Misattribution bias of COVID-19 hospitalizations in Alberta using an admission algorithm
Background:
With initial waves of COVID-19, many public health systems assumed each COVID-19 positive hospitalization was a direct cause from COVID-19 infection. Since January 2022, Alberta Health Services Communicable Disease Control Hospitalization Team (CDC-HT) implemented an admission criteria algorithm to adjudicate COVID-19 as a direct, contributing, or unrelated cause for all COVID-19 admissions in Alberta.
Methods:
This quality improvement initiative sought to improve the admission algorithm's precision in reporting COVID-19 admissions. Patient hospitalization records from January-February 2022 with a positive COVID-19 test in the last 30 days were proportionally sampled in a geographically stratified manner across Alberta health zones. 261 patient records were sampled and determination of COVID-19 attribution by CDC-HT algorithm was compared to adjudication by a panel of infectious diseases physicians with extensive COVID-19 clinical experience.
Results:
Of 261 sampled COVID-19 admissions, blinded physician adjudication determined 39.9% were direct-cause, 17.2% contributing-cause, and 37.6% unrelated-cause. Within the same cohort the CDC-HT admission algorithm adjudicated 42.9% direct-cause, 24.5% contributing-cause, and 30.3% unrelated-cause. Cohen's kappa was 0.475, providing only moderate agreement. The majority of discrepancy was from over-attribution of unrelated hospitalizations as contributing cause. Implementation of this algorithm in Alberta throughout 2022 showed a fluctuating proportion of direct plus contributing COVID-19 hospitalizations as low as 40%.
Conclusion:
There was misattribution bias in COVID-19 hospitalization determination using the admission algorithm. The findings from this analysis led to improvements in the algorithm to improve precision. Public health jurisdictions should review their COVID-19 hospitalization reporting approaches to ensure validity and consideration of incidental cases.
