DLSU faculty build model to simulate optimal allocation of COVID-19 antivirals

Developing an effective treatment against COVID-19 is one challenge, but ensuring access to these medical services is another altogether.

The latter problem is what a team of DLSU faculty and researchers—led by Dr. Charlle Sy from the Industrial Engineering Department and Dr. Kathleen Aviso from the Chemical Engineering Department—has come together to address. By building a computational model, they can show possible scenarios and determine how best to allocate COVID-19 antiviral drugs.

Sy discussed their project in “How to distribute COVID-19 Drugs?”, a webinar hosted by the DLSU Data Science Institute (DSI) and Animo Labs Foundation Inc. last October 22 via Zoom and Facebook Live. Other team members then joined in a panel discussion following Sy’s presentation, with Department of Science and Technology Undersecretary Dr. Rowena Guevarra also in attendance.

Multifaceted issue

In the healthcare sector’s efforts to mitigate the detrimental impacts of COVID-19 on infected individuals, several “readily available” medications were repurposed to alleviate the effects of the disease. “But even then these drugs are considered limited resources…Even if we want to accommodate all patients, there is simply not enough supply to go around,” prefaced Sy. “Do we give [these drugs] to the most vulnerable? Or to the ones with the highest chance of surviving disease?”

To answer such questions, complex decisions must be made which account for the severity of conditions, risk profiles or vulnerability to infection, availability of health services, and even ethical facets. The team’s model was thus framed as an “analytical approach to address this [issue] in a more systematic manner” in consideration of several factors, according to Sy.

Input parameters included the severity of the cases, from mild, moderate to critical status; the required treatment approach, such as admittance to an intensive care unit (ICU) for critical cases and stay-at-home measures for mild cases; fatality rates per severity level; access to proper care and the availability of hospital beds in different areas; and the hypothetical effectiveness of antiviral drugs.

Several assumptions, Sy noted, were also necessary for the model to be useful. “We assumed that if you’re part of the critical cases, then you need [to go to] the ICU facility [to] get intubated. But [for] moderate [cases], probably a normal hospital bed will suffice,” she explained. Further, patients given different antivirals were assumed, based on the efficacy of these medications, to have a corresponding “chance to get downgraded from critical to moderate or mild [status].”

With all these parameters, the linear programming allocation model utilized several mathematical functions to test and represent various situations including: patient access to proper care, such as being admitted at a medical center versus isolating at home; types of treatment services received by infected individuals; allocation of antiviral drugs according to severity; and mortality rates differentiated according to quality of treatment.

Validating the solution

Applying the model to case studies, the team utilized data published by the Department of Health and worked within the limits of what was known about COVID-19 back in April. “No safe and effective antivirals were known yet, so the efficacy rates [used] are fictitious,” Sy said.

Three scenarios were then considered. The baseline, with no antiviral drugs administered, resulted in a 5.5 percent fatality rate and an estimated 80 out of 1,000 people “without access to needed care.” Sy added, “There [would be] a chance of overwhelming the healthcare system.”

If two antiviral drugs are to be used interchangeably—that is,  without regard to severity status and efficacy of the medications—case mortality rates were simulated to drop to 1.3 percent. “Now everyone is able to access needed care because you free up the resources of the hospital because [patients given antivirals] may get downgraded [to less severe levels],” Sy said, citing that all 1,000 patients were provided adequate medical services based on the model’s predictions.

The situation was shown to improve if antivirals are to be allocated according to priority, with higher efficacy antivirals intended particularly for critical cases. Mortalities were reduced further to a 0.9 percent fatality rate, demonstrating the importance of optimizing the distribution of COVID-19 drugs.

When applied to a larger scale, and with data coming from multiple cities in the metro, the team’s model showed that “even a 10 percent increase in availability will significantly decrease the number of deaths,” Sy described. As the availability of resources continued to rise, the effects were not as distinct, though marginal reductions in fatality rates were still observed.

Nevertheless, Sy highlighted that prioritizing and providing severe patients with proper treatment enabled “freeing up hospital resources” which could then be allotted for other patients in need, thereby avoiding overwhelming the healthcare system.

Transitioning to mainstream use

While the model so far only serves as a “proof of concept” using hypothetical data, Vice Chancellor for Research and Innovation Dr. Raymond Tan emphasized the value of pursuing such endeavors, “We should be one step ahead…to have the model ready, rather than rush a model when the data [arrives].”

To expand the scale and complexity of the project, another model is also in the works, this time for vaccine distribution. Also being eyed are “simulation models for what-if situations”, according to Animo Labs Executive Director Dr. Frederico Gonzalez. These additions could help account for social determinants such as policy decisions, stigma, and vaccine hesitancy.

The team ultimately envisions to provide a tool that can inform policies and prioritization at the level of government agencies and other organizations, as well as individual hospitals.

“We will look at options on how to make it accessible,” Aviso explained on translating the program to other platforms such as Excel as well as for clinician or public use. “Maybe [we can] have workshops and training programs once these [applications] have been developed in a user-friendly interface.”

Touting the project as an effort to contribute to nation-building, DSI Dean Dr. Macario Cordel closed the session by affirming, “[An effective model is] highly dependent on [the] validity of [the] data, and [we must] regularly update the model to ensure the outputs are valid and relevant to [what is happening] on the ground.”

By Erinne Ong

Leave a Reply