The collection, analysis and reporting of traditional public health data has been used for years to track infectious diseases and other public health threats, like H1N1 influenza, SARS and Ebola. Now we have the tools to track and stop COVID-19, thanks to big data sentinel sources.
Mark Lambrecht, PhD, Director of the Global Health and Life Sciences Practice at analytics software leader SAS, takes us through how this data can be leveraged during the crisis.
“For decades, SAS has provided analytics software to public health and government agencies around the world, helping them improve the health and well-being of their citizens,” says Lambrecht. “Algorithms and models that help prevent epidemics have been around for many years as well. They are used by research teams at universities, private companies and government institutions.
“While advanced analytics techniques have been deployed in hospitals and within public health organizations, the difference in the benefits achieved is often related to the maturity of the e-health system, the readiness of the health care system to execute decisions and measure the outcomes, and the ability to gather high quality digital information.”
With this varied level in preparedness between the health organisations, hitting as close to the bullseye as possible is vital.
“The effectiveness of a predictive model (hitting as close to the bullseye as possible) increases if the algorithm has access to official statistics from the first responders or hospitals. Artificial intelligence algorithms do not work autonomously – they are supported by entire teams of doctors, epidemiologists and other scientists who are constantly working on data using the help of computer programs.
“To increase accuracy and precision, diverse information sources are combined into analytical data sets, e.g. official incidence records, clinical emergency data, physician’s records, social media, flight records, school absence, and sales data of anti-fever medication. Advanced analytics can help to complement the clinical findings, specific adverse events, and model characteristics of a new viral epidemic or pandemic such as COVID-19. These early findings are crucial to ready the health care system and ensure the right capacity to put patients in quarantine or have enough antiviral medicines and materials ready.
Although getting it right first time is the goal, COVID-19 is a new virus with new nuances in its data modelling, says Lamprecht.
“Because of the dynamic nature of disease spread, particularly for new, previously unseen viruses, as well as the unknown impacts of potential future government and public health interventions, complete precision in epidemic modeling is usually impossible. There is always uncertainty.
“The goal of any epidemiological model isn’t necessarily to get the predictions exactly right, but rather to help provide insights about the epidemic that can facilitate effective, rapid decision making for public health officials, hospital operations, and policymakers. When evaluating the utility of a good model, it is important to use great care when assessing predicted future spread based on historical information.
“Some of the most advanced computational methods, applied by some of the smartest scientists in the world, still get these predictions wrong – often by large margins. The most useful models are those that instead highlight areas where there may be known weaknesses or vulnerabilities that could illuminate opportunities for action.
“For example, regardless of the specific accuracy of predicted case counts, we could use travel patterns and demographic data to highlight areas that could be at increased risk of a new introduced infection. Identifying strong physical connectedness between population centers can help understand where the next center of the outbreak might happen. We can then increase public health surveillance and other interventions in those areas, deploying resources optimally.
Lamprecht says the world is in a great position to be leveraging these data sources, compared to previous pandemics.
“When SARS emerged in 2002, there were fewer data sources that could be leveraged, such as social media, Internet of Things (IoT) devices and technologies to help with diagnostics. Phone apps for tracking of health data and diagnostics were not yet present because smartphones weren’t as connected and widely accessible as they are today. With the advent of the iPhone and new types of apps and technologies, scientists can leverage a lot more data for analysis in addition to the available sentinel sources.”
Regardless of model, developing a data culture is crucial to being prepared for the next public health crisis.
“Rather than looking for specific and precise predictions, models that seek to reveal where effective interventions can be taken are helpful in saving lives,” says Lamprecht. “Expertise into these techniques does not come overnight – it requires knowledge, organized data streams and a proactive attitude that deliver great value when a pandemic such as COVID-19 strikes.
“When starting to implement analytics technology into a health care organization, it is essential that an analytics culture needs to be established and impactful use cases need to be identified. This requires considerable investment - a true data-driven analytics strategy supported by hospital management over years.”