Social inequalities in medical rehabilitation outcomes—a registry-based study on 219 584 insured persons in Germany

Abstract

Background

Given limited knowledge on the extent of social inequalities in longer-term work ability of people with a chronic disease, this study analyzes social inequalities of three consecutive indicators of work ability following medical rehabilitation in a large sample of insured employees.

Methods

Based on data from the German statutory pension insurance, a representative 20% random sample of all employed persons undergoing medical rehabilitation between 2006 and 2008 was included in a longitudinal analysis (n=219 584 persons). Three measures of consecutive work-related outcomes (physicians’ assessment of work ability at discharge; return to work in the year thereafter; disability pension during follow-up) and socioeconomic position (SEP) (education, occupational position and income) were assessed. Adjusted relative risks (RRs) for each outcome were calculated according to SEP, applying Poisson regression analysis.

Results

The measures of SEP were associated with all three outcomes of work ability in the fully adjusted models. Relatively strongest relationships were observed for education as SEP measure, and they were particularly pronounced for ‘low work ability’ (RR=2.38 for lower secondary education compared to tertiary education; 95% CI: 2.26–2.51). Based on average marginal effects, absolute differences of work ability by SEP indicate a socially graded pattern, with only few exceptions.

Conclusions

Despite Germany’s universal access to medical and vocational rehabilitation social inequalities in longer-term work ability following chronic disease persist, thus calling for targeted programmes of prevention and occupational health promotion.

Introduction

Scientific evidence indicates that social gradients of morbidity and mortality persist in many countries with developed social and health policies, leaving people in lower socioeconomic positions (SEPs) at higher risk to develop a disease.¹ Less knowledge, though, is available on social inequalities of the longer-term consequences for living and working conditions among those who already developed or survived an incident chronic disease, such as cardiovascular disease, cancer or depression. Recent data from the German micro census (2013), e.g. suggests that 12.7 million people with impairments are living in Germany. They are 30 percentage points less likely to participate on the labour market and 7 percentage points more likely to experience poverty compared to the population without impairments.² While several investigations documented social gradients of success in return to work following hospitalization due to a stroke,³ coronary heart disease,^4,5 cancer^6,7 or other long-standing illnesses,⁸ few studies only analyzed longer-term outcomes, such as disability pensions, for these patient groups. A recent report from Finland provides an exception as it explored whether social inequalities of disability pension differed between hospitalized persons with severe disease and persons without severe disease.⁹ Although this study observed social gradients in disability pensions in both groups, the association of SEP with disability pension was somewhat stronger among persons with hospitalization, in particular in case of injuries and musculoskeletal disorders.⁸ Yet, a comprehensive study on socioeconomic differences of trajectories of work ability from hospitalization to differential opportunities of returning to work and to longer-term risks of early exit from labour market is still missing.

To fill this gap, administrative data derived from the German health and social security system are of interest. There are almost no social differences in access to, and treatment obtained in medical rehabilitation clinics for persons insured under the German pension insurance scheme. The pension funds have established a detailed registry of patients’ trajectories following medical rehabilitation, including data on sociodemographic characteristics, employment histories (return to work and early exit from paid work) and survival.¹⁰

With this retrospective observational study, we set out to analyze social inequalities of trajectories following medical rehabilitation from chronic disease in a large sample of insured men and women in Germany, using this administrative dataset. More specifically, three steps of these trajectories are studied, (i) the patients’ work ability, as assessed by physicians before leaving the rehabilitation clinic, (ii) the patients’ return to work during the first year after medical rehabilitation and (iii) whether patients left the labour market due to disability pension during a follow-up observation period (up to 7 years). As SEP was recorded by occupational position, educational level and income, we further explored the relative strength of impact of each indicator on the outcomes under study.

Methods

Data

This analysis relies on administrative data from the German statutory pension insurance [i.e. ‘Deutsche Rentenversicherung’ (DRV)].¹⁰ Beyond its responsibility for the administration of insurance contributions and pension insurance payments, the DRV has been charged with the implementation, financing and control of medical rehabilitation across the country.¹¹ This policy was introduced several decades ago to save costs by reducing premature exit from labour market through strengthening the work ability of persons affected by chronic diseases and disabilities. In this context, the DRV has established an elaborated administrative database that can be used for scientific purposes, with anonymized information complying with established standards of data protection. Our analysis relies on information from a 20% sample of all cases of medical rehabilitation occurring between January 2006 and December 2008, financed by the DRV.

Setting

Germany offers extensive medical rehabilitation services to all persons covered by the national statutory pension insurance following hospitalization or onset of a major chronic disease, e.g. stroke, coronary heart disease, cancer, mental disorders and musculoskeletal disorders. Services are typically provided in specialized rehabilitation clinics as inpatient treatment for several weeks, with the aim of improving the opportunities of returning to work.¹¹

Sample

For the purpose of this study, the 20% sample of the total population of insured persons undergoing medical rehabilitation between 2006 and 2008 (487 241 observations, 417 465 persons) was further reduced by excluding persons who were not regularly employed during the whole year before the rehabilitation (225 260 observations), and by excluding persons aged below 19 years or beyond 60 years (12 821 observations) that did not meet all inclusion criteria (e.g. long-term return to work can no longer be assessed after age 60). Furthermore, we excluded observations with missing data on occupation (n = 3478) or medical diagnosis (n = 168), resulting in a sample size of 245 514 observations from 219 584 persons. Given an additional high percentage of observations with missing information on education (n = 24 826, 10.1%), we decided not to exclude this subgroup, but rather to integrate it as a distinct category (‘education: unknown’) into the analysis. The difference between number of persons and number of observations is due to the fact that the same person can undergo more than one rehabilitation.

Measures

SEP: the following three indicators are available from this database: education, income and occupational position. Educational degree is measured by four categories ranging from lower secondary education (ISCED-2011 Level 2) and higher secondary education (ISCED-2011 Level 3) to post-secondary non-tertiary education and short-cycle tertiary education (ISCED-2011 Level 4–5) to tertiary education (ISCED-2011 Level 6–8).¹² Information on income is based on the yearly labour wage of insured persons, categorized into ‘low’, ‘medium’ and ‘high’ tertiles, separately for part-time and full-time employees. According to a previous approach,¹³ we classified information on occupational position into five distinct groups: (i) non-skilled manual, (ii) skilled manual, (iii) non-skilled non-manual, (iv) skilled non-manual and (v) highly-qualified jobs. All indicators were obtained from data available during the year before onset of medical rehabilitation.

Outcome measures: we use three consecutive binary measures of work-related outcomes, reflecting occupational trajectories following medical rehabilitation. These measures are (i) the patients’ work ability, as assessed by physicians before leaving the rehabilitation clinic, (ii) the patients’ return to work (full or part-time) during the first year after medical rehabilitation and (iii) the patients’ early exit from work due to disability pension during a follow-up observation period up to 7 years. While the last two measures are administrative registry data, the measure of work ability represents physicians’ evaluation of the patients’ ability to meet their occupational demands, given their current functional limitation. Clearly, these assessments may not be free from biased decision-making.

Additional measures

To estimate the impact of disease severity on the outcomes under study, we introduced the following three proxy measures of severity of disease at onset of medical rehabilitation: (i) the patients’ duration of sickness absence during the year before rehabilitation; (ii) the patients’ first versus recurrent rehabilitation; (iii) the patients’ diagnosis at onset of medical rehabilitation. As an additional variable, age was included as continuous variable. Moreover, all analyses were adjusted for sex as we did not find systematic differences between men and women in the associations under study.

Statistical analysis

Following a basic sample description (table 1), we give a brief overview of the socioeconomic distribution of the three indicators of rehabilitation outcomes, ‘work inability’, ‘failed return to work’ and ‘disability pension’. Separate distributions are displayed for occupational position, education and income (see table 2). Next, we use Poisson regression analysis to calculate adjusted relative risks (RRs) for each indicator of rehabilitation outcomes according to the three measures of SEP. In a first step, these risks are estimated separately for each measure (model 1), and in a second step, they are analyzed jointly for the three measures with respective adjustments (model 2). We calculated RR instead of odds ratios as they provide less biased estimates in studies with frequent outcomes.¹⁴ All estimations were adjusted for age, sex and the three proxy measures of disease severity, i.e. medical diagnosis, first vs. repeated rehabilitation and duration of sickness absence before rehabilitation. In all models, robust estimators were used to account for potential clustering of observations in individuals. In addition, to visualize absolute differences between socioeconomic groups, predicted values based on average marginal effects were calculated (see below figure 1).

Figure 1

Predicted frequencies, based on average marginal effects, adjusted for age, sex, diagnosis, repeated rehabilitation and sickness absence before rehabilitation (n=245 514)

Table 1

Sample description (N=245 514)

	N	%
Low work ability
No	197 779	80.6
Yes	47 735	19.4
Failed return to work
No	150 835	61.4
Yes	94 679	38.6
Disability pension
No	216 076	88.0
Yes	29 438	12.0
Education
Unknown	24 826	10.1
Lower secondary	32 515	13.2
Higher secondary	1238	0.5
Post-secondary/short-cycle tertiary	172 482	70.3
Tertiary	14 453	5.9
Occupation
Unskilled, manual.	40 261	16.4
Skilled, manual	51 421	20.9
Unskilled, non-manual	56 358	23.0
Skilled, non-manual	87 159	35.5
Highly skilled	10 315	4.2
Income
Low	81 845	33.3
Middle	81 834	33.3
High	81 835	33.3
Diagnosis
Musculoskeletal system	136 666	55.7
Cardiovascular system	21 365	8.7
Neoplasm	13 289	5.4
Mental	32 955	13.4
Others	41 239	16.8
Repeated treatment
No	154 011	62.7
Yes	91 503	37.3
Sickness absence before rehabilitation
None	50 439	20.5
Below 3 months	142 338	58.0
3–6 months	34 040	13.9
Over 6 months	18 697	7.6
Sex
Male	132 682	54.0
Female	112 832	46.0
Age
19–39	35 210	14.3
40–54	147 249	60.0
55–60	63 055	25.7
Total	245 514	100.0

Table 2

Absolute and relative frequencies of three indicators of ‘low work ability’, ‘failed return to work’ and ‘disability pension’ following medical rehabilitation along the SEP indicators and the additional measures (N =245 514)

	Low work ability		Failed return to work		Disability pension
	N	%	N	%	N	%
Sex
Male	27 630	20.8	51 435	38.8	13 952	10.5
Female	20 105	17.8	43 244	38.3	15 486	13.7
Education
Unknown	6450	26.0	12 528	50.5	3455	13.9
Lower secondary	8464	26.0	14 961	46.0	4430	13.6
Higher secondary	203	16.4	408	33.0	124	10.0
Post-secondary/short-cycle tertiary	31 194	18.1	63 358	36.7	20 433	11.8
Tertiary	1424	9.9	3424	23.7	996	6.9
Occupation
Unskilled, manual	10 018	24.9	17 659	43.9	4768	11.8
Skilled, manual	11 071	21.5	20 268	39.4	5508	10.7
Unskilled, non-manual	12 980	23.0	25 450	45.2	7525	13.4
Skilled, non-manual	12 616	14.5	28 731	33.0	10 936	12.5
Highly skilled	1050	10.2	2571	24.9	701	6.8
Income
Low	21 864	26.7	41 645	50.9	12 230	14.9
Middle	15 691	19.2	30 703	37.5	10 173	12.4
High	10 180	12.4	22 331	27.3	7035	8.6
Diagnosis
Musculoskeletal system	26 142	19.1	49 385	36.1	13 243	9.7
Cardiovascular system	4216	19.7	8569	40.1	2509	11.7
Neoplasm	3456	26.0	6958	52.4	1993	15.0
Mental	5226	15.9	14 070	42.7	5197	15.8
Others	8695	21.1	15 697	38.1	6496	15.8
Age
19–39	7110	20.2	14 449	41.0	2454	7.0
40–54	26 267	17.8	51 472	35.0	20 957	14.2
55–60	14 358	22.8	28 758	45.6	6027	9.6
Repeated treatment
No	27 461	17.8	55 333	35.9	15 573	10.1
Yes	20 274	22.2	39 346	43.0	13 865	15.2
Sickness absence before rehabilitation
None	16 498	32.7	14 644	29.0	4664	9.2
Below 3 months	15 373	10.8	45 857	32.2	14 552	10.2
3–6 months	8877	26.1	20 084	59.0	6208	18.2
Over 6 months	6987	37.4	14 094	75.4	4014	21.5
Total	47 735	19.4	94 679	38.6	29 438	12.0

Results

As obvious from table 1, similar proportions of men and women were represented in this sample (54% male, 46% female). Concerning education, the large majority of respondents had post-secondary education (70.3%), reflecting the high relevance of vocational training in the German education system. When considering the distribution of occupations, manual occupations were still rather frequent in this cohort (37.3%), but the small number of persons belonging to highly skilled occupation is noticeable (4.2%).

This fact that may be attributed to the underrepresentation of civil servants and self-employed persons as they are generally covered by different pension systems. Moreover, people with high SEP generally prefer outpatient treatment rather than rehabilitation in clinical settings. With regard to age, a large majority belonged to the category 40–54 years (60.0%). The most frequent diagnosis leading to medical rehabilitation was ‘musculoskeletal disorders’ (55.7%). It is also of interest to see that for more than a third of the sample, this was not the first medical rehabilitation in their occupational career, a fact that points to a high burden of chronic health problems in the German middle-aged employed population.

Table 2 displays the distribution of main variables of the study population, stratified according to the three outcome measures. There are no pronounced differences in the frequency of the three outcomes between men and women (e.g. 38.8% male, 38.3% female in case of ‘failed return to work’). In all three measures of SEP, persons with higher positions had somewhat lower percentages of the outcomes. Persons diagnosed with neoplasms or mental illness were more likely not to return to work than those with other diagnoses, and these groups are likely to have suffered from long-term sickness absence (>3 months) before onset of medical rehabilitation. Nevertheless, overall rates of negative work-related outcomes are rather low in this sample, supporting the aim of this extensive social and labour market policy in Germany.

In table 3, adjusted RRs of the three outcomes according to SEP are given. If measured by income and education, social inequalities in these associations are rather consistent and statistically significant, even in the estimates of the fully adjusted models. The group with missing data on education deviates from this pattern and seems to contain a socially disadvantaged group. Those with higher secondary education display slightly better outcomes than those with post-secondary education. In fact, their level of school training is higher, but does not contain the duration of vocational training of the latter group. Associations according to occupational position are less consistent. Overall, the strength of associations is rather modest, as RRs were mostly below a level of 2 (with exception of ‘low work ability’). Relatively strongest relationships were observed for education as an SEP measure, and they were particularly pronounced with regard to ‘low work ability’, as exemplified in a RR of 2.38 for the group with lower secondary education compared to the one with tertiary education (95% CI: 2.26–2.51).

Table 3

Associations of three SEP indicators with ‘work inability’, ‘failed return to work’ and ‘disability pension’ (N=245 514)

	Low work ability				Failed return to work				Disability pension
	Model 1		Model 2		Model 1		Model 2		Model 1		Model 2
	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI	RR	95% CI
Education: unknown	2.35	2.23–2.48	1.40	1.32–1.48	1.90	1.84–1.97	1.34	1.30–1.39	1.79	1.67–1.92	1.42	1.32–1.54
Education: lower secondary	2.38	2.26–2.51	1.42	1.34–1.50	1.73	1.68–1.79	1.25	1.20–1.29	1.66	1.55–1.78	1.38	1.28–1.49
Education: higher secondary	1.64	1.44–1.87	1.24	1.09–1.42	1.33	1.22–1.45	1.12	1.03–1.22	1.39	1.15–1.69	1.22	1.01–1.48
Education: post-secondary/ short-cycle tertiary	1.75	1.67–1.84	1.23	1.16–1.29	1.46	1.42–1.50	1.17	1.14–1.21	1.61	1.51–1.72	1.38	1.29–1.48
Education: tertiary	Reference		Reference		Reference		Reference		Reference		Reference
Occupation: unskilled, manual.	2.27	2.14–2.41	1.53	1.44–1.63	1.60	1.54–1.66	1.20	1.15–1.24	1.52	1.40–1.64	1.10	1.01–1.20
Occupation: skilled, manual	2.00	1.89–2.12	1.48	1.39–1.57	1.47	1.42–1.53	1.18	1.13–1.22	1.45	1.34–1.57	1.10	1.01–1.20
Occupation: unskilled, non-manual	2.10	1.98–2.23	1.39	1.31–1.48	1.62	1.56–1.67	1.18	1.14–1.23	1.61	1.49–1.75	1.15	1.06–1.25
Occupation: skilled, non-manual	1.42	1.34–1.50	1.19	1.12–1.26	1.23	1.19–1.28	1.09	1.05–1.13	1.59	1.47–1.72	1.31	1.21–1.42
Occupation: highly skilled	Reference		Reference		Reference		Reference		Reference		Reference
Income: low	2.12	2.08–2.17	1.86	1.82–1.91	1.77	1.74–1.79	1.63	1.61–1.65	1.54	1.50–1.59	1.54	1.49–1.59
Income: middle	1.55	1.51–1.58	1.42	1.39–1.45	1.33	1.31–1.35	1.27	1.25–1.29	1.35	1.31–1.40	1.33	1.29–1.37
Income: high	Reference		Reference		Reference		Reference		Reference		Reference
N	245 514		245 514		245 514		245 514		245 514		245 514

Notes: Relative risks (RRs) and 95% confidence intervals (CI) for three measures of longer-term work ability, based on a Poisson analysis, adjusted for age, sex, diagnosis, repeated rehabilitation and sickness absence before rehabilitation. Model 1 is estimated separately for the indicators of SEP. In Model 2 the effect of the respective SEP indicator is adjusted for the remaining indicators.

As evident from figure 1, absolute differences are nevertheless relevant, in particular in case of ‘failed return to work’ and ‘education’, as mentioned above.

Discussion

In this study, associations of three established measures of SEP with indicators of longer-term work ability following medical rehabilitation were observed in a large sample of insured persons in Germany. These associations followed a social gradient with higher probabilities of work ability among those in higher positions. The outcome indicators reflected three stages of persons’ trajectories following medical rehabilitation, first, work ability at the end of medical rehabilitation, assessed by a physician, second, return to work during the first year after rehabilitation, as evident from registry data and third, absence of a disability pension during a follow-up period up to 7 years. Although increases of RRs were relatively small, they were consistent in terms of income and education, but less so in terms of occupation. ‘Low work ability’ was the indicator with relatively strongest socioeconomic differences. To our knowledge, this is one of the first studies that analyzed social inequalities of three consecutive outcomes assessing work ability among persons who underwent medical rehabilitation.

The findings confirm previous knowledge obtained from other high- income countries. Opportunities of returning to work following hospitalization were shown to follow a social gradient among persons who were hospitalized for distinct chronic diseases.^4,6,8 The same was true for disability pensions due to medically certified diseases. For instance, in a comparative analysis of seven longitudinal aging studies from Finland, France, UK and USA, RRs of health-related work exit (mostly disability pensions) among individuals with lowest level of education or occupational class were three- to four times as high as those of individuals in the highest social class.¹⁵ Moreover, detailed studies from Finland demonstrated particularly high social inequalities of disability pension among persons with coronary heart disease,¹⁶ musculoskeletal disorders and injuries,^9,17 but less so among persons with mental disorders.¹⁷ Yet, different from this study, these investigations used one single workability-related outcome measure only, return to work or early work exit (disability pension). In view of reported differences of workability according to disease category, we repeated all calculations for the major diagnostic groups that were documented as prerequisites of accessing medical rehabilitation. While, in general, socioeconomic differences in these subgroups were similar to those observed in the total study population, there was one obvious exception: the group diagnosed with neoplasms exhibited substantially larger social inequalities of workability, compared to the remaining groups (results not shown). As access to medical rehabilitation is open to every entitled and diagnosed person within the German retirement insurance system, social selection into hospitalization is unlikely to exert a significant effect. The same holds true for treatment outcomes of rehabilitation, given standardized diagnosis-specific in-hospital treatment programmes. We cannot exclude the possibility that physicians’ assessments of workability before hospital discharge were biased towards more favourable results among patients with higher social standing.¹⁸ Yet, it is unlikely that this bias accounts for the observed social inequalities of our consecutive outcome measures. In this context, it is important to mention that we adjusted our analyses for the impact of disease severity on social differences in workability outcomes, using three proxy measures of this important confounding factor. Comparing the three outcome measures, social inequalities were strongest in case of ‘low work ability’ and weakest in case of ‘disability pension’. In this latter case, selective early mortality could partly explain this finding, but there was no evidence of socially selective in mortality in our sample. Concerning the observed marked social differences in return to work, differential job opportunities available in the labour market may be as important as restricted personal capabilities.

This study has several strengths and limitations. The strengths of this report are, first, the large study population of a nationally representative sample of persons who underwent medical rehabilitation in a specialized clinic, second, the availability of three consecutive outcome indicators with relevance to workability, retrieved from administrative records and from physicians’ assessments and third, the combined analysis of three relevant measures of SEP, education, occupational class and income. The long follow-up period up to 7 years, from hospital discharge up to the fact of being or not being granted a disability pension, is considered a further strength. Yet, several weaknesses are also obvious. Due to lack of more precise data, the validity of our proxy measures of disease severity before admission to a rehabilitation clinic is limited. Given reported social gradients of disease severity,¹⁹ we cannot rule out its significant effect on reported social inequalities of all three outcome indicators. Further limitations of administrative data concern the absence of information on treatment quality in rehabilitation clinics and on the quality of working conditions among persons who returned to work. This latter gap of knowledge is particularly critical given a strong impact of poor physical and psychosocial quality of work on opportunities of return to work and on risks on health-related work exit.^20,21 Income was restricted to the mean yearly wage or salary of the registered person and did not include household income or wealth. It may therefore not represent an accurate estimation of people’s financial situation. Furthermore, it should be taken into consideration that civil servants and self-employed persons generally are covered by different pension systems and, therefore, are excluded from our analysis, with only few exceptional cases¹¹ and that persons with a higher education and a higher occupational status are underrepresented in our analysis. Finally, it should be mentioned, that there are alternative explanations for the observed inequalities in work ability. For those who work in occupations with high physical demands it could be more difficult to return to work. Likewise, the labour market is more receptive for persons who worked in high skilled occupations. Lastly, financial incentives for return to work can differ between income groups.

In conclusion, this study based on administrative data documents social inequalities of three consecutive indicators related to work and health in a large representative sample of middle-aged to early old age men and women who underwent medical rehabilitation due to occurrence of a chronic disease. The lower their educational degree, income or occupational position, the higher was their probability of being unable to return to work and of leaving work due to disability pension. Interestingly, these inequalities persisted despite of extensive investments in universal access to medical and vocational rehabilitation in Germany. On the one hand, this suggests that policy measures should not only focus on universal access to health care and health promotion, and that other policy measures are equally important, such as more targeted active labour market policy. On the other hand, this points to increased efforts needed to promote healthy working and living conditions among less privileged occupational groups.

Acknowledgement

We furthermore thank the scientific advisory board of the underlying project (Prof. Andreas Müller, University of Duisburg-Essen; Dr Kerstin Hofreuter-Gätgens, Techniker Krankenkasse; Dr Wolfgang Bödeker, Epicurus) for helpful comments.

Funding

This research was supported by the Gesellschaft für Rehabilitationswissenschaften (GfR) NRW e. V. [Society for Rehabilitation Sciences] (Grant number: GFR 16001). It is also conducted in the frame of the German initiative ‘labor market participation at older ages’ which is financed by the ‘Fund for the Future’ of the Ministeriums für Innovation, Wissenschaft und Forschung des Landes NRW [Ministry of Innovation, Science and Research, North Rhine-Westphalia] from 2016 to 2019. The authors are responsible for the content of the publication.

Ethical approval

All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. The Ethics Committee of the Duesseldorf University Clinic approved the study and confirmed that it complies with the above mentioned guidelines (Study Nr. 6162R).

Conflicts of interest: None declared.

Key points

• Chances for work ability following medical rehabilitation from chronic disease are significantly lower among persons with low compared to those with high SEP.

• This holds true in the German system of medical rehabilitation despite its universal access and provision of standardized treatment.

• Efforts to strengthen work ability following rehabilitation should target primarily people in lower SEPs.