AUSTIN (KXAN) — With active cases of COVID-19 in slight decline around Texas, a projection by modelers at the University of Texas at Austin raises new concerns. It suggests there will be 15,000 more deaths in Texas over the next three weeks.

The model by UT’s College of Natural Sciences originally predicted between 20,655 and 26,726 Texans will die as a result of COVID-19 by Aug. 31.

On Aug. 11, the University revised their model with new data. The new projections saw between 28,963 and 12,638 deaths by Aug. 31, with a final prediction of 16,019 deaths.

UT modelers base their projection on current social distancing data they glean from geolocation cell phone data. An anomaly in the data accounted for the earlier high prediction.

As of Sunday, 8,459 Texans have died since the pandemic began. The predicted escalation would represent a 177% increase by Aug. 31.

The Modeling Consortium at UT uses the S-E-I-R method for predicting the number of COVID deaths over several months. The model takes into account the number of Susceptible people, how many individuals in the population have been Exposed, the number of Infected and the number of people who are in Recovery.

Additionally, the number of confirmed and possible COVID-19 deaths reported by Johns Hopkins University are factored in. Johns Hopkins bases their number on several factors including the number of deaths reported by the state.

Finally, the model factors in social distancing. The researchers determine how much distancing is occurring in an area by tracking the GPS on people’s phones. If you’re concerned about privacy, the GPS tracking does not follow you constantly. Instead it looks at locations and basis the information on that. The researchers gave looking at the number of people who visit a coffee shop in a given period as an example.

On July 27, the Texas Department of Health and Human Services changed the way they reported COVID-19 deaths. Previously, DSHS. counted a death when it was reported by local health departments. Now deaths are counted only if the death certificate lists COVID-19 as the cause of death. D.S.H.S. says this method results in fatalities being counted quicker.

When the new method went into effect, 675 new deaths were added to the state’s total that day. Plus, an automation error occurred a few days later. This caused 225 more deaths to be added to the count. Those numbers have since been removed from the state’s total, but not before warping UT’s projection for the next month.

Finally, the Consortium says that the projections do not reflect the slowdown in transmissions and hospitalizations that began in mid-June following policy changes at the state capitol. If mortality date begins to decline, the Consortium says the model will reflect that.

The state reported 116 new COVID-19 deaths Sunday, along with 4,800 new cases. It would need to average more than 700 new deaths a day to make the UT prediction true.