Non applicability of validated predictive models for intensive care admission and death of COVID-19 patients in a secondary care hospital in Belgium

Objective To set up simple and reliable predictive scores for intensive care admissions and deaths in COVID-19 patients. These scores adhere to the TRIPOD (transparent reporting of a multivariable prediction model for individual prognosis or diagnosis) reporting guidelines. Design Monocentric retrospective cohort study run from early March to end of May in Clinique Saint-Pierre Ottignies, a secondary care hospital located in Ottignies-Louvain-la-Neuve, Belgium. The outcomes of the study are (i) admission in the Intensive Care Unit and (ii) death. Data sources All patients admitted in the Emergency Department with a positive RT-PCR SARS-CoV-2 test were included in the study. Routine clinical and laboratory data were collected at their admission and during their stay. Chest X-Rays and CT-Scans were performed and analyzed by a senior radiologist. Methods A recently published predictive score conducted on a large scale was used as a benchmark value (Liang score)1. Logistic regressions were used to develop predictive scores for (i) admission to ICU among emergency ward patients; (ii) death among ICU patients on 40 clinical variables. These models were based on medical intuition and simple model selection tools. Their predictive capabilities were then compared to Liang score. Results Our results suggest that Liang score may not provide reliable guidance for ICU admission and death. Moreover, the performance of this approach is clearly outperformed by models based on simple markers. For example, a logistic regression considering only the LDH yields to similar sensitivity and greater specificity. Finally, all models considered in this study lead to levels of specificity under or equal to 50%. Conclusions In our experience, the results of a predictive score based on a large-scale Chinese study cannot be applied in the Belgian population. However, in our small cohort it appears that LDH above 579 UI/L and venous lactate above 3.02 mmol/l may be considered as good predictive biological factors for ICU admission. With regard to death risk, NLR above 22.1, tobacco abuse status and 80 % of respiratory impairment appears to be relevant predictive factors. A predictive score for admission to ICU or death is desperately needed in secondary hospitals. Optimal allocation of resources guided by evidence-based indicators will best guide patients at time of admission and avoid futile treatments in intensive care units.