Nonresponse Bias Confounds Self-Reported Mistreatment by Diverse Physician Associate Students

INTRODUCTION: This study aimed to evaluate whether physician assistant/associate (PA) students' sociodemographic factors were predictors of risk for mistreatment. A secondary analysis aimed to evaluate whether sociodemographic features were evenly distributed among respondents who answered mistreatment items. METHODS: Data originated from the PA Education Association End of Program Surveys (2018, 2019, 2021, and 2022). Independent variables were gender, race, ethnicity, and sexual orientation. Using logistic regression, odds ratios were calculated for 2 separate dependent variables: whether respondents experienced mistreatment and whether respondents completed mistreatment items. RESULTS: Surveys included 11,461 respondents, 3218 (28.1%) of whom experienced mistreatment; however, 3258 (28.4%) of respondents did not answer any mistreatment items. Analysis showed statistically significant sociodemographic factors for risk for mistreatment, but the model had poor fit ( P -value = 0; receiver operating characteristic [ROC] 0.553), possibly related to partial nonresponse bias. Completing mistreatment items was statistically significantly more likely for respondents who indicated they were gay or lesbian (odds ratio [OR] 1.52) or bisexual (OR 1.82) and less likely for respondents who indicated they were male (OR 0.68), sexual orientation "I don't know/prefer not to answer" (OR 0.65), Hispanic (OR 0.79), or not White (OR 0.49). DISCUSSION: The sociodemographic factors evaluated were inadequate to predict mistreatment of PA students, but sociodemographic factors were associated with willingness to complete mistreatment questions. Qualitative research is needed to determine why respondents who are male, Hispanic, or not White are reluctant to complete mistreatment questions. Findings could inform survey improvements to more accurately measure health professions student mistreatment.