Two American studies reveal the effectiveness of these systems in determining the risk of complications and helping doctors plan treatments.
The first study was conducted in eastern Massachusetts to find out whether the sociodemographic characteristics, laboratory values and comorbidities of patients hospitalised with the 2019 coronavirus could allow a serious disease course to be predicted.
The rationale for the study lies in the fact that, since resources for the treatment of COVID-19 are limited, particularly with regard to mechanical ventilation, simple approaches are needed to stratify morbidity and mortality risk at the time of hospitalisation.
The retrospective study was conducted on a cohort of 2,511 patients hospitalised positive for SARS-CoV-2 in six hospitals, of whom 215 (8.6%) were admitted to the intensive care unit, 164 (6.5%) required mechanical ventilation and 292 (11.6%) died. In a risk prediction model, 212 deaths (78%) occurred in the quintile with the highest mortality risk.
The researchers used L1 regression models to determine whether abnormal haematological results and decreased renal function were associated with an increased risk of a severe hospital course.
Researchers say that predictions may be more useful during the initial week of hospitalization and that a further study would be useful to see if the repetition of models with more laboratory data, or the incorporation of other biomarkers, could improve long-term forecasting.
The second study was conducted by Duke University to find out if it would be possible to develop and then evaluate the performance of a clinical decision support tool (CDSS) to predict the use of resources and thus prioritise elective surgical procedures in recovery after the 2019 coronavirus pandemic (COVID-19).
Researchers developed predictive models to estimate the overall length of hospitalisation, intensive care, mechanical ventilator requirements and discharge disposition in a specialised nursing facility, using historical case data extracted from the electronic medical records of 42,199 patients. These models have been integrated into an interactive online dashboard with end-user input and tested iteratively.
The data that have been used include case types, patient demographics, patient history, comorbidities and drugs. The average length of stay was 2.3 days (range, 1.3-4.2 days), 6416 patients (15.2%) were admitted to intensive care, 1624 (3.8%) received mechanical ventilation and 2843 (6.7%) were discharged to a specialist care facility.
The performance of the CDSS was very good, with an area below the receiving operator characteristic ranging from 0.76 to 0.93. The sensitivity of high and medium risk groups was set at 95%. The negative predictive value of the low risk grouping was 99%.
The researchers integrated the models into a daily Tableau dashboard to guide decision-making. The CDSS is currently used by surgical management to inform case planning.
Digital-health 