Study design and study population

Data were derived from the cross-sectional “German Health Update” (GEDA), a national telephone health interview survey among adults in Germany. The GEDA study is part of the nationwide Health Monitoring System administered by the Robert Koch Institute, Berlin. The Robert Koch Institute is a federal institution within the portfolio of the German Federal Ministry of Health responsible for disease control and prevention. The aim of the regularly conducted GEDA surveys is to provide current data on population health, health determinants, and use of health services for national and European health reporting systems, health policies, and public health research.

Data were collected by means of a standardised computer-assisted telephone interview (CATI). Besides information on health status, health determinants, and use of health services, data on sociodemographic characteristics were gathered in the survey. Interviews took an average of approximately 30 minutes. The study was approved by The Federal Commissioner for Data Protection and Freedom of Information, and verbal informed consent was obtained from all of the participants in advance.

Health check attendance

Self-reported attendance at health checks was the outcome measure of this study. Respondents with an age of 35 years or older were initially asked whether they had ever attended the health check recommended by the statutory health insurance funds (“yes”, “no”, and “not sure”). Those persons who answered “yes” were asked a second question on attendance in the past 2 years (“yes” or “no”). We created a binary variable indicating health check attendance in the past 2 years (yes = 1, no = 0).

Predisposing factors

Independent variables were selected by referring to the Behavioral Model of Health Services Use [31]. Predisposing factors encompassed age (35–44, 45–54, 55–64, 65–74, and 75+ years), socioeconomic status, and marital status. Socioeconomic status was measured with an additive multidimensional index based on school and vocational education, occupational status, and net equivalent household income [37]. This index was classified as low (quintile 1), medium (quintiles 2–4), and high (quintile 5) by quintiles of the whole population. A more detailed description of the socioeconomic status index, its sub-dimensions, and its calculation can be found elsewhere [37]. Marital status was grouped as married and living together, separated/divorced/widowed, and single.

Enabling factors

In Germany, every person covered by statutory health insurance aged 35 years or older is eligible to have a biennial health check. However, the type of statutory health insurance was used in this study to consider possible differences in sickness funds’ strategies to inform their insurants about this early detection measure. We differentiated between local health insurance funds (Allgemeine Ortskrankenkasse, AOK), alternative health insurance funds (Ersatzkassen, EK), company health insurance funds (Betriebskrankenkassen, BKK), and others. The Oslo Three-Item Social Support Scale (OSS-3) was used to measure social support, which was assumed to be another factor enabling people to use health services. The OSS-3 sum score was classified as poor (3–8), intermediate (9–11), and strong (12–14) social support [38]. The type of residential area was considered to refer to the population density of the area in which the study participants lived. Based on classification by the German Federal Institute for Research on Building, Urban Affairs and Spatial Development, we assigned administrative districts to four types of areas: metropolitan, urban, rural with concentration signs, as well as rural and sparsely populated.

Need factors

We used information on behavioural risk factors for cardiovascular diseases and diabetes, self-perceived health, and use of outpatient care to assess respondents’ need for health checks. By this means, perceived and evaluated need components were considered. Obesity was assessed by body mass index (BMI) using the World Health Organization criteria (BMI ≥ 30 kg/m²) [39]. BMI was calculated using self-reported height and weight. Physical activity was ascertained with two questions on the number of days per week and minutes per day of physical activity, which was defined as starting to sweat and getting out of breath. We classified physical activity as active (≥150 minutes/week), slightly active (<150 minutes/week), and not active at all (no weekly physical activity). We assessed respondents’ smoking status with one question on their smoking behaviour (current smokers, ex-smokers, and never-smokers). Alcohol consumption was measured using the Alcohol Use Disorders Identification Test–Consumption (AUDIT–C) [40]. According to internationally used cut points [41], the AUDIT–C sum score (0–12) was classified as hazardous consumption (men: ≥5, women: ≥4), moderate consumption (men: 1–4, women: 1–3), and never-drinking (0). We considered fruit and vegetable consumption in the analysis because fruit and vegetable intake is inversely associated with the risk of cardiovascular diseases [42, 43]. Respondents were asked for the number of fruit and vegetable portions they take in daily. We categorised the number of portions per day as low (quintile 1), moderate (quintiles 2–4), and high (quintile 5). Self-rated health was measured on a five-point scale and dichotomised as very good/good or fair/poor/very poor. Use of outpatient care was assessed by the number of resident physician visits during the past 12 months, and was classified as low (quintile 1), medium (quintiles 2–4), and high (quintile 5).

Statistical analysis

Health check attendance rates were calculated by predisposing, enabling, and need factors. We used Pearson’s χ²-test and 95% confidence intervals (CIs) to examine statistically significant differences. Logistic regression models were fitted to estimate adjusted odds ratios (ORs) of attendance in the past 2 years with 95% CIs and p-values. In the first step, we computed age-adjusted ORs of attendance by each of the above-described factors in separate logistic regression models. In the second step, we used a full model containing all predisposing, enabling, and need factors as described above. All analyses were conducted separately for men and women to identify sex-specific factors associated with health check attendance and to prevent potential gender bias. Weighting factors were used to account for the sampling design and to adjust for distribution of the sample by sex, age, education, and region to match the German population. All statistical analyses were performed using STATA 12.0 survey data procedures.

Bivariate analysis

Age-adjusted analysis

Multivariate analysis

After mutual adjustment for all predisposing, enabling, and need factors (Table 3), men and women aged 45 years and older had an increased odds of attendance compared with those aged less than 45 years. The association between higher socioeconomic status and health check attendance remained after multivariate adjustment, and married people who lived with their spouse had higher odds of attendance than single persons. In women, having statutory company health insurance (BKK) was associated with an increased odds of using health checks compared with those having statutory local health insurance (AOK). Social support remained as a predictor of health check attendance, while the type of residential area was not related to attendance behaviour after multiple adjustment. Obesity did not predict attendance behaviour in men or women. Physical inactivity and smoking remained associated with a lower odds of attendance in both sexes. Alcohol consumption persisted as a predictor of using health checks only in women. In men, poorer self-rated health was related to lower attendance. Higher fruit and vegetable consumption, as well as higher use of outpatient care, remained associated with a higher odds of health check attendance in men and women after adjusting for all factors.

Comparison with previous research

The findings of the GEDA study are in line with previous studies from Germany and other countries showing that younger people are less likely to participate in health checks than older people [1, 29, 30, 44]. However, some studies did not find any relationship between age and participation [24, 45]. Moreover, our findings support evidence from previous German studies, which found that the uptake of prevention services varies with socioeconomic factors. Janßen et al. [17] and Klein et al. [19] conducted systematic literature reviews on socioeconomic differences in the use of (preventive) health services in Germany. Both reviews showed that socioeconomically disadvantaged groups are less likely to use preventive services than socioeconomically privileged groups. This relationship has also been found in other countries and different health care systems [1, 15, 16, 20, 21, 24, 46]. However, the above-mentioned reviews [17, 19] also showed that the pertinent evidence from Germany is more consistent for cancer screening programmes, check-ups for children, or primary prevention measures than for the statutory health check for adults [18, 26, 28–30]. This finding could be explained by methodological issues, such as the use of different indicators of socioeconomic status, or the fact that, in some studies on health check attendance in Germany, people with private health insurance had not been excluded from the analysis [26, 29]. The present findings add to previous evidence on social inequalities in health check attendance derived from the German GEDA study [4] that the socioeconomic patterning of attendance behaviour remains after adjustment for enabling factors, which were not considered in the preceding analyses.

Our findings are consistent with previous studies, which showed that married people are more likely to use preventive measures than single persons [1, 46, 47]. A possible explanation for this finding could be that living together with a partner is associated with a mutual “health monitoring” [48], which could result in increased attention to health or symptoms, and a higher participation in prevention measures.

Previous research has suggested that social support is a promoter for the uptake of early detection measures [14, 49, 50], which is also supported by our findings. Social relations might facilitate the use of secondary prevention services (e.g., friends taking people to the doctor [51] or neighbours undertaking tasks of daily living, such as caring for children or relatives, when people visit a health professional). Additionally, the prospect of not being alone after obtaining a diagnosis could reduce concerns about participating in a medical check-up. These assumptions should be investigated prospectively. While the present study considered general social support, future studies on health care-seeking behaviour could also assess specific support with reference to the use of health services.

Our findings are also consistent with previous studies on the role of need factors as determinants of health check attendance. Dryden et al. [1] showed in a literature review that non-attenders have a less favourable cardiovascular risk factor profile than attenders, especially smokers are less likely to attend health checks than non-smokers. Moreover, other studies have shown a facilitating effect of regular doctor visits on the use of early detection measures [15, 47, 52, 53]. This finding is consistent with our study, which showed a strong association between higher use of outpatient care and the uptake of health checks. Frequent consulters of primary care are probably more likely to receive invitations to secondary prevention services from health professionals than people outside the healthcare system. However, Dryden and colleagues [1] have also reported that ongoing contact with physicians might be a reason for not attending health checks. Because of this inconsistency between studies, further research on the relation between primary care use and health check attendance is required.

The present findings also support evidence suggesting that determinants of care-seeking behaviour differ between men and women, apparently in close connection with societal gender roles [54, 55]. Although there is a growing body of studies showing differences in care-seeking behaviour between men and women, the number of studies explicitly analysing the sex or gender differences in health care utilisation is still low [54]. In future research, a stronger distinction between sex as a biological and gender as a social category may help to better understand these differences.

The observed differences in women's health check attendance by the type of statutory health insurance may reflect potential differences in the way health insurance funds inform their insurants about health checks or encourage them to use this service. Possibly, information provided by company health insurance funds is more appealing to women than men.

Effects of health checks

Future studies also need to examine the effects of health checks on population health. A systematic review on the benefits and harms of general health checks indicated that such preventive measures do not reduce morbidity or mortality from disease, although the number of new diagnoses has increased [56]. Potential harms of health checks, such as overdiagnosis or negative psychological consequences for patients, have rarely been studied. The authors of the systematic review mentioned above acknowledged that many trials had methodological problems and most of the trials were old [56]. The oldest study was from the 1960s and the most recent was from the 2000s. Within this time span, there have been many changes in medical practice [57]. Therefore, the transferability of the results to current settings might be limited. Another systematic review and meta-analysis showed that health checks carried out in general practice are associated with statistically significant, but clinically small, improvements in surrogate outcomes, such as blood pressure and BMI [58]. Accordingly, beneficial effects of health checks are still controversial [59, 60]. To the best of our knowledge, no evidence is available yet concerning the effects of the statutory health check for adults in Germany. Currently, there is a lively and controversial debate on the benefits and harms of cancer screening programmes in Germany. However, the potential effects of the statutory health check seem to be barely on the agenda, although this early detection measure also may potentially lead to overdiagnosis and overtreatment. Therefore, population-based measures for early diagnosis, such as the statutory health check, should be supervised scientifically to monitor the expected effects on population health, as well as potential harms for the participants and the health system.

Methodological considerations

The strengths of this study include the use of up-to-date data from large national samples and the wide range of explanatory variables considered in the analyses. The large sample size also allowed separate analyses for men and women. Nevertheless, there are several limitations to our study. We used self-reported attendance at health checks as the outcome measure. Self-reports on the use of health checks may be biased by social desirability and by confusing other medical services with health checks. We accounted for the latter possibility by adding the category “not sure” to the response options (treated as missing data in the analysis). The low percentage of respondents who used this category (2.6%) indicated that potential bias due to uncertainty of respondents should be small. Moreover, the recall period of 2 years may have led to recall bias in the assessment of health check attendance. Social desirability might also have affected responses on questions regarding social support and health behaviour. This may have contributed to associations between these indicators and health check attendance [61].

Because of the cross-sectional nature of the study, no causal relationships could be established and there may have been potential bias related to reverse causality. If respondents who had attended a health check had a more favourable risk factor profile as a result of their health check participation, the effects of behavioural risk factors on attendance may have been overestimated. Furthermore, results may have been affected by selection bias due to underrepresentation of certain population groups in the sample, such as people with a migration background, because interviews were only conducted in German. For that reason, migration or ethnic background was not considered in the analyses. Other limitations arise from the selection and measurement of factors considered in the study. The type of residential area may only be a proxy measure for regional density of primary care physicians. This might have contributed to our finding that there was no regional variation in health check attendance. A study which examined the use of cancer screening in Germany, for instance, revealed that screening rates are higher in areas with higher physician density [62]. Psychological factors, such as health beliefs, and contextual factors, such as regional physician density, were not considered in the analysis. However, such factors are considered to determine health care use in the Behavioral Model of Health Services Use [31, 63]. The assessment of physical activity and its definition in the questionnaire may have been imprecise; weekly metabolic equivalents may have been a better indicator of physical activity. Moreover, non-behavioural need factors, such as family history of diseases or psychosocial stress, could not be controlled for. This may have led to residual or unmeasured confounding. Besides, it has to be considered that failing to see an association (beta error) may have been more likely in men than in women due to a lower number of men in the study population.