Development of a Decision Tree Analysis model that predicts recovery from acute brain injury

Authors


  • This work was supported by an Inha University Research Grant.

Wha Sook Seo, Department of Nursing, College of Medicine, Inha University, Yong Hyun Dong 253, Incheon 402-751, Korea. Email: wschang@inha.ac.kr

Abstract

Aim:  The present study was conducted to identify significant demographic, illness-related, and physiological factors associated with recovery at 1 month after brain injury, and to develop and evaluate a Decision Tree Analysis-based prediction model.

Methods:  This study was conducted using a prospective study design. The study subjects were 190 adult patients with brain injury admitted to the neurological intensive care unit of a university hospital located in Incheon, South Korea. Degree of recovery from brain injury was evaluated 1 month after admission using the Glasgow Outcome Scale.

Results:  A prediction model was developed using Decision Tree Analysis. The most significant predictor of recovery at 1 month after brain injury using the devised model was Glasgow Coma Scale score on admission, and its optimum cut-off value for the differentiation of good and poor recovery was 8.5. The next best predictors were age and blood glucose level on admission, which had cut-off values of 49.5 years and 155 mg/dL, respectively. In addition, hematoma volume, systolic blood pressure, and respiratory rate on admission were also found to predict 1 month recovery significantly, with corresponding cut-off values of 44 cc, 168 mmHg, and 29/min, respectively.

Conclusion:  The developed model includes common and routinely used monitoring parameters as significant predictors, and proposes cut-off values for these predictors. This model appears to be useful for predicting 1 month recovery in various brain injury types. The authors believe that the devised model provides a basis for the evidence-based nursing care of brain injured patients during the acute stage.

Ancillary