Performance of shock index, lactate and physician’s gestalt in predicting adverse events among critically ill adult patients at an Emergency Department in Tanzania

Abstract

Introduction Physicians working in Emergency Departments (ED) must be able to identify critically ill patients likely to deteriorate and prioritize their care. While shock index, point of lactate and physician gestalt are commonly used in high-income countries to predict adverse events their applicability has not been widely tested in low-middle income countries. Methods A prospective cohort study of all adult patients (≥18 years) triaged to the resuscitation area of Muhimbili ED in Tanzania from 1 st September - 31 st December 2022. Shock index (SI) on arrival, initial point of care lactate and the physician’s estimates of the probability of an adverse event were recorded. The discriminatory ability of these index tests was compared for prediction of an adverse event in the ED, using DeLong’s test. Sensitivity, specificity, positive and negative predictive values were calculated. For SI, a standard cut off ≥ 1 was used and for lactate a standard cut off > 2 mmol/L was used. For physician gestalt, the optimum cut off of a 70 % probability of adverse event was determined based on Youdens’ index. Results A total of 320 critically ill patients were enrolled. SI (AUC = 0.734, 95 % CI 0.658 to 0.810) achieved a sensitivity of 59.3 % and a specificity of 83.1 %. Elevated lactate (AUC = 0.819, 95 % CI 0.76 to 0.88) achieved a sensitivity of 89.8 % and a specificity of 58.2 %. Physician gestalt had an AUC = 0.877, (95 % CI 0.833 to 0.922) using the optimal probability of ≥70 % and achieved a sensitivity of 88.1 % and a specificity of 76.2 %. The AUC of physician’s gestalt was significantly higher than that of SI ( p = 0.0008), but was not significantly different from lactate ( p = 0.067). The difference in AUC for lactate and SI was also not statistically significant ( p = 0.10). Discussion Physicians gestalt in our setting demonstrated superior predictive performance compared to shock index and a comparable performance to lactate. This supports the use of gestalt as a low-resource risk stratification tool in emergency settings. However, accuracy of gestalt may depend on physician experience.