Aims This study aimed to review the performance of machine learning (ML) methods compared with conventional statistical models (CSMs) for predicting readmission and mortality in patients with heart failure (HF) and to present an approach to formally evaluate the quality of studies using ML algorithms for prediction modelling. Methods and results Following Preferred Reporting Items for Systematic Reviews and Meta‐Analyses guidelines, we performed a systematic literature search using MEDLINE, EPUB, Cochrane CENTRAL, EMBASE, INSPEC, ACM Library, and Web of Science. Eligible studies included primary research articles published between January 2000 and July 2020 comparing ML and CSMs in mortality and readmission prognosis of initially hospitalized HF patients. Data were extracted and analysed by two independent reviewers. A modified CHARMS checklist was developed in consultation with ML and biostatistics experts for quality assessment and was utilized to evaluate studies for risk of bias. Of 4322 articles identified and screened by two independent reviewers, 172 were deemed eligible for a full‐text review. The final set comprised 20 articles and 686 842 patients. ML methods included random forests (n = 11), decision trees (n = 5), regression trees (n = 3), support vector machines (n = 9), neural networks (n = 12), and Bayesian techniques (n = 3). CSMs included logistic regression (n = 16), Cox regression (n = 3), or Poisson regression (n = 3). In 15 studies, readmission was examined at multiple time points ranging from 30 to 180 day readmission, with the majority of studies (n = 12) presenting prediction models for 30 day readmission outcomes. Of a total of 21 time‐point comparisons, ML‐derived c‐indices were higher than CSM‐derived c‐indices in 16 of the 21 comparisons. In seven studies, mortality was examined at 9 time points ranging from in‐hospital mortality to 1 year survival; of these nine, seven reported higher c‐indices using ML. Two of these seven studies reported survival analyses utilizing random survival forests in their ML prediction models. Both reported higher c‐indices when using ML compared with CSMs. A limitation of studies using ML techniques was that the majority were not externally validated, and calibration was rarely assessed. In the only study that was externally validated in a separate dataset, ML was superior to CSMs (c‐indices 0.913 vs. 0.835). Conclusions ML algorithms had better discrimination than CSMs in most studies aiming to predict risk of readmission and mortality in HF patients. Based on our review, there is a need for external validation of ML‐based studies of prediction modelling. We suggest that ML‐based studies should also be evaluated using clinical quality standards for prognosis research. Registration: PROSPERO CRD42020134867

The aim of the present study was to evaluate the interaction between depressive symptoms and metabolic dysregulations as risk factors for type 2 diabetes. The sample comprised of 2525 adults who participated in a baseline and a follow-up assessment over a 4.5-year period in the Emotional Health and Wellbeing Study (EMHS) in Quebec, Canada. A two-way stratified sampling design was used, on the basis of the presence of depressive symptoms and metabolic dysregulation (obesity, elevated blood sugar, high blood pressure, high levels of triglycerides and decreased high-density lipoprotein). A total of 87 (3.5%) individuals developed diabetes. Participants with both depressive symptoms and metabolic dysregulation had the highest risk of diabetes (adjusted odds ratio=6.61, 95% confidence interval (CI): 4.86-9.01), compared with those without depressive symptoms and metabolic dysregulation (reference group). The risk of diabetes in individuals with depressive symptoms and without metabolic dysregulation did not differ from the reference group (adjusted odds ratio=1.28, 95% CI: 0.81-2.03), whereas the adjusted odds ratio for those with metabolic dysregulation and without depressive symptoms was 4.40 (95% CI: 3.42-5.67). The Synergy Index (SI=1.52; 95% CI: 1.07-2.17) suggested that the combined effect of depressive symptoms and metabolic dysregulation was greater than the sum of individual effects. An interaction between depression and metabolic dysregulation was also suggested by a structural equation model. Our study highlights the interaction between depressive symptoms and metabolic dysregulation as a risk factor for type 2 diabetes. Early identification, monitoring and a comprehensive management approach of both conditions might be an important diabetes prevention strategy.

High depressive symptoms and cardiometabolic abnormalities are independently associated with an increased risk of diabetes. The purpose of this study was to assess the association of co-occurring depressive symptoms and cardiometabolic abnormalities on risk of diabetes in a representative sample of the English population aged 50 years and older. Data were from the English Longitudinal Study of Ageing. The sample comprised of 4454 participants without diabetes at baseline. High depressive symptoms were based on a score of 4 or more on the 8-item binary Centre for Epidemiologic Studies–Depression scale. Cardiometabolic abnormalities were defined as 3 or more cardiometabolic risk factors (hypertension, impaired glycemic control, systemic inflammation, low high-density lipoprotein cholesterol, high triglycerides, and central obesity). Cox proportional hazards regressions assessed the association between co-occurring depressive symptoms and cardiometabolic abnormalities with incidence of diabetes. Multiple imputation by chained equations was performed to account for missing data. Covariates included age, sex, education, income, smoking status, physical activity, alcohol consumption, and cardiovascular comorbidity. The follow-up period consisted of 106 months, during which 193 participants reported a diagnosis of diabetes. Diabetes incidence rates were compared across the following four groups: 1) no or low depressive symptoms and no cardiometabolic abnormalities (reference group, n = 2717); 2) high depressive symptoms only (n = 338); 3) cardiometabolic abnormalities only (n = 1180); and 4) high depressive symptoms and cardiometabolic abnormalities (n = 219). Compared to the reference group, the hazard ratio for diabetes was 1.29 (95% CI 0.63, 2.64) for those with high depressive symptoms only, 3.88 (95% CI 2.77, 5.44) for those with cardiometabolic abnormalities only, and 5.56 (95% CI 3.45, 8.94) for those with both high depressive symptoms and cardiometabolic abnormalities, after adjusting for socio-demographic, lifestyle and clinical variables. These findings suggest that those with high depressive symptoms and cardiometabolic abnormalities are at a particularly increased risk of type 2 diabetes.