Abstract
•In our cohort of patients hospitalised due to COVID-19, 6.3% died•CURB-65, 4C Mortality and PRIEST scores all accurately predicted mortality•Risk prediction scores were accurate for Māori and Pacific peoples
COVID-19 severity prediction scores need further validation due to evolving COVID-19 illness. We evaluated existing COVID-19 risk prediction scores in Aotearoa New Zealand, including for Māori and Pacific peoples who have been inequitably effected by COVID-19.
We conducted a multicentre retrospective cohort study in adults hospitalised with COVID-19 from January-May 2022, including all Māori and Pacific patients, and every second non-Māori, non-Pacific (NMNP) patient to achieve equal analytic power by ethnic grouping. We assessed the accuracy of existing severity scores (4C Mortality, CURB-65, PRIEST and VACO) to predict death in hospital or within 28 days.
Of 2,319 patients, 582 (25.1%) identified as Māori, 914 (39.4%) Pacific, and 862 (37.2%) NMNP. There were 146 (6.3%, 95% confidence interval [CI] 5·4%, 7·4%) deaths, with predicted probability of death higher than observed mortality for VACO (10·4%), modified PRIEST (15·1%) and 4C mortality (15·5%) scores, but lower for CURB-65 (4·5%). C-statistics (95% CI) of severity scores were: 4C mortality: Māori 0·82 (0·75, 0·88), Pacific 0·87 (0·83, 0·90), NMNP 0·90 (0·86, 0·93); CURB-65: Māori 0·83 (0·69, 0·92), Pacific 0·87 (0·82, 0·91), NMNP 0·86 (0·80, 0·91); modified PRIEST: Māori 0·85 (0·79, 0·90), Pacific 0·81 (0·76, 0·86), NMNP 0·83 (0·78, 0·87); and VACO: Māori 0·79 (0·75, 0·83), Pacific 0·71 (0·58, 0·82), NMNP 0·78 (0·73, 0·83).
Following re-calibration, existing risk prediction scores accurately predicted mortality.