Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2021 Jan 9;9(1):6.
doi: 10.1186/s40560-021-00525-z.

Machine learning-based prediction models for accidental hypothermia patients

Yohei Okada  1   2   3 , Tasuku Matsuyama  4 , Sachiko Morita  5 , Naoki Ehara  6 , Nobuhiro Miyamae  7 , Takaaki Jo  8 , Yasuyuki Sumida  9 , Nobunaga Okada  4   10 , Makoto Watanabe  4 , Masahiro Nozawa  11 , Ayumu Tsuruoka  12 , Yoshihiro Fujimoto  13 , Yoshiki Okumura  14 , Tetsuhisa Kitamura  15 , Ryoji Iiduka  16 , Shigeru Ohtsuru  17
Affiliations

Machine learning-based prediction models for accidental hypothermia patients

Yohei Okada et al. J Intensive Care. .

Abstract

Background: Accidental hypothermia is a critical condition with high risks of fatal arrhythmia, multiple organ failure, and mortality; however, there is no established model to predict the mortality. The present study aimed to develop and validate machine learning-based models for predicting in-hospital mortality using easily available data at hospital admission among the patients with accidental hypothermia.

Method: This study was secondary analysis of multi-center retrospective cohort study (J-point registry) including patients with accidental hypothermia. Adult patients with body temperature 35.0 °C or less at emergency department were included. Prediction models for in-hospital mortality using machine learning (lasso, random forest, and gradient boosting tree) were made in development cohort from six hospitals, and the predictive performance were assessed in validation cohort from other six hospitals. As a reference, we compared the SOFA score and 5A score.

Results: We included total 532 patients in the development cohort [N = 288, six hospitals, in-hospital mortality: 22.0% (64/288)], and the validation cohort [N = 244, six hospitals, in-hospital mortality 27.0% (66/244)]. The C-statistics [95% CI] of the models in validation cohorts were as follows: lasso 0.784 [0.717-0.851] , random forest 0.794[0.735-0.853], gradient boosting tree 0.780 [0.714-0.847], SOFA 0.787 [0.722-0.851], and 5A score 0.750[0.681-0.820]. The calibration plot showed that these models were well calibrated to observed in-hospital mortality. Decision curve analysis indicated that these models obtained clinical net-benefit.

Conclusion: This multi-center retrospective cohort study indicated that machine learning-based prediction models could accurately predict in-hospital mortality in validation cohort among the accidental hypothermia patients. These models might be able to support physicians and patient's decision-making. However, the applicability to clinical settings, and the actual clinical utility is still unclear; thus, further prospective study is warranted to evaluate the clinical usefulness.

Keywords: Accidental hypothermia; Artificial intelligence; Gradient boosting tree; Lasso; Machine learning; Prediction; Random forest.

PubMed Disclaimer

Conflict of interest statement

The authors declare that they have no competing interests.

Figures

Fig. 1
Fig. 1
Study flowchart
Fig. 2
Fig. 2
The features of the models. Beta coefficients value in lasso and importance of variables in random forest and gradient boosting tree were shown. ADL: activity of daily living, BT: body temperature, SBP: systolic blood pressure, GCS: Glasgow coma scale, WBC: white blood cell count, Hgb: hemoglobin, Hct: hematocrit, PLT: platelet count, BUN: blood urea nitrogen, TP: Total protein, Alb: serum albumin, T-bil: Total bilirubin, CK: creatine kinase
Fig. 3
Fig. 3
ROC of the models. The C-statistics [95% CI] of the models; lasso: 0.784 [0.717–0.851], random forest: 0.794 [0.735–0.853], boosting tree: 0.780 [0.714–0.847], SOFA: 0.787 [0.722–0.851], and 5A score: 0.750 [0.681–0.820]. CI: confidence interval, ROC: receiver operating curve
Fig. 4
Fig. 4
Calibration plot and decision curve analysis. Left: calibration plot, right: decision curve analysis. Calibration plot, X-axis: predicted probability, Y-axis: observed frequency in validation cohort. Decision curve analysis, X-axis: threshold probability, y-axis: net-benefit. The detail of net-benefit and decision curve analysis is described in Supplementary Appendix 3 in Additional file 1
See this image and copyright information in PMC

References

    1. Brown DJ, Brugger H, Boyd J, Paal P. Accidental hypothermia. N Engl J Med. 2012;367(20):1930–1938. doi: 10.1056/NEJMra1114208. - DOI - PubMed
    1. Matsuyama T, Morita S, Ehara N, Miyamae N, Okada Y, Jo T, Sumida Y, Okada N, Watanabe M, Nozawa M, Tsuruoka A, Fujimoto Y, Okumura Y, Kitamura T, Ohta B. Characteristics and outcomes of accidental hypothermia in Japan: the J-Point registry. Emerg Med J. 2018;35(11):659–66. 10.1136/emermed-2017-207238. Epub 2018 Jun 9. PMID: 29886414. - PubMed
    1. Medicine JAfA: The clinical characteristics of hypothermic patients in the winter of Japan—the final report of Hypothermia STUDY 2011. J Jpn Assoc Acute Med. 2013;24:12.
    1. Vassal T, Benoit-Gonin B, Carrat F, Guidet B, Maury E, Offenstadt G. Severe accidental hypothermia treated in an ICU: prognosis and outcome. Chest. 2001;120(6):1998–2003. doi: 10.1378/chest.120.6.1998. - DOI - PubMed
    1. Okada Y, Matsuyama T, Morita S, Ehara N, Miyamae N, Jo T, Sumida Y, Okada N, Watanabe M, Nozawa M, et al. The development and validation of a "5A" severity scale for predicting in-hospital mortality after accidental hypothermia from J-point registry data. J Intensive Care. 2019;7:27. doi: 10.1186/s40560-019-0384-2. - DOI - PMC - PubMed