Search

Background Emergency admissions are a major source of healthcare spending. We aimed to derive, validate, and compare conventional and machine learning models for prediction of the first emergency admission. Machine learning methods are capable of capturing complex interactions that are likely to be present when predicting less specific outcomes, such as this one. Methods and findings We used longitudinal data from linked electronic health records of 4.6 million patients aged 18-100 years from 389 practices across England between 1985 to 2015. The population was divided into a derivation cohort (80%, 3.75 million patients from 300 general practices) and a validation cohort (20%, 0.88 million patients from 89 general practices) from geographically distinct regions with different risk levels. We first replicated a previously reported Cox proportional hazards (CPH) model for prediction of the risk of the first emergency admission up to 24 months after baseline. This reference model was then compared with 2 machine learning models, random forest (RF) and gradient boosting classifier (GBC). The initial set of predictors for all models included 43 variables, including patient demographics, lifestyle factors, laboratory tests, currently prescribed medications, selected morbidities, and previous emergency admissions. We then added 13 more variables (marital status, prior general practice visits, and 11 additional morbidities), and also enriched all variables by incorporating temporal information whenever possible (e.g., time since first diagnosis). We also varied the prediction windows to 12, 36, 48, and 60 months after baseline and compared model performances. For internal validation, we used 5-fold cross-validation. When the initial set of variables was used, GBC outperformed RF and CPH, with an area under the receiver operating characteristic curve (AUC) of 0.779 (95% CI 0.777, 0.781), compared to 0.752 (95% CI 0.751, 0.753) and 0.740 (95% CI 0.739, 0.741), respectively. In external validation, we observed an AUC of 0.796, 0.736, and 0.736 for GBC, RF, and CPH, respectively. The addition of temporal information improved AUC across all models. In internal validation, the AUC rose to 0.848 (95% CI 0.847, 0.849), 0.825 (95% CI 0.824, 0.826), and 0.805 (95% CI 0.804, 0.806) for GBC, RF, and CPH, respectively, while the AUC in external validation rose to 0.826, 0.810, and 0.788, respectively. This enhancement also resulted in robust predictions for longer time horizons, with AUC values remaining at similar levels across all models. Overall, compared to the baseline reference CPH model, the final GBC model showed a 10.8% higher AUC (0.848 compared to 0.740) for prediction of risk of emergency admission within 24 months. GBC also showed the best calibration throughout the risk spectrum. Despite the wide range of variables included in models, our study was still limited by the number of variables included; inclusion of more variables could have further improved model performances. Conclusions The use of machine learning and addition of temporal information led to substantially improved discrimination and calibration for predicting the risk of emergency admission. Model performance remained stable across a range of prediction time windows and when externally validated. These findings support the potential of incorporating machine learning models into electronic health records to inform care and service planning., Author(s): Fatemeh Rahimian 1,2,*, Gholamreza Salimi-Khorshidi 1,2,*, Amir H. Payberah 1,2, Jenny Tran 1,2, Roberto Ayala Solares 1,2, Francesca Raimondi 1,2, Milad Nazarzadeh 1,2, Dexter Canoy 1,2, Kazem Rahimi 1,2,3 [...]

Background Multimorbidity in people with cardiovascular disease (CVD) is common, but large-scale contemporary reports of patterns and trends in patients with incident CVD are limited. We investigated the burden of comorbidities in patients with incident CVD, how it changed between 2000 and 2014, and how it varied by age, sex, and socioeconomic status (SES). Methods and findings We used the UK Clinical Practice Research Datalink with linkage to Hospital Episode Statistics, a population-based dataset from 674 UK general practices covering approximately 7% of the current UK population. We estimated crude and age/sex-standardised (to the 2013 European Standard Population) prevalence and 95% confidence intervals for 56 major comorbidities in individuals with incident non-fatal CVD. We further assessed temporal trends and patterns by age, sex, and SES groups, between 2000 and 2014. Among a total of 4,198,039 people aged 16 to 113 years, 229,205 incident cases of non-fatal CVD, defined as first diagnosis of ischaemic heart disease, stroke, or transient ischaemic attack, were identified. Although the age/sex-standardised incidence of CVD decreased by 34% between 2000 to 2014, the proportion of CVD patients with higher numbers of comorbidities increased. The prevalence of having 5 or more comorbidities increased 4-fold, rising from 6.3% (95% CI 5.6%-17.0%) in 2000 to 24.3% (22.1%-34.8%) in 2014 in age/sex-standardised models. The most common comorbidities in age/sex-standardised models were hypertension (28.9% [95% CI 27.7%-31.4%]), depression (23.0% [21.3%-26.0%]), arthritis (20.9% [19.5%-23.5%]), asthma (17.7% [15.8%-20.8%]), and anxiety (15.0% [13.7%-17.6%]). Cardiometabolic conditions and arthritis were highly prevalent among patients aged over 40 years, and mental illnesses were highly prevalent in patients aged 30-59 years. The age-standardised prevalence of having 5 or more comorbidities was 19.1% (95% CI 17.2%-22.7%) in women and 12.5% (12.0%-13.9%) in men, and women had twice the age-standardised prevalence of depression (31.1% [28.3%-35.5%] versus 15.0% [14.3%-16.5%]) and anxiety (19.6% [17.6%-23.3%] versus 10.4% [9.8%-11.8%]). The prevalence of depression was 46% higher in the most deprived fifth of SES compared with the least deprived fifth (age/sex-standardised prevalence of 38.4% [31.2%-62.0%] versus 26.3% [23.1%-34.5%], respectively). This is a descriptive study of routine electronic health records in the UK, which might underestimate the true prevalence of diseases. Conclusions The burden of multimorbidity and comorbidity in patients with incident non-fatal CVD increased between 2000 and 2014. On average, older patients, women, and socioeconomically deprived groups had higher numbers of comorbidities, but the type of comorbidities varied by age and sex. Cardiometabolic conditions contributed substantially to the burden, but 4 out of the 10 top comorbidities were non-cardiometabolic. The current single-disease paradigm in CVD management needs to broaden and incorporate the large and increasing burden of comorbidities., Author(s): Jenny Tran 1,2, Robyn Norton 1,3, Nathalie Conrad 1, Fatemeh Rahimian 1,2, Dexter Canoy 1,2, Milad Nazarzadeh 1,4, Kazem Rahimi 1,2,5,* Introduction Comorbidities, defined as the presence of additional [...]