Racial Data Errors Are Likely to Occur in Electronic Medical Records
THURSDAY (HealthDay News), September 5, 2024 A study released online in JAMA Network Open on September 3rd, according to children’s electronic medical records (EMRs) across health systems, reveals substantial inaccuracies in the recording of racial data.
Researchers Gary L. Freed, M.D., of the University of Michigan in Ann Arbor, and associates looked at the rate of racial and ethnic attribution errors in electronic medical records from the state’s three biggest pediatric health systems. Comparisons of the parent or guardian report of race and ethnicity with the electronic medical records (EMRs) for 4,333 children seen between September 1, 2023, and January 31, 2024, at outpatient clinics, emergency departments, and inpatient units were part of the analysis.
The researchers discovered that the highest mistake rate (ranging from 41 to 78 percent) was observed when the parental report’s racial classification exactly matched the EMR. With the consolidation of racial alternatives, the matching percentage increased, narrowing the disparities between the health systems to 78–88 percent matching. Between 65 and 95 percent of health systems’ EMRs and parental reports matched in terms of ethnicity. The percentage of missing racial or ethnic data in the EMR that was considered a nonmatch varied between health systems, ranging from 2 to 10%. The health system with the most options for race and ethnicity also had the highest error rates.
Large error rates cast doubt on the findings, implying that there may or may not be injustices and disparities in particular clinical care domains, and they may jeopardize efforts to enhance treatment, according to the authors.
Acceptable Responses to Half of Patient Urology Questions Are Provided by ChatGPT
THURSDAY, August 29, 2024 (NewsDay) — A research released in the Sept. 1 edition of urological Practice found that ChatGPT can satisfactorily answer almost half of patient urological inquiries.
The quality of ChatGPT’s answers to actual urology patient messages was assessed by Michael Scott, M.D., and associates from Stanford University School of Medicine in Palo Alto, California. A total of 100 electronic patient messages were analyzed, and five urologists independently assessed the ChatGPT responses.
47 percent of the responses were deemed suitable for patient distribution, according to the researchers’ findings. On easy compared to difficult questions, ChatGPT performed better (56 versus 34 percent acceptable). Easy queries received responses that were more precise, thorough, beneficial, and comprehensible than those on more challenging topics.
Nearly 50% of patient messages received satisfactory responses from ChatGPT, with easy inquiries doing better than complex ones. The authors state that using ChatGPT to reply to patient communications can help to enhance wellness and lessen the time commitment for the care team. “As generative artificial intelligence technology evolves, artificial intelligence performance is projected to continue to increase.
