University of Arizona College of Medicine – Phoenix is a 4-year allopathic school with an abbreviated preclinical curriculum and entry into clerkships 3 months earlier than students at many other medical schools. Traditionally, the Year 3 curriculum comprised six core clerkships — two of which were 12 weeks in length (Internal Medicine and Surgery) and four of which were 6 weeks in length (Obstetrics and Gynecology (OB/GYN); Family, Community, and Preventive Medicine (Family Medicine or FM); Pediatrics; and Psychiatry). Core clerkships in Neurology and Emergency Medicine occurred in Year 4; each was 4 weeks long. Beginning in the 2019–2020 academic year, the Internal Medicine (IM) and Surgery clerkships were shortened to 8 weeks in order to permit movement of the 4-week Neurology clerkship and one 4-week elective from year 4 into year 3. These curricular changes were made to allow earlier individual career exploration through an elective and to provide better preparation for USMLE Step 2 CK through earlier Neurology exposure. To accommodate the reduction in IM and Surgery clerkship length, time within ambulatory IM and surgical subspecialties was reduced, with the option for students to obtain additional experience in these areas through year 4 electives. There were no additional changes to the structure of the core clerkship curricula. A longitudinal patient care course spanning years 3 and 4 similarly remained unchanged in the new clinical curriculum. The Emergency Medicine (EM) clerkship remained in year 4 without change.

All medical students at the University of Arizona College of Medicine – Phoenix who completed a clinical clerkship and NBME subject examination between the periods of 2017–2018, 2018–2019, 2019–2020, and 2021–2022 were included as participants, provided they had signed consent to participate in medical education research under the Research Office for Medical Education (ROME) Umbrella IRB for medical student data at our institution. Students completing core clerkships during academic year 2020–2021 were excluded because of significant pandemic-related alteration in clerkship length and curricula during that year. Data from the six core clerkships occurring in year 3 (FM, IM, OB/GYN, Pediatrics, Psychiatry, and Surgery) were included. EM and Neurology clerkship data were excluded because these clerkships were not taken consistently in the third year, and more clinical experience was considered a confounding factor. All research conducted within this study was approved under the ROME umbrella IRB and the Committee for Assessment and Evaluation of Submission Approval for Research (CAESAR) governed by ROME.

The first hypothesis tested in the study is that reducing clerkship length does not impact student performance on the associated NBME subject exam. This metric, rather than clinical performance ratings or final clerkship grades, was chosen to avoid potential bias introduced by subjective rating patterns of individual faculty or by changes in faculty across academic years. Students from the traditional longer clerkship cohorts (academic years 2017–2018 and 2018–2019) served as the control or pre-intervention group. Students from the 2019–2020 and 2021–2022 cohorts were in the experimental or post-intervention group, with the IM and Surgery clerkship lengths each shortened from 12 to 8 weeks for this group. Other clerkship exam scores (OB/GYN, FM, Pediatrics, and Psychiatry) within these four cohorts of students served as internal controls, as their lengths remained the same.

We also tested a second hypothesis: the timing of an individual clerkship within the academic year’s sequence of core clerkships does impact student performance on the corresponding NBME subject exam. When testing this hypothesis, the IM, Surgery, Pediatrics, FM, OB/GYN, and Psychiatry subject exam scores were included for all four cohorts of students. In addition, individual student MCAT exam performance was examined as a potential confounding variable (data included for all four cohorts of students).

Information was extracted pertaining to NBME subject exam scores, MCAT performance, and the sequence of NBME subject exams taken for four cohorts of medical students at the University of Arizona College of Medicine – Phoenix. Descriptive statistics (namely, mean and standard deviation for continuous measures, and frequency and percentage for nominal measures) were calculated for these extracted measures. Differences across cohorts in these measures were assessed via one-way ANOVAs for continuous measures and Pearson’s chi-square tests for nominal measures. For each of these statistical analyses, and those discussed below, significance was set at p ≤ 0.05, and the analyses were completed using STATA 18 [15].

During this data extraction process, investigators were blinded to student identity and gender to maintain student anonymity. While certain demographic characteristics (age and race/ethnicity) were available, the inclusion of this data in the analyses discussed below was withheld to avoid identifying individual subjects and maintain student privacy protections afforded by the Family Educational Rights and Privacy Act (FERPA). Differences in these measures across cohorts were analyzed and identified when statistically significant.

To test our first hypothesis (that shortened clerkship length does not impact subject exam scores), a two-sample t-test comparing scores on IM and Surgery subject exams between the control and experimental groups was completed. The two-sample t-test assumed homogeneity of variances, with the independent variable set as the student group (pre- or post-intervention) and the dependent variable as average score on the individual NBME subject exam (IM and Surgery considered separately). We also completed one-way ANOVAs comparing exam scores across all four cohorts (academic years) for IM and Surgery. For this single-factor ANOVA model, the independent variable was set as the cohort, and the dependent variable was the average score on the individual NBME subject exam. This second analysis was performed to evaluate for any meaningful differences in cohort exam performance within the pre- and post-intervention groups that would be missed with just the pre- vs. post-intervention comparison.

While the length of the remaining year 3 core clerkships did not change, we also completed the same two-sample t-test and one-way ANOVA analyses on these subject exams to evaluate for any effect of shortened IM and Surgery clerkships on NBME subject exam performance in other areas. This approach of evaluating the potential effect of shortened IM and Surgery clerkships within all year 3 core clerkships was then extended through the inclusion of restricted maximum likelihood (REML) mixed-effects regression analyses that evaluated the impact of this clerkship change across all six NBME subject exam score areas. The first (of two) of these mixed-effects models included our measure of before and after the clerkship change intervention, subject exam area, and the interaction between the two as fixed effects, with the individual student set as the random effect. The second, full model also included MCAT performance and block sequence as additional covariates.

Individual student MCAT performance was an important inclusion as a covariate, as the literature demonstrates that MCAT scores are predictive of performance on USMLE and NBME exams and of year 3 grades [16,17,18,19]. MCAT scores reported on the old reporting scale (prior to April 2015) were converted to the new scoring scale via an online score converter to enable comparison across the four cohorts [20]. The correlation between MCAT score and NBME exam performance (and its relevance here) was assessed through (1) ANCOVAs evaluating the association between MCAT scores and NBME subject exam scores when factoring in differences pre- vs. post- intervention for each clerkship; and (2) subsequent regression analyses, including ultimately the full mixed-effects model discussed above. Each of these regression analyses assumed a linear relationship between MCAT performance and subject exam score, as shown in Fig. 1a and b found in the “Results” section [16,17,18,19].

a Linear regression model of IM NBME subject exam score on MCAT score. b Linear regression model of Surgery NBME subject exam score on MCAT score

Additionally, to test our secondary hypothesis (that NBME subject exam scores could be impacted by timing of the associated clerkship within the sequence of the academic year), the block sequence in which a given NBME subject exam was taken was included as a covariate as well. The correlation between this sequence measure and NBME exam performance was assessed through two-way ANOVAs evaluating the association between sequence and NBME subject exam scores when factoring in differences pre- vs. post- intervention for each clerkship and through subsequent regression analyses, which include the full mixed-effects model discussed previously. Each of these regression analyses assumed a unique, block-specific relationship with subject exam score, and that NBME exam order reflected clerkship sequence, since students take the exam at the end of the clerkship block before starting their next rotation [4].