Artificial intelligence in Brazilian Primary Health Care: scoping review.
Dany Alexandra Morales Botero, Rafaele Bonfim, Karina Fonseca and 3 others
PMID 41370529WHAT IT FOUND
Across 27 Brazilian primary care AI studies, computer models were used mainly for early diagnosis, screening and risk prediction.
Language-based AI showed no applications, and weak data, internet, cost and access limit use.
Key findings
01The scoping review included 27 publications after screening 981 identified studies.
02Fourteen included studies used machine learning, 13 used deep learning, and natural language processing did not demonstrate applicability results.
03Nineteen different AI applications were reported in Brazilian primary care, with strengths and challenges including early diagnosis, risk stratification, limited data, low internet connection, high cost and access problems.
STILL TO COME
How it was doneWhat they foundWhat it means for RNs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The search was conducted by a single researcher in February 2025, so some relevant publications may have been missed. The review did not search Google Scholar or manually screen references of included studies. Four included publications were conference abstracts rather than journal articles. The included studies had heterogeneous objectives, designs and populations, so the review could not combine them into a single estimate of effect. It is a scoping review: it mapped evidence and reported strengths and challenges, but did not test whether AI improved patient outcomes. Natural language processing did not demonstrate applicability results, so the review gives no practical information about language-based AI in primary care.
The easy way to misread this
Do not read this review as evidence that AI improves patient care in primary health care. It mapped 27 publications and reported applications and challenges, but it did not test effectiveness, and natural language processing showed no applicability results.