Prediction of in‐hospital hypokalemia using machine learning and first hospitalization day records in patients with traumatic brain injury

• DOI: 10.1111/cns.13993
• Publication Date: 2022-10-19T04:14:48Z
• Authors: Zhengyu Zhou; Chiungwei Huang; Pengfei Fu; Hong Huang; Qi Zhang; Xuehai Wu; Qiong Yu; Yirui Sun
• Keywords: Medical record

Abstract Aims Hypokalemia is a common complication following traumatic brain injury, which may complicate treatment and lead to unfavorable outcomes. Identifying patients at risk of hypokalemia on the first day admission helps implement prophylactic treatment, reduce complications, improve prognosis. Methods This multicenter retrospective study was performed between January 2017 December 2020 using electronic medical records admitted due injury. A propensity score matching approach adopted with ratio 1:1 overcome overfitting data imbalance during subgroup analyses. Five machine learning algorithms were applied generate best‐performed prediction model for in‐hospital hypokalemia. The internal fivefold cross‐validation external validation demonstrate interpretability generalizability. Results total 4445 TBI recruited analysis generation. occurred in 46.55% incidences mild, moderate, severe 32.06%, 12.69%, 1.80%, respectively. associated increased mortality, while cast greater impacts. logistic regression algorithm had best performance predicting decreased serum potassium moderate‐to‐severe hypokalemia, an AUC 0.73 ± 0.011 0.74 0.019, further verified two datasets, including our previous published open‐assessed Medical Information Mart Intensive Care database. Linearized calibration curves showed no statistical difference ( p > 0.05) perfect predictions. Conclusions occurrence injury can be predicted by hospitalization algorithms. optimal both validation.