Search

Objective To investigate whether inverse probability of participation weighting (IPPW) using register data on sociodemographic and disease history variables can improve external validity in a cohort study with selective participation. Study Design and Setting We fitted various IPPW models by logistic regression using register data for the participants (n = 1,111) and nonparticipants (n = 1,132) of a Swedish cohort study. For each of six diagnostic groups, we then estimated (1) weighted disease prevalence proportions and (2) weighted cross-sectional associations (odds ratios) between sociodemographic variables and disease prevalence. Using register data on the remaining individuals of the entire study population of men and women aged 50-64 years (n = 22,259), we addressed how the choice of variables used for IPPW influenced estimation errors. Results Disease prevalence proportions were generally underestimated in the absence of IPPW but became markedly closer to population values after IPPW using sociodemographic variables. We found limited evidence of selective participation bias in association estimates, but IPPW improved external validity when bias was present. Conclusions IPPW using sociodemographic register data can improve the external validity of disease prevalence estimates in cohort studies with selective participation. The performance of IPPW for association estimates merits further investigations in longitudinal settings and larger cohorts.

Objectives Understanding attitudes to mental health issues can inform public health interventions. However, low response rates may contribute to nonresponse bias. In a randomized controlled trial we examined the effect of sending a prenotification postcard before the questionnaire and the placement of a short message on the survey envelope (teaser) on response rates to a mailed questionnaire about bulimia nervosa “mental health literacy”. Study Design and Setting Questionnaires were mailed to 3,010 adults (50.6% female and 49.4% male) aged 18–65 years. In a 2 (pre-notification–present; absent) by 2 (teaser–present; absent) design, questionnaire recipients were randomly allocated to the experimental strategies. Outcomes considered were response rate, response time, and cost. Results The overall response rate was 22.0%. Significant main effects showed higher response rates for the use of prenotification (present = 23.6%; absent = 20.3%), among female participants, and older participants. A significant interaction of teaser by gender indicated lower response rates for men who received the teaser but not for women. Older participants returned the questionnaire more promptly than younger participants. Females—but not males—who received the teaser were slower to return the questionnaire. Higher response rates for participants receiving the postcard compensated for increased costs, particularly for males and older participants. Conclusion Response rates to a mental health postal survey can be increased through the use of prenotification.

Objectives To display and discuss the reasons and consequences of length and time-dependent bias. They might occur in presence of a time-dependent study entry or a time-dependent exposure which might change from unexposed to exposed. Study Design and Setting Recalling the popular study of Oscar nominees and using a real-data example from hospital epidemiology, we give innovative and easy-to-understand graphical presentations of how these biases corrupt the analyses via distorted time-at-risk. Cumulative hazard plots and Cox proportional hazards models were used. We are building bridges to medical disciplines such as critical care medicine, hepatology, pharmaco-epidemiology, transplantation medicine, neurology, gynecology and cardiology. Results In presence of time-dependent bias, the hazard ratio (comparing exposed with unexposed) is artificially underestimated. The length bias leads to an artificial underestimation of the overall hazard. When both biases coexist it can lead to different directions of biased hazard ratios. Conclusion Since length and time-dependent bias might occur in several medical disciplines, we conclude that understanding and awareness are the best prevention of survival bias.

Objectives To examine nonparticipation in a survey by linking it with register information and identify potential nonresponse bias of inequalities in health status among aging people. Study Design and Setting Cross-sectional questionnaire survey with clinical checkups carried out in 2002 among persons born in 1926–1930, 1936–1940, and 1946–1950 in Southern Finland. The sample was linked with register information from Statistics Finland and analyzed in terms of participation and health status as measured by medicine reimbursements. Results Participation in the survey was more frequent among those who were older, female, married or cohabiting, higher educated and nonurban residents, and those with higher income and moderate health. Among nonrespondents, women were less healthy than men, whereas among respondents, the results were reversed. Among nonrespondents, better income was associated with unfavorable health. Poor health was generally more common among nonrespondents than respondents in several subgroups. Conclusion Differences in response rates were found in sociodemographic factors, health, and socioeconomic position. Favorable health was generally more frequent among respondents than nonrespondents. In particular, health inequalities by gender and income differed between respondents and nonrespondents. Thus, nonresponse may lead to bias in analyses of health inequalities among aging people.

Objective To assess whether a mixed-mode survey design reduced bias and enhanced methods commonly used to correct for bias (poststratification weighting). Study Design and Setting The data for this paper are from a study of 1,900 adult patients enrolled in a randomized controlled trial to promote repeat treatment for relapsed smokers at five Veteran's Affairs Medical Centers. A sequential mixed-mode design was used for data collection whereby the initial attempt was conducted using phone administration, with mail follow-up for nonresponders. Analyses examined demographic, health, and smoking cessation treatment seeking differences between telephone responders, mail responders, and nonresponders and compared the relative effectiveness of global vs. targeted poststratification weighting adjustments for correcting for response bias. Results The findings suggest (1) that responders to the additional survey mode (mail) did not significantly differ from responders to the first mode (phone) or nonresponders and (2) that poststratification weighting adjustments that take this additional information into account perform better than the standard global adjustments. Conclusions A mixed-mode design can improve survey representativeness and enhance the performance of poststratification weighting adjustments.

Objective To assess demographic predictors of response for specialists who were mailed a postal questionnaire on drug treatments for Alzheimer disease. Study design and setting The questionnaire was sent to 317 specialists in Quebec, Canada. Demographic variables included specialty, urban/rural practice, language, sex, and ‘number of years since receiving a medical license.’ The specialists were stratified according to responder status (i.e., respondent/nonrespondent), and respondents were further stratified as early or late responders. Variables differing between these strata were entered into logistic regression models to see if they predicted response. Results Only ‘female sex’ was a predictor of response in the respondent/nonrespondent analysis (OR 2.03; 95% CI 1.17, 3.53). No demographic variables predicted early or late response. Conclusion Researchers planning postal questionnaires should target male specialists with modified or additional mailings to increase response and reduce the potential for nonresponse bias. Caution should be exercised when comparing early vs. late responders as a means of assessing nonresponse bias.

Objectives To show how reweighting can correct for unit nonresponse bias in an occupational health surveillance survey by using data from administrative databases in addition to classic sociodemographic data. Study Design and Setting In 2010, about 10,000 workers covered by a French health insurance fund were randomly selected and were sent a postal questionnaire. Simultaneously, auxiliary data from routine health insurance and occupational databases were collected for all these workers. To model the probability of response to the questionnaire, logistic regressions were performed with these auxiliary data to compute weights for correcting unit nonresponse. Corrected prevalences of questionnaire variables were estimated under several assumptions regarding the missing data process. The impact of reweighting was evaluated by a sensitivity analysis. Results Respondents had more reimbursement claims for medical services than nonrespondents but fewer reimbursements for medical prescriptions or hospitalizations. Salaried workers, workers in service companies, or who had held their job longer than 6 months were more likely to respond. Corrected prevalences after reweighting were slightly different from crude prevalences for some variables but meaningfully different for others. Conclusion Linking health insurance and occupational data effectively corrects for nonresponse bias using reweighting techniques. Sociodemographic variables may be not sufficient to correct for nonresponse.

Our objectives were (1) to analyze the refusal bias associated with postal questionnaires and face-to-face interviews and (2) to compare responses elicited from a postal questionnaire with those provided by a face-to-face interview in subjects over 75 years of age. Our methods included a postal questionnaire sent to a representative sample of community-dwelling elderly individuals (n = 842). All subjects were also invited to participate in an in-home interview conducted by a trained nurse. One hundred and six subjects (12.6%) were nonrespondents to the postal questionnaire. These nonrespondents were more cognitively impaired, more disabled, and showed a higher 1-year mortality rate (13.2 versus 5.2%) than respondents. Late respondents who needed a reminder letter were similar to nonrespondents. One hundred and eighty-seven subjects (22.2%) refused to participate in the home interview. These nonparticipants were similar to the participants on age, sex, 1-year mortality rate, and responses given to all but two postal questions. Comparison of postal questionnaire responses and face-to-face responses revealed that agreement was substantial to moderate for factual information (kappa = 0.41 to 0.80) but poor for clinical information (depression, cognitive impairment, disability). A short index including age, sex, and six questions on disability was useful in identifying disabled subjects by a postal questionnaire. Our conclusion is that nonresponse bias is evidenced when postal questionnaires are utilized in the very elderly population and caution should be used when interpreting health data obtained by this technique. Refusal bias from face-to-face interviewing is less important in this population.

Although mailed surveys are an important component of epidemiological research, results from mailed surveys are often suspect because of poor response rates and the potential for nonresponse bias. Previous work has demonstrated that paying subjects to complete questionnaires increases response rates, but this work has not well addressed the impact of the timing of incentives on total cost, cost effectiveness, and response bias. We surveyed 400 university employees about health benefits. By random allocation, half received a check for $5 along with the mailed survey, and the other half received the promise of $5 on return of a completed survey. The response rates for both groups were about the same (64 and 59%, respectively), but prepayment was less expensive in aggregate and less expensive per response. In addition, we found that subjects with lower salaries were more likely to respond when paid in advance. We conclude that prepayment may actually be less expensive and more cost effective than payment on completion, but that the timing of payment may influence the profile of respondents.

Nonresponse to patient-reported outcome (PRO) questionnaires after surgery might bias the results. Our aim was to gauge the potential impact of nonresponse bias by comparing the outcomes of early and late responders.This study compares 59,565 early and 20,735 late responders who underwent a hip or knee replacement, hernia repair, or varicose vein (VV) surgery. The association between timeliness of response and three outcomes (the mean postoperative disease-specific PRO and generic PRO scores and the proportion reporting a fair or poor result) was examined by regression analysis.Late responders were slightly more likely to be young, nonwhite, deprived, and have a more severe preoperative condition with poorer quality of life. Late response was associated with a slightly poorer outcome in all four procedures although not statistically significant (P 0.05) for VV surgery. Adjusting for preoperative characteristics reduced the strength of the associations, but they remained statistically significant.As nonresponse to PRO questionnaires introduces slight bias, differences in response rates between hospitals should be taken into account when making comparisons so as to avoid overestimating the performance of those with lower response rates and failing to detect poor performing hospitals.

Non-response can bias studies of disease conditions but its influence has rarely been evaluated due to limitations of available data on the non-respondents. Because of a detailed medical record review for eligibility, we were able to compare clinical as well as demographic characteristics of respondents and non-respondents in a population-based study of diabetic complications among Rochester, Minnesota residents. Non-respondents were older, less well educated, more likely to be widowed and more often retired. They were much more likely to have cardiovascular disease at baseline, but the prevalence of retinopathy, nephropathy and diabetic neuropathy was similar for respondents and non-respondents, who were also comparable with regard to type of diabetes and diabetic therapy. While these findings indicate that data from the Rochester Diabetic Neuropathy Study can probably be generalized to diabetic residents generally, they reemphasize the potential for non-response bias in epidemiologic studies of clinical conditions, especially cardiovascular disease.

During 1981–1982, a cohort of elderly Japanese Americans living in Hawaii was recruited for an epidemiologic study of osteoporosis. The male subjects were simultaneously being examined for an epidemiologic study of heart disease. Baseline data collected from both the men and women at a previous heart disease examination were used to compare responders vs nonresponders. The target population for the osteoporosis study consisted of 1685 men and 1594 women. Of these, 1379 men (81.8%) and 1105 women (72.0%) participated in the initial osteoporosis examination. For each sex, nonrespondents were older and had higher systolic blood pressure levels than did the respondents. Male nonresponders had a higher stroke prevalence and more frequent recent use of vasodilator medicine. Female nonresponders had a less frequent history of having ever taken female hormones than did the responders. The responders and nonresponders were reasonably similar in other respects, as indicated by the comparison of more than 40 other variables. This suggests that nonresponse bias is probably not a major influence in exposure-disease associations in this osteoporosis cohort. We believe this is the first published report dealing with nonresponse characteristics in a cohort study of osteoporosis.

This paper describes a basic investigation of possible non-response bias in a mail survey. We compare characteristics of responders and non-responders to a mail survey of health outcomes among participants of a longitudinal study of physical activity, physical fitness, and health. Results indicate that, at the first clinic visit, the responders were essentially the same as the non-responders on personal health history and laboratory measurements, while reporting significantly more family history of specific chronic diseases (cardiovascular disease, hypertension, stroke). The male responders were younger and reported more positive health behaviors as well as better weight and treadmill times at the first clinic visit. These results suggest that both response groups were equally healthy at entry, and that individuals who had family members with certain chronic conditions and who had positive health behaviors were more likely to respond (participate) in this health-related survey. Differences of this type could affect interpretation of future analyses. This work illustrates the importance of incorporating methods to examine non-response into any epidemiologic study.

Objective To determine the degree to which mailed survey response rates, response times, and nonresponse bias are affected by questionnaire size and color. Study Design and Setting Questionnaires were mailed to a random sample of 2,000 Mayo Clinic patients in one of four size/color “test” groups. One thousand three hundred nine surveys were completed, approximately two-thirds in each group. Results A small (6⅛ × 8¼ in) questionnaire booklet on white paper had a higher response rate (68.4%) than a similarly sized questionnaire on blue paper (62.3%). A large (8¼ × 11 in) questionnaire on white paper had a 62.7% rate, whereas a large, blue questionnaire had a response rate of 68.6%. Median response times did not differ by questionnaire size/color. No evidence of differential nonresponse bias was observed across the four test groups. Conclusion This study supports the use of a small/white questionnaire format advocated by the Total Design Method advanced by Don Dillman at Washington State University. We observed a favorable response rate for a large questionnaire printed on blue paper; however, if time and resources are limited, use of a small/white questionnaire appears preferable.

The SF-12 Health Survey is a 12-item questionnaire that yields two summary scores (physical and mental health). Neither score can be computed when an item is missing. We explored imputation methods for missing scores for this instrument.Using data from a population-based survey, we tested several ways of imputing simulated missing data.Among 1250 participants, 118 (9.6%) had at last one missing SF-12 item. Missing data were more common among women, older respondents, non-Swiss nationals, and health service users. Among the 1132 respondents with complete data, replacement of any item with the mean population item weight yielded good results: the mean correlation between imputed and true score was 0.979 for both the physical and mental score. Results remained satisfactory when up to three of the six key items for each score (items that contribute predominantly to a given score), and any number of non-key items, were replaced by the mean. Application of this imputation algorithm to the original survey reduced the proportion of missing scores to1%. Respondents with incomplete surveys, hence imputed scores, had lower scores than respondents with complete data (physical score: 44.9 vs. 49.8, p0.001, mental score: 44.4 vs. 46.3, p=0.064).A simple imputation algorithm can substantially reduce the proportion of missing scores for the SF-12 health survey, and consequently reduce non-response bias.

Background and objective This study demonstrates the impact of survey nonresponse bias on conclusions from a mammography trial targeting a disadvantaged population. Methods The trial randomized 1558 women to three interventions designed to promote repeat mammography: mailed reminder (minimum group); mailed thank-you card, patient newsletters, and reminder (maximum group); and no mailings (control group). The primary outcome, repeat mammogram within 15 months, was assessed from administrative and phone survey data. Results Administrative estimates revealed a statistically significant difference of 7% between the maximum and control groups on the primary outcome. Survey estimates (response rate 80%) revealed no significant differences. The differences by data source were traced to a survey nonresponse bias. There was a statistically significant difference of 16% between the maximum and control groups among survey nonrespondents for the primary outcome, but there were no differences among survey respondents. Conclusion The findings reiterate that even a low survey nonresponse rate can bias study conclusions and suggest studies targeting disadvantaged populations should avoid relying solely on survey data for outcome analyses.

The objective of this study was to identify characteristics of non-respondents and late respondents to a mailed health survey. Persons who returned and those who did not return the questionnaire were compared using health insurance data, which indicated their age, sex, and health care expenditures in the previous year. Insurance and questionnaire data were used to compare early and late survey respondents and to compare categories of non-respondents. Questions covered use of health services, health status, and sociodemographic characteristics. Participants were members of health insurance plans in Geneva, Switzerland, 19-45 years old (n = 1822). Respondents (n = 1424) and non-respondents (n = 398) were of similar age and sex. The proportion of persons who had health care expenditures greater than zero Swiss francs (SFr) was higher among respondents (75%) than among non-respondents (69%, p = 0.03). Among non-respondents, expenditures of persons who explicitly refused to participate (2378 SFr) were higher than expenditures of persons who moved out of Geneva (1085 SFr) or who failed to return the questionnaire (1592 SFr, p = .02). Among respondents, being born in a Switzerland, having completed elementary school, having generated health care expenditures, and reporting good physical health were independent predictors of early response. In conclusion, low response rates to mailed health surveys may result in overestimating the utilization of health services. However, non-respondents did not constitute a homogeneous group, and the strength and even direction of non-response bias depended on the mechanisms of non-response.