Performance of drug sales and primary care encounters data for early detection of influenza-like illness surges in Brazil: a national time series analysis

Read the full article See related articles

Listed in

This article is not in any list yet, why not save it to one of your lists.
Log in to save this article

Abstract

Background: Effective pandemic preparedness requires integration and continuous evaluation of multiple data sources, including clinical and alternative data. This nationwide study with a large population evaluated the value of over the counter (OTC) drug sales and primary health care (PHC) encounter data for the early detection of surges in respiratory disease related hospitalization in Brazil. Methods: From November 2022 to December 2024, we analysed weekly time series data on PHC encounters, OTC drug sales used for influenza like illness (ILI) treatment, and respiratory diseases related hospitalizations across all 510 immediate regions in Brazil. Health data were provided by the Brazilian Ministry of Health, and drug sales data were sourced from IQVIA. We applied negative binomial autoregressive moving average models in the context of Statistical Process Control techniques to identify early warnings in PHC and OTC time series. Surges in respiratory disease related hospitalizations were defined as periods of weekly counts exceeding a defined threshold, enabling evaluation of each surveillance stream's performance. Findings: During the study period, Brazil recorded 1.8 million respiratory diseases related hospitalizations, 51 million PHC encounters and 584 million OTC drug units sold for ILI. In general, all three data shows increases in cases in March, with ILI associated PHC encounters and hospitalizations peaking in May, and ILI associated OTC sales drug sales peaking in July. A total of 13,106 and 13,066 warnings were detected in PHC encounters and OTC drug sales timeseries, respectively. OTC data anticipated 63.8% of 556 hospitalization surges events 1 to 3 weeks in advance, detected 6.5% concurrently, and missed 29.7%. PHC data anticipated 54.3%, detected 8.5% concurrently, and missed 37.2%. OTC data showed higher sensitivity (70.3%) but lower specificity (48.8%) compared to PHC data (62.8% and 52.8%, respectively) to anticipate respiratory diseases-related hospitalization surges. Performance varied regionally: both data streams performed best in the Center West, while OTC underperformed in urban areas and PHC in smaller populations. In 68.8% of regions, at least one stream provided high precision (Pr 51.8% for PHC encounters data and Pr 46.8% for OTC drug sales), and in 62.6% of regions, they anticipated at least 60% of hospitalizations surges. Interpretation: This study demonstrates the reliability and value of using OTC drug sales data, complementing PHC based surveillance to enhance early detection of respiratory diseases-related hospitalizations surges in a large upper middle income country.

Article activity feed