Examination performance with flipped classroom as instructional strategy in the carbohydrate metabolism course unit at a Philippine medical school: Estimation of average treatment effect from observational data

Michael Van Haute, De La Salle Medical and Health Sciences Institute
Jan David Monzon, De La Salle Medical and Health Sciences Institute


While the innovativeness of the flipped classroom (FC) approach promotes active participation and higher-order thinking among students, there are concerns about its effectiveness in terms of knowledge retention. Currently, there are no studies involving medical school biochemistry that evaluate this aspect of effectiveness. Thus, we conducted a historical control study that analyzed observational data from two freshman batches of the Doctor of Medicine program in our institution. Class 2021 (n = 250) served as the traditional lecture (TL) group while Class 2022 (n = 264) served as the FC group. Data on relevant observed covariates (age, sex, National Medical Admission Test or NMAT score, undergraduate degree) and the outcome variable (carbohydrate metabolism course unit examination percentage scores, as indicator of knowledge retention) were included in the analysis. Propensity scores were calculated using logit regression conditional on these observed covariates. Afterwards, 1:1 nearest-neighbor propensity score matching (PSM) was performed to produce an estimated average treatment effect (ATE) measure afforded by FC (as adjusted mean difference in examination scores between the two batches), balancing on the covariates. Nearest-neighbor matching using the calculated propensity scores effectively balanced the two groups (standardized bias <10%), producing 250 matched student-pairs that received either TL or FC. Following PSM, the FC group was found to have a significantly higher adjusted mean examination score compared to the TL group (adjusted mean difference = 5.62%, 95% CI: 2.54%, 8.72%; p < 0.001). Using this approach, we were able to demonstrate benefit of FC over TL in terms of knowledge retention, as reflected by the estimated ATE.