This study was approved by the Institutional Review Board and Ethics Committee of Tohoku University School of Medicine (#2015-1-787, approved on March 17, 2016). We applied opt-out consent according to the recruitments of the human research ethics committees at the institutions and local law.
We retrospectively reviewed 8032 consecutive adult patients who underwent surgery in the operating room of Tohoku University Hospital between April 2014 and November 2015. Of these patients, 2669 whose body temperature at the bladder during surgery was not measured (e.g., for surgeries less than 1 h in duration) and 300 with inaccurate body temperature measurements during surgery (i.e., measurement of less than 30.0°C) were excluded from the study. The remaining 5063 patients were enrolled in the study (Fig. 1).
Warming of patients during surgery
After anesthetic induction, we initiated forced-air warming using two types of warming power units: the Bair Hugger Model 775 (3M Company, St. Paul, MN, USA) and Warm Touch 5300A (Medtronic, Minneapolis, MN, USA). We used the power units with one of four types of blankets: under full body (Bair Hugger Models 545, 585, and 635), over full body (Warm Touch Lower Body Blanket), over upper body (Warm Touch Upper Body Blanket), and over lower body (Warm Touch Lower Body Blanket) blankets. We defined each warming method as follows: under full cover (under full body warming by an under full body blanket), over full cover (over full body warming by an over full body blanket), over upper cover (over upper body warming by an over upper body blanket), over lower cover (over lower body warming by an over lower body blanket). In some patients, the over full body blanket was used to cover only the right or left half of the body (over right cover and over left cover, respectively). Resistive heating blankets (SmartCare, Geratherm Medical AG, Geschwenda, Germany) were used during surgeries performed in a bioclean room (over heating cover). Thus, we used the following seven warming methods: under full cover, over full cover, over upper cover, over lower cover, over right cover, over left cover, and over heating cover. Nurses selected these seven warming methods, either alone or in combination, according to the surgical procedure performed and the appropriate patient position.
Raw measurements of vital signs, including body temperature, were transferred onto a server (PRIMERGY TX200 S3, 2 Intel Xenon X5335 processors, 2 GB DIMM, 300 GB HDD) and saved in text format. An electronic anesthesia recording system (PrimeGaia, Nihon Kohden, Tokyo, Japan) was used to retrieve the data from the server at 2.5-min intervals and to display them, together with the patient’s medical information obtained from the hospital information system, as an anesthesia record. Background information on age, sex, height, weight, American Society of Anesthesiology physical status, and type of surgery was obtained. All surgeries were classified into 11 categories defined by the Japanese Society of Anesthesiologists Committee on Operating Room Safety for closed claim studies . The duration of anesthesia from start to end, method of warming (e.g., under full body, over lower body), and body temperature measurement at the bladder were obtained in CSV format from the data warehouse on the server. We assigned the patients warmed by the under full cover method to the Under group and those warmed by the other methods (i.e., all other patients except for those in the Under group) to the Control group (Fig. 1).
We used propensity score matching to reduce potential bias caused by the lack of randomization in this study. The probability (from 0 to 1) of allocation of use of the under full cover was estimated in each patient as a propensity score, based on a multivariate logistic regression model incorporating age, sex, height, weight, American Society of Anesthesiology physical status, type of surgery, duration of anesthesia, and use of one of the other six cover methods (excluding under full cover) as variables. Propensity score matching was performed by random selection of a patient in the Under group and identifying the patient who had the closest propensity score (within 0.03 on a scale of 0 to 1) in the Control group, as described previously . Both before and after propensity score matching, numerical data such as body temperature were compared between the Under and Control groups using Student’s t-test or the Mann–Whitney test. Categorical data were compared between the two groups using the chi square test. The primary endpoint was the incidence of hypothermia at the end of surgery (i.e., a body temperature < 36.0°C, defined by the National Institute for Health and Care Excellence guidelines ), which was compared between the two groups using the chi square test. All statistical calculations were performed using SPSS software (ver. 22, IBM Corp., Chicago, IL, USA). A p value < 0.05 was considered to indicate statistical significance.