Leveraging the power of digital health to fight pandemics: The example of ÆSOPAbstract
Emerging outbreaks highlight the need for early warning systems, but low-resource centers often face challenges to maintain surveillance capabilities. Administrative data-based systems offer a cost-efficient approach to strengthening surveillance. The present study evaluated whether a primary health care (PHC)-based early warning system could anticipate respiratory outbreak detection, when compared to traditional surveillance. Weekly counts of influenza-like illness PHC encounters in Rio de Janeiro were analyzed from October 2019 to May 2020 and from October 2021 to May 2022. PHC data was compared to weekly surveillance notifications and used time series regression to estimate predicted counts of PHC encounters. Subsequent outbreak warnings were then issued. Our study identified 659,230 influenza-like illness PHC encounters in the first period, and 702,886 in the second period. In the first period, PHC data deviated from baseline two weeks before the rise in notifications during the first COVID-19 wave and one week earlier in the second period. The PHC-based system successfully triggered warnings capable of anticipating the surveillance system. Our findings show PHC-based early warning systems can anticipate outbreaks earlier than traditional surveillance, supporting their role in enhancing surveillance in low-resource settings.
Key words:
Early Warning System; Primary Health Care; Digital Health
Resumo
Surtos de doenças emergentes apontam a necessidade de sistemas de alerta precoce, mas locais com recursos limitados têm dificuldades em manter sistemas de vigilância ativa. Sistemas de alerta precoce baseados em dados administrativos são uma abordagem custo-efetiva para fortalecer ações de vigilância. Avaliar a tempestividade de um sistema de alerta precoce baseado em dados de atenção primária à saúde (APS) na detecção de surtos respiratórios em comparação à vigilância epidemiológica. Analisamos dados semanais de atendimentos na APS do Rio de Janeiro entre outubro/2019 a maio/2020 e outubro/2021 a maio/2022. A distribuição dos dados da APS foi comparada com notificações da vigilância. Um modelo de regressão de séries temporais foi aplicado para emitir alertas de surtos a partir dos dados da APS. No primeiro período, os dados da APS indicaram anteciparam em duas semanas o aumento das notificações resultante da primeira onda da COVID-19. No segundo período, essa antecipação foi de uma semana. O sistema baseado na APS emitiu alertas antecipando o sistema de vigilância. Sistemas de alerta precoce baseados na APS podem antecipar surtos antes da vigilância tradicional, fortalecendo a vigilância em locais com recursos limitados.
Palavras-chave:
Sistema de Alerta Precoce; Atenção Primária à Saúde; Saúde Digital
Resumen
Brotes emergentes resaltan la necesidad de sistemas de alerta temprana. Regiones con recursos limitados enfrentan dificultades para mantener una vigilancia efectiva. Los sistemas basados en datos administrativos ofrecen un enfoque coste-efectivo para fortalecer la vigilancia. Evaluamos si un sistema de alerta temprana basado en datos de la atención primaria de salud (APS) podría anticipar la detección de brotes respiratorios en comparación con la vigilancia tradicional. Analizamos los recuentos semanales de consultas por enfermedades tipo influenza en APS en Río de Janeiro, en dos periodos: octubre-2019 a mayo-2020 y octubre-2021 a mayo-2022. Comparamos los datos de APS y de notificaciones semanales de vigilancia y utilizamos regresión de series temporales para emitir alertas de brotes con base en los datos de APS. En el primer periodo, los datos de APS anticiparan en dos semanas el aumento de las notificaciones durante la primera ola de COVID-19. El sistema basado en APS emitió alertas anticipándose al sistema de vigilancia tradicional. Los sistemas de alerta temprana basados en APS pueden anticipar brotes antes que la vigilancia tradicional, mejorando de la vigilancia en localidades con recursos limitados.
Palabras clave:
Sistema de Alerta Temprana; Atención Primaria de Salud; Salud Digital
Introduction
Epidemiological surveillance is crucial for planning and informing decisions related to preparedness and response to health emergencies1. Recent episodes of emerging and re-emerging zoonotic outbreaks highlight the importance of developing efficient early warning systems (EWS)2. However, the reality of frequently ill-equipped and overburdened health services hinders the development of such EWS and the refinement of epidemiological surveillance capabilities3,4.
Investing in automated processes for disease surveillance represents an achievable solution for establishing an EWS, particularly in low-resource settings. The growing availability of electronic health records (EHR) and administrative health databases provide valuable epidemiological information5,6. In this context, the Municipal Health Department of Rio de Janeiro recently established the Epidemiological Intelligence Center (EIC), which incorporates technological resources aligned with epidemic intelligence for the efficient use of data to enhance epidemiological surveillance7.
The epidemiological surveillance system in Brazil was established in the early 1970s and is regarded as a comprehensive and well-organized system in which data is compiled and secured in a stable and structured database8. It relies on the active notification of a pre-specified list of diseases and conditions, which imposes timeliness and cost-efficiency limitations. Brazil’s Primary Health Care (PHC) system provides comprehensive health care with great granularity as part of the Unified Health System (SUS), the largest public and universal health system in the world. For funding and planning purposes, all PHC encounters are regularly registered in a database managed by the Ministry of Health (MoH). Integrating the PHC database into epidemiological surveillance activities ensures timeliness to the system, as no duplication of registry is required. This cost-efficient approach might allow the development of an EWS even in low-resource settings4,9.
The present study explores the use of the PHC database in anticipating the detection of acute respiratory infection (ARI) outbreaks when compared to traditional epidemiological surveillance in the city of Rio de Janeiro. This system was developed as part of the ÆSOP initiative, a previously described EWS10.
Methods
The timeliness of detection of ARI outbreaks when using the PHC database was compared to the traditional epidemiological surveillance system (SIVEP-Gripe) in the city of Rio de Janeiro.
Data sources
Data referent to PHC encounters were extracted from the National Information System on PHC (Sistema de Informação em Saúde para a Atenção Básica - SISAB). The SISAB database harbors data on all publicly funded PHC encounters, coded by either the International Classification of Diseases (ICD-10) or the International Classification of Primary Care (ICPC-2). Our study used weekly counts of PHC encounters due to influenza-like illness (ILI) in the city of Rio de Janeiro. A list of 50 diagnostic codes was included to identify encounters possibly related to ILI (database description and the scripts are available at https://github.com/cidacslab/AESOP-Data-Documentation/tree/main/DataPipeline).
The traditional epidemiological surveillance system for acute respiratory infections in Brazil is based on the mandatory reporting of all hospitalized or deceased cases of Severe Acute Respiratory Syndrome (SARS) to the Notifiable Diseases Information System (SIVEP-Gripe). Our study extracted weekly counts of all SARS cases in Rio de Janeiro from the openly available, non-identified SIVEP-Gripe database provided by the MoH. A SARS case is defined as any patient presenting at least two of the following ILI symptoms: fever; cold shiver; sore throat; headache; cough; runny nose; loss of smell; loss of taste, as well as any of the following signs of case severity: dyspnea or shortness of breath; chest pressure or persistent chest pain; oxygen saturation below 95% in ambient air; and cyanosed lips or face. Additionally, all death cases due to SARS are reported, regardless of hospitalization.
Analysis
To evaluate whether modeling/monitoring the PHC database anticipates the detection of ARI outbreaks by the traditional epidemiological surveillance system, the proportion of PHC encounters due to ILI was compared to the total number of SARS reports, per week in two distinct periods with well-documented ARI outbreaks: the first period ranging from October 6, 2019, to May 2, 2020 (epidemiological weeks 41-2019 to 18-2020), which encompasses the first wave of COVID-19 cases in Brazil11, and the second period ranging from October 3, 2021, to April 30, 2022 (epidemiological weeks 40-2021 to 17-2022), which encompassed a massive H3N2 outbreak, followed by the first Omicron-variant COVID-19 wave in Rio de Janeiro12.
Additionally, our work tested whether an EWS model applied to the PHC time series would trigger timely warnings in relation to the first pandemic wave in 2020, using data from January 1, 2017, to December 31, 2019, to establish a baseline of ILI-related PHC encounters by fitting a general linear model with negative binomial distribution, with one term per year to control for annual trends, and harmonic terms to control for seasonality. An offset term was also added to the total count of PHC encounters during the week to account for changes in the health-seeking behavior4. The baseline was used to predict the expected number of ILI-related PHC encounters from January 1 to May 2, 2020. The predicted counts with the corresponding 95% confidence interval (95%CI) defined the threshold for the EWS to trigger a warning. The sustained high fluctuations of PHC encounters due to the COVID-19 pandemic from 2020 to 2022, along with the misleadingly low numbers due to lockdowns, hindered us from using a similar methodology to evaluate the PHC-based EWS capabilities for triggering timely and accurate warnings in the second study period.
All analyses were conducted using R software, version 4.3.1, and the surveillance package13.
Research ethics
This study is based on secondary, aggregated, non-identified data, and was approved by the Ethical Review Board of the Oswaldo Cruz Foundation, Instituto Gonçalo Moniz, logged under CAAE 61444122.0.0000.0040.
Results
Our study identified 659,230 ILI-related PHC encounters from October 6, 2019, to May 2, 2020, and 702,886 from October 3, 2021, to April 30, 2022, which corresponds to a median of 17,804 (Interquartile Range - IQR: 12,910-27,250) and 13,908 (IQR: 9,637-22,672) encounters per week in the first and second study periods, respectively. The proportion of ILI-related PHC encounters ranged from 6.5% to 25.0% per week in the first period, and from 4.6% to 47.5%, in the second period. There were 7,581 SARS reports in the first period, 93% of which were registered between the 13th and 17th epidemiological weeks (March 22 to April 25). In the second period, from October 2021 to January 2022, 13,660 SARS reports were found (median: 351 IQR: 278-516 per week).
Figure 1 presents the weekly time series of the proportion of ILI-related PHC encounters and SARS reports in Rio de Janeiro across both study periods. In the first period (Figure 1A), representing the initial wave of the COVID-19 pandemic, both curves followed a similar pattern until early March 2020, when the proportion of ILI-related PHC encounters departed from the baseline, occurring approximately two weeks earlier than the corresponding rise in SARS reports.
Proportion of ILI-related PHC encounters and absolute number of SARS cases, per epidemiological week, Rio de Janeiro. Figure 1A shows data from October 6, 2019, to May 2, 2020 (epidemiological weeks 41-2019 to 18-2020). Figure 1B shows data from October 3, 2021, to April 30, 2022 (epidemiological weeks 40-2021 to 17-2022).
Figure 1B exhibits two episodes of upward departures from the baseline level: a first peak in mid-November 2021, followed by a second peak in early January 2022. In both episodes, the rise in the proportion of ILI-related PHC encounters anticipates the increase in SARS reports by one week.
Figure 2 shows the EWS based on PHC encounters. The model identified warnings of an ILI-related outbreak since the last week of February, which anticipates the identification of a rising slope in the SARS notification time series.
Observed number and estimated threshold of weekly counts of ILI-related PHC encounters in the baseline (January 2017 to December 2019) and predicted (January to May 2020) time series, Rio de Janeiro. The black line shows the weekly counts of observed ILI-related PHC encounters. The light gray area shows the 95% predicted threshold. Black dots show warnings triggered by the EWS. The vertical dashed line separates the baseline from the predicted time periods.
Discussion
Our findings demonstrate that leveraging PHC administrative data provides timely and valuable insights for epidemiological surveillance. Systematic monitoring of ILI-related PHC encounters in ARI outbreaks anticipates the current in-place epidemiological surveillance system by one to two weeks. Additionally, the use of a PHC-based EWS provided timely warnings that anticipated the rise of the first COVID-19 wave in Rio de Janeiro. The unusually high and low case numbers during the COVID-19 pandemic precluded us from evaluating the EWS capabilities for the period from October 3, 2021, to April 30, 2022 (epidemiological weeks 40-2021 to 17-2022).
Similar to previous studies, we found that using data-based syndromic surveillance offers useful information for early outbreak detection4,9,14. A comprehensive review, including 68 scientific publications, found that 42 of those studies concluded that data-based EWS successfully functioned independently as surveillance systems, while 16 reported that EWS contributed to the existing surveillance systems14. A report on the establishment of a syndromic surveillance system in Liberia during the COVID-19 pandemic concluded that the existing infrastructure of health data collection can be leveraged to monitor a variety of diseases with pandemic potential4.
When discussing the establishment of innovative EWS, it is important to address the needs of low and middle-income countries (LMIC). In this context, relying on an automated, administrative data-based EWS, such as that used in this study10, offers a cost-efficient approach to overcoming infrastructure limitations4,9. Moreover, incorporating PHC data into EWS promotes significant advantages for surveillance, such as detecting unusual symptom patterns before a surge in severe cases. This enables the timely deployment of response measures, potentially averting health system overloads14. Moreover, the PHC system in Brazil also offers great granularity, reaching underserved populations, even where more advanced healthcare facilities are lacking4.
Our study presents potential limitations. The EWS presented here relies on the continuous availability of data; therefore, the existence of areas or periods with substantial data gaps or fluctuations in data quality over time could impact the system’s performance4,15. Additionally, health-seeking behavior and data collection practices may suffer significant changes due to unforeseen external factors, such as the lockdowns enacted during COVID-19, or natural disasters and holiday periods, all of which will affect the stability of the baseline, thus hindering an accurate estimation of the EWS.
Conclusion
The primary healthcare data-based early warning system anticipates outbreak detection when compared to traditional epidemiological surveillance. These findings support the benefits of leveraging administrative health data to enhance surveillance in low-resource settings through a cost-efficient approach. This, in turns, allows for a more rapid response and control measures, ultimately enhancing the resilience of health systems and enhancing preparedness for future threats.
Acknowledgements
This study is part of the Alert-Early System of Outbreaks with Pandemic Potential (ÆSOP, http://aesop.health), an initiative under development by Brazil’s Fundação Oswaldo Cruz (Fiocruz) and the Universidade Federal do Rio de Janeiro, and financially supported by Rockefeller Foundation’s Health Initiative. We wish to thank Kate T. de Souza and the Primary Healthcare Department of the Ministry of Health (SAPS-MoH) for data acquisition and transfer to Fiocruz. We would also like to thank Gislani Mateus Aguilar for kindly presenting and discussing the Epidemiological Intelligence Center (CIE) of Rio de Janeiro’s structure and datasets to the AESOP team.
References
-
1 World Health Organization (WHO). COVID 19 Strategic Preparedness and Response Plan. Operational Planning Guidelines to Support Country Preparedness and Response [Internet]. 2020 [cited 2024 out 31]. Available from: https://www.who.int/publications/i/item/draft-operational-planning-guidance-for-un-country-teams
» https://www.who.int/publications/i/item/draft-operational-planning-guidance-for-un-country-teams - 2 Baker RE, Mahmud AS, Miller IF, Rajeev M, Rasambainarivo F, Rice BL, Takahashi S, Tatem AJ, Wagner CE, Wang LF, Wesolowski A, Metcalf CJE. Infectious disease in an era of global change. Nat Rev Microbiol 2022; 20:193-205.
- 3 Richards CL, Iademarco MF, Atkinson D, Pinner RW, Yoon P, Mac Kenzie WR, Lee B, Qualters JR, Frieden TR. Advances in Public Health Surveillance and Information Dissemination at the Centers for Disease Control and Prevention. Public Health Rep 2017; 132(4):403-410.
- 4 Fulcher IR, Boley EJ, Gopaluni A, Varney PF, Barnhart DA, Kulikowski N, Mugunga J-C, Murray M, Law MR, Hedt-Gauthier B, the Cross-site COVID-19 Syndromic Surveillance Working Group. Syndromic surveillance using monthly aggregate health systems information data: methods with application to COVID-19 in Liberia. Int J Epidemiol 2021; 50(4):1091-1102.
- 5 Lewis AE, Weiskopf N, Abrams ZB, Foraker R, Lai AM, Payne PRO, Gupta A. Electronic health record data quality assessment and tools: a systematic review. J Am Med Inform Assoc 2023; 30:1730-1740.
-
6 European Centre for Disease Prevention and Control (CDC). Data quality monitoring and surveillance system evaluation - A handbook of methods and applications [Internet]. 2014 [cited 2024 out 31]. Available from: https://data.europa.eu/doi/10.2900/35329
» https://data.europa.eu/doi/10.2900/35329 - 7 Cruz DMO, Ferreira CD, Carvalho LF, Saraceni V, Durovni B, Cruz OG, Garcia MHO, Aguilar GMO. Inteligência epidemiológica, investimento em tecnologias da informação e as novas perspectivas para o uso de dados na vigilância em saúde. Cad. Saude Publica 2024; 40(8):e00160523.
- 8 Teixeira MG, Penna GO, Risi JB, Penna ML, Alvim MF, Moraes JC, Luna E. Seleção das doenças de notificação compulsória: critérios e recomendações para as três esferas de governo. Inf Epidemiol Sus 1998; 7(1):7-28.
- 9 Bagaria J, Jansen T, Marques DF, Hooiveld M, McMenamin J, Lusignan S, Vilcu AM, Meijer A, Rodrigues AP, Brytting M, Mazagatos C, Cogdale J, Werf S, Dijkstra F, Guiomar R, Enkirch T, Valenciano M; I-MOVE-COVID-19 study team. Rapidly adapting primary care sentinel surveillance across seven countries in Europe for COVID-19 in the first half of 2020: strengths, challenges, and lessons learned. Eurosurveillance 2022; 27(26):2100864.
- 10 Ramos PIP, Marcilio I, Bento AI, Penna GO, Oliveira JF, Khouri R, Andrade RFS, Carreiro RP, Oliveira VA, Galvão LAC, Landau L, Barreto ML, van der Horst K, Barral-Neto M, AESOP Collaborating Teams. Combining Digital and Molecular Approaches Using Health and Alternate Data Sources in a Next-Generation Surveillance System for Anticipating Outbreaks of Pandemic Potential. JMIR Public Health Surveill 2024; 10:e47673.
- 11 Souza WM, Buss LF, Candido DDS, Carrera JP, Li S, Zarebski AE, Pereira RHM, Prete CA Jr, Souza-Santos AA, Parag KV, Belotti MCTD, Vincenti-Gonzalez MF, Messina J, Sales FCS, Andrade PDS, Nascimento VH, Ghilardi F, Abade L, Gutierrez B, Kraemer MUG, Braga CKV, Aguiar RS, Alexander N, Mayaud P, Brady OJ, Marcilio I, Gouveia N, Li G, Tami A, Oliveira SB, Porto VBG, Ganem F, Almeida WAF, Fantinato FFST, Macário EM, Oliveira WK, Nogueira ML, Pybus OG, Wu CH, Croda J, Sabino EC, Faria NR. Epidemiological and clinical characteristics of the COVID-19 epidemic in Brazil. Nat Hum Behav 2020; 4(8):856-865.
- 12 Nott R, Fuller TL, Brasil P, Nielsen-Saines K. Out-of-Season Influenza during a COVID-19 Void in the State of Rio de Janeiro, Brazil: Temperature Matters. Vaccines 2022; 10(5):821.
- 13 Meyer S, Held L, Höhle M. Spatio-Temporal Analysis of Epidemic Phenomena Using the R Package surveillance. J Stat Software 2017; 77(11):1-55.
- 14 Meckawy R, Stuckler D, Mehta A, Al-Ahdal T, Doebbeling BN. Effectiveness of early warning systems in the detection of infectious diseases outbreaks: a systematic review. BMC Public Health 2022; 22(1):2216.
- 15 Florentino P, Bertoldo-Junior J, Barbosa GCG, Cerqueira-Silva T, Oliveira VA, Garcia MHO, Penna GO, Boaventura V, Ramos PIP, Barral-Neto M, Marcilio I. Impact of Primary Health Care data quality on their use for infectious disease surveillance. JMIR Public Health Surveill 2025; 11:e67050.
-
Funding
This study is financially supported by the Rockefeller Foundation’s Health Initiative (Grant 2023-PPI-007 awarded to M Barral-Netto). M Barral-Netto, PIP Ramos, and V Boaventura are Research Fellows from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq, Brazil). The funding institutions did not interfere in the analysis, interpretation, or decision to submit the manuscript for publication.
Source: Authors.
Source: Authors.