Abstract
Objectives:
Participant retention is essential in prospective studies to minimize bias and maintain validity. This study aimed to estimate retention rates and identify factors influencing attrition in a multicenter cohort (PROCEED) of U.S. adults with acute/recurrent acute pancreatitis (AP/RAP) and chronic pancreatitis (CP).
Methods:
Participants from the PROCEED study with AP/RAP or CP eligible for follow-up within the designated visit window were included. Retention was defined as completing at least one follow-up visit during the observation period. Retention was analyzed by follow-up mode (in-person and/or medical record review) and disease type (AP/RAP and CP). Retention factors were assessed using univariate analyses (Wilcoxon rank-sum for continuous variables, Chi-square/Fisher's exact for categorical) and multivariable logistic regression. Missing data were adjusted using multiple imputations in statistical modeling.
Results:
Among 1,279 participants (AP/RAP: n=632, median age 47, 48.9% female; CP: n=647, median age 54, 46.5% female), we observed high cumulative and annualized retention rates, with improved retention when incorporating medical record reviews. In multivariable regression analyses, older age was statistically significantly associated with in-person retention in the AP/RAP group (p=0.0001), while a higher self-reported physical health (assessed with PROMIS Global Health Physical Health T-score) was a key predictor of retention in CP (p=0.034).
Conclusions:
Integrating medical records review substantially enhanced retention rates, enabling robust analyses of pancreatitis outcomes and disease progression. Older age and better self-reported physical health were key predictors of retention, highlighting the need for targeted strategies to enhance engagement among younger participants and those with poorer health in long-term studies.
Keywords: Participant retention, Pancreatitis, PROCEED, Longitudinal cohort studies
Introduction
Longitudinal cohort studies are critical in clinical research as they allow investigation of the natural history of diseases, enhance understanding of factors associated with disease progression over time, and improve knowledge of time-varying factors.1, 2 Participant retention in cohort studies is particularly important because selective attrition can compromise study validity and introduce bias.1, 3, 4 Moreover, participant attrition may lead to a prolonged study duration, requiring an additional investment of resources (financial and personnel).5
While some loss to follow-up is often unavoidable in cohort studies, no universal standard exists for acceptable rates of follow-up and attrition due to variations in study design, study populations, and outcome measures across studies. Some authors suggest acceptable follow-up rates for epidemiologic cohorts range between 50 and 80%, with an acceptable threshold for attrition around 20%.6, 7 However, these benchmarks have not been extensively researched or validated.
Research on factors associated with participant retention remains limited. Attrition can result from various factors, including patient-level characteristics such as sociodemographic factors and disease-specific characteristics, and high disease burden.5 In cohorts of older individuals, morbidity and mortality are often the primary reasons for loss to follow-up.4 In cohorts of adult participants, factors associated with attrition may be influenced by the nature and design of the study, which may or may not be modifiable by researchers. Regardless, recognizing and accounting for potential sociodemographic and clinical determinants of attrition is essential in data analyses.
Retention in studies of conditions such as acute or recurrent acute (AP/RAP) and chronic pancreatitis (CP) may present a unique set of challenges. Patients often experience multiple AP attacks, persistent symptoms, such as severe abdominal pain, debilitating complications, and prolonged hospital stays, which may temper enthusiasm for ongoing participation and hinder attendance at follow-up visits. These challenges can often disrupt efforts by research team members to retain and follow patients in longitudinal research studies. While recurrent attacks may motivate some patients to attend follow-ups, frequent episodes could disrupt visit scheduling. Some patients develop long-term adverse outcomes and conditions like diabetes, exocrine pancreatic insufficiency, and osteopathy, which may necessitate more frequent clinical follow-ups but simultaneously make attendance more challenging. In addition, patients with CP often have multiple chronic comorbidities, such as coronary artery disease, chronic obstructive pulmonary disease (COPD), and chronic heart failure (CHF), requiring additional medical attention, which may impact retention.8
The Prospective Evaluation of Chronic Pancreatitis for Epidemiologic and Translational Studies (PROCEED) is an ongoing multicenter, observational cohort study (ClinicalTrials.gov NCT03099850) in the United States, conducted as part of the Consortium for the Study of Chronic Pancreatitis, Diabetes, and Pancreas Cancer (CPDPC) consortium.9 PROCEED aims to enroll and follow a well-phenotyped population across different stages of CP to define disease progression. PROCEED represents the first longitudinal research cohort to study CP as well as AP/RAP in the US. Participant retention is, therefore, of critical importance to the study investigators. Thus, the primary objective of our study was to estimate retention rates within the PROCEED cohort and identify factors associated with participant retention.
Materials and Methods
Participants and Procedures
This longitudinal analysis utilized data from the PROCEED study collected between June 2017 and January 2025. Adults (≥18 years) with AP/RAP were enrolled between June 2017 and July 2023, and those with CP were enrolled between July 2017 and December 2021. All participants provided informed consent, and institutional review board (IRB) approval was obtained at each participating site. Individuals included in this analysis were patients enrolled in the following PROCEED groups: (1) those with confirmed diagnoses of AP or RAP and (2) those with definite CP. Detailed eligibility criteria for enrollment for individual groups have been previously described.9 In brief, eligibility criteria for the AP group were (1) one documented attack of AP in the preceding 18 months based on Revised Atlanta Classification;10 (2) Cambridge grade < III on CT scan and MRI/MRCP; (4) Pancreatic necrosis, if present, is <50%; (5) No prior pancreatic surgery. For the RAP group: (1) Two or more documented attacks of AP; (2) Cambridge grade < III on CT scan and MRI/MRCP; (3) No prior pancreatic surgery. Eligibility criteria for enrollment in the CP group were (1) Presence of unequivocal CP, that is, Cambridge grade 3 or 4, and/or parenchymal and/or ductal calcifications by CT scan or MRI/MRCP, or (2) Pancreatic histology diagnostic of CP (including findings of fibrosis [Ammann's ≥ 6], chronic inflammation, and acinar loss).9
PROCEED data were collected using structured case report forms (CRFs) completed by participants, study coordinators, and enrolling physicians, as previously described.9 After the baseline assessment, all participants were asked to return for annual in-person follow-up visits within a ±6-month visit window. The preferred method of follow-up was an in-person visit, which included the completion of CRFs by participants, study coordinators, and physicians, as well as the collection of biological samples (e.g., blood and urine).
During the COVID-19 pandemic, an additional follow-up method was introduced, allowing study coordinators to conduct telephone interviews with participants to administer CRFs. This approach aimed to reduce participant burden and improve data collection. A completed telephone interview was considered sufficient to fulfill the in-person follow-up visit requirement. For the purposes of this study, we included telephone interviews within the in-person retention category. If neither in-person nor telephone follow-up was possible, available information was extracted from the participant's medical records.
Data collected using structured CRFs included sociodemographic information, disease-related variables (e.g., age at diagnosis, etiology, disease duration), clinical characteristics (e.g., exocrine pancreatic dysfunction (EPD), diabetes, AP attacks), imaging findings, treatment history, hospitalizations, and patient-reported outcomes, such as quality of life and pain.
Measures
For these analyses, retention was defined as completing at least one follow-up visit during the observation period. Participants who did not complete any follow-up visits were classified as not retained. Individuals who completed or passed at least one follow-up visit window were included in these analyses. Retention was evaluated based on two modes of follow-up: (1) follow-up visits completed in-person or by telephone (primary outcome), and (2) follow-up visits completed using all available data (in-person/telephone interview or medical records review) (secondary outcome). Retention rates were analyzed separately for the AP/RAP and CP groups.
Demographic variables included in the analyses were age, sex, education, income, employment, marital status, and the distance between the participant's residence and the clinical center. Age was analyzed both as a continuous variable and as a categorical variable (<35 years, 35–50, 50–65, and >65 years) measured at baseline. Distance to the clinical center was measured in miles and categorized into four groups (0–25, 25–50, 50–100, and ≥100 miles).
Baseline clinical characteristics potentially relevant to retention included disease-specific factors (e.g., pancreatitis duration, baseline exocrine pancreatic dysfunction, diabetes), lifestyle risk factors (e.g., tobacco and alcohol use), hospitalizations, emergency department visits, Charlson Comorbidity Index (CCI),11 pain, and quality of life. The CCI was categorized as follows: Category 1, score 0–1; Category 2, score 2–3; and Category 3, score >3, where a higher CCI score indicates a greater burden of comorbidities. Quality of life was evaluated using Patient-Reported Outcomes Measurement Information System (PROMIS) instruments, as previously reported.12, 13 PROMIS Global Physical and Mental Health T-scores, which are standardized to a mean of 50 (SD=10) in the U.S. general population, were used, with higher scores indicating better physical and mental health.
Statistical analysis
Summary statistics were computed using frequencies and proportions for categorical variables and means, standard deviations (SD), medians, and interquartile ranges (IQR) for continuous variables. Baseline characteristics were compared across retention categories using the Wilcoxon rank-sum test for continuous variables and either the chi-square test or Fisher's exact test for categorical variables. Missing data in baseline predictors were addressed through multiple imputations using the PROC MI procedure in SAS, generating five imputed datasets. This approach was limited to regression analysis and enhanced the effective sample size by retaining observations that would otherwise be excluded in a complete case analysis.
Multivariable logistic regression models were developed separately for each retention category. All analyses accounted for missing data through multiple imputation, using five independently imputed datasets per retention category. Predictors were selected using a stepwise regression method (α=0.05 for both entry and exit) applied independently to each imputed dataset. Variables demonstrating statistical significance (p<0.05) in any individual imputed dataset were retained for final model inclusion. Final models containing these variables were subsequently refitted to all five imputed datasets, and results were pooled using Rubin's rules14 through SAS PROC MIANALYZE, ensuring proper accounting for between-imputation variability.
Annualized and cumulative retention rates were computed for AP/RAP and CP groups by retention type (e.g., in-person retention and all available data, including review of medical records).
Data management and statistical analyses were performed using SAS version 9.4 (SAS Institute, Cary, NC). As each association was assessed independently, adjustments for multiple comparisons (e.g., familywise hypothesis testing) were not applied.
Results
Study Cohort
The analysis included 1,279 individuals who met the study inclusion criteria: 632 with AP/RAP and 647 with CP (Tables 1 and 2). The AP/RAP group had a median age of 47.0 years at enrollment, with nearly half (48.9%) of the participants being female. Approximately 44% had post-secondary education, 43.9% reported a combined family income exceeding $100,000, and 65% were employed. In comparison, the CP group had a median age of 54.0 years, with a higher proportion of males (53.5%). Only 30% of the participants in the CP group had post-secondary education, nearly half (47.4%) reported a combined family income below $50,000, and over one-third were unemployed. CP participants had high rates of current cigarette smoking (36.5%), with a mean of 23.8 pack-years, whereas the majority of AP/RAP participants (55.4%) had never smoked. Most CP cases were attributed to alcohol-related etiology (41.6%), while most AP/RAP cases were idiopathic (56.3%).
Table 1.
Descriptive characteristics of the sociodemographic and clinical characteristics among participants with AP/RAP by in-person retention status.
| In-person Retention | ||||
|---|---|---|---|---|
| No (N=159) |
Yes (N=473) |
Total (N=632) |
P-value | |
|
Age at enrollment, years Median (IQR) |
42.0 (32.0, 53.0) |
49.0 (39.0, 59.0) |
47.0 (37.0, 57.0) |
<.00011 |
| Age categories, years, n (%) | <.00012 | |||
| <35 | 54 (34.0%) | 76 (16.1%) | 130 (20.6%) | |
| 35–50 | 55 (34.6%) | 168 (35.5%) | 223 (35.3%) | |
| 50–65 | 41 (25.8%) | 179 (37.8%) | 220 (34.8%) | |
| >65 | 9 (5.7%) | 50 (10.6%) | 59 (9.3%) | |
| Sex, n (%) | 0.61552 | |||
| Female | 75 (47.2%) | 234 (49.5%) | 309 (48.9%) | |
| Education, n (%) | 0.01682 | |||
| HS/GED or less | 39 (28.1%) | 94 (20.6%) | 133 (22.3%) | |
| Some training past HS | 53 (38.1%) | 147 (32.2%) | 200 (33.6%) | |
| Bachelor's or graduate school | 47 (33.8%) | 216 (47.3%) | 263 (44.1%) | |
| Missing | 20 | 16 | 36 | |
| Combined family income, $, n (%) | 0.00092 | |||
| >50k | 50 (40.7%) | 105 (25.4%) | 155 (28.9%) | |
| 50–100k | 35 (28.5%) | 111 (26.8%) | 146 (27.2%) | |
| ≥100k | 38 (30.9%) | 198 (47.8%) | 236 (43.9%) | |
| Missing | 36 | 59 | 95 | |
| Employment status, n (%) | 0.02782 | |||
| Unemployed | 41 (28.5%) | 84 (18.4%) | 125 (20.8%) | |
| Retired | 16 (11.1%) | 69 (15.1%) | 85 (14.2%) | |
| Employed | 87 (60.4%) | 303 (66.4%) | 390 (65.0%) | |
| Missing | 15 | 17 | 32 | |
| Marital status, n (%) | 0.00082 | |||
| Married | 77 (51.7%) | 310 (66.2%) | 387 (62.7%) | |
| Single | 54 (36.2%) | 99 (21.2%) | 153 (24.8%) | |
| Divorced/Separated/Widowed | 18 (12.1%) | 59 (12.6%) | 77 (12.5%) | |
| Missing | 10 | 5 | 15 | |
| Cigarette smoking status, n (%) | 0.08892 | |||
| Never | 80 (53.0%) | 259 (56.2%) | 339 (55.4%) | |
| Past | 42 (27.8%) | 146 (31.7%) | 188 (30.7%) | |
| Current | 29 (19.2%) | 56 (12.1%) | 85 (13.9%) | |
| Missing | 8 | 12 | 20 | |
| Pack-years among cigarette smokers | 0.04441 | |||
| Median (IQR) | 9.5 (4.0, 24.0) | 8.0 (2.5, 17.5) | 8.4 (3.0, 20.0) | |
| Alcohol use, n (%) | 0.78132 | |||
| Never | 23 (15.2%) | 61 (13.2%) | 84 (13.7%) | |
| Past | 90 (59.6%) | 275 (59.7%) | 365 (59.6%) | |
| Current | 38 (25.2%) | 125 (27.1%) | 163 (26.6%) | |
| Missing | 8 | 12 | 20 | |
| Pain pattern, n (%) | 0.13992 | |||
| No Pain | 7 (4.7%) | 19 (4.2%) | 26 (4.3%) | |
| Intermittent Mild | 24 (16.2%) | 116 (25.5%) | 140 (23.2%) | |
| Intermittent Severe | 58 (39.2%) | 148 (32.5%) | 206 (34.2%) | |
| Constant Mild | 12 (8.1%) | 25 (5.5%) | 37 (6.1%) | |
| Constant Severe | 47 (31.8%) | 147 (32.3%) | 194 (32.2%) | |
| Missing | 11 | 18 | 29 | |
| Duration of pancreatitis prior to enrollment | 0.41131 | |||
| Median (IQR) | 3.0 (1.0, 7.0) | 2.0 (1.0, 7.0) | 2.0 (1.0, 7.0) | |
| N (Missing) | 159 (0) | 473 (0) | 632 (0) | |
| PROMIS Global Health Physical score 3 | 0.00491 | |||
| Median (IQR) | 42.3 (37.4, 50.8) | 44.9 (39.8, 50.8) | 44.9 (39.8, 50.8) | |
| N (Missing) | 141 (18) | 454 (19) | 595 (37) | |
| PROMIS Global Health Mental score 4 | 0.10531 | |||
| Median (IQR) | 48.3 (41.1, 53.3) | 48.3 (41.1, 56.0) | 48.3 (41.1, 53.3) | |
| N (Missing) | 141 (18) | 454 (19) | 595 (37) | |
| Pancreatitis etiology, n (%) | 0.56582 | |||
| Idiopathic | 90 (56.6%) | 266 (56.2%) | 356 (56.3%) | |
| Alcohol | 31 (19.5%) | 78 (16.5%) | 109 (17.2%) | |
| All other(s) | 38 (23.9%) | 129 (27.3%) | 167 (26.4%) | |
| Exocrine pancreatic dysfunction present at baseline, n (%) | 0.58842 | |||
| No | 59 (83.1%) | 232 (80.3%) | 291 (80.8%) | |
| Yes | 12 (16.9%) | 57 (19.7%) | 69 (19.2%) | |
| Not tested5 | 88 | 184 | 272 | |
| Diabetes present at baseline, n (%) | 0.19022 | |||
| No | 105 (73.9%) | 350 (79.2%) | 455 (77.9%) | |
| Yes | 37 (26.1%) | 92 (20.8%) | 129 (22.1%) | |
| Not tested6 | 17 | 31 | 48 | |
| Charlson Comorbidity Index, n (%) | 0.44002 | |||
| 0–1 | 135 (84.9%) | 389 (82.2%) | 524 (82.9%) | |
| ≥2 | 24 (15.1%) | 84 (17.8%) | 108 (17.1%) | |
| Hospitalizations for pancreatitis in the past 1 year, n (%) | 0.79512 | |||
| No | 44 (28.6%) | 127 (27.5%) | 171 (27.8%) | |
| Yes | 110 (71.4%) | 335 (72.5%) | 445 (72.2%) | |
| Missing | 5 | 11 | 16 | |
| ED visits for pancreatitis in the past 1 year, n (%) | 0.57342 | |||
| No | 84 (54.5%) | 264 (57.1%) | 348 (56.5%) | |
| Yes | 70 (45.5%) | 198 (42.9%) | 268 (43.5%) | |
| Missing | 5 | 11 | 16 | |
| Distance between residence and clinical center, miles, n (%) | 0.04742 | |||
| 0–25 | 31 (21.5%) | 148 (32.0%) | 179 (29.5%) | |
| 25–50 | 28 (19.4%) | 94 (20.3%) | 122 (20.1%) | |
| 50–100 | 28 (19.4%) | 85 (18.4%) | 113 (18.6%) | |
| ≥100 | 57 (39.6%) | 135 (29.2%) | 192 (31.7%) | |
| Missing | 15 | 11 | 26 | |
Wilcoxon rank sum p-value;
Chi-Square p-value;
Physical Component Score from PROMIS Global Health;
Mental Component Score from PROMIS Global Health;
Patients with no known EPD who did not have clinical steatorrhea and did not undergo per-protocol fecal elastase testing;
Patients with no known diabetes did not undergo per-protocol diabetes testing.
Table 2.
Descriptive characteristics of the sociodemographic and clinical characteristics among participants with CP by in-person retention status.
| In-person Retention | ||||
|---|---|---|---|---|
| No (N=119) |
Yes (N=528) |
Total (N=647) |
P-value | |
|
Age at enrollment, years Median (IQR) |
52.0 (43.0, 61.0) |
55.0 (44.0, 64.0) |
54.0 (44.0, 63.0) |
0.05141 |
| Age categories, years, n (%) | 0.01842 | |||
| <35 | 15 (12.6%) | 51 (9.7%) | 66 (10.2%) | |
| 35–50 | 33 (27.7%) | 149 (28.2%) | 182 (28.1%) | |
| 50–65 | 62 (52.1%) | 228 (43.2%) | 290 (44.8%) | |
| >65 | 9 (7.6%) | 100 (18.9%) | 109 (16.8%) | |
| Sex, n (%) | 0.59142 | |||
| Female | 58 (48.7%) | 243 (46.0%) | 301 (46.5%) | |
| Education, n (%) | 0.01742 | |||
| HS/GED or less | 50 (45.5%) | 163 (31.5%) | 213 (34.0%) | |
| Some training past HS | 34 (30.9%) | 187 (36.2%) | 221 (35.2%) | |
| Bachelor's or graduate school | 26 (23.6%) | 167 (32.3%) | 193 (30.8%) | |
| Missing | 9 | 11 | 20 | |
| Combined family income, $, n (%) | 0.13262 | |||
| <50k | 56 (56.6%) | 213 (45.5%) | 269 (47.4%) | |
| 50–100k | 23 (23.2%) | 132 (28.2%) | 155 (27.3%) | |
| ≥100k | 20 (20.2%) | 123 (26.3%) | 143 (25.2%) | |
| Missing | 20 | 60 | 80 | |
| Employment status, n (%) | 0.00792 | |||
| Unemployed | 52 (46.4%) | 162 (31.4%) | 214 (34.1%) | |
| Retired | 19 (17.0%) | 129 (25.0%) | 148 (23.6%) | |
| Employed | 41 (36.6%) | 225 (43.6%) | 266 (42.4%) | |
| Missing | 7 | 12 | 19 | |
| Marital status, n (%) | 0.70062 | |||
| Married | 63 (53.8%) | 303 (57.9%) | 366 (57.2%) | |
| Single | 28 (23.9%) | 118 (22.6%) | 146 (22.8%) | |
| Divorced/Separated/Widowed | 26 (22.2%) | 102 (19.5%) | 128 (20.0%) | |
| Missing | 2 | 5 | 7 | |
| Cigarette smoking status, n (%) | 0.20982 | |||
| Never | 38 (33.0%) | 174 (33.2%) | 212 (33.2%) | |
| Past | 28 (24.3%) | 166 (31.7%) | 194 (30.4%) | |
| Current | 49 (42.6%) | 184 (35.1%) | 233 (36.5%) | |
| Missing | 4 | 4 | 8 | |
| Pack-years among cigarette smokers | 0.14081 | |||
| Median (IQR) | 24.0 (10.5, 36.0) | 19.0 (6.7, 34.0) | 19.5 (7.5, 34.0) | |
| Alcohol use, n (%) | 0.71042 | |||
| Never | 17 (14.9%) | 66 (12.6%) | 83 (13.0%) | |
| Past | 69 (60.5%) | 337 (64.3%) | 406 (63.6%) | |
| Current | 28 (24.6%) | 121 (23.1%) | 149 (23.4%) | |
| Missing | 5 | 4 | 9 | |
| Pain pattern, n (%) | 0.06152 | |||
| No Pain | 13 (11.3%) | 80 (15.4%) | 93 (14.7%) | |
| Intermittent Mild | 21 (18.3%) | 106 (20.5%) | 127 (20.1%) | |
| Intermittent Severe | 21 (18.3%) | 120 (23.2%) | 141 (22.3%) | |
| Constant Mild | 6 (5.2%) | 43 (8.3%) | 49 (7.7%) | |
| Constant Severe | 54 (47.0%) | 169 (32.6%) | 223 (35.2%) | |
| Missing | 4 | 10 | 14 | |
| Duration of pancreatitis prior to enrollment | 0.12221 | |||
| Median (IQR) | 6.0 (2.0, 14.0) | 5.0 (1.0, 10.0) | 5.0 (2.0, 11.0) | |
| N (Missing) | 118 (1) | 527 (1) | 645 (2) | |
| PROMIS Global Health Physical score 3 | 0.01631 | |||
| Median (IQR) | 42.3 (34.9, 47.7) | 42.3 (37.4, 50.8) | 42.3 (37.4, 50.8) | |
| N (Missing) | 117 (9) | 508 (13) | 625 (22) | |
| PROMIS Global Health Mental score 4 | 0.10221 | |||
| Median (IQR) | 45.8 (38.8, 50.8) | 48.3 (41.1, 53.3) | 48.3 (41.1, 53.3) | |
| N (Missing) | 111 (8) | 516 (12) | 627 (20) | |
| Pancreatitis etiology, n (%) | 0.31022 | |||
| Idiopathic | 45 (37.8%) | 199 (37.7%) | 244 (37.7%) | |
| Alcohol | 55 (46.2%) | 214 (40.5%) | 269 (41.6%) | |
| All other(s) | 19 (16.0%) | 115 (21.8%) | 134 (20.7%) | |
| Exocrine pancreatic dysfunction present at baseline, n (%) | 0.06002 | |||
| No | 15 (26.3%) | 138 (39.3%) | 153 (37.5%) | |
| Yes | 42 (73.7%) | 213 (60.7%) | 255 (62.5%) | |
| Not tested5 | 62 | 177 | 239 | |
| Diabetes present at baseline, n (%) | 0.28222 | |||
| No | 60 (52.2%) | 289 (57.7%) | 349 (56.7%) | |
| Yes | 55 (47.8%) | 212 (42.3%) | 267 (43.3%) | |
| Not tested6 | 4 | 27 | 31 | |
| Charlson comorbidities Index, n (%) | 0.05432 | |||
| 0–1 | 77 (64.7%) | 388 (73.5%) | 465 (71.9%) | |
| ≥2 | 42 (35.3%) | 140 (26.5%) | 182 (28.1%) | |
| Hospitalizations for pancreatitis in the past 1 year, n (%) | 0.06722 | |||
| No | 43 (37.7%) | 247 (47.1%) | 290 (45.5%) | |
| Yes | 71 (62.3%) | 277 (52.9%) | 348 (54.5%) | |
| Missing | 5 | 4 | 9 | |
| ED visits for pancreatitis in the past 1 year, n (%) | 0.00422 | |||
| No | 56 (49.1%) | 333 (63.5%) | 389 (61.0%) | |
| Yes | 58 (50.9%) | 191 (36.5%) | 249 (39.0%) | |
| Missing | 5 | 4 | 9 | |
| Distance between residence and medical center, miles, n (%) | 0.74342 | |||
| 0–25 | 29 (27.6%) | 167 (33.2%) | 196 (32.2%) | |
| 25–50 | 32 (30.5%) | 142 (28.2%) | 174 (28.6%) | |
| 50–100 | 21 (20.0%) | 93 (18.5%) | 114 (18.8%) | |
| ≥100 | 23 (21.9%) | 101 (20.1%) | 124 (20.4%) | |
| Missing | 14 | 25 | 39 | |
Wilcoxon rank sum p-value;
Chi-Square p-value; The univariate p-values in this table are from complete case analysis without missing data imputation.
Physical Component Score from PROMIS Global Health;
Mental Component Score from PROMIS Global Health;
Patients with no known EPD who did not have clinical steatorrhea and did not undergo per protocol fecal elastase testing;
Patients with no known diabetes did not undergo per-protocol diabetes testing.
Cumulative and Annualized Retention Rates
Figures 1 and 2 present the cumulative and annualized retention rates for the AP/RAP and CP groups, respectively. Among AP/RAP participants, the cumulative retention rate using all available data was 88.9% at year 1, increasing to 97.9% by year 5. When considering only in-person follow-up data, the cumulative retention rate ranged from 58.9% to 75.5% over the five-year period. Cumulative retention rates in the CP group were similar to those observed in the AP/RAP group for both all available data and in-person follow-up, as shown in Figure 2.
Figure 1.
Annualized and cumulative retention rates in the AP/RAP group
Figure 2.
Annualized and cumulative retention rates in the CP group
Annualized retention rates, displayed as stacked bars in Figures 1 and 2, illustrate the improvement in retention rates achieved through the inclusion of medical record reviews alongside in-person follow-up data.
Demographic and Clinical Characteristics by In-person Retention Status
Table 1 compares the demographic and clinical characteristics of participants with AP/RAP based on in-person retention. Participants with AP/RAP who returned for in-person follow-up visits were older (median age: 49 vs. 42 years, p<.0001), married (66.2% vs. 51.7%), had higher educational attainment (47.3% with a bachelor's degree or higher vs. 33.8%) and higher family incomes (47.8% with ≥$100,000 vs. 30.9%). They also lived closer to the clinical center than those without in-person retention (Table 1). There were no significant differences in disease-specific and other clinical characteristics, except for a higher median PROMIS Global Health Physical T-score in the group with in-person retention, indicating better overall physical health (median T-score: 44.9 vs. 42.3).
In the CP group, participants with in-person retention were older (18.9% aged >65 years vs. 7.6%), had higher educational attainment (32.3% with a bachelor's degree or higher vs. 23.6%), and a lower proportion of them were unemployed (31.4% vs. 46.4%), compared to participants with no retention. Additionally, a lower proportion of the CP group with retention reported ED visits for pancreatitis (36.5% vs. 50.9%) and had better overall physical health, as measured by PROMIS Global Health Physical score, compared to those with no retention.
Multivariable Logistic Regression Analysis
Table 3 presents the multivariable logistic regression analysis for the AP/RAP group, including variables that showed significant associations with in-person retention in at least one multiple imputation dataset. These variables included age, combined family income, the distance between residence and medical center, and PROMIS Global Health Physical T-score. The final model revealed that only age remained significantly associated with in-person retention (p=0.0001), with older participants having higher odds of in-person retention.
Table 3.
Results of the multivariable logistic regression analyses of factors associated with in-person retention in the AP/RAP group
| Parameter | Odds Ratio | 95% CI | P-value | Overall P-value |
|---|---|---|---|---|
| Age categories, years | 0.0001 | |||
| <35 | Ref | |||
| 35–50 | 1.011 | [0.728, 1.403] | 0.950 | |
| 50–65 | 1.397 | [0.992, 1.968] | 0.056 | |
| >65 | 1.525 | [0.866, 2.684] | 0.144 | |
| Combined family income, $ | 0.173 | |||
| <50k | Ref | |||
| 50–100k | 1.026 | [0.759, 1.386] | 0.867 | |
| ≥100k | 1.243 | [0.930, 1.661] | 0.141 | |
| Distance between residence and medical center, miles | 0.083 | |||
| 0–25 | Ref | |||
| 25–50 | 1.013 | [0.699, 1.468] | 0.945 | |
| 50–100 | 0.888 | [0.609, 1.295] | 0.536 | |
| ≥100 | 0.750 | [0.557, 1.009] | 0.057 | |
| PROMIS Global Health Physical T-score | 1.022 | [0.998, 1.046] | 0.076 | 0.076 |
Note: The risk factors in this table were identified from the model selection procedure. Intercept=1.197 (95% CI: 0.401–3.569)
In the CP group, variables significantly associated with in-person retention in at least one multiple imputation dataset included age, presence of exocrine pancreatic dysfunction at baseline, ED visits for pancreatitis in the past year, and PROMIS Global Health Physical T-score (Table 4). However, in the final multivariable analysis, only the PROMIS Global Health Physical T-score remained significantly associated with in-person retention (OR = 1.026; 95% CI: 1.002–1.051), with the effect size representing the impact of a one-point increase in the continuous T-score.
Table 4.
Results of the multivariable logistic regression analyses of factors associated with in-person retention in the CP group
| Parameter | Odds Ratio | 95% CI | P-value | Overall P-value |
|---|---|---|---|---|
| Age categories, years | 0.058 | |||
| <35 | Ref | |||
| 35–50 | 1.014 | [0.694,1.480] | 0.943 | |
| 50–65 | 0.771 | [0.556,1.069] | 0.119 | |
| >65 | 1.963 | [1.126,3.422] | 0.017 | |
| Exocrine pancreatic dysfunction present at baseline | ||||
| No | Ref | |||
| Yes | 0.861 | [0.622,1.193] | 0.347 | 0.335 |
| ED visits for pancreatitis in the past 1 year | ||||
| No | Ref | |||
| Yes | 0.826 | [0.667,1.023] | 0.079 | 0.079 |
| PROMIS Global Health Physical T-score | 1.026 | [1.002,1.051] | 0.034 | 0.034 |
Note: The risk factors in this table were identified from the model selection procedure. Intercept=1.646 (95% CI: 0.567–4.780)
Retention using all available data
Supplemental Tables 1 and 2 compare the sociodemographic and clinical characteristics in AP/RAP and CP groups based on the secondary outcome: retention using all available data (i.e., in-person follow-up/telephone interview and review of records). In the AP/RAP group, retention using all available data was significantly associated with age, marital status, and the presence of diabetes at baseline (Supplemental Table 1). In the CP group, the distance between the participant's residence and the medical center was the only factor significantly associated with retention using all available data sources (Supplemental Table 2).
Discussion
This analysis provides valuable insights into retention rates and factors associated with retention in the PROCEED study—a large, multicenter cohort of patients with pancreatitis. We observed high cumulative and annualized retention rates among participants with AP/RAP and CP in this study. Notably, incorporating medical record reviews into the follow-up process significantly enhanced data quality, allowing for a more comprehensive analysis of pancreatitis-related outcomes and disease progression—key objectives of the PROCEED study.
Despite the importance of participant retention in longitudinal research, studies examining determinants of retention remain limited, particularly in pancreatitis-related research. Socioeconomic factors—education level, income, and access to healthcare—are commonly associated with retention in clinical research.5, 15 Studies have linked certain participant characteristics- including older age, being a non-white male, low educational attainment, multiple comorbidities, and irregular healthcare utilization- to study attrition.16 However, research on these factors in pancreatic diseases is scarce. In our study, older age, higher education, employment or retirement, and higher family income were associated with in-person retention among AP/RAP participants in unadjusted analyses. These findings suggest that financial stability and flexible schedules may facilitate engagement in clinical research. Similarly, in the CP group, older age, higher education and income, and employment/retirement were associated with in-person retention, reinforcing the positive role of socioeconomic stability in sustained study participation. Interestingly, multivariable analyses demonstrated that older age was significantly associated with in-person retention only in the AP/RAP group.
Our findings align, in part, with a recent Nationwide Inpatient Sample (NIS) analysis of the Healthcare Cost and Utilization Project (HCUP), which identified younger age, male sex, Black race, lower socioeconomic status, alcohol etiology of pancreatitis, smoking, and depression as factors associated with discharge against medical advice (AMA) in pancreatitis hospitalizations.17 The overlap between these AMA-related characteristics in pancreatitis-related hospitalizations and our findings highlights the complex interplay of socioeconomic and behavioral factors influencing healthcare engagement in this patient population, underscoring the need for investigators to develop strategies for retention that target these potential barriers.
Health status, including comorbidities and severity of symptoms, has been shown to influence participants' ability to remain in longitudinal studies.18 In our study, participants with AP/RAP and CP who returned for in-person follow-up reported a better perception of physical health, as indicated by higher PROMIS Global Physical Health T-scores, compared to those who did not return for follow-up. Specifically, in-person retention among participants with CP was significantly associated with better physical health in multivariable analyses. Interestingly, multiple comorbidities were not significantly associated with retention in either disease group. Prior research suggests that comorbid conditions and age-related health issues often hinder study participation and retention.19 Moreover, certain comorbidities are common exclusion criteria in many research studies. In the AP/RAP group, we observed a low prevalence of comorbidities at enrollment, which may explain the lack of significant association with in-person retention.
Our study supports prior research demonstrating that shorter distances to clinical sites are associated with higher retention rates.20, 21 In unadjusted analyses, AP/RAP participants who returned for in-person follow-up tended to live closer to clinical centers than those lost to follow-up. This finding aligns with studies showing that transportation challenges and financial burden are key considerations for study participation.22, 23 However, the association between proximity to the clinical center and study retention was not observed in the CP group, underscoring the varying challenges of participant retention across the spectrum of pancreatic diseases.
Recruiting and retaining participants from socioeconomically diverse backgrounds remains a challenge in clinical research. Contributing factors include mistrust in research, economic constraints, and negative healthcare experiences.24 CP patients are particularly vulnerable, as disease-related treatment costs can impose a significant financial burden, leading to adverse health outcomes.25 In our study, the majority of CP participants reported low household incomes, with over half of those who did not return for in-person follow-up earning less than $50,000 annually. These findings highlight the need for additional research on healthcare disparities in pancreatitis, which are driven by both patient-level and healthcare system factors and likely contribute to attrition in longitudinal studies.26
Maintaining high retention rates in prospective cohorts, such as PROCEED, is essential for advancing our understanding of pancreatitis disease progression and informing clinical care strategies. Successful retention relies on strong participant engagement, community support, and dedicated study personnel.2 Recent studies and systematic reviews highlight several effective retention strategies, including personalized participant outreach (e.g., reminders, flexible scheduling); reducing participant burden (e.g., electronic data collection as an alternative to in-person visits); emphasizing the significance of study participation; and engagement with patient advocacy group/representatives during study planning.1, 3, 27–30
These strategies must also address the increasing impact of study burden on participants, which can significantly impact retention. Factors such as the length and frequency of study commitments, travel challenges (e.g., distance, transportation access, parking difficulties), scheduling constraints (e.g., fasting requirements, weekday appointments), and the complexity or perceived risk of study procedures may deter continued participation. A comprehensive retention strategy should anticipate, address, and accommodate potential socioeconomic and logistical barriers to support continued participation and long-term engagement.31
In PROCEED study, ongoing efforts to promote participant retention include research payments for follow-up visits, travel reimbursements informed by study coordinator feedback, the dissemination of a study newsletter to all participants, and coverage of costs for certain protocol-mandated tests such as MRI, CT, and/or DXA scans. Additionally, PROCEED has established a Recruitment and Retention Committee—a dedicated team of investigators and study coordinators that continuously monitors and optimizes study follow-up.
Strengths and Limitations
A key strength of this study includes a large, well-characterized, multicenter prospective cohort, which allowed for a robust analysis of retention factors in patients with different clinical stages of pancreatitis. The collection of comprehensive patient- and disease-specific data using standardized methods further strengthened this study. Our analyses demonstrated that medical record reviews enhanced data accuracy, providing a more comprehensive assessment of pancreatitis-related outcomes and disease progression. This approach could be valuable in other research studies as a strategy to optimize data collection when in-person visits are not feasible. However, several limitations should be acknowledged. Some data were self-reported, which may introduce recall or response bias. Nevertheless, the inclusion of physician-reported and medical record data helps mitigate this concern. An inherent limitation of any analysis of participant attrition is the incomplete collection of time-varying data that may influence continued participation among individuals who did not return for follow-up, such as quality of life, pain levels, and new or competing comorbidities. Future analyses will focus on characterizing comorbidities within this cohort. Additionally, data loss may occur if patients receive care at institutions other than the participating site, where medical records are not accessible. Furthermore, our analysis primarily focused on baseline characteristics associated with retention, and future research will be needed to explore time-varying factors that may influence study participation over time. Finally, our study was unable to assess the potential influence of the study team, including study coordinators, on retention.
Conclusions
We observed that incorporating medical record reviews enhanced data availability, which will support a more comprehensive assessment of pancreatitis-related outcomes and disease progression in future analyses. Sociodemographic factors—particularly age—were key predictors of in-person retention in AP/RAP participants, while patient-reported physical health perception was a significant predictor in the CP group. These findings suggest that tailored strategies targeting these factors could enhance long-term participant engagement in pancreatitis research.
Supplementary Material
Acknowledgments
This study was supported by the National Cancer Institute and National Institute of Diabetes and Digestive and Kidney Diseases under the following award numbers: U01DK108288 (S.T.C., M.T.), U01DK108300 (W.G.P.), U01DK108306 (D.Y.), U01DK108314 (C.Y.J., S.J.P.), UO1DK108300 (C.E.F., S.J.H.), U01DK108323 (E.F., J.E.), U01DK108326 (W.E.F., M.O.O.), U01DK108327 (P.H., D.L.C.), U01DK108332 (S.K.V), and U01DK108328 (Z.F., L.L.). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. The other authors declare no conflict.
REFERENCES
- 1.Booker CL, Harding S, Benzeval M. A systematic review of the effect of retention methods in population-based cohort studies. BMC public health. 2011;11:1–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Price S, Park C, Brenner R, et al. Participant retention in a longitudinal study: Do motivations and experiences matter? Survey Practice 9 (4): 1–10. 2016. [Google Scholar]
- 3.Abshire M, Dinglas VD, Cajita MI, Eakin MN, Needham DM, Himmelfarb CD. Participant retention practices in longitudinal clinical research studies with high retention rates. BMC Med Res Methodol. Feb 20 2017;17(1):30. doi: 10.1186/s12874-017-0310-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Deeg DJ. Attrition in longitudinal population studies: does it affect the generalizability of the findings?: an introduction to the series. Journal of Clinical Epidemiology. 2002;55(3):213–215. [Google Scholar]
- 5.Wheeler AJ, Garg H, Kaushik D, Mansour A, Pruthi D, Liss MA. Predictors of participants' retention-socioeconomic factors or nonadherence: insights from a urological clinical prospective study. Trials. Dec 2 2022;23(1):970. doi: 10.1186/s13063-022-06901-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Fewtrell MS, Kennedy K, Singhal A, et al. How much loss to follow-up is acceptable in long-term randomised trials and prospective studies? Archives of disease in childhood. 2008;93(6):458–461. [DOI] [PubMed] [Google Scholar]
- 7.Kristman V, Manno M, Côté P. Loss to follow-up in cohort studies: how much is too much? European journal of epidemiology. 2004;19:751–760. [DOI] [PubMed] [Google Scholar]
- 8.Bang UC, Benfield T, Hyldstrup L, Bendtsen F, Jensen JEB. Mortality, cancer, and comorbidities associated with chronic pancreatitis: a Danish nationwide matched-cohort study. Gastroenterology. 2014;146(4):989–994. e1. [DOI] [PubMed] [Google Scholar]
- 9.Yadav D, Park WG, Fogel EL, et al. PROspective Evaluation of Chronic Pancreatitis for EpidEmiologic and Translational StuDies (PROCEED): Rationale and Study Design From the Consortium for the Study of Chronic Pancreatitis, Diabetes, and Pancreatic Cancer. Pancreas. 2018;47(10):1229. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Banks PA, Bollen TL, Dervenis C, et al. Classification of acute pancreatitis—2012: revision of the Atlanta classification and definitions by international consensus. Gut. 2013;62(1):102–111. [DOI] [PubMed] [Google Scholar]
- 11.Roffman CE, Buchanan J, Allison GT. Charlson comorbidities index. Journal of physiotherapy. 2016;62(3):171. [DOI] [PubMed] [Google Scholar]
- 12.Measures Health. Transforming How Health is Measured. Accessed March 17, 2025. https://www.healthmeasures.net/explore-measurement-systems/promis [Google Scholar]
- 13.Yadav D, Askew RL, Palermo T, et al. Association of chronic pancreatitis pain features with physical, mental, and social health. Clinical Gastroenterology and Hepatology. 2023;21(7):1781–1791. e4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Little R, Rubin D. Multiple imputation for nonresponse in surveys. Wiley. 1987;10:9780470316696. [Google Scholar]
- 15.Sharrocks K, Spicer J, Camidge D, Papa S. The impact of socioeconomic status on access to cancer clinical trials. British journal of cancer. 2014;111(9):1684–1687. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Davis LL, Broome ME, Cox RP. Maximizing retention in community-based clinical trials. Journal of Nursing Scholarship. 2002;34(1):47–53. [DOI] [PubMed] [Google Scholar]
- 17.Chaudhary F, Albeiruti R, Alqahtani F, Alhajji M, Lerfald N, Hutson W. Temporal trends and predictors of pancreatitis patients who leave against medical advice: a nationwide analysis. Gastroenterology Research. 2020;13(2):58. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Zhang Y, Pratap A, Folarin AA, et al. Long-term participant retention and engagement patterns in an app and wearable-based multinational remote digital depression study. NPJ digital medicine. 2023;6(1):25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Forsat ND, Palmowski A, Palmowski Y, Boers M, Buttgereit F. Recruitment and retention of older people in clinical research: a systematic literature review. Journal of the American Geriatrics Society. 2020;68(12):2955–2963. [DOI] [PubMed] [Google Scholar]
- 20.Spring B, Sohn M-W, Locatelli SM, Hadi S, Kahwati L, Weaver FM. Individual, facility, and program factors affecting retention in a national weight management program. BMC public health. 2014;14:1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Beardsley K, Wish ED, Fitzelle DB, O'Grady K, Arria AM. Distance traveled to outpatient drug treatment and client retention. Journal of Substance Abuse Treatment. 2003;25(4):279–285. [DOI] [PubMed] [Google Scholar]
- 22.Odierna DH, Bero LA. Retaining participants in outpatient and community-based health studies: researchers and participants in their own words. SAGE open. 2014;4(4):2158244014554391. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Winkfield KM, Phillips JK, Joffe S, Halpern MT, Wollins DS, Moy B. Addressing financial barriers to patient participation in clinical trials: ASCO policy statement. Journal of Clinical Oncology. 2018;36(33):3331–3339. [DOI] [PubMed] [Google Scholar]
- 24.Ejiogu N, Norbeck JH, Mason MA, Cromwell BC, Zonderman AB, Evans MK. Recruitment and Retention Strategies for Minority or Poor Clinical Research Participants: Lessons From the Healthy Aging in Neighborhoods of Diversity Across the Life Span Study. The Gerontologist. 2011;51(suppl_1):S33–S45. doi: 10.1093/geront/gnr027 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Choate R, Wasilchenko C, Thakur K, Hill R, Wright E, Conwell DL. Financial Toxicity in Patients With Chronic Pancreatitis. Pancreas. 2024;53(9):e774–e779. doi: 10.1097/mpa.0000000000002384 [DOI] [PubMed] [Google Scholar]
- 26.Choate R, Bradley D, Conwell D, Yazici C. Healthcare disparities in pancreatitis: knowledge gaps and next steps. Current Opinion in Gastroenterology. 2024;40(5):422–430. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Costello L, Dare J, Dontje M, Lambert C, Straker L. Applying the 4Ps of social marketing to retain and engage participants in longitudinal cohort studies: generation 2 Raine study participant perspectives. BMC Medical Research Methodology. 2022;22(1):288. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Robinson KA, Dinglas VD, Sukrithan V, et al. Updated systematic review identifies substantial number of retention strategies: using more strategies retains more study participants. Journal of clinical epidemiology. 2015;68(12):1481–1487. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Teague S, Youssef GJ, Macdonald JA, et al. Retention strategies in longitudinal cohort studies: a systematic review and meta-analysis. BMC medical research methodology. 2018;18:1–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Bower P, Brueton V, Gamble C, et al. Interventions to improve recruitment and retention in clinical trials: a survey and workshop to assess current practice and future priorities. Trials. 2014;15:1–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Dimakos C, Loomis C, Gilmer A, et al. Locating longitudinal study participants 10 years after last contact: Contemporary approaches to sample retention. Journal of Community Psychology. 2023;51(4):1540–1559. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.