Public Health and Epidemiology Informatics: Recent Research Trends Moving toward Public Health Data Science.

OBJECTIVES: To introduce and summarize current research in the field of Public Health and Epidemiology Informatics.
METHODS: PubMed searches of 2019 literature concerning public health and epidemiology informatics were conducted and the returned references were reviewed by the two section editors to select 14 candidate best papers. These papers were then peer-reviewed by external reviewers to allow the Editorial Committee a curated selection of the best papers.
RESULTS: Among the 835 references retrieved from PubMed, two were finally selected as best papers. The first best paper leverages satellite images and deep learning to identify remote rural communities in low-income countries; the second paper describes the development of a worldwide human disease surveillance system based on near real-time news data from the GDELT project. Internet data and electronic health records are still widely used to detect and monitor disease activity. Identifying and targeting specific audiences for public health interventions is a growing subject of interest.
CONCLUSIONS: The ever-increasing amount of data available offers endless opportunities to develop methods and tools that could assist public health surveillance and intervention belonging to the growing field of public health Data Science. The transition from proofs of concept to real world applications and adoption by health authorities remains a difficult leap to make. Georg Thieme Verlag KG Stuttgart.

Introduction

The increasing digitization of health data and the recent advances in several fields of computer science such as natural language processing and deep learning offer more opportunities for applications in the domain of public health and epidemiology.

Data generated on the Internet can be used to measure the prevalence and incidence of diseases and allows the development of real-time applications to serve the early detection of epidemics 1 . Although easy and cheap to access, Internet data is often noisy and extracting good quality data for decision makers is often very challenging and requires strong and multidisciplinary expertise.

Harder to access, electronic health records (EHRs) and clinical data registries contain very high quality data generated by health professionals. Several international initiatives like the Observational Health Data Sciences and Informatics (OHDSI, https://ohdsi.org) aim at facilitating the interoperability and the exploitation of clinical data while guaranteeing data protection and ownership. Recently, the feasibility of building a cohort of hundreds of millions patients across the globe has been demonstrated 2 and the activity opens up new research perspectives at the global scale.

A promising technology for public health is the increasing use of mobile phones. The surge of computing power and the ubiquity of mobile phones around the globe make it possible for large populations to participate in public health surveillance and prevention campaigns 3 . Further research is expected to fully leverage this technology for the benefit of public health.

This synopsis looks at the literature published in 2019 in the domain of medical informatics applied to public health and epidemiology. The aim is to identify new topics and trends as compared to previous years and describe the selection process of the best papers published in 2019 based on quality and originality of articles.

Methods

A comprehensive literature search was performed using PubMed/Medline database from NCBI, National Center for Biotechnology Information. Using a large set of MeSH descriptors, the queries targeted public health or epidemiological journal articles over the year 2019 that included medical informatics topics. Returned references addressing topics of the other sections of the Yearbook, e.g ., those related to sensors, were excluded from our search. The study was performed at the beginning of January 2020, and the search returned a total of 835 references.

Articles were separately reviewed by the two section editors and were first classified into three categories: keep, discard, or leave pending with the BibReview tool 4 . Then, the two lists of references were merged yielding 90 references that were retained by at least one reviewer or classified as “pending” by both of them. The two section editors jointly reviewed the 90 references and selected a consensual list of 15 candidate best papers. Two candidate best papers were removed from the list because the papers were selected by other sections and one paper was drafted to obtain a final list of 14 candidate best papers. All of these papers were then peer-reviewed by editors and external reviewers. Each paper was reviewed by at least four reviewers. Two papers were finally selected as best papers by the Yearbook Editorial Committee ( Table 1 ). A content summary of these selected papers can be found in the appendix of this synopsis.

Table 1

Best paper selection of articles for the IMIA Yearbook of Medical Informatics 2020 in the section ‘Public Health and Epidemiology Informatics’. The articles are listed in alphabetical order of the first author’s surname.

SectionPublic Health and Epidemiology Informatics

▪ Bruzelius E, Le M, Kenny A, Downey J, Danieletto M, Baum A, Doupe P, Silva B, Landrigan PJ, Singh P. Satellite images and machine learning can identify remote communities to facilitate access to health services. J Am Med Inform Assoc 2019;26(8-9):806-12. ▪ Feldman J, Thomas-Bachli A, Forsyth J, Patel ZH, Khan K. Development of a global infectious disease activity database using natural language processing, machine learning, and human expertise. J Am Med Inform Assoc 2019;26(11):1355-9.

Results

The trend towards the increase in the number of publications in infodemiology noticed in 2018 5 continued in 2019 with new emerging use cases like the monitoring of physical activity using Twitter Data 6 , the surveillance of plague outbreak with Google Trends 7 , the identification of patients with diabetes or the detection of conjunctivitis epidemics worldwide based on search engine queries 8 9 . One selected best paper describes the development of a global infectious disease database using natural language processing, machine learning, and human expertise 10 . The original idea of this paper was to exploit the publicly available data of the GDELT project that monitors in near real time the world’s broadcast, print, and web news. The system developed was capable of analyzing news in 65 languages to early detect onset of epidemics worldwide.

Disease surveillance systems based on social media and search queries aim to measure current disease activity, aka nowcasting , but are still prone to errors due to the imperfect features of the models they rely upon. Priedhorsky et al. , 11 proposed the metric of deceptiveness which quantifies the noise in the features of a model. This metric could improve in the future the measurement of disease prevalence and incidence. In order to be adopted by health authorities, there is much room for further research to improve the performance of these statistical models as accurate and reliable estimations of disease activity based on Internet data.

EHRs are still a source of high quality information for public health researchers. Post marketing drug surveillance 12 13 and healthcare-associated outbreaks detection 14 15 continue to be hot topics of research.

Also, as quoted by the survey paper of the Public Health and Epidemiology Informatics section of the 2020 International Medical Informatics Association (IMIA) Yearbook 16 , the targeting of sub-populations for dedicated public health interventions is a growing subject of interest. Digital segmentation aims to reach audiences using digital technologies offering new opportunities to deliver appropriate prevention messages 17 . Several studies have already shown the interest of social media for public health campaigns such as smoking cessation 18 . A way to maximize their impact and efficiency could be to identify and target specific audiences. To do so, natural language processing, data mining, and machine learning have been used to classify user traits 19 . The strategy is similar to that of online targeted advertising except that the goal is to deliver dedicated public health, rather than advertising, messages. The second selected best paper applied deep learning on satellite images to identify rural and hard-to-reach remote communities in low-income countries and help community health workers deliver health services 20 . The geographical segmentation of population based on their access to healthcare is needed to organize specific healthcare delivery and to reduce inequalities.

Despite the obvious need of these new approaches, the use of phenotyping algorithms to classify individuals raises ethical issues about data privacy, confidentiality, and informed consent. Interestingly, these issues are rarely discussed when information is retrieved from social media, unlike from EHRs where an institutional review board authorization is often mandatory to carry out such analyses. A consensus has yet to emerge to handle Internet data 21 . In the meantime, public health researchers must do their utmost to protect user data and to keep confidential the models of individual prediction.

Conclusion

The huge amount of data available from Internet, from EHRs, and upcoming from mobile phones, is the fuel for a lot of research on different topics covering statistics, informatics, and epidemiology defining public health Data Science.