Artificial intelligence-driven prediction of COVID-19-related hospitalization and death: a systematic review

Abstract

AimTo perform a systematic review on the use of Artificial Intelligence (AI) techniques for predicting COVID-19 hospitalization and mortality using primary and secondary data sources.Study eligibility criteriaCohort, clinical trials, meta-analyses, and observational studies investigating COVID-19 hospitalization or mortality using artificial intelligence techniques were eligible. Articles without a full text available in the English language were excluded.Data sourcesArticles recorded in Ovid MEDLINE from 01/01/2019 to 22/08/2022 were screened.Data extractionWe extracted information on data sources, AI models, and epidemiological aspects of retrieved studies.Bias assessmentA bias assessment of AI models was done using PROBAST.ParticipantsPatients tested positive for COVID-19.ResultsWe included 39 studies related to AI-based prediction of hospitalization and death related to COVID-19. The articles were published in the period 2019-2022, and mostly used Random Forest as the model with the best performance. AI models were trained using cohorts of individuals sampled from populations of European and non-European countries, mostly with cohort sample size <5,000. Data collection generally included information on demographics, clinical records, laboratory results, and pharmacological treatments (i.e., high-dimensional datasets). In most studies, the models were internally validated with cross-validation, but the majority of studies lacked external validation and calibration. Covariates were not prioritized using ensemble approaches in most of the studies, however, models still showed moderately good performances with Area under the Receiver operating characteristic Curve (AUC) values >0.7. According to the assessment with PROBAST, all models had a high risk of bias and/or concern regarding applicability.ConclusionsA broad range of AI techniques have been used to predict COVID-19 hospitalization and mortality. The studies reported good prediction performance of AI models, however, high risk of bias and/or concern regarding applicability were detected.

Mapping the Heartbeat of America with ChatGPT-4: Unpacking the Interplay of Social Vulnerability, Digital Literacy, and Cardiovascular Mortality in County Residency Choices

“…While there was no significant change in classification (the national percentage of rural and urban populations remained consistent with the 2010 definition), the underlying population shift by the new definition could have introduced bias in our current findings [66]. Even though there has been a demonstration of good prediction with AI models when studying mortality and hospitalization rates for certain health conditions such as COVID-19, there could be concerns about replicability and the bias inherent in the study [7], including ours.…”

“…AI tools could also play a big role in identifying high-risk populations for hospital injury and guide interventions and preventive actions to improve patient safety [6]. Recent applications of AI in this domain have been in predicting hospitalizations and determining the mortality associated with COVID-19 [7,8]. The prospects and benefits of AI and machine learning use are well recognized and accepted in healthcare today concerning disease diagnosis, risk assessment associated with morbidity and mortality, surveillance of infectious diseases and outbreak prediction, and policy implications [8][9][10].…”

Transformative Potential of AI in Healthcare: Definitions, Applications, and Navigating the Ethical Landscape and Public Perspectives

“…AI-powered epidemiological models have become invaluable tools in predicting the transmission of viruses, including COVID-19. Furthermore, AI-based techniques can be utilized to track and monitor the spread of viruses at different scales, ranging from individual to population levels [99].…”

Machine learning-driven development of a disease risk score for COVID-19 hospitalization and mortality: a Swedish and Norwegian register-based study

“…Heterogeneity of populations and risk factors across geographical settings ( 10 ), including the effects of social determinants and their interplay ( 14 ) and the lack of validation of prognostic tools in multiple cohorts ( 17 ) has played a key role for lack of robustness and generalizability in performance across populations and settings. We recently conducted a systematic review and found that previous machine learning and artificial intelligence (AI)-based predictive models for COVID-19 hospitalization and mortality were affected by a high risk of bias or lack of applicability, especially due to lack of external validation of prognostic models ( 18 ). Of note, there are examples of studies that have developed AI-driven models for COVID-19 hospitalization or death, which underwent external validation ( 19 , 20 ).…”

“…Of note, there are examples of studies that have developed AI-driven models for COVID-19 hospitalization or death, which underwent external validation ( 19 , 20 ). However, it is worth mentioning that we consider these studies as having a high risk of bias ( 18 ). Therefore, this study aimed at overcoming the limitations of the previously developed AI models by more stringently identifying predictors of COVID-19 severity and using them to develop a disease risk score (DRS) for COVID-19-related hospitalization and for COVID-19 death – overall and across the COVID-19 waves – for residents in Sweden, and externally validate the DRS in Norway.…”