Predictive Analytics Detects COVID-19 Infection
Predictive analytics indicates that seven joint symptoms could determine COVID-19 infection in a community.
Source: CDC
By - Using predictive analytics, a set of seven symptoms together can be used to maximize the detection of COVID-19 in a community. The paper was published in PLOS Medicine by Elliott of Imperial College London's Marc Chadeau-Hyam, Paul Elliott, and colleagues.
Although the research was conducted in England, information regarding the spread of COVID-19 can provide critical insights for public health officials worldwide.
The rapid detection of COVID-19 infection in a community is key in slowing the spread and controlling the transmission. When testing capacity is limited, tests must be used as efficiently as possible, including using the most informative symptoms for test allocation.
In the study, researchers obtained throat and nose swabs with valid COVID-19 PCR test results from 1,147,345 volunteers in England aged 5 years and older.
The data was collected over eight rounds of testing conducted between June 2020 and January 2021 as part of the REal-time Assessment of Community Transmission-1 (REACT-1) study. Participants were asked about symptoms they experienced the week before testing.
The research team developed a model based on the data obtained during rounds two to seven, with seven symptoms selected as jointly positively predictive of PCR positivity. The symptoms included loss or change of smell, loss or change of taste, fever, new persistent cough, chills, appetite loss, and muscle aches.
The first four of those symptoms are currently used in the United Kingdom to determine eligibility for community PCR testing. In round eight of testing, the resulting model used predictive analytics to determine PCR positivity.
The modeling suggested that using the seven symptoms identified for PRC test allocation would result in 30 to 40 percent of symptomatic individuals in England being eligible for a test, compared to 10 percent currently. If everyone eligible is tested, it could detect 70 to 75 percent of positive cases.
"In order to improve PCR positivity detection rates and consequently improve control of viral transmission via isolation measures, we would propose to extend the list of symptoms used for triage to all 7 symptoms we identified," the authors said in a press release.
"These findings suggest many people with COVID-19 won't be getting tested – and therefore won't be self-isolating – because their symptoms don't match those used in current public health guidance to help identify infected people," Elliott explained.
"We understand that there is a need for clear testing criteria, and that including lots of symptoms which are commonly found in other illnesses like seasonal flu could risk people self-isolating unnecessarily. I hope that our findings on the most informative symptoms mean that the testing program can take advantage of the available evidence, helping to optimize the detection of infected people."
