Identification of Patients with COVID-19 Infection Prior to the New COVID-19 Diagnostic Code - a Premier Database Analysis.
Author(s)
Holy C1, Shah S2, Elangovanraaj N2, Krishnan D3, Gupta S4, Trivedi P2, Devulapally M2, Mohapatra A2, Johnston S5, Coplan P6
1Johnson & Johnson, Somerville, MA, USA, 2Mu Sigma, Bangalore, KA, India, 3Mu Sigma, Bengaluru, KA, India, 4Mu Sigma, Bangalore, India, 5Johnson & Johnson, Annapolis, MD, USA, 6Johnson & Johnson, New Brunswick, NJ, USA
OBJECTIVES: A new diagnostic code was released in April 2020 for COVID-19 infection. Prior to that, database research used CDC-listed symptoms to identify COVID-19 cases. This study evaluates whether use of these symptoms can accurately identify patients with COVID-19, and at what time an increase in mortality was first observed using combinations of these symptoms. METHODS: Patients from the Premier Healthcare Database with an inpatient event during which patients deceased, with COVID-19 related symptoms and a DRG (diagnostic related group) of respiratory infection or sepsis and, from April 2020 onwards, a COVID-19 diagnosis, were identified, from September 2019 to most recent. Comorbidities and symptoms of COVID-19 and related treatments were used as model variables. Patients with and without COVID-19 diagnosis were matched using propensity score matching (model: logit, method: nearest neighbor, caliper: 0.1). Time of discharge was analyzed to evaluate changes in mortality in the matched cohorts. RESULTS: Pre-match, 3,923 patients without (noDiag) and 4,903 patients with COVID-specific diagnosis codes (withDiag) were identified. In the noDiag group, there were 52.7% males and 11.8% African Americans (AA). In the withDiag group, there were 56.4% males and 21.1% AA. In the noDiag cohort, 71.2% patients had a DRG code of septicemia vs 52.9% in the withDiag group. After matching, there were 1,340 patients in each group, 60% males, 21.7% AA. In the NoDiag matched cohort, 59% death occurred in 2020, with a clear spike in March (30%). The remaining 41% were identified prior to 2020, as early as October 2020, suggesting potential contamination with mortality due to other causes. CONCLUSIONS: Due to the similarity of COVID-symptoms to those observed with other common respiratory diseases, and concurrent timing of COVID mortality with that of viral pneumonia, matching on symptoms and treatments identified patients prior to first US-reported case and therefore included patients with infections other than COVID-19.
Conference/Value in Health Info
2020-11, ISPOR Europe 2020, Milan, Italy
Value in Health, Volume 23, Issue S2 (December 2020)
Code
PIN117
Topic
Epidemiology & Public Health
Disease
Infectious Disease (non-vaccine)