Selective Nonparticipation Research Articles

Prediction of (Non)Participation of Older People in Digital Health Research: Exergame Intervention Study.

BackgroundThe use of digital technologies is increasing in health care. However, studies evaluating digital health technologies can be characterized by selective nonparticipation of older people, although older people represent one of the main user groups of health care.ObjectiveWe examined whether and how participation in an exergame intervention study was associated with age, gender, and heart failure (HF) symptom severity.MethodsA subset of data from the HF-Wii study was used. The data came from patients with HF in institutional settings in Germany, Italy, the Netherlands, and Sweden. Selective nonparticipation was examined as resulting from two processes: (non)recruitment and self-selection. Baseline information on age, gender, and New York Heart Association Functional Classification of 1632 patients with HF were the predictor variables. These patients were screened for HF-Wii study participation. Reasons for nonparticipation were evaluated.ResultsOf the 1632 screened patients, 71% did not participate. The nonrecruitment rate was 21%, and based on the eligible sample, the refusal rate was 61%. Higher age was associated with lower probability of participation; it increased both the probabilities of not being recruited and declining to participate. More severe symptoms increased the likelihood of nonrecruitment. Gender had no effect. The most common reasons for nonrecruitment and self-selection were related to physical limitations and lack of time, respectively.ConclusionsResults indicate that selective nonparticipation takes place in digital health research and that it is associated with age and symptom severity. Gender effects cannot be proven. Such systematic selection can lead to biased research results that inappropriately inform research, policy, and practice.Trial RegistrationClinicalTrial.gov NCT01785121, https://clinicaltrials.gov/ct2/show/NCT01785121

Nonparticipation Selection Bias in the MOBI-Kids Study.

MOBI-Kids is a 14-country case-control study designed to investigate the potential effects of electromagnetic field exposure from mobile telecommunications devices on brain tumor risk in children and young adults conducted from 2010 to 2016. This work describes differences in cellular telephone use and personal characteristics among interviewed participants and refusers responding to a brief nonrespondent questionnaire. It also assesses the potential impact of nonparticipation selection bias on study findings. We compared nonrespondent questionnaires completed by 77 cases and 498 control refusers with responses from 683 interviewed cases and 1501 controls (suspected appendicitis patients) in six countries (France, Germany, Israel, Italy, Japan, and Spain). We derived selection bias factors and estimated inverse probability of selection weights for use in analysis of MOBI-Kids data. The prevalence of ever-regular use was somewhat higher among interviewed participants than nonrespondent questionnaire respondents 10-14 years of age (68% vs. 62% controls, 63% vs. 48% cases); in those 20-24 years, the prevalence was ≥97%. Interviewed controls and cases in the 15- to 19- and 20- to 24-year-old age groups were more likely to have a time since start of use of 5+ years. Selection bias factors generally indicated a small underestimation in cellular telephone odds ratios (ORs) ranging from 0.96 to 0.97 for ever-regular use and 0.92 to 0.94 for time since start of use (5+ years), but varied in alternative hypothetical scenarios considered. Although limited by small numbers of nonrespondent questionnaire respondents, findings generally indicated a small underestimation in cellular telephone ORs due to selective nonparticipation.

Adjusting for selective non-participation with re-contact data in the FINRISK 2012 survey.

A common objective of epidemiological surveys is to provide population-level estimates of health indicators. Survey results tend to be biased under selective non-participation. One approach to bias reduction is to collect information about non-participants by contacting them again and asking them to fill in a questionnaire. This information is called re-contact data, and it allows to adjust the estimates for non-participation. We analyse data from the FINRISK 2012 survey, where re-contact data were collected. We assume that the respondents of the re-contact survey are similar to the remaining non-participants with respect to the health given their available background information. Validity of this assumption is evaluated based on the hospitalisation data obtained through record linkage of survey data to the administrative registers. Using this assumption and multiple imputation, we estimate the prevalences of daily smoking and heavy alcohol consumption and compare them to estimates obtained with a commonly used assumption that the participants represent the entire target group. When adjusting for non-participation using re-contact data, higher prevalence estimates were observed compared to prevalence estimates based on participants only. Among men, the smoking prevalence estimate was 28.5% (23.2% for participants) and heavy alcohol consumption prevalence was 9.4% (6.8% for participants). Among women, smoking prevalence was 19% (16.5% for participants) and heavy alcohol consumption was 4.8% (3% for participants). The utilisation of re-contact data is a useful method to adjust for non-participation bias on population estimates in epidemiological surveys.

A Novel Approach to Assessing Professionalism in Preclinical Medical Students Using Multisource Feedback Through Paired Self- and Peer Evaluations

ABSTRACTPhenomenon: Professionalism is integral to the role of the physician. Most professionalism assessments in medical training are delayed until clinical rotations where multisource feedback is available. This leaves a gap in student assessment portfolios and potentially delays professional development. Approach: A total of 246 second-year medical students (2013–2015) completed self- and peer assessments of professional behaviors in 2 courses following a series of Team-Based Learning exercises. Correlation and regression analyses were used to examine the alignment or misalignment in the relationship between the 2 types of assessments. Four subgroups were formed based on observed patterns of initial self- and peer assessment alignment or misalignment, and subgroup membership stability over time was assessed. A missing data analysis examined differences between average peer assessment scores as a function of selective nonparticipation. Findings: Spearman correlation demonstrated moderate to strong correlation between self-assessments completed alone (no simultaneous peer assessment) and self-assessments completed at the time of peer assessments (ρ = .59, p < .0001) but weak correlation between the two self-assessments and peer assessments (alone: ρ = .13, p < .013; at time of peer: ρ = .21, p < .0001). Generalized estimating equation models revealed that self-assessments done alone (p < .0001) were a significant predictor of self-assessments done at the time of peer. Course was also a significant predictor (p = .01) of self-assessment scores done at the time of peer. Peer assessment score was not a significant predictor. Bhapkar's test revealed subgroup membership based on the relationship between self- and peer ratings was relatively stable across Time 1 and Time 2 assessments (χ2 = 0.83, p = .84) for all but one subgroup; members of the subgroup with initially high self-assessment and low peer assessment were significantly more likely to move to a new classification at the second measurement. A missing data analysis revealed that students who completed all self-assessments had significantly higher average peer assessment ratings compared to students who completed one or no self-assessments with a difference of –0.32, 95% confidence interval [–0.48, –0.15]. Insights: Multiple measurements of simultaneous self- and peer assessment identified a subgroup of students who consistently rated themselves higher on professionalism attributes relative to the low ratings given by their peers. This subgroup of preclinical students, along with those who elected to not complete self-assessments, may be at risk for professionalism concerns. Use of this multisource feedback tool to measure perceptual stability of professionalism behaviors is a new approach that may assist with early identification of at-risk students during preclinical years.

Using Community Feedback to Improve Community Interventions: Results From the Deep South Network for Cancer Control Project.

The Deep South Network for Cancer Control (DSNCC), initiated in 2000, is a dual-state, community-based participatory research infrastructure composed of academic and community partners committed to reducing cancer disparities among underserved African Americans in 12 designated counties of the Alabama Black Belt and the Mississippi Delta, 2 historically underserved areas of the country. Local residents trained as Community Health Advisors as Research Partners implemented a 3-tier community action plan (CAP) focused on promoting cancer screening, physical activity, and nutrition. Breast, cervical and colorectal cancer screening, healthy eating habits, and physical activity levels increased among many, but not all, African American women in the 12-county DSNCC coverage area. Seeking to improve our reach to include participants who reported they had never heard of the DSNCC or participated in the CAP, we conducted in-depth conversations with community residents about reasons for selective nonparticipation and ways to improve participation in the DSNCC community health interventions. Three patterns and their associated themes described ways to improve the penetration of CAP strategies and tailor them to effectively reach underserved African Americans in the intervention counties. We conclude with lessons learned for future interventions.

Developing Professionalism via Multisource Feedbackin Team-Based Learning

CGEA 2015 CONFERENCE ABSTRACT (EDITED) A Novel Approach to Assessing Professionalism in Preclinical Medical Students Using Paired Self- and Peer Evaluations. Amanda R. Emke, Steven Cheng, and Carolyn Dufault Construct: This study sought to assess the professionalism of 2nd-year medical students in the context of team-based learning. Background: Professionalism is an important attribute for physicians and a core competency throughout medical education. Preclinical training often focuses on individual knowledge acquisition with students working only indirectly with faculty assessors. As such, the assessment of professionalism in preclinical training continues to present challenges. We propose a novel approach to preclinical assessment of medical student professionalism to address these challenges. Approach: Second-year medical students completed self- and peer assessments of professionalism in two courses (Pediatrics and Renal/Genitourinary Diseases) following a series of team-based learning exercises. Assessments were composed of nearly identical 9-point rating scales. Correlational analysis and linear regression were used to examine the associations between self- and peer assessments and the effects of predictor variables. Four subgroups were formed based on deviation from the median ratings, and logistic regression was used to assess stability of subgroup membership over time. A missing data analysis was conducted to examine differences between average peer-assessment scores as a function of selective nonparticipation. Results: There was a significant positive correlation (r = .62, p < .0001) between self-assessments completed alone and those completed at the time of peer assessment. There was also a significant positive correlation between average peer-assessment and self-assessment alone (r = .19, p < .0002) and self-assessment at the time of peer assessment (r = .27, p < .0001). Logistic regression revealed that subgroup membership was stable across measurement at two time points (T1 and T2) for all groups, except for members of the high self-assessment/low peer assessment at T1, who were significantly more likely to move to a new group at T2, χ2(3, N = 129) = 7.80, p < .05. Linear regression revealed that self-assessment alone and course were significant predictors of self-assessment at the time of peer assessment (Fself_alone = 144.74, p < .01 and Fcourse = 4.70, p < .05), whereas average peer rating, stage (T1, T2) and academic year (13–14, 14–15) were not. Linear regression also revealed that students who completed both self-assessments had significantly higher average peer assessment ratings (average peer rating in students with both self-assessments = 8.42, no self-assessments = 8.10, self_at_peer = 8.37, self_alone = 8.28) compared to students who completed one or no self-assessments (F = 5.34, p < .01). Conclusions: When used as a professionalism assessment within team-based learning, stand-alone and simultaneous peer and self-assessments are highly correlated within individuals across different courses. However, although self-assessment alone is a significant predictor of self-assessment made at the time of assessing one's peers, average peer assessment does not predict self-assessment. To explore this lack of predictive power, we classified students into four subgroups based on relative deviation from median peer and self-assessment scores. Group membership was found to be stable for all groups except for those initially sorted into the high self-assessment/low peer assessment subgroup. Members of this subgroup tended to move into the low self-assessment/low peer assessment group at T2, suggesting they became more accurate at self-assessing over time. A small group of individuals remained in the group that consistently rated themselves highly while their peers rated them poorly. Future studies will track these students to see if similar deviations from accurate professional self-assessment persist into the clinical years. In addition, given that students who fail to perform self-assessments had significantly lower peer assessment scores than their counterparts who completed self-assessments in this study, these students may also be at risk for similar professionalism concerns in the clinical years; follow-up studies will examine this possibility.

Differences in participation rates and lessons learned about recruitment of participants--the European Health Examination Survey Pilot Project.

In the 1980s, participation rates in health interview and health examination surveys were around 80% while now they are around 50-60%. There is also evidence that non-participation is selective. Low participation rates and selective non-participation may cause bias to our survey results based on participants alone. We aim to increase knowledge on cultural differences in acceptance and feasibility of different recruitment methods. The European Health Examination Survey Pilot Project, conducted in 2009-2012, included pilot surveys in 12 countries among people aged 25-64 years. Information about recruitment methods and participation rates in these surveys was collected. Participation rates ranged from 16% to 57% for men and from 31% to 74% for women, where in most surveys women had higher participation rates than men. A variety of recruitment and promotion methods were used to obtain as high participation rates as possible. Combinations of phone calls, invitation letter and home visits were used to recruit invitees. Obtaining valid phone numbers for survey invitees was difficult in several countries. Incentives, websites and promotion in local media were used to promote the surveys. The European Health Examination Survey Pilot surveys showed that obtaining a participation rate above 50% for a representative population sample is possible but it requires hard work and a well-planned recruitment strategy. Recruitment methods used in one country may not be possible to use in another country due to cultural norms and national regulations.

The influence of non-response in a population-based cohort study on type 2 diabetes evaluated by the Swedish Prescribed Drug Register

Bias due to non-response in observational studies may lead to false risk estimates. We evaluated potential selective non-response in Stockholm Diabetes Prevention Program (SDPP) using a population-based drug register. A cohort of 12,952 men and 19,416 women, aged 35-56 years, was screened for diabetes and family history of diabetes (FHD). Response rate: 79% in men and 85% in women. Of the 4,209 men and 6,916 women that were invited, 70 and 68% participated in a baseline health examination including oral glucose tolerance test. A follow-up was performed in 79% of male and 70% of female baseline participants. We used data from the prescription register to estimate absolute risks and ORs for drug-treated diabetes in the non-response/non-participation groups. At both screening and baseline steps absolute risks of drug-treated diabetes were equal in non-participants and participants. Adjusted ORs were 0.9-1.0. At follow-up, absolute risks were higher among non-participants than participants, men 6.2/4.4% and women 2.6/1.6%, adjusted ORs 1.4 (0.9-2.3) and 1.5 (0.9-2.4), suggesting selective non-participation. Further analyses of FHD, smoking, physical activity, socioeconomic position and psychological distress demonstrated no previous false risk estimates for type 2 diabetes. However, for body mass index, there were indications of overestimation in women. We conclude that bias from non-response was not present at screening or baseline steps, suggesting that diabetes prevalence and risks may be estimated from a population-based cohort study with high attendance rate such as the SDPP. However, follow-up data should be treated with some caution, since the sample may have become biased.

Correcting HIV Prevalence Estimates for Survey Nonparticipation Using Heckman-type Selection Models

HIV prevalence estimates from population-based surveys are vulnerable to selection bias if HIV status is missing for a proportion of the eligible population. Standard approaches, such as imputation, to correct prevalence estimates for selective nonparticipation assume that data are "missing at random." These approaches lead to biased estimates, if unobserved factors are associated with both survey participation and HIV status. We use Heckman-type selection models to test and correct for selection on unobserved factors (separately for men and women) in the 2007 Zambia Demographic and Health Survey, in which 28% of the 7146 eligible men and 23% of the 7408 eligible women did not participate in HIV testing. Performance of these models depends crucially on selection variables that determine survey participation but do not independently affect HIV status. We identify 2 highly-plausible selection variables that are statistically significant determinants of survey participation: interviewer identity, and visit on the first day of fieldwork in a survey cluster. HIV-positive status was negatively correlated with consent to test in men (ρ = -0.75 [95% confidence interval = -0.94 to -0.18]), but not in women. Adjusting for selection on unobserved variables substantially increased the HIV prevalence estimate for men from 12% (based on measured HIV status alone) and 12% (based on imputation) to 21%. In addition, the adjustment for selection substantially changed the estimated effects of HIV risk factors. Studies of HIV prevalence and risk factors based on surveys with substantial nonparticipation should routinely use Heckman-type selection models to correct for selection on unobserved variables.

Obtaining active parental consent for school‐based research: a guide for researchers

Schools increasingly require researchers to obtain active parental consent for students to participate in health research. We sought to identify effective strategies for the recruitment of child research participants through schools. A search of Medline, PsycINFO, Educational Resources Information Center, ProQuest 5000 and the Cochrane Library electronic databases was conducted for the period 1988 to 2008. The review found evidence that the following strategies may be effective in enhancing participation rates: 1) promotion of the research to school principals, teachers, parents and students; 2) dissemination of study information using methods allowing direct contact with parents (i.e. telephone or face-to-face); 3) provision of incentives to teachers, students and at a class level; 4) making reminder contacts; and 5) having a member of the research team co-ordinate and closely monitor the recruitment process. Application of these strategies should reduce the risk of non-response and other biases that result from selective non-participation. Further randomised controlled trials of these and other strategies are required to strengthen the evidence base.

Study Participation and Nonresponse in a Population of Adolescents and Adults with Operated Congenital Heart Disease (GUCH Patients)

A group of patients after surgery of congenital heart defects was examined as to whether participants and nonparticipants (refusers, nonresponders, moved away, and deceased subjects) differed in terms of the type of congenital heart disease, the type of surgery (curative, reparative, and palliative), age and gender. A group of 698 subjects between 15 and 45 years were invited to participate in a study where a standardized interview was combined with a series of medical examinations. Finally, 361 patients participated, 121 explicitly refused, 92 did not respond, 91 had moved away, and 33 had died. Comparing participants and nonparticipants at the level of bivariate analyses using type of surgery, type of congenital malformation, gender, and age did not yield statistically significant differences. The final analysis using multivariate logistic regression revealed that individuals who underwent palliative treatment, the most severe type of surgery, had the same likelihood of not participating than patients with curative treatment. Classifying patients by the type of heart defect did not reveal group differences in the likelihood to participate. No differences for gender and age emerged. The analyses have shown that a worse health status may not lead to nonparticipation. As a consequence, we may conclude that selective nonparticipation because of the measures considered may not have occurred, and the likelihood for obtaining biased results can be considered as low.

Deviant Research: Deception, Male Internet Escorts, and Response Rates

Some social scientists who study deviant groups assume that it is necessary to use deception in order to avoid the bias associated with the decision of some subjects to not participate in such studies. Selective non-participation can lead to low response rates and, thereby, adversely affect the representativeness of the sample. This study directly tests for a differential in response rates to data gathered from an e-mail request for information on male-for-male escort fees sent by a sociologist with data gathered from an e-mail that appeared to be from a potential client asking for rates. Response rates for the social science and “client” e-mails are 15% and 60%, respectively.

Sampling irregularities, or why the present estimate for risk of HIV casual transmission is probably an underestimate

Updating the list of casual transmission studies yielded a minor revision of the recently reported estimate of the risk of HIV casual household transmission, estimating it as 0.3% per year of contact. A meta-analysis of the studies used to calculate this risk indicated that the estimated risk is probably an underestimate for four main sampling irregularities: (a) The studies employed samples which were based on selective non-participation of subjects including refusals and unlocatable subjects at average rates of 36% and 14%, respectively; (b) only 17% of the studies reported full numerical details of the population studied; (c) the studies were based on unjustified exclusions of subjects by the investigators at an average rate of 31%; (d) the sampling of studies cited in the transmission studies is consistent with the hypothesis that the studies were biased against reporting cases of casual transmission. In the 13% of the studies that reported full details of exclusions, the average rate of all exclusions combined was 84%. Since it is likely that cases of casual transmission were mostly included among the exclusions, this may have resulted in an unrepresentatively low frequency of casual infection among the 16% that were left to be studied.