BMJ

Confidential: For Review Only Prevalence of potentially inappropriate prescribing in older people in primary care and its association with hospital admission: a lognitudinal study

Journal: BMJ

Manuscript ID BMJ.2018.046067

Article Type: Research

BMJ Journal: BMJ

Date Submitted by the 18-Jul-2018 Author:

Complete List of Authors: Peréz, Teresa; Universidad Complutense de Madrid Facultad de Estudios Estadisticos, Department of Statistics and Operational Research III; Royal College of Surgeons in Ireland, HRB Centre for Primary Care Research, Department of General Practice Moriarty, Frank; Royal College of Surgeons in Ireland, HRB Centre for Primary Care Research, Department of General Practice Wallace, Emma; Royal College of Surgeons in Ireland, HRB Centre for Primary Care Research, Department of General Practice McDowell, Ronald; Royal College of Surgeons in Ireland, HRB Centre for Primary Care Research, Department of General Practice Redmond, Patrick ; Royal College of Surgeons in Ireland, HRB Centre for Primary Care Research, Department of General Practice; University of Cambridge Department of Public Health and Primary Care Fahey, Tom ; Royal College of Surgeons in Ireland, HRB Centre for Primary Care Research, Department of General Practice

Keywords: older patients, prescribing, STOPP, primary care, hospitalisation

https://mc.manuscriptcentral.com/bmj Page 1 of 42 BMJ

1 2 3 4 Appendix 1 STOPP version 2 criteria that could not be applied 5 6 Section A: Indication of medication 7 8 1. Any drug prescribed without an evidence-based clinical indication. 9 10 SectionConfidential: B: Cardiovascular System For Review Only 11 1. Digoxin for heart failure with normal systolic ventricular function (no clear evidence of benefit). 12 13 2. Verapamil or diltiazem with NYHA Class III or IV heart failure (may worsen heart failure). 14 4. Beta blocker with bradycardia (< 50/min), type II heart block or complete heart block (risk of 15 complete heart block, asystole). 16 17 7. Loop diuretic for dependent ankle oedema without clinical, biochemical evidence or radiological 18 evidence of heart failure, liver failure, nephrotic syndrome or renal failure (leg elevation and /or 19 compression hosiery usually more appropriate). 20 9. Loop diuretic for treatment of hypertension with concurrent urinary incontinence (may 21 exacerbate incontinence). 22 23 10. Centrally-acting antihypertensives (e.g. methyldopa, clonidine, moxonidine, rilmenidine, 24 guanfacine), unless clear intolerance of, or lack of efficacy with, other classes of antihypertensives 25 (centrally-active antihypertensives are generally less well tolerated by older people than younger 26 people). 27 28 11. ACE inhibitors or Angiotensin Receptor Blockers in patients with hyperkalaemia. 29 12. Aldosterone antagonists (e.g. spironolactone, eplerenone) with concurrent potassium conserving 30 drugs (e.g. ACEI’s, ARB’s, amiloride, triamterene) without monitoring of serum potassium (risk of 31 dangerous hyperkalaemia i.e. > 6.0 mmol/l – serum K should be monitored regularly, i.e. at least 32 every 6 months). 33 34 Section C: Antiplatelet/Anticoagulant Drugs 35 36 3. Aspirin, clopidogrel, dipyridamole, vitamin K antagonists, direct thrombin inhibitors or factor Xa 37 inhibitors with concurrent significant bleeding risk, i.e. uncontrolled severe hypertension, bleeding 38 diathesis, recent non-trivial spontaneous bleeding) (high risk of bleeding). 39 4. Aspirin plus clopidogrel as secondary stroke prevention, unless the patient has a coronary stent(s) 40 inserted in the previous 12 months or concurrent acute coronary syndrome or has a high grade 41 42 symptomatic carotid arterial stenosis (no evidence of added benefit over clopidogrel monotherapy). 43 5. Aspirin in combination with vitamin K antagonist, direct thrombin inhibitor or factor Xa inhibitors 44 in patients with chronic atrial fibrillation (no added benefit from aspirin) 45 46 8. Vitamin K antagonist, direct thrombin inhibitor or factor Xa inhibitors for first deep venous 47 thrombosis without continuing provoking risk factors (e.g. thrombophilia) for > 6 months, (no proven 48 added benefit). 49 9. Vitamin K antagonist, direct thrombin inhibitor or factor Xa inhibitors for first pulmonary embolus 50 without continuing provoking risk factors (e.g. thrombophilia) for > 12 months (no proven added 51 benefit). 52 53 Section D: Central Nervous System and Psychotropic Drugs 54 55 4. Selective serotonin re-uptake inhibitors (SSRI’s) with current or recent significant hyponatraemia 56 i.e. serum Na+ < 130 mmol/l (risk of exacerbating or precipitating hyponatraemia). 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 2 of 42

1 2 3 9. Neuroleptic antipsychotic in patients with behavioural and psychological symptoms of dementia 4 (BPSD) unless symptoms are severe and other non-pharmacological treatments have failed 5 (increased risk of stroke). 6 10. Neuroleptics as hypnotics, unless sleep disorder is due to psychosis or dementia (risk of 7 confusion, hypotension, extra-pyramidal side effects, falls). 8 9 13. Levodopa or dopamine agonists for benign essential tremor (no evidence of efficacy) 10 11 SectionConfidential: E: Renal System. The following drugs For are potentially Review inappropriate inOnly older people with 12 acute or chronic kidney disease with renal function below particular levels of eGFR (refer to 13 summary of product characteristics datasheets and local formulary guidelines) 14 15 1. Digoxin at a long-term dose greater than 125µg/day if eGFR < 30 ml/min/1.73m2 (risk of digoxin 16 toxicity if plasma levels not measured). 17 2. Direct thrombin inhibitors (e.g. dabigatran) if eGFR < 30 ml/min/1.73m2 (risk of bleeding). 18 19 3. Factor Xa inhibitors (e.g. rivaroxaban, apixaban) if eGFR < 15 ml/min/1.73m2 (risk of bleeding). 20 4. NSAID’s if eGFR < 50 ml/min/1.73m2 (risk of deterioration in renal function). 21 22 5. Colchicine if eGFR < 10 ml/min/1.73m2 (risk of colchicine toxicity). 23 6. Metformin if eGFR < 30 ml/min/1.73m2 (risk of lactic acidosis). 24 25 Section F: Gastrointestinal System 26 27 3. Drugs likely to cause constipation (e.g. antimuscarinic/anticholinergic drugs, oral iron, opioids, 28 verapamil, aluminium antacids) in patients with chronic constipation where nonconstipating 29 alternatives are available (risk of exacerbation of constipation). 30 31 Section G: Respiratory System 32 5. Benzodiazepines with acute or chronic respiratory failure i.e. pO2 < 8.0 kPa ± pCO2 > 6.5 kPa (risk 33 of exacerbation of respiratory failure). 34 35 Section H: Musculoskeletal System 36 37 1. Non-steroidal anti-inflammatory drug (NSAID) other than COX-2 selective agents with history of 38 peptic ulcer disease or gastrointestinal bleeding, unless with concurrent PPI or H2 antagonist (risk of 39 peptic ulcer relapse). 40 41 5. Corticosteroids (other than periodic intra-articular injections for mono-articular pain) for 42 osteoarthritis (risk of systemic corticosteroid side-effects). 43 44 Section I: Urogenital System 45 2. Selective alpha-1 selective alpha blockers in those with symptomatic orthostatic hypotension or 46 micturition syncope (risk of precipitating recurrent syncope). 47 48 Section J. Endocrine System 49 50 3. Beta-blockers in diabetes mellitus with frequent hypoglycaemic episodes (risk of suppressing 51 hypoglycaemic symptoms). 52 5. Oral oestrogens without progestogen in patients with intact uterus (risk of endometrial cancer). 53 54 6. Androgens (male sex hormones) in the absence of primary or secondary hypogonadism (risk of 55 androgen toxicity; no proven benefit outside of the hypogonadism indication). 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 3 of 42 BMJ

1 2 3 Section K: Drugs that predictably increase the risk of falls in older people 4 1. Benzodiazepines (sedative, may cause reduced sensorium, impair balance). 5 6 2. Neuroleptic drugs (may cause gait dyspraxia, Parkinsonism). 7 3. Vasodilator drugs (e.g. alpha-1 receptor blockers, calcium channel blockers, long-acting nitrates, 8 ACE inhibitors, angiotensin I receptor blockers, ) with persistent postural hypotension i.e. recurrent 9 drop in systolic blood pressure ≥ 20mmHg (risk of syncope, falls). 10 11 4. HypnoticConfidential: Z-drugs e.g. zopiclone, zolpidem, For zaleplon (mayReview cause protracted daytimeOnly sedation, 12 ataxia). 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 4 of 42

taken into taken account Previous Previous information Prescriptions weeks4 prior to period analysed Prescriptions weeks3 prior to period analysed Prescriptions 3 priormonths to period analysed Prescriptions 13weeks prior to period analysed Prescriptions weeks2 prior to period analysed Prescriptions weeks2 prior to period analysed Previous relatedinformation to prescriptions.

(1) (1) (1) (1) diltiazem within diltiazem weeks.2 weeks2 Patients with Patients eachin PIP year Prescribed verapamil or Prescribed beta-blocker within

Prescribed for doses weeks4 > Prescribed for doses weeks3 > Prescribed from drug samedrug within class weeks.2 • • prescribedantiarrhythmicNot / beta-blockers the months in 3 theirpreceding amiodarone first prescription. BMJ

https://mc.manuscriptcentral.com/bmj Patients Patients risk ofat PIP blocker verapamil diltiazemor a for Have prescription a for Have prescription Beta- Have a a forHave prescription NSAID Prescribed for doses months3 >

Have a a forHave prescription orZolpidem Zopiclone a forHave prescription Chloral hydrate a forHave prescription at least one the drugs theof in list. • • first theamiodaroneHave the prescription in analysed year

Confidential: For Review Only recommended duration, where treatment wellis duration defined ZopicloneZolpidem(up & to4 weeks) Chloral (upweeks)tohydrate 3 to Melatonin (up 13 weeks) toNSAIDs (up 3 months) a for Have prescription Melatonin Prescribed for doses 13 > weeks prescription verapamil diltiazemor antiarrhythmictherapy in supraventriculartachyarrhythmias 1. 1. drug Any prescribed beyondthe 2. duplicate Any classdrug 3. Beta-blocker combination in with 4. Amiodarone first-lineas ! 5     {Çhtt      {Çhtt    5  ! descriptionCriteria Indication of medication ofDuplicationdrug class therapy ofdrugs: List Stimulant laxatives ||Thiazide diuretics || Loop diuretics Beta-blockers || inhibitorsACE || || II Angiotensin Receptorantagonists Calcium-|| blockerschannel Oral||anticoagulants || Statins Benzodiazepines|| Z-drugs || hypnotics TCAs|| || Opioid|| SSRIs analgesics NSAIDs.|| Cardiovascular system 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Previous Previous relatedinformation to anddiagnosis prescriptions. Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Prescriptions weeks4 prior to period analysed all and available informationrelated to diagnosis Prescriptions weeks2 prior to period analysed Prescriptions weeks4 prior to period analysed all and available informationrelated to diagnosis Prescriptions weeks4 prior to period analysed in in

within within weeks2 weeks within weeks4 of antiplatelet prescription weeks of anticoagulant prescription Prescribed anticoagulant Prescribedwithin NSAID 2 Prescribed anticoagulant Prescribed antiplatelet within 4 Not prescribedanNot antihypertensive blocker, (beta ARB,ACE, CCB) 3the theirprecedingmonths first diureticloop prescription Prescribed diureticthiazide after diagnosis (anythiazide date diuretic prescriptionin thefor year, patients regularconsidered user) Prescribed phosphodiesterase type-5 inhibitors and afternitrates diagnosis (anyprescription date of both medications the in year, for patients regularconsidered user) Prescribed greater doses than 150 mg Prescribed ticlopidine prescribedNot PPI within weeks4 aspirin of prescription • • • • prescribedNot PPI within weeks4 antiplatelet of prescription

(3)

(2) BMJ

(4) (6)

(5) https://mc.manuscriptcentral.com/bmj anticoagulant a for Have prescription NSAID a for Have prescription atherosclerosis Have atherosclerosis Have Coded as havingas Coded hypertension and subsequently ahaving first prescription loopfor a diuretic analysed gout diagnosisHave code be or usersregular of for drugs gout diagnosis angina Have code be or usersregular of for drugs angina term Long users of aspirin peptic Have diagnosisulcer code prescribed and afterwardsaspirin • • • regular andcode diagnosis users of antiplatelet afterwards • regular andcode diagnosis users of anticoagulant afterwards a forHave prescription antiplatelet prescribedandNSAIDagent within weeks2 All patientsAll

Confidential: For Review Only for for hypertension gout with concurrent nitrate therapy for severeangina in heartfailure than 150mg per day withoutdisease ulcer concomitant PPI directantagonist, thrombin inhibitor Xafactor orinhibitors) in combination anticoagulant (vitamin Kantagonist, thrombin direct inhibitor factoror Xa inhibitors) within patients coronary/cerebrovascular/arterial disease withoutagent(s) PPI prophylaxis 5. 5. Loop diuretic first-lineas treatment 6. Thiazide with diuretic historya of 7. Phosphodiesterase type-5 inhibitors 8. Long-term aspirin at doses greater 9. any Ticlopidine in circumstances 10. with Aspirin pasta history ofpeptic 11. and NSAID anticoagulant (vitamin K 12. Antiplatelet agents with 13. with NSAID antiplateletconcurrent Antiplatelet/Anticoagulant drugs Page 5 of 42 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 6 of 42 Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Prescriptions weeks4 prior to period analysed Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis.

(1) Prescribed after TCA diagnosis (any date prescriptionTCA the in patientsyear, for considered user)regular Prescribed after TCA diagnosis (any date prescriptionTCA the in patientsyear, for considered user)regular Prescribed after TCA diagnosis date Prescribed after TCA diagnosis (any date prescriptionTCA the in patientsyear, for considered user)regular Prescribed afterneuroleptics diagnosis (anyneuroleptics date prescription in the for year, patients regularconsidered user) Prescribed for doses ≥ weeks4 Prescribed antipsychotics after diagnosis (anyantipsychotics date prescription in the for year, patients regularconsidered user) Prescribed anticholinergics / antimuscarinics after diagnosis (any date anticholinergics / antimuscarinics prescription in the patientsyear, for considered user)regular

(9) (9) BMJ

(8)

(7) https://mc.manuscriptcentral.com/bmj

(7) (10)

dementiaHave diagnosis code be or usersregular of antidementia drugs narrow Have glaucomaangle orregularcode diagnosis be users antiglaucoma of drugs cardiacHave conduction abnormalities diagnosis code BPH Have diagnosis orcode be usersregular of for drugs BPH BPH Have diagnosis orcode be usersregular of for drugs BPH a forHave prescription Benzodiazepines parkinsonismHave diagnosis code regular be or usersof antiparkinson drugs dementia/deliriumHave diagnosis be or regularcode usersof antidementia drugs

Confidential: For Review Only with any of with belowany the conditions: Dementia Narrow glaucoma angle conductionCardiac abnormalities urinaryProstatism and retention (BPH) antimuscarinicanticholinergic / effects with historyofa prostatism previous retentionor urinary quetiapine or those clozapine) in with parkinsonism with patients dementia delirium/ 14. Tricyclic antidepressants (TCAs) 15. with Neuroleptics moderate-marked 16. ≥4 Benzodiazepines weeksfor 17. Antipsychotics (i.e. otherthan 18. Anticholinergics antimuscarinics / in Central nervous Central andsystem drugs psychotropic 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Previous Previous relatedinformation to prescriptions. Previous relatedinformation to prescriptions. Previous informationrelated to diagnosis. Prescriptions weeks8 prior to analysedperiod (1) (1) Not prescribingNot any other theantidepresant in months 3 theirpreceding first TCA prescription. Prescribed weeks4within Prescribed anticholinesterase within weeks2 prescribingNot any other antipsychotic in3 the months theirpreceding phenothiazine first prescription. Prescribed first-generation antihistamines Prescribed prochlorperazineor metoclopramideafter diagnosis (any date prochlorperazine or metoclopramideprescription in the patientsyear, for considered user)regular Prescribed for doses weeks8 > Prescribed greater doses than 200 daily mg Anticholinergics/antimuscarinics BMJ

(11) https://mc.manuscriptcentral.com/bmj

(10) Have a first a Have prescriptionTCA the in year analysed a forHave prescription neuroleptic a ofHave prescription drug to heartrate reduce firsta Have phenothiazine the prescription in analysedyear parkinsonismHave diagnosis code regular be or usersof antiparkinson drugs a forHave prescription PPI a forHave prescription oral elemental iron All patientsAll

Confidential: For Review Only antidepressant treatment extra-pyramidaltreat side-effects of medicationsneuroleptic treatmentconcurrent with drugs reduce that heart rate beta- such as blockers, diltiazem,digoxin, verapamil psychotic treatment metoclopramide with Parkinsonism ordisease erosive peptic oesophagitis at full therapeutic for >dosage weeks 8 mg than 200daily Antidepressants (TCAs) as first-line 19. Initiation of TriCyclic 20. Anticholinergics/antimuscarinics to 21. Anticholinesterase withinhibitors 22. Phenothiazines first-lineas anti- 23. First-generation antihistamines 24. Prochlorperazine or 25. uncomplicated PPI for ulcerpeptic 26. elemental Oral dosesiron greater Gastro-Intestinal system Respiratory system Page 7 of 42 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 8 of 42 Previous Previous relatedinformation to diagnosis. Previous relatedinformation to anddiagnosis asthma prescription. Previous relatedinformation to diagnosis Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Previous relatedinformation to anddiagnosis hypertension prescription. Previous relatedinformation to anddiagnosis hypertension prescription. Prescriptions 3 priormonths to period analysed Not prescribed any Not other COPD treatment Prescribed for non-selective beta- blockereither diagnosis after date prescriptionasthmaor prescribedNot inhaled steroids Prescribed antimuscarinic bronchodilators diagnosisafter (any date antimuscarinic bronchodilators prescriptionin the patientsyear, for considered user)regular Prescribed antimuscarinic bronchodilators diagnosisafter (any date antimuscarinic bronchodilators prescriptionin the patientsyear, for considered user)regular Prescribed NSAID after diagnosis date Prescribed NSAID after diagnosis date prescribingNot paracetamol the in precedingmonths3 their first

(9) BMJ

(8) https://mc.manuscriptcentral.com/bmj Coded as havingas Coded COPD and subsequently ahaving first theophylline prescription for asthma Have andcodediagnosis prescribed treatmentasthma afterwards havingas Coded COPD and subsequently atleasthaving two prescription for systemic steroids narrow-angleHave glaucoma orregularcode diagnosis be users antiglaucoma of drugs BPH Have diagnosis orcode be usersregular of for drugs BPH hypertensionHave diagnosis code prescribed orand 3 more antihypertensive afterwardsclasses heart failure Have codediagnosis prescribed orand 3 more antihypertensive afterwardsclasses term Long user of NSAID (> 3 months)

Confidential: For Review Only Chronic obstructive Chronic pulmonary (COPD) disease (oral/topical) with history of asthma requiring treatment inhaled corticosteroids for maintenance therapy in moderate- COPD severe tiotropium)ipratropium, with a the history of below conditions: Narrow-angle glaucoma outflowBladderobstruction severe heart failure Severe hypertension failureSevere heart formonths) symptom relief of 27. as Theophylline for monotherapy 28. beta-blocker Non-selective 29. Systemic corticosteroidsof instead 30. Antimuscarinicbronchodilators (e.g. 31. NSAID with severe hypertension or 32. Long-term use of NSAID(>3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Previous Previous relatedinformation to diagnosis, DMARD prescriptions,and corticosteroids prescriptions priormonths3 to period analysed Previous relatedinformation to diagnosis, xanthine- oxidase inhibitor prescriptions,and NSAID prescriptions months3 prior to period analysed Previous relatedinformation to diagnosis. Prescriptions 3 priormonths to period analysed Previous relatedinformation to diagnosis. weeks of corticosteroids weeks of NSAID prescribed Not PPI within 4 prescribed Not PPI within 4 NSAID prescription.NSAID prescribeddiseaseNot modifying (DMARD)antirheumatic drug prescribedNot xanthine-oxidase inhibitor Prescribed COX-2 selective NSAIDs diagnosisafter date (any COX-2 selective NSAIDs prescription in the for year, patients regularconsidered user) • • Prescribed bisphosphonatesoral diagnosisafter date (any oral bisphosphonates prescription in the for patients year, considered user)regular , , (2) BMJ https://mc.manuscriptcentral.com/bmj

(12) (13) months) and and months) prescribed corticosteroids at the same time weeks)2 (within (>3corticosteroids months) and NSAIDatprescribed the same (withinweeks)time 2 term Long users of NSAID (>3 term Long users of Have rheumatoidarthritisHave diagnosis prescribedand code corticosteroids for more than 3 afterwards months gout diagnosisHave code be or usersregular of for drugs gout prescribed and NSAID colchicine or for more than 3 afterwards months cardiovascularHave disease orregularcode diagnosis be users fordrugs cardiovascularof disease • • diseaseGI Have diagnosis or code regularbe users of GI disease drugs

Confidential: For Review Only osteoarthritis osteoarthritis wherepain paracetamol has been triednot months) formonotherapyas rheumatoid arthritis formonths) chronic treatmentof where gout nothereis contraindication to a xanthine- inhibitoroxidase concurrent cardiovascular disease withoutcorticosteroids PPI prophylaxis with a current/recent history of upper gastrointestinal (GI) disease, upper bleeding)orGI belowthe conditions: 33. Long-term (>3corticosteroids 34. Long-term NSAIDor (>3colchicine 35. COX-2 selective NSAIDs with 36. with NSAID concurrent 37. bisphosphonates Oral in patients 38. Antimuscarinicdrugs with any of Urogenital Urogenital system Page 9 of 42 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 10 of 42 Previous Previous relatedinformation to diagnosis. Previous Previous relatedinformation to diagnosis. Previous Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Previous relatedinformation to diagnosis. Previous relatedinformation to prescriptions. Prescribed oestrogens after diagnosis date Prescribed oestrogens after diagnosis date Prescribed antimuscarinic after diagnosis (anyantimuscarinic date prescription in the for year, patients regularconsidered user) Prescribed antimuscarinic after diagnosis (anyantimuscarinic date prescription in the for year, patients regularconsidered user) Prescribed antimuscarinic after diagnosis (anyantimuscarinic date prescription in the for year, patients regularconsidered user) Prescribed Sulphonylureas Prescribed thiazolidenediones diagnosisafter date (any prescription ofthiazolidenediones the in for patients year, considered user)regular prescribedpainNot medications in 3the theirprecedingmonths first or transdermaloral strong opioid prescription. BMJ

(8)

(14) https://mc.manuscriptcentral.com/bmj

(3) (7) (9) Have venousthromboembolismHave codediagnosis Have dementiaHave diagnosis code be or usersregular of antidementia drugs narrow-angleHave glaucoma orregularcode diagnosis be users antiglaucoma of drugs chronicHave prostatism diagnosis be or regularcode usersof for drugs BPH heart failure Have codediagnosis or regularbe users of drugs for heart failure first theoralHave transdermalor prescription strong opioids in the year analysed Have breast cancerHave diagnosis code All patientsAll

Confidential: For Review Only

Dementia, or or Dementia, chronic cognitive impairment Narrow-angle glaucoma prostatismChronic with of typeaction diabetes 2 mellitus failure heart cancer venous thromboembolismor Breast cancer Venous thromboembolism (morphine,opioids oxycodone, fentanyl, buprenorphine, diamorphine, methadone, tramadol, pethidine, pentazocine) first as line 39. Sulphonylureas with a long duration 40. Thiazolidenediones with patients in 41. Oestrogens with history ofa breast 42. oral Use of or transdermal strong Endocrinesystem Analgesic drugs 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Laxativeprescriptions 6 tomonths prior period analysed Short-acting opioids prescriptions monthprior1 to period analysed Prescriptions weeks2 prior to period analysed Prescriptions weeks2 prior to period analysed Prescriptions weeks2 prior to period analysed Prescriptions weeks2 prior to period analysed antihistamines or antimuscarinics or anticholinergics within 2 weeks antimuscarinics or anticholinergics within 2 weeks antihistaminesgeneration anticholinergicsorwithin 2 weeks antihistaminesgeneration antimuscarinicsor within 2 weeks Prescribedgeneration first Prescribed orTCA Prescribed orTCA first Prescribed orTCA first • • • • Not prescribedNot laxative within 6 of months opioids prescribedNot short-acting opioids within monthof 1 long-acting opioids BMJ

(15) https://mc.manuscriptcentral.com/bmj generation antihistaminesgeneration antimuscarinics anticholinergics a for Have prescription TCA a for Have prescription first a for Have prescription a for Have prescription • • • •

Regular usersofRegular opioids a Have opioidslong-acting prescription

Confidential: For Review Only therapy fortherapy painmild opioidswithout PRN) concomitant laxative opioids actingfor break-through pain with drugs antimuscarinic/anticholinergic (e.g.bladderproperties antispasmodics, intestinal antispasmodics, tricyclic antidepressants, first generation antihistamines) Prescribed treatmentthat exceed the recommended duration or second prescriptionof this medication withinrecommended the duration, andtotal, in user Regular gout:of drugs whofor patients athave least prescription one gout of treatment (allopurinol,tisopurine, febuxostat, and colchicine, user Regular angina:of drugs for whopatients have atleast one prescription of diuretics loop ACEand the inhibitorfirst during months5 oneand prescription ofloop diureticsand inhibitor ACE during lastthe7 inmonths the year analysed. Long users term aspirin:of patients prescribed aspirin the first monthsduring 5 theduring and last months7 in analysed.the year usersRegular of antiplatelet: to rule cases ofout switching from an antiplatelet anticoagulant, to an patients tohavehad further a antiplatelet prescription weeksin4 the following their antitcoagulant prescription the number of the treatment days of the recommended exceed duration. agents)uricosuricduring the5 first months, one prescription theduring 7 last months only or prescription one at with least two in issues the year analysed. 43. regular Use of (as fromdistinct 44. Long-actingwithout opioids short- 45. Concomitant oruse of two more (1) (2) (3) (4) (5) Antimuscarinic/Anticholinergic drugburden Page 11 of 42 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 12 of 42

BMJ https://mc.manuscriptcentral.com/bmj

Confidential: For Review Only Regular usersRegular of anticoagulant:to outrule of cases switching an from anticoagulant to an antiplatelet, patients tohadfurther aanticoagulanthave user Regular of antidementia drugs:patientswho have at least one prescription ofantidementia drug first during the months, 5 prescription one of user Regular of antiglaucomadrugs: patientswho have at prescription least oneantiglaucoma of drug during first the5 one months, of prescription user Regular of drugs patientsfor BPH:who at have least prescriptionone for BPH theduring first months,5 prescription one for BPH during last the ormonths7 one prescription with atleast issues two inthe analysed. year user Regular of antiparkinsondrugs: patientswho have at prescription least oneantiparkinson of drug during first the5 one months, of prescription withPatients Zollinger Ellison and syndrome Barrett's were oesophagus excluded. user Regular of drugs for cardiovascular disease: patients havewho at least one prescription of drugs for cardiovasculardisease during first the 5 user Regular GIof drugs for disease: patients whohave at least one prescription fordrugs diseaseGIof the first months,5during prescriptionone of withPatients methadone were excluded. usersRegular patientsof opioids:who have at least 3 prescriptionsof theduring opioidsfirst half of the prescriptionsyear 3 and of opioids the during prescription in the 4 prescription weeksin4 the following their antiplatelet prescription antidementia thedrug during last months 7 prescriptiononeor with leasttwo issues at the in analysed. year antiglaucoma drug theduring last 7 prescription oneor months with at leasttwo issues the in analysed. year antiparkinson drug theduring last 7 oneor prescriptionmonths with at leasttwo issues the in analysed. year months, prescriptionone forofdrugs cardiovascular disease during last one the ormonths7 prescription with at issuesleast two in the year for GIdrugs disease the during7 last or months one prescription with leasttwo issues at the in analysed. year second half, a or prescription with at least 3 inissues first theand half a prescription with least at issues3 in the second half analysed.

(6) (7) (8) (9) (10) (11) (12) (13) (14)
 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 Page 13 of 42 BMJ

1 2 3 Prevalence of potentially inappropriate prescribing in older people in primary 4 5 care and its association with hospital admission: a longitudinal study 6 7 Teresa Pérez, associate professor in statistics1,2 8 9 Frank Moriarty, postdoctoral research fellow2 10 Confidential: For Review Only 11 2 12 Emma Wallace, senior lecturer in general practice 13 14 Ronald McDowell, postdoctoral research fellow in biostatistics2 15 16 Patrick Redmond, systematic review project manager2,3 17 18 Tom Fahey, professor of general practice2 19 20 1. Department of Statistics and Operational Research III, Complutense University of Madrid, 21 22 Madrid, Spain. 23 24 2. HRB Centre for Primary Care Research, Royal College of Surgeons in Ireland, Dublin, 25 26 Ireland. 27 28 3. Department of Public Health and Primary Care, School of Clinical Medicine, University of 29 30 Cambridge, Cambridge, UK 31 32 33 34 Main text: 4,391 words 35 36 37 Abstract: 307 words 38 39 Corresponding author: 40 41 Dr Frank Moriarty 42 43 HRB Centre for Primary Care Research, Department of General Practice, Royal College of 44 Surgeons in Ireland, Mercer Building, Lower Mercer Street, Dublin 2 45 46 Tel: 353-1-402-8575 47 48 Email: [email protected] 49 Twitter: @FrankMoriarty 50 51 ORCID ID: http://orcid.org/0000-0001-9838-3625 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 14 of 42

1 2 3 Abstract 4 5 Objective: To determine if hospitalisation is associated with potentially inappropriate prescribing 6 7 (PIP) among older primary care patients (aged ≥65 years), and if PIP was more likely after 8 hospitalisation relative to before. Our research question was how does hospitalisation impact on 9 10 appropriate prescribing. 11 Confidential: For Review Only 12 Design: Longitudinal study of retrospectively extracted data from general practice records. 13 14 Setting: Forty-four general practices in Ireland in the years 2012-2015. 15 16 Participants: Adults aged ≥65 years attending participating practices. 17 18 Exposure: Admission to hospital (any hospitalisation versus none, and post-hospitalisation versus 19 pre-hospitalisation). 20 21 Main outcome measures: Prevalence of PIP assessed using 45 criteria from the Screening Tool for 22 23 Older Persons’ Prescription (STOPP) version 2. This was analysed both as number of distinct PIP 24 25 (Poisson regression) and binary presence of PIP (logistic regression). We adjusted for patient 26 characteristics, and also conducted a sensitivity analysis matching with propensity scores based on 27 28 patient characteristics and diagnoses. 29 30 Results: Overall 38,229 patients were included, and during 2012, the mean age was 76.8 years (SD 31 8.2), and 43% were male. Each year, 10.4%-15.0% of patients had ≥1 hospital admission. The overall 32 33 prevalence of PIP ranged from 45.3% of patients in 2012 to 50.9% in 2015. Independent of age, 34 35 gender, number of prescription items, co-morbidity, and health cover, hospitalisation was associated 36 with higher number of distinct PIP e.g. adjusted rate ratio for hospitalisation was 1.26 (95%CI 1.22 to 37 38 1.30). Among participants who were hospitalised, the likelihood of PIP post-hospitalisation was 39 higher than pre-hospitalisation independent of patient characteristics e.g. adjusted odds ratio for 40 41 post-hospitalisation was 1.72 (95%CI 1.63 to 1.84). Analysis of propensity score-matched pairs 42 showed a slight reduction in the rate ratio for hospitalisation to 1.21 (95%CI 1.17 to 1.25). 43 44 45 Conclusions: Hospitalisation was independently associated with PIP. It is important to determine 46 how hospital admission may impact on appropriateness of prescribing for older people and how 47 48 potential adverse consequences of hospitalisation can be minimised. 49 50 51 52 53 54 55 56 57 58 1 59 60 https://mc.manuscriptcentral.com/bmj Page 15 of 42 BMJ

1 2 3 What is already known on this subject 4 5 • Potentially inappropriate prescribing is common among older people. 6 • It is associated with adverse outcomes including emergency hospital attendances and 7 admissions, adverse drug events, and poorer quality of life. 8 9 • Research to date has focussed on patient and GP characteristics as risk factors for poor 10 prescribing quality. 11 Confidential: For Review Only 12 13 14 What this study adds 15 16 • Our study suggests hospitalisation is independently associated with an increased number of 17 potentially inappropriate prescriptions. 18 • Patients who were hospitalised were more likely to have potentially inappropriate 19 20 prescribing post-hospitalisation compared to pre-hospitalisation, independent of patient 21 characteristics. 22 • This illustrates the need to consider and address potential adverse effects of hospitalisation 23 on prescribing appropriateness among older patients. 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 2 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 16 of 42

1 2 3 Print abstract 4 5 Study question Is hospitalisation associated with potentially inappropriate prescribing (PIP) among 6 7 older primary care patients (aged ≥65 years). 8 9 Methods This is a longitudinal study of retrospectively extracted data from general practice records 10 from forty-four general practices in Ireland in the years 2012-2015. Participants were adults aged 11 Confidential: For Review Only 12 ≥65 years attending participating practices. We examined the relationship between hospital 13 admission (both any hospitalisation versus none, and post-hospitalisation versus pre- 14 15 hospitalisation), and the outcome of PIP prevalence (assessed using 45 criteria from the Screening 16 Tool for Older Persons’ Prescription (STOPP) version 2). This was analysed both as number of distinct 17 18 PIP (Poisson regression) and binary presence of PIP (logistic regression). We adjusted for patient 19 20 characteristics, and also conducted a sensitivity analysis matching with propensity scores based on 21 patient characteristics and diagnoses. 22 23 Study answer and limitations Independent of age, gender, number of prescription items, co- 24 25 morbidity, and health cover, hospitalisation was associated with higher number of distinct PIPs 26 (adjusted rate ratio 1.26, 95%CI 1.22 to 1.30), and among hospitalised participants, the likelihood of 27 28 PIP post-hospitalisation was higher than pre-hospitalisation independent of patient characteristics 29 30 (adjusted odds ratio 1.72, 95%CI 1.63 to 1.84). As this is an observational study, there is potential for 31 unmeasured confounding which may partly or fully explain the relationship between hospitalisation 32 33 and PIP, and we were unable to examine whether cases of PIP may have been clinically justified. 34 35 What this study adds Our study suggests hospitalisation is independently associated with an 36 increased number of potentially inappropriate prescriptions following discharge back to primary 37 38 care. 39 40 Funding, competing interests, data sharing This research was supported by the Health Research 41 42 Board (HRB) in Ireland through grant no. HRC/2014/1, and the Spanish Ministry of Economy and 43 Competitiveness through grant MTM2016-75351-R. The authors have no other competing interests 44 45 to declare. No additional data are available. 46 47 48 49 [Please include Figure 1 as a small/simple figure if a figure is being used] 50 51 52 53 54 55 56 57 58 3 59 60 https://mc.manuscriptcentral.com/bmj Page 17 of 42 BMJ

1 2 3 Introduction 4 5 Adults aged ≥65 years are a growing population and represent the largest consumers of prescribed 6 7 medications.[1,2] Although optimal prescribing aims to maximise patients benefits while minimising 8 harms and cost, achieving this balance when caring for older patients in primary care can be 9 10 challenging.Confidential: Physiological changes in ageing For can impair Review metabolism and excretion Only of drugs and 11 increase sensitivity to their effects.[3] In addition, older patients tend to have a higher burden of 12 13 multimorbidity and so take more medications, contributing to both increased treatment burden, 14 and potential drug-drug and drug-disease interactions.[4] Lastly, although most prescribing in 15 16 primary care is repeat prescribing,[5] such medications are often initiated in secondary care which 17 18 can be problematic as the general practitioner (GP) is responsible for co-ordination and managing all 19 prescriptions.[6] This can be even more challenging for patients with multimorbidity who attend 20 21 multiple healthcare providers. 22 23 Medication use among older adults is increasing, despite the high risk of adverse drug events (ADEs) 24 and resultant morbidity and mortality.[1,2,7] A recent systematic review focusing on ADEs in 25 26 ambulatory care found prevalence rates ranging from 2.8%-34.7%, up to a quarter of which were 27 28 judged to be preventable.[8] Another systematic review reported that 9.9% of all hospital 29 admissions in those aged ≥65 years were as a result of an ADE.[9] 30 31 Appropriateness of prescribing can be assessed by process (i.e. what providers do) or outcome 32 33 measures (i.e. patient outcomes). These can be either implicit (judgment-based) or, more often, 34 explicit (criterion-based).[10] Examples of explicit measures include the Beers criteria, and Screening 35 36 Tool of Older Person’s potentially inappropriate Prescribing (STOPP) and the Screening Tool to Alert 37 doctors to the right treatment (START).[11] Explicit measures have the advantage of being based on 38 39 literature review and expert consensus and are both reliable and have content validity, although 40 41 they do periodically require revision to reflect new evidence. In 2015, the STOPP/START criteria were 42 updated to add new and remove obsolete criteria.[11] In STOPP/START 2, the final list of 114 criteria 43 44 was agreed after two Delphi validation rounds, including 80 STOPP criteria and 34 START criteria.[11] 45 The STOPP/START 2 criteria can be used to examine potentially inappropriate prescribing (PIP) in 46 47 older people. 48 49 The adverse outcomes associated with the STOPP criteria are well established, including ADEs, 50 51 emergency admissions or emergency department visits, and poorer quality of life.[12–15] Previous 52 studies have examined predictors of PIP, such as patient characteristics (e.g. multimorbidity, age, 53 54 and number of prescribed medications), and GP practice characteristics (e.g. deprivation of 55 catchment area).[16,17] There has been less focus on how health system factors, such as 56 57 58 4 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 18 of 42

1 2 3 hospitalisation or care transitions, may contribute to the appropriateness of prescribing for 4 ambulatory care patients. 5 6 Therefore, the objectives of this study are: 7 8 (i) to estimate the annual prevalence of PIP in older community-dwelling people in Ireland 9 10 using the revised STOPP criteria, 11 Confidential:(ii) to examine any association between For hospital Review admission and PIP, andOnly 12 (iii) to compare PIP prevalence pre- and post-hospitalisation. 13 14 We hypothesised that levels of PIP among older adults may be significantly associated with hospital 15 16 admission and, among patients who were hospitalised, levels of PIP may differ post-admission and 17 pre-admission. 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 5 59 60 https://mc.manuscriptcentral.com/bmj Page 19 of 42 BMJ

1 2 3 Methods 4 5 Study population and study design 6 7 This is a longitudinal study of a dataset collected retrospectively which included GP patients aged 8 ≥65 years between 2012 and 2015. Data were collected from 44 general practices in Ireland using 9 10 the patient software management system Socrates (www.socrates.ie) and included prescribing, 11 Confidential: For Review Only 12 demographic, clinical and hospitalisation records. Socrates is one of four electronic health record 13 vendor systems accredited by the Irish College of General Practitioners. The majority of general 14 15 practices in Ireland (94%) are computerised and electronic morbidity coding and prescribing occurs 16 in over 90% of these computerised practices.[18] Although the validity of morbidity recording in 17 18 Ireland is not as good at the UK, recent initiatives have improved both completeness and validity of 19 morbidity coding.[19] Socrates has created quality indicator tools used for audit and also in research, 20 21 such as a study of resistance patterns of urinary tract infections.[20] Ethical approval was obtained 22 23 from the Irish College of General Practitioners. The STROBE (STrengthening the Reporting of 24 OBservational studies in Epidemiology) statement was used in the conduct and reporting of this 25 26 study.[21] 27 28 Explanatory variables and outcomes of interest 29 30 Potentially inappropriate prescribing, according to 45 STOPP 2 criteria, was identified using 31 information on medications and diagnoses for each patient in the dataset, in each of the four years A 32 33 total of 35 criteria could not be applied (44%), for example due to lack of information on laboratory 34 monitoring, history of falls, or prescribing indication (see Appendix 1). Where necessary, prescribing 35 36 and diagnosis information prior to 2012 was included in estimating PIP prevalence e.g. for criteria 37 38 relating to first-line treatment. An extensive description of criteria and their application is provided 39 in Appendix 2. For each patient the total number of distinct PIP per calendar year was calculated. 40 41 We analysed this either as a count variable, i.e. number of distinct PIPs per year, or a dichotomous 42 variable, i.e. at least one PIP in the period considered or no PIP. 43 44 The STOPP criteria, as explicit measures of inappropriate prescribing, have been used extensively in 45 46 research as process measures of care. Their validity has been established in multiple studies 47 48 demonstrating their relationship with important outcomes for patients. In terms of predictive 49 validity, STOPP modestly discriminates for outcomes such as ADEs, emergency department visits and 50 51 hospitalisations (c-indices of 0.65-0.70).[22] Other observational studies have found consistent 52 associations between the STOPP criteria and avoidable ADEs relevant to the index admission among 53 54 hospital inpatients,[12] poorer quality of life,[14] emergency department visits,[14,15] and 55 unplanned hospital re-admission.[23] Prescribing included in STOPP was considered causal in 30% of 56 57 58 6 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 20 of 42

1 2 3 adverse drug reactions in a Swedish older population study,[24] while in a study of definitely or 4 possibly avoidable ADEs which caused hospitalisation, 62.2% were listed in the STOPP criteria.[12] 5 6 On this basis, the STOPP criteria can be considered a valid process measure of quality care, and they 7 have been used as primary outcomes in interventional trials aiming to improve prescribing.[25,26] 8 9 10 To examineConfidential: the association between hospitalisation For and Review PIP, the independent Only variable 11 hospitalisation was defined as a dichotomous variable (no hospitalisation versus any hospitalisation). 12 13 For all practices, elective and emergency admissions to public hospitals were included, and four of 14 the 44 practices additionally included emergency department attendances. While comparing PIP 15 16 pre- and post-hospital admission, the independent variable was time period (after hospitalisation 17 versus before). The post-hospitalisation period starts on the day following hospital admission. For 18 19 those patients hospitalised more than once in the same year, only the first admission was 20 21 considered. 22 23 Categorical covariates adjusted for in all models were gender and health cover type (with 4 24 categories: General Medical Services (GMS) scheme, Doctor Visit Card (DVC), private patients (PRV), 25 26 and other). Continuous covariates were age (years), number of prescription items in that period and 27 multimorbidity. The GMS and DVC schemes are types of public health coverage, providing eligible 28 29 patients with a range of health services including GP visits free of charge. These are means-tested, 30 31 with eligibility based on household income and age. The GMS scheme covers the most 32 socioeconomically deprived individuals, approximately one-third of the population, and 90% of 33 34 those aged ≥70 years where a lower income threshold applies.[27] The DVC scheme covers 35 individuals with higher, but still limited, means. Other individuals pay out-of-pocket for primary care 36 37 services such as GP visits and medications; hence Ireland has a mixed public-private healthcare 38 system. The number of prescription items was assessed as the total number of items prescribed to a 39 40 patient per year, not accounting for multiple issues/repeats on prescriptions. Multimorbidity was 41 42 assessed using the Charlson comorbidity index (CCI).[28] The CCI is based on 17 conditions weighted 43 by 1-year mortality risk, and a higher score indicates more severe comorbidity. 44 45 46 Statistical analyses 47 Annual prevalence of potentially inappropriate prescribing (PIP) 48 49 Demographic and clinical characteristics of patients (e.g. age, gender, health cover type, number of 50 51 prescription items and multimorbidity) and the overall prevalence of PIP are described for each year 52 considered. Data are expressed as mean (standard deviation), median (interquartile range), and 53 54 proportions (absolute and relative frequencies) as appropriate. 55 56 Association between PIP and hospitalisation 57 58 7 59 60 https://mc.manuscriptcentral.com/bmj Page 21 of 42 BMJ

1 2 3 We examined the relationship between PIP and hospitalisation adjusted for age, gender, health 4 cover type, number of prescription items and multimorbidity. Both mixed-effect logistic regression 5 6 models (where the outcome was defined as dichotomous (0 without any PIP in that year or 1 7 otherwise)) and mixed-effect Poisson regression models (where the count outcome was the number 8 9 of distinct PIPs observed in each patient per calendar year) were fitted. These models extend the 10 11 generalConfidential: linear model (GLM) by incorporating For correlations Review among the outcomes Only (multiple 12 observations per patient). This can be accomplished by including random effects. In this study, we 13 14 introduced two random effects representing the patient and the year. We used MCMCglmm 15 package in R,[29] because models obtained using the glmer function of the lme4 package did not 16 17 converge. We addressed potential overdispersion in the Poisson models in the MCMCglmm function 18 19 by obtaining robust standard error for the parameter estimates. In order to avoid double counting, 20 criteria 32 was omitted from the Poisson analysis because it overlapped fully with criteria 1 (both 21 22 relate to long-term use of NSAIDs, see Appendix 2 for further details). 23 24 PIP pre-and post-admission to hospital 25 26 A second analysis was performed comparing PIP pre- and post- hospitalisation among only those 27 patients who were admitted to hospital during a study year. Paired-sample tests (i.e. having two 28 29 observations per patient: one for presence/absence of PIP before hospitalisation and one following 30 31 hospitalisation) allowed the temporality of the relationship between hospitalisation and PIP to be 32 assessed and accounted for between-patient variability. We fitted a mixed-effect logistic regression 33 34 model, and included a random intercept for each patient to allow between-patient variability in the 35 outcome and for each year, using the MCMCglmm package in R.[30] The outcome was whether or 36 37 not the patient had any PIP in the time period considered. The independent variable was time 38 period (after hospitalisation, relative to before hospitalisation), with adjustments made for the 39 40 covariates listed above. A count outcome was not analysed as it could not be assumed that the 41 42 distribution of number of PIPs was well approximated by the Poisson distribution due to the smaller 43 sample size and shorter observation period. In all analyses we defined p<0.05 as statistically 44 45 significant. 46 47 Sensitivity analyses 48 49 Firstly, we repeated each of the above analyses separately by calendar year to assess the consistency 50 of observed associations over the study period. Secondly, due to some missing data for the CCI, we 51 52 also repeated analyses using an alternative measure of multimorbidity. We used RxRisk-V, a 53 prescription-based measure of morbidity including medication proxies for 45 conditions, which has 54 55 shown criterion validity and reliability compared to patients' medical diagnoses.[31] Prescription 56 57 58 8 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 22 of 42

1 2 3 data was available for all included participants and RxRisk was adjusted for in models as a binary 4 indicator of multimorbidity (i.e. 2 or more conditions). 5 6 Lastly, as patients were not randomly allocated to being hospitalised or not being hospitalised, these 7 8 groups may have differences in their characteristics which could bias estimates. A sensitivity analysis 9 10 usingConfidential: propensity score-matching was conducted For to assess Review if the association between Only hospitalisation 11 and PIP could be due to unmeasured confounders. [32] The propensity score, defined as the 12 13 conditional probability of hospitalisation given the measured covariates, was used to balance 14 covariates in the two groups. Using the MatchIt package in R,[33] a logistic regression model was 15 16 first fitted to estimate propensity scores. We modelled the conditional probability of hospitalisation 17 as a function of baseline and those clinical characteristics associated with hospitalisation that were 18 19 also independent risk factors of PIP. These variables included: age, gender, health cover type, 20 21 number of prescription items, CCI, and if the patient had been diagnosed with any of the five most 22 common conditions (diabetes, COPD, any type of tumour, a myocardial infraction or cerebrovascular 23 24 disease). Each patient with a hospitalisation was randomly selected and then matched with the 25 patient with no hospitalisation with the nearest propensity score. Finally the same mixed-effect 26 27 models were fitted considering only the matched pairs. 28 29 Patient involvement 30 31 Patients were not involved in the conception, design, or conduct of this research. We plan to 32 disseminate the findings to the public and patients through our contacts in patient representative 33 34 bodies, the popular media, and through the participating general practices. 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 9 59 60 https://mc.manuscriptcentral.com/bmj Page 23 of 42 BMJ

1 2 3 Results 4 5 Descriptive statistics 6 7 A total of 38,229 patients were included in the dataset over the period 2012 to 2015. Demographics 8 and clinical characteristics of this sample, by year, are presented in Table 1. We excluded patients 9 10 without prescriptions during the period analysed. During 2012, the mean age of included patients 11 Confidential: For Review Only 12 was 76.8 years (standard deviation 8.2), and 43% were male. For each study year, 10.4%-15.0% of 13 patients had at least one hospitalisation. 14 15 16 Annual prevalence of potentially inappropriate prescribing (PIP) 17 The overall prevalence of PIP ranged from 45.3% of patients in 2012 to 50.9% in 2015 (Appendix 3). 18 19 The individual criteria with the highest prevalence in 2015 included proton pump inhibitor (PPI) for 20 uncomplicated peptic ulcer disease or erosive peptic oesophagitis at full therapeutic dosage for >8 21 22 weeks (26.9%), benzodiazepines for ≥4 weeks (19.1%), and drugs prescribed beyond the 23 recommended duration (13.7%, primarily driven by Z-drug hypnotics for >4 weeks), and this was 24 25 observed in each calendar year (Appendix 3). 26 27 Association between PIP and hospitalisation 28 29 In mixed-effect Poisson regression models, hospitalisation, higher age, number of prescription items 30 and multimorbidity were all associated with a higher number of PIPs. Rates ratios for hospital 31 32 admission, shown in Figure 1 with 95% credible intervals (CIs), is 1.26 (95%CI 1.22 to 1.30) when 33 34 controlling for the other covariates. Thus, the expected number of distinct PIPs per year increased by 35 at least 26% if a patient had been admitted to hospital. For gender, the number of PIPs per year is 36 37 approximately 19% lower for men. Health cover status is also significantly associated with PIP; the 38 incident rate ratio is 32% higher for GMS (i.e. public) patients compared to private patients. The 39 40 number of distinct PIP observed in one year also increases as age, number of prescription items and 41 42 multimorbidity increase. 43 44 Results obtained from the mixed-effect logistic regression model are analogous, although in this 45 model age is not significant (see Appendix 4). The odds ratio (OR) for hospital admission is 1.49 46 47 (95%CI 1.42 to 1.58), i.e. the probability of at least one PIP during a year increases by 49% for 48 hospitalised patients, after adjusting for relevant covariates. 49 50 51 52 53 54 55 56 57 58 10 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 24 of 42

1 2 3 PIP pre- and post-admission to hospital 4 5 Having analysed PIP in patients hospitalised compared to those who were not, we determined the 6 7 impact of hospitalisation on an individual’s likelihood of having PIP among only those patients who 8 were admitted. Figure 2 shows the estimated ORs with 95% CIs. Among patients that had at least 9 10 one hospital admission in a year, the risk of having any PIP increases after being admitted to hospital 11 Confidential: For Review Only by 72%. Patients with greater numbers of prescription items and females are more likely to have 12 13 PIPs. Again, associations were consistent across study years. 14 15 Sensitivity Analysis 16 17 When analyses were repeated on a year-by-year basis, the relationship between hospitalisation and 18 19 PIP was consistent over time (Appendix 5). Adjusting for multimorbidity using RxRisk instead of the 20 CCI (Table 2), therefore including participants for whom diagnostic coding may have been missing, 21 22 resulted in little change in the magnitude of the parameter estimates for hospitalisation. Lastly, 23 propensity score-matching was conducted to compare hospitalised patients to those who were not 24 25 hospitalised using both the Poisson model (Figure 3) and the logistic model (Appendix 6). These 26 analyses still showed a statistically significant association between hospital admission and the 27 28 outcome of PIP (adjusted rate ratio 1.21; 95%CI 1.17 to 1.25 and adjusted odds ratio 1.48; 95%CI 29 30 1.37 to 1.58). 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 11 59 60 https://mc.manuscriptcentral.com/bmj Page 25 of 42 BMJ

1 2 3 Discussion 4 5 6 Statement of principal findings 7 This study found a substantial proportion of community-dwelling older people had at least one PIP 8 9 defined by the STOPP 2 criteria, and that hospitalisation was a significant marker of PIP. Set against a 10 general increase in PIP and patients on multiple PIPs, we determined that after controlling for the 11 Confidential: For Review Only 12 characteristics assessed in this study (such as age, number of prescriptions and multimorbidity) 13 14 hospitalisation was associated with a higher rate of PIP. Furthermore, for patients who are 15 hospitalised, their likelihood of having PIP increased after admission compared to before by 72%, 16 17 independent of other patient factors. These relationships were consistent across study years and 18 were robust to different analytical approaches in sensitivity analyses. 19 20 21 Strengths and weaknesses of the study 22 This study includes a large sample of community-dwelling older adults and uses the most recent 23 24 version of the STOPP criteria to assess PIP. Using two different approaches (unpaired and paired 25 samples), consistent conclusions were obtained. However, due to the secondary nature of this 26 27 analysis, 35 of 80 STOPP criteria (44%) for which relevant patient information was unavailable could 28 29 not be applied (see Appendix 1). Like other explicit measure of PIP, STOPP does not account for 30 clinical judgement and individual clinical circumstances where prescribing may be justified and 31 32 appropriate in certain cases, however despite any clinical rationale that may be present, STOPP has 33 consistently been associated with poorer patient outcomes.[13] The quality of clinical coding of 34 35 diagnoses was somewhat variable, which precluded application of the START criteria to identify 36 prescribing omissions. This may be of particular interest for future research assessing the impact of 37 38 hospitalisation on prescribing appropriateness, as unintentional omission of treatments is noted as 39 40 the most common medication error at transitions of care.[34] We addressed the quality of clinical 41 and diagnostic coding by conducting a sensitivity analysis using a prescription-based measure of 42 43 multimorbidity for adjustment and demonstrated little impact on the results. While GP practices 44 were recruited from a wide geographical area, they may not be representative of all practices with 45 46 the potential for volunteer bias. However, the analysis includes patients with any type of health 47 48 cover, compared to other studies limited to participants eligible for the means-tested GMS scheme. 49 There was variability in hospitalisation coding between practices, and private hospital admissions 50 51 were not captured, hence there is potential for misclassification of exposure status for patients. 52 However the vast majority of secondary care interactions for older patients would be with public 53 54 hospitals and so this is unlikely to significantly impact the findings. Although we adjusted for a range 55 of patient characteristics, as with any observational study there is potential for unmeasured 56 57 58 12 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 26 of 42

1 2 3 confounding, which may partly or fully explain the relationship between hospitalisation and PIP. We 4 assessed the robustness of our result to different adjustment method using a propensity-matched 5 6 sensitivity analysis, however there may still be residual confounding due to other factors, such as the 7 illness severity. 8 9 10 Comparison with previous studies 11 Confidential: For Review Only The literature examining the impact of hospitalisation on PIP is limited. Some studies have compared 12 13 medication appropriateness at hospital admission and discharge, including PIP defined by Beers 14 15 criteria alone,[35] or in addition to STOPP/START.[36,37] In these studies, either no difference,[36] 16 or a small reduction in PIP was found between admission and discharge.[35,37] However, only the 17 18 relatively short period of hospitalisation was considered and the impact on primary care prescribing 19 after discharge was not assessed. These studies included between approximately 180 and 2,000 20 21 patients, so in contrast to the present study of over 40,000 individuals, may have been 22 underpowered to detect an association. 23 24 25 A previous study assessed the prevalence of PIP among 1,016 older GMS scheme patients in Ireland 26 presenting at one emergency department following a fall.[38] The overall prevalence of both the 27 28 STOPP (version 1) and Beers criteria (2012) did not change in the 12 months post-fall compared to 29 pre-fall. Prescribing of some medications associated with falls, such as neuroleptics and 30 31 benzodiazepines, did decrease however. Discordance between these findings and the present study 32 may be because these patients were attending hospital for a specific adverse event (i.e. a fall) and 33 34 therefore it is likely an assessment of risk factors contributing to this, including medications, was 35 36 conducted during/after hospital discharge. 37 38 The present study applied the recently revised 2015 STOPP criteria, i.e. the most current definition 39 of PIP. The prevalence here is closely comparable to estimates from other studies using STOPP 2 40 41 which ranged from 40.4% and 56% among community-dwelling people aged ≥65 years, and ≥80 42 years respectively,[39,40] to between 41.5% and 71.5% in older patients being discharged from 43 44 hospital.[36,41] Like the present study, long-term prescription of benzodiazepines and Z drugs was 45 common in a number of other studies using STOPP 2.[36,39–42] In contrast, the long-term use of 46 47 PPIs, the most common criterion in the present study, was only noted as particularly prevalent in 48 49 two previous studies using STOPP 2.[36,42] 50 51 Implications for clinicians, research and policy 52 53 Inpatient admissions can provide the opportunity for specialist teams to review and optimise 54 management of older patients’ chronic conditions, including their medications.[43] Although 55 56 hospitalisations have the potential to improve management of medications, this study suggests 57 58 13 59 60 https://mc.manuscriptcentral.com/bmj Page 27 of 42 BMJ

1 2 3 these possible benefits to prescribing appropriateness after discharge to primary care are not being 4 realised. In fact, our findings suggest that hospitalisation (which may result from a change in a 5 6 patient’s clinical status and may result in an intensification of healthcare) is an important driver of 7 PIP and the overuse and/or misuse of medications. Medicines management services for inpatients in 8 9 Ireland are broadly similar to the UK, however the extent to which they are provided in practice is 10 11 variableConfidential: due to hospital pharmacy resourcing. For In approximately Review 40% of Irish hospitalsOnly pharmacists 12 conduct admission medication reconciliation and review, which is similar to the proportion in UK 13 14 hospitals, although fewer Irish hospitals involve pharmacists in emergency department and acute 15 medical assessment units.[44] The majority provide inpatient clinical pharmacy services, however 16 17 unlike the UK, in most Irish hospitals (86%) pharmacists had no formal involvement in the discharge 18 19 prescribing process. The vast majority also do not supply medications to patients on discharge, while 20 about half provide pharmacist counselling on discharge medications.[44] The 2017 National Patient 21 22 Experience Survey report underlines the need for improved medication management services at 23 discharge, where 40% of patients reported not being advised about medication side effects to be 24 25 aware of.[45] 26 27 Poor coordination of transitions between care settings, i.e. from secondary to primary care, can put 28 patients at increased risk of medication errors, ADEs and readmissions.[46–48] Improving 29 30 coordination of care, particularly for older patients with complex care needs, has been identified as 31 32 an international policy priority.[49,50] Transitional care interventions for older patients with chronic 33 disease discharged from hospital to primary care have been evaluated in a recent systematic 34 35 review.[51] Evidence suggests these interventions can reduce mortality, hospital readmissions, and 36 number of readmission days after 3-18 months (for example a mortality risk difference at 18 months 37 38 of −0.07 [95%CI −0.12 to −0.02]), however no evidence of a quality of life benefit was demonstrated 39 40 in meta-analysis.[51] A recent quasi-experimental study evaluated the effect of a medication 41 management system (Pharm2Pharm) provided by hospital and community pharmacists for older 42 43 adults at risk of medication problems.[52] The intervention appeared to reduce the medication- 44 related hospitalisation rate and provide cost savings. 45 46 More effective means of medicines reconciliation in hospital and primary care, for example through 47 48 the availability of a summary care record, may allow for more clinician time to be focussed on 49 assessment of appropriateness of medications.[53] Similarly, implementing a standardised electronic 50 51 format for discharge summaries could improve their quality and reduce discrepancies arising from 52 53 transitions between hospital and primary care.[54] As well as addressing deficits in communication, 54 a robust electronic record system could also incorporate decision support to aid clinicians in 55 56 reviewing prescriptions, which combined with incentives and professional education has been 57 58 14 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 28 of 42

1 2 3 shown to effectively reduce high-risk prescribing and associated adverse events.[55] A large-scale 4 study of almost 1 million patients in UK general practice has demonstrated that there is high 5 6 variation between practices in the prevalence of such high-risk prescribing,[56] suggesting that 7 practice-level interventions to improve prescribing should be targeted. Variation among practices in 8 9 the impact of hospitalisation on appropriate prescribing also warrants examination to help inform 10 11 strategiesConfidential: to address this. For Review Only 12 13 There are a number of potential solutions individual clinicians may consider. A recent systematic 14 review identified incomplete clinical picture (information deficits due to poor communication among 15 16 multiple prescribers and fragmentation at care interfaces) as a barrier to minimising inappropriate 17 medications by prescribers.[57] Many of the common STOPP criteria in our study relate to 18 19 inappropriate duration of use, so documenting and clearly communicating the intended prescription 20 21 duration or planned review date would ensure other clinicians such as GPs have complete 22 information for reviewing and stopping such prescriptions. Similarly, documentation of the 23 24 indication for a medication will facilitate review of appropriateness and continued need.[57] The 25 indication and duration should also be discussed with patients, which would mean they expect 26 27 future review or stopping of medicines and thus reduce patient ambivalence/resistance to change as 28 a barrier to appropriate prescribing.[57,58] Prescribers have also cited a lack of evidence and 29 30 difficulty in assessing the benefits/harms of therapy as a barrier.[57] A number of evidence-based 31 32 guidelines have recently been developed to support decisions on deprescribing specific medications, 33 including PPIs, benzodiazepines and Z-drugs which were among the most prevalent issues identified 34 35 in our study.[59,60] Deprescribing algorithms and patient information leaflets and decision aids as 36 companions to these guidelines are also available from www.deprescribing.org. 37 38 It is not possible to attribute whether the observed increase in PIP is a consequence of illness that 39 40 prompted hospital admission or whether PIP is a consequence of further medical intervention during 41 42 hospital stay. Future research should identify the mechanisms by which hospitalisation is associated 43 with PIP, including detailed review of patient clinical records to explore how PIP may have been 44 45 contributory or causal in hospital admissions, and to understand the clinical decisions (both in 46 primary and secondary care) that resulted in PIP among patients following hospital discharge. 47 48 Research should also evaluate how to address these to enhance prescribing appropriateness for 49 older patients after discharge. This may include better continuity of information through improved 50 51 health information and communication technology infrastructure, as well as formal transitional care 52 53 programmes.[51,61] In addition, hospital-based interventions to enhance appropriateness of 54 prescribing for older patients should be evaluated, such as reviews using prescribing criteria like 55 56 57 58 15 59 60 https://mc.manuscriptcentral.com/bmj Page 29 of 42 BMJ

1 2 3 STOPP or alignment of clinical pharmacists with medical teams to provide integrated medicines 4 management.[62,63] 5 6 7 Conclusions 8 This study demonstrates that PIP is becoming increasingly prevalent among community-dwelling 9 10 older people according to the most recent STOPP criteria. Furthermore, hospitalisation is 11 Confidential: For Review Only independently associated with an increased risk of PIP following discharge back to primary care. It is 12 13 vital to identify optimal management strategies for older people to ensure the risk of inappropriate 14 15 medications is minimised following transitions of care. 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 16 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 30 of 42

1 2 3 Acknowledgements 4 5 Contributors: All authors contributed to the conception and design of this study. RMcD, PR, and TF 6 acquired the data. TP, FM, and RMcD analysed the data, and all authors interpreted the data. TP and 7 FM drafted the manuscript, and all authors were involved in critical revision and approval of the final 8 9 manuscript. TF is the guarantor. 10 11 CompetingConfidential: interests: All authors have completed For the ICMJEReview uniform disclosure Only form at 12 www.icmje.org/coi_disclosure.pdf and declare: support from the Health Research Board (HRB) in 13 Ireland through grant no. HRC/2014/1 (TF), and the Spanish Ministry of Economy and 14 Competitiveness through grant MTM2016-75351-R (TP); no financial relationships with any 15 organisations that might have an interest in the submitted work in the previous three years; and no 16 17 other relationships or activities that could appear to have influenced the submitted work. 18 19 Funding: The funders had no role in the design and conduct of the study; collection, management, 20 analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and 21 decision to submit the manuscript for publication. 22 23 Acknowledgements: The authors gratefully acknowledge the contributions of Dr Fiona Boland and 24 Dr Tamasine Grimes to the initial study concept, and all of the participating general practitioners and 25 patients. 26 27 Data sharing: No additional data are available. 28 29 Transparency statement: The lead author (FM) affirms that the manuscript is an honest, accurate, 30 31 and transparent account of the study being reported; that no important aspects of the study have 32 been omitted; and that any discrepancies from the study as planned have been explained. 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 17 59 60 https://mc.manuscriptcentral.com/bmj Page 31 of 42 BMJ

1 2 3 References 4 5 1 Qato DM, Wilder J, Schumm LP, et al. Changes in Prescription and Over-the-Counter 6 Medication and Dietary Supplement Use Among Older Adults in the United States, 2005 vs 7 2011. JAMA Intern Med 2016;176:473–82. doi:10.1001/jamainternmed.2015.8581 8 9 2 Moriarty F, Hardy C, Bennett K, et al. Trends and interaction of polypharmacy and potentially 10 inappropriate prescribing in primary care over 15 years in Ireland: a repeated cross-sectional 11 Confidential:study. BMJ Open 2015;5:e008656. doi:10.1136/bmjopen-2015-008656For Review Only 12 3 Mangoni A, Jackson S. Age-related changes in pharmacokinetics and pharmacodynamics: 13 basic principles and practical applications. Br J Clin Pharmacol 2003;57:6–14. 14 doi:10.1046/j.1365-2125.2003.02007.x 15 16 4 Mallet L, Spinewine A, Huang A. The challenge of managing drug interactions in elderly 17 people. Lancet 2007;370:185–91. 18 5 Petty DR, Zermansky AG, Alldred DP. The scale of repeat prescribing – time for an update. 19 BMC Health Serv Res 2014;14:76. doi:10.1186/1472-6963-14-76 20 21 6 Fahey T, Sinclair H. Hospital initiated prescribing in the General Medical Services scheme. Ir 22 Med J 1993;86:122–4. 23 24 7 Fick D, Waller J, Maclean J, et al. Potentially inappropriate medication use in a Medicare 25 managed care population: Association with higher costs and utilization. J Manag Care Pharm 26 2001;7:407–13. 27 8 Tache S, Sonnichsen A, Ashcroft D. Prevalence of adverse drug events in ambulatory care: a 28 systematic review. Ann Pharmacother 2011;45:977–89. doi:10.1345/aph.1P627 29 30 9 Kongkaew C, Noyce PR, Ashcroft DM. Hospital Admissions Associated with Adverse Drug 31 Reactions: A Systematic Review of Prospective Observational Studies. Ann Pharmacother 32 2008;42:1017–25. doi:10.1345/aph.1L037 33 10 Spinewine A, Schmader KE, Barber N, et al. Appropriate prescribing in elderly people: how 34 well can it be measured and optimised? Lancet 2007;370:173–84. doi:10.1016/S0140- 35 6736(07)61091-5 36 37 11 O’Mahony D, O’Sullivan D, Byrne S, et al. STOPP/START criteria for potentially inappropriate 38 prescribing in older people: version 2. Age Ageing 2015;44:213–8. doi:10.1093/ageing/afu145 39 40 12 Hamilton H, Gallagher P, Ryan C, et al. Potentially inappropriate medications defined by 41 STOPP criteria and the risk of adverse drug events in older hospitalized patients. Arch Intern 42 Med 2011;171:1013–9. doi:10.1001/archinternmed.2011.215 43 13 Hill-Taylor B, Sketris I, Hayden J, et al. Application of the STOPP/START criteria: a systematic 44 review of the prevalence of potentially inappropriate prescribing in older adults, and 45 evidence of clinical, humanistic and economic impact. J Clin Pharm Ther 2013;38:360–72. 46 doi:10.1111/jcpt.12059 47 48 14 Wallace E, McDowell R, Bennett K, et al. Impact of Potentially Inappropriate Prescribing on 49 Adverse Drug Events, Health Related Quality of Life and Emergency Hospital Attendance in 50 Older People Attending General Practice: A Prospective Cohort Study. Journals Gerontol Ser A 51 Biol Sci Med Sci 2017;72:271–7. doi:10.1093/gerona/glw140 52 53 15 Moriarty F, Bennett K, Cahir C, et al. Potentially inappropriate prescribing according to STOPP 54 and START and adverse outcomes in community-dwelling older people: a prospective cohort 55 study. Br J Clin Pharmacol 2016;:[Epub ahead of print]. doi:10.1111/bcp.12995 56 16 Cahir C, Fahey T, Teeling M, et al. Potentially inappropriate prescribing and cost outcomes for 57 58 18 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 32 of 42

1 2 3 older people: a national population study. Br J Clin Pharmacol 2010;69:543–52. 4 doi:10.1111/j.1365-2125.2010.03628.x 5 17 Vezmar Kovačević S, Simišić M, Stojkov Rudinski S, et al. Potentially inappropriate prescribing 6 in older primary care patients. PLoS One 2014;9:e95536. doi:10.1371/journal.pone.0095536 7 8 18 O’Kelly M, Teljeur C, O’Kelly F, et al. Structure of General Practice in Ireland 1982 — 2015. 9 Dublin, Ireland: 2016. 10 19 Sweeney J, Kearney P, Redmond P, et al. Point of Care Morbidity Coding; a Feasibility Study in 11 Confidential: For Review Only 12 Primary Care. Forum J Irish Coll Gen Pract 2017;34:52–4. 13 20 Galvin S, Callan A, Cormican M, et al. Improving antimicrobial prescribing in Irish primary care 14 through electronic data collection and surveillance: a feasibility study. BMC Fam Pract 15 2015;16:77. doi:10.1186/s12875-015-0280-3 16 17 21 von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational 18 Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. 19 PLoS Med 2007;4:e296. doi:10.1371/journal.pmed.0040296 20 22 Brown JD, Hutchison LC, Li C, et al. Predictive Validity of the Beers and Screening Tool of 21 Older Persons’ Potentially Inappropriate Prescriptions (STOPP) Criteria to Detect Adverse 22 Drug Events, Hospitalizations, and Emergency Department Visits in the United States. J Am 23 Geriatr Soc 2016;64:22–30. doi:10.1111/jgs.13884 24 25 23 Lau M, Tenney J. Evaluation of Drug-Disease Interactions and Their Association with 26 Unplanned Hospital Readmission Utilizing STOPP Version 2 Criteria. Geriatrics 2017;2:33. 27 doi:10.3390/geriatrics2040033 28 29 24 Hedna K, Hakkarainen KM, Gyllensten H, et al. Potentially inappropriate prescribing and 30 adverse drug reactions in the elderly: a population-based study. Eur J Clin Pharmacol 31 2015;71:1525–33. doi:10.1007/s00228-015-1950-8 32 25 Walsh KA, O’Riordan D, Kearney PM, et al. Improving the appropriateness of prescribing in 33 older patients: a systematic review and meta-analysis of pharmacists’ interventions in 34 secondary care. Age Ageing 2016;45:201–9. doi:10.1093/ageing/afv190 35 36 26 Clyne B, Fitzgerald C, Quinlan A, et al. Interventions to address potentially inappropriate 37 prescribing in community-dwelling older adults: a systematic review of randomized 38 controlled trials. J Am Geriatr Soc 2016;64:1210–22. doi:10.1111/jgs.14133 39 27 Sinnott S-J, Bennett K, Cahir C. Pharmacoepidemiology resources in Ireland—an introduction 40 41 to pharmacy claims data. Eur J Clin Pharmacol 2017;73:1449–55. doi:10.1007/s00228-017- 42 2310-7 43 28 Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in 44 longitudinal studies: development and validation. J Chronic Dis 1987;40:373–83. 45 46 29 R Core Team. R: A language and environment for statistical computing. R Foundation for 47 Statistical Computing, Vienna, Austria. 2013.http://www.r-project.org/. 48 30 Hadfield JD. MCMC Methods for Multi-Response Generalized Linear Mixed Models: The 49 MCMCglmm R Package. J Stat Softw 2010;33:1–22. doi:10.18637/jss.v033.i02 50 51 31 Huntley AL, Johnson R, Purdy S, et al. Measures of multimorbidity and morbidity burden for 52 use in primary care and community settings: a systematic review and guide. Ann Fam Med 53 2012;10:134–41. doi:10.1370/afm.1363 54 32 D’Agostino RB. Propensity score methods for bias reduction in the comparison of a treatment 55 56 to a non-randomized control group. Stat Med 1998;17:2265–81. 57 58 19 59 60 https://mc.manuscriptcentral.com/bmj Page 33 of 42 BMJ

1 2 3 33 Ho DE, Imai K, King G, et al. Matching as Nonparametric Preprocessing for Reducing Model 4 Dependence in Parametric Causal Inference. Polit Anal 2007;15:199–236. 5 doi:10.1093/pan/mpl013 6 34 Almanasreh E, Moles R, Chen TF. The medication reconciliation process and classification of 7 discrepancies: a systematic review. 2016;82. doi:10.1111/bcp.13017 8 9 35 Laroche M-L, Charmes J-P, Nouaille Y, et al. Impact of hospitalisation in an acute medical 10 geriatric unit on potentially inappropriate medication use. Drugs Aging 2006;23:49–59. 11 Confidential: For Review Only 12 36 Gutiérrez-Valencia M, Izquierdo M, Malafarina V, et al. Impact of hospitalization in an acute 13 geriatric unit on polypharmacy and potentially inappropriate prescriptions: A retrospective 14 study. Geriatr Gerontol Int 2017;17:2354–60. doi:10.1111/ggi.13073 15 37 Rodríguez del Río E, Perdigones J, Fuentes Ferrer M, et al. [Impact of medium-term outcomes 16 of inappropriate prescribing in older patients discharged from a short stay unit]. Atención 17 Primaria 2017;S0212-6567:30277–3. doi:10.1016/j.aprim.2017.03.018 18 19 38 McMahon CG, Cahir CA, Kenny RA, et al. Inappropriate prescribing in older fallers presenting 20 to an Irish emergency department. Age Ageing 2014;43:44–50. doi:10.1093/ageing/aft114 21 39 Blanco-Reina E, García-Merino MR, Ocaña-Riola R, et al. Assessing Potentially Inappropriate 22 Prescribing in Community-Dwelling Older Patients Using the Updated Version of STOPP- 23 START Criteria: A Comparison of Profiles and Prevalences with Respect to the Original 24 25 Version. PLoS One 2016;11:e0167586. doi:10.1371/journal.pone.0167586 26 40 Wauters M, Elseviers M, Vaes B, et al. Too many, too few, or too unsafe? Impact of 27 inappropriate prescribing on mortality, and hospitalization in a cohort of community-dwelling 28 oldest old. Br J Clin Pharmacol 2016;82:1382–92. doi:10.1111/bcp.13055 29 30 41 Pardo-Cabello A, Manzano-Gamero V, Zamora-Pasadas M, et al. Potentially inappropriate 31 prescribing according to STOPP-2 criteria among patients discharged from Internal Medicine: 32 prevalence, involved drugs and economic cost. Arch Gerontol Geriatr 2017;74:150–4. 33 doi:10.1016/J.ARCHGER.2017.10.009 34 42 Kimura T, Ogura F, Yamamoto K, et al. Potentially inappropriate medications in elderly 35 Japanese patients: effects of pharmacists’ assessment and intervention based on Screening 36 Tool of Older Persons’ Potentially Inappropriate Prescriptions criteria ver.2. J Clin Pharm Ther 37 2017;42:209–14. doi:10.1111/jcpt.12496 38 39 43 Van Craen K, Braes T, Wellens N, et al. The Effectiveness of Inpatient Geriatric Evaluation and 40 Management Units: A Systematic Review and Meta-Analysis. J Am Geriatr Soc 2010;58:83–92. 41 doi:10.1111/j.1532-5415.2009.02621.x 42 43 44 Grimes T, Duggan C, Delaney T. Pharmacy services at admission and discharge in adult, acute, 44 public hospitals in Ireland. Int J Pharm Pract 2010;18:346–52. doi:10.1111/j.2042- 45 7174.2010.00064.x 46 45 The National Patient Experience Survey Findings of the 2017 inpatient survey. 47 48 46 Moore C, Wisnivesky J, Williams S, et al. Medical errors related to discontinuity of care from 49 an inpatient to an outpatient setting. J Gen Intern Med 2003;18:646–51. doi:10.1046/J.1525- 50 1497.2003.20722.X 51 47 Kripalani S, LeFevre F, Phillips CO, et al. Deficits in Communication and Information Transfer 52 Between Hospital-Based and Primary Care Physicians. JAMA 2007;297:831. 53 54 doi:10.1001/jama.297.8.831 55 48 Witherington EMA, Pirzada OM, Avery AJ. Communication gaps and readmissions to hospital 56 for patients aged 75 years and older: observational study. Qual Saf Heal Care 2008;17:71–5. 57 58 20 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 34 of 42

1 2 3 doi:10.1136/qshc.2006.020842 4 49 Coleman EA. Falling through the cracks: challenges and opportunities for improving 5 transitional care for persons with continuous complex care needs. J Am Geriatr Soc 6 2003;51:549–55. 7 8 50 Goodwin N, Dixon A, Poole T, et al. Improving quality of care in General practice: Report of an 9 independent inquiry commissioned by The King’s Fund. London, UK: 2011. 10 51 Le Berre M, Maimon G, Sourial N, et al. Impact of Transitional Care Services for Chronically Ill 11 Confidential: For Review Only 12 Older Patients: A Systematic Evidence Review. J Am Geriatr Soc 2017;65:1597–608. 13 doi:10.1111/jgs.14828 14 52 Pellegrin KL, Krenk L, Oakes SJ, et al. Reductions in Medication-Related Hospitalizations in 15 Older Adults with Medication Management by Hospital and Community Pharmacists: A 16 Quasi-Experimental Study. J Am Geriatr Soc 2017;65:212–9. doi:10.1111/jgs.14518 17 18 53 Bowden T, Coiera E. The role and benefits of accessing primary care patient records during 19 unscheduled care: a systematic review. BMC Med Inform Decis Mak 2017;17:138. 20 doi:10.1186/s12911-017-0523-4 21 54 Maurice AP, Chan S, W Pollard C, et al. Improving the quality of hospital discharge summaries 22 utilising an electronic prompting system. BMJ Qual Improv Reports 2014;3:u200548.w2201. 23 doi:10.1136/bmjquality.u200548.w2201 24 25 55 Dreischulte T, Donnan P, Grant A, et al. Safer Prescribing — A Trial of Education, Informatics, 26 and Financial Incentives. N Engl J Med 2016;374:1053–64. doi:10.1056/NEJMsa1508955 27 28 56 Stocks S, Kontopantelis E, Akbarov A, et al. Examining variations in prescribing safety in UK 29 general practice : Cross sectional study using the Clinical Practice Research Datalink. BMJ 30 2015;351:h5501. doi:10.1136/bmj.h5501 31 57 Anderson K, Stowasser D, Freeman C, et al. Prescriber barriers and enablers to minimising 32 potentially inappropriate medications in adults: a systematic review and thematic synthesis. 33 BMJ Open 2014;4:e006544. doi:10.1136/bmjopen-2014-006544 34 35 58 Reeve E, To J, Hendrix I, et al. Patient barriers to and enablers of deprescribing: a systematic 36 review. Drugs Aging 2013;30:793–807. doi:10.1007/s40266-013-0106-8 37 59 Farrell B, Pottie K, Thompson W, et al. Deprescribing proton pump inhibitors: Evidence-based 38 clinical practice guideline. Can Fam Physician 2017;63:354–64. 39 40 60 Pottie K, Thompson W, Davies S, et al. Deprescribing benzodiazepine receptor agonists: an 41 evidence-based clinical practice guideline. Can Fam Physician 2018;64. 42 43 61 Redmond P, Grimes T, McDonnell R, et al. Tackling transitions in patient care: the process of 44 medication reconciliation. Fam Pract 2013;30:483–4. doi:10.1093/fampra/cmt051 45 62 O’Connor MN, O’Sullivan D, Gallagher PF, et al. Prevention of Hospital-Acquired Adverse Drug 46 Reactions in Older People Using Screening Tool of Older Persons’ Prescriptions and Screening 47 Tool to Alert to Right Treatment Criteria: A Cluster Randomized Controlled Trial. J Am Geriatr 48 Soc 2016;64:1558–66. doi:10.1111/jgs.14312 49 50 63 Grimes TC, Deasy E, Allen A, et al. Collaborative pharmaceutical care in an Irish hospital: 51 uncontrolled before-after study. BMJ Qual Saf 2014;23:574–83. doi:10.1136/bmjqs-2013- 52 002188 53 54 55 56 57 58 21 59 60 https://mc.manuscriptcentral.com/bmj Page 35 of 42 BMJ

1 2 3 Table 1. Demographics and main clinical characteristics by year. 4 5 Demographic 2012 2013 2014 2015 Missing 6 clinical (n=30753) (n=30789) (n=30231) (n=29077) data 7 characteristics (%) 8 Age mean (SD), 76.8 (8.2) 76.4 (8.1) 75.9 (7.8) 75.0 (7.6) 0.08 9 years 10 Male n (%) 13212 (43.0) 13335 (43.3) 13176 (43.5) 12687 (43.6) 0.08 11 Confidential: For Review Only 12 Number of patients 13 with hospitalisation 4151 (13.5) 4496 (14.6) 4537 (15.0) 3015 (10.4) 0 14 n (%) 15 Health Cover n (%) 0.03 16 17 General Medical 21053 (68.4) 21472 (69.7) 21202 (70.1) 20859 (71.7) 18 Services scheme 19 Doctor Visit Card 3029 (9.8) 3153 (10.2) 3201 (10.5) 3280 (11.2) 20 21 Private patients 6518 (21.2) 6004 (19.5) 5705 (18.8) 4817 (16.5) 22 23 Other 153 (0.5) 160 (0.5) 123 (0.4) 71 (0.2) 24 Number of 25 prescription items 22 (9-42) 22 (9-43) 23 (10-44) 21 (9-40) 0 26 per patient median 27 (IQR) 28 29 CCI mean (SD) 0.89 (1.23) 0.94 (1.27) 1 (1.31) 1 (1.31) 24.2 30 Prevalence of PIPs 0

31 n (%) 32 1 PIP 6452 (21%) 6843 (22.2%) 6771 (22.4%) 6857 (23.5%) 33 34 2 PIPs 4171 (13.5%) 4254 (13.8%) 4429 (14.6%) 4220 (14.5%) 35 36 ≥3 PIPs 3317 (10.7%) 3654 (11.8%) 3762 (12.4%) 3746 (12.8%) 37 38 SD: standard deviation; IQR: interquartile range; CCI: Charlson comorbidity index 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 22 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 36 of 42

1 2 3 4 5 6 7 8 9 10 11 Confidential: For Review Only 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 23 59 60 https://mc.manuscriptcentral.com/bmj Page 37 of 42 BMJ

1 2 3 Table 2. Comparison of models adjusted for morbidity using Charlson Comorbidity Index (standard 4 analysis) and RxRisk (sensitivity analysis) 5 6 7 8 n Rate Ratio (95% CI) a Odds ratio (95% CI) a 9 Estimate for hospitalised 10 11 (relativeConfidential: to not hospitalised) For Review Only 12 Adjusted for CCI 28,831 1.27 (1.23 to 1.30) 1.49 (1.42 to 1.59) 13 Adjusted for RxRisk 38,169 1.28 (1.24 to 1.31) 1.55 (1.47 to 1.64) 14 Estimate for post-admission

15 (relative to pre-admission) 16 Adjusted for CCI 9,549 - 1.72 (1.63 to 1.84) 17 18 Adjusted for RxRisk 11,277 - 1.71 (1.63 to 1.81) a 19 Additionally adjusted for age, sex, number of prescriptions items, and health cover type. CI = 20 credible interval 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 24 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 38 of 42

Figure 1. Estimated rate ratios for the number of distinct PIP among all participants with 95% credible intervals 1 2 The reference groups used are no hospital admission, female gender, and private patient (health cover type) 3 4 Figure 2. Estimated odds ratios for the presence of PIP among hospitalised participants only, wih 95% credible 5 intervals. 6 7 The reference groups are before hospitalisation and female gender. Also adjusted for patient health cover type, which did not 8 show any significant association 9 10 Figure 3. Estimated rate ratios for the number of distinct PIP among propensity-score matched participants, with 11 Confidential: For Review Only 12 95% credible intervals. 13 The reference groups used are no hospital admission, female gender, and private patient (health cover type) 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 25 59 60 https://mc.manuscriptcentral.com/bmj Page 39 of 42 BMJ

Appendix 1 STOPP version 2 criteria that could not be applied 1 2 3 Appendix 2 Description of the 45 STOPP 2 criteria applied 4 5 Appendix 3. Prevalence of PIP by year according to the STOPP Version 2 criteria. 6 7 Appendix 4. Estimated odds ratios for the presence of PIP among all participants, with 95% credible intervals. 8 9 The reference groups used are no hospital admission, female gender, and private patient (health cover type) 10 11 Appendix 5.Confidential: Sensitivity analyses fitting models by calendarFor year Review Only 12 13 Appendix 6. Propensity score-matched sensitivity analysis to estimate odds ratios for the presence of PIP, with 14 15 95% credible intervals. 16 The reference groups used are no hospital admission, female gender, and private patient (health cover type) 17

18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 26 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 40 of 42

1 2 3 4 5 6 7 8 9 10 11 Confidential: For Review Only 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 134x124mm (300 x 300 DPI) 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj Page 41 of 42 BMJ

1 2 3 4 5 6 7 8 9 10 11 Confidential: For Review Only 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 134x124mm (300 x 300 DPI) 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj BMJ Page 42 of 42

1 2 3 4 5 6 7 8 9 10 11 Confidential: For Review Only 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 134x124mm (300 x 300 DPI) 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 https://mc.manuscriptcentral.com/bmj