Researchers from Iceland used artificial intelligence to develop a machine-learning model to evaluate patients with respiratory symptoms before they visited a primary care clinic. The researchers utilized only questions that a patient could be questioned about before a clinic appointment to build the machine-learning model.
Image Credit: Blue Planet Studio/Shutterstock.com
The data came from 1,500 clinical text notes that contained a physician’s evaluation of the patient’s symptoms and indicators, as well as the reasoning for clinical decisions taken during the consultation, such as imaging referrals and prescriptions. Based on clinical notes, patients were assigned to one of five diagnostic groups.
Patients from every primary care facility in Iceland’s capital region were included. Each patient was given a score in two external databases, and patients were then categorized into ten risk categories. The researchers next looked at selected results from each group.
In comparison to higher risk groups 6–10, patients in risk groups 1–5 were younger, had lower rates of lung inflammation, were less likely to undergo reevaluation in primary and emergency care, and were less likely to be referred for chest X-Rays or antibiotic prescriptions.
None of the chest X-Rays or medical diagnoses of pneumonia were present in the bottom five groups. The model can decrease the amount of chest X-Ray referrals by removing them in risk categories 1 through 5, according to the study’s findings.
What is Known
People often see primary care doctors for respiratory issues. Many of their symptoms, however, go away on their own. Prioritizing patients before doctor appointments, according to researchers, could reduce the need for pointless diagnostic tests, medical expenses, and the overprescription of medicines, which can increase bacterial resistance.
What This Study Adds
Researchers discovered that a machine learning model can successfully group patients into ten risk groups, enabling clinicians to communicate with lower-risk patients in a way that does not add to their already busy work schedules and to focus on higher-risk patients and those who have severe respiratory symptoms.
According to the researchers, the machine learning model could result in lower expenses for society as a whole, the healthcare system, and patients.
Ellertsson, S., et al. (2023) Triaging Patients with Artificial Intelligence for Respiratory Symptoms in Primary Care to Improve Patient Outcomes: A Retrospective Diagnostic Accuracy Study. Annals of Family Medicine. doi:10.1370/afm.2970.