Prediction of SARS-CoV-2 infection with a Symptoms-Based model to aid public health decision making in Latin America and other low and middle income settings

Implementing Symptom-Based Predictive Models for Early Diagnosis of Pediatric Respiratory Viral Infections

(1) Background: Respiratory viral infections, including those caused by SARS-CoV-2, respiratory syncytial virus (RSV), influenza viruses, rhinovirus, and adenovirus, are major causes of acute respiratory infections (ARIs) in children. Symptom-based predictive models are valuable tools for expediting diagnoses, particularly in primary care settings. This study assessed the effectiveness of machine learning-based models in estimating infection probabilities for these common pediatric respiratory viruses, using symptom data. (2) Methods: Data were collected from 868 children with ARI symptoms evaluated across 14 primary care centers, members of COPEDICAT (Coronavirus Pediatria Catalunya), from October 2021 to October 2023. Random forest and boosting models with 10-fold cross-validation were used, applying SMOTE-NC to address class imbalance. Model performance was evaluated via area under the curve (AUC), sensitivity, specificity, and Shapley additive explanations (SHAP) values for feature importance. (3) Results: The model performed better for RSV (AUC: 0.81, sensitivity: 0.64, specificity: 0.77) and influenza viruses (AUC: 0.71, sensitivity: 0.70, specificity: 0.59) and effectively ruled out SARS-CoV-2 based on symptom absence, such as crackles and wheezing. Predictive performance was lower for non-enveloped viruses like rhinovirus and adenovirus, due to their nonspecific symptom profiles. SHAP analysis identified key symptoms patterns for each virus. (4) Conclusions: The study demonstrated that symptom-based predictive models effectively identify pediatric respiratory infections, with notable accuracy for those caused by RSV, SARS-CoV-2, and influenza viruses.

Development of a diagnostic multivariable prediction model of a positive SARS-CoV-2 RT-PCR result in healthcare workers with suspected SARS-CoV-2 infection in hospital settings

BACKGROUND

Despite declining COVID-19 incidence, healthcare workers (HCWs) still face an elevated risk of Severe Acute Respiratory Syndrome Coronavirus 2 (SARS-CoV-2) infection. We developed a diagnostic multivariate model to predict positive reverse transcription polymerase chain reaction (RT-PCR) results in HCWs with suspected SARS-CoV-2 infection.

METHODS

We conducted a cross-sectional study on episodes involving suspected SARS-CoV-2 symptoms or close contact among HCWs in Bogotá, Colombia. Potential predictors were chosen based on clinical relevance, expert knowledge, and literature review. Logistic regression was used, and the best model was selected by evaluating model fit with Akaike Information Criterion (AIC), deviance, and maximum likelihood.

RESULTS

The study included 2498 episodes occurring between March 6, 2020, to February 2, 2022. The selected variables were age, socioeconomic status, occupation, service, symptoms (fever, cough, fatigue/weakness, diarrhea, anosmia or dysgeusia), asthma, history of SARS-CoV-2, vaccination status, and population-level RT-PCR positivity. The model achieved an AUC of 0.79 (95% CI 0.77-0.81), with 93% specificity, 36% sensitivity, and satisfactory calibration.

CONCLUSIONS

We present an innovative diagnostic prediction model that as a special feature includes a variable that represents SARS-CoV-2 epidemiological situation. Given its performance, we suggest using the model differently based on the level of viral circulation in the population. In low SARS-CoV-2 circulation periods, the model could serve as a replacement diagnostic test to classify HCWs as infected or not, potentially reducing the need for RT-PCR. Conversely, in high viral circulation periods, the model could be used as a triage test due to its high specificity. If the model predicts a high probability of a positive RT-PCR result, the HCW may be considered infected, and no further testing is performed. If the model indicates a low probability, the HCW should undergo a COVID-19 test. In resource-limited settings, this model can help prioritize testing and reduce expenses.

Integrating Interpretability in Machine Learning and Deep Neural Networks: A Novel Approach to Feature Importance and Outlier Detection in COVID-19 Symptomatology and Vaccine Efficacy

In this study, we introduce a novel approach that integrates interpretability techniques from both traditional machine learning (ML) and deep neural networks (DNN) to quantify feature importance using global and local interpretation methods. Our method bridges the gap between interpretable ML models and powerful deep learning (DL) architectures, providing comprehensive insights into the key drivers behind model predictions, especially in detecting outliers within medical data. We applied this method to analyze COVID-19 pandemic data from 2020, yielding intriguing insights. We used a dataset consisting of individuals who were tested for COVID-19 during the early stages of the pandemic in 2020. The dataset included self-reported symptoms and test results from a wide demographic, and our goal was to identify the most important symptoms that could help predict COVID-19 infection accurately. By applying interpretability techniques to both machine learning and deep learning models, we aimed to improve understanding of symptomatology and enhance early detection of COVID-19 cases. Notably, even though less than 1% of our cohort reported having a sore throat, this symptom emerged as a significant indicator of active COVID-19 infection, appearing 7 out of 9 times in the top four most important features across all methodologies. This suggests its potential as an early symptom marker. Studies have shown that individuals reporting sore throat may have a compromised immune system, where antibody generation is not functioning correctly. This aligns with our data, which indicates that 5% of patients with sore throats required hospitalization. Our analysis also revealed a concerning trend of diminished immune response post-COVID infection, increasing the likelihood of severe cases requiring hospitalization. This finding underscores the importance of monitoring patients post-recovery for potential complications and tailoring medical interventions accordingly. Our study also raises critical questions about the efficacy of COVID-19 vaccines in individuals presenting with sore throat as a symptom. The results suggest that booster shots might be necessary for this population to ensure adequate immunity, given the observed immune response patterns. The proposed method not only enhances our understanding of COVID-19 symptomatology but also demonstrates its broader utility in medical outlier detection. This research contributes valuable insights to ongoing efforts in creating interpretable models for COVID-19 management and vaccine optimization strategies. By leveraging feature importance and interpretability, these models empower physicians, healthcare workers, and researchers to understand complex relationships within medical data, facilitating more informed decision-making for patient care and public health initiatives.

The application of artificial intelligence in health policy: a scoping review

BACKGROUND

Policymakers require precise and in-time information to make informed decisions in complex environments such as health systems. Artificial intelligence (AI) is a novel approach that makes collecting and analyzing data in complex systems more accessible. This study highlights recent research on AI's application and capabilities in health policymaking.

METHODS

We searched PubMed, Scopus, and the Web of Science databases to find relevant studies from 2000 to 2023, using the keywords "artificial intelligence" and "policymaking." We used Walt and Gilson's policy triangle framework for charting the data.

RESULTS

The results revealed that using AI in health policy paved the way for novel analyses and innovative solutions for intelligent decision-making and data collection, potentially enhancing policymaking capacities, particularly in the evaluation phase. It can also be employed to create innovative agendas with fewer political constraints and greater rationality, resulting in evidence-based policies. By creating new platforms and toolkits, AI also offers the chance to make judgments based on solid facts. The majority of the proposed AI solutions for health policy aim to improve decision-making rather than replace experts.

CONCLUSION

Numerous approaches exist for AI to influence the health policymaking process. Health systems can benefit from AI's potential to foster the meaningful use of evidence-based policymaking.