We used claims data of the German pension fund for the present analyses. The Scientific Use File (SUF) “completed rehabilitation measures 2010–2017” (Abgeschlossene Rehabilitationen 2010–2017 [in German]) is available upon request for scientific institutions. The dataset contains person-based data with records on insured persons and their eligible relatives. Included are persons with one of the following characteristics within the observed period (2010–2017): completed rehabilitation, approved pension or belonging to a specific demographic cohort (death before or at the age of 75 or belonging to a certain year of birth). Persons whose applications for rehabilitation or for any pension had finally been denied are excluded. For each rehabilitation, pension, or demographic event, included persons are being observed over an 8-year period, with minor exceptions (e.g. date of death information is available up to 9 years after the first documented event.) The SUF is a complex random sample regarding the aforementioned events and comprises a random sample of 20% of all insured persons (N ≈ 3.7 million). We used the given weighting factor to infer from the random sample to the target population. Given that the DRV is the main payer for rehabilitation services for the working population of Germany, the age range of the participants in this dataset is typically between 16 and 66 years. A detailed description of the dataset, including the sampling design, can be found elsewhere [17, 18]. Due to its longitudinal design (with individual data from 2010 to 2017), the SUF is an appropriate source for the analysis of the effects and time trends of rehabilitation services.
The study adheres to the reporting guidelines of the STROBE Statement for observational studies .
Participants and Outcomes
Out of the complete dataset of 3.7 million insured persons, we only used data with information on medical rehabilitation measures for our analyses. For these approximately 2.2 million patients we defined the following inclusion and exclusion criteria.
In the first step, we included rehabilitation patients with a primary diagnosis of CHD or ischemic heart disease (ICD10 I20.–I25.) who had completed medical rehabilitation (Phase II). In the next step, we matched EBRP-III to our study population. Participation had to start within six months after completing Phase II rehabilitation.
People who died within a 12-month period after completing Phase II rehabilitation were excluded to avoid selection bias due to the inclusion of severely ill patients. We also excluded persons without regularly completed rehabilitation phase II because of medical problems or premature termination of any other cause.
The time of study entry was determined by the date of the first use of Phase II rehabilitation services, which could have been any date after January 1, 2010. Patients were followed up for mortality or reduced working capacity as primary study outcomes until December 31, 2017. The loss of working capacity was determined by the payment of a reduced earning capacity pension by the DRV. Based on the actual duration of EBRP-III participation, we defined three separate study groups: 1) no participation, 2) short participation (< 90 days), and 3) long participation (≥ 90 days). A participation duration of 90 days indicated an EBRP-III program participation rate of a minimum of 50%, which was defined as acceptable adherence (long participation).
In our analyses we used a stepwise selection procedure for the selection of covariates. Education level was determined using the available data on the highest level of schooling and/or professional training. Education was categorized as lower secondary or elementary school with or without professional training, higher education entrance qualification with or without professional training, university degree, or no information available.
Subjective medical rehabilitation outcome was determined based on the physician’s assessment indicated in the Phase II rehabilitation discharge letter. Patients could rate their rehabilitation outcome as worse, unchanged, or better.
Information regarding sick leave was based on DRV recordings and expressed as days of absence from work within the last 12 months prior to rehabilitation entry. This data was categorized as no days of absence, less than 3 months of absence, 3–6 months of absence, more than 6 months of absence, or not employed.
The number of comorbidities represented the number of documented physician-based diagnoses next to CHD or ischemic heart disease to represent the degree of multi-morbidity.
For unadjusted analyses and baseline characteristics, a Chi2 test was applied for associations between sex, education, subjective medical rehabilitation outcome (Phase II), sick leave, age at rehabilitation entry, number of comorbidities, and participation in EBRP-III (3 groups). For adjusted analyses, Cox proportional hazards models were applied to account for potential confounding in the relationship between program participation and mortality or reduced earning capacity, respectively . The exposure variable was program participation, with “long participation” as the reference category.
For sensitivity analyses, we used a shorter period of only six months of minimum survival after completing medical rehabilitation (Phase II).
All analyses were performed using the mentioned weighting factor which is provided within the SUF to account for a disproportional sampling procedure . Furthermore, all analyses were performed with the statistical software package SAS University Edition (SAS Institute, Cary, NC, USA). A p-value < 0.05 was considered to indicate statistical significance. For the reporting of patient pathways and outcomes within the German rehabilitation system we used a Sankey diagram. Sankey diagrams are a type of flow diagram in which the width of the arrows is proportional to the flow rate. In our case all CHD patients were divided in two groups (participators vs. non-participators) and the connections (arrows) show the evolution between different stages of the rehabilitation process.