Accidental Hypothermia: Factors Related to a Prolonged Hospital Stay - Nationwide Observational Study in Japan Shuhei

Background: Accidental hypothermia (AH) has a low frequency, and the length of hospital stay in patients with AH remains poorly understood. The present study explored which factors were related to long-term hospitalization among patients with AH using Japan’s nationwide registry data. Methods: The data from the Hypothermia STUDY 2018, which included patients ≥ 18 years old with a body temperature ≤ 35°C, were obtained from a multicenter registry for AH conducted at 89 institutions throughout Japan, collected from December 1, 2018, to February 28, 2019. The patients were divided into a “short-stay patients” group (within 7 days) and “long-stay patients” group (more than 7 days). A logistic regression analysis after multiple imputation was performed to obtain odds ratios (ORs) for prolonged hospitalization with age, frailty, cause of hypothermia, mechanism underlying hypothermia, temperature, pH, potassium level, and disseminated intravascular coagulation (DIC) score as independent variables. Results: In total, 656 patients were included in the study, of which 362 were eligible for the analysis. The median length of hospital stay was 17 days. Of the 362 patients, 265 (73.2%) stayed in the hospital for more than 7 days. The factors associated with prolonged hospitalization were frailty (OR, 2.11; 95% condence interval [CI], 1.09-4.10; p = 0.027), the occurrence of indoor (OR, 3.20; 95% CI, 1.58-6.46; p = 0.001), alcohol intoxication (OR, 0.17; 95% CI, 0.05-0.56; p = 0.004), pH (OR, 0.07; 95% CI, 0.01-0.76; p = 0.029), potassium level (OR, 1.36; 95% CI, 1.00-1.85; p = 0.048), and DIC score (OR, 1.54; 95% CI, 1.13-2.10; p = 0.006).


Background
Accidental hypothermia (AH) is de ned by a body core temperature of < 35 °C [1,2]. While it has a low frequency, the mortality rate is high for cases of severe hypothermia [3,4]. Previous studies regarding AH have shown that prognostic factors in patients with AH were the age, pH, and potassium level [3][4][5][6][7].
However, these factors were insu cient for predicting the long-term prognosis of patients with AH or prescribing prophylactic measures to prevent hypothermia. In contrast, studies focusing on the length of hospital stay have proven bene cial for preventing AH and considering the public health regarding AH. For instance, shortening hospitalization can increase the bed turnover rate, thereby reducing medical costs. However, few studies have investigated which factors are associated with the length of hospital stay [8].
The characteristics of patients with AH in Japan include an older age, high rate of occurring indoors, and high mortality rate [9] [10]. Japan is the most rapidly aging society in the world, so we believe that analyzing the AH registry in Japan will meaningfully in uence the situation in other countries that are also experiencing the aging of their societies.

Page 4/21
The present study investigated which factors are associated with long-term hospitalization in patients with AH using Japan's nationwide registry data of hypothermia.

Study design and setting
We performed a prospective, observational, multi-center registry of hypothermia: the Hypothermia STUDY 2018. This study was conducted from December 1, 2018, to February 28, 2019, among a consortium of 89 academic and community medical centers from different geographic regions across Japan. The study was approved by the institutional review board of each hospital listed in the acknowledgements, and the requirement for informed consent was waived due to the observational nature of the study.

Patient selection and data collection
The present study included consecutive patients whose body temperature measured by emergency medical services (EMS) or at emergency department was less than 35 °C. Patients younger than 18 years old were excluded. The following data were collected: age, sex, any pre-existing conditions, activities of daily living, lifestyle, mechanism underlying hypothermia (acute medical illness, trauma [subimmersion, distress], alcohol intoxication, others [including drugs]), cause of hypothermia, Charlson comorbidity index, Glasgow coma scale (GCS), Sequential Organ Failure Assessment (SOFA) score [11], disseminated intravascular coagulation (DIC) score [12], laboratory data, temperature, dysrhythmias, cardiac arrest during pre-hospital, venous-arterial extracorporeal membrane oxygenation (V-A ECMO) rewarming, intubation, hospital periods, survival, and Cerebral Performance Category (CPC) score at 30 days after admission.
Acute medical illnesses were as follows: cerebrovascular disease, cardiac disease, infectious disease, undernutrition, epilepsy, renal disease, endocrine disease, gastrointestinal disease. The clinical frailty scale (CFS) score was determined using the activities of daily living and pre-existing conditions, as described previously [13]. Patients were de ned as frail if they had a CFS score of ≥ 5 before hospital admission. Temperature was recorded as the core temperature from the rectum, urinary bladder, or esophagus if available; otherwise, the peripheral temperature from the axilla or ear was noted.
The laboratory data included the pH, potassium level, platelet count, and CPK level at the emergency department. The pH value in principle was evaluated by an arterial blood gas analysis, and the pH measured using the venous blood gas was adjusted as described in a previous study [14]. Complications during hospitalization were recorded and classi ed as pneumonia, pancreatitis, or other. Pneumonia was de ned as an obvious shadow on chest radiography or computed tomography (CT). Pancreatitis was de ned as cases meeting at least two of the following conditions: 1) abdominal pain, 2) elevation of ultrasound/CT. In the present study, patients were excluded from the analysis if they were 30-day nonsurvivors, lacked a hospital stay, or had an unknown length of hospital stay or an unknown body temperature, as these patients had potential biases.
Previous studies showed that the median length of hospital stay in patients with AH was 4-9 days [10] [4] [8]. We therefore divided the patients with AH into 2 groups: the "short-stay patients" group (≤ 7 days) and "long-stay patients" group (> 7 days).

Data analyses
Data are expressed as the number (%) or median (interquartile range), as appropriate. Intergroup comparisons were made using Fisher's exact test for categorical data and the Mann-Whitney U test for continuous data. For a further evaluation, we conducted multivariate logistic regression analyses to control for the potentially confounding roles of age, frailty, cause of hypothermia, mechanism underlying hypothermia, temperature, pH, potassium, and DIC score. These potential confounders were selected based on a previous study regarding AH [8] or the consideration of clinically signi cant variables. Furthermore, a multiple linear regression analysis including similar variables was performed. Missing data were managed with multiple imputation by chained equations [15,16]. The variables included in the imputation model were those from the logistic regression analysis. Twenty-ve dataset were imputed with 10 iterations each. Multivariate logistic regression and multiple liner regression analyses were applied to the 25 imputed datasets, and nal estimates were obtained by averaging the 25 estimates according to Rubin's rules. Furthermore, a complete data set was used for the sensitive analysis. All tests were two-sided with P values of less than 0.05 considered statistically signi cant.
Statistical analyses were performed with the EZR software program (Saitama Medical Center, Jichi Medical University, Saitama, Japan), a graphical user interface for R software. Multiple imputation was performed using mice package in the R software program, version 4.0.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results
Of the 656 patients with hypothermia included in the Hypothermia STUDY 2018, 294 were excluded from the present study because of death within 30 days (N = 160), non-hospital stay (N = 110), unknown length of hospital stay (N = 21), or others, including an unknown temperature (N = 3). The remaining 362 patients were eligible for the present analysis. The patient ow diagram is shown in Fig. 1. Depending on the length of hospital stay, the 362 patients were divided into the "short-stay patients" group (N = 97) and "long-stay patients" group (N = 265). The overall mortality rate within 30 days was 160 among 656 patients (24.4%). Among the remaining 362 patients, 265 (73.2%) stayed in the hospital for more than 7 days.
Baseline characteristics of the study population Table 1 shows the baseline characteristics of the study population and a comparison of the clinical characteristics between the "short-stay patients" and "long-stay patients" groups. The median patient age was 78 years old, and 75.1% of AH cases occurred indoors. The age, SOFA score, and DIC score were higher in the "long-stay patients" group than in the "short-stay patients" group. There was a signi cant difference in the CFS score, cause of hypothermia, and mechanism of hypothermia between the "shortstay patients" and "long-stay patients" groups. In the "long-stay patients" group, the rate of occurrence indoors was higher than that in the "short-stay patients" group. Regarding the cause of hypothermia, acute illness was the predominant cause in the "long-stay patients" group, whereas the proportion of cases caused by alcohol intoxication was higher in "short-stay patients" group than in the "long-stay patients" group.

Clinical and laboratory data
The clinical and laboratory data are presented in Table 2. There were no signi cant differences in the severity grade of temperature, GCS, blood pressure, heart rate, respiratory rate, platelet level, or rate of intubation between the "short-stay patients" and "long-stay patients" groups. However, the pH values were signi cantly lower and the potassium and CPK levels higher in the "long-stay patient" group than in the "short-stay patient" group. In addition, the "long-stay patient" group showed a signi cantly higher incidence of arrhythmia than the "short-stay patient" group. A total of 222 (61.3%) patients were admitted to the intensive-care unit (ICU), although there was no signi cant difference in the rate of ICU admission between the groups.
Complications, neurological score, and hospital length of stay There was no signi cant difference in the incidence of complications between the "short-stay patients" and "long-stay patients" groups (Table 3). However, at the neurological assessment, the "long-stay patients" group showed a higher rate of worsened neurological score (CPC 3-5) at 30 days after admission than the "short-stay patients" group, while the "short-stay patients" group showed a signi cantly higher rate of patients with a favorable neurological outcome (CPC 1-2) than the "long-stay patients" group. The median length of ICU stay was 4 days, and the median length of hospital stay was 17 days for the 362 total patients. Additional le 1 shows the distribution of the hospital stay duration. Table 4 shows that the factors associated with the length of hospital stay, based on a multivariate logistic regression analysis after multiple imputation, were frailty (odds ratio [OR], 2.11; 95% con dence interval [CI], 1.09-4.10; p = 0.027), indoor occurrence (OR, 3.20; 95% CI, 1.58-6.46; p = 0.001), mechanism underlying hypothermia (alcohol intoxication) (OR, 0.17; 95% CI, 0.05-0.56; p = 0.004), pH (OR, 0.07; 95% CI, 0.01-0.76; p = 0.029), potassium levels (OR, 1.36; 95% CI, 1.00-1.85; p = 0.048), and the DIC score (OR, 1.54; 95% CI, 1.13-2.10; p = 0.006). In the sensitivity analysis, pH (OR, 0.10; 95% CI, 0.01-1.11, p = 0.061) was not signi cantly associated with the length of hospital stay, but the result was close to signi cance, and the sensitivity analysis showed similar results to the analysis after multiple imputation. Thus, the results of the analysis after multiple imputation were considered robust. Table 5 shows the correlation between variables and the length of hospital stay, based on a multiple liner regression analysis after multiple imputation. Indoor occurrence (β = 7.19; 95% CI, 1.14-13.24, P = 0.020), mechanism underlying hypothermia (alcohol intoxication) (β = -10.82; 95% CI, -20.92-0.71, P = 0.036), and the DIC score (β = 2.84; 95% CI, 0.41-5.28, P = 0.022) were signi cantly correlated with the length of hospital stay. In the sensitivity analysis, alcohol intoxication (β = -8.95; 95% CI, -20.82-2.91, p = 0.139) was not signi cantly associated with the length of hospital stay, but the result was close to signi cance, so the interpretation of results did not make much difference. Thus, the results of the analysis after multiple imputation were considered robust.

Discussion
The present nationwide study showed that a total of 75.1% of patients with AH developed AH indoors. A total of 73.2% of patients with AH stayed in hospitals for more than 7 days. The median length of hospital stay was 17 days. The rate of complications was low, there was no association between complications and the length of hospital stay. The factors related to long-term hospitalization were frailty, indoor occurrence, alcohol intoxication, pH, potassium levels, and the DIC score.
Among these factors related to a prolonged hospital stay, the pH, potassium levels, and DIC score were not preventable, while frailty and an indoor occurrence were potentially preventable with intervention.
Previous studies have shown many patients with AH in urban settings develop AH indoors [9] [10], which was in agreement with the present ndings. In addition, the patients who develop AH indoors have a higher death rate and longer hospital stay because of their tendency to have an advanced age [17] [10]. However, of note, the present study showed that the factors associated with the length of hospital stay were not just the age but the frailty. Recently, frailty, which is highly prevalent in older individuals and confers a high risk of falls, disability, hospitalization, and mortality [18], has also been noted in critically ill patients [19] [20]. To our knowledge, limited data exist regarding the relationship between AH and frailty. The present ndings are of critical importance. Clinically, the risk factors of frailty and an indoor occurrence may be useful for the early detection of AH patients likely to require a prolonged hospital stay. In the future, preventing frailty and maintaining adequate activities of daily living may help suppress the occurrence of AH and reduce the length of hospital stay. This approach is certain to have an incredible impact on the aging populations of Japan as well as other industrialized countries.
In the present study, alcohol intoxication, pH, potassium levels, and the DIC score were shown to be associated with length of hospital stay. In general, the DIC score is increased by sepsis and organ failure. Accordingly, AH patients complicated by such issues may have a signi cantly prolonged hospital stay. Alcohol intoxication was conversely associated with a short hospital stay. We speculate that these patients had acute alcohol intoxication, which resulted in an early discharge from the hospital after rewarming and awakening. However, a previous study showed that a lower core temperature, lower degree of consciousness, and lower platelet count were associated with a prolonged hospital stay, although these factors were not found to be associated in the present study. Several reasons may explain this discrepancy. First, the previous study differed from the present study with regard to the patient background. Previous studies regarding AH patients from Europe and North Europe showed that these patients were mostly healthy, younger individuals, and AH occurred outdoors, such as in cases of avalanche or submersion in freezing water [3,21,22], resulting in an early discharge from the hospital. In contrast, the patients in the present study were older individuals with underlying diseases, and AH tended to occur indoors. These differences may affect the factors associated with the hospital stay. Second, the previous study was performed at a single institution and extracted from a univariate analysis, so potential confounders were not considered for adjustment. The present study was a multicenter, nationwide study associated with AH in Japan and involved a multivariate analysis of credible data.
The present study had several limitations. First, the length of hospitalization is in uenced by the social environment and medical care system. Therefore, it may be not appropriate to compare the results of this study with those from other countries. Second, this registry did not include data on the CFS score, and instead, the CFS score was evaluated using the activities of daily living and pre-existing conditions. As a result, the frailty not have been accurately classi ed. Finally, although we adjusted for variables, the possibility of residual confounding factors cannot be avoided.

Conclusions
This study found, that among patients with AH, three-quarters needed to be hospitalized for more than 7 days. The factors related to a prolonged hospital stay were frailty, indoor occurrence, pH, potassium levels, and the DIC score. Conversely, alcohol intoxication was associated with a short hospital stay. Taken together, these ndings suggest that reducing frailty and maintaining adequate activities of daily living may help reduce the length of hospital stay in patients with AH. This is certain to in uence the outcomes of aging populations in Japan as well as other industrialized countries. Availability of data and materials: No data is available Ethics approval and consent to participate: The institutional review board of each hospital listed in the acknowledgements approved the study, and the requirement for informed consent was waived due to the observational nature of the study.

Consent for publication:
Not applicable