Objective To construct a prediction model for the transfer-out rate of patients in the intensive care unit (ICU) based on frequent pattern growth (FP-Growth) algorithm and evaluate its application value.
Methods A total of 4, 000 ICU patients were selected as study subjects and divided into model construction group and validation group. Clinical data from both groups were collected and compared. Association rule analysis based on the FP-Growth algorithm was performed in the model construction group. A prediction model for the ICU patient transfer-out rate was established by calculating effective strong association rules among the elements in the final scanned set of the model construction group. In internal validation, the model′s consistency was evaluated using calibration curves and other metrics. In external validation, the area under the receiver operating characteristic (ROC) curve (AUC) for predicting the ICU patient transfer-out rate was compared between the model construction group and the validation group.
Results In the model construction group, with corresponding clinical data available, the transfer-out rates within 7 days, > 7 to14 days, and > 14 to 21 days were 71%, 40% and 18%, respectively. During internal validation, the calibration curve demonstrated satisfactory consistency between predicted and observed values. In external validation, the AUC values for the model′s predictions of transfer-out rates at 7 days, > 7 to14 days, and > 14 to 21 days in the model construction group were 0.880, 0.861 and 0.654, respectively.
Conclusion The prediction model for the ICU patient transfer-out rate exhibits favorable short-term (within 14 days) predictive performance, and its application holds certain reference value for optimizing the overall treatment efficacy in the ICU and the allocation of medical resources.
DownLoad: