A Validation Study: How Predictive Is a Diagnostic Coding Algorithm at Identifying Rheumatic Heart Disease in Western Australian Hospital Data?

BACKGROUND: International Classification of Diseases codes for rheumatic heart disease (RHD) (ICD-10 I05-I08) include valvular heart disease of unspecified origin, limiting their usefulness for estimating RHD burden. An expert opinion-based algorithm was developed to increase their accuracy for epidemiological case ascertainment. The algorithm included codes not defaulting to RHD ('probable') plus selected codes pertaining to mitral valve involvement in patients <60 years ('possible'). We aimed to determine the positive predictive value (PPV) for RHD of algorithm-selected hospital admissions. METHODS: Chart reviews of RHD-coded admissions (n=368) to Western Australian tertiary hospitals (2009-2016) authenticated RHD diagnosis. We selected all cases with algorithm-positive codes from populations at high-risk of RHD and an age-stratified random sample from low-risk groups. RHD status was determined from echocardiographic reports or clinical diagnosis in charts. PPVs were compared by population risk status (high-risk/low-risk), age group, gender, principal/secondary diagnosis and probable/possible codes. RESULTS: High-risk patients had higher PPVs than low-risk patients (83.8% vs 54.9%, p<0.0001). PPVs were 91.5% and 51.5% respectively for algorithm-defined 'probable RHD' and 'possible' codes (p<0.0001). The PPVs in low-risk patients were higher for principal diagnoses than secondary diagnoses (84.5% vs 44.8%, weighted p<0.0001) but were similar in high-risk patients (92.5% vs 81.7%, p=0.096). CONCLUSION: The algorithm performs well for RHD coded as a principal diagnosis, 'probable' codes or in populations at high risk of RHD. Refinement is needed for identifying true RHD in low-risk groups.

Machine learning in heart failure: ready for prime time.

PURPOSE OF REVIEW: The aim of this review is to present an up-to-date overview of the application of machine learning methods in heart failure including diagnosis, classification, readmissions and medication adherence. RECENT FINDINGS: Recent studies have shown that the application of machine learning techniques may have the potential to improve heart failure outcomes and management, including cost savings by improving existing diagnostic and treatment support systems. Recently developed deep learning methods are expected to yield even better performance than traditional machine learning techniques in performing complex tasks by learning the intricate patterns hidden in big medical data. SUMMARY: The review summarizes the recent developments in the application of machine and deep learning methods in heart failure management.

Associations between Helicobacter pylori infection, co-morbid infections, gastrointestinal symptoms, and circulating cytokines in African children.

BACKGROUND: Refugee children have complex medical needs and often have multiple infections. The relationship between infection, gastrointestinal symptoms, and systemic inflammation is poorly understood. We investigated these parameters in refugee children with a high prevalence of Helicobacter pylori, helminth, and malaria infection. MATERIALS AND METHODS: African refugee children were recruited at resettlement health screening. Data were collected on demography, gastrointestinal symptoms, co-morbid infection, and serum for peripheral cytokine levels. Helicobacter pylori infection was diagnosed by a fecal-based immunoassay. RESULTS: Data from 163 children were analyzed, of which 84.0% were positive for H. pylori. Infected children were significantly older (9.2 years +/- 3.7 vs 7.1 years +/- 3.9, p = .01). Half the cohort (84/163, 51.5%) described gastrointestinal symptoms but these were not strongly associated with co-morbid infections. Helicobacter pylori-infected children had significantly lower circulating log-interleukin-8 (IL-8) (odds ratio 0.61, 95% confidence interval (CI) 0.40, 0.94, p = .025). Helminth infections were common (75/163, 46%) and associated with elevated log-IL-5 (beta: 0.42, 95% CI 0.077, 0.76). Children with malaria (15/163, 9.2%) had elevated log-tumor necrosis factor-alpha (TNFalpha) and log-IL-10 (beta: 0.67, 95% CI 0.34, 1.0 and beta: 1.3, 95% CI 0.67, 1.9, respectively). IL-10 : IL-12 ratios were increased in H. pylori-infected children with malaria or helminth infections. Symptoms were generally not associated with levels of circulating peripheral cytokines irrespective of co-morbid infection diagnosis. CONCLUSIONS: There is a high prevalence of asymptomatic H. pylori infection in recently resettled African refugee children. Gastrointestinal symptoms were not predictive of H. pylori nor of helminth infections. Serum cytokines, particularly IL-5, IL-10, and TNFalpha, were significantly elevated in children with malaria and helminth infections but not in those with H. pylori infection.

Machine learning-based prediction of heart failure readmission or death: implications of choosing the right model and the right metrics.

AIMS: Machine learning (ML) is widely believed to be able to learn complex hidden interactions from the data and has the potential in predicting events such as heart failure (HF) readmission and death. Recent studies have revealed conflicting results likely due to failure to take into account the class imbalance problem commonly seen with medical data. We developed a new ML approach to predict 30 day HF readmission or death and compared the performance of this model with other commonly used prediction models. METHODS AND RESULTS: We identified all Western Australian patients aged above 65 years admitted for HF between 2003 and 2008 in the linked Hospital Morbidity Data Collection. Taking into consideration the class imbalance problem, we developed a multi-layer perceptron (MLP)-based approach to predict 30 day HF readmission or death and compared the predictive performances using the performance metrics, that is, area under the receiver operating characteristic curve (AUC), area under the precision-recall curve (AUPRC), sensitivity and specificity with other ML and regression models. Out of the 10 757 patients with HF, 23.6% were readmitted or died within 30 days of hospital discharge. We observed an AUC of 0.55, 0.53, 0.58, and 0.54 while an AUPRC of 0.39, 0.38, 0.46, and 0.38 for weighted random forest, weighted decision trees, logistic regression, and weighted support vector machines models, respectively. The MLP-based approach produced the highest AUC (0.62) and AUPRC (0.46) with 48% sensitivity and 70% specificity. CONCLUSIONS: We show that for the medical data with class imbalance, the proposed MLP-based approach is superior to other ML and regression techniques for the prediction of 30 day HF readmission or death.