4060 — Identifying Postoperative Infections with Trends in Inpatient Temperature

Lead/Presenter: Laura Graham,  COIN - Palo Alto
All Authors: Graham LA (Center for Innovation to Implementation, VA Palo Alto Health Care System, Palo Alto, CA; Stanford-Surgical Policy Improvement Research Education, Stan), Wagner TH (Center for Innovation to Implementation, VA Palo Alto Health Care System, Palo Alto, CA; Stanford-Surgical Policy Improvement Research Education, Stanford University, Stanford, CA), Hawn MT (Department of Surgery, Stanford University Medical School, Stanford, CA)

Objectives:
Postoperative infections are prevalent, costly, and a major concern for healthcare systems. Developing infections are often identified by subtle changes in a patient's vital signs but no infection prediction tools include vital signs. The aim of this study is to examine the association between postoperative inpatient temperature and developing infections.

Methods:
We use a prospective cohort study design to identify Veterans undergoing inpatient surgery at a VA medical center between October 1, 2007 and September 30, 2014. All inpatient temperature assessments were obtained from the VA Corporate Data Warehouse Vital Signs domain. These recordings are collected on all patients at routine times, avoiding bias that could be due to clinician judgment. Postoperative infections were defined as wound infections, pneumonia, urinary tract infections, or sepsis identified by a trained VA Surgical Quality Improvement Program nurse. Smoothed plots were used to visualize postoperative trajectories of inpatient temperature among patients developing an infection within 24 hours as compared to patients with no postoperative infection. Multivariate logistic models were used to estimate the association of temperature with developing an infection within 24 hours of the temperature assessment.

Results:
We identified 175,882 inpatient surgical procedures (48.4% general, 22.1% vascular, 19.6% urologic, 9.8% thoracic) with 3,496,767 postoperative inpatient temperature readings. Patients spent an average of 6.1 days in the hospital after surgery (Standard Deviation = 4.6). The overall rate of postoperative infection was 10.6% (N = 18,565). The four most common infections were: wound infection (44.3% of all infections), pneumonia (30.3%), urinary tract infection (23.0%), and sepsis (21.2%). On average, patients developing an infection within 24 hours had a temperature 0.5 degrees Fahrenheit higher than patients who did not develop an infection (p < 0.001). After adjusting for observed confounders, and diurnal variation, a 1-degree Fahrenheit increase in temperature was associated with a 46% increased risk of an infection diagnosis within 24 hours of the temperature reading (95% Confidence Interval = 1.42-1.49).

Implications:
Routinely collected biometric data may offer new insights into developing postoperative infections.

Impacts:
Incorporating real-time vital sign assessments, namely body temperature, into an infection risk prediction tool could result in earlier identification of infections and patients at risk for developing an infection at hospital discharge.