Area Deprivation and Regional Disparities in Treatment And

Diabetes Care Volume 41, December 2018 2517

Marie Auzanneau,1,2 Stefanie Lanzinger,1,2 Area Deprivation and Regional Barbara Bohn,1,2 Peter Kroschwald,3 Ursula Kuhnle-Krahl,4 Disparities in Treatment and Paul Martin Holterhus,5 Kerstin Placzek,6 Johannes Hamann,7 Rainer Bachran,8 Outcome Quality of 29,284 Joachim Rosenbauer,2,9 and Werner Maier,2,10 on behalf of the Pediatric Patients With Type 1 DPV Initiative Diabetes in Germany: A Cross- sectional Multicenter DPV PDMOOYHAT EVCSRESEARCH SERVICES EPIDEMIOLOGY/HEALTH Analysis Diabetes Care 2018;41:2517–2525 | https://doi.org/10.2337/dc18-0724

1Institute of Epidemiology and Medical Biome- try, ZIBMT, University of Ulm, Ulm, Germany 2German Center for Diabetes Research (DZD), Neuherberg, Germany 3Children’s Hospital, Ruppiner Kliniken GmbH, OBJECTIVE Hochschulklinikum der Medizinischen Hochschule This study analyzed whether area deprivation is associated with disparities in health Brandenburg, Neuruppin, Germany 4 care of pediatric type 1 diabetes in Germany. Practice for Pediatric Endocrinology and Dia- betology, Gauting, Germany 5 RESEARCH DESIGN AND METHODS Division of Pediatric Endocrinology and Diabe- tes, Department of Pediatrics, University Hospital We selected patients <20 years of age with type 1 diabetes and German residence of Schleswig-Holstein, Campus Kiel/Christian- documented in the “diabetes patient follow-up” (Diabetes-Patienten-Verlaufsdo- Albrechts University of Kiel, Kiel, Germany 6 kumentation [DPV]) registry for 2015/2016. Area deprivation was assessed by Pediatric and Adolescent Medicine, University Hospital, Martin-Luther University, Halle, Germany quintiles of the German Index of MultipleDeprivation (GIMD 2010) at the district level 7Department of Pediatrics, St. Marien Hospital and was assigned to patients. To investigate associations between GIMD 2010 and Landshut, Landshut, Germany indicators of diabetes care, we used multivariable regression models (linear, 8Pediatric Practice, Oberhausen, Germany 9 logistic, and Poisson) adjusting for sex, age, migration background, diabe- Institute for Biometrics and Epidemiology, Ger- man Diabetes Center, Leibniz Center for Diabe- tes duration, and German federal state. tes Research at Heinrich Heine University, Dusseldorf,¨ Germany RESULTS 10Institute of Health Economics and Health Care We analyzed data from 29,284 patients. From the least to the most deprived Management, Helmholtz Zentrum Munchen,¨ quintile, use of continuous glucose monitoring systems (CGMS) decreased from German Research Center for Environmental Health (GmbH), Neuherberg, Germany 6.3 to 3.4% and use of long-acting insulin analogs from 80.8 to 64.3%, whereas Corresponding author: Marie Auzanneau, marie. use of rapid-acting insulin analogs increased from 74.7 to 79.0%; average HbA1c [email protected]. increasedfrom7.84to8.07%(62to65mmol/mol),andtheprevalenceofoverweight Received 3 April 2018 and accepted 12 Septem- from 11.8 to 15.5%, but the rate of severe hypoglycemia decreased from 12.1 to ber 2018. 6.9 events/100 patient-years. Associations with other parameters showed a more This article contains Supplementary Data online complex pattern (use of continuous subcutaneous insulin infusion [CSII]) or were at http://care.diabetesjournals.org/lookup/suppl/ not signiﬁcant. doi:10.2337/dc18-0724/-/DC1. J.R. and W.M. share last authorship. CONCLUSIONS © 2018 by the American Diabetes Association. Area deprivation was associated not only with key outcomes in pediatric type 1 Readers may use this article as long as the work diabetes but also with treatment modalities. Our results show, in particular, that is properly cited, the use is educational and not for proﬁt, and the work is not altered. More infor- the access to CGMS and CSII could be improved in the most deprived regions in mation is available at http://www.diabetesjournals Germany. .org/content/license. 2518 Area Deprivation and Type 1 Diabetes Diabetes Care Volume 41, December 2018

Despite considerable advances in the statutory health insurance (covering ;90% and German residence documented in management of pediatric type 1 diabetes of the population) if poor glycemic con- the DPV for 2015 and 2016. The definition over the last two decades, major geo- trol persists despite intensified conven- of type 1 diabetes in the DPV database is graphic variations in metabolic control tional insulin treatment (14). Patients based on a physician’s diagnosis accord- and diabetes-related complications have and diabetologists have to apply to ing to the international guidelines (17) persisted between countries around the health insurance company for reim- and can be revised based on the course the world (1). Treatment and outcome bursement by providing comprehensive of the disease. For each patient, we quality of patients with type 1 diabetes documentation of the blood glucose aggregated clinical data for the years also vary within countries. In Brazil, levels and insulin therapy over the last 2015 and 2016 as median, percentage, large discrepancies were found in clinical 3 months. Exigent documentation and or rate per 1 or 100 patient-years (PYs) care across different regions (2). In Ger- uncertainty of reimbursement may dis- for continuous, categorical, and event many, significant disparities in the use of courage some families in socioeconom- variables, respectively. insulin pumps and rapid-acting or long- ically disadvantaged areas. Application Area Deprivation acting analogs, HbA levels, the preva- for reimbursement of real-time CGM by 1c Area deprivation was assessed using the lence of overweight, and the rate of statutory health insurance is also neces- German Index of Multiple Deprivation severe hypoglycemia have been reported sary and only possible since 2016. For from 2010 (GIMD 2010). This index was between the federal states (3). However, patients covered by private health in- developed by Maier et al. (18,19) and is a regional variations in treatment and surance (;10% of the population), re- validated measure of area deprivation for outcome quality of care of patients imbursement depends on specifications Germany (5,8,19,20). The GIMD includes with type 1 diabetes are not completely in the insurance contract. Different seven domains of deprivation with dif- explained. proportions of patients with private ferent weighting: income (25%), employ- Relative material and social depriva- versus statutory health insurance be- ment (25%), education (15%), municipal/ tion (i.e., the lack of resources for people tweenareasinGermanycouldalso district revenue (15%), social capital compared with the societies to which lead to regional variation in diabetes (10%), environment (5%), and security they belong) show significant area-level treatment (15). (5%) (18,19). The GIMD 2010 was gen- disparities associated with health (4). The objective of our study was there- erated for all 412 districts of Germany Therefore, indices of multiple depriva- fore to analyze whether area deprivation (boundaries at 31 December 2010). Dis- tion have been used increasingly since is associated with regional disparities in tricts were categorized into deprivation 2000 to assess area deprivation, notonly the treatment and outcome quality of quintiles, with quintile 1 (Q1) represent- for epidemiological research but also for pediatric patients with type 1 diabetes in ing the least deprived and quintile 5 (Q5) public health policy (5). According to Germany. the most deprived districts. We used the Noble et al. (6), area deprivation refers five-digit postal code of the patient’s not merely to the proportion of deprived RESEARCH DESIGN AND METHODS residence to assign the district of resi- people in an area but also to an “area Study Population dence. The postal code of residence was effect” and to the negative consequences We used data from the multicenter not available for 2.6% of the patients (n = of “the lack of facilities in that area.” “diabetes patient follow-up” registry 766), so we used the postal code of the Correspondingly, indices of multiple dep- (Diabetes-Patienten-Verlaufsdokumentation treating diabetes center as proxy. rivation provide multidimensional in- [DPV]). Currently, 459 diabetes care formation on living conditions at the centers, mainly in Germany (n = 416) and Indicators of Diabetes Care regional level. Austria (n = 40), participate in the DPV Indicators of medical treatment in our Concerning type 2 diabetes, a notable initiative and prospectively document analysis were use of insulin pump ther- number of studies have shown that area demographic and clinical data on treat- apy (CSII), use of CGMS, frequency of self- deprivation is associated with worse ment and outcome quality. Twice a year, monitoring of blood glucose (SMBG), use indicators of outcome quality, such as centers transmit locally collected and of rapid-acting insulin analogs and use of BMI, HbA1c, lipid profile, and short-term anonymized data to the University of long-acting insulin analogs in patients on or long-term diabetes-related complica- Ulm, Germany, for central analysis and injection therapy, and participation in tions (7,8). However, evidence is weaker quality assurance (16). Inconsistent or diabetes education programs. CGMS with regard to type 1 diabetes (9–13). implausible data are reported back to includes real-time CGM and CGM with Moreover, to date, studies on type 1 centers for verification or correction. intermittent scanning, also called “flash diabetes focused on associations be- Data collection and analysis of anony- glucose monitoring.” Diabetes educa- tween area deprivation and metabolic mized data from the DPV registry were tion was documented if a teaching ses- control but not medical treatment approved by the Medical Faculty Ethics sion lasted for at least 45 min and if the (9–13). Committee of the University of Ulm, patient and/or members of his or her Nevertheless, regional socioeconomic Germany, and by the local review boards family or other caregivers participated disparities may be a major determinant of participating centers. (21). of the use of insulin pump therapy (con- As of March 2017, 484,365 patients Indicators of outcome quality were tinuous subcutaneous insulin infusion with any type of diabetes were docu- BMI, presence of overweight or obesity, [CSII]), continuous glucose monitoring mented in the DPV database. We in- HbA1c, rates of severe hypoglycemia systems (CGMS), and insulin analogs. cluded only patients younger than (with or without coma) and of severe In Germany, CSII is reimbursed by the 20 years of age with type 1 diabetes hypoglycemia with coma, rates of care.diabetesjournals.org Auzanneau and Associates 2519

diabetic ketoacidosis (DKA) and of severe with district as the categorical indepen- the study population stratified by GIMD DKA, and number of hospital days per dent variable, adjusting for sex, age 2010 quintiles are given in Table 1. person and year (/PY). BMI values, ex- group (,6 years, 6 to ,12 years, 12 Results of regression models for CSII, pressed as weight in kilograms/squared to ,20 years), migration background HbA1c, prevalence of overweight, rate height in meters (kg/m2), were trans- (defined as at least one parent or the of severe hypoglycemia, and rate of formed to a BMI SD score (BMI SDS) using child born outside Germany), and diabe- DKA are illustrated graphically (Fig. 2); national reference data from the German tes duration (,2 years, $2 years). Ad- results for other outcomes are pre- Health Interview and Examination Survey justed mean estimates for districts were sented in Table 2. for Children and Adolescents (KiGGS) then categorized into outcome quintiles. (22). A BMI above the 90th or 97th To investigate the association be- Medical Treatment Visual comparison of the regional distri- percentile of this reference population tween the GIMD 2010 quintiles and butions of CSII and GIMD 2010 (Fig. 1) was defined as overweight (including indicators of diabetes care, we per- indicated that CSII was used less fre- obesity) or obesity, respectively (22), formed multivariable regression models quently in the least deprived districts. according to the German national guide- (linear, logistic, or Poisson considering Regression analyses with and without line (Arbeitsgemeinschaft Adipositas im overdispersion) with GIMD 2010 quintiles adjusting for federal state confirmed Kindes- und Jugendalter [AGA]) (23) and as the categorical independent variable this impression (CSII use was 41.7% in the European Childhood Obesity Group and adjusting for sex, age group, migra- Q1 and 42.4–48.0% in other quintiles, in (ECOG) guideline (24). HbA was stan- tion background, and diabetes duration. 1c the model adjusting for federal state), dardized to the Diabetes Control and In a second step, we also adjusted for but showed further that use of CSII de- Complications Trial (DCCT) reference of German federal state in regression models creased from Q2 to Q5 (Fig. 2A). Re- 4.05–6.05% (21–43 mmol/mol), applying to investigate whether the effects of area gression analyses, with and without the “multiple-of-the-mean” transformation deprivation were independent of the adjusting for federal state, showed method to adjust for differences be- federal structure of Germany. All analy- that CGMS was used less frequently in tween local laboratories (25). Severe ses were repeated stratified by sex to districts with higher deprivation (3.4% in hypoglycemia (with or without coma) examine possible differences in the as- Q5 vs. 6.3% in Q1 in the model adjusting was defined as self-reported uncon- sociations of GIMD 2010 with indicators for federal state) (Table 2). Rapid-acting sciousness, convulsion, or being unable of care between girls and boys. insulin analogs among patients on in- to take glucose without third-party as- The number of cases used in the jection therapy tended to be used more sistance(26)or, inpreschoolchildren,as analysis of each variable is indicated in frequently with increasing area depriva- an altered mental status and an inability the tables and figures. Results of regres- tion according to the model not consid- to assist in hypoglycemia treatment sion analyses are presented as ad- ering federal states. However, differences (27). DKA was defined as pH ,7.3 and/ justed mean estimates (least square between deprivation quintiles became or requirement of hospital treatment; se- means) with 95% CIs. P values were smaller after adjusting for federal state vere DKA was defined as pH ,7.1. DKA at adjusted for multiple testing using the (79.0% in Q5 vs. 74.7% in Q1). In the model diabetes onset was not considered in this false discovery rate (FDR)-controlling without federal states, the pattern of analysis. Benjamini-Hochberg procedure (30). association between long-acting insulin The level of significance of two-sided analogs and area deprivation appeared Statistical Analysis tests was set at P , 0.01. Statistical to be more complex (highest use in Q1 We present descriptive data as me- analysis was performed using SAS 9.4 and Q5, lowest use in Q2 and Q3). After dian (lower–upper quartile), percentage, software (SAS Institute, Cary, NC). Cho- adjustment for federal state, long- or rate per 1 or 100 PYs for continu- ropleth maps were created using QGIS actinginsulinanalogstendedtobe ous, categorical, and event variables, 2.14 open source software. used less frequently with increasing respectively. area deprivation (64.3% in Q5 vs. To illustrate the regional distribution RESULTS 80.8% in Q1 and Q3). In all models, of CSII, HbA , prevalence of overweight, The study population comprised 29,284 1c associations with frequency of SMBG rate of severe hypoglycemia, and rate children and adolescents with type 1 were not significant. With increasing of DKA at district level in Germany, we diabetes (selection presented in Supple- area deprivation, patients and their created quintile-based choropleth maps mentary Fig. 1). Of all subjects included, families participated more often in di- (Fig. 1B–F). Choropleth maps display 45.6% used CSII, 6.3% used CGMS, abetes education programs, but these areas that are shaded or patterned in and 46.8% had participated in a di- associations were no longer significant relation to the level of the variable of abetes education program. Median after additional adjustment for federal interest. They are frequently used to HbA was 7.62% (60 mmol/mol; inter- 1c state. visualize the geographical distribution quartile range 6.94–8.50% [52–69 mmol/ of health outcomes (28) and also in mol]). The rate was 10.2 events/100 Outcome Quality the field of diabetes research (29). For PYs for severe hypoglycemia and 1.8 Visual comparison of the regional distri- this purpose, we derived district-specific events/100 PYs for DKA. Data showed butions of HbA1c and GIMD 2010 (Fig. 1) adjusted mean estimates (least square that 13.4% of the patients were over- indicated that HbA1c was higher in means) for each of these outcomes from weight (including obesity) and 3.5% the most deprived districts. Regression multivariable regression models (linear, lo- were obese. The number of hospital analyses with and without adjusting gistic, or Poisson considering overdispersion) days was 4.9/PY. Demographic data of for federal state confirmed this finding. 2520 Area Deprivation and Type 1 Diabetes Diabetes Care Volume 41, December 2018

Figure 1—Quintile-based distribution of the GIMD 2010 (A) and of selected indicators of type 1 diabetes care at district level (B–F). B–F: Adjusted mean estimates (least square means) from regression models (linear, logistic, and Poisson), adjusting for sex, age-group, migration, and diabetes duration, with district as the categorical independent variable, categorized into outcome quintiles.

Average HbA1c increased almost linearly after adjusting for federal state) (Fig. 2B). area deprivation (from 12.1 events/ from the least to the most deprived In contrast to HbA1c, the rate of severe 100 PYs to 6.9 events/100 PYs in the districts (from 7.84% [62 mmol/mol] hypoglycemia (with or without coma) model adjusted for federal state) in Q1 to 8.07% [65 mmol/mol] in Q5, decreased in all models with higher (Fig. 2C), whereas the rate of severe care.diabetesjournals.org Auzanneau and Associates 2521

Table 1—Characteristics of the study population by GIMD 2010 quintiles All patients Q1 Q2 Q3 Q4 Q5 n = 29,284 n = 7,109 n = 7,541 n = 5,353 n = 5,804 n = 3,477 Girls, % 47.2 46.7 48.1 48.2 46.2 46.6 Age, years* 13.4 (9.8–16.2) 13.5 (9.9–16.3) 13.4 (9.9–16.2) 13.3 (9.8–16.2) 13.3 (9.7–16.2) 13.1 (9.7–16.0) Age at onset, years* 7.7 (4.4–11.1) 7.8 (4.4–11.2) 7.6 (4.4–11.1) 7.8 (4.4–11.1) 7.6 (4.4–11.1) 7.7 (4.5–11.1) Diabetes duration, years* 4.0 (1.3–7.5) 4.0 (1.4–7.5) 4.1 (1.4–7.6) 4.0 (1.3–7.5) 3.9 (1.2–7.5) 3.7 (1.2–7.3) Migration background, % 21.6 21.1 23.7 22.5 23.9 13.3 East German residence (new federal states), % 15.9 0.0 0.4 3.1 30.5 77.3 Unadjusted data. *Data are median (lower–upper quartile). hypoglycemia with coma did not vary certain medical criteria have been met for federal state. Here, many factors significantly with area deprivation level (leading to approval by the health in- may interact in a complex manner. Pos- (Table 2). Positive associations between surance company), for instance, if in- sible explanations include differences in area deprivation and DKA (Fig. 2D)or tensified conventional insulin therapy patients’ health insurance (private vs. severe DKA (pH ,7.1) (Table 2) were not is not sufficient to achieve goals for statutory) or regionally different local significant. The prevalence of overweight glycemic control (14). We found the discount agreements with pharmaceu- (including obesity) increased steadily lowest HbA1c levels in the least deprived tical companies (15). with area deprivation, and this associa- districts (Q1) where pump use was also With regard to indicators of outcome tion was stronger when additionally ad- less frequent. It is possible that HbA1c quality, our results concerning the asso- justingforfederal state(from11.8%inQ1 goals in these districts (Q1) are more ciation between area deprivation and to 15.5% in Q5) (Fig. 2E). The pattern of often achieved with intensified conven- HbA1c are in line with the findings association was similar for BMI SDS (Table tional insulin therapy compared with from previous studies. Several reports 2). The increase in obesity prevalence was more deprived districts, so that medical on patients with type 1 diabetes have not significant. The number of hospital criteria for reimbursement of CSII are less shown significant associations between days (rate/PY) increased with higher area frequently met. Further, in districts in higher area deprivation and poorer met- deprivation in the model not adjusting deprivation quintiles Q2 to Q5, CSII was abolic control in children (9) and adults for federal state, but this association was used less frequently with increasing area (11). no longer significant after controlling for deprivation. This pattern may be associ- We also found a positive association federal state (Table 2). ated with the uncertainty of reimburse- between area deprivation and over- ment of the insulin pump, which may weight or BMI SDS, and these findings Analysis by Sex constitute an obstacle for some families are also consistent with previous reports Considering the model adjusting for in more deprived regions. Associations in the general population (8,31). For federal state, stratified by sex, the re- between socioeconomic factors and example, significant associations be- sults were similar in boys and girls ex- the use of CSII have been rarely investi- tween area deprivation and obesity cept for a slightly but significantly less gated. However, some studies have have been reported in adults in Germany, frequent SMBG only in boys in Q5 com- indicated that individuals in higher so- after controlling for education (8). A pared with other deprivation quintiles cioeconomic groups injected insulin more strong association between area depri- (Supplementary Table 2). frequently and were also more likely to vation and weight status was also con- use insulin pumps (13). firmed in British children: children living CONCLUSIONS We found that CGMS was used less in more deprived locations had both We found that area deprivation was in more deprived districts. Associations greater waist circumference and greater associated with the use of CSII, CGMS, between area deprivation or individual body mass, even after controlling for rapid-acting or long-acting insulin ana- socioeconomic status (SES) and CGMS confounders (age, sex, stature, hip cir- logs, HbA1c levels, the rate of severe have not been investigated yet. Since cumference) (31). hypoglycemia, BMI SDS, and the preva- June 2016 only, real-time CGM but not In contrast to previous reports (32), we lence of overweight, independently of intermittent scanning CGM has been found a negative association between the federal states. Associations of other reimbursed by statutory health insur- area deprivation and the rate of severe factors with area deprivation were not ance in Germany. Absence of reim- hypoglycemia (with or without coma). significant regardless of the model con- bursement until this date may have Recent studies have demonstrated that sidered or no longer significant after led to avoidance of CGMS use, particu- the evidence for an association between adjustment for federal state. larly in more deprived regions. low HbA1c and hypoglycemia risk in Our analysis showed a significantly Use of rapid-acting insulin analogs was type 1 diabetes no longer exists (33). less frequent use of CSII in the least positively associated with area depriva- However, we cannot exclude the possi- deprived districts (Q1) compared with tion, whereas long-acting insulin analogs bility that in our setting, the lower rate of others (Q2–Q5). In Germany, CSII is re- were used less frequently with increas- severe hypoglycemia in the most de- imbursed on a case-by-case basis, if ing area deprivation, after adjustment prived districts is associated with higher 2522 Area Deprivation and Type 1 Diabetes Diabetes Care Volume 41, December 2018

Figure 2—Multiple adjusted mean estimates of indicators of type 1 diabetes care by GIMD 2010 quintiles: CSII (A), HbA1c (B), severe hypoglycemia (C), DKA (D), and overweight (E). Model 1 (triangles): Adjusted mean estimates (least square means) from regression models (linear, logistic, and Poisson), with GIMD 2010 quintiles as the categorical independent variable, adjusting for sex, age-group, migration, and diabetes duration. Model 2 (circles): Adjusted mean estimates (least square means) from regression models (linear, logistic, and Poisson), with GIMD 2010 quintiles as the categorical independentvariable,adjustingforsex,agegroup,migration,diabetesduration,andfederalstate.*P valueswereadjustedformultipletestingusingthe FDR-controlling Benjamini-Hochberg procedure (30).

HbA1c, which is related to higher area with type 1 diabetes living in more de- the medical relevance or social desirabil- deprivation in our study. Another hy- prived areas tend to underreport se- ity bias) compared with parents of chil- pothesis could be that parents of children vere hypoglycemia (minimization of dren living in less deprived districts. In care.diabetesjournals.org Auzanneau and Associates 2523

Table 2—Multiple adjusted mean estimates (95% CI) of indicators of type 1 diabetes care by GIMD 2010 quintiles Outcome n Model Q1 Q2 Q3 Q4 Q5 P value* Treatment CGMS, % 29,284 1 7.3 (6.7–7.9) 5.6 (5.2–6.2) 5.6 (5.1–6.3) 4.8 (4.3–5.4) 4.5 (3.9–5.2) ,0.001 2 6.3 (5.7–7.0) 5.6 (5.1–6.2) 5.7 (5.1–6.4) 5.3 (4.7–6.0) 3.4 (2.7–4.3) 0.002 Rapid-acting insulin analogs, % 15,719** 1 66.8 (65.3–68.3) 70.4 (68.8–71.9) 66.7 (64.8–68.5) 78.0 (76.5–79.5) 87.8 (86.2–89.2) ,0.001 2 74.7 (73.1–76.2) 75.9 (74.3–77.4) 70.9 (68.9–72.7) 76.7 (74.9–78.3) 79.0 (75.8–81.8) ,0.001 Long-acting insulin analogs, % 15,719** 1 77.8 (76.5–79.2) 71.5 (69.9–73.0) 75.2 (73.4–76.8) 72.5 (70.8–74.1) 81.2 (79.4–82.9) ,0.001 2 80.8 (79.4–82.2) 77.3 (75.8–78.8) 80.8 (79.3–82.3) 72.4 (70.5–74.3) 64.3 (60.4–68.0) ,0.001 SMBG, times/day 27,335 1 5.8 (5.7–5.8) 5.7 (5.7–5.8) 5.8 (5.7–5.8) 5.7 (5.7–5.8) 5.6 (5.6–5.7) 0.02 2 5.7 (5.7–5.8) 5.7 (5.7–5.8) 5.7 (5.7–5.8) 5.8 (5.8–5.9) 5.7 (5.6–5.8) 0.03 Diabetes education program, % 29,284 1 44.2 (43.0–45.4) 46.8 (45.7–48.0) 46.1 (44.8–47.5) 47.7 (46.4–49.0) 51.7 (50.0–53.5) ,0.001 2 46.0 (44.6–47.4) 48.2 (47.0–49.5) 46.6 (45.1–48.1) 46.6 (45.1–48.1) 46.0 (43.4–48.7) 0.18 Outcome quality Severe hypoglycemia with coma, 29,284 1 1.8 (1.5–2.2) 2.1 (1.8–2.5) 2.5 (2.1–3.0) 2.0 (1.7–2.4) 1.6 (1.3–2.2) 0.06 events/100 PYs 2 1.9 (1.6–2.3) 1.9 (1.6–2.3) 2.2 (1.8–2.7) 1.9 (1.5–2.3) 1.8 (1.2–2.6) 0.76 Severe DKA (pH ,7.1), 28,965 1 0.2 (0.1–0.3) 0.2 (0.1–0.3) 0.3 (0.2–0.4) 0.2 (0.2–0.4) 0.4 (0.3–0.7) 0.04 events/100 PYs 2 0.2 (0.1–0.3) 0.1 (0.1–0.3) 0.2 (0.1–0.5) 0.2 (0.1–0.5) 0.3 (0.1–0.8) 0.48 BMI SDS 28,327 1 0.28 (0.26–0.30) 0.33 (0.31–0.35) 0.35 (0.33–0.37) 0.33 (0.31–0.35) 0.36 (0.33–0.39) ,0.001 2 0.26 (0.24–0.29) 0.29 (0.27–0.32) 0.33 (0.31–0.36) 0.35 (0.33–0.38) 0.46 (0.41–0.50) ,0.001 Obesity, % 28,327 1 3.2 (2.8–3.6) 3.0 (2.6–3.4) 3.7 (3.2–4.2) 3.6 (3.2–4.2) 3.8 (3.2–4.5) 0.07 2 3.2 (2.8–3.7) 2.8 (2.5–3.3) 3.6 (3.1–4.2) 3.7 (3.2–4.3) 3.9 (3.0–5.0) 0.10 Number of hospital days/1 PY 29,284 1 3.9 (3.3–4.6) 4.5 (3.9–5.3) 4.5 (3.8–5.4) 4.7 (4.0–5.6) 6.8 (5.7–8.2) ,0.001 2 4.2 (3.5–5.0) 4.7 (4.0–5.5) 4.5 (3.8–5.5) 4.7 (3.9–5.6) 5.1 (3.8–7.0) 0.85 Model 1: Adjusted mean estimates (least square means) with respective 95% CI derived from logistic regression analysis (for outcomes use of CGMS, use of rapid-acting insulin analogs, use of long-acting insulin analogs, participation in diabetes education program, prevalence of obesity), linear regression analysis (for outcomes SMBG, BMI SDS), or Poisson regression analysis considering overdispersion (for outcomes rate of severe hypoglycemia with coma, rate of severe DKA [pH ,7.1], number of hospital days). All regression models were performed with GIMD 2010 quintiles as the categorical independent variable and adjusting for sex, age group, migration background, and diabetes duration. Model 2: Estimates from regression models additionally adjusted for German federal state. *P value of test of no difference in outcome distribution across GIMD quintiles. P values were adjusted for multiple testing using the FDR-controlling Benjamini-Hochberg procedure (30). ** Only patients without CSII.

fact, in contrast to DKA, which requires a country (Table 1 and Fig. 1A). Although thus, independently of East–West dis- visit to the diabetes care center, severe the living conditions in former East and parities. hypoglycemia can be treated by patients West Germany have slowly converged The major strength of this study is its or parents themselves and may easily since German reunification (35), eco- very large sample size with patients be forgotten until the next medical visit. nomic performance is still lower and from a large number of diabetes care In accordance with this explanation, the proportion of people affected by centers throughout the country. We no association was observed between poverty and unemployment remains used a nationwide diabetes follow-up area deprivation and severe hypoglyce- higher in the eastern part of the country registry covering more than 85% of mia with coma, where underreporting (36). The health status of children and the pediatric subjects with type 1 di- is less likely. adolescents has become more similar, abetes in Germany, so that the results In our results, higher area deprivation but some important differences in health can be considered as representative of tended to be associated with higher risk behavior still remain. In particular, com- this population. Moreover, detailed in- of hospital admission for DKA, and this is pared with peers living in the western formation on the patients’ demographic consistent with previous findings (34). part of the country, more adolescents in and clinical characteristics was available, Overall, many factors may contribute the eastern part regularly drink alcohol or which allows comprehensive control of to the differences in treatment and out- smoke, and fewer children are members potential confounders. come quality in pediatric patients with of a sports club (37). However, our study One limitation of this study is that type 1 diabetes within Germany. The indicates that half of the analyzed di- analyses could not consider individual- GIMD 2010 partly reflects East–West abetes-related outcomes (use of CSII, level SES. In DPV, education level is inequalities in Germany: districts in CGMS, or insulin analogs; HbA1c; rate incompletely documented, and house- less deprived quintiles were mostly lo- of severe hypoglycemia; BMI SDS; and hold income is not available. Studies cated in the western part, whereas dis- prevalence of overweight) were signifi- on patients with type 2 diabetes have tricts in the most deprived quintiles were cantly associated with area deprivation demonstrated that the effect of area mostly located in the eastern part of the independently of the federal state and, deprivation remains significant after 2524 Area Deprivation and Type 1 Diabetes Diabetes Care Volume 41, December 2018

controlling for individual SES (8,19). quality of care of pediatric type 1 di- 3. Bohn B, Rosenbauer J, Icks A, et al.; DPV Maier et al. (19) argue that individual SES abetes, even in high-income countries. Initiative. Regional disparities in diabetes care and area deprivation may “act through for pediatric patients with type 1 diabetes. A cross-sectional DPV multicenter analysis of different pathways.” For instance, a strong Acknowledgments. Special thanks to A. Hun- 24,928 German children and adolescents. Exp – social safety net, as well as dedicated re- gele and R. Ranz for support and the develop- Clin Endocrinol Diabetes 2016;124:111 119 sources through social spending to “stable ment of the DPV documentation software; to K. Fink 4. Townsend P, Phillimore P, Beattie A. Health housing, educational opportunities, nutri- and E. Bollow for the DPV data management; and to and Deprivation: Inequality and the North. Lon- tion, and transportation,” is considered to R.W. Holl for data management, initiation of the don, Routledge, 1988 5. Fairburn J, Maier W, Braubach M. Incorpo- play a decisive role in enhancing the DPV collaboration, and being the principal inves- tigator of the DPV registry (all from the Institute of rating environmental justice into second gener- quality of care, especially for populations Epidemiology and Medical Biometry,ZIBMT,Uni- ation indices of multiple deprivation: lessons with lower income, lower educational versity of Ulm). The authors thank G.G. Greiner from the UK and progress internationally. Int level, or minority status (38). (Institute of Health Economics and Health Care J Environ Res Public Health 2016;13:750 Another weakness is the heterogene- Management, Helmholtz Zentrum Munchen)¨ for 6. Noble M, Wright G, Smith G, Dibben C. Mea- suring multiple deprivation at the small-area ity of German districts: they are admin- assistance in creating the maps. Furthermore, the authors thank all participating centers in the level. Environ Plann A 2006;38:169–185 istrative units that vary considerably in DPVinitiative, especially the centers contributing 7. Grintsova O, Maier W, Mielck A. Inequalities in area and population size (from ;35,000 data to this investigation and their patients. A health care among patients with type 2 diabetes up to more than 1 million inhabitants). detailed list of the centers contributing data to by individual socio-economic status (SES) and re- We assume that the analysis could be less this analysis can be found in the online Supple- gional deprivation: a systematic literature review. Int J Equity Health 2014;13:43 sensitive in larger districts than in smaller mentary Data. Funding. The DPV registry and this analysis 8. Maier W, Scheidt-Nave C, Holle R, et al. Area ones. However, the influence of extreme are supported by the German Center for Diabe- level deprivation is an independent determi- values in single domains of the GIMD is tes Research (DZD). Further financial support nant of prevalent type 2 diabetes and obesity limited because a ranking transformation for the DPV registry was provided by the German at the national level in Germany. Results from was used in the algorithm for the index Diabetes Association (DDG) and by the European the National Telephone Health Interview Sur- Foundation for the Study of Diabetes (EFSD). The veys ‘German Health Update’ GEDA 2009 and calculation. Furthermore, because pedi- DPV registry receives funding from the Innova- 2010. PLoS One 2014;9:e89661 atric diabetes health care in Germany is tive Medicines Initiative 2 Joint Undertaking 9. Zuijdwijk CS, Cuerden M, Mahmud FH. Social organized at the district level, heteroge- (INNODIA) under grant agreement 115797, determinants of health on glycemic control in pe- neity within districts may play a less supported by the European Commission’s diatric type 1 diabetes. J Pediatr 2013;162:730–735 10. Clarke ABM, Daneman D, Curtis JR, Mahmud important role. Horizon 2020 Research and Innovation Program and the European Federation of Pharmaceutical FH. Impact of neighbourhood-level inequity Further shortcomings of this study are Industries and Associations, JDRF, and The Leona on paediatric diabetes care. Diabet Med 2017; that complete data were not available M. and Harry B. Helmsley Charitable Trust. 34:794–799 for each patient, and variability in the Duality of Interest. No potential conflicts of 11. Collier A, Ghosh S, Hair M, Waugh N. Impact measurements of clinical characteristics interest relevant to this article were reported. of socioeconomic status and gender on glycaemic control, cardiovascular risk factors and di- cannot be completely excluded because Author Contributions. M.A. wrote the manuscript and created the figures. M.A., S.L., B.B., abetes complications in type 1 and 2 diabetes: of the multicenter design. However, we J.R., and W.M. designed the study. S.L., J.R., and a population based analysis from a Scottish standardized locally measured HbA1c val- W.M. analyzed the study data and reviewed and region. Diabetes Metab 2015;41:145–151 ues to the DCCT standard. Furthermore, edited the manuscript. B.B., P.K., U.K.-K., P.M.H., 12. Lindner LME, Rathmann W, Rosenbauer J. because of the cross-sectional design, K.P., J.H., R.B., J.R., and W.M. contributed to Inequalities in glycaemic control, hypoglycaemia and diabetic ketoacidosis according to socio- this study does not allow us to draw the discussion and reviewed and edited the manuscript. W.M. created the maps. S.L. is economic status and area-level deprivation in any causal interpretation. Finally, the the guarantor of this work and, as such, had type 1 diabetes mellitus: a systematic review. nature of the database does not allow full access to all the data in the study and takes Diabet Med 2018;35:12–32 in-depth analysis of all possibly impor- responsibility for the integrity of the data and 13. Scott A, Chambers D, Goyder E, O’Cathain A. tant determinants (e.g., individual so- the accuracy of the data analysis. Socioeconomic inequalities in mortality, mor- Prior Presentation. Parts of this study were bidity and diabetes management for adults cioeconomic data), and the nature of presented in abstract form at the 43rd Annual with type 1 diabetes: a systematic review. PLoS the German diabetes care system limits Conference of the International Society for Pe- One 2017;12:e0177210 generalizability of the findings. diatric and Adolescent Diabetes (ISPAD), Inns- 14. Neu A, Burger-B¨ using¨ J, Danne T, et al. In conclusion, we showed that in pe- bruck, Austria, 18–21 October 2017, and at Diagnosis, therapy and follow-up of diabetes diatric patients with type 1 diabetes in the European Congress of Epidemiology, Lyon, mellitus in children and adolescents. Diabetol – – fi France, 4 6July2018. Stoffwechs 2016;11:S35 S116 [in German] Germany, area deprivation was signi - 15. Wild F. Supply of insulins in private health cantly associated with many indicators References insurance compared to statutory health insur- of treatment and outcome quality, inde- 1. de Beaufort CE, Swift PG, Skinner CT, et al.; ance. Gesundheitsokonomie¨ und Qualitatsma-¨ pendently of the federal state. In partic- Hvidoere Study Group on Childhood Diabetes nagement 2009;14:200–203 [in German] ular, our findings suggest that a focus on 2005. Continuing stability of center differences 16. Bohn B, Karges B, Vogel C, et al.; DPV Initiative. 20 years of pediatric benchmarking equal access to diabetes treatment, such in pediatric diabetes care: do advances in diabetes treatment improve outcome? The Hvi- in Germany and Austria: age-dependent analy- as CGMS and CSII, is important because sis of longitudinal follow-up in 63,967 children doere Study Group on Childhood Diabetes. treatment is a directly modifiable factor. and adolescents with type 1 diabetes. PLoS One Diabetes Care 2007;30:2245–2250 2016;11:e0160971 Moreover, diabetes technology may im- 2. Gomes MB, Cobas RA, Matheus AS, et al. prove metabolic control regardless of 17. Craig ME, Jefferies C, Dabelea D, Balde N, Regional differences in clinical care among pa- Seth A, Donaghue KC; International Society educational level (39). Consequently, bet- tients with type 1 diabetes in Brazil: Brazilian for Pediatric and Adolescent Diabetes. ISPAD ter access to diabetes technology in the Type 1 Diabetes Study Group. Diabetol Metab Clinical Practice Consensus Guidelines 2014. most deprived areas may improve the Syndr 2012;4:44 Definition, epidemiology, and classification care.diabetesjournals.org Auzanneau and Associates 2525

of diabetes in children and adolescents. Pediatr Network Diabetes Mellitus. Improved metabolic hypoglycemia rates are not associated with Diabetes 2014;15(Suppl. 20):4–17 control in children and adolescents with type 1 HbA1c:across-sectionalanalysisof3contemporary 18. Maier W, Fairburn J, Mielck A. Regional diabetes: a trend analysis using prospective pediatric diabetes registry databases. Pediatr deprivation and mortality in Bavaria. Develop- multicenter data from Germany and Austria. Diabetes 2017;18:643–650 ment of a community-based index of multiple Diabetes Care 2012;35:80–86 34. Govan L, Maietti E, Torsney B, et al.; Scottish deprivation. Gesundheitswesen 2012;74:416–425 26. Ly TT, Maahs DM, Rewers A, Dunger D, Diabetes Research Network Epidemiology [in German] Oduwole A, Jones TW; International Society Group. The effect of deprivation and HbA1c 19. Maier W, Holle R, Hunger M, et al.; DIAB- for Pediatric and Adolescent Diabetes. ISPAD on admission to hospital for diabetic ketoacidosis CORE Consortium. The impact of regional dep- Clinical Practice Consensus Guidelines 2014. in type 1 diabetes. Diabetologia 2012;55:2356– rivation and individual socio-economic status on Assessment and management of hypoglyce- 2360 the prevalence of type 2 diabetes in Germany. A miainchildrenandadolescentswithdiabe- 35. Bundeszentrale fur¨ politische Bildung. Dos- pooled analysis of five population-based studies. tes. Pediatr Diabetes 2014;15(Suppl. 20):180– sier: Lange Wege der Deutschen Einheit [Inter- Diabet Med 2013;30:e78–e86 192 20. Schafer¨ T, Pritzkuleit R, Jeszenszky C, et al. 27. Clarke W, Jones T, Rewers A, Dunger D, net], 2017. Available from http://www.bpb.de/ Trends and geographical variation of primary hip Klingensmith GJ. Assessment and management geschichte/deutsche-einheit/lange-wege-der- and knee joint replacement in Germany. Oste- of hypoglycemia in children and adolescents deutschen-einheit/. Accessed 2 November oarthritis Cartilage 2013;21:279–288 with diabetes. Pediatr Diabetes 2009;10(Suppl. 2017 21. Konrad K, Vogel C, Bollow E, et al.; German/ 12):134–145 36. Bundesagentur fur¨ Arbeit. Statistik. Avail- Austrian DPV Initiative and the Competence 28. Cromley RG, Cromley EK. Choropleth map able from https://statistik.arbeitsagentur.de/ Network of Diabetes. Current practice of diabe- legend design for visualizing community health Navigation/Statistik/Statistik-nach-Regionen/ tes education in children and adolescents with disparities. Int J Health Geogr 2009;8:52 Politische-Gebietsstruktur/Ost-West-Nav.html. type 1 diabetes in Germany and Austria: analysis 29. Kirtland KA, Burrows NR, Geiss LS. Diabetes Accessed 19 October 2017 based on the German/Austrian DPV database. interactive atlas. Prev Chronic Dis 2014;11: 37. Lampert T. 20 years of German unity: are Pediatr Diabetes 2016;17:483–491 130300 there still differences between east and west as 22. Rosario AS, Kurth BM, Stolzenberg H, Ellert 30. Benjamini Y, Hochberg Y. Controlling the regards the health of children and adolescents? U, Neuhauser H. Body mass index percentiles false discovery rate: a practical and powerful GBE kompakt 4/2010 [Internet], 2010. Berlin, for children and adolescents in Germany based approach to multiple testing. J Roy Stat Soc B Robert Koch Institute. Available from http:// – on a nationally representative sample (KiGGS 1995;57:289 300 www.rki.de/EN/Content/Health_Monitoring/ – 2003-2006). Eur J Clin Nutr 2010;64:341 349 31. Nevill AM, Duncan MJ, Lahart I, Sandercock Health_Reporting/GBEDownloadsK/2010_4_ 23. Wabitsch M, Kunze D. Konsensbasierte (S2) G. Modelling the association between weight 20_years_german_unity.pdf?__blob=publicationFile. Leitlinie zur Diagnostik, Therapie und Pravention¨ status and social deprivation in English school Accessed 22 February 2018 von Ubergewicht¨ und Adipositas im Kindes- children: can physical activity and fitness affect 38. Schneider EC, Squires D. From last to firstd und Jugendalter (version 15.10.2015) [Internet], the relationship? Ann Hum Biol 2016;43:497– could the U.S. health care system become the 2015. Available from http://www.aga.adipositas- 504 – gesellschaft.de/fileadmin/PDF/Leitlinien/AGA_S2_ 32. Leese GP, Wang J, Broomhall J, et al.; DARTS/ best in the world? N Engl J Med 2017;377:901 Leitlinie.pdf. Accessed 9 April 2018 MEMO Collaboration. Frequency of severe hy- 904 24. Poskitt EM. Defining childhood obesity: poglycemia requiring emergency treatment in 39. Senniappan S, Hine P, Tang W, et al.; North the relative body mass index (BMI). European type 1 and type 2 diabetes: a population-based West England Paediatric Diabetes Network. The Childhood Obesity Group. Acta Paediatr 1995; study of health service resource use. Diabetes effect of socioeconomic deprivation on efficacy 84:961–963 Care 2003;26:1176–1180 of continuous subcutaneous insulin infusion: 25. Rosenbauer J, Dost A, Karges B, et al.; DPV 33. Haynes A, Hermann JM, Miller KM, et al.; T1D a retrospective paediatric case-controlled sur- Initiative and the German BMBF Competence Exchange, WACDD and DPV Registries. Severe vey. Eur J Pediatr 2012;171:59–65