Development and evaluation of a machine learning-based in-hospital COVID-19 disease outcome predictor (CODOP): A multicontinental retrospective study.

New SARS-CoV-2 variants, breakthrough infections, waning immunity, and sub-optimal vaccination rates account for surges of hospitalizations and deaths. There is an urgent need for clinically valuable and generalizable triage tools assisting the allocation of hospital resources, particularly in resource-limited countries. We developed and validate CODOP, a machine learning-based tool for predicting the clinical outcome of hospitalized COVID-19 patients. CODOP was trained, tested and validated with six cohorts encompassing 29223 COVID-19 patients from more than 150 hospitals in Spain, the USA and Latin America during 2020-22. CODOP uses 12 clinical parameters commonly measured at hospital admission for reaching high discriminative ability up to 9 days before clinical resolution (AUROC: 0·90-0·96), it is well calibrated, and it enables an effective dynamic risk stratification during hospitalization. Furthermore, CODOP maintains its predictive ability independently of the virus variant and the vaccination status. To reckon with the fluctuating pressure levels in hospitals during the pandemic, we offer two online CODOP calculators, suited for undertriage or overtriage scenarios, validated with a cohort of patients from 42 hospitals in three Latin American countries (78-100% sensitivity and 89-97% specificity). The performance of CODOP in heterogeneous and geographically disperse patient cohorts and the easiness of use strongly suggest its clinical utility, particularly in resource-limited countries.

While COVID-19 vaccines have saved millions of lives, new variants, waxing immunity, unequal rollout and relaxation of mitigation strategies mean that the pandemic will keep on sending shockwaves across healthcare systems. In this context, it is crucial to equip clinicians with tools to triage COVID-19 patients and forecast who will experience the worst forms of the disease. Prediction models based on artificial intelligence could help in this effort, but the task is not straightforward. Indeed, the pandemic is defined by ever-changing factors which artificial intelligence needs to cope with. To be useful in the clinic, a prediction model should make accurate prediction regardless of hospital location, viral variants or vaccination and immunity statuses. It should also be able to adapt its output to the level of resources available in a hospital at any given time. Finally, these tools need to seamlessly integrate into clinical workflows to not burden clinicians. In response, Klén et al. built CODOP, a freely available prediction algorithm that calculates the death risk of patients hospitalized with COVID-19 (https://gomezvarelalab.em.mpg.de/codop/). This model was designed based on biochemical data from routine blood analyses of COVID-19 patients. Crucially, the dataset included 30,000 individuals from 150 hospitals in Spain, the United States, Honduras, Bolivia and Argentina, sampled between March 2020 and February 2022 and carrying most of the main COVID-19 variants (from the original Wuhan version to Omicron). CODOP can predict the death or survival of hospitalized patients with high accuracy up to nine days before the clinical outcome occurs. These forecasting abilities are preserved independently of vaccination status or viral variant. The next step is to tailor the model to the current pandemic situation, which features increasing numbers of infected people as well as accumulating immune protection in the overall population. Further development will refine CODOP so that the algorithm can detect who will need hospitalisation in the next 24 hours, and who will need admission in intensive care in the next two days. Equipping primary care settings and hospitals with these tools will help to restore previous standards of health care during the upcoming waves of infections, particularly in countries with limited resources.

Incidence of COVID-19 in outpatients with cancer receiving active treatment in the context of a pandemic: An Andalusian cohort study.

BACKGROUND: Leading scientific societies have recommended delaying and/or suspending active cancer treatment during the COVID-19 pandemic. Nevertheless, data on this novel infection in patients with a diagnosis of cancer receiving active treatment are scarce and it is unknown if these recommendations could have repercussions on future progress of the disease. The main objective of this study is to learn the COVID-19 incidence rate in outpatients with cancer receiving active treatment. METHODS: This work is a retrospective cohort study that included all patients with a diagnosis of cancer who received active cancer treatment in two Andalusian hospitals between February 26 and May 13, 2020. Variables regarding the patient, tumor, and development of COVID-19 were collected. A descriptive analysis was performed and the cumulative incidence of COVID-19 in these patients was evaluated. RESULTS: A total of 673 patients were included. The median age was 62 years. There was a low rate of comorbidity and 12.1% had an ECOG >2. Breast cancer was the most common cancer (41%), followed by colorectal and lung cancer. Stage IV cancer was reported in 52.7% of patients. The most common treatment was chemotherapy (53.9%). Treatment was delayed or suspended in 6% of patients. Only three patients developed COVID-19. The cumulative incidence was 0.44% and one person died due to infection. CONCLUSIONS: In the present retrospective cohort study we found a low incidence of COVID-19 infection in patients with cancer receiving active treatment in an outpatient setting. The sociodemographic factors of Andalusia may explain why these results differ from those presented by other colleagues in Spain, but raise questions about whether universal recommendations may put the benefits of antineoplastic therapy at risk.