Researchers in Iceland have trained machine learning models using artificial intelligence to triage patients with respiratory symptoms before they visit primary care clinics. To train the machine learning model, the researchers used only questions that patients might be asked before visiting the clinic. Information was extracted from 1,500 clinical text notes, including physician interpretations of patient symptoms and signs and the reasons for clinical decisions made during the consultation, such as imaging and prescriptions. Patients were classified into one of five diagnostic categories based on information from clinical notes. Patients from all primary care clinics in the Icelandic metropolitan area were included. The model scored each patient on two external datasets and divided the patient into her ten risk her groups. The researchers then analyzed the selected results in each group.
Patients in risk groups 1–5 are younger, have lower pulmonary inflammation rates, are less likely to be reassessed in first-line and emergency care, and are less likely to prescribe antibiotics and chest x-rays than high-risk patients. were less likely to receive Groups 6-10. Her bottom five groups did not include chest radiographs showing signs of pneumonia or a doctor’s diagnosis of pneumonia. The researchers concluded that the model could reduce the number of referrals by excluding referrals for chest radiography in risk groups 1–5.
What we know: Respiratory symptoms are a common reason people visit primary care clinicians. However, many of the symptoms resolve on their own. Researchers argue that triaging patients before seeing a doctor can reduce unnecessary diagnostic tests. Medical bills. Overprescription of antibiotics can increase bacterial resistance.
What this study adds: The researchers found that a machine learning model could effectively classify patients into 10 risk groups. This will enable clinicians to communicate with low-risk patients in a way that does not add to heavy work schedules and enables patient care. High-risk patients and patients with severe respiratory symptoms. The research team claims that machine learning models can reduce costs for patients, healthcare systems and society.
Improving patient outcomes by triaging patients with artificial intelligence for respiratory symptoms in primary care: a retrospective diagnostic accuracy study
Emil L. Sigurdsson, M.D., et al.
Metropolitan primary health care.Icelandic Primary Health Care Development Center and University of Iceland Department of Family Medicine, Reykjavik, Iceland
permanent link
