# Evaluating antimicrobial prescriptions in primary health care across an entire Brazilian city through the analysis of electronic medical records: where public health and data science converge

## TL;DR

This study uses electronic health records from a Brazilian city to analyze antimicrobial prescriptions, revealing patterns and opportunities for better public health decisions.

## Contribution

The novel contribution is the application of data science to analyze antimicrobial prescriptions at a city-wide scale in primary care settings.

## Key findings

- Upper respiratory infections were the most common reason for antimicrobial prescriptions.
- Amoxicillin was the most frequently prescribed antimicrobial.
- The study identified gaps in documentation and non-first-line treatments for some infections.

## Abstract

Exploring records from entire cities to make decisions, particularly within public health systems, remains challenging.

This study investigates the public health data of São Caetano do Sul (SCS), in Brazil, to uncover patterns of antimicrobial prescriptions for infectious diseases using electronic health system records from primary care. Data science techniques such as preprocessing, transformation, loading, and analytics were also applied to achieve this goal.

From January to September 2023, a total of 575,616 records of medical appointments were analyzed, and 67,023 patients underwent one or more medical appointments of which 16,572 had infectious diagnoses. There were 7,938 prescriptions of antimicrobials for infections of which the most frequent were upper respiratory infections (37%), gingivitis/periodontal disease (20%), and urinary tract infections (9%). The most frequently prescribed antimicrobials were amoxicillin (23%), azithromycin (15%), amoxicillin/clavulanate (13%), ciprofloxacin (11%), and cephalexin (11%). A preliminary evaluation of the data highlighted several points for targeted interventions, as well as challenges in obtaining certain information. For instance, some infections lacked documented antimicrobial treatment, while others were managed with medications not considered first-line options.

Implementing a system that can extract data directly from electronic records and automatically present it in a logical and relevant way to health professionals—including policymakers and administrators—would enable the identification of potential problems, the planning of interventions to improve antimicrobial use, and the monitoring of their impact. Our findings highlight opportunities to improve antimicrobial prescribing through data-driven tracking, analysis, and feedback mechanisms.

The online version contains supplementary material available at 10.1186/s12911-025-03260-9.