What Patients Reward and Punish in Physician Communication: A Framework-Based Analysis of 62,319 Indian Reviews

Background: Patient-generated online reviews contain detailed accounts of physician communication that remain systematically underused in health services research. Prior computational analyses have either applied service quality frameworks without grounding in clinical communication theory, or have been geographically and specialty-restricted. Objective: To examine which physician communication dimensions predict patient recommendation, switching intent, and retention across specialties and cities in India. Methods: We analyzed 62,319 patient reviews from the Practo platform (10 cities, 8 specialties) coded against established clinical communication frameworks, the Kalamazoo Consensus Statement, Calgary-Cambridge Guide, and SPIKES protocol, adapted into a 12-dimension codebook. Expert confirmatory review on 200 stratified reviews yielded 98.5% agreement between reviewer and model labels across 11 communication dimensions. Binary recommendation was designated the primary outcome; recommendation intensity is reported as secondary. Logistic regression with 1,000 bootstrap iterations examined associations between communication dimensions and recommendation and switching intent, adjusting for treatment outcome, cost, wait time, and specialty. Results: Reassurance (OR 2.07, 95% CI 1.86-2.42) and empathy (OR 1.89, 95% CI 1.79-2.00) were the strongest positive predictors of recommendation. Excessive directiveness (OR 0.62, 95% CI 0.60-0.64) and rushedness (OR 0.65, 95% CI 0.64-0.67) were the dominant negative predictors; directiveness was present in 33.3% of switching-intent reviews. Cost asymmetry was pronounced: cost concerns reduced recommendation rates by 54.2 percentage points; positive cost comments increased recommendation by 6.2 points. Specialty variance in recommendation intensity exceeded city variance (ratio 2.63). Conclusions: Excessive directiveness and rushedness were the communication behaviors most strongly associated with adverse behavioral outcomes in this corpus; reassurance and empathy are the strongest drivers of recommendation. These effects are modified by specialty context in ways that align with Kalamazoo, Calgary-Cambridge, and SPIKES theoretical predictions. The specialty-stratified findings provide a direct empirical basis for competency-based communication curricula in undergraduate and postgraduate medical education.