Baseline participation in a health examination survey of the population 65 years and older: who is missed and why?

In recent decades, population-based health surveys have been facing decreasing participation rates [1]. Especially health studies of the older population are plagued by high non-response rates [e.g., 2‐4]. This could compromise the validity of study results. At the same time valid estimates of health status, health risks and health care needs in the population 65+ years are urgently needed due to population aging.

Studies of non-participation conducted up to the 1990ies have often not included or sufficiently reported results for individuals 65+ years [e.g., 5‐7], or left out the oldest old, i.e. those 80/85+ years [e.g., 8, 9]. The number of studies dealing with non-participation in health studies of older adults has increased since the 1990ies [2‐4, 10‐27], although many of these studies have focused on varying subsets of older persons with specific health problems, e.g. falls [4, 13], respiratory health [15], visual impairment [21] or rheumatoid arthritis [24].

The majority of health studies of older persons that investigated non-participation found non-participants are less well educated, have lower income, and live more often in residential care than participants [2‐4, 12, 14, 15, 21, 25, 27]. However, regarding sex, marital status and subjective health status, findings have been inconsistent. For example, married individuals were found to participate less often than [17, 26], more often than [3, 4, 16, 25, 26] and to the same extent than non-married individuals [18, 20, 22, 26, 27]. Immigration background was rarely considered; one Dutch study reported no difference in response rates according to first language [18].

Although qualitative studies have generated lists of potential reasons for non-participation in intervention studies among older individuals, the frequency of single reasons could not be quantified [13, 23]. Reasons for non-participation of the older population appear to have changed over time. Studies from the 1960ies and 1970ies noted negative opinions about the health care system and health research studies in general [11, 25]; since the 1990ies, ill health and lacking time or interest have been the predominant reasons for non-participation [4, 8, 10, 12, 16, 19, 20, 26].

The aims of our study were twofold: (1) to assess baseline differences between participants and non-participants in a population register-based health examination study of adults 65+ years (including the oldest old), and (2) to analyze reasons for non-participation at baseline.

The sample is part of the research project ‘Operationalizing Multimorbidity and Autonomy for Health Services Research in Aging Populations’ (OMAHA). The project was conducted as part of the German collaborative research initiative on health in older populations supported by the Federal Ministry of Research and Education (Germany). The project was approved by the local ethics committee at Charité – Universitätsmedizin Berlin (EA2/066/08) and was conducted in compliance with data protection and privacy regulations, as requested by the Federal and Berlin Offices for the Protection of Data. All procedures performed were in accordance with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed written consent was obtained from all participants.

This register-based study of adults 65 years and older aimed to estimate baseline response biases and to highlight the diversity of non-participants. We applied a three-step approach to collect information and differentiated between study participants, non-participants who at least answered a standardized health questionnaire (NP+), non-participants who actively declined to provide any information (NP-), and non-responders (NR). Information on main study characteristics (age, sex, non-German citizenship as an indicator of immigration background, living in a socially deprived area) and nursing home residence was available for the full sample. At the second level, information on education, living arrangements, health status, health-related quality of life, and need for assistance was available for a total of 683/1308 or 52.2 % of the net sample, including study participants and NP+. Finally, detailed information on reasons for non-participation was obtained for a total of (786/1009) or 77.9 % of all non-participants including 183 non-responders for whom this information could be obtained retrospectively.

As expected, the proportion of eligible persons who participated in the interview and examination was small (299/1308). Compared to the net sample, persons with non-German citizenship, persons 85 years and older, those living in deprived neighborhoods, and nursing home residents were underrepresented among study participants. Non-German citizenship was the single most important independent determinant of non-response or non-participation with decline to provide any health information.

In accordance with others [e.g., 20, 26], non-participants were a heterogeneous group. Ill health, limited German language proficiency, dementia frailty and being in the hospital ranked among the top 10 reasons, along with restraints to participate in scientific studies in principle or in this particular study and lack of interest or time.

Compared with participants, NP+ lived more often in residential care and reported lower health-related quality of life; they were also less likely to report at least one chronic disease or polypharmacy. One possible explanation for these seemingly contradictory results may be lower health service utilization among non-participants compared with participants. This could reduce the probability to receive a medical diagnosis or treatment despite functional limitations. Less education, as in our study, could be an underlying determinant and act as a barrier in health care utilization in this subgroup [38]. More research is needed to explain and verify this finding. However, we cannot rule out that measurement and mode effects caused these findings because the wording of the questions on chronic disease and need for support varied slightly between the two groups. In addition, participants took part in a brown bag medication review. They could have been more aware of their medications intake than NP+.

NP- and NR belonged more often to disadvantaged groups (non-German citizenship; living in deprived neighborhoods) compared with participants at baseline. Socially disadvantaged individuals are at higher risk of diseases [e.g., 39]. In longitudinal studies, higher rates of morbidity [40‐42] and mortality [4, 20, 26, 40, 42] have been reported for non-participants vs. participants. Subgroup analyses found especially worse outcomes for those non-participants who were too ill and for NR compared with participants [26, 40]. Only eight participants died within the first year of follow-up of our study [see 28]. Unfortunately, it was not possible to conduct a mortality follow-up of the entire study cohort (i.e. the subgroups of non-participants).

Obviously, there are difficulties in recruiting representative baseline samples of the population 65+ years for population register-based health studies. In particular, individuals 85+ years and those from disadvantaged subgroups are missing from the studies. In Germany as in other high-income countries, persons 80+ years are the fasted growing group of the population [43]. Currently, the risk of poverty among older people is relatively low, but socioeconomic projections indicate, that poverty rates among older people might increase in future cohorts [44]. In order to provide valid estimates of morbidity, functional capacities, health risks and health care needs in the population 65+ years, it is necessary to develop specific recruitment strategies tailored to the oldest old and to underrepresented groups; existing efforts do not appear to be sufficient. As a first example, a brief description of our study was offered in several languages to overcome language barriers, but this description was rarely requested. As a second example, only 55 individuals used the home visit offer. Possibly, tailoring recruitment strategies to older persons and including gatekeepers, such as family physicians, home care nurses or social workers could be helpful [19, 45]. However, in health studies aimed at the population 65 years and older at large, register-based sampling strategies may not be effective at all to achieve representative and sufficiently large samples of subgroups that are difficult to reach. Sampling strategies therefore need reconsideration. Multiple sampling frames may be useful to estimate a small set of key health indicators for the older population at large and to collect additional information relevant to specific subgroups also using proxy information [19, 46]. Register-based samples could be complemented by additional samples drawn at places where sufficiently large numbers of individuals can be approached in an atmosphere of trust and care, such as nursing homes, home nursing care networks, family physicians, dentists, adult day care centers for seniors or regular meetings for older persons organized by churches.

To obtain a better understanding of the moderate total participation rate in our study at baseline, the analysis of reasons for non-participation is helpful. Of all non-participants, no interest was the main reason for non-participation, followed by ill health, which is in line with the literature [8, 10, 12, 16]. In contrast to previous studies [20, 26], we found that only a marginal percentage of non-participants described themselves as ‘too healthy’ (n = 7). Although a group of ‘too healthy’ individuals might exist across countries, its percentage and influence on bias could vary. The frequency of reasons differed by non-participant subgroup. ‘Refusal to participate in scientific studies on principle’ and ‘no interest’ were more frequently cited among NR and NP-. In contrast, ‘being too ill’ was more often reported by NP+, which is in line with our observation of lower health-related quality of life in this group compared to study participants. The living situations of older individuals are heterogeneous and complex. When planning health studies including older individuals, this circumstance should be considered, and field workers should be provided with appropriate strategies.

The major strength of our study is that we were able to estimate baseline non-response bias and to highlight the diversity of non-participants by applying a three-step approach to collect information. Some limitations apply to our study. First, we obtained a total baseline participation rate of 52 % and a participation rate for the complete study protocol of only 23 %. A number of factors are likely to have contributed to this finding: (a) the application of very few exclusion criteria, e.g., language problems and dementia were not exclusion criteria; (b) the inner-city resident sample, others found worse participation rates in urban vs. rural areas [e.g., 3, 47‐49] - especially in inner cities [50]; and (c) the oversampling of the oldest old, who have the lowest participation rates [see also 15]. Second, some of the reasons for non-participation were obtained retrospectively. Akhtar [11] reported that only 30 % indicated the same reason for non-participation one year later. However, individuals in this previous study were interviewed once by a health visitor and once by a physician, which might have added to this effect and which was not the case in our study. Third, the diagnosis of dementia was not verified by medical records; only proxy information was given.

Our analyses on baseline non-participation in a health examination study among individuals 65+ years had some advantages. First, we obtained complete register-based information on demographics for the total sample. Therefore, our subgroup analysis also included a non-responder group. Second, the majority of all non-participants (78 %) provided a reason for non-participation, including even 61 % of NR. Third, only a small number of exclusion criteria were applied. In contrast to others [e.g., 20, 38], residential care, insufficient knowledge of the language, and having a terminal illness or dementia were not exclusion criteria. Fourth, the oldest old were included and even oversampled in the study. Finally, a home visit was offered as a standard procedure.