0 1 1 1 1 0 41 E12.1 E12.1 E12.1 E12.1 1 1 1 1 0 42 E12.2 E12.2 E12.2 E12.2 1 1 1 1 0 43 E12.3 E12.3 E12.3 E12.3 1 1 1 1 0 44 E12.4 E12.4 E12.4 E12.4 1 1 1 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 E12.5 E12.5 E12.5 E12.5 1 1 1 1 0 47 E12.6 E12.6 E12.6 E12.6 1 1 1 1 0 48 E12.7 E12.7 E12.7 E12.7 1 1 1 1 0 49 E12.8 E12.8 E12.8 E12.8 1 1 1 1 0 50 E12.9 E12.9 E12.9 E12.9 1 1 1 1 0 51 52 E13.0 E13.0 E13.0 E13.0 1 1 1 1 0 53 E13.1 E13.1 E13.1 E13.1 1 1 1 1 0 54 E13.2 E13.2 E13.2 E13.2 1 1 1 1 0 55 E13.3 E13.3 E13.3 E13.3 1 1 1 1 0 56 57 E13.4 E13.4 E13.4 E13.4 1 1 1 1 0 58 E13.5 E13.5 E13.5 E13.5 1 1 1 1 0 59 E13.6 E13.6 E13.6 E13.6 1 1 1 1 0 60 E13.7 E13.7 E13.7 E13.7 1 1 1 1 0 E13.8 E13.8 E13.8 E13.8 1 1 1 1 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 74 of 90
1 2 E13.9 E13.9 E13.9 E13.9 1 1 1 1 0
3 E14.0 E14.0 E14.0 E14.0 1 1 1 1 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 E14.1 E14.1 E14.1 E14.1 1 1 1 1 0 5 6 E14.2 E14.2 E14.2 E14.2 1 1 1 1 0 7 E14.3 E14.3 E14.3 E14.3 1 1 1 1 0 8 E14.4 E14.4 E14.4 E14.4 1 1 1 1 0 9 E14.5 E14.5 E14.5 E14.5 1 1 1 1 0 10 11 E14.6 E14.6 E14.6 E14.6 1 1 1 1 0 12 E14.7 E14.7 E14.7 E14.7 1 1 1 1 0 13 E14.8 E14.8 E14.8 E14.8 1 1 1 1 0 14 E14.9 E14.9 E14.9 E14.9 1 1 1 1 0 15 E16.0 E16.0 E16.0 E16.0 1 1 0 1 0 16 17 E16.1 E16.1 E16.1 E16.1 0 1 0 1 0 18 E16.2 E16.2 E16.2 ForE16.2 peer0 review1 only0 1 0 19 E16.3 E16.3 E16.3 E16.3 0 1 0 1 0 20 E16.4 E16.4 E16.4 E16.4 0 1 0 1 0 21 E16.5 E16.5 E16.5 E16.5 0 1 0 1 0 22 23 E16.6 E16.6 E16.6 E16.6 0 1 0 1 0 24 E16.7 E16.7 E16.7 E16.7 0 1 0 1 0 25 E16.8 E16.8 E16.8 E16.8 0 1 0 1 0 26 E16.9 E16.9 E16.9 E16.9 0 1 0 1 0 27 E66 E66 E66 E66 1 0 0 0 0 28 29 E73 E73 E73 E73 1 0 0 0 0 30 E78 E78 E78 E78 1 0 0 0 0 31 E79 E79 E79 E79 1 0 0 0 0 32 F00 F00 F00 F00 0 0 0 0 1 33 34 F01 F01 F01 F01 0 0 0 0 1 35 F02 F02 F02 F02 0 0 0 0 1 36 F03 F03 F03 F03 0 0 0 0 1 37 F05.1 F05.1 F05.1 F05.1 0 0 0 0 1 http://bmjopen.bmj.com/ 38 F10 F10 F10 F10 1 0 0 0 0 39 40 F13 F13 F13 F13 1 0 0 0 0 41 F17 F17 F17 F17 1 0 0 0 0 42 F20 F20 F20 F20 0 0 0 1 0 43 F21 F21 F21 F21 0 0 0 1 0 44 F22 F22 F22 F22 0 0 0 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 F23 F23 F23 F23 0 0 0 1 0 47 F24 F24 F24 F24 0 0 0 1 0 48 F25 F25 F25 F25 0 0 0 1 0 49 F28 F28 F28 F28 0 0 0 1 0 50 F29 F29 F29 F29 0 0 0 1 0 51 52 F30 F30 F30 F30 0 0 0 1 0 53 F31 F31 F31 F31 0 0 0 1 0 54 F32 F32 F32 F32 1 1 0 1 0 55 F33 F33 F33 F33 1 1 0 1 0 56 57 F34 F34 F34 F34 0 0 0 1 0 58 F38 F38 F38 F38 0 0 0 1 0 59 F39 F39 F39 F39 0 0 0 1 0 60 F40 F40 F40 F40 1 0 0 1 0 F41 F41 F41 F41 1 0 0 1 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 75 of 90 BMJ Open
1 2 F45 F45 F45 F45 1 0 0 0 0
3 F51 F51 F51 F51 1 0 0 0 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 F52 F52 F52 F52 1 0 0 0 0 5 6 G20 G20 G20 G20 1 0 0 1 0 7 G21 G21 G21 G21 1 0 0 1 0 8 G22 G22 G22 G22 1 0 0 1 0 9 G30 G30 G30 G30 0 0 0 0 1 10 11 G31 G31 G31 G31 0 0 0 0 1 12 G35 G35 G35 G35 0 0 0 1 0 13 G40 G40 G40 G40 1 0 0 1 0 14 G41 G41 G41 G41 1 0 0 1 0 15 G43 G43 G43 G43 1 0 0 1 0 16 17 G44 G44 G44 G44 1 0 0 1 0 18 G45 G45 G45 ForG45 peer1 review1 only1 1 0 19 G47 G47 G47 G47 1 0 0 0 0 20 G50 G50 G50 G50 1 0 0 0 0 21 G51 G51 G51 G51 1 0 0 0 0 22 23 G52 G52 G52 G52 1 0 0 0 0 24 G53 G53 G53 G53 1 0 0 0 0 25 G54 G54 G54 G54 1 0 0 0 0 26 G55 G55 G55 G55 1 0 0 0 0 27 G56 G56 G56 G56 1 0 0 0 0 28 29 G57 G57 G57 G57 1 0 0 0 0 30 G58 G58 G58 G58 1 0 0 0 0 31 G59 G59 G59 G59 1 0 0 0 0 32 G60 G60 G60 G60 1 0 0 0 0 33 34 G61 G61 G61 G61 1 0 0 0 0 35 G62 G62 G62 G62 1 0 0 0 0 36 G63 G63 G63 G63 1 0 0 0 0 37 G64 G64 G64 G64 1 0 0 0 0 http://bmjopen.bmj.com/ 38 G80.0 G80.0 G80.0 G80.0 0 0 1 1 0 39 40 G80.1 G80.1 G80.1 G80.1 0 0 1 1 0 41 G80.2 G80.2 G80.2 G80.2 0 0 1 1 0 42 G81 G81 G81 G81 0 0 1 1 0 43 G82 G82 G82 G82 0 0 1 1 0 44 G83.0 G83.0 G83.0 G83.0 0 0 1 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 H01.1 H01.1 H01.1 H01.1 1 0 0 0 0 47 H17 H17 H17 H17 1 1 0 1 0 48 H18 H18 H18 H18 1 1 0 1 0 49 H25 H25 H25 H25 1 1 0 1 0 50 H26 H26 H26 H26 1 1 0 1 0 51 52 H27 H27 H27 H27 1 1 0 1 0 53 H28 H28 H28 H28 1 1 0 1 0 54 H31 H31 H31 H31 1 1 0 1 0 55 H33 H33 H33 H33 1 0 0 1 0 56 57 H34.1 H34.1 H34.1 H34.1 1 1 0 1 0 58 H34.2 H34.2 H34.2 H34.2 1 1 0 1 0 59 H34.8 H34.8 H34.8 H34.8 1 1 0 1 0 60 H34.9 H34.9 H34.9 H34.9 1 1 0 1 0 H35 H35 H35 H35 1 1 0 1 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 76 of 90
1 2 H36 H36 H36 H36 1 1 0 1 0
3 H40 H40 H40 H40 1 1 0 1 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 H43 H43 H43 H43 1 1 0 1 0 5 6 H47 H47 H47 H47 1 1 0 1 0 7 H54 H54 H54 H54 1 1 0 1 0 8 H81 H81 H81 H81 1 0 0 0 0 9 H82 H82 H82 H82 1 0 0 0 0 10 11 H90 H90 H90 H90 1 1 0 0 0 12 H91.0 H91.0 H91.0 H91.0 1 1 0 0 0 13 H91.1 H91.1 H91.1 H91.1 1 1 0 0 0 14 H91.3 H91.3 H91.3 H91.3 1 1 0 0 0 15 H91.8 H91.8 H91.8 H91.8 1 1 0 0 0 16 17 H91.9 H91.9 H91.9 H91.9 1 1 0 0 0 18 I10 I10 I10 ForI10 peer1 review1 only0 1 0 19 I11 I11 I11 I11 1 1 0 1 0 20 I12 I12 I12 I12 1 1 0 1 0 21 I13 I13 I13 I13 1 1 0 1 0 22 23 I15 I15 I15 I15 1 1 0 1 0 24 I20 I20 I20 I20 1 1 0 1 0 25 I21 I21 I21 I21 1 1 1 1 0 26 I22 I22 I22 I22 1 0 0 1 0 27 I23 I23 I23 I23 1 0 0 1 0 28 29 I24 I24 I24 I24 1 0 0 1 0 30 I25 I25 I25 I25 1 1 0 1 0 31 I26 I26 I26 I26 1 0 0 0 0 32 I34 I34 I34 I34 1 0 0 0 0 33 34 I35 I35 I35 I35 1 0 0 0 0 35 I36 I36 I36 I36 1 0 0 0 0 36 I37 I37 I37 I37 1 0 0 0 0 37 I44 I44 I44 I44 1 0 0 1 0 http://bmjopen.bmj.com/ 38 I45 I45 I45 I45 1 0 0 1 0 39 40 I46.0 I46.0 I46.0 I46.0 1 0 0 1 0 41 I46.9 I46.9 I46.9 I46.9 1 0 0 1 0 42 I47 I47 I47 I47 1 0 0 1 0 43 I48 I48 I48 I48 1 0 0 1 0 44 I49.1 I49.1 I49.1 I49.1 1 0 0 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 I49.2 I49.2 I49.2 I49.2 1 0 0 1 0 47 I49.3 I49.3 I49.3 I49.3 1 0 0 1 0 48 I49.4 I49.4 I49.4 I49.4 1 0 0 1 0 49 I49.5 I49.5 I49.5 I49.5 1 0 0 1 0 50 I49.8 I49.8 I49.8 I49.8 1 0 0 1 0 51 52 I49.9 I49.9 I49.9 I49.9 1 0 0 1 0 53 I50 I50 I50 I50 1 1 1 1 0 54 I60 I60 I60 I60 1 1 1 1 0 55 I61 I61 I61 I61 1 1 1 1 0 56 57 I62 I62 I62 I62 1 1 1 1 0 58 I63 I63 I63 I63 1 1 1 1 0 59 I64 I64 I64 I64 1 1 1 1 0 60 I65 I65 I65 I65 1 1 1 1 0 I66 I66 I66 I66 1 1 1 1 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 77 of 90 BMJ Open
1 2 I67.0 I67.0 I67.0 I67.0 0 1 0 0 0
3 I67.2 I67.2 I67.2 I67.2 1 1 1 1 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 I67.4 I67.4 I67.4 I67.4 1 1 0 1 0 5 6 I67.8 I67.8 I67.8 I67.8 0 1 0 0 0 7 I67.9 I67.9 I67.9 I67.9 0 1 0 0 0 8 I69 I69 I69 I69 1 1 1 1 0 9 I70 I70 I70 I70 1 1 1 1 0 10 11 I71 I71 I71 I71 1 1 1 1 0 12 I72 I72 I72 I72 1 1 1 1 0 13 I73.9 I73.9 I73.9 I73.9 1 1 1 1 0 14 I80 I80 I80 I80 1 0 0 0 0 15 I83 I83 I83 I83 1 0 0 0 0 16 17 I87.0 I87.0 I87.0 I87.0 1 0 0 0 0 18 I87.2 I87.2 I87.2 ForI87.2 peer1 review0 only0 0 0 19 I95 I95 I95 I95 1 0 0 0 0 20 J30 J30 J30 J30 1 0 0 0 0 21 J40 J40 J40 J40 1 1 1 0 0 22 23 J41 J41 J41 J41 1 1 1 1 0 24 J42 J42 J42 J42 1 1 1 1 0 25 J43 J43 J43 J43 1 1 1 1 0 26 J44 J44 J44 J44 1 1 1 1 0 27 J45 J45 J45 J45 1 1 1 1 0 28 29 J47 J47 J47 J47 1 1 1 1 0 30 K21 K21 K21 K21 1 0 0 1 0 31 K25.4 K25.4 K25.4 K25.4 1 0 1 1 0 32 K25.5 K25.5 K25.5 K25.5 1 0 1 1 0 33 34 K25.6 K25.6 K25.6 K25.6 1 0 1 1 0 35 K25.7 K25.7 K25.7 K25.7 1 0 1 1 0 36 K25.9 K25.9 K25.9 K25.9 1 0 1 1 0 37 K26.4 K26.4 K26.4 K26.4 1 0 1 1 0 http://bmjopen.bmj.com/ 38 K26.5 K26.5 K26.5 K26.5 1 0 1 1 0 39 40 K26.6 K26.6 K26.6 K26.6 1 0 1 1 0 41 K26.7 K26.7 K26.7 K26.7 1 0 1 1 0 42 K26.9 K26.9 K26.9 K26.9 1 0 1 1 0 43 K27.4 K27.4 K27.4 K27.4 1 0 1 1 0 44 K27.5 K27.5 K27.5 K27.5 1 0 1 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 K27.6 K27.6 K27.6 K27.6 1 0 1 1 0 47 K27.7 K27.7 K27.7 K27.7 1 0 1 1 0 48 K27.9 K27.9 K27.9 K27.9 1 0 1 1 0 49 K28.4 K28.4 K28.4 K28.4 1 0 1 1 0 50 K28.5 K28.5 K28.5 K28.5 1 0 1 1 0 51 52 K28.6 K28.6 K28.6 K28.6 1 0 1 1 0 53 K28.7 K28.7 K28.7 K28.7 1 0 1 1 0 54 K28.9 K28.9 K28.9 K28.9 1 0 1 1 0 55 K29.2 K29.2 K29.2 K29.2 1 0 1 1 0 56 57 K29.3 K29.3 K29.3 K29.3 1 0 1 1 0 58 K29.4 K29.4 K29.4 K29.4 1 0 1 1 0 59 K29.5 K29.5 K29.5 K29.5 1 0 1 1 0 60 K29.6 K29.6 K29.6 K29.6 1 0 1 1 0 K29.7 K29.7 K29.7 K29.7 1 0 1 1 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 78 of 90
1 2 K29.8 K29.8 K29.8 K29.8 1 0 1 1 0
3 K29.9 K29.9 K29.9 K29.9 1 0 1 1 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 K52.2 K52.2 K52.2 K52.2 1 0 0 0 0 5 6 K57 K57 K57 K57 1 0 0 0 0 7 K58 K58 K58 K58 1 0 0 0 0 8 K64 K64 I84 I84 1 0 0 0 0 9 K70.0 K70.0 K70.0 K70.0 1 0 1 1 0 10 11 K70.1 K70.1 K70.1 K70.1 1 0 1 1 0 12 K70.2 K70.2 K70.2 K70.2 1 0 0 1 0 13 K70.3 K70.3 K70.3 K70.3 1 0 0 1 0 14 K70.4 K70.4 K70.4 K70.4 1 0 1 1 0 15 K70.5 K70.5 K70.5 K70.5 1 0 0 1 0 16 17 K70.6 K70.6 K70.6 K70.6 1 0 0 1 0 18 K70.7 K70.7 K70.7 ForK70.7 peer1 review0 only0 1 0 19 K70.8 K70.8 K70.8 K70.8 1 0 0 1 0 20 K70.9 K70.9 K70.9 K70.9 1 0 1 1 0 21 K71.0 K71.0 K71.0 K71.0 0 0 1 0 0 22 23 K71.1 K71.1 K71.1 K71.1 0 0 1 0 0 24 K71.2 K71.2 K71.2 K71.2 0 0 1 0 0 25 K71.3 K71.3 K71.3 K71.3 1 0 0 1 0 26 K71.4 K71.4 K71.4 K71.4 1 0 0 1 0 27 K71.5 K71.5 K71.5 K71.5 1 0 0 1 0 28 29 K71.6 K71.6 K71.6 K71.6 0 0 1 0 0 30 K71.7 K71.7 K71.7 K71.7 1 0 0 1 0 31 K71.8 K71.8 K71.8 K71.8 0 0 1 0 0 32 K72.0 K72.0 K72.0 K72.0 0 0 1 0 0 33 34 K72.1 K72.1 K72.1 K72.1 1 0 1 1 0 35 K72.2 K72.2 K72.2 K72.2 0 0 1 0 0 36 K72.3 K72.3 K72.3 K72.3 0 0 1 0 0 37 K72.4 K72.4 K72.4 K72.4 0 0 1 0 0 http://bmjopen.bmj.com/ 38 K72.5 K72.5 K72.5 K72.5 0 0 1 0 0 39 40 K72.6 K72.6 K72.6 K72.6 0 0 1 0 0 41 K72.7 K72.7 K72.7 K72.7 1 0 1 1 0 42 K72.8 K72.8 K72.8 K72.8 0 0 1 0 0 43 K72.9 K72.9 K72.9 K72.9 1 0 1 1 0 44 K73 K73 K73 K73 1 0 1 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 K74 K74 K74 K74 1 0 1 1 0 47 K75.0 K75.0 K75.0 K75.0 0 0 1 0 0 48 K75.1 K75.1 K75.1 K75.1 0 0 1 0 0 49 K75.2 K75.2 K75.2 K75.2 0 0 1 0 0 50 K75.3 K75.3 K75.3 K75.3 0 0 1 0 0 51 52 K75.4 K75.4 K75.4 K75.4 0 0 1 0 0 53 K75.8 K75.8 K75.8 K75.8 0 0 1 0 0 54 K75.9 K75.9 K75.9 K75.9 0 0 1 0 0 55 K76.0 K76.0 K76.0 K76.0 0 0 1 1 0 56 57 K76.1 K76.1 K76.1 K76.1 0 0 1 1 0 58 K76.2 K76.2 K76.2 K76.2 1 0 0 1 0 59 K76.3 K76.3 K76.3 K76.3 1 0 0 1 0 60 K76.4 K76.4 K76.4 K76.4 1 0 0 1 0 K76.5 K76.5 K76.5 K76.5 1 0 0 1 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 79 of 90 BMJ Open
1 2 K76.6 K76.6 K76.6 K76.6 1 0 0 1 0
3 K76.7 K76.7 K76.7 K76.7 1 0 0 1 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 K76.8 K76.8 K76.8 K76.8 1 0 0 1 0 5 6 K76.9 K76.9 K76.9 K76.9 1 0 0 1 0 7 K77 K77 K77 K77 0 0 1 0 0 8 K80 K80 K80 K80 1 0 0 1 0 9 K81.1 K81.1 K81.1 K81.1 1 0 0 1 0 10 11 K90.0 K90.0 K90.0 K90.0 1 0 0 0 0 12 L23 L23 L23 L23 1 0 0 0 0 13 L27.2 L27.2 L27.2 L27.2 1 0 0 0 0 14 L40 L40 L40 L40 1 0 0 0 0 15 L56.4 L56.4 L56.4 L56.4 1 0 0 0 0 16 17 M05 M05 M05 M05 1 1 0 1 0 18 M06 M06 M06 ForM06 peer1 review1 only0 1 0 19 M07 M07 M07 M07 1 1 0 1 0 20 M10 M10 M10 M10 1 0 0 0 0 21 M15 M15 M15 M15 1 1 0 1 0 22 23 M16 M16 M16 M16 1 1 0 1 0 24 M17 M17 M17 M17 1 1 0 1 0 25 M18 M18 M18 M18 1 1 0 1 0 26 M19 M19 M19 M19 1 1 0 1 0 27 M30 M30 M30 M30 1 0 1 1 0 28 29 M31 M31 M31 M31 1 0 1 1 0 30 M32 M32 M32 M32 1 0 1 1 0 31 M33 M33 M33 M33 1 0 1 1 0 32 M34 M34 M34 M34 1 0 1 1 0 33 34 M35 M35 M35 M35 1 0 1 1 0 35 M36 M36 M36 M36 1 0 1 1 0 36 M40 M40 M40 M40 1 0 0 1 0 37 M41 M41 M41 M41 1 0 0 1 0 http://bmjopen.bmj.com/ 38 M42 M42 M42 M42 1 0 0 1 0 39 40 M43 M43 M43 M43 1 0 0 1 0 41 M45 M45 M45 M45 1 0 0 1 0 42 M47 M47 M47 M47 1 0 0 1 0 43 M48.0 M48.0 M48.0 M48.0 1 0 0 1 0 44 M48.1 M48.1 M48.1 M48.1 1 0 0 1 0 45 on September 26, 2021 by guest. Protected copyright. 46 M48.2 M48.2 M48.2 M48.2 1 0 0 1 0 47 M48.5 M48.5 M48.5 M48.5 1 0 0 1 0 48 M48.8 M48.8 M48.8 M48.8 1 0 0 1 0 49 M48.9 M48.9 M48.9 M48.9 1 0 0 1 0 50 M50 M50 M50 M50 1 0 0 1 0 51 52 M51 M51 M51 M51 1 0 0 1 0 53 M53 M53 M53 M53 1 0 0 1 0 54 M54 M54 M54 M54 1 0 0 1 0 55 M79.0 M79.0 M79.0 M79.0 1 1 0 0 0 56 57 M80 M80 M80 M80 1 1 0 0 0 58 M81 M81 M81 M81 1 1 0 0 0 59 M82 M82 M82 M82 1 1 0 0 0 60 N00 N00 N00 N00 1 1 1 0 0 N01 N01 N01 N01 1 1 1 0 0
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 80 of 90
1 2 N02 N02 N02 N02 1 1 1 0 0
3 N03 N03 N03 N03 1 1 1 0 0 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 N04 N04 N04 N04 1 1 1 0 0 5 6 N05 N05 N05 N05 1 1 1 0 0 7 N06 N06 N06 N06 1 1 1 0 0 8 N07 N07 N07 N07 1 1 1 0 0 9 N08 N08 N08 N08 1 1 1 0 0 10 11 N18 N18 N18 N18 1 1 1 0 0 12 N19 N19 N19 N19 1 1 1 0 0 13 N20 N20 N20 N20 1 0 0 0 0 14 N25 N25 N25 N25 1 1 1 0 0 15 N26 N26 N26 N26 1 1 1 0 0 16 17 N27 N27 N27 N27 1 1 1 0 0 18 N28 N28 N28 ForN28 peer1 review1 only1 0 0 19 N29 N29 N29 N29 1 1 1 0 0 20 N39.3 N39.3 N39.3 N39.3 1 0 0 0 0 21 N39.4 N39.4 N39.4 N39.4 1 0 0 0 0 22 23 N40 N40 N40 N40 1 0 0 0 0 24 N48.4 N48.4 N48.4 N48.4 1 0 0 0 0 25 N81 N81 N81 N81 1 0 0 0 0 26 N84 N84 N84 N84 1 0 0 0 0 27 N85 N85 N85 N85 1 0 0 0 0 28 29 N86 N86 N86 N86 1 0 0 0 0 30 N87 N87 N87 N87 1 0 0 0 0 31 N88 N88 N88 N88 1 0 0 0 0 32 N89 N89 N89 N89 1 0 0 0 0 33 34 N90 N90 N90 N90 1 0 0 0 0 35 N93 N93 N93 N93 1 0 0 0 0 36 N95 N95 N95 N95 1 0 0 0 0 37 R00 R00 R00 R00 1 0 0 1 0 http://bmjopen.bmj.com/ 38 R03.1 R03.1 R03.1 R03.1 1 0 0 0 0 39 40 R32 R32 R32 R32 1 0 0 0 0 41 R42 R42 R42 R42 1 0 0 0 0 42 R51 R51 R51 R51 1 0 0 1 0 43 R54 R54 R54 R54 0 0 0 0 1 44 T78.1 T78.1 T78.1 T78.1 1 0 0 0 0 45 on September 26, 2021 by guest. Protected copyright. 46 T78.4 T78.4 T78.4 T78.4 1 0 0 0 0 47 T88.7 T88.7 T88.7 T88.7 1 0 0 0 0 48 U85 U85 U85 U85 0 0 1 0 0 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Page 81 of 90 BMJ Open
1 2 3 4 5 6 7 Persons insured by 8 CRT participants TK in 2012 9 (n=505) 10 For peer(n=8,109,215) review only 11 12 http://bmjopen.bmj.com/ 13 14 Excluded (n=7,516,759) 15 - Age <60 years (n=6,572,610) 16 17 - Not continuously insured by TK between 18Excluded (n=3) 01/12 and 12/14 and/or no primary care 19 20- Age <60 years (n=3) contact in 2012 (n=158,688) on September 26, 2021 by guest. Protected copyright. 21 - <3 chronic diseases (n=268,319) 22 - <5 chronic medications (n=454,758) 23 24 - Dementia diagnosis (n =27,115) 25 - Under legal guardianship (n=35,269) 26 27 28 29 No. of participants No. of persons 30 included in analyses insured by TK and 31 32 (n=502) included in analyses 33 (n=592,456) 34 35 36 37 38 39 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
BMJ Open Page 82 of 90
Additional file 5: Univariate Analyses: Association between predictor variables and combined outcome in claims data 1 2 combined outcome reached combined outcome reached combined outcome reached 3 4 after 6 months after 9 months after 24 months 5
6 Yes No Total Yes No Total Yes No Total 7 8 (n = (n = (n = (n = (n = (n = (n = (n = (n = 9 10 192357) 400099) 592456) 244190) 348266) 592456) 387951) 204505) 592456) 11 12 Mean MeanFor Mean peer p-value Mean review Mean Mean only p-value Mean Mean Mean p-value 13 14 (SD) (SD) (SD) (SD) (SD) (SD) (SD) (SD) (SD) 15 16
Age 72.2 (7.3) 70.8 (6.9) 71.3 (7.1) <0.001 72.1 (7.3) 70.7 (6.9) 71.3 (7.1) <0.001 http://bmjopen.bmj.com/ 71.8 (7.2) 70.2 (6.7) 71.3 (7.1) <0.001 17 18 Disease count 10.5 (4.0) 9.3 (3.6) 9.7 (3.8) <0.001 10.4 (3.9) 9.2 (3.5) 9.7 (3.8) <0.001 10.2 (3.9) 8.8 (3.4) 9.7 (3.8) <0.001 19 20 CCI 3.4 (2.8) 2.8 (2.4) 3.0 (2.5) <0.001 3.4 (2.7) 2.8 (2.4) 3.0 (2.5) <0.001 3.2 (2.7) 2.6 (2.3) 3.0 (2.5) <0.001 21 22 No. of specific chronic 4.7 (2.1) 4.1 (1.9) 4.3 (2.0) <0.001 4.7 (2.1) 4.1 (1.9) 4.3 (2.0) <0.001 4.6 (2.0) 3.9 (1.8) 4.3 (2.0) <0.001 23 24 diseases (Diederichs) on September 26, 2021 by guest. Protected copyright. 25 26 No. of PIM (EU-PIM) 1.3 (1.2) 1.1 (1.1) 1.1 (1.2) <0.001 1.3 (1.2) 1.1 (1.1) 1.1 (1.2) <0.001 1.2 (1.2) 1.0 (1.0) 1.1 (1.1) <0.001 27 28 ADS 1.2 (1.6) 0.9 (1.4) 1.0 (1.5) <0.001 1.2 (1.6) 0.9 (1.3) 1.0 (1.5) <0.001 1.1 (1.5) 0.8 (1.3) 1.0 (1.5) <0.001 29 30 DBI 0.9 (1.2) 0.7 (1.0) 0.8 (1.0) <0.001 0.9 (1.1) 0.7 (0.9) 0.8 (1.0) <0.001 0.8 (1.1) 0.6 (0.9) 0.8 (1.0) <0.001 31 32 Number of involved 11.0 (5.6) 9.4 (5.0) 10.0 (5.3) <0.001 10.9 (5.6) 9.3 (4.9) 10.0 (5.3) <0.001 10.6 (5.4) 8.8 (4.7) 10.0 (5.3) <0.001 33 34 physicians 35 36 Abbreviations: ACh burden – Anticholinergic drug burden, ADS – Anticholinergic Drug Scale, CCI - Charlson Comorbidity Index, DBI – Drug 37 38 Burden Index, PIM – Potentially Inappropriate Medication. 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
Page 83 of 90 BMJ Open
combined outcome reached combined outcome reached combined outcome reached 1 2 after 6 months after 9 months after 24 months 3 4 Yes No Total Yes No Total Yes No Total 5 6 (n = (n = (n = (n = (n = (n = (n = (n = (n = 7 8 192357) 400099) 592456) 244190) 348266) 592456) 387951) 204505) 592456) 9 10 n (%) n (%) n (%) p-value n (%) n (%) n (%) p-value n (%) n (%) n (%) p-value 11 12 Sex Female 90924 182079For 273003 peer 114859 review 158144 273003 only 180222 92781 273003 13 14 (47.3%) (45.5%) (46.1%) (47.0%) (45.4%) (46.1%) (46.5%) (45.4%) (46.1%) 15 16 Male 101433 218020 319453 <0.001 129331 190122 319453 <0.001 207729 111724 319453 <0.001 http://bmjopen.bmj.com/ 17 18 (52.7%) (54.5%) (53.9%) (53.0%) (54.6%) (53.9%) (53.5%) (54.6%) (53.9%) 19 20 Previous No 107627 289845 397472 141326 256146 397472 240676 156796 397472 21 22 hospitalis (56%) (72.4%) (67.1%) (57.9%) (73.5%) (67.1%) (62%) (76.7%) (67.1%) 23 ation 24 yes 84730 110254 194984 <0.001 102864 92120 194984 <0.001 147275 47709 194984 <0.001 on September 26, 2021 by guest. Protected copyright. 25 (44%) (27.6%) (32.9%) (42.1%) (26.5%) (32.9%) (38%) (23.3%) (32.9%) 26 27 Previous No 103734 325335 429069 142477 286592 429069 255691 173378 429069 28 29 falls (53.9%) (81.3%) (72.4%) (58.3%) (82.3%) (72.4%) (65.9%) (84.8%) (72.4%) 30 31 yes 88623 74764 163387 <0.001 101713 61674 163387 <0.001 132260 31127 163387 <0.001 32 33 (46.1%) (18.7%) (27.6%) (41.7%) (17.7%) (27.6%) (34.1%) (15.2%) (27.6%) 34 35 36 37 38 39 40 41 42 43 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 BMJ Open Page 84 of 90
1 2
3 Additional file 6: Univariate analyses BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 Table A6.1: Association between predictor variables and EQ5D-3L (Pearson 6 7 8 Correlation) 9 10 Predictor variable EQ5D-3L (T1) 11 12 13 Core predictors 14 15 . Age -0.09 16 17 . Disease count -0.24 18 For peer review only 19 . Charlson Comorbidity Index (CCI) -0.07 20 21 22 . No. of specific chronic diseases (Diederichs) -0.19 23 24 . No. of drugs -0.26 25 26 . No. of PIM (EU-PIM) -0.18 27 28 29 . ACh burden (ADS) -0.16 30 31 . Mod. Drug Burden Index -0.20 32 33 . No. of involved physicians -0.06 34 35 36 Additional predictors
37 http://bmjopen.bmj.com/ 38 . No. of persons living in household 0.04 39 40 . CASMIN 0.09 41 42 . Alcohol intake (AUDIT C) 0.13 43 44
45 . Body Mass Index -0.15 on September 26, 2021 by guest. Protected copyright. 46 47 . MAI -0.24 48 49 . CIRS sum score -0.27 50 51 52 . CIRS, no. of organ systems -0.22 53 54 . HRQoL-CI, mental -0.24 55 56 . HRQoL-CI, physical -0.20 57 58 59 . Depressive Symptoms (GDS) -0.52 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 85 of 90 BMJ Open
1 2
3 . EQ5D-3L (Baseline) 0.68 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from 4 5 6 7 8 Abbreviations: ACh burden – Anticholinergic drug burden, ADS – Anticholinergic 9 10 Drug Scale, AUDIT - Alcohol Use Disorders Identification Test (WHO), CASMIN - 11 12 Comparative Analysis of Social Mobility in Industrial Nations, CCI - Charlson 13 14 15 Comorbidity Index, CIRS – Cumulative Illness Rating Scale, GDS – Geriatric 16 17 Depression Scale, HRQoL – Health-Related Quality of Life, HRQoL-CI – Health- 18 For peer review only 19 Related Quality of Life Comorbidity Index, MAI – Medication Appropriateness Index, 20 21 22 PIM – Potentially Inappropriate Medication. 23 24 25 26 27 28 Table A6.2: Association between predictor variables and EQ5D-3L (T-Test) 29 30 31 Predictor variable Mean (SD) Mean (SD) p-value 32 33 34 Core predictors 35 36 . Sex (female / male) 67.7 (25.77) 78.5 (22.97) <0.001
37 http://bmjopen.bmj.com/ 38 . Previous hospitalisation (yes / no) 71.2 (26.76) 73.3 (24.67) 0.531 39 40 41 . Previous falls (yes/no) 65.9 (25.28) 74.3 (24.76) 0.011 42 43 Additional predictors 44 45 . Smoker (yes / no) 74.1 (25.81) 73.1 (25.07) 0.807 on September 26, 2021 by guest. Protected copyright. 46 47 Abbreviations: SD – Standard Deviation 48 49 50 51 52 53 54 55 56 57 58 59 60
For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 86 of 90 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 3 4 Reporting checklist for prediction model 5 6 7 development and validation study. 8 9 10 11 Based on the TRIPOD guidelines. 12 13 14 15 Instructions to authors 16 For peer review only 17 Complete this checklist by entering the page numbers from your manuscript where readers will find 18 19 20 each of the items listed below. 21 22 23 Your article may not currently address all the items on the checklist. Please modify your text to 24 25 include the missing information. If you are certain that an item does not apply, please write "n/a" and 26 27 provide a short explanation. 28 29 30 Upload your completed checklist as an extra file when you submit to a journal. 31 32 http://bmjopen.bmj.com/ 33 In your methods section, say that you used the TRIPODreporting guidelines, and cite them as: 34 35 36 37 Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent reporting of a multivariable prediction 38 39 model for individual prognosis or diagnosis (TRIPOD): The TRIPOD statement. 40 41 on September 26, 2021 by guest. Protected copyright. 42 Page 43 44 Reporting Item Number 45 46 47 Title 48 49 50 #1 Identify the study as developing and / or validating a 1 51 52 53 multivariable prediction model, the target population, and 54 55 the outcome to be predicted. 56 57 58 Abstract 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 87 of 90 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 #2 Provide a summary of objectives, study design, setting, 2 3 4 participants, sample size, predictors, outcome, statistical 5 6 analysis, results, and conclusions. 7 8 9 Introduction 10 11 12 #3a Explain the medical context (including whether diagnostic 4 13 14 15 or prognostic) and rationale for developing or validating 16 For peer review only 17 the multivariable prediction model, including references 18 19 to existing models. 20 21 22 #3b Specify the objectives, including whether the study 5 23 24 describes the development or validation of the model or 25 26 27 both. 28 29 30 Methods 31 32 33 Source of data #4a Describe the study design or source of data (e.g., 5/6 http://bmjopen.bmj.com/ 34 35 randomized trial, cohort, or registry data), separately for 36 37 the development and validation data sets, if applicable. 38 39 40 41 Source of data #4b Specify the key study dates, including start of accrual; 6 on September 26, 2021 by guest. Protected copyright. 42 43 end of accrual; and, if applicable, end of follow-up. 44 45 46 Participants #5a Specify key elements of the study setting (e.g., primary 6 47 48 care, secondary care, general population) including 49 50 number and location of centres. 51 52 53 54 Participants #5b Describe eligibility criteria for participants. 6 55 56 57 Participants #5c Give details of treatments received, if relevant n/a 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 88 of 90 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Outcome #6a Clearly define the outcome that is predicted by the 7 3 4 prediction model, including how and when assessed. 5 6 7 Outcome #6b Report any actions to blind assessment of the outcome n/a 8 9 to be predicted. 10 11 12 Predictors #7a Clearly define all predictors used in developing or 8 13 14 15 validating the multivariable prediction model, including 16 For peer review only 17 how and when they were measured 18 19 20 Predictors #7b Report any actions to blind assessment of predictors for n/a 21 22 the outcome and other predictors. 23 24 25 Sample size #8 Explain how the study size was arrived at. 6 26 27 28 Missing data #9 Describe how missing data were handled (e.g., 11 29 30 31 complete-case analysis, single imputation, multiple 32 33 imputation) with details of any imputation method. http://bmjopen.bmj.com/ 34 35 36 Statistical #10a If you are developing a prediction model describe how 11 37 38 analysis methods predictors were handled in the analyses. 39 40 41 Statistical #10b If you are developing a prediction model, specify type of 11-13 on September 26, 2021 by guest. Protected copyright. 42 43 44 analysis methods model, all model-building procedures (including any 45 46 predictor selection), and method for internal validation. 47 48 49 Statistical #10c If you are validating a prediction model, describe how the 11-13 50 51 analysis methods predictions were calculated. 52 53 54 Statistical #10d Specify all measures used to assess model performance 11-13 55 56 57 analysis methods and, if relevant, to compare multiple models. 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 89 of 90 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Statistical #10e If you are validating a prediction model, describe any 11-13 3 4 analysis methods model updating (e.g., recalibration) arising from the 5 6 validation, if done 7 8 9 Risk groups #11 Provide details on how risk groups were created, if done. n/a 10 11 12 Development vs. #12 For validation, identify any differences from the 11-13 13 14 15 validation development data in setting, eligibility criteria, outcome, 16 For peer review only 17 and predictors. 18 19 20 Results 21 22 23 Participants #13a Describe the flow of participants through the study, Figure 2 24 25 including the number of participants with and without the 26 27 28 outcome and, if applicable, a summary of the follow-up 29 30 time. A diagram may be helpful. 31 32 33 Participants #13b Describe the characteristics of the participants (basic Table 1, http://bmjopen.bmj.com/ 34 35 demographics, clinical features, available predictors), Additional 36 37 including the number of participants with missing data for file 2 38 39 40 predictors and outcome. 41 on September 26, 2021 by guest. Protected copyright. 42 43 Participants #13c For validation, show a comparison with the development Additional 44 45 data of the distribution of important variables file 2 46 47 (demographics, predictors and outcome). 48 49 50 Model #14a If developing a model, specify the number of participants Additional 51 52 53 development and outcome events in each analysis. files 3-5 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml BMJ Open Page 90 of 90 BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Model #14b If developing a model, report the unadjusted association, Additional 3 4 development if calculated between each candidate predictor and files 3-5 5 6 outcome. 7 8 9 Model #15a If developing a model, present the full prediction model to Additional 10 11 specification allow predictions for individuals (i.e., all regression files 3-5 12 13 14 coefficients, and model intercept or baseline survival at a 15 16 Forgiven peer time point). review only 17 18 19 Model #15b If developing a prediction model, explain how to the use 17 20 21 specification it. 22 23 24 Model #16 Report performance measures (with CIs) for the Additional 25 26 27 performance prediction model. files 3-5 28 29 30 Model-updating #17 If validating a model, report the results from any model 17 31 32 updating, if done (i.e., model specification, model 33 http://bmjopen.bmj.com/ 34 performance). 35 36 37 Discussion 38 39 40 41 Limitations #18 Discuss any limitations of the study (such as 23 on September 26, 2021 by guest. Protected copyright. 42 43 nonrepresentative sample, few events per predictor, 44 45 missing data). 46 47 48 Interpretation #19a For validation, discuss the results with reference to 22 49 50 performance in the development data, and any other 51 52 53 validation data 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml Page 91 of 90 BMJ Open BMJ Open: first published as 10.1136/bmjopen-2020-039747 on 22 October 2020. Downloaded from
1 2 Interpretation #19b Give an overall interpretation of the results, considering 23 3 4 objectives, limitations, results from similar studies, and 5 6 other relevant evidence. 7 8 9 Implications #20 Discuss the potential clinical use of the model and 23 10 11 implications for future research 12 13 14 15 Other information 16 For peer review only 17 18 Supplementary #21 Provide information about the availability of 25 19 20 information supplementary resources, such as study protocol, Web 21 22 calculator, and data sets. 23 24 25 Funding #22 Give the source of funding and the role of the funders for 26 26 27 28 the present study. 29 30 31 None The TRIPOD checklist is distributed under the terms of the Creative Commons Attribution 32 33 License CC-BY. This checklist can be completed online using https://www.goodreports.org/, a tool http://bmjopen.bmj.com/ 34 35 made by the EQUATOR Network in collaboration with Penelope.ai 36 37 38 39 40 41 on September 26, 2021 by guest. Protected copyright. 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 For peer review only - http://bmjopen.bmj.com/site/about/guidelines.xhtml