Predictors of Early Clinical Deterioration from the Emergency Department and Clinical Gestalt: A Prospective Case-Control Study

Study Objectives: Rapid response teams (RRT) aim to reduce morbidity and mortality of hospitalized patients through early intervention on those who are clinically deteriorating. Identifying predictors of early deterioration of patients may improve quality and safety. Abnormal vital signs (VS) have previously been studied as predictors of early deterioration and increased RRT activation after admission from the ED. The primary objective is to identify these predictors of RRT activation within 24 hours of admission. Secondary objectives are to detect differences in hospital length-of-stay (LOS), admitting diagnoses, and 30-day mortality in those with and without RRT (+RRT and –RRT). Last, we examine the predictive value of physician clinical gestalt on RRTs. Methods: This was a prospective, observational case-control study by chart review of adult patients admitted at Kaweah Health Medical Center between December 2020 and March 2021. Exclusion criteria were age <18, admission to ICU, and direct transfer out of the ED. At time of admission, we performed chart review to collect eight VS used to activate a RRT at our hospital. To determine clinician gestalt, the physician was also asked if RRT would occur within 24 hours. Each patient’s hospital course was later reviewed for the occurrence of RRT, admitting diagnoses, LOS, and 30-day mortality. Categorical variables were analyzed using Fisher’s exact test. Noncategorical variables were analyzed using Wilcoxon rank sum test, independent T-test, and Pearson correlation coefficient. Predictive value was analyzed with negative and positive predictive values (NPV and PPV), sensitivity, and specificity. Results: 199 patients met inclusion/exclusion criteria for analysis (+RRT N=3, -RRT N=196). No significant differences were detected in sex, age, or level of care between our groups. The groups differed in median heart rate and respiratory rate, but the difference was insignificant. There was a marginally significant association between COVID-19 as admitting diagnosis and RRT (p=0.052). There was no significant association between RRT and 30-day mortality. Mean LOS did not differ between the groups (p=0.297). The mean number of abnormal VS in those deceased at 30 days (1.2) was significantly higher than those alive at 30 days (0.7) (p=0.047, correlation coefficient r=0.14). Analysis of clinical gestalt on RRT showed PPV 3.2%, NPV 98.8%, sensitivity 33.3%, and specificity 84.5%. Conclusion: Due to small sample size, our results did not show significant differences in sex, age, level of care, heart rate, respiratory rate, or LOS between the +RRT and –RRT groups. However, our study was significant for three findings. First, there was a marginally significant association between an admitting diagnosis of COVID-19 and RRT. Second, patients deceased within 30 days had a significantly higher number of abnormal VS than patients who were alive at 30 days, suggesting a positive correlation. Third, results suggest that the clinical gestalt of emergency physicians at predicting who will not have an RRT is reasonably good, but may not be as good at predicting who will have an RRT. Further studies determining other factors contributing to early deterioration can help craft interventions to improve patient safety. [Formula presented]