Forecasting the Next Outbreak: Making PHEOCs More Efficient Using Predictive Analytics

Imagine a small, primary care medical clinic near the border between two developing countries. A young woman, our archetypal client, Sara, has just arrived, complaining of trouble breathing. Immediately, she is carefully distanced from other patients, given a mask and hand sanitizer, screened for a fever, then tested for COVID-19—which comes back positive. As Sara is treated, her diagnosis is captured by a healthcare worker who reports the notifiable disease using the Digital Health Information System (DHIS2) app installed on her tablet. This data point is about to become part of a composite picture where private sector data collection is seamlessly shared with the government to save lives across the region.

Sara’s diagnosis flows – anonymously and in real-time – to a database housed within the National Health Management Information System (HIS) and monitored by the Public Health Emergency Operations Center, or PHEOC. There, it is analyzed alongside other test results from her region, local social media mentions of COVID-19 and its symptoms, World Bank and WHO population statistics, as well as weather and migration patterns. Thanks to a constantly improving algorithm, the PHEOC staff can predict that this outbreak is about to rapidly spread and that it presents a potential cross-border risk. The growth rate of positive tests, the social media mentions of fatigue regarding masking and social distancing, and the drought decreasing access to water for handwashing stations all together indicate the need for a rapid response. Ministry of Health (MOH) officials can now efficiently deploy resources including more tests, disinfectants, PPE, and behavioral guidance to this region—and warn their bordering neighbor to support them in protecting their collective regional population. Meanwhile, Sara is being interviewed for contact tracing purposes, and receiving the care she needs early enough in her infection cycle to save her life and protect others from ongoing community transmission.

Now imagine we can run this process for multiple diseases at once. This hypothetical chain of events is rapidly becoming possible in the Lao PDR, with Cambodia and Myanmar to follow, through efforts executed by PSI’s field programs and the Digital Health and Monitoring team alongside the PHEOCs we support.

Predictive analytics is hardly a new field. However, its application to public health in the developing world has only emerged over the past five years to better track notifiable diseases such as malaria. Its public health debut has lagged behind other industries like financial services and marketing, due to challenges both human and digital. To stand up PHEOCs where trained analysts can best apply this discipline on a national scale, PSI has expanded relationships with MOHs, private sector healthcare providers, and a wide range of technical partners (public and private). This helps to ensure that PHEOCs receive the local, regional and national, country-specific information needed for reactive decision making.

To collect this data, we have worked to increase digital infrastructure across health systems, and technology proficiency across providers and consumers. PSI has deployed tablets, apps, chatbots, digital signposting, and other tools into the hands of Sara and the health providers who serve her, then collected and collated their queries, concerns and responses. This informational ecosystem is now primed for a future where it could be leveraged alongside multi-source streams of structured, unstructured, historical and transactional data relevant to healthcare, to a) predict future outbreaks, and b) proactively plan for them accordingly.

PSI’s 50 years of health programs in over 50 countries has afforded us public and private health industry relationships, and an understanding of national to regional healthcare systems—across areas where a PHEOC could mean the difference between local containment or a cross-border outbreak. PSI’s long-term operational experience in developing healthcare markets has allowed us to produce and deploy at scale multiple digital tools that assess population health and needs and ultimately strengthen healthcare nationwide. A few examples:

  • DHIS2 mobile apps render paper files obsolete and potentially speed up electronic medical record (EMR) sharing between public and private health providers as well as national health systems.
  • Chatbots interact directly with Sara, helping her self-diagnose, and signposting her to clinics when needed. They also stream critical health information from providers into national networks. In early 2020, PSI successfully deployed chatbots in Facebook Messenger in Lao PDR, and in the Zalo messaging app in Vietnam. Both are currently used by doctors and pharmacists to report malaria cases. Another Facebook Messenger bot is under development in Myanmar and Lao PDR, initially designed for malaria but soon to be expanded to other notifiable diseases—meanwhile, Vietnam’s Zalobot is expanding to support reporting suspected COVID-19 cases seen in private clinics and pharmacies.
  • Social listening dashboards consolidate information from Facebook, WhatsApp and other social media, and are primed to capture mentions of and render sentiment analysis for health topics designated by the MOH. These have included menstruation, pregnancy, and preferred condom brands—the goal being to inform reactive, social behavioral change (SBC) campaigns around family planning in West Africa, for example, and PSI has deployed social listening tools in Angola, Guatemala, and Vietnam as well. Social listening also captures and identifies information that could indicate outbreaks; algorithms combined with machine learning comb searches and chats from individuals for mentions of symptoms. This information, coupled with real-time reporting through MOH systems and DHIS2 apps, will allow PSI-led projects and PHEOCs alike to react quickly to outbreaks and travel patterns, as well as anticipate hotspots.

In working to bring these tools to scale, we have increasingly witnessed when, where and how to structure the data such that the resulting predictive analytics could shift national health response planning from reactive to proactive. The goal is to build PHEOCs with integrated, multi-source surveillance data and analytics for proactive planning to address any eventual health crises. This would de-silo our response to any one disease, allowing PHEOCs to respond holistically to other related health threats at the same time, thus saving more lives. Imagine being able to detect, track and respond to malaria, dengue, Ebola, and COVID-19 outbreaks all at once.

Another win is being able to predict a disease-specific outbreak, e.g. measles, days or weeks before the formal health system picks it up, enabling the quick mobilization and deployment of teams to respond to the outbreak at hand. With malaria entering an elimination phase in many countries, every case that is detected and responded to in a timely fashion brings us closer to ending the disease. Meanwhile, the efficiencies realized by simultaneously deploying COVID-19 prevention measures alongside malaria medication in an area containing outbreaks of both, means that the resources saved in delivery costs could be applied to other crises.

This work is never done in isolation; we partner with smart, strategic experts who understand the possibilities that digital tools, and ultimately predictive analytics represent to public health. In order to strengthen the PHEOCs that stand and those that have yet to be built; in order to assure efficient allocation of often limited healthcare resources, we are constantly seeking partners who can help us build on our foundational experience and benefit from our global market penetration. We are well positioned to collaborate on proactive tools that listen to Sara, capture her health needs and concerns, and map them against a macrocosm of data that ultimately can inform a safer world that delivers her the care she needs—before she needs it.

If you or your organization are interested in working with PSI to support PHEOCs and build predictive analytics capabilities for national health systems, please contact Martin Dale at martind@psi.org–or for more information about our field operations contact Eric Seastedt at eseastedt@psi.org.

Sign up to
Receive Updates

Donate to
Support Our Work

Related