Patient-specific versus non-patient-specific alerts in decision support system to prevent contrast-induced nephropathy: A randomized controlled trial



'''Purpose:'''  Physicians ordering contrast imaging should identify patient risks and prevent contrast-induced nephropathy (CIN). Our aim is to determine whether patient-specific or non-patient-specific alerts in clinical decision support system triggers higher compliance with guidelines to prevent CIN.      

'''Materials and Methods:'''  A 3-arm cluster randomized controlled trial was conducted in two university hospitals using the same computerized physician order entry. Eligible physicians were randomized to receive patient-specific alerts, non-patient-specific alerts or no intervention (groups 1–3 respectively). Patient-specific alerts automatically prompted only when CIN risk was encountered and provides patient-specific data to physicians, while non-patient-specific alerts always prompted regardless of risk without providing patient-specific data. CIN risk was stratified as high-risk, low-risk and minimal-risk according to patients renal function. Contrast imaging order-cancellation rate was measured as primary outcome.      

'''Results:'''  Orders for 5372 patients from 99 physicians were analyzed. Renal function and risk distributions of patients were not statistically different among groups. Order-cancellation rates were 32.1%, 14.3%, 1.7% for high risk patients, and 7.2%, 3.0%, 1.3% for low risk patients in groups 1–3, respectively. Using generalized linear model, significant order-cancellation factors in at-risk patients were non-patient-specific alert (p = 0.04), patient-specific alert (p < 0.0001), high CIN risk (p = 0.003), and physicians with low contrast imaging ordering volume (p < 0.0001). Order-cancellation effects were compared and measured by odds ratio (OR). For high risk patients, patient-specific versus non-patient-specific alerts OR was significant at 2.73 (95%CI 1.09–6.84), while patient-specific vs. control OR = 28.32 (95%CI 3.21–249.65), and non-patient-specific vs. control OR = 9.17 (95%CI 0.96–87.55).      

'''Conclusion:'''  Patient-specific alert significantly outperformed non-patient-specific alert in physician compliance with guidelines.