Dr. Lauren Neal, chief of analysis and consulting agency Booz Allen’s well being AI observe, is a proponent of taking a multi-method strategy to modeling COVID-19 illness dynamics in synthetic intelligence.
She believes a multi-method strategy gives a greater understanding of COVID-19 and different infectious illnesses – how they unfold and impression communities, with the aim of being higher ready for future public well being threats.
She additionally believes a “digital laboratory” can be utilized to analyze a variety of what-if situations, and simply tailored to future high-consequence public well being threats.
Healthcare IT Information sat down with Neal to speak about these approaches and the way AI will help with the COVID-19 pandemic.
Q. In terms of synthetic intelligence and COVID-19, how is a multi-method strategy to modeling COVID-19 illness dynamics higher than different approaches?
A. We now have lengthy employed simulation modeling to additional enhance our understanding of complicated infectious illnesses in addition to their growth, unfold dynamics and potential remedies. Examples embody fashions for zoonotic illnesses similar to Zika, Ebola, West Nile Virus, SARS, MERS and the current COVID-19.
Two modeling methods, system dynamics (SD) and agent-based modeling (ABM), have been steadily used in recent times to analyze the complicated nature of infectious illnesses regardless of their limitations. For instance, SD operates at a excessive degree of abstraction by compartmentalizing the human inhabitants into completely different illness levels similar to inclined (S), contaminated (I) and recovered (R), amongst others whereas assuming everybody behaves the identical manner inside every compartment.
ABMs have a tendency to deal with this limitation by monitoring every particular person member of the inhabitants and simulating granular profiles of particular person interactions and actions inside the inhabitants. Nonetheless, this excessive degree of mannequin constancy comes with a handful of trade-offs, together with intensive value of computation for giant populations in addition to elevated mannequin uncertainty on account of a myriad of mannequin assumptions.
We consider that successfully selecting between modelling strategies is a query of minimizing trade-offs within the mannequin creation, verification and validation course of. The concept of multi-method modeling is to combine completely different strategies of modeling to beat the constraints of particular person strategies and get essentially the most from each.
Booz Allen’s multi-method mannequin for COVID-19 combines the benefits of SD and ABM, permitting the simulation of spatially specific situations representing future states of illness transmission inside completely different native communities and testing threat administration insurance policies throughout a variety of situations utilizing “what-if” evaluation.
Q. What’s a digital laboratory and the way can it’s used to analyze public well being threats?
A. Traditionally, randomized management trials, cohort research and case-control research have been generally used strategies to analyze the epidemiology of public well being threats in addition to potential intervention choices to mitigate the dangers. Nonetheless, performing giant trials and research to realize generalizability and adequate statistical energy is sort of tough, time-consuming and dear.
Subsequently, a comparable, dependable and easy-to-use planning instrument is required to evaluate interventions and their impacts. A digital laboratory is a particular kind of simulation mannequin that can be utilized to characterize the dynamics of COVID-19 unfold inside a neighborhood and facilitate “what-if” simulations that explicitly characterize the uncertainty in supporting information and assumptions about threat components related to onset of the illness inside the neighborhood.
A digital laboratory is a risk-free setting, by which concepts on intervention methods for a selected public well being risk (for instance, social distancing, partial lockdown and vaccination, amongst others) will be examined in a scientific method with out the time, prices and dangers related to experiments carried out in a real-world setting.
Digital laboratories can have many makes use of, and current many prospects for innovation, however it’s their functionality to supply real-time perception, allow forecasting and supply determination assist for dwell operations that’s most instantly accessible. With these skills, neighborhood, state and federal public well being decision-makers will be simpler, enhance effectivity and ship value financial savings whereas defending lives.
Q. What’s a multi-criteria determination evaluation (MCDA) framework, and the way is it used with synthetic intelligence and COVID-19?
A. Resolution-making relating to implementation of public well being interventions can generally be heuristic, and it may be argued that selections based mostly on a single criterion disregard essential details about different related associated outcomes. In managing the COVID-19 pandemic, a number of compelling narratives appear to have performed a major position within the decision-making processes relating to which threat intervention and administration measures to implement.
Throughout the pandemic, public authorities have needed to make selections based mostly on unsure quantitative proof and professional scientific proof (for instance, potential future situations), on assessments of the well being system capability (for instance, ICU beds)Â and on anticipated public adoption of roughly restrictive measures similar to social distancing and lockdown measures in addition to reopening of native communities and companies.
When empowered by real-time information harnessed utilizing synthetic intelligence and machine studying methods, in addition to forecasted illness dynamics based mostly on simulation modeling, multi-criteria determination evaluation (MCDA) will help decision-makers make data-driven selections based mostly on a number of, generally conflicting standards in a clear and systematic method.
For instance, Booz Allen has used an MCDA framework contemplating native determination standards similar to new each day infections, decline in new each day deaths, new hospitalizations and ICU mattress utilizations to systematically analyze simulated forecasts obtained from our multi-method mannequin and generate threat maps for particular person states.
These threat maps may probably be utilized by public well being decision-makers to focus on out there surveillance and an infection management measures based mostly on the perceived ranges of COVID-19 dangers in native communities.
Q. How does all of this apply to the work of healthcare supplier organizations’ C-suite executives and caregivers on the frontlines of the pandemic?
A. The COVID-19 pandemic has introduced us unprecedented and evolving challenges since its onset. We now have made appreciable efforts to deal with these challenges utilizing a set of data-driven instruments, together with synthetic intelligence and simulation modeling.
Whereas early efforts have been centered on epidemiological modeling of COVID-19 unfold at world, nationwide and state ranges, the pandemic has raised many extra localized challenges that our data-driven approaches may also tackle.
For instance, the fast onset of the COVID-19 disaster has proven elevated demand and dangers for healthcare supplier organizations on account of repeatedly altering and unpredictable circumstances. Simulation modeling and digital laboratories will be utilized to proactively handle threat to healthcare organizations through the present pandemic and future large-scale public well being threats.
We are able to examine a variety of situations to reinforce our preparedness by optimizing hospital workflow buildings, creating new processes, managing staffing ranges, procuring tools, mattress administration, and implementing consistency of medical administration of sufferers, amongst others.
In these methods, a digital laboratory can be utilized as each a studying instrument (for instance, higher understanding how a hospital in addition to frontline healthcare suppliers operate below a area people COVID-19 outbreak) and an analysis instrument (for instance, testing complicated situations like optimum affected person throughput for an emergency division).
Digital laboratories can successfully assist executive-level selections made on the healthcare supplier organizational degree to create capability and handle scarce assets for the efficient care of critically unwell sufferers, whereas testing situations to guage the flexibility of the well being system capability to deal with anticipated and sudden calls for through the pandemic.
Twitter:Â @SiwickiHealthIT
Electronic mail the author:Â bsiwicki@himss.org
Healthcare IT Information is a HIMSS Media publication.
