Assessing Social Determinants of Health Training in Graduate Medical Education: A Narrative Review Using Kirkpatrick’s Model

TL;DR

This paper reviews how medical training on social determinants of health is evaluated, finding that most studies focus on basic outcomes like satisfaction rather than real-world impact.

Contribution

The paper introduces a novel integration of Kirkpatrick’s model, GRADE quality assessment, and bibliometric analysis to evaluate SDOH training in medical education.

Findings

79% of studies focused on Kirkpatrick Levels 1/2 (Reaction and Learning), with limited emphasis on behavioral or patient outcomes.

Most studies were rated low quality by GRADE due to observational or quasi-experimental designs.

Bibliometric analysis revealed three institutional clusters of collaboration in SDOH research.

Abstract

Social determinants of health (SDOH) training is a critical component of medical education, equipping physicians to address factors such as economic stability, education, and health care access. Despite its importance, the evaluation of SDOH training often focuses on lower-level outcomes, such as learner satisfaction and knowledge assessment, with limited emphasis on behavioral change and patient outcomes. Kirkpatrick’s model offers a validated framework for addressing this gap by categorizing training outcomes into four levels: Reaction, Learning, Behavior, and Results.

By applying Kirkpatrick’s model in combination with Grading of Recommendations Assessment, Development and Evaluation (GRADE) quality assessment and bibliometric network analysis, this review seeks to assess SDOH training evaluation and provide insights into designing curricula that produce measurable improvements in physician behavior and patient care outcomes.

This narrative review analyzes the efficacy of SDOH training programs in graduate medical education using studies identified in two recent scoping reviews (2019, 2024). Studies were categorized by their highest reported Kirkpatrick level: Levels 1/2 (Reaction and Learning) or Levels 3/4 (Behavior and Results). If a study spanned multiple levels, it was categorized at the highest level. This review further uniquely integrates GRADE and bibliometric co-authorship network analysis to explore evidence quality and author collaborations.

A total of 33 studies were analyzed, with 79% classified as Levels 1/2 and 21% as Levels 3/4. Most studies used Level 1 evaluation methods, primarily subjective surveys. Among Level 3/4 studies, 71% used career tracking as a proxy for behavioral change, and only three assessed patient outcomes (Level 4). Most studies also reported positive results, with two noting alternate outcomes related to program length and parental trust.

GRADE analysis rated all studies as low or very low quality, reflecting the predominance of observational or quasi-experimental designs without randomization or blinding. Only two studies incorporated control groups but still received low ratings due to high risk of bias and lack of replication. Bibliometric co-authorship network analysis identified three distinct institutional clusters of collaboration: one centered around Cincinnati Children’s Hospital (2014-2018), one around Montefiore Medical Center (2010-2012), and a dyad at Johns Hopkins Hospital (2016).

Most SDOH training evaluations focused on Kirkpatrick Levels 1/2, offering limited insight into real-world impact. While Kirkpatrick’s framework helps organize outcomes, it is limited, as interventions with multifactorial effects may not directly link to training. Consistently low GRADE ratings and limited collaboration across institutions highlight the need for stronger study designs and more coordinated research efforts.

Most SDOH training evaluations rely on subjective outcomes and are supported by low-quality evidence, limiting conclusions about impact on clinical practice. Without rigorous, higher-level evaluation, the true impact of SDOH training in graduate medical education on clinical practice and health equity remains uncertain. Advancing SDOH education will require longitudinal curricula, objective and patient-centered outcomes, and institutionally supported, multi-site study designs to generate generalizable evidence.

Figures