23. Data Science in Global Health—Highlighting the Burdens of Human Papillomavirus and Cervical Cancer in the MENA Region Using Open Source Data and Spatial Analysis

We sourced data on the regional burden of cervical cancer from the International Agency for Research on Cancer (IARC) Global Cancer Observatory (GCO) estimates of incidence, mortality and prevalence for the year 2018 in 185 countries or territories for 36 cancer types by sex and age group. After collecting annual cervical cancer incidence and mortality rates from each country in the MENA region, as defined by UNAIDS, from this database, as well as projections for each of these data points through 2040, we calculated regional totals to use in our data visualisation. We also used country reports put together by the HPV Information Centre and news coverage to determine which countries across the region have implemented the HPV vaccine.

In order to combine the different data sources, we used the international country code ISO 3166-1 alpha-3 (ISO-3) developed by the International Standard Organization. In case a data source did not include the ISO-3 code, we used the R-Package ‘countrycode’ (version 1.1.0) which converts the country’s name to ISO-3. We relied on the countrycode package to reduce human errors and ensure that the process of generating the dataset is automated. This enables reproducibility and replicability of our data pipeline.

To develop the GIS, we used Carto technology (https://​carto.​com/​). Carto is a cloud-based platform used to build powerful “Location Intelligence” applications. It includes a data repository that converts ISO-3 codes to spatial data and generates a spatial data representation on an interactive map. Carto GIS interface is customisable and includes several features such as multi-layered mapping, dynamic statistics and filtering.

After we developed the interactive map, the second phase of the project was the development of a regional page and country profile web pages, each with three main sections, to provide a narrative and additional data points for viewers. The first section of each of these pages includes a “dashboard” which highlights the current status of several key performance indicators (e.g. new cases of cervical cancer in 2018, deaths from cervical cancer in 2018 (total), the status of the national HPV vaccination programme). These data points came from the IARC GCO estimates and the HPV Information Centre’s database. The second section of each page provides a short description of the current epidemiological challenges facing that country in terms of the burdens of HPV and cervical cancer. The third section lists the peer-reviewed publications and news articles that relate to the context of these preventable diseases in each specific country, touching on current levels of awareness, effective interventions and actions at the local and national levels and future approaches to prevention.

Data points on HPV prevalence, as well as the lists of peer-reviewed publications from across the region, were collected through a thorough internet search of recent academic journal articles discussing “HPV” and/or “cervical cancer” in each country. Bringing together over 250 academic publications and data sources, the regional and country profile pages serve as a useful repository of information and insight into the current landscape of disease burden, infection and awareness in the MENA region.

We sourced relevant news articles on each regional and country profile page from internet searches for pieces covering “HPV” and/or “cervical cancer” in Arabic, English and French. Given the large quantity of media coverage, we have included the latest and most relevant pieces to each country and the region’s HPV and cervical cancer prevention efforts. An important point to consider is that although the inclusion of news articles could provide a more up-to-date view of current realities, the validation of the content in the news articles may be challenging especially with the rise of false or misleading news. While, we understand these limitations, we identified a process of quasi-peer-review as a workable solution. In this quasi-peer-review process, two independent members of the group validated the content of the articles. In a later stage, we consider including a time-dependent factor and excluding news articles that were not published within the last year. This process is time consuming, and other alternative such as crowd-sourcing could be potentially explored.

The dashboards and profile pages were developed by the engineering team at the Tunisian Center for Public Health and uses a Django framework, which can be used to develop websites and web applications and uses the Python programming language. The data enrichment pipeline is based on R programming language.

Our core challenges in developing the interactive map and regional and country profile pages centered on the lack of national-level data on HPV prevalence and the relative prevalence of each HPV strain. Academic studies on community and city-level burdens of HPV infection were incorporated into our platform, but because there are few national-level data points available, any analysis using the interactive map and regional and country profile pages is limited to what little is known about HPV prevalence. In the next iteration of the map, we may explore integrating data from the Institute for Health Metrics and Evaluation on Disability Adjusted Life Years due to cervical cancer in the MENA region. Similarly, few academic studies on the cost-effectiveness of and potential pathways to the rollout of nationwide HPV vaccination and cervical cancer screening programs exist. Future research in this area is needed to elucidate the context of HPV and cervical cancer prevention at the country level and inform policy. With additional data and analysis on options for the way forward for HPV vaccination and cervical cancer prevention, the map and profile pages could provide a much more complete picture for policymakers and advocates who might use it.

Another key challenge was the fact that the data we highlight on annual and projected country-level cervical cancer incidence and mortality rates come from a repository of estimates, rather than concrete datapoints—the IARC GCO database. On the GCO online platform, the authors highlight that the data points they present “are the best available for each country worldwide. However, caution must be exercised when interpreting the data, recognizing the current limitations in the quality and coverage of cancer data, particularly in low-and middle-income countries.” Our interactive map and profile pages are therefore limited and should be used carefully given the possibility that the IARC GCO estimates of cervical cancer incidence and mortality in countries in the MENA region could be inaccurate to the reality on the ground.

A few more limitations do exist in the current iteration of the interactive map and regional and country profile pages. The process for data validation and quality control is difficult and not totally transparent. Moreover, while the application is online, policymakers and researchers might lack the technical expertise to analyse the data, navigate the GIS system and derive accurate information. An effort to educate, train and support users continuously is important. Therefore, an important feature to include in the upcoming release is the development of an online help center and a repository of training resources. The map and profile pages are also intended to be a platform, rather than a definitive authority on HPV and cervical cancer in the MENA region. Into the future, this platform will grow and our team will continue to collect relevant data points from international and national sources, academic articles and news pieces.