Routine Laboratory Blood Tests Predict SARS-CoV-2 Infection Using Machine Learning.

Area of Science:

• Medical Diagnostics
• Machine Learning in Healthcare
• Infectious Disease Management

Background:

• Accurate and rapid diagnostics for SARS-CoV-2 are crucial for patient care and healthcare worker safety.
• Reverse transcription-polymerase chain reaction (RT-PCR) is the predominant diagnostic test but often has delays.
• Routine laboratory tests offer a readily available alternative with faster turnaround times.

Purpose of the Study:

• To develop and validate a machine learning model for predicting SARS-CoV-2 infection status.
• To assess the model's performance using routine laboratory test results.
• To provide a rapid diagnostic tool for early identification of SARS-CoV-2 positive individuals.

Main Methods:

• A gradient boosting decision tree (GBDT) model was trained using demographic data and 27 routine laboratory tests.
• Data from 3,356 patients (1,402 SARS-CoV-2 positive) were used for model training.
• The model was validated on an independent patient dataset from a separate hospital.

Main Results:

• The GBDT model achieved an area under the receiver operating characteristic curve (AUC) of 0.854 in the training set.
• Validation on an independent dataset yielded a comparable AUC of 0.838.
• The model successfully predicted initial SARS-CoV-2 RT-PCR positivity in 66% of individuals who later tested positive.

Conclusions:

• Routine laboratory tests, analyzed by a machine learning model, can facilitate early and rapid identification of SARS-CoV-2 infection.
• This approach can assist in managing patient care and protecting healthcare personnel.
• The model offers a valuable tool for SARS-CoV-2 detection in resource-limited settings where RT-PCR is inaccessible.