Regularly scheduled educational conferences, such as case-based morning reports, have been a standard part of internal medicine residencies for decades.1-4 In addition to better patient care from the knowledge gained at educational conferences, attendance by interns and residents (collectively called house staff) may be associated with higher in-service examination scores.5 Unfortunately, competing priorities, including patient care and trainee supervision, may contribute to an action-intention gap among house staff that reduces attendance.6-8 Low attendance at morning reports represents wasted effort and lost educational opportunities; therefore, strategies to increase attendance are needed. Of several methods studied, more resource-intensive interventions (eg, providing food) were the most successful.6,9-12
Using the behavioral economics framework of nudge strategies, we hypothesized that a less intensive intervention of a daily reminder text page would encourage medical students, interns, and residents (collectively called learners) to attend the morning report conference.8,13 However, given the high cognitive load created by frequent task switching, a reminder text page could disrupt workflow and patient care without promoting the intended behavior change.14-17 Because of this uncertainty, our objective was to determine whether a preconference text page increased learner attendance at morning report conferences.
Methods
This study was a single-center, multiple-crossover cluster randomized controlled trial conducted at the Veteran Affairs Boston Healthcare System (VABHS) in Massachusetts. Study participants included house staff rotating on daytime inpatient rotations from 4 residency programs and students from 2 medical schools. The setting was the morning report, an in-person, interactive, case-based conference held Monday through Thursday, from 8:00 am to 8:45 am. On Friday mornings, the morning report was replaced with a medical Jeopardy game-style conference. Historically, attendance has not been recorded for these conferences.
Learners assigned to rotate on the inpatient medicine, cardiology, medicine consultation, and patient safety rotations were eligible to attend these conferences and for inclusion in the study. Learners rotating in the medical intensive care unit, on night float, or on day float (an admitting shift for which residents are not on-site until late afternoon) were excluded. Additional details of the study population are available in the supplement (eAppendix). The study period was originally planned for September 30, 2019, to March 31, 2020, but data collection was stopped on March 12, 2020, due to the COVID-19 pandemic and suspension of in-person conferences. We chose the study period, which determined our sample size, to exclude the first 3 months of the academic year (July-September) because during that time learners acclimate to the inpatient workflow. We also chose not to include the last 3 months of the academic year to provide time for data analysis and preparation of the manuscript within the academic year.
Intervention and Outcome Assessment
Each intervention and control period was 3 weeks long; the first period was randomly determined by coin flip and alternated thereafter. Additional details of randomization are available in the supplement (Appendix 1). During intervention periods, all house staff received a page at 7:55 am that listed the time and location of the upcoming morning report or Jeopardy conference. Medical students do not carry pagers and did not receive reminder pages; however, we included these learners because changes in their conference attendance behavior would indicate an extension of the effect of reminder pages beyond the individual learner who received the page.
A daily facesheet (a roster of house staff names and photos) was used to identify learners for conference attendance. This facesheet was already used for other purposes at VABHS. At 8:00 am and 8:10 am, a chief medical resident who was not blinded to the intervention or control period recorded the attendance of each eligible learner as present or absent; learners were unaware that their attendance was being recorded. This approach to data collection was selected to minimize the likelihood that the behavior of the study participants would be influenced.
During control periods, no text page reminder of upcoming conferences was sent, but the attendance of total learners at 8:00 am and 8:10 am was recorded by a chief medical resident who used the same method as during the intervention periods. Attendance at 8:10 am was chosen as the primary outcome to account for the possibility that learners may arrive after a conference begins. Attendance at 8:00 am also was recorded to assess the effect of reminder pages on attendance at the start of morning reports.
Statistical Analysis
The primary outcome was the proportion of eligible learners present at 8:10 am at the morning report, expressed as the risk difference for attendance between intervention and control periods. Secondary outcomes included the proportion of learners present at 8:00 am (on-time attendance), the proportion of learners present by type (student vs house staff), and the proportion of learners present at the Friday Jeopardy conference. Two preplanned subgroup analyses were performed: one assessing the impact of rotating on clinical services with lighter workloads, and the other assessing the impact of the number of overnight admissions received on the relationship between receipt of a reminder page and conference attendance.
To estimate the primary outcome, we modeled the risk difference adjusted for covariates using a generalized estimating equation accounting for the clustering of attendance behavior within individuals and controlling for date and team. Secondary outcomes were estimated similarly. To evaluate the robustness of the primary outcome, we performed a sensitivity analysis using a multilevel generalized linear model with clustering by individual learner and team. Additional details on our statistical analysis plan, including accessing our raw data and analysis code, are available in Appendices 2 and 3. Categorical variables were compared using the χ2 or Fisher exact test. Continuous variables were compared using the t test or Wilcoxon rank-sum tests. All P values were 2-sided, and a significance level of ≤ .05 was considered statistically significant. Analysis was performed in Stata v16.1. Our study was deemed exempt by the VABHS Institutional Review Board, and this article was prepared following the CONSORT reporting guidelines. The trial protocol has been registered with the International Standard Randomized Controlled Trial Number registry (ISRCTN14675095).