| 摘要: | Importance: Laboratory tests consider an essential part of patient safety as patients’ screening, diagnosis, and follow-up are solely based on laboratory tests. Diagnosis of patients could be wrong, missed, or delayed if laboratory tests are performed erroneously. However, recognizing the value of correct laboratory test ordering remains underestimated by policymakers and clinicians. Nowadays, artificial intelligence methods such as machine learning and deep learning (DL) have been extensively used as powerful tools for pattern recognition in large data sets.
Objective: Our objective was to develop a deep learning-based automated model that can provide laboratory tests recommendation based on simple variables available in EHRs.
Design, Settings and Participants: A retrospective analysis of the National Health Insurance database between January 1, 2013, and December 31, 2013, was performed. We reviewed the record of 530,050 patients who visited the hospital and ordered at least one laboratory test. The entire dataset was partitioned into training (70%) and testing (30%) to develop a prediction model. In the internal validation, 20% of data were randomly selected from the train set to evaluate our prediction model.
Main Outcomes and Measures: Diagnostic accuracy measures, including area under the receiver operating characteristic curve (AUROC), precision, recall, F1score, and hamming loss were considered to evaluate our prediction model while predicting laboratory tests.
Results: There were 1,463,837 prescriptions with 315 different types of laboratory tests. The deep learning model for predicting laboratory tests had a higher area under the receiver operating characteristic curve (AUROC micro = 0.98 and AUROC macro =0.94). The precision, recall, and hamming loss were different in various cutoff values. Using low cutoff value, the deep learning model identifies appropriate laboratory tests with 0.96% sensitivity. However, the model showed balanced precision, recall, F1 score, and hamming loss at cutoff value 0.35 (Precision: 0.57, recall: 0.57, F1 score: 0.55, and hamming loss: 0.017).
Conclusion and Relevance: The findings suggest that DL exhibited good discriminative capability for predicting laboratory tests using routinely collected EHR data. Utilization of DL approaches can facilitate optimal laboratory test selection for patients, which may contribute to reducing under and over-utilization of laboratory tests. |
