Study design and setting
This multi-center retrospective study evaluated the J-point registry database [2, 8], which has been previously described [2, 8–10]. The registry includes data regarding patients with an unknown BT or those with a BT of ≤ 35.0˚C who were treated for AH in emergency departments between April 1, 2011 and March 31, 2016. The registry contains data from eight critical care centers and four non-critical care centers in the Osaka, Kyoto, and Shiga prefectures of Japan. In Japan, a critical care center generally serves a population of 500,000 people and is certified by the Ministry of Health, Labor and Welfare of Japan to provide 24h/day care to patients with severe trauma, severe shock, stroke, and acute coronary syndrome. The median number of patients who visit the emergency department at each facility is 19,651 (IQR: 13,281–27,554). Patients were excluded from the J-point registry if they or their family members explicitly refused to be included in the registry. For the present study, clinical data were retrospectively extracted by emergency physicians using a predefined data extraction sheet. The J-point registry protocol and retrospective analysis of anonymized data were approved by the ethics committee of Saiseikai Shiga Hospital (approval ID: 244). Each hospital also approved the registry protocol as necessary.
Study patients
This study included the patients from the J-point registry who were > 18 years old and diagnosed with hypothermia. Based on a previous report [1], hypothermia was defined as having a core BT of ≤ 35˚C. Patients were excluded if they had cardiac arrest at the hospital arrival or had no record of consciousness level.
Data collection
We obtained clinical data from the J-point registry database regarding age, sex, activities of daily living (ADL), medical history, vital signs at hospital arrival, blood test findings, comorbidities, and in-hospital mortality. Age was categorized as 18–64 years, 65–74 years, and ≥ 75 years, based on age definitions reported from government statistical methods [11]. We also categorized the vital signs based on the Japan Fire and Disaster Management Agency protocol for triage [12]. Patients whose systolic blood pressure was difficult to measure were assigned a value of ≤ 60 mmHg and patients whose heart rate was difficult to measure were assigned a value of ≤ 30 beats/min. Data regarding respiratory status were not included in the analysis, as it is difficult to measure percutaneous arterial oxygen saturation in patients with hypothermia [13] and many patients were missing data regarding their respiratory rate.
Exposure
The primary exposure of interest was the level of consciousness at hospital arrival. The level of impaired consciousness was evaluated using the Glasgow Coma Scale (GCS) and the results were classified three groups as mild (GCS of 13–15), moderate (GCS of 9–12), or severe (GCS of 3–8). In Japan, the level of impaired consciousness is also evaluated using the Japan Coma Scale (JCS), which consists of eye-opening to stimuli, similar to the E component of the GCS. The JCS is widely accepted by paramedics and nurses working in the emergency department, and is reportedly correlated with the GCS [14]. Therefore, when GCS data were missed, we imputed values for mildly impaired consciousness (JCS level 1, 1–3), moderately impaired consciousness (JCS level 2, 10–30), and severely impaired consciousness (JCS level 3, 100–300), based on previous reports [8, 9].
Outcomes
The primary outcome was defined as severe hypothermia at hospital arrival. According to the Swiss staging system, hypothermia was classified as stage 1 (BT: 32–35.0˚C), stage 2 (BT: 28–31.9˚C), stage 3 (BT: 24–27.9˚C), and stage 4 (BT: <24˚C) [1]. For the present study, severe hypothermia was defined as Stage 3–4 cases [1]. The secondary outcome was defined as all-cause in-hospital mortality.
Statistical analysis
Patient characteristics at hospital arrival were analyzed according to the level of consciousness. Continuous variables were reported as medians and IQRs, and categorical variables were reported as numbers and percentages. As a primary analysis, the association between the level of impaired consciousness and severe hypothermia was evaluated by multivariable logistic model adjusting potential confounders as following covariates that are generally available in the out-of-hospital setting: age, sex, ADL, consciousness, systolic blood pressure, and heart rate. Logistic regression analysis was used to evaluate the risk of severe hypothermia (BT: <28.0˚C) according to level of impaired consciousness, and the results were reported as COR or AOR with the corresponding 95% CI. Further, we investigated the ability of impaired level of consciousness to predict severe hypothermia was evaluated based on the values for sensitivity, specificity, positive LR and negative LR. As a secondary analysis, the association between the level of impaired consciousness and in-hospital mortality was investigated by logistic regression analysis using the same covariates as potential confounders for the primary analysis. All analyses were performed using JMP Pro 14 for Windows software (SAS Institute, Tokyo, Japan).