Incidence and Risk Factors for Postoperative Hypothermia After Orthopaedic Surgery

John P. Kleimeyer, MD; Alex H. S. Harris, PhD; Joseph Sanford, MD; William J. Maloney, MD; Bassam Kadry, MD; Julius A. Bishop, MD


J Am Acad Orthop Surg. 2018;26(24):e497-e503. 

In This Article


Study Design

The institutional review board approved the protocol for this study. We completed a retrospective review of prospectively collected data from 6,950 patients undergoing orthopaedic surgery at a tertiary care academic medical center over a 30-month period (November 2012 to May 2014). The procedure type was identified by the primary Current Procedural Terminology (CPT) code and associated with a subspecialty and anatomic location. Demographic data were extracted from the medical record. Temperature measurements were obtained in routine perioperative patient care. The measurement site was chosen by the anesthesiologist or by the recovery area protocol with esophageal (33.5%), nasal (23.5%), skin (10.2%), and bladder (6.8%) being used most frequently. Rectal or catheter temperatures were obtained less often (combined <1.0%). All surgical cases in the 30-month period were included.


The primary outcome was postoperative hypothermia, defined as postoperative temperature <36.0°C on arrival to the postanesthesia care unit or next level of care (eg, intensive care unit). This value is commonly defined in the literature,[2] has been associated with increased morbidity,[6,7,11,12] and is consistent with quality performance measures implemented by the Centers for Medicare and Medicaid Services and National Quality Forum.[1] A mixed-effects multivariable logistic regression including random effect for surgeon was performed to identify notable correlates for hypothermia. To operationalize low temperature in a continuous rather than dichotomous approach, a separate mixed-effect multivariable linear regression was used to examine predictors of postoperative temperature. The glmmPQL function in the MASS package of R version 3.0 was used. In both models, candidate predictors were procedure type and duration, patient age, sex, body mass index (BMI), preoperative (holding area) temperature, intraoperative hypothermia (ie, minimum single measurement intraoperative temperature <36.0°C), anesthesia type, and American Society of Anesthesiology (ASA) status. BSA was considered a candidate predictor using the Mosteller method; however, this is highly correlated with BMI (r = 0.70) and was therefore not included in the analysis. Subspecialties were defined as adult reconstruction, arthroscopy, foot and ankle, hand, spine, sports (excluding arthroscopy), trauma, and tumor. Fracture treatment proximal to the wrist or ankle was defined as trauma procedures. Anatomic locations included ankle, arm, calf, elbow, foot, forearm, hand, hip, knee, pelvis, shoulder, spine (ie, grouped by cervical, thoracic, or lumbar), thigh, wrist, and indeterminate. Locations were grouped by region for analysis as follows: upper extremity (ie, arm, elbow, forearm, hand, shoulder, and wrist), core (ie, spine and all associated thoracic or cervical procedures), hip and pelvis (hip and pelvis), proximal leg (thigh and knee), and distal leg (calf, ankle, and foot).

Outliers, including patients with BMI > 100 (n = 1), preoperative hypothermia (temperature <36.0°C; n = 4), initial postoperative temperature <30.0°C (n = 12), or postoperative discharge temperature <30.0°C (n = 1) were excluded. Procedures with missing covariate data (n = 3,110), were also excluded, including unrecorded preoperative temperature (n = 1,689), undefined subspecialty or location based on CPT code (n = 1,109), BMI (n = 302), ASA status (n = 185), or case duration (n = 27). Procedures were considered undefined if their primary CPT code was not associated with a specific subspecialty or the anatomic location was indeterminate. After application of exclusion criteria, the final sample size for analysis was 3,822 procedures. Rates of hypothermia (P = 0.98), preoperative temperature (P = 0.93), BSA (P = 0.98), and sex (P = 0.15) were not different in the excluded group. Postoperative temperature, age, and BMI were higher in the excluded group (P's < 0.01).

Miscellaneous specialty cases and indeterminate anatomic location cases that could not be categorized were excluded from subgroup analysis. Hand as a subspecialty (n = 45) had no observed cases of postoperative hypothermia and was therefore excluded from the regression model. The hand and pelvis anatomic locations were similarly excluded from the regression model, although all regions were included. Because subspecialty and anatomic location were highly overlapping, separate models were estimated for each with 3,777 and 3,414 procedures included, respectively. The subspecialty and anatomic region with the lowest hypothermia rates were chosen as the reference groups (ie, trauma and upper extremity, respectively).