• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of amjphAmerican Journal of Public Health Web SiteAmerican Public Health Association Web SiteSubmissionsSubscriptionsAbout Us
Am J Public Health. 2002 March; 92(3): 445–450.
PMCID: PMC1447097

Subgroups of Refusers in a Disability Prevention Trial in Older Adults: Baseline and Follow-Up Analysis

Abstract

Objectives. This study explored differences between refusers and participants in a longitudinal study with extensive baseline and follow-up information.

Methods. Results of a trial comparing 791 participants and 401 community-residing older adults who refused to participate in a study concerning preventive home visits were examined. Information was collected from interviews, insurance records, and government files.

Results. Despite similarities in terms of age, sex, and self-perceived health at baseline, 3-year follow-up data indicated that refusers had a 1.58-fold higher risk of entering a nursing home than participants. There were additional differences between refusers and participants when refusers were categorized in 4 subgroups based on self-reported reason for refusal (too ill, too healthy, no interest, and other reasons).

Conclusions. Future studies should include follow-up data to allow comparisons between refusers and participants and should address the presence of multiple subgroups of refusers. (Am J Public Health. 2002;92:445–450)

Concerns about biases introduced by refusers in studies of the health of the elderly have been raised repeatedly.1 In descriptive studies, refusal or nonresponse can lead to selection bias if disease or exposure status is associated with selection of subjects.2,3 In randomized, controlled studies, refusal can decrease the generalizability of the findings if subgroups with characteristics associated with intervention effects are misrepresented. Both consequences can have serious ramifications.

In a systematic literature search, we found 13 published studies of health or health care use among older people that had as a main topic analysis of differences between refusers and participants.4–16 Of these 13 studies, 4 included both baseline and follow-up data,4–7 and 9 focused on baseline data alone.8–16 Several of the studies reported (mostly minor) differences in sociodemographic characteristics between refusers and participants. The most frequently reported differences were as follows: older age,5,10,12,14 lower educational level,8,10–13 and higher proportion of women among refusers than among participants.5,7

With regard to health, health behavior, and health care use, the studies reported variable findings. Several authors reported that refusers were worse in regard to health-related status (e.g., lower levels of self-perceived health8,12,15 and higher rates of mortality,6 cognitive impairment,6,9,13 health problems or risks,6,11,13,15 and health care use15) than participants. On the other hand, some studies reported that refusers had a lower prevalence of health problems or risks7,10,14 than participants. Overall, in terms of health status, most studies identified one subgroup of older individuals who were underrepresented in the study, and the presence of this group was usually interpreted as not having an effect on the study conclusions.

Several facts may account for these diverse research findings. First, intervention intensity and type of subject involvement varied between the studies and might have affected differences between refusers and participants. For example, one study6 showed that if individuals were asked to respond in a personal interview, there were only minor differences between refusers and participants, but if the same individuals were asked to return a mail questionnaire, refusers differed significantly from participants. Those who refused to return mail questionnaires had lower levels of cognitive functioning than participants, probably because cognitively impaired individuals have difficulties answering and returning such a questionnaire.

Second, the health risk factors targeted influenced participation. For example, individuals without vision problems were more likely to refuse to participate in a study of an ophthalmologic intervention,12 individuals without oral health problems were more likely to refuse to participate in a dental intervention program,10 and individuals who perceived few health benefits from seeing a geriatric health specialist were more likely to refuse to participate in a geriatric assessment study.16

Third, methodological factors might account for the differences in previous results. The studies varied in regard to the nature and completeness of the information collected. For example, none of the studies with follow-up information included combined data on demographics, health status, and health care use among refusers and participants. Thus, the diverse findings may have derived from the use of selected subsets of variables, thereby limiting the multidimensionality of the analyses.

Finally, completeness of information on refusers varied. For example, only 2 studies reported that the percentage of missing health status or health care use information was lower than 20% among refusers,5,7 and in most studies a proportion of the original population could never be contacted, thus increasing the number of potentially eligible persons with missing information.

In summary, our review of previous research on refusal to participate among older adults shows that these studies involved different methods, strategies, age groups, and interventions, making conclusions difficult. The discordant findings between the studies are partly explained by the intensity and type of intervention offered; however, there is no explanation as to why some studies show that refusers tend to be healthier than participants while others show that refusers tend to be sicker than participants. It has been suggested, but not confirmed by empirical data, that sick persons do not participate in preventive trials and that healthy persons do not participate in therapeutic trials.17 On the basis of this information, it is quite likely that refusers are a heterogeneous group of people who refuse for different reasons and are likely to differ in varying ways depending on reasons for nonparticipation.

To resolve the problems of previous studies, we analyzed the database of a recent randomized, controlled intervention trial for which we had detailed baseline and long-term follow-up information with little missing data on either participants or refusers.18 The completeness of this database allowed us to (1) determine whether conclusions based on comparisons of baseline data between refusers and participants differ from those drawn from baseline and follow-up data combined and (2) explore factors involved in refusal to participate to improve understanding of the unexplained health differences between refusers and participants revealed in previous studies.

METHODS

The study was approved by the institutional review committee at the University of Bern and was based on a secondary analysis of a randomized trial conducted during 1993 through 1997.18

Study Population

The study population was drawn from a health insurance list of community-living individuals who were 75 years and older and were residents of Bern. Because more than 95% of the city's inhabitants have health insurance, a list of the clients of a single health insurance company is fairly representative of the inhabitants of this area. Of a random sample of 1998 individuals generated from this list, 806 were excluded for the following reasons: 165 had died, 595 were living in nursing homes or board and care facilities, 17 had relocated, 17 did not speak German, and 12 reported that they had a terminal disease. Overall, among the eligible 1192 subjects, 401 (34%) declined and 791 (66%) agreed to participate in the study. The 791 participants were randomly allocated to intervention (n = 264) and control (n = 527) groups.

Available Data

During the recruitment telephone call, individuals were asked whether they would be willing to participate in a study aimed at helping people remain independent and stay in their own home as long as possible. Persons who were willing to participate signed an informed-consent form and completed a baseline interview within 2 weeks after recruitment. Those who refused to participate were asked to provide information on their selfperceived health status and living arrangements. Overall, 65 persons among the 401 refusers were not willing to provide this information; these data were available from the remaining 336 refusers. During the recruitment telephone call, refusers were also asked about their main reason for refusal; reasons were listed and later categorized into 4 main categories of refusal by 2 researchers working independently. The rate of interrater agreement for these categorizations exceeded 90%. In instances of disagreement, a third person was asked to categorize the reasons, and then a decision was made.

In the case of both refusers and participants, information on demographic characteristics, survival, and living status in the community, as well as on health care usage, use of home help–home care, and medication costs, was obtained from health insurance records. Numbers and types (general practitioners vs specialists) of physicians visited were extracted from health insurance billing records. Information about the economic status of refusers and participants was obtained from income tax files. In Switzerland, information on taxable income and assets is publicly available. This information was missing for 4 individuals (3 refusers, 1 participant) because no tax file could be found. In addition, information on receipt of financial assistance from the government was obtained from government files. These data were entered into a file that protected individual study participants from identification.

Statistical Analysis

Bivariate statistical comparisons between refusers and participants were based on the 2-group t test for continuous variables and on χ2 and Fisher exact tests for percentages. Cox proportional hazards regression was used in multivariate analyses of survival as well as of the probability of nursing home admission.19 In the latter analysis, deaths of community-living persons (i.e., those who died without residing in a nursing home) were treated as censored observations. Number of physician visits was calculated according to period at risk. Thus, data for subjects who died during the first year were included in first-year ambulatory care analyses by taking into account number of days survived. Subjects assigned to the intervention group were not included in the analyses of follow-up data, because the intervention had significant effects on health status and health care use outcomes.

RESULTS

Overall Comparisons

Table 1 [triangle] presents comparisons of baseline characteristics between participants and refusers. There were some statistically significant but relatively small differences between refusers and participants in regard to number of homeowners, number not seeing a physician before baseline, and number of different physicians seen for ambulatory care. Figure 1 [triangle] depicts the results of bivariate Cox analyses of survival and nursing home admissions among refusers vs participating control subjects. The mortality rate for refusers was significantly (P = .0005) higher than that for participants (survival analysis risk ratio [RR] = 1.53, 95% confidence interval [CI] = 1.20, 1.94). In a multivariate model including refusal status and baseline characteristics, the mortality risk ratio for refusers remained highly significant (RR = 1.49, 95% CI = 1.15, 1.93; P = .002). Refusers also had a significantly (Cox survival analysis; P = .03) higher risk of entering a nursing home than participants (RR = 1.58, 95% CI = 1.05, 2.36). In a multivariate analysis including refusal status and baseline characteristics, the risk ratio remained high (1.42, 95% CI = 0.92, 2.20) but became statistically nonsignificant (P = .11).

FIGURE 1
Survival probabilities for refusers (n = 401) and participating control subjects (n = 527) (a) and probabilities of nursing home admission for refusers (n = 401) and participating control subjects (n = 527) (b): Bern, Switzerland, 1993–1997.
TABLE 1
Comparison of Refusers and Participants at Baseline: Demographic, Socioeconomic, and Health Care Use Characteristics (Bivariate Analyses), Bern, Switzerland, 1993–1994

Comparisons Involving Subgroups of Refusers

Table 2 [triangle] presents the results for the 4 subgroups of refusers distinguished by selfreported reason for refusal. Those who reported that they were too healthy to take part generally indicated that they did not need help or that they were healthy and therefore believed it was not necessary to participate in a preventive program. At baseline, these “too healthy” individuals had better self-perceived health and lower rates of physician use than participants. Those who reported that they were too ill to participate generally were concerned that the program would be an unnecessary burden. These individuals had significantly worse self-perceived health and higher rates of physician use than participants. Those who reported that they had no interest were similar to participants in terms of selfperceived health but tended to be of lower socioeconomic status and to have lower rates of physician use. The remaining group of refusers, who indicated a variety of other reasons for refusal, had characteristics similar to those of participants.

TABLE 2
Comparison of Subgroups of Refusers and Participants at Baseline: Demographic, Socioeconomic, and Health Care Use Characteristics (Bivariate Analyses), Bern, Switzerland, 1993–1994

Table 3 [triangle] presents the comparison of follow-up data between refuser subgroups and participating controls. The “too healthy” subgroup tended to have a lower nursing home admission rate than participants but exhibited a similar 5-year mortality rate. The “too ill” subgroup had significantly higher rates of mortality and nursing home admissions than participants. The “no interest” group exhibited a significantly higher mortality rate than participants. The remaining group of refusers tended to have a higher nursing home admission rate than participants.

TABLE 3
Comparison of Subgroups of Refusers and Participanting Controls: Mortality Over 5-Year Follow-Up and Nursing Home Admissions Over 3-Year Follow-Up (Bivariate and Multivariate Analyses), Bern, Switzerland, 1993–1997

DISCUSSION

The finding of major mortality and nursing home admission rate differences between refusers and participants (Figure 1 [triangle]) was unexpected, because baseline comparisons between the 2 groups did not reveal significant differences in terms of age, sex, health status, or use of home help–home care. As shown in the multivariate analyses, this mortality difference could not be explained by baseline demographic or health-related differences between refusers and participants. Thus, our study demonstrates that, even in the presence of comparable baseline characteristics, follow-up health status measures may differ substantially between refusers and participants.

Refuser Subgroups

Analyses revealed that differences between refusers and participants varied substantially according to the former's reason for refusal. In some cases, the differences were in opposite directions. For example, the “too healthy” refuser subgroup had better self-perceived health than participants, while the “too ill” subgroup had worse self-perceived health. It is interesting to note that the aggregate comparison presented in Table 1 [triangle] did not reveal a significant difference in self-perceived health between refusers and participants, indicating that such subgroup differences may not be apparent in aggregate analyses. Similarly, the fact that the “too ill” subgroup had higher physician use rates and the “too healthy” and “no interest” subgroups had lower rates is not seen in the aggregate analysis of Table 1 [triangle].

The finding of at least 3 subgroups of refusers might best be explained by the concept of utility, defined as ratio of self-perceived cost of adherence to self-perceived benefit of adherence.16 One might hypothesize that the “too ill” group considered the cost of participating in a preventive intervention (i.e., taking the time to undergo a 2-hour interview and examination) too high relative to the self-perceived benefit. On the other hand, the “too healthy” group might have considered the benefit of the intervention too low relative to the cost. Finally, the “no interest” group might have been more skeptical toward health care use and would accept an intervention only if the ratio of cost to benefit were very low. This explanation based on personal utility is also compatible with previous studies indicating that the health risk factors targeted by the program in question influence subject participation, and it is supported by the findings of a study that examined the factors explaining adherence with a geriatric assessment program.16

The fact that previous studies identified only 1 subgroup of older persons exhibiting differences in health status between refusers and participants is probably explained by the limited availability of data. It is likely that overall differences in health status between refusers and participants in previous studies represented the net difference resulting from the underrepresentation of 2 or more subgroups, each with different health characteristics.

Implications

The present findings have implications for the conclusions of the primary study from which our analysis was derived.18 For example, the underrepresentation of individuals of better health status relative to the participants is relevant. It is likely that this subgroup would have benefited from the intervention, because the overall study showed that preventive home visits reduced the onset of disability in the subgroup of older persons at low baseline risk. Thus, the underrepresentation of the “too healthy” subgroup calls for better strategies in future preventive home visitation programs to ensure that groups of relatively healthy older adults can be reached. On the other hand, the finding that the “too ill” group was underrepresented in the study is probably not relevant, because the primary study showed that high-risk older adults do not benefit from preventive home visits.

Finally, the finding that the “no interest” group tended to refuse participation is again relevant. The expected intervention effect in this subgroup was unpredictable. One might hypothesize that the members of the “no interest” group tended to have lower adherence rates and thus would not have experienced favorable intervention effects, or, alternatively, they might have had a high number of previously undetected risk factors for disability that were potentially amenable to successful preventive interventions. However, because this was a relatively small subgroup (12% of eligible persons), it is unlikely that its underrepresentation influenced the overall project findings, which were based on the 67% of persons who agreed to participate in the study.

The generalizability of this study is limited in that it was based on a preventive intervention offered to a selected sample of individuals who were 75 years or older and were living at home in a suburban area. Although in the case of refusers some of the information collected via interviews was missing (Table 1 [triangle]), the present analysis was based on an extensive database with complete follow-up of both groups in terms of survival and nearly complete follow-up in terms of nursing home admissions and health care use. The amount of available information at baseline was limited, and it is therefore conceivable that there were undetected differences in health status between refusers and participants.

It is unlikely that the differences in follow-up data between refusers and participants were due to a residual beneficial intervention effect for the participants (as described earlier, participants were those assigned to the control group in the randomized study). Although control subjects underwent a multidimensional geriatric evaluation at baseline and the 2-year follow-up, they received usual care without recommendations or other interventions based on the evaluation.

This study has several other implications for future research. First, the findings might be used in developing strategies to minimize refusal rates among older adults. Because perceived utility seems to be the overarching concern in regard to study participation, the benefits of participation should be made clear in the recruitment process. These benefits should be substantial and visible, and personal costs (in terms of time and effort) should be as limited as possible. Second, our findings suggest that reasons for refusal should be recorded and reported in studies of older adults, because they might provide valuable insights into the types of underrepresented subgroups of nonparticipants. Third, the findings show that comparisons of baseline data between refusers and participants might significantly underestimate differences in health risk between the groups.

Finally, because several subgroups of refusers might be underrepresented in a single study, our findings indicate that simple overall comparisons of refusers with participants may mask true differences. For example, if those who perceive themselves as “too healthy” and those who perceive themselves as “too ill” are less likely to participate in a study, the average health status of refusers might be similar to that of participants; in fact, however, both groups are underrepresented. Therefore, multiple comparisons are needed to detect underrepresented subgroups of refusers in studies of older adults and, probably, individuals in other age groups as well.

Acknowledgments

This project was supported by grants from the Swiss National Science Foundation (4032-35637 and 32-52 804.97), Novartis Foundation for Gerontological Research and the Velux Foundation.

The assistance of B. Lüthi (data checking) is gratefully acknowledged. We would like to thank Robert E. Bjork, Arizona Center for Medieval and Renaissance Studies, Arizona State University, for his editorial review of the manuscript.

Notes

C. E. Minder and A. E. Stuck designed the study, planned the data analysis, and wrote the manuscript. T. Müller conducted the literature search and contributed to data management and manuscript revision. G. Gillmann performed the statistical analyses, constructed the graphs, and contributed to the revision of the tables and figures. J. C. Beck contributed to the design and interpretation of the study and to revision of the manuscript.

Peer Reviewed

References

1. Klesges RC, Williamson JE, Somes GW, Talcott GW, Lando HA, Haddock CK. A population comparison of participants and nonparticipants in a health survey. Am J Public Health. 1999;89:1228–1231. [PMC free article] [PubMed]
2. Cochran WG. Sampling Techniques.3rd ed. New York, NY: John Wiley & Sons Inc; 1977.
3. Kish L. Survey Sampling.New York, NY: John Wiley & Sons Inc; 1965.
4. Donald IP, Bulpitt CJ. The Gloucestershire Longitudinal Study of Disability: outcomes in nonresponders, responders, and subsequent defaulters. J Clin Epidemiol. 1998;51:1305–1310. [PubMed]
5. Ganguli M, Mendelsohn A, Lytle M, Dodge H. A follow-up comparison of study participants and refusers within a rural elderly population. J Gerontol. 1998;53A:M465–M470. [PubMed]
6. Hébert R, Bravo G, Korner-Bitensky N, Voyer L. Refusal and information bias associated with postal questionnaires and face-to-face interviews in very elderly subjects. J Clin Epidemiol. 1996;49:373–381. [PubMed]
7. Ives DG, Traven ND, Kuller LH, Schulz R. Selection bias and nonresponse to health promotion in older adults. Epidemiology. 1994;5:456–461. [PubMed]
8. Augustsson O, Eriksson BG, Rosenhall U, Rothenberg E, Wärne B, Steen B. The Johanneberg Study—a sociomedical survey in an urban elderly population, I: general presentation of the study including an analysis of non-response and identification of risk groups. Scand J Soc Med. 1994;22:283–292. [PubMed]
9. Launer LJ, Wind AW, Deeg DJH. Nonresponse pattern and bias in a community-based cross-sectional study of cognitive functioning among the elderly. Am J Epidemiol. 1994;139:803–812. [PubMed]
10. Locker D. Effects of non-response on estimates derived from an oral health survey of older adults. Community Dent Oral Epidemiol. 1993;21:108–113. [PubMed]
11. Matthis C, Schlaich C, Scheidt-Nave C, Raspe A, Raspe H. Die europäische Studie zur vertebralen Osteoporose (EVOS): Teilnahmebereitschaft und Selektionsverzerrung in Deutschland [European Vertebral Osteoporosis Study Group (EVOS): cooperation of participation behavior and selection bias in Germany]. Med Klin. 1998;93(suppl 2):18–25. [PubMed]
12. Muñoz B, West S, Rubin GS, Schein OD, Fried LP, Bandeen-Roche K. Who participates in population based studies of visual impairment? The Salisbury Eye Evaluation Project experience. Ann Epidemiol. 1999;9:53–59. [PubMed]
13. Norton MC, Breitner JCS, Welsh KA, Wyse BW. Characteristics of nonresponders in a community survey of the elderly. J Am Geriatr Soc. 1994;42:1252–1256. [PubMed]
14. O'Neill TW, Marsden D, Silman AJ. Differences in the characteristics of responders and non-responders in a prevalence survey of vertebral osteoporosis. Osteoporos Int. 1995;5:327–334. [PubMed]
15. Osler M, Schroll M. Differences between participants and non-participants in a population study on nutrition and health in the elderly. Eur J Clin Nutr. 1992;46:289–295. [PubMed]
16. Reuben DB, Posey E, Hays RD, Lim ME. Predictors of patient refusal to participate in ambulatory-based comprehensive geriatric assessment. J Gerontol. 1994;49:M209–M215. [PubMed]
17. McKee M, Britton A, Black N, McPherson K, Sanderson C, Bain C. Interpreting the evidence: choosing between randomised and non-randomised studies. BMJ. 1999;319:312–315. [PMC free article] [PubMed]
18. Stuck AE, Minder CE, Peter-Wüest I, et al. A randomized trial of in-home visits for disability prevention in community-dwelling older people at low and at high risk for nursing home admission. Arch Intern Med. 2000;160:977–986. [PubMed]
19. Kalbfleisch JD, Prentice RL. The Statistical Analysis of Failure Time Data.New York, NY: John Wiley & Sons Inc; 1980.

Articles from American Journal of Public Health are provided here courtesy of American Public Health Association

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

  • MedGen
    MedGen
    Related information in MedGen
  • PubMed
    PubMed
    PubMed citations for these articles

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...