Provider decisions to treat respiratory illnesses with antibiotics: Insights from a randomized controlled trial

Abstract

Rationale: Lower respiratory tract illness (LRTI) frequently causes adult hospitalization and antibiotic overuse. Procalcitonin (PCT) treatment algorithms have been used successfully in Europe to safely reduce antibiotic use for LRTI but have not been adopted in the United States. We recently performed a feasibility study for a randomized clinical trial (RCT) of PCT and viral testing to guide therapy for non-pneumonic LRTI. Objective: The primary objective of the current study was to understand factors influencing PCT algorithm adherence during the RCT and evaluate factors influencing provider antibiotic prescribing practices for LRTI. Study Design: From October 2013-April 2014, 300 patients hospitalized at a community teaching hospital with non-pneumonic LRTI were randomized to standard or PCT-guided care with viral PCR testing. Algorithm adherence data was collected and multivariate stepwise logistic regression of clinical variables used to model algorithm compliance. 134 providers were surveyed anonymously before and after the trial to assess knowledge of biomarkers and viral testing and antibiotic prescribing practices. Results: Diagnosis of pneumonia on admission was the only variable significantly associated with non-adherence [7%(adherence) vs. 26% (nonadherence), p = 0.01]. Surveys confirmed possible infiltrate on chest radiograph as important for provider decisions, as were severity of illness, positive sputum culture, abnormal CBC and fever. However, age, patient expectations and medical-legal concerns were also at least somewhat important to prescribing practices. Physician agreement with the importance of viral and PCT testing increased from 42% to 64% (p = 0.007) and 49% to 74% (p = 0.001), respectively, after the study. Conclusions: Optimal algorithm adherence will be important for definitive PCT intervention trials in the US to determine if PCT guided algorithms result in better outcomes than reliance on traditional clinical variables. Factors influencing treatment decisions such as patient age, presence of fever, patient expectations and medical legal concerns may be amenable to education to improve PCT algorithm compliance for LRTI.