Medication Adherence and Healthcare Costs in Chronically Ill Patients Using German Claims Data.

BACKGROUND AND OBJECTIVE: Despite the importance of medication adherence for chronically ill patients and the vast literature on its relationship to costs, this field suffers from methodological limitations. These are caused, amongst others, by the lack of generalizability of data sources, varying definitions of adherence, costs, and model specification. We aim to address this with different modeling approaches and to contribute evidence on the research question. METHODS: We extracted large cohorts of nine chronic diseases (n = 6747-402,898) from German claims data of stationary health insurances between 2012 and 2015 (t0-t3). Defined as the proportion of days covered by medication, we examined the relationship of adherence using several multiple regression models at baseline year t0 with annual total healthcare costs and four sub-categories. Models with concurrent, and differently time-lagged measurements of adherence and costs were compared. Exploratively, we applied non-linear models. RESULTS: Overall, we found a positive association between the proportion of days covered by medication and total costs, a weak association with outpatient costs, positive with pharmacy costs, and frequently negative with inpatient costs. There were major differences by disease and its severity but little between years, provided adherence and costs were not measured concurrently. The fit of linear models was mainly not inferior to that of non-linear models. CONCLUSIONS: The estimated effect on total costs differed from most other studies, which highlights concerns about generalizability, although effect estimates in sub-categories were as expected. Comparison of time lags indicates the importance of avoiding concurrent measurement. A non-linear relationship should be considered. These methodological approaches are valuable in future research on adherence and its consequences.

Identification of target groups and individuals for adherence interventions using tree-based prediction models.

Background: In chronically ill patients, medication adherence during implementation can be crucial for treatment success and can decrease health costs. In some populations, regression models do not show this relationship. We aim to estimate subgroup-specific and personalized effects to identify target groups for interventions. Methods: We defined three cohorts of patients with type 1 diabetes (n = 12,713), type 2 diabetes (n = 85,162) and hyperlipidemia (n = 117,485) from German claims data between 2012 and 2015. We estimated the association of adherence during implementation in the first year (proportion of days covered) and mean total costs in the three following years, controlled for sex, age, Charlson's Comorbidity Index, initial total costs, severity of the disease and surrogates for health behavior. We fitted three different types of models on training data: 1) linear regression models for the overall conditional associations between adherence and costs, 2) model-based trees to identify subgroups of patients with heterogeneous adherence effects, and 3) model-based random forests to estimate personalized adherence effects. To assess the performance of the latter, we conditionally re-estimated the personalized effects using test data, the fixed structure of the forests, and fixed effect estimates of the remaining covariates. Results: 1) our simple linear regression model estimated a positive adherence effect, that is an increase in total costs of 10.73 Euro per PDC-point and year for diabetes type 1, 3.92 Euro for diabetes type 2 and 1.92 Euro for hyperlipidemia (all p ≤ 0.001). 2) The model-based tree detected subgroups with negative estimated adherence effects for diabetes type 2 (-1.69 Euro, 24.4% of cohort) and hyperlipidemia (-0.11 Euro, 36.1% and -5.50 Euro, 5.3%). 3) Our model-based random forest estimated personalized adherence effects with a significant proportion (4.2%-24.1%) of negative effects (up to -8.31 Euro). The precision of these estimates was high for diabetes type 2 and hyperlipidemia patients. Discussion: Our approach shows that tree-based models can identify patients with different adherence effects and the precision of personalized effects is measurable. Identified patients can form target groups for adherence-promotion interventions. The method can also be applied to other outcomes such as hospitalization risk to maximize positive health effects of an intervention.

Compensating for missing data in the OHRKAN cohort study examining total leisure noise exposure among adolescents.

OBJECTIVE: Investigating determinants of total leisure noise (TLN) exposure among adolescents over 7.5 years and compensating for missing data due to loss to follow-up. DESIGN: In the OHRKAN cohort study, data were collected by questionnaires at four waves. TLN was calculated from self-reported duration spent participating in 18 leisure activities. High exposure was defined as exceeding 85 dB(A) of equivalent continuous average sound pressure level (SPL) during a 40-h week. Multiple imputation (MI) and generalised estimating equations (GEE) were used to analyse odds ratios (OR) of determinants of TLN exposure and compared to complete-case analysis. STUDY SAMPLE: Closed cohort of 2148 students enrolled in grade 9 of any school in Regensburg (Germany), recruited from 2009 to 2011. RESULTS: Up to 74% of adolescents had risky TLN exposure, depending on wave. The most significant sources were discotheques, portable listening devices (PLD) and stereo systems. Higher TLN exposure was associated with time point, education, single-parent households and gender. MI under MAR assumption changed results only slightly compared to complete-case analysis. CONCLUSIONS: Prevalence of risky TLN is high during adolescence. MI reinforced trends detected in former results of OHRKAN. Preventive measures should consider the main drivers of noise exposure including changes by age and high risks groups.