Hypertension screening, prevalence, treatment, and control at a large private hospital in Kampala, Uganda: A retrospective analysis.

Adult hypertension prevalence in Uganda is 27%, but only 8% are aware of their diagnosis, accordingly treatment and control levels are limited. The private sector provides at least half of care nationwide, but little is known about its effectiveness in hypertension control. We analyzed clinical data from 39 235 outpatient visits among 17 777 adult patients from July 2017 to August 2018 at Uganda's largest private hospital. We calculated blood pressure screening rate at every visit, and hypertension prevalence, medication treatment, and control rates among the 5 090 patients with two or more blood pressure checks who received any medications from the hospital's pharmacy. We defined hypertension in this group as 1) an average of two blood pressure measurements at separate consecutive visits, higher than 140 mm Hg systolic or 90 mm Hg diastolic, 2) receipt of any antihypertensive medication, or 3) the use of a hypertension electronic medical record code. We deemed hypertension control as normotensive at the most recent check. 12 821 (72.1%) of patients received at least 1 blood pressure check. Among the 5 090 patients above, 2 121 (41.6%) had hypertension (33.4% age-standardized to a world population standard): 1 915 (37.6%) with elevated blood pressure, and 170 (3.3%) were normotensive but receiving medication. 838 (39.4%) of patients with hypertension received medication at least once. Overall, 18.3% of patients achieved control (27% of treated patients, and 15% of untreated patients). Hypertension is common and incompletely controlled in this Ugandan private-sector population, suggesting several avenues for novel interventions.

Machine Learning Methods for Precision Medicine Research Designed to Reduce Health Disparities: A Structured Tutorial.

Precision medicine research designed to reduce health disparities often involves studying multi-level datasets to understand how diseases manifest disproportionately in one group over another, and how scarce health care resources can be directed precisely to those most at risk for disease. In this article, we provide a structured tutorial for medical and public health researchers on the application of machine learning methods to conduct precision medicine research designed to reduce health disparities. We review key terms and concepts for understanding machine learning papers, including supervised and unsupervised learning, regularization, cross-validation, bagging, and boosting. Metrics are reviewed for evaluating machine learners and major families of learning approaches, including tree-based learning, deep learning, and ensemble learning. We highlight the advantages and disadvantages of different learning approaches, describe strategies for interpreting "black box" models, and demonstrate the application of common methods in an example dataset with open-source statistical code in R.

Multiple chronic conditions at a major urban health system: a retrospective cross-sectional analysis of frequencies, costs and comorbidity patterns.

OBJECTIVE: To (1) examine the burden of multiple chronic conditions (MCC) in an urban health system, and (2) propose a methodology to identify subpopulations of interest based on diagnosis groups and costs. DESIGN: Retrospective cross-sectional study. SETTING: Mount Sinai Health System, set in all five boroughs of New York City, USA. PARTICIPANTS: 192 085 adult (18+) plan members of capitated Medicaid contracts between the Healthfirst managed care organisation and the Mount Sinai Health System in the years 2012 to 2014. METHODS: We classified adults as having 0, 1, 2, 3, 4 or 5+ chronic conditions from a list of 69 chronic conditions. After summarising the demographics, geography and prevalence of MCC within this population, we then described groups of patients (segments) using a novel methodology: we combinatorially defined 18 768 potential segments of patients by a pair of chronic conditions, a sex and an age group, and then ranked segments by (1) frequency, (2) cost and (3) ratios of observed to expected frequencies of co-occurring chronic conditions. We then compiled pairs of conditions that occur more frequently together than otherwise expected. RESULTS: 61.5% of the study population suffers from two or more chronic conditions. The most frequent dyad was hypertension and hyperlipidaemia (19%) and the most frequent triad was diabetes, hypertension and hyperlipidaemia (10%). Women aged 50 to 65 with hypertension and hyperlipidaemia were the leading cost segment in the study population. Costs and prevalence of MCC increase with number of conditions and age. The disease dyads associated with the largest observed/expected ratios were pulmonary disease and myocardial infarction. Inter-borough range MCC prevalence was 16%. CONCLUSIONS: In this low-income, urban population, MCC is more prevalent (61%) than nationally (42%), motivating further research and intervention in this population. By identifying potential target populations in an interpretable manner, this segmenting methodology has utility for health services analysts.

Satellite images and machine learning can identify remote communities to facilitate access to health services.

OBJECTIVE: Community health systems operating in remote areas require accurate information about where people live to efficiently provide services across large regions. We sought to determine whether a machine learning analyses of satellite imagery can be used to map remote communities to facilitate service delivery and planning. MATERIALS AND METHODS: We developed a method for mapping communities using a deep learning approach that excels at detecting objects within images. We trained an algorithm to detect individual buildings, then examined building clusters to identify groupings suggestive of communities. The approach was validated in southeastern Liberia, by comparing algorithmically generated results with community location data collected manually by enumerators and community health workers. RESULTS: The deep learning approach achieved 86.47% positive predictive value and 79.49% sensitivity with respect to individual building detection. The approach identified 75.67% (n = 451) of communities registered through the community enumeration process, and identified an additional 167 potential communities not previously registered. Several instances of false positives and false negatives were identified. DISCUSSION: Analysis of satellite images is a promising solution for mapping remote communities rapidly, and with relatively low costs. Further research is needed to determine whether the communities identified algorithmically, but not registered in the manual enumeration process, are currently inhabited. CONCLUSIONS: To our knowledge, this study represents the first effort to apply image recognition algorithms to rural healthcare delivery. Results suggest that these methods have the potential to enhance community health worker scale-up efforts in underserved remote communities.

Machine Learning for Health Services Researchers.

BACKGROUND: Machine learning is increasingly used to predict healthcare outcomes, including cost, utilization, and quality. OBJECTIVE: We provide a high-level overview of machine learning for healthcare outcomes researchers and decision makers. METHODS: We introduce key concepts for understanding the application of machine learning methods to healthcare outcomes research. We first describe current standards to rigorously learn an estimator, which is an algorithm developed through machine learning to predict a particular outcome. We include steps for data preparation, estimator family selection, parameter learning, regularization, and evaluation. We then compare 3 of the most common machine learning methods: (1) decision tree methods that can be useful for identifying how different subpopulations experience different risks for an outcome; (2) deep learning methods that can identify complex nonlinear patterns or interactions between variables predictive of an outcome; and (3) ensemble methods that can improve predictive performance by combining multiple machine learning methods. RESULTS: We demonstrate the application of common machine methods to a simulated insurance claims dataset. We specifically include statistical code in R and Python for the development and evaluation of estimators for predicting which patients are at heightened risk for hospitalization from ambulatory care-sensitive conditions. CONCLUSIONS: Outcomes researchers should be aware of key standards for rigorously evaluating an estimator developed through machine learning approaches. Although multiple methods use machine learning concepts, different approaches are best suited for different research problems.

Assessment of Risk of Harm Associated With Intensive Blood Pressure Management Among Patients With Hypertension Who Smoke: A Secondary Analysis of the Systolic Blood Pressure Intervention Trial.

Importance: The randomized Systolic Blood Pressure Intervention Trial (SPRINT) showed that lowering systolic blood pressure targets for adults with hypertension reduces cardiovascular morbidity and mortality in general. However, whether the overall benefit from intensive blood pressure control masks important heterogeneity in risk is unknown. Objective: To test the hypothesis that the overall benefit observed in SPRINT masked important heterogeneity in risk from intensive blood pressure control. Design, Setting, and Participants: In this exploratory, hypothesis-generating, ad hoc, secondary analysis of data obtained from 9361 participants in SPRINT, a random forest-based analysis was used to identify potential heterogeneous treatment effects using half of the trial data. Cox proportional hazards regression models were applied to test potential heterogeneous treatment effects on the remaining data. The original trial was conducted at 102 sites in the United States between November 2010 and March 2013. This analysis was conducted between November 2016 and August 2017. Interventions: Participants were assigned a systolic blood pressure target of less than 120 mm Hg (intervention treatment) or of less than 140 mm Hg (standard treatment). Main Outcomes and Measures: The primary composite cardiovascular outcome was myocardial infarction, other acute coronary syndromes, stroke, heart failure, or death from cardiovascular causes. Results: Of 9361 participants in SPRINT, 466 participants (5.0%) were current smokers with systolic blood pressure greater than 144 mm Hg at baseline, with 230 participants (49.4%) randomized to the training data set and 236 participants (50.6%) randomized to the testing data set; 286 participants (61.4%) were male, and the mean (SD) age was 60.7 (7.2) years. Combinations of 2 covariates (ie, baseline smoking status and systolic blood pressure) distinguished participants who were differentially affected by the intervention. In the testing data, Cox proportional hazards models for the primary outcome revealed a number needed to harm of 43.7 to cause 1 event across 3.3 years among participants who, at baseline, were current smokers with systolic blood pressure greater than 144 mm Hg (10.9% [12 of 110] of primary outcome events for intervention treatment vs 4.8% [6 of 126] for standard treatment; hazard ratio, 10.6; 95% CI, 1.3-86.1; P = .03). This subgroup was also associated with a higher likelihood to experience acute kidney injury under intensive blood pressure control (with a frequency of 10.0% [11 of 110] of acute kidney injury events for intervention treatment vs 3.2% [4 of 126] for standard treatment; hazard ratio, 9.4; 95% CI, 1.2-77.3; P = .04). Conclusions and Relevance: In this secondary analysis of SPRINT data, current smokers with a baseline systolic blood pressure greater than 144 mm Hg had a higher rate of cardiovascular events in the intensive treatment group vs the standard treatment group. Further research is needed to evaluate the potential tradeoffs of intensive blood pressure control in hypertensive smokers.