Decision region analysis to deconstruct the subgroup influence on AI/ML predictions

Assessing the generalizability of deep learning algorithms based on the size and diversity of the training data is not trivial. This study uses the mapping of samples in the image data space to the decision regions in the prediction space to understand how different subgroups in the data impact the neural network learning process and affect model generalizability. Using vicinal distribution-based linear interpolation, a plane of the decision region space spanned by the random 'triplet' of three images can be constructed. Analyzing these decision regions for many random triplets can provide insight into the relationships between distinct subgroups. In this study, a contrastive self-supervised approach is used to develop a 'base' classification model trained on a large chest x-ray (CXR) dataset. The base model is fine-tuned on COVID-19 CXR data to predict image acquisition technology (computed radiography (CR) or digital radiography (DX) and patient sex (male (M) or female (F)). Decision region analysis shows that the model's image acquisition technology decision space is dominated by CR, regardless of the acquisition technology for the base images. Similarly, the Female class dominates the decision space. This study shows that decision region analysis has the potential to provide insights into subgroup diversity, sources of imbalances in the data, and model generalizability. © 2023 SPIE.

Learning from real world data about combinatorial treatment selection for COVID-19.

COVID-19 is an unprecedented global pandemic with a serious negative impact on virtually every part of the world. Although much progress has been made in preventing and treating the disease, much remains to be learned about how best to treat the disease while considering patient and disease characteristics. This paper reports a case study of combinatorial treatment selection for COVID-19 based on real-world data from a large hospital in Southern China. In this observational study, 417 confirmed COVID-19 patients were treated with various combinations of drugs and followed for four weeks after discharge (or until death). Treatment failure is defined as death during hospitalization or recurrence of COVID-19 within four weeks of discharge. Using a virtual multiple matching method to adjust for confounding, we estimate and compare the failure rates of different combinatorial treatments, both in the whole study population and in subpopulations defined by baseline characteristics. Our analysis reveals that treatment effects are substantial and heterogeneous, and that the optimal combinatorial treatment may depend on baseline age, systolic blood pressure, and c-reactive protein level. Using these three variables to stratify the study population leads to a stratified treatment strategy that involves several different combinations of drugs (for patients in different strata). Our findings are exploratory and require further validation.

Mycotic infection as a risk factor for COVID-19: A meta-analysis.

More than 405 million people have contracted coronavirus disease 2019 (COVID-19) worldwide, and mycotic infection may be related to COVID-19 development. There are a large number of reports showing that COVID-19 patients with mycotic infection have an increased risk of mortality. However, whether mycotic infection can be considered a risk factor for COVID-19 remains unknown. We searched the PubMed and Web of Science databases for studies published from inception to December 27, 2021. Pooled effect sizes were calculated according to a random-effects model or fixed-effect model, depending on heterogeneity. We also performed subgroup analyses to identify differences in mortality rates between continents and fungal species. A total of 20 articles were included in this study. Compared with the controls, patients with mycotic infection had an odds ratio (OR) of 2.69 [95% confidence interval (CI): 2.22-3.26] for mortality and an OR of 2.28 (95% CI: 1.65-3.16) for renal replacement therapy (RRT). We also conducted two subgroup analyses based on continent and fungal species, and we found that Europe and Asia had the highest ORs, while Candida was the most dangerous strain of fungi. We performed Egger's test and Begg's test to evaluate the publication bias of the included articles, and the p-value was 0.423, which indicated no significant bias. Mycotic infection can be regarded as a risk factor for COVID-19, and decision makers should be made aware of this risk.

Efficacy of early use of remdesivir: a systematic review of subgroup analysis.

OBJECTIVE: A possible benefit has been suggested for early treatment of severe coronavirus disease 2019 (COVID-19) with remdesivir. The efficacy of this drug is controversial and could significantly influence the efficiency in healthcare systems. The objective is the methodological interpretation of subgroup analyzes according to starting of remdesivir treatment with respect to symptom onset of COVID-19. METHODS: A search in Pubmed® database was performed. Randomized clinical trials (RCTs) with subgroup analysis regarding early and late use of remdesivir were selected. All endpoints were assessed using two methodologies. First methodology considered statistical interaction, pre-specification, biological plausibility, and consistency of results. Second methodology was a validated tool with preliminary questions to discard subset analysis without relevant minimum conditions, and a checklist with recommendations for applicability. RESULTS: A total of 54 results were found and five RCTs were selected. According first methodology, consistent heterogeneity was only found in time to clinical improvement and better clinical status score at day 15 for patients with severe COVID-19 and <7 days of symptoms. About second methodology, these results about early use of remdesivir may be applied to clinical practice with caution. CONCLUSIONS: We developed a systematic search and application of an established methodology for interpretation of subgroup analysis about early use of remdesivir. Results in severe COVID-19 suggested that early use of remdesivir provides a greater benefit in <7 days of symptoms for time to clinical improvement and better clinical status score at day 15. Future studies could use 7-day cut-off of symptoms to evaluate remdesivir.

Clinical characteristics of COVID-19 patients with complications: implications for management.

BACKGROUND: A novel coronavirus disease 2019 (COVID-19) has caused outbreaks worldwide, and the number of cases is rapidly increasing through human-to-human transmission. Because of the greater transmission capacity and possible subsequent multi-organ damage caused by the virus, it is crucial to understand precisely and manage COVID-19 patients. However, the underlying differences in the clinical features of COVID-19 with and without comorbidities are not fully understood. AIM: The objective of this study was to identify the clinical features of COVID-19 patients with and without complications to guide treatment and predict the prognosis. METHOD: We collected the clinical characteristics of COVID-19 patients with and without different complications, including hypertension, cardiovascular disease and diabetes. Next, we performed a baseline comparison of each index and traced the dynamic changes in these factors during hospitalization to explore the potential associations. RESULT: A clinical index of differential expression was used for the regression to select top-ranking factors. The top-ranking clinical characteristics varied in each subgroup, such as indices of liver function, renal function and inflammatory markers. Among them, the indices of renal function were highly ranked in all subgroups and displayed significant differences during hospitalization. CONCLUSION: Organ functions of COVID-19 patients, particularly renal function, should be cautiously taken care of during management and might be a crucial factor for a poor prognosis of these patients with complications.

Effectiveness of convalescent plasma in Indian patients with COVID-19.

BACKGROUND: Convalescent plasma (CP) is being used as a treatment option in hospitalized patients with COVID-19. Till date, there is conflicting evidence on efficacy of CP in reducing COVID-19 related mortality. OBJECTIVE: To evaluate the effect of CP on 28-day mortality reduction in patients with COVID-19. METHODS: We did a multi-centre, retrospective case control observational study from 1st May 2020 to 31st August 2020. A total of 1079 adult patients with moderate and severe COVID-19 requiring oxygen, were reviewed. Of these, 694 patients were admitted to ICU. Out of these, 333 were given CP along with best supportive care and remaining 361 received best supportive care only. RESULTS: In the overall group of 1079 patients, mortality in plasma vs no plasma group was statistically not significant (22.4% vs 18.5%; p = 0.125; OR = 1.27, 95% CI: 0.94--1.72). However, in patients with COVID-19 admitted to ICU, mortality was significantly lower in plasma group (25.5% vs 33.2%; p = 0.026; OR = 0.69, 95%CI: 0.50-0.96). This benefit of reduced mortality was most seen in age group 60 to 74 years (26.7% vs 43.0%; p = 0.004; OR = 0.48, 95% CI: 0.29-0.80), driven mostly by females of this age group (23.1% vs 53.5%; p = 0.013; OR = 0.26, 95% CI: 0.09-0.78). Significant difference in mortality was observed in patients with one comorbidity (22.3% vs 36.5%; p = 0.004; OR = 0.50, 95% CI: 0.31-0.80). Moreover, patients on ventilator had significantly lower mortality in the plasma arm (37.2% vs 49.3%; p = 0.009; OR = 0.61, 95% CI: 0.42-0.89); particularly so for patients on invasive mechanical ventilation (63.9% vs 82.9%; p = 0.014; OR = 0.37, 95% CI: 0.16-0.83). CONCLUSION: The use of CP was associated with reduced mortality in COVID-19 elderly patients admitted in ICU, above 60 years of age, particularly females, those with comorbidities and especially those who required some form of ventilation.