3 launch points to get your data modeling project started

Health data analytics holds great promise in helping care facilities prepare for what lies ahead. Hospitals and health systems using COVID-related data models, for example, gain capacity-planning insight into expected case counts, bed and ICU availability, timing of patient surges and medical supply needs.

Similarly, vulnerability maps use social and economic population variables (e.g., density, travel distance to treatment centers, underlying chronic illnesses and economic status) to identify at-risk communities. From a broader perspective, case mapping and projection tools visually depict disease incidence at national, state or county levels.

Here’s a sampling of COVID-specific data-modeling projects in use across the country:

Harvard University’s Regional Hospital Capacity Calculator outputs “heat maps” for Hospital Referral Regions (HRRs) showing projected occupancy levels under segmented infection rates for time periods of 6, 12 and 18 months.

Rush University’s Hospital Resource Calculator allows users to choose from variable growth models to predict the number of cases up to 60 days in the future. It also provides prior predictions from the 10 most recent days so that users can compare projections to actual case counts.

Penn Medicine’s Hospital Impact Model for Epidemics projects the daily number of new COVID hospital admissions, along with a census of current COVID patients by cohort (e.g., hospitalized, in ICU, on ventilator).

The American Hospital Association’s Bed Occupancy Projection Tool enables users to select a future date and the percentage of hospital or ICU beds already occupied by non-COVID patients to forecast when bed demand will exceed capacity. A regional mapping dashboard displays adult bed-capacity measures at state, HRR or Hospital Service Area levels, and can be filtered by population age groups, poverty rates and insured status.

Setting the stage for innovation

Beyond COVID applications, many healthcare decision-makers are looking closely at emerging technology for predictive analytics with particular interest in artificial intelligence (AI) for early detection of illness and health conditions.

IT executives looking for project launch points should consider the following perspectives:

• Pursue preventative medicine. Author and futurist Bernard Marr advocates for initiatives that inhibit contagious disease outbreaks, but also recommends efforts where lifestyle factors like diet, exercise and environment likely lead to health issues in different populations or geographical areas. “AI makes it possible to create tools that can spot patterns across huge datasets far more effectively that traditional analytics processes, leading to more accurate predictions and ultimately better patient outcomes,” explains Marr.

• Build on existing models. The CDC’s Center for Forecasting and Outbreak Analytics (CFA) shows where federal funds are flowing — $26 million allocated to development of next-generation systems with expanded capabilities for data sharing and integration. Focus should be on maximizing interoperability with data standards and utilizing open-source software and application programming interface capabilities to merge existing and new data streams, according to CFA.

• Work with diverse data sources. “The fundamental limit to forecasting is human behavior,” notes Marc Lipsitch, CFA’s science director. “Small sets of data from very well studied populations are more valuable in some ways than larger amounts of data [where] you don’t know as much about the people.” Hospitalizations and case counts are almost always important, he adds, but non-traditional sources of data such as mobility or prescription/ non-prescription drug sales should also have key roles.

With incremental improvements in data flow and quality, predictive models will leverage computational skills to anticipate human response to disease spread. In the meantime, “the thing to do is start generating predictions now,” says Lipsitch, to set the stage for greater utility in forthcoming applications.

Frank Irving is a Philadelphia-based content writer and communications consultant with specialties in healthcare, technology and sports. When not following those beats, he writes creative fiction.