BACKGROUND
Artificial intelligence is rapidly gaining traction in healthcare. Mediverse, an AI-symptom checker constructed using medical knowledge database, predictive algorithm, Bayesian inference, feedback loop and clinical validation, aimed to aid in the initial diagnosis and provide indications for referral for patients with common symptoms, thereby improving efficiency of triaging and diagnostic process with concurrent patient education.
OBJECTIVE
We conducted a study on the Mediverse AI Symptom Checker to assess its concordance rate with the diagnoses formulated by the physicians. Additionally, we assessed patients' and doctors' satisfaction with the system's patient education on diseases. User-friendliness and acceptance levels of the system were also evaluated among both patients and doctors.
METHODS
This is a prospective validation study. Patients presenting to both outpatient clinic and emergency department of Hospital Putrajaya, Malaysia, covering study period from 1st of January 2023 to 31st March 2023, were first assessed and treated by doctors. They were then given access to Mediverse AI symptom checker to key in their main presenting symptoms. Diagnosis accuracy was evaluated by comparing the AI generated diagnosis with the diagnosis of the treating doctor. An independent assessor evaluated the concordance of the diagnosis offered by Mediverse AI symptom checker and the doctor’s diagnosis and classified the results as “complete match” (the exact same diagnosis), “partial match” (not the exact same diagnosis but a similar diagnosis that would result in the same management strategy) and “mismatch” (different category of diagnosis that would result in a different management strategy). User’s satisfaction was also evaluated.
RESULTS
The study involved 75 doctors and 325 patients from general outpatient clinic and emergency department. In 90.5% of the encounters, the top three diagnoses offered by Mediverse AI symptom checker matched the doctor’s diagnosis, with 87.3% of complete match or partial match. Among the participating doctors, the majority (n=180, 55.4%) strongly agreed that the triage advice provided by the Mediverse AI symptom checker for each diagnosis was appropriate.
CONCLUSIONS
The Mediverse AI symptom checker demonstrates promise as a tool for assisting initial diagnosis, exhibiting a high level of concordance with human diagnosis and receiving a high user satisfaction rate. Further research is warranted to assess its effectiveness in a diverse patient population encompassing a wide range of symptoms and conditions.
CLINICALTRIAL
nil