Predicting COVID-19 Mortality
A research team developed an algorithm review estimating not only hospital admissions, but also mortality outcomes in adults with COVID-19.
Over the last 7 days on average, the United States has seen daily cases soar beyond 192000. Over 82000 people are hospitalized and we are inching closer to 2 million cases a day. Hospitalizations are a lagging indicator and will continue to rise with the cases. Assuming cases only worsen in the coming weeks during the holiday season, this will mean unprecedented strains on healthcare personnel and hospitals.
When hospitals become overwhelmed, both in terms of resources and staff, crisis care is often enacted. This ultimately recognizes the limitations of how much care can be given in an over-strained healthcare system. More recently, a research team worked to study a predictive algorithm that would estimate not only hospital admissions, but also mortality outcomes in adults with COVID-19.
Pulling data from over 1200 general practices in England, hospital episode statistics, and death registry data, the team looked at information on over 6 million adults in a derivation dataset and then 2 million within the validation dataset. Overall, this review looked at those adults aged 19-100 years from mid-January to late July 2020.
The authors noted, “the primary outcome was time to death from COVID-19, defined as death due to confirmed or suspected COVID-19 as per the death certification or death occurring in a person with confirmed severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection in the period January 24 to April 30. The secondary outcome was time to hospital admission with confirmed SARS-CoV-2 infection. Models were fitted in the derivation cohort to derive risk equations using a range of predictor variables.”
Upon review of 4384 deaths in the derivation cohort and then 1722 in the validation cohort (621 in the second validation cohort period), the algorithm had considerable success with the first validation cohort, explaining 73% of the variation in time of death in men.
For women, similar results were found but the algorithm also noted that in those patients with the highest predicted risks of death, the sensitivity of identifying death with 97 days was nearly 76%. Per NHS guidance, the conditions that they were also including as they were moderately associated with increased risk of complications were things like chronic, non-severe respiratory diseases, chronic kidney disease, chronic neurological conditions, diabetes, etc. Overall, the algorithm included a substantial amount of data and variables like chemotherapy, race, living situation, learning disability, etc.
As algorithms are important pieces of public health and policy making, it is important they are continuously improved, revised, and reviewed. This particular one is promising and could be used in understanding future waves of COVID-19 cases, hospitalizations, and deaths, which would allow for additional resources, better planning, etc.
The authors did emphasize in their study though that,“the QCOVID population based risk algorithm performed well, showing very high levels of discrimination for deaths and hospital admissions due to covid-19. The absolute risks presented, however, will change over time in line with the prevailing SARS-C0V-2 infection rate and the extent of social distancing measures in place, so they should be interpreted with caution.”
As we learn more about the virus that causes COVID-19 and the transmission dynamics and pieces of patient care that make it challenging, it will be important to continuously review and update predictive models for better application.