A novel study appearing as a preprint on the medRxiv* server describes the performance of a new risk prediction model based on a wide range of patient variables for death and intensive care unit (ICU) admission.
The role of risk stratification
Risk stratification is essential in the management of COVID-19 due to the need to prioritize critical care services in situations of overwhelming numbers of patients. Since a subgroup of patients with this illness are severely sick and rapidly progress to life-threatening illness, it is necessary to have a reliable system to distinguish patients who will develop such symptoms and require mechanical ventilation and ICU admission, from low-risk patients who can be discharged to their homes.
Of the many prognostic models now described, many are subject to bias because of the population in which they were developed. Moreover, external validation is lacking for most cohorts.
In other models, the predictors are dependent on an accurate recording of comorbidities. Not only are such illnesses known to be poorly recorded in health records, but the predictive value for deterioration after admission is doubtful for such models.
This news article was a review of a preliminary scientific report that had not undergone peer-review at the time of publication. Since its initial publication, the scientific report has now been peer reviewed and accepted for publication in a Scientific Journal. Links to the preliminary and peer-reviewed reports are available in the Sources section at the bottom of this article. View Sources
Study aims
The current study explores the additional value of comorbidities and other patient information in such a model and updates the model with biomarkers measured in real-time. The researchers aimed to develop a new model to provide a predicted risk for death or ICU admission, to validate other earlier models (the UK International Severe Acute Respiratory and Emerging Infection Consortium (ISARIC) model and simplified 4C score), and to compare the performances of the three models.
They also set up daily models incorporating data from the first eight days post-admission to understand how the predictors change over the course of the disease.
Predictors
The researchers used routine data in over 1,000 patients with COVID-19 in a single center to create their model. The University Hospitals Birmingham (UHB) dataset is among the largest and most ethnically heterogeneous cohorts in the UK. Moreover, early research interest ensured that all patients were widely tested on admission to hospital, supporting the current study’s aims.
The predictors explored included demographics such as age and sex, clinical symptoms and signs, and the patient’s history of medical conditions. Laboratory and imaging results were also incorporated, for a total of 63 predictors in all.
Promising performance
The performance of these models was promising. They were able to predict mortality within 28 days of admission in 78% of cases, and 89% of ICU admissions in this period, for both males and females.
Since all the variables used as predictors are not routinely collected at admission in many hospitals, the researchers developed more economical models that use variables found in the CovidCollab external validation dataset. These were found to predict 79% of mortality and 81% of ICU admission.
The current model out-performs the ISARIC and 4C score models, which predicted 76% and 75% of mortality with the same data. Moreover, it uses only routinely collected data obtained at admission, obviating the need for additional evaluations and history-taking.
The model predictions agreed well with the observed probabilities of admission, particularly at lower probabilities. This is precisely the range of probability where it would be most useful to know the risk of disease progression.
The prognostic model developed here was also an external validation of models used in a large cohort of patients from all over the world, from a heterogeneous background, including patients from low- and middle-income countries.
The researchers failed to find any significant changes in the variables over the course of time. A recent Hong Kong study did report the ability to predict 91% of severe outcomes in COVID-19 using time-dependent variables, but external validation is still to be performed.
What are the implications?
These novel models show good discrimination, even beyond the accepted ISARIC and 4C score models, and maintained their performance even on external validation. These models can be used in tandem with hospital electronic medical records systems to predict the probability of mortality or ICU admission within 28 days of admission.
The UHB models use only routinely obtained admission data. This is helpful in that it can allow the prognosis to be predicted for each patient at the time of admission.
This will be helpful not just for triaging patients for critical care but to explain the patient’s condition and future measures that may be required for patient care, during acute severe infection. The implementation of these models, following their validation in other cohorts and populations, should be examined.
This news article was a review of a preliminary scientific report that had not undergone peer-review at the time of publication. Since its initial publication, the scientific report has now been peer reviewed and accepted for publication in a Scientific Journal. Links to the preliminary and peer-reviewed reports are available in the Sources section at the bottom of this article. View Sources
Article Revisions
- Apr 4 2023 - The preprint preliminary research paper that this article was based upon was accepted for publication in a peer-reviewed Scientific Journal. This article was edited accordingly to include a link to the final peer-reviewed paper, now shown in the sources section.
Incidence and risk factors of an invasive fungal lung infection among COVID-19 patients