Abstract P92: Provider Intuition Predicts Clinical Deterioration in Ward Patients
Introduction: In-hospital cardiac arrest is rarely sudden and may be preventable if patients are intervened on in the early stages of clinical deterioration. However, identifying high-risk patients remains a significant challenge. Scoring tools based on vital signs and laboratory findings have yielded limited accuracy and require labor or technology to capture prospectively. We hypothesized that an intuition-based scoring tool would predict clinical deterioration on the wards.
Methods: Intern, resident and attending physicians and midlevel providers (nurse practitioners or physician assistants) on nine medical ward services at a tertiary care hospital were prospectively surveyed every call cycle and asked to rate each of their patients on a scale from 1 to 7 based on the patient’s likelihood of suffering a cardiac arrest or requiring emergent transfer to the ICU within the next 24 hours. Inter-rater reliability was assessed between providers, for the same patient day, using weighted Kappa statistics. A Receiver Operator Characteristics (ROC) curve was constructed for each type of provider and the area under each curve calculated to measure the accuracy of the scoring system. These areas were compared to one another using a Chi-squared statistic.
Results: Between 1/14/08 and 5/31/08, 129 (88%) eligible providers consented to participate and provided 5320 individual scores on 1490 patients, with a data capture rate of 75%. Kappa coefficients ranged from 0.32 to 0.43 between the different types of providers, with the greatest agreement occurring between the attendings and midlevels. The risk of clinical deterioration increased exponentially with increasing scores, from 0% (for a score of 1) to 21–34% (depending upon the type of provider, for a score of 7). Area under the ROC curves ranged from 0.70 for residents to 0.85 for attendings (p=0.01). The other groups were not statistically different from one another.
Conclusions: A 7-point scale of provider intuition yielded fair to moderate inter-rater reliability and fair to good accuracy in predicting cardiac arrest and emergent ICU transfer. Attending physician scores correlated best with patient outcomes. This simple tool may be useful for identifying ward patients at risk of clinical deterioration.