771. Prospective validation of the universal vital assessment (UVA) score to predict the in-hospital mortality of patients with acute illness admitted to a government district hospital in KwaZulu-Natal, South Africa

Background: Critical illness is a frequent cause of mortality in resource-limited settings Improved triage on admission could improve mortality, but existing tools depend on variables that often are not available We prospectively evaluated the universal vital assessment (UVA) score to predict mortality among patients admitted to a district hospital in rural, highly HIV-prevalent South Africa Figure 1 Receiver operator characteristic (ROC) curves for the UVA and qSOFA scoring tools Methods: In February-March 2020, adults admitted to the medical wards were enrolled, prior to interruption by covid19, and 30-day mortality assessed Clinical parameters including temperature, heart and respiratory rates, systolic blood pressure, oxygen saturation, Glasgow Coma Scale score, and HIV status were collected within 24 hours of admission as part of routine care Lower respiratory tract infections (LRTI) included pneumonia and suspected pulmonary tuberculosis To evaluate the predictive ability of the UVA score, area under the receiver operating characteristic curve (aROC) and age-sex adjusted binary logistic regression models were generated, and compared to the sequential organ failure assessment (qSOFA) Results: Sixty one patients were enrolled;outcomes were available for 56 patients Patients had a mean age of 52 (SD+17), 51% were women, and 46% were HIV infected The 30-day mortality was 16 1% (9/56) with infections and non-communicable diseases comprising 47% and 47% of admission diagnoses, respectively The most common admitting diagnosis was LRTI (24 6%) The median (+IQR) UVA score was 2 (+3) accounting for 36% of participants Medium-risk (2-4) and high-risk (>4) UVA groups were not associated with 30-day mortality, although the high-risk score trended towards significance (p=0 07) However, a UVA score > 3 was significantly associated with 30-day mortality, both alone and after adjusting for age and sex (aOR 6 2, 95% CI 1 2-33 1;p=0 03) The aROC (95% CI) for the UVA score was 0 74 (0 55 - 0 93), which performed better than qSOFA (aROC 0 59, 95% CI 0 37-0 81) and is shown in Figure 1 Conclusion: In this resource-limited, HIV-prevalent setting, the UVA score predicted mortality better than the qSOFA score A moderate-risk UVA score (>3) was predictive of 30-day mortality, though needs to be confirmed in larger studies