280 likes | 408 Views
“Maximising Return from Cohorts” Project: Prevention of Attrition Findings. Cara Booker, Ph.D. MRC SPHSU October 12, 2009. Overview. Systematic Review Survey of Retention Among MRC Cohort Studies Preliminary Findings Future Analyses. Systematic Review. Objectives
E N D
“Maximising Return from Cohorts” Project: Prevention of Attrition Findings Cara Booker, Ph.D. MRC SPHSU October 12, 2009
Overview • Systematic Review • Survey of Retention Among MRC Cohort Studies • Preliminary Findings • Future Analyses
Systematic Review • Objectives • To determine the effectiveness of retention strategies in improving response rates in prospective population-based cohort studies. • To identify possible characteristics (i.e. locality, age, etc.) of studies that may have affected the retention of cohort members.
Population-based cohort study: “Any well-defined population defined by geographic boundaries, membership or occupation”.Szklo M. Population-based cohort studies. Epidemiologic Reviews. 1998;20(1):81-90. Exclusion Criteria: Clinical or non-clinical trials Non-population-based cohort studies Cohorts with record linkage as the only method of follow-up Studies which only looked at effectiveness of tracking methods Definition, Inclusion & Exclusion Criteria • Inclusion Criteria: • At least one wave of follow-up data collection in which the participant was personally contacted by the study • Retention rates were reported • Some description of the retention methods was available
Review Methods • Electronic Database Search • 13 databases searched • 5 search terms • Manual searches of bibliographies • Internet and website searches for technical reports and internal documents • Contact with Principal Investigators and other study personnel for access to unpublished manuscripts and grey literature
Total papers found from search (n=17,210) Excluded: Trials & non-population-based cohort studies (n=12,596) Excluded: Recruitment (n=3,701) Potentially relevant prospective population-based cohort studies (n=913) Excluded (n=623) No information on strategies (n=577) No information on follow-up (x=46) Papers retrieved for further evaluation (n=290) Working Papers /Handbooks/Users Guide (n=5) Unpublished papers (n=1) Book chapter (n=1) Excluded from review (n=265) No evaluation (n=245) Reviews (n=13) Tracing methods only (n=7) 28 Studies in 32 Papers Flowchart
Results • 28 Studies were identified • Countries • 15 in the US • 5 in the UK • 3 in Canada • 2 in Sweden • 1 in Norway, Spain & Australia • 11 were randomized trials • Response rates ranges from 34% - 98%
Incentives • Randomized Trials (n=8) • Incentives associated with increased response rates • Average increase ranged from 4% to 27% • Type of incentive (i.e. gift, money, cash card) did not appear to have an effect on response rate • Use of cash cards and percentage of checks cashed varied • Previous round responders had higher rates than non-responders • Non-randomized studies (n=2) • Comparing monetary to gifts, monetary had increased response • Non-monetary and informational incentives compared against each other produced no difference in response • Greater value of incentive great response, less money spent overall
Reminder Letters/Calls • Randomized Trials (n=1) • Higher response among participants with second reminder of phone call • Non-randomized studies (n=8) • Reminder letters increased response • However compared to a second questionnaire there was lower response from the reminder letters • Overall, use of multiple reminders or methods of data collection increased response • Average initial response rate was 61% • The average increase of response rate via • Reminder letters was 17% • Reminder calls was 11%
Multiple Posting/Calls/Visits • 8 studies posted multiple questionnaires, 1 had multiple visits and 2 with multiple calls • Posting additional questionnaires increased response by an average of 15% • Costs increase with subsequent posting • Multiple visits to schools increased response by 34% • More call attempts appeared to increase response rates
Alternative Data Collection Method • 10 studies offered alternative methods • Postal studies that offered telephone interviews increased response by an average of 5% • Face-to-face interviews which offered alternatives increased response by 25% • Two studies started in a central location, one started with clinic visits • Telephone studies that sent postal questionnaires increased response by 1% • Costs are higher for postal vs. in-school questionnaires (costs of mail & project coordinator) and telephone interviews are more expensive than post (due to personnel costs, and tracing costs)
Other • 2 Randomized Trials • Length of questionnaire • Received higher response with short form of questionnaire (not significant) • Postal method • Randomized type of envelope and certified mail vs. UPS • Certified mail had better response • Handwritten envelopes had better response than other types of envelopes
Study Characteristics • Different designs, sample sizes, reporting etc make it difficult to address our secondary aims. Broad pattern of: • Lowest response rates in US-based studies • Higher response rates in studies started in the 1980s and 1990s • In general, participants who were ethnic minorities, younger and of lower SES had lower response regardless of retention method • Higher education associated with earlier response • Postal studies with males only had higher response rates than mixed and female only studies • In mixed gender studies females had higher response rates in 8 of 11 studies. • Proportion of males increased with increasing contacts
Conclusions • Incentives were the only strategy that had a clear and positive association with response rates • Multiple mailings of questionnaires and reminder letters also appear to increase response • Alternative data collection methods had minimal increases for postal questionnaire studies, but large increases for in-person interview studies • Use of multiple strategies increases overall response rate
Discussion/Limitations • Very few cohort studies with analyses of retention methods • Reporting of attrition is not standardized • Often have some mention in various manuscripts detailing characteristics of attriters, which may vary by wave of data collection • More often don’t have technical reports, or detailed manuscript about the strategies used the characteristics of the attriters across the study • Calculation of response rates is also difficult due to eligibility criteria, definition of denominator and booster samples
Recommendations • Pilot studies or sub-samples to evaluate retention strategies • Focus more expensive strategies to non-responders • Most initial responders will do so without the need for incentives or other more vigorous reminders • Explore effectiveness and ethical issues associated with internet searches and use of social networking sites in relation to participation not just for tracking
Survey of Retention Among MRC Cohort Studies • Objectives • To examine what retention methods are associated with reductions in attrition • To explore what other study design features, if any, influence minimisation of attrition • Sample • Drawn from studies which are housed in one of the 12 MRC Population Health Research Network (PHSRN) units • Thirty-eight studies were identified
Data collection Conducted between December 2007 and August 2008 Reminder methods 1 Reminder telephone call 3 Reminder emails Questionnaire Contents General study design issues Recruitment Consent issues Compensation/Incentives/Reimbursement Tracking of study members Attrition Methods of Survey
Retention, Tracking, Tracing... • Retention Strategies • Used to increase response at time of data collection • Tracking/Tracing Strategies • Used to find participants between methods of data collection • Ambiguous Strategies • Often used between data collection but often used to increase loyalty & not specifically to track/trace
General Study Design • 24 studies responded (75% response rate) • 7 were dropped from analyses • One study had two distinct samples, increasing analysed studies to 25 • 4 settings • Data presented here is from 18 studies
Data Collection • How studies collected data • Face-to-face: Over 90% • Post: 70% • Telephone: 40% • Internet: 10% • Studies collected between four and eight different types of data
Retention Strategies • 6 studies allocated funding for retention • 10 studies offered incentives • 1 study evaluated use of incentives • Unconditional gift voucher found to be the most successful • Retention Methods • 50% of the studies used 2-6 different retention methods • Most commonly used • Email, newsletters, newspapers/magazines • GPs/CCs, schools • Key Leaders/Gatekeepers, administrative/supervisory bodies, parent assessments
Tracking Methods • Tracking/Tracking Methods • Minimum number of methods used was 1 & 14 was maximum • Most used: • Change of address cards, GP records, routinely registered events, stable address of close relative • Ambiguous methods • 80% provided findings • 40% provided holiday cards & 20% provided birthday cards
Strategies: Numbers, Barriers & Successes • Successful Retention Strategies • Family/friend contact, change of address card and telephone/directory assistance • Barriers to retention • Residential mobility, disinterest in issues covered by study, incorrect address and people too busy to join study.
Further Analyses • Analyze the impact of retention methods and different data collection methods on retention • What retention methods are the most significant in reducing attrition? • What is the impact on attrition when more invasive data collection methods are used (i.e. Vene-puncture, tissue collection, etc.) • Use a multi-level modelling approach • Waves nested within studies • Wave variables include: retention methods, data collection methods • Study level variables include: setting, average age of sample at baseline, gender of sample, year study started
Future Analytic Methods • Multivariate modelling • Combine methods into larger categories • i.e. Radio + Internet = Media • Examine the effects of different sample population characteristics on subsequent retention rates • Data collection types • Age • Gender • Study setting