Predictive model validation for Covid-19 hospital mortality
Research type
Research Study
Full title
External validation of the Oldham mortality model for patients hospitalised with Covid-19.
IRAS ID
288939
Contact name
Georges Ng Man Kwong
Contact email
Sponsor organisation
The Pennine Acute Hospitals NHS Trust
Duration of Study in the UK
0 years, 2 months, 30 days
Research summary
Summary of Research
This study aims to externally validate a model that predicts death in hospital or within 30 days of discharge for those admitted with Covid-19. The model was derived through multivariate analysis of a wide range of admission sociodemographic, clinical, biochemical, and radiological factors from patients admitted with Covid-19 to Royal Oldham Hospital prior to 19th May 2020.The factors included in the model are Rockwood clinical frailty scale, history of hypertension, history of current or prior cancer, presence of acute kidney injury, >50% new total lung field infiltrates on chest radiograph, C-reactive protein >100, and platelets <150. The area under the curve (AUC) of the receiver operating characteristic in the training cohort is 0.83 indicating excellent discriminatory ability.
If successfully externally validated, this model will aid decision making regarding which patients with Covid-19 require admission to hospital, and which patients may benefit from early respiratory support and/or the use of potential novel therapeutic agents.
Multiple sites within the North West of England will contribute to the study, such that the findings are transferable to similar settings. Data will be collected retrospectively from electronic patient records such that each site can contribute 50 patients who have been admitted since October 1st 2020 such that the findings are contemporaneous given that the treatment of Covid-19 is rapidly evolving. The expected number of subjects is 500, to allow at least 100 patients in each outcome group of death and survival, given that mortality in the modelling cohort was 36%.
Summary of Results
Aims: This study examined whether death among those hospitalised with Covid-19 could be predicted using the Oldham Composite Covid-19 associated Mortality Model (OCCAM). This predictive model is comprised of seven pieces of information routinely available at the time of hospital admission – patient age, history of high blood pressure, history of cancer, low platelet count (blood cells that play an important part in clotting), raised C-reactive protein (blood marker of inflammation), the presence of acute kidney injury (damage to the kidneys), and the extent of damaged lung tissue evident on the patients’ chest x-ray. This was undertaken in a different group of patients to the group that informed development of the model – this is known as external validation and aims to provide assurance that the model can be generalised to other populations.Background: The OCCAM model was derived from data collected about patients admitted to Royal Oldham Hospital with Covid-19 between March and May 2020. Prior analysis of this data suggested that the factors included in the model were independent predictors for death in hospital or within 30 days of discharge. In this model development group 36% of patients died. The ability to predict poor outcomes has important implications for decisions around hospital admission, the targeted and timely use of medicines which may be in short supply or are very expensive, and shared decision making with patients and their families.
Methods: This study looked at how well the OCCAM model predicted death during admission or within 30 days of discharge from hospital among 300 patients admitted for the treatment of Covid-19 to six different hospitals in North West England between September 2020 and February 2021. Data was collected retrospectively, after patients’ hospital admissions had ended.
Key findings:297 patients were included in the final analysis for external validation. The OCCAM model demonstrated excellent ability to predict death during admission or within 30 days of discharge and performed equally well across groups of patients estimated to be at low, medium, and high levels of risk for death. It was equally effective at predicting risk of death in the external validation group as the group of patients from which the model was derived.
Unlike other types of infection, the severity of disease and risk of death from Covid-19 has lessened. This is a result of the emergence of different variants (sorts) of virus as well as high levels of vaccine uptake and better medicines for treating Covid-19 infection. As such, further work should be undertaken to re-evaluate the performance of the OCCAM model.
REC name
Yorkshire & The Humber - South Yorkshire Research Ethics Committee
REC reference
21/YH/0037
Date of REC Opinion
5 Feb 2021
REC opinion
Favourable Opinion