Performance of physician’s clinical gestalt, shock index and lactate in predicting adverse events among critically ill adult patients presenting at Emergency Medicine Department.
Abstract
Title: Performance of physician’s clinical gestalt, shock index and lactate in predicting adverse events among critically ill adult patients presenting at Emergency Medicine Department.
Background: The burden of critically ill patients in low- and middle-income (LMICs) is higher. Given the number of critically ill patients seen at Emergency department, physicians need to be able to determine who is most likely to deteriorate in order to prioritize their care and dispose to intensive care unit (ICU).
Methodology: A prospective cohort study of all critically ill adult patients presenting to the Muhimbili Emergency Medicine Department from September-, 2022 to December-, 2022. Physicians were asked for their clinical gestalt probability of how likely the patient was to experience an adverse event. Shock index was determined by the investigators using initial vital signs; lactate if drawn was recorded for all patients. Patients were observed at ED until disposition for outcome. For physician gestalt probability, a cut off was determined using a ROC curve, and based on Youdens’ index, a probability of ≥70% maximized sensitivity and specificity, and this estimate was used as a cut off to define a physician’s prediction of high probability of an adverse event to occur. For shock index a standard cut off of ≥1 was used and for lactate a standard cut off of >2 mmol/L was used. Sensitivity, specificity, positive and negative predictive values and the area under the curve (AUC) were calculated to evaluate each test’s ability to predict adverse events.
Results: A total of 320 critically ill patients who met the inclusion criteria were enrolled. Physician clinical gestalt had an AUC=0.877, (95% CI 0.833 to 0.922); - An optimal cut off probability of ≥70% predicted the occurrence of an adverse events in 88.1% of patients, and correctly identified those who did not have adverse event 76.2% of the time. Shock index had the ability to detect the occurrence of adverse events with a sensitivity of 59.3% and identify those who did not have adverse events in 83.1% of cases (AUC= 0.734, 95% CI 0.658 to 0.810). Elevated lactate predicted the occurrence of adverse events with a sensitivity of 89.8% and identified those who did not have an adverse event with a specificity of 58.2% (AUC=0.819, 95% CI 0.76 to 0.88). A combination of both physician gestalt probability and shock index predicted the occurrence of adverse events with a sensitivity of 52.5% and identified those who did not adverse events with a specificity of 92%. Of the three index tests; the cut off used will affect their sensitivity and specificity in predicting adverse events, depending on the priorities of the setting.
Conclusion: Clinician gestalt has excellent discriminatory ability, indicating that physicians can rely on their clinical judgement to determine which critically ill patients are likely to have adverse events as well as those who will need continuum of care in intensive care unit/high dependent unit (ICU/HDU). And in our setting, it was superior to other frequently recommended guides.
