A research project headed by St. Clair College data analytics professor John Ulakovich shows promising results in accurately forecasting the number of COVID-19 cases in Windsor-Essex.
“The forecasting is very close,” Ulakovich said. “We will be using this work to help healthcare providers and decision makers better understand the past, current and future COVID-19 epidemiological trends and resource requirements.”
For the last few months, Ulakovich and his student researcher, Tejas Deepak Shinde, have been collecting data and creating analytical dashboards that display various epidemiological metrics.
Ulakovich said the forecasting seems to be very accurate for the prototypes that have been developed. For instance, on Tuesday Shinde created a chart that estimated the number of confirmed cases to be 2,273. On Wednesday, the Windsor-Essex County Health United reported 2,275 positive cases.
“We’re testing the model. We’re trying to gauge it. We’re trying to predict what the trend is going to be one week or two weeks into the future,” Ulakovich said.
“For example, based on historical trends we can create a chart that can forecast the number of positive cases the next 12 days,” Ulakovich said. “This can be a valuable resource because if the number of positive cases begins to climb, we know that a certain percentage will require hospitalization
The forecasting is based on the actual daily rate over the previous 10 to 14 days. So, for instance, after Windsor-Essex moved into Phase 2, the number of cases began to rise after a couple of weeks. Ulakovich said it stands to reason that there might be another jump in that trend when the area moves to Phase 3. “So, what we're doing is looking at the estimated values that we're getting and comparing the actual daily rate to see how accurate our forecasting is…It’s fairly close now.”
The data that’s being collected can be used to predict other factors, such as the rate of hospitalizations . The overall patient numbers can then be used to forecast essential resources and supply needs like ICU beds, staffing, biomedical equipment and PPE’s. “
Ulakovich said ultimately, the goal is to help local healthcare providers make decisions as they continue to deal with the healthcare crisis. “If they can increase the volume of tests with faster results, then our charts will be more accurate, and we can make better predictions.”
The researchers are ready to share the prototype with healthcare representatives to get feedback on the metrics. Ulakovich said custom reports could then be created and given to healthcare administrators to help them with treatments and patient recovery.
Ulakovich’s work is being funded by WE-SPARK Health Institute, a collaboration of local health and education institutions to foster research projects.