Methods: We evaluated all patients who had elective knee arthroplasty and were discharged within 5 days at NorthShore University Health System in Illinois between January 1, 2007 and December 31, 2013. Inpatient temperature, gender, age and BMI measurements were extracted from the data warehouse. Group based trajectory modeling was performed to identify clusters of postoperative patients following similar progression of average temperature mapping over time with gender, age and BMI as covariates.
Results: Five thousand six hundred patients’ postoperative temperature curves were evaluated. Three hundred sixty six (6.5%), 372 (6.7%), 225 (4.4%) and 30 (4.3%) patients had a Tmax greater than 100.4 on postoperative days 0, 1, 2 and 3 respectively (patients censored on discharge). Average temperature trajectories were created over 24 time periods (4 days), with each time period representing a 4-hour interval when temperatures were likely recorded (Figure 1). Three distinct trajectory clusters were identified with 41.1% displaying group membership to trajectory 1 (red), 49.3% to trajectory 2 (green) and 9.6% to trajectory 3 (blue). All 3 trajectories displayed cubic functional forms. Patients in trajectory 2 and 3 were younger than those in trajectory 1 (Likelihood estimates (LE) -0.027 and -0.020 respectively; p <0.001 for both) and had higher BMI (LE 0.036 and 0.061 respectively, p<0.001 for both).
Conclusion: Patients display different temperature trajectories following elective knee arthroplasty. Further work is needed to determine whether different trajectories indicate different likelihoods of an adverse event occurring. Group based trajectory modeling may allow for personalizing the interpretation of a temperature in the postoperative setting.
J. P. Ridgway, None
R. Padman, None
A. Robicsek, None