Research Attitudes Questionnaire scores and retention in a recruitment registry

Abstract

Background

Recruitment registries are maximally effective when registrants are retained to the point of referral. The Research Attitudes Questionnaire (RAQ) has previously been shown to predict research participation behaviors, including Alzheimer's disease clinical trial completion.

Objective

To test the hypothesis that RAQ score is associated with retention behaviors in a local recruitment registry.

Methods

Using data from the UC Irvine Consent-to-Contact Registry, a recruitment registry that enrolls adults 18 years and older, we used logistic regression to quantify the association of RAQ score and the odds of first-year non-renewal. Covariates included demographic variables, comorbidities, and recruitment source. In longitudinal analyses, we used discrete proportional hazards and Cox proportional hazards models to quantify the relationship between RAQ score and time to non-renewal and time to active withdrawal, respectively.

Results

Among n = 4663 participants, we estimated that a 5-point higher baseline RAQ score was associated with a 15% lower odds of first-year non-renewal, after adjustment for potential confounding factors (OR: 0.85, 95% CI: (0.79, 0.92), p < 0.001). Older age and higher education were also associated with lower odds of non-renewal while Asian race, Hispanic ethnicity, and certain recruitment sources (e.g., doctor or friend referral) were associated with higher odds of non-renewal. Higher baseline RAQ and higher annually updated RAQ were both significantly associated with lower odds of non-renewal longitudinally. Age, education, and some recruitment sources, but not RAQ, were associated with active withdrawal.

Conclusions

Opportunities exist to identify predictors of registry retention behaviors and possible targets for intervention to improve related outcomes.

Keywords

Alzheimer's disease recruitment registries renewal research attitudes retention withdrawal

Introduction

Recruitment registries are tools designed to accelerate enrollment into clinical research, particularly preclinical Alzheimer's disease clinical trials.^1–3 Retention of participants who enroll in a registry is essential to ensure that adequate pools of potential participants are available for recruitment. Reports from large national registries, however, suggest that retention is challenging, with annual renewal rates in the range of 30–38%.^{1, 2, 4}

Identifying factors associated with dropout and retention in registries is essential for helping investigators better understand the causes of attrition, developing retention interventions, and engaging participants who are most likely to be lost to follow up. The Research Attitudes Questionnaire (RAQ) is a short, validated scale used to assess perspectives towards research participation.⁵ Higher scores on the RAQ indicate a greater value placed on medical research and have been shown to correlate with several measures of willingness to participate in research, increased study compliance, and higher propensity for study completion.^6–9

In this study, we examined whether RAQ scores were associated with retention behaviors in the UC Irvine Consent-to-Contact (C2C) Registry. The C2C is a local recruitment registry established in 2016 with the goal of improving recruitment to clinical research studies at UC Irvine.¹⁰ We tested the hypotheses that higher RAQ scores are associated with better registry retention.

Methods

Participants

Adults aged 18 years and older were eligible to enroll online or by phone into the UC Irvine C2C Registry.¹⁰ The study was approved by the UC Irvine Institutional Review Board. Participants provided informed consent electronically or via telephone, granting permission to be contacted about recruiting studies and to have their data used in additional projects such as this one. They were then asked to complete a survey of demographics, family history of chronic disease, health history, willingness to participate in research activities, and validated instruments assessing diet, physical activity, sleep, subjective memory complaints, and the RAQ. Data were captured through REDCap.¹¹

Primary and secondary outcomes

The primary response variable in the current study was an indicator of whether each participant failed to renew enrollment in the registry. We invited participants to renew their enrollment annually, including an option to update their contact information, demographics, and health history. At renewal, registrants repeated assessment of willingness to participate and some validated instruments, including the RAQ. An email invitation to renew enrollment was sent 344 days after initial enrollment or previous renewal, with 3 email reminders sent 7 days apart if renewal was not completed. The outcome of “renewal” at all timepoints was defined as providing a signature on the consent form. We included participants who completed the initial enrollment process in these analyses. If participants failed to renew, they were kept active in the registry for 3 years after enrollment, at which point they were removed by the research team. If participants indicated that they wished to withdraw from the registry by contacting the research team, they were withdrawn, and we defined this event as an “active withdrawal”, which was a secondary outcome for this study.

Covariates

Demographic variables collected included age, sex, race, ethnicity, and years of education. We collapsed and categorized race and ethnicity into mutually exclusive categories including Hispanic/Latino (any race), non-Hispanic (NH) Asian, NH Black, NH White, and other (which included NH American Indian or Alaska Native (n = 5), NH Native Hawaiian or Other Pacific Islander (n = 8), other (where the participant could further specify their race, n = 72), and multiracial (where the participant selected multiple responses, n = 103)). Years of education was collected as a continuous variable but truncated at 26 years for this study. Twenty-five participants reported > 26 years of education in which case a value of 27 years was imputed. Participants who identified their sex as “other” (n = 15), did not specify their race (n = 93), or declined to disclose their race (n = 48) were categorized as having missing data for those respective fields. Participants self-reported their current chronic comorbidities at baseline, which included diabetes, coronary artery disease, congestive heart failure, stroke, liver disease, kidney disease, hypertension, hypercholesterolemia, and emphysema. For analysis, we categorized the number of comorbidities into 0, 1, 2, and ≥3 among the above conditions. Recruitment method, i.e., the means by which participants learned about the registry, was captured at baseline and included email, community talk, postcard, a doctor's referral, a friend's referral, newspaper, TV, radio, online search, social media, other sources, multiple sources, and none reported. We combined recruitment through newspaper, TV, radio, online search, and social media into the category “media”, and combined other sources, multiple sources, and no sources reported into the category “other”.

Predictor of interest

At enrollment and at each renewal, participants were asked to complete the 7-item RAQ (Supplemental Table 1). The RAQ items assess general feelings about medical research, with 5-point Likert-type responses ranging from strongly disagree to strongly agree. Total RAQ score ranges from 7 to 35 points, with higher scores representing more positive research attitudes.⁵

Study cohort

The cohort for the current study included 4663 participants who enrolled between June 1, 2016 and August 4, 2022. Only subjects that entered a valid contact email address upon signing up for the registry were included. For first-year non-renewal and time to non-renewal models, we only considered participants who enrolled before May 11, 2021, to allow for 15 months (including a three-month grace period from final email reminder) to renew. For these models, the datasets included a subset of 4232 participants. The analysis of active withdrawal included n = 4663 registrants, with 330 active withdrawals.

Missing data

Figure 1 presents the data screening flow chart. Partially or completely missing RAQ data were observed for 267 (2.65%) of 10,062 longitudinally collected scores. Of these, 161 were missing a response for one of the seven RAQ questions, 18 were missing entries for two questions, 7 were missing entries for three to six questions, and 81 were missing entries for all seven questions. We did not impute missing baseline RAQ values (n = 111); those participants were not included in the final analysis. Missing longitudinal RAQ values (n = 147) were imputed using baseline or the most recent previous scores collected. Missing baseline values for time-invariant demographic covariates were backfilled with data entered at any renewal. In cases where we could not impute or backfill, participants were removed from statistical analyses (n = 262, 5.32%).

Figure 1.

Data screening flow chart.

Statistical methods

We used means and standard deviations to summarize continuous baseline covariates and frequencies and percentages to summarize discrete baseline covariates, with stratification by baseline RAQ score. The primary outcome was non-renewal within 15 months of enrollment. Secondary outcomes included time to non-renewal and time to active withdrawal from the registry.

To assess whether baseline RAQ was associated with first-year non-renewal, we utilized a logistic regression model. We adjusted for a priori identified potential confounding factors including baseline age, sex, years of education, race and ethnicity, recruitment method, and number of comorbidities. Estimated odds ratios along with corresponding 95% Wald-based confidence intervals and p-values were computed for each covariate included in the model.

To evaluate the relationship between RAQ and time to non-renewal, we used a discrete proportional hazards model to estimate the discrete hazard for non-renewal (no response to renew received within 15 months of previous enrollment/renewal) at a given year, conditional upon renewing at all prior years.¹² We modeled the RAQ in three ways: first only considering baseline RAQ score as the predictor of interest, then with time-varying RAQ (the most recently completed RAQ score prior to current measure of renewal) and the change in RAQ from the previous renewal year as predictors of interest. Each of these models were adjusted for the covariates described in the primary analysis. Estimated discrete hazard ratios for non-renewal along with corresponding 95% Wald-based confidence intervals and p-values were computed for each covariate included in the model.

We modeled the time to active withdrawal via a Cox proportional hazards model. Participants were censored at 15 months since enrollment or last renewal if there was no response throughout that period, as they did not get further reminder emails after this point. We again adjusted for all potential confounding factors that were included in the primary analysis described previously.

For each model, we reported p-values for likelihood ratio tests of the significance of the entire construct of each discrete variable with more than two categories. Residual diagnostics were performed for each regression model, with particular focus on the existence of influential observations and deviations from the proportional hazards assumption. No highly influential observations were identified in any of the analyses and no overt suggestions of a departure from the proportional hazards assumption were observed.

Finally, we conducted a sensitivity analysis in the active withdrawal model where we censored participants at 39 months since enrollment or last renewal, since they were removed from the registry if they did not renew for three years. The results from this sensitivity analysis, shown in Supplemental Table 2, did not qualitatively differ from the original analysis where we censored participants at 15 months.

Results

Table 1 depicts the sample baseline characteristics stratified by quartiles of baseline RAQ (for 4663 participants in the cohort, the quartile ranges were 7 to 26, 27 to 28, 29 to 32, and 33 to 35). With increasing baseline RAQ, participants generally had higher years of education, and fewer were recruited through email and doctor's referral. Age, sex, and comorbidity distributions were largely similar across these groups.

Table 1.

Descriptive statistics for the sample.

	Baseline RAQ
Covariate	[7, 26] (n = 1223)	[27, 28] (n = 1117)	[29, 32] (n = 1333)	[33, 35] (n = 990)
First-year Non-renewal
Renewed	462 (37.8%)	481 (43.1%)	647 (48.5%)	488 (49.3%)
Did not renew	663 (54.2%)	529 (47.4%)	551 (41.3%)	411 (41.5%)
<15 mo. follow-up	98 (8.0%)	107 (9.6%)	135 (10.1%)	91 (9.2%)
Age	56.2 (16.5)	57.4 (16.6)	57.9 (17.2)	57.4 (17.1)
Sex
Male	429 (35.1%)	403 (36.1%)	479 (35.9%)	338 (34.1%)
Female	794 (64.9%)	714 (63.9%)	854 (64.1%)	652 (65.9%)
Education	16.0 (2.5)	16.3 (2.8)	16.6 (2.8)	16.6 (2.9)
Race and Ethnicity
NH White	904 (73.9%)	861 (77.1%)	1052 (78.9%)	751 (75.9%)
Hispanic/Latino	137 (11.2%)	103 (9.2%)	109 (8.2%)	114 (11.5%)
NH Asian	105 (8.6%)	105 (9.4%)	103 (7.7%)	64 (6.5%)
NH Black	21 (1.7%)	16 (1.4%)	20 (1.5%)	14 (1.4%)
Other*	56 (4.6%)	32 (2.9%)	49 (3.7%)	47 (4.7%)
Recruitment Method
Email	645 (52.7%)	540 (48.3%)	618 (46.4%)	434 (43.8%)
Doctor	62 (5.1%)	44 (3.9%)	40 (3.0%)	39 (3.9%)
Community Talk	82 (6.7%)	102 (9.1%)	153 (11.5%)	115 (11.6%)
Media	166 (13.6%)	165 (14.8%)	217 (16.3%)	178 (18.0%)
Friend	82 (6.7%)	81 (7.3%)	103 (7.7%)	75 (7.6%)
Postcard	79 (6.5%)	79 (7.1%)	67 (5.0%)	46 (4.6%)
Other **	107 (8.7%)	106 (9.5%)	135 (10.1%)	103 (10.4%)
No. of Comorbidities
0	723 (59.1%)	672 (60.2%)	794 (59.6%)	573 (57.9%)
1	293 (24.0%)	268 (24.0%)	308 (23.1%)	237 (23.9%)
2	136 (11.1%)	121 (10.8%)	171 (12.8%)	124 (12.5%)
≥ 3	71 (5.8%)	56 (5.0%)	60 (4.5%)	56 (5.7%)

Table 1 displays summary statistics of 4663 participants, reporting mean (standard deviation) for continuous variables and N (%) for discrete variables. Participants with less than 15 months follow-up enrolled after 05/11/2021. * Race defined as “Other” included NH American Indian or Alaska Native, NH Native Hawaiian or Other Pacific Islander, multi-racial, or race other than those prespecified. ** Recruitment method defined as “Other” included sources not pre-specified, multiple sources, and no sources indicated.

First-year non-renewal

Figure 2 displays a forest plot describing the proportion of participants who did not renew at the first year for each quartile of baseline RAQ. We observed that less participants in the third and fourth quartiles did not renew (46.0% and 45.7%, respectively) than those in the first and second quartiles (58.9% and 52.4%, respectively) and those missing RAQ entries (58.5%).

Figure 2.

Displays a forest plot illustrating the proportion of participants who did not renew (estimate and 95% confidence interval (CI)) at the first year for each quartile of baseline RAQ and those who had missing baseline RAQ entries. The diamond and vertical line illustrate the estimate and 95% CI for the overall cohort.

Table 2 displays the results of the logistic regression model of the odds of first-year non-renewal. Overall, there were significant associations of baseline RAQ, age, education, race and ethnicity, and recruitment method with the odds of non-renewal. We estimated that the odds of first-year non-renewal were approximately 15% lower comparing groups of participants with 5 points higher baseline RAQ scores to those 5 points lower, after adjustment for all covariates listed in the table (OR: 0.85, 95% CI: (0.79, 0.92), p < 0.001). The odds of first-year non-renewal were estimated to be 9% lower comparing groups of participants 5 years older (OR: 0.91, 95% CI: (0.89, 0.93), p < 0.001). Hispanic/Latino participants were estimated to have a 35% higher odds of first-year non-renewal (OR: 1.35, 95% CI: (1.08, 1.69), p = 0.008) and NH Asian participants had an estimated 45% higher odds of first-year non-renewal (OR: 1.45, 95% CI: (1.13, 1.85), p = 0.003) relative to NH White participants. Compared to email, participants referred by a doctor, a media source, or through friends had an estimated 69% (OR: 1.69, 95% CI: (1.21, 2.34), p = 0.002), 29% (OR: 1.29, 95% CI: (1.07, 1.57), p = 0.008), and 43% (OR: 1.43, 95% CI: (1.11, 1.83), p = 0.006) higher odds of non-renewal, respectively.

Table 2.

Logistic regression model of baseline RAQ and first-year non-renewal.

		Unadjusted			Adjusted
Covariate	n	Est. OR	95% CI	p	Est. OR	95% CI	p
Baseline RAQ (5 pt)	4232	0.83	(0.77, 0.89)	<0.001	0.85	(0.79, 0.92)	<0.001
Age (5 y)	4232	0.90	(0.88, 0.91)	<0.001	0.91	(0.89, 0.93)	<0.001
Sex
Male	1509	Referent	-	-	Referent	-	-
Female	2723	1.08	(0.95, 1.22)	0.247	0.96	(0.84, 1.10)	0.544
Education	4232	0.93	(0.91, 0.95)	<0.001	0.94	(0.92, 0.97)	<0.001
Race and Ethnicity							0.003
NH White	3276	Referent	-	-	Referent	-	-
Hispanic/Latino	420	1.92	(1.56, 2.37)	<0.001	1.35	(1.08, 1.69)	0.008
NH Asian	318	1.77	(1.40, 2.24)	<0.001	1.45	(1.13, 1.85)	0.003
NH Black	59	1.85	(1.09, 3.15)	0.023	1.48	(0.86, 2.55)	0.156
Other	159	1.63	(1.18, 2.26)	0.003	1.23	(0.88, 1.72)	0.232
Recruitment Method							<0.001
Email	2153	Referent	-	-	Referent	-	-
Doctor	171	1.65	(1.20, 2.27)	0.002	1.69	(1.21, 2.34)	0.002
Community Talk	405	0.72	(0.58, 0.89)	0.002	0.98	(0.78, 1.23)	0.862
Media	580	1.22	(1.01, 1.47)	0.034	1.29	(1.07, 1.57)	0.008
Friend	296	1.35	(1.05, 1.72)	0.017	1.43	(1.11, 1.83)	0.006
Postcard	269	0.89	(0.69, 1.15)	0.377	1.08	(0.83, 1.40)	0.578
Other	358	1.25	(1.00, 1.57)	0.050	1.43	(1.13, 1.80)	0.003
Comorbidities							0.533
None	2482	Referent	-	-	Referent	-	-
1	1009	0.78	(0.67, 0.90)	0.001	0.98	(0.84, 1.14)	0.764
2	516	0.82	(0.68, 0.99)	0.039	1.08	(0.88, 1.33)	0.443
≥3	225	0.94	(0.72, 1.24)	0.665	1.18	(0.89, 1.58)	0.247

Table 2 displays results of the logistic regression model of the odds of first-year non-renewal associated with baseline RAQ (by 5 points) and adjustment covariates.

Time to non-renewal

Table 3 displays results of the discrete proportional hazards model for time to non-renewal. Baseline RAQ, age, education, race and ethnicity, and recruitment method were found to be significantly associated with the hazard for non-renewal over time. We estimated that a 5-point increase in baseline RAQ was associated with a 9% decrease in the hazard for non-renewal over time (HR: 0.91; 95% CI: (0.88, 0.95), p < 0.001). Notably, the estimated hazard for non-renewal in the second year was 42% lower (HR: 0.58; 95% CI: (0.53, 0.64), p < 0.001), and the estimated hazard in the third year was 59% lower (HR: 0.41; 95% CI: (0.36, 0.46), p < 0.001). The effects of age, education, race and ethnicity, and recruitment method remained generally similar to the primary model considering non-renewal at only the first year.

Table 3.

Discrete proportional hazards model of time to non-renewal with baseline RAQ.

		Unadjusted			Adjusted
Covariate	n	Est. HR	95% CI	p	Est. HR	95% CI	p
Year
1	2154	Referent	-	-	Referent	-	-
2	743	0.55	(0.51, 0.60)	<0.001	0.58	(0.53, 0.64)	<0.001
3	432	0.38	(0.33, 0.42)	<0.001	0.41	(0.36, 0.46)	<0.001
4	637	0.09	(0.06, 0.11)	<0.001	0.09	(0.07, 0.12)	<0.001
5	266	0.13	(0.08, 0.19)	<0.001	0.14	(0.09, 0.22)	<0.001
Baseline RAQ (5 pt)	4232	0.88	(0.84, 0.91)	<0.001	0.91	(0.88, 0.95)	<0.001
Age (5 y)	4232	0.91	(0.90, 0.92)	<0.001	0.94	(0.93, 0.95)	<0.001
Sex
Male	1509	Referent	-	-	Referent	-	-
Female	2723	1.07	(1.00, 1.15)	0.064	0.99	(0.92, 1.07)	0.859
Education	4232	0.95	(0.94, 0.96)	<0.001	0.97	(0.95, 0.98)	<0.001
Race and Ethnicity							0.001
NH White	3276	Referent	-	-	Referent	-	-
Hispanic/Latino	420	1.63	(1.45, 1.82)	<0.001	1.20	(1.06, 1.35)	0.003
NH Asian	318	1.53	(1.34, 1.74)	<0.001	1.23	(1.08, 1.41)	0.002
NH Black	59	1.56	(1.17, 2.09)	0.003	1.28	(0.96, 1.72)	0.096
Other	159	1.64	(1.37, 1.95)	<0.001	1.21	(1.01, 1.45)	0.044
Recruitment Method							<0.001
Email	2153	Referent	-	-	Referent	-	-
Doctor	171	1.56	(1.31, 1.85)	<0.001	1.43	(1.20, 1.71)	<0.001
Community Talk	405	0.75	(0.66, 0.85)	<0.001	0.97	(0.85, 1.11)	0.698
Media	580	1.18	(1.06, 1.31)	0.002	1.22	(1.09, 1.35)	<0.001
Friend	296	1.25	(1.09, 1.44)	0.001	1.27	(1.10, 1.46)	0.001
Postcard	269	0.84	(0.73, 0.98)	0.024	1.04	(0.89, 1.21)	0.647
Other	358	1.11	(0.98, 1.26)	0.111	1.22	(1.07, 1.39)	0.003
Comorbidities							0.141
None	2482	Referent	-	-	Referent	-	-
1	1009	0.81	(0.74, 0.88)	<0.001	0.97	(0.89, 1.06)	0.508
2	516	0.88	(0.79, 0.99)	0.026	1.07	(0.95, 1.20)	0.280
≥3	225	1.03	(0.88, 1.20)	0.712	1.16	(0.99, 1.37)	0.069

Table 3 displays results of the discrete proportional hazards model for the time to non-renewal associated with baseline RAQ and adjustment covariates.

Finally, when considering RAQ as a time-varying covariate, we estimated that the hazard for non-renewal was 10% lower when comparing groups of participants with time-varying RAQ 5 points higher to those 5 points lower (HR = 0.90; 95% CI: (0.87, 0.93); p < 0.001). In addition, the estimated relative hazard for non-renewal for participants with a 1-point increase in their RAQ between adjacent visits was 1.00 (95% CI: (0.99, 1.01); p = 0.792). Other results for the above two models are displayed in Tables 4 and 5.

Table 4.

Discrete proportional hazards model of time to non-renewal with time-varying RAQ.

		Unadjusted			Adjusted
Covariate	n	Est. HR	95% CI	p	Est. HR	95% CI	p
Year
1	2154	Referent	-	-	Referent	-	-
2	743	0.55	(0.51, 0.60)	<0.001	0.59	(0.54, 0.64)	<0.001
3	432	0.38	(0.33, 0.42)	<0.001	0.41	(0.37, 0.47)	<0.001
4	637	0.09	(0.06, 0.11)	<0.001	0.09	(0.07, 0.12)	<0.001
5	266	0.13	(0.08, 0.19)	<0.001	0.14	(0.10, 0.22)	<0.001
Time-varying RAQ (5 pt)	4232	0.86	(0.83, 0.89)	<0.001	0.90	(0.87, 0.93)	<0.001
Age (5 y)	4232	0.91	(0.90, 0.92)	<0.001	0.94	(0.93, 0.95)	<0.001
Sex
Male	1509	Referent	-	-	Referent	-	-
Female	2723	1.07	(1.00, 1.15)	0.064	0.99	(0.92, 1.07)	0.816
Education	4232	0.95	(0.94, 0.96)	<0.001	0.97	(0.95, 0.98)	<0.001
Race and Ethnicity							<0.001
NH White	3276	Referent	-	-	Referent	-	-
Hispanic/Latino	420	1.63	(1.45, 1.82)	<0.001	1.20	(1.07, 1.36)	0.003
NH Asian	318	1.53	(1.34, 1.74)	<0.001	1.24	(1.08, 1.41)	0.002
NH Black	59	1.56	(1.17, 2.09)	0.003	1.28	(0.96, 1.72)	0.098
Other	159	1.64	(1.37, 1.95)	<0.001	1.20	(1.00, 1.45)	0.048
Recruitment Method							<0.001
Email	2153	Referent	-	-	Referent	-	-
Doctor	171	1.56	(1.31, 1.85)	<0.001	1.44	(1.20, 1.72)	<0.001
Community Talk	405	0.75	(0.66, 0.85)	<0.001	0.97	(0.85, 1.11)	0.678
Media	580	1.18	(1.06, 1.31)	0.002	1.21	(1.09, 1.35)	<0.001
Friend	296	1.25	(1.09, 1.44)	0.001	1.26	(1.10, 1.45)	0.001
Postcard	269	0.84	(0.73, 0.98)	0.024	1.03	(0.88, 1.20)	0.720
Other	358	1.11	(0.98, 1.26)	0.111	1.22	(1.07, 1.39)	0.003
Comorbidities							0.121
None	2482	Referent	-	-	Referent	-	-
1	1009	0.81	(0.74, 0.88)	<0.001	0.97	(0.89, 1.06)	0.528
2	516	0.88	(0.79, 0.99)	0.026	1.07	(0.95, 1.20)	0.252
≥ 3	225	1.03	(0.88, 1.20)	0.712	1.17	(0.99, 1.38)	0.058

Table 4 displays results of the discrete proportional hazards model for the time to non-renewal associated with time-varying RAQ and adjustment covariates.

Table 5.

Discrete proportional hazards model of time to non-renewal with change in RAQ.

		Unadjusted			Adjusted
Covariate	n	Est. HR	95% CI	p	Est. HR	95% CI	p
Year
2	743	Referent	-	-	Referent	-	-
3	432	0.68	(0.60, 0.78)	<0.001	0.70	(0.61, 0.80)	<0.001
4	637	0.15	(0.12, 0.20)	<0.001	0.16	(0.12, 0.21)	<0.001
5	266	0.23	(0.15, 0.35)	<0.001	0.24	(0.16, 0.36)	<0.001
Change in RAQ (1 pt)	2078	1.00	(0.99, 1.02)	0.769	1.00	(0.99, 1.01)	0.792
Age (5 yr)	2078	0.94	(0.92, 0.96)	<0.001	0.95	(0.93, 0.98)	<0.001
Sex
Male	759	Referent	-	-	Referent	-	-
Female	1319	1.07	(0.94, 1.21)	0.289	1.03	(0.90, 1.17)	0.662
Education	2078	0.97	(0.95, 1.00)	0.018	0.98	(0.96, 1.00)	0.119
Race and Ethnicity							0.267
NH White	1717	Referent	-	-	Referent	-	-
Hispanic/Latino	153	1.38	(1.11, 1.71)	0.004	1.15	(0.92, 1.44)	0.232
NH Asian	122	1.31	(1.02, 1.67)	0.032	1.16	(0.90, 1.49)	0.248
NH Black	22	1.21	(0.67, 2.21)	0.523	1.03	(0.56, 1.88)	0.924
Other	64	1.69	(1.23, 2.33)	0.001	1.38	(1.00, 1.92)	0.053
Recruitment Method							0.046
Email	1083	Referent	-	-	Referent	-	-
Doctor	65	1.49	(1.07, 2.08)	0.018	1.45	(1.04, 2.02)	0.029
Community Talk	237	0.81	(0.66, 0.99)	0.044	0.95	(0.77, 1.18)	0.633
Media	263	1.20	(1.00, 1.43)	0.049	1.26	(1.05, 1.52)	0.012
Friend	127	1.18	(0.92, 1.51)	0.192	1.23	(0.96, 1.58)	0.109
Postcard	143	0.83	(0.64, 1.06)	0.135	0.98	(0.76, 1.26)	0.877
Other	160	0.97	(0.77, 1.23)	0.830	1.08	(0.85, 1.37)	0.529
Comorbidities							0.162
None	1165	Referent	-	-	Referent	-	-
1	536	0.84	(0.72, 0.97)	0.016	0.94	(0.81, 1.10)	0.443
2	268	0.97	(0.81, 1.16)	0.736	1.10	(0.91, 1.33)	0.347
≥3	109	1.15	(0.89, 1.50)	0.284	1.27	(0.97, 1.67)	0.079

Table 5 displays results of the discrete proportional hazards model for the time to non-renewal associated with change in RAQ from the previous year and adjustment covariates.

Active withdrawal

Figure 3 displays a Kaplan-Meier plot for the time to active withdrawal. The survival function of participants with baseline RAQ in the first quartile was generally lowest, while the survival function of those in the fourth quartile was generally highest. Table 6 displays results of the Cox proportional hazards model for the time to active withdrawal. We estimated that the hazard ratio for active withdrawal for participants with a 5-point increase in their baseline RAQ was 0.94 (95% CI: (0.83, 1.06); p = 0.328). Statistically significant covariates included age, years of education, and recruitment source. We estimated that a 5-year increase in age was associated with a 6% higher hazard for active withdrawal (HR: 1.06; 95% CI: (1.01, 1.10), p = 0.012), and a 1-year increase in education was associated with an estimated 4% lower hazard (HR: 0.96; 95% CI: (0.92, 1.00), p = 0.037). Compared to registrants recruited through email, we estimated that the hazard for active withdrawal among participants recruited via community talks was 49% lower (HR: 0.51; 95% CI: (0.33, 0.78), p = 0.002), and for those recruited through media, the hazard for active withdrawal was 31% lower (HR: 0.69; 95% CI: (0.49, 0.97), p = 0.035).

Figure 3.

Displays a Kaplan-Meier plot of the survival functions for participants in each quartile of baseline RAQ for the event of active withdrawal. The number of participants in the risk set (and the number of cumulative events for times) are reported.

Table 6.

Cox proportional hazards model of active withdrawal censoring at 15 months.

			Unadjusted			Adjusted
Covariate	n	Nw	Est. HR	95% CI	p	Est. HR	95% CI	p
Baseline RAQ (5 pt)	4663	330	0.92	(0.82, 1.04)	0.206	0.94	(0.83, 1.06)	0.328
Age (5 y)	4663	330	1.04	(1.01, 1.08)	0.026	1.06	(1.01, 1.10)	0.012
Sex
Male	1649	118	Referent	-	-	Referent	-	-
Female	3014	212	1.00	(0.80, 1.25)	0.985	1.04	(0.83, 1.32)	0.708
Education	4663	330	0.95	(0.92, 0.99)	0.022	0.96	(0.92, 1.00)	0.037
Race and Ethnicity								0.939
NH White	3568	271	Referent	-	-	Referent	-	-
Hispanic/Latino	463	24	0.94	(0.62, 1.42)	0.761	1.04	(0.67, 1.61)	0.862
NH Asian	377	22	1.03	(0.67, 1.60)	0.886	1.25	(0.80, 1.95)	0.331
NH Black	71	5	1.40	(0.58, 3.39)	0.457	1.48	(0.61, 3.60)	0.390
Other	184	8	0.80	(0.39, 1.61)	0.530	0.89	(0.44, 1.81)	0.742
Recruitment Method								1.000
Email	2237	164	Referent	-	-	Referent	-	-
Doctor	185	10	0.95	(0.50, 1.81)	0.882	0.88	(0.46, 1.67)	0.699
Community Talk	452	26	0.56	(0.37, 0.86)	0.008	0.51	(0.33, 0.78)	0.002
Media	726	46	0.74	(0.52, 1.04)	0.080	0.69	(0.49, 0.97)	0.035
Friend	341	30	1.12	(0.75, 1.66)	0.572	1.15	(0.78, 1.71)	0.481
Postcard	271	26	0.95	(0.63, 1.45)	0.823	0.86	(0.56, 1.31)	0.485
Other	451	28	0.86	(0.57, 1.29)	0.467	0.85	(0.56, 1.27)	0.426
Comorbidities								0.986
None	2762	173	Referent	-	-	Referent	-	-
1	1106	77	0.99	(0.76, 1.30)	0.970	0.92	(0.70, 1.22)	0.569
2	552	58	1.44	(1.07, 1.94)	0.016	1.32	(0.97, 1.81)	0.077
≥ 3	243	22	1.53	(0.98, 2.39)	0.059	1.32	(0.83, 2.08)	0.238

Table 6 displays results of the Cox proportional hazards model of the relative hazard for active withdrawal, censoring participants at 15 months since enrollment or last renewal. Nw is the number of active withdrawals in the category.

Discussion

Our analyses revealed that attitudes towards biomedical research were associated with retention behaviors in this recruitment registry. More specifically, higher scores on the RAQ assessed at baseline were associated with lower odds of non-renewal after one year and lower risk of non-renewal during subsequent years in our registry. Additionally, higher longitudinal RAQ scores were associated with lower risk of non-renewal after successful renewal. These results suggest a relationship between participants’ attitudes toward medical research and their decisions to stay enrolled in recruitment registries.

These results are consistent with previous studies demonstrating significant relationships between research attitudes and participatory behaviors in medical research. Numerous studies have shown that more positive attitudes toward research are associated with greater willingness to consider research participation.^{8, 9, 13–17} More directly relevant to the current results, previous studies have observed associations between higher RAQ scores and compliance with taking study medication and study completion in an AD clinical trial.⁷

Retaining participants in longitudinal research requires great effort and carries a cost. The C2C Registry for example, uses email reminders, annual newsletters and messages encouraging continued participation. An examination of retention in the NIH-funded Alzheimer's Disease Research Centers (ADRCs) revealed that centers employed an average of 42 retention tactics per site and centers that utilized more retention tactics had higher rates of retention.¹⁸ Designing more targeted approaches, however, such as implementing the most effective interventions specifically with participants at highest risk for dropout, may have the potential to lower cost without sacrificing study integrity, though further research will be needed to instruct such efforts. Our results suggest that the RAQ is a low burden tool that could be used to target interventions to those who may be at risk for loss-to-follow up in longitudinal research.

Our analyses identified additional factors associated with renewal behaviors. Older age and higher education were both associated with lower odds of non-renewal. Although the data linking retention behaviors and age are mixed in longitudinal research,^18–21 advancing age may have been a predictor of renewal in this study for a number of reasons, including a higher likelihood of reaching retirement age and having more time to dedicate to research.²² In support of this conjecture, insufficient time was cited previously as a barrier to research participation for younger registry participants.¹⁴ Higher educational attainment was associated with greater “registry engagement” in one study and greater odds of being retained in the ADRCs.^{18, 23, 24} We found that Asian participants and those who identify as Hispanic /Latino were at higher odds of non-renewal relative to NH White participants. It is not clear from our analysis why Hispanic/Latino and NH Asian participants were at greater odds of non-renewal. Demographic variables have been shown previously to associate with retention in registries and in longitudinal cohort studies.^{2, 23, 25} People of Asian race or Hispanic ethnicity may have more barriers to overcome to enroll and stay enrolled in research than do NH White participants ²⁶ and this may contribute to differential retention rates.²⁷ A review of the perceived barriers and facilitators to research participation among underrepresented groups identified convenience and sustained engagement as important factors when considering retention of minoritized participants.²⁸ Further research, including qualitative assessments in those declining to renew, will be necessary to understand the underlying mechanisms driving these observations more fully.

We observed that recruitment method was related to renewal behaviors in this analysis. Relative to email invitations, people who were recruited to the C2C Registry by referral through their doctor, through a media source, a friend or through other non-categorized methods were at greater odds of non-renewal after one year, even after successful renewal. This was not the case for registrants who received a postcard or heard about the C2C Registry through community talks. While one other group found that recruitment source was not related to rates of attrition in a randomized controlled trial of cognitive training,²⁹ a secondary analysis of REVEAL study of AD genetic risk disclosure showed that people actively recruited, for example being approached by their doctor or by study personnel, were more likely to withdraw than participants who self-identified through a website search or community presentation.³⁰ Although not entirely in line with those results, we did find that some of the more active forms of recruitment to the C2C Registry conferred with higher odds of non-renewal. We also observed that baseline RAQ varied by recruitment source such that the groups referred from some sources (e.g., doctor, email) tended to have lower RAQ scores while the groups recruited from other sources (e.g., community talk media source) tended to score higher. These data may highlight the difference between self-motivated participants and those who may be convinced to enroll. Though requiring further study, this could create an interesting and important conflict, whereby particular recruitment approaches may be most beneficial to increase representation of particular subgroups but also be associated with lower retention rates.

Active withdrawal, in contrast to non-renewal, was significantly associated with age and recruitment source. Active withdrawal from the C2C Registry requires that a person contact the research team to be removed while non-renewal is defined in this study by inaction or failure to renew within a timeframe; two distinct behaviors. Active withdrawal may result from a dissatisfactory experience, while non-renewal may result from being too busy or forgetting to respond.³¹ In this study, participants who were younger or heard about the study from a community talk or media source were at significantly lower odds of withdrawing from the registry, factors distinctly different than those observed in the models of non-renewal. Other groups have similarly reported that predictors of withdrawal differ from those of loss-to-follow-up or longitudinal task completion.^{22, 23} One study found that while younger participants were more likely to fail to complete longitudinal surveys in an online study, older adults were more likely to actively withdraw from the study.³² Older age is associated with added health problems, decreased mobility, degradation of technical competency, and changes in family dynamics that may impact one's ability to participate in studies. Older age, particularly in the context of deteriorating health, has been linked with higher likelihood of dropout in longitudinal cohort studies.^33–35 Though our data do not explain the seeming incongruence between our non-renewal and active dropout models, older participants may also feel more compelled to inform registry operators of their desire to no longer participate, compared to the passive option of simply not renewing when no longer interested.

Overall, these results may suggest that developing methods to either improve research attitudes or targeting those with lower RAQ scores for enhanced engagement could lead to better retention outcomes. For example, interventions to make research participants feel valued could improve participants’ attitudes towards research and make them more likely to want to remain actively engaged in a recruitment registry. Future research should test interventions to improve RAQ scores and whether this affects renewal rates

We acknowledge that this study has multiple limitations. First, the UCI C2C Registry has a unique process of enrollment and renewal. The time it takes to complete enrollment and update information differs from one registry to the next, as may the retention strategies employed. Hence the results here may not directly generalize to other registries. In addition, participants in the C2C Registry can choose to skip questions at enrollment and at any renewal instance. Because this is an online self-administered survey, missing data occurs frequently and limited our sample size for some analyses and may have led to bias in our results. Furthermore, we do not have data on why participants chose to renew or not to renew their enrollment. These data would have provided context to help better understand the relationship between research attitudes and the decision whether to renew. We observed some differences in our outcomes based on race and ethnicity. These are largely social constructs, however, and this registry includes limited assessments of social determinants of health that might be more important than these constructs. It is important to note that the RAQ is self-reported and may be subject to response bias. Though this study adds to the growing body of evidence that RAQ is a potentially important tool, prospective studies of its value remain few.

Registries may be most impactful in recruiting to AD studies enrolling cognitively unimpaired participants such as preclinical Alzheimer's disease clinical trials. These trials generally screen large populations of individuals in order to identify individuals meeting AD biomarker enrollment criteria³⁶ and recruitment registries can supply robust participant pools for screening. Such efforts, however, will be most effective if registries successfully retain large populations of participants to the point of active recruitment. In this study, favorable research attitudes predicted renewal behaviors in our local recruitment registry. Other important predictors of retention in this study included demographic variables and referral source. These data may be useful to develop strategies to improve retention in this and other recruitment registries. Future research should seek to replicate these findings in other registries, including other local as well as larger national registries. It will also be important to better understand the reasons why people withdraw from registries.

Supplemental Material

sj-docx-1-alz-10.1177_13872877241302422 - Supplemental material for Research Attitudes Questionnaire scores and retention in a recruitment registry

Supplemental material, sj-docx-1-alz-10.1177_13872877241302422 for Research Attitudes Questionnaire scores and retention in a recruitment registry by Megan Witbracht, Yiren Xu, Olivia B Morgan, Christian R Salazar, Dan Hoang, Amy Kind, Daniel L Gillen and Joshua D Grill in Journal of Alzheimer's Disease

Footnotes

Acknowledgments

We would like to thank the participants of the UC Irvine Consent-to-Contact Registry.

ORCID iD

Megan Witbracht

Author contributions

Megan Witbracht (Conceptualization; Writing – original draft; Writing – review & editing); Yiren Xu (Formal analysis; Methodology; Writing – original draft; Writing – review & editing); Olivia B Morgan (Methodology; Writing – review & editing); Christian R Salazar (Writing – review & editing); Dan Hoang (Data curation; Methodology; Project administration); Amy Kind (Conceptualization; Methodology; Writing – review & editing); Daniel L Gillen (Conceptualization; Formal analysis; Methodology; Supervision; Writing – original draft); Joshua D Grill (Conceptualization; Methodology; Supervision; Writing – original draft).

Funding

Creation of the registry was made possible by a donation from HCP, Inc. and is supported by NIA P30 AG016573, R01 AG077628, and NCATSUL1 TR001414. Christian Salazar is supported by NIA K01AG076811. Yiren Xu is supported by the UCI Steckler Center for Responsible, Ethical, and Accessible Technology (CREATE) Fellowship. Olivia Morgan was supported by the National Science Foundation Graduate Research Fellowship under grant number DGE-1839285.

Declaration of conflicting interests

The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Joshua D Grill received funding from Eli Lilly, Genentech, Biogen and Eisai and receives consulting fees from SiteRx (2021–2024). Daniel L Gillen serves on the Data Safety Monitoring Board or Advisory Board of the following companies, Novo Nordisk, Astrazeneca, Novartis, Genentech, Amgen, Editas, Seattle Genetics, Intellia, Generate, CRISPR, Biomarin, Bristol Meyers Squib, Eli Lilly, and Celgene (2021–2024). The remaining authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Data availability

The data supporting the findings of this study are available to requesting investigators. Please contact the corresponding author.

Supplemental material

Supplemental material for this article is available online.

References

Weiner

Nosheny

Camacho

, et al. The Brain Health Registry: an internet-based platform for recruitment, assessment, and longitudinal monitoring of participants for neuroscience studies. Alzheimers Dement 2018; 14: 1063–1076.

Langbaum

High

Nichols

, et al. The Alzheimer's Prevention Registry: a large internet-based participant recruitment registry to accelerate referrals to Alzheimer's-focused studies. J Prev Alzheimers Dis 2020; 7: 242–250.

Grill

. Recruiting to preclinical Alzheimer's disease clinical trials through registries. Alzheimers Dement (N Y) 2017; 3: 205–212.

Walter

Langford

Clanton

, et al. The Trial-Ready Cohort for Preclinical and Prodromal Alzheimer's Disease (TRC-PAD): experience from the first 3 years. J Prev Alzheimers Dis 2020; 7: 234–241.

Rubright

Cary

Karlawish

, et al. Measuring how people view biomedical research: reliability and validity analysis of the Research Attitudes Questionnaire. J Empir Res Hum Res Ethics 2011; 6: 63–68.

Sewell

Neugroschl

Umpierre

, et al. Research attitudes and interest among elderly Latinxs: the impact of a collaborative video and community peers. J Alzheimers Dis 2021; 82: 771–779.

Stites

Turner

Gill

, et al. Research Attitudes Questionnaire scores predict Alzheimer's disease clinical trial dropout. Clin Trials 2021; 18: 237–244.

Grill

Zhou

Elashoff

, et al. Disclosure of amyloid status is not a barrier to recruitment in preclinical Alzheimer's disease clinical trials. Neurobiol Aging 2016; 39: 147–153.

Salazar

Hoang

Gillen

, et al. Racial and ethnic differences in older adults’ willingness to be contacted about Alzheimer's disease research participation. Alzheimers Dement (N Y) 2020; 6: e12023.

10.

Grill

Hoang

Gillen

, et al. Constructing a local potential participant registry to improve Alzheimer's disease clinical research recruitment. J Alzheimers Dis 2018; 63: 1055–1063.

11.

Harris

Taylor

Thielke

, et al. Research electronic data capture (REDCap)–a metadata-driven methodology and workflow process for providing translational research informatics support. J Biomed Inform 2009; 42: 377–381.

12.

Kalbfleisch

Prentice

. The statistical analysis of failure time data. 2nd ed. Hoboken, NJ: John Wiley, 2002, p.xiii, 439 p.

13.

Gucciardi

Di Liao

Cameron

. Determining the factors associated with health research participation. J Ambul Care Manage 2010; 33: 272–284.

14.

Jefferson

Lambe

Chaisson

, et al. Clinical research participation among aging adults enrolled in an Alzheimer's Disease Center research registry. J Alzheimers Dis 2011; 23: 443–452.

15.

Karlawish

Rubright

Casarett

, et al. Older adults’ attitudes toward enrollment of non-competent subjects participating in Alzheimer's research. Am J Psychiatry 2009; 166: 182–188.

16.

Lott

Kirby

Doran

, et al. Research attitudes in families of individuals with Down syndrome: importance for clinical trials. Alzheimers Res Ther 2022; 14: 176.

17.

Cary

Rubright

Grill

, et al.

Why are spousal caregivers more prevalent than nonspousal caregivers as study partners in AD dementia clinical trials?

Alzheimer Dis Assoc Disord 2015; 29: 70–74.

18.

Grill

Kwon

Teylan

, et al. Retention of Alzheimer disease research participants. Alzheimer Dis Assoc Disord 2019; 33: 299–306.

19.

Okpara

Adachi

Papaioannou

, et al. Exploring participant attrition in a longitudinal follow-up of older adults: the Global Longitudinal Study of Osteoporosis in Women (GLOW) Hamilton cohort. BMJ Open 2023; 13: e066594.

20.

Bernstein

Grill

Gillen

. Recruitment and retention of participant and study partner dyads in two multinational Alzheimer's disease registration trials. Alzheimers Res Ther 2021; 13: 16.

21.

Burke

Naseh

, et al. Factors influencing attrition in 35 Alzheimer's Disease Centers across the USA: a longitudinal examination of the National Alzheimer's Coordinating Center's Uniform Data Set. Aging Clin Exp Res 2019; 31: 1283–1297.

22.

Mein

Johal

Grant

, et al. Predictors of two forms of attrition in a longitudinal health study involving ageing participants: an analysis based on the Whitehall II study. BMC Med Res Methodol 2012; 12: 164.

23.

Ashford

Eichenbaum

Williams

, et al. Effects of sex, race, ethnicity, and education on online aging research participation. Alzheimers Dement (N Y) 2020; 6: e12028.

24.

Beavers

Pettinger

Espeland

, et al. The evolution of the WHI 80+cohort. J Gerontol A Biol Sci Med Sci 2016; 71: S13–S22.

25.

Hall

Ficker

Chadiha

, et al. Promoting retention: African American older adults in a research volunteer registry. Gerontol Geriatr Med 2016; 2: 2333721416677469.

26.

Milani

Swain

Otufowora

, et al.

Willingness to participate in health research among community-dwelling middle-aged and older adults: does race/ethnicity matter?

J Racial Ethn Health Disparities 2021; 8: 773–782.

27.

Salazar

Ritchie

Gillen

, et al. Strategies associated with retaining participants in the Longitudinal National Alzheimer's Coordinating Center Uniform Data Set Study. J Alzheimers Dis 2022; 87: 1557–1566.

28.

Wong

Amano

Lin

, et al. Strategies for the recruitment and retention of racial/ethnic minorities in Alzheimer disease and dementia clinical research. Curr Alzheimer Res 2019; 16: 458–471.

29.

Julbe-Delgado

O'Brien

Abdulkarim

, et al. Quantifying recruitment source and participant communication preferences for Alzheimer's disease prevention research. J Prev Alzheimers Dis 2021; 8: 299–305.

30.

Christensen

Roberts

Zikmund-Fisher

, et al. Associations between self-referral and health behavior responses to genetic risk information. Genome Med 2015; 7: 10.

31.

Katsuno

Bourbeau

, et al. Factors associated with attrition in a longitudinal cohort of older adults in the community. Chronic Obstr Pulm Dis 2023; 10: 178–189.

32.

Rubsamen

Akmatov

Castell

, et al. Factors associated with attrition in a longitudinal online study: results from the HaBIDS panel. BMC Med Res Methodol 2017; 17: 132.

33.

Jacobsen

Ran

Liu

, et al. Predictors of attrition in a longitudinal population-based study of aging. Int Psychogeriatr 2021; 33: 767–778.

34.

Sharma

Tobin

Brant

. Factors affecting attrition in the Baltimore Longitudinal Study of Aging. Exp Gerontol 1986; 21: 329–340.

35.

Chatfield

Brayne

Matthews

. A systematic literature review of attrition between waves in longitudinal studies in the elderly shows a consistent pattern of dropout between differing studies. J Clin Epidemiol 2005; 58: 13–19.

36.

Raman

Quiroz

Langford

, et al. Disparities by race and ethnicity among adults recruited for a preclinical Alzheimer disease trial. JAMA Netw Open 2021; 4: e2114364.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB