Patient acuity rating: quantifying clinical judgment regarding inpatient stability

Abstract

Background: New resident work-hour restrictions are expected to result in further increases in the number of handoffs between inpatient care providers, a known risk factor for poor outcomes. Strategies for improving the accuracy and efficiency of provider sign-outs are needed.

Objective: To develop and test a judgment-based scale for conveying the risk of clinical deterioration.

Design: Prospective observational study.

Setting: University teaching hospital.

Subjects: Internal medicine clinicians and patients.

Measurements: The Patient Acuity Rating (PAR), a 7-point Likert score representing the likelihood of a patient experiencing a cardiac arrest or intensive care unit (ICU) transfer within the next 24 hours, was obtained from physicians and midlevel practitioners at the time of sign-out. Cross-covering physicians were blinded to the results, which were subsequently correlated with outcomes.

Results: Forty eligible clinicians consented to participate, providing 6034 individual scores on 3419 patient-days. Seventy-four patient-days resulted in cardiac arrest or ICU transfer within 24 hours. The average PAR was 3 ± 1 and yielded an area under the receiver operator characteristics curve (AUROC) of 0.82. Provider-specific AUROC values ranged from 0.69 for residents to 0.85 for attendings (P = 0.01). Interns and midlevels did not differ significantly from the other groups. A PAR of 4 or higher corresponded to a sensitivity of 82% and a specificity of 68% for predicting cardiac arrest or ICU transfer in the next 24 hours.

Conclusions: Clinical judgment regarding patient stability can be reliably quantified in a simple score with the potential for efficiently conveying complex assessments of at-risk patients during handoffs between healthcare members.

Figure 1

Patient Acuity Rating scale.

Figure 2

Percent of patients showing clinical deterioration to the point of cardiac arrest or ICU transfer by Patient Acuity Rating and provider. Ordinal trend p-value was <0.001 for each provider.