Many clinical research studies collect data repeatedly overtime. A participant maybe screened, enrolled, reviewed on day 7, assessed on day 28, followed monthly, and closed out at the end of the study. Longitudinal data collection requires careful planning because repeated measurements create design and analysis complexities.
The data manager must identify which variables are collected once and which are repeated.
Date of birth, sex, and baseline demographics are usually collected once. Weight, symptoms, laboratory values, medication adherence, adverse events, and outcome measures may be collected repeatedly. If repeated variables are not structured properly, the final dataset may become difficult to analyze.
Visit schedules should define expected visit windows. For example, a day 28 visit may allow visits from day 25 to day 35. The CRF should capture the actual visit date, not merely whether the visit occurred. The actual date allows the statistician to determine whether the visit was within window, calculate time intervals, and handle late or early visits appropriately. Without actual dates, data managers lose important information. Longitudinal studies must also handle missed visits, unscheduled visits, and early termination. A participant may miss a scheduled follow-up, return to clinic unexpectedly, withdraw consent, transfer to another facility, or die. The database should distinguish these scenarios rather than treating them all as missing data. Missing because a participant died is very different from missing because a form was not entered.
In REDCap, longitudinal design may be implemented using events, repeating instruments, or both. Events are useful when visits are scheduled and predictable, such as screening, enrollment, day 7, and day 28. Repeating instruments are useful when the number of observations is unknown, such as adverse events, concomitant medications, laboratory samples, or hospital admissions. The design choice should follow the protocol and analysis requirements.
Visit schedule design should involve investigators, coordinators, data managers, and statisticians. Coordinators understand workflow, investigators understand clinical relevance, statisticians understand analysis needs, and data managers understand database implications. When these perspectives are integrated early, the resulting CRFs are more likely to be feasible anduseful.