Factors associated with patient trust in physicians: domain-specific association analyses and exploratory prediction in a cross-sectional web panel survey in Japan

Patient trust in physicians is central to the physician–patient relationship. However, the extent to which readily observable care-process, linguistic, and physician-related variables explain trust remains unclear. We examined three conceptually motivated domains and the broader predictability of trust from questionnaire data. We conducted a cross-sectional survey using a non-probability commercial web panel with nationwide geographic coverage among adults in Japan who had a usual physician. Trust100, a 0–100 score derived from five items based on the Japanese version of the Abbreviated Wake Forest Interpersonal Trust in Physician Scale, was the outcome. Domain-specific association analyses examined waiting time, physician–respondent local-dialect use, and attitudes toward and awareness of the physician’s medical school. Adjusted linear models used heteroskedasticity-robust covariance estimates, with Holm correction across three global domain-level tests. Separately, mean-prediction, ordinary least-squares, ridge, and Extra Trees models were compared using repeated nested cross-validation. Consultation time was examined post hoc after showing the highest mean permutation importance. The study was not prospectively registered, although a dated study protocol and questionnaire had been prepared before data collection. Among 2,173 respondents, mean Trust100 was 66.8. Waiting time showed a weak negative association, but the adjusted association did not remain statistically significant after Holm correction (adjusted p = 0.073; partial R² = 0.0043). No clear association was observed for physician–respondent dialect-use combinations in the pooled region-adjusted analysis ( p = 0.783; partial R² = 0.0005), with no statistical evidence of regional heterogeneity. The global test for the medical-school attitude/awareness domain remained statistically significant after Holm correction, but explanatory contribution was limited (partial R² = 0.022), the pattern was non-monotonic, and no interaction was observed. In a post hoc descriptive comparison, respondents reporting consultation times of ≥ 3 min had a mean Trust100 score 7.69 points higher than those reporting < 3 min (95% confidence interval, 6.10 to 9.27). Patient trust was only weakly explained and predicted by the measured questionnaire variables. Unmeasured relational or contextual factors and measurement limitations may both contribute.