Intervention-driven predictive framework for modeling healthcare data

Assessing prognostic risk is crucial to clinical care, and critically dependent on both diagnosis and medical interventions. Current methods use this augmented information to build a single prediction rule. But this may not be expressive enough to capture differential effects of interventions on prognosis. To this end, we propose a supervised, Bayesian nonparametric framework that simultaneously discovers the latent intervention groups and builds a separate prediction rule for each intervention group. The prediction rule is learnt using diagnosis data through a Bayesian logistic regression. For inference, we develop an efficient collapsed Gibbs sampler. We demonstrate that our method outperforms baselines in predicting 30-day hospital readmission using two patient cohorts - Acute Myocardial Infarction and Pneumonia. The significance of this model is that it can be applied widely across a broad range of medical prognosis tasks. © 2014 Springer International Publishing.

Regularizing topic discovery in emrs with side information by using hierarchical bayesian models

We propose a novel hierarchical Bayesian framework, word-distance-dependent Chinese restaurant franchise (wd-dCRF) for topic discovery from a document corpus regularized by side information in the form of word-to-word relations, with an application on Electronic Medical Records (EMRs). Typically, a EMRs dataset consists of several patients (documents) and each patient contains many diagnosis codes (words). We exploit the side information available in the form of a semantic tree structure among the diagnosis codes for semantically-coherent disease topic discovery. We introduce novel functions to compute word-to-word distances when side information is available in the form of tree structures. We derive an efficient inference method for the wddCRF using MCMC technique. We evaluate on a real world medical dataset consisting of about 1000 patients with PolyVascular disease. Compared with the popular topic analysis tool, hierarchical Dirichlet process (HDP), our model discovers topics which are superior in terms of both qualitative and quantitative measures.

Is demography destiny? Application of machine learning techniques to accurately predict population health outcomes from a minimal demographic dataset

For years, we have relied on population surveys to keep track of regional public health statistics, including the prevalence of non-communicable diseases. Because of the cost and limitations of such surveys, we often do not have the up-to-date data on health outcomes of a region. In this paper, we examined the feasibility of inferring regional health outcomes from socio-demographic data that are widely available and timely updated through national censuses and community surveys. Using data for 50 American states (excluding Washington DC) from 2007 to 2012, we constructed a machine-learning model to predict the prevalence of six non-communicable disease (NCD) outcomes (four NCDs and two major clinical risk factors), based on population socio-demographic characteristics from the American Community Survey. We found that regional prevalence estimates for non-communicable diseases can be reasonably predicted. The predictions were highly correlated with the observed data, in both the states included in the derivation model (median correlation 0.88) and those excluded from the development for use as a completely separated validation sample (median correlation 0.85), demonstrating that the model had sufficient external validity to make good predictions, based on demographics alone, for areas not included in the model development. This highlights both the utility of this sophisticated approach to model development, and the vital importance of simple socio-demographic characteristics as both indicators and determinants of chronic disease.

Global, regional, and national incidence and mortality for HIV, tuberculosis, and malaria during 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013

BACKGROUND: The Millennium Declaration in 2000 brought special global attention to HIV, tuberculosis, and malaria through the formulation of Millennium Development Goal (MDG) 6. The Global Burden of Disease 2013 study provides a consistent and comprehensive approach to disease estimation for between 1990 and 2013, and an opportunity to assess whether accelerated progress has occured since the Millennium Declaration. METHODS: To estimate incidence and mortality for HIV, we used the UNAIDS Spectrum model appropriately modified based on a systematic review of available studies of mortality with and without antiretroviral therapy (ART). For concentrated epidemics, we calibrated Spectrum models to fit vital registration data corrected for misclassification of HIV deaths. In generalised epidemics, we minimised a loss function to select epidemic curves most consistent with prevalence data and demographic data for all-cause mortality. We analysed counterfactual scenarios for HIV to assess years of life saved through prevention of mother-to-child transmission (PMTCT) and ART. For tuberculosis, we analysed vital registration and verbal autopsy data to estimate mortality using cause of death ensemble modelling. We analysed data for corrected case-notifications, expert opinions on the case-detection rate, prevalence surveys, and estimated cause-specific mortality using Bayesian meta-regression to generate consistent trends in all parameters. We analysed malaria mortality and incidence using an updated cause of death database, a systematic analysis of verbal autopsy validation studies for malaria, and recent studies (2010-13) of incidence, drug resistance, and coverage of insecticide-treated bednets. FINDINGS: Globally in 2013, there were 1·8 million new HIV infections (95% uncertainty interval 1·7 million to 2·1 million), 29·2 million prevalent HIV cases (28·1 to 31·7), and 1·3 million HIV deaths (1·3 to 1·5). At the peak of the epidemic in 2005, HIV caused 1·7 million deaths (1·6 million to 1·9 million). Concentrated epidemics in Latin America and eastern Europe are substantially smaller than previously estimated. Through interventions including PMTCT and ART, 19·1 million life-years (16·6 million to 21·5 million) have been saved, 70·3% (65·4 to 76·1) in developing countries. From 2000 to 2011, the ratio of development assistance for health for HIV to years of life saved through intervention was US$4498 in developing countries. Including in HIV-positive individuals, all-form tuberculosis incidence was 7·5 million (7·4 million to 7·7 million), prevalence was 11·9 million (11·6 million to 12·2 million), and number of deaths was 1·4 million (1·3 million to 1·5 million) in 2013. In the same year and in only individuals who were HIV-negative, all-form tuberculosis incidence was 7·1 million (6·9 million to 7·3 million), prevalence was 11·2 million (10·8 million to 11·6 million), and number of deaths was 1·3 million (1·2 million to 1·4 million). Annualised rates of change (ARC) for incidence, prevalence, and death became negative after 2000. Tuberculosis in HIV-negative individuals disproportionately occurs in men and boys (versus women and girls); 64·0% of cases (63·6 to 64·3) and 64·7% of deaths (60·8 to 70·3). Globally, malaria cases and deaths grew rapidly from 1990 reaching a peak of 232 million cases (143 million to 387 million) in 2003 and 1·2 million deaths (1·1 million to 1·4 million) in 2004. Since 2004, child deaths from malaria in sub-Saharan Africa have decreased by 31·5% (15·7 to 44·1). Outside of Africa, malaria mortality has been steadily decreasing since 1990. INTERPRETATION: Our estimates of the number of people living with HIV are 18·7% smaller than UNAIDS's estimates in 2012. The number of people living with malaria is larger than estimated by WHO. The number of people living with HIV, tuberculosis, or malaria have all decreased since 2000. At the global level, upward trends for malaria and HIV deaths have been reversed and declines in tuberculosis deaths have accelerated. 101 countries (74 of which are developing) still have increasing HIV incidence. Substantial progress since the Millennium Declaration is an encouraging sign of the effect of global action. FUNDING: Bill & Melinda Gates Foundation.

Global, regional, and national levels and causes of maternal mortality during 1990-2013: a systematic analysis for the Global Burden of Disease Study 2013

BACKGROUND: The fifth Millennium Development Goal (MDG 5) established the goal of a 75% reduction in the maternal mortality ratio (MMR; number of maternal deaths per 100,000 livebirths) between 1990 and 2015. We aimed to measure levels and track trends in maternal mortality, the key causes contributing to maternal death, and timing of maternal death with respect to delivery. METHODS: We used robust statistical methods including the Cause of Death Ensemble model (CODEm) to analyse a database of data for 7065 site-years and estimate the number of maternal deaths from all causes in 188 countries between 1990 and 2013. We estimated the number of pregnancy-related deaths caused by HIV on the basis of a systematic review of the relative risk of dying during pregnancy for HIV-positive women compared with HIV-negative women. We also estimated the fraction of these deaths aggravated by pregnancy on the basis of a systematic review. To estimate the numbers of maternal deaths due to nine different causes, we identified 61 sources from a systematic review and 943 site-years of vital registration data. We also did a systematic review of reports about the timing of maternal death, identifying 142 sources to use in our analysis. We developed estimates for each country for 1990-2013 using Bayesian meta-regression. We estimated 95% uncertainty intervals (UIs) for all values. FINDINGS: 292,982 (95% UI 261,017-327,792) maternal deaths occurred in 2013, compared with 376,034 (343,483-407,574) in 1990. The global annual rate of change in the MMR was -0·3% (-1·1 to 0·6) from 1990 to 2003, and -2·7% (-3·9 to -1·5) from 2003 to 2013, with evidence of continued acceleration. MMRs reduced consistently in south, east, and southeast Asia between 1990 and 2013, but maternal deaths increased in much of sub-Saharan Africa during the 1990s. 2070 (1290-2866) maternal deaths were related to HIV in 2013, 0·4% (0·2-0·6) of the global total. MMR was highest in the oldest age groups in both 1990 and 2013. In 2013, most deaths occurred intrapartum or postpartum. Causes varied by region and between 1990 and 2013. We recorded substantial variation in the MMR by country in 2013, from 956·8 (685·1-1262·8) in South Sudan to 2·4 (1·6-3·6) in Iceland. INTERPRETATION: Global rates of change suggest that only 16 countries will achieve the MDG 5 target by 2015. Accelerated reductions since the Millennium Declaration in 2000 coincide with increased development assistance for maternal, newborn, and child health. Setting of targets and associated interventions for after 2015 will need careful consideration of regions that are making slow progress, such as west and central Africa. FUNDING: Bill & Melinda Gates Foundation.

HealthMap: a visual platform for patient suicide risk review

Misjudging suicide risk can be fatal. Risk assessment is complicated by multiplicity of risk factors, none of which individually can reliably predict risk. This paper addresses the need for better clinical support, visualising risk factors scattered in raw electronic medical records. HealthMap is a visual tool that helps clinicians effectively examine patient histories during a suicide risk assessment. We characterise the information visualisation problems accompanying suicide risk assessments. A design driven by visualisation principles was implemented. The prototype was evaluated by clinicians and accepted into daily clinical work-flow.

The role of the middle class in economic development: What do Cross-Country data show?

This paper investigates the channels through which the middle class may matter for consumption growth and development. Determinants of the size and the growth of the middle class are also examined. Using several different middle class measures and a panel of 72 developing countries spanning the period 1985-2006, we find that a larger middle class influences growth primarily through higher levels of human capital investment. We also find that large governments, higher levels of urbanization, greater democracy, ethnic concentration, and sea access are all associated with a larger middle class. © 2011 by Asian Development Bank.

Web search activity data accurately predict population chronic disease risk in the USA

BACKGROUND: The WHO framework for non-communicable disease (NCD) describes risks and outcomes comprising the majority of the global burden of disease. These factors are complex and interact at biological, behavioural, environmental and policy levels presenting challenges for population monitoring and intervention evaluation. This paper explores the utility of machine learning methods applied to population-level web search activity behaviour as a proxy for chronic disease risk factors. METHODS: Web activity output for each element of the WHO's Causes of NCD framework was used as a basis for identifying relevant web search activity from 2004 to 2013 for the USA. Multiple linear regression models with regularisation were used to generate predictive algorithms, mapping web search activity to Centers for Disease Control and Prevention (CDC) measured risk factor/disease prevalence. Predictions for subsequent target years not included in the model derivation were tested against CDC data from population surveys using Pearson correlation and Spearman's r. RESULTS: For 2011 and 2012, predicted prevalence was very strongly correlated with measured risk data ranging from fruits and vegetables consumed (r=0.81; 95% CI 0.68 to 0.89) to alcohol consumption (r=0.96; 95% CI 0.93 to 0.98). Mean difference between predicted and measured differences by State ranged from 0.03 to 2.16. Spearman's r for state-wise predicted versus measured prevalence varied from 0.82 to 0.93. CONCLUSIONS: The high predictive validity of web search activity for NCD risk has potential to provide real-time information on population risk during policy implementation and other population-level NCD prevention efforts.

Exploiting side information in Bayesian nonparametric models and their applications

 My research is to exploit side information into advanced Bayesian nonparametric models. We have developed some novel models for data clustering and medical data analysis and also have made our methods scalable for large-scale data. I have published my research in several journal and conference papers.

A predictive framework for modeling healthcare data with evolving clinical interventions

Medical interventions critically determine clinical outcomes. But prediction models either ignore interventions or dilute impact by building a single prediction rule by amalgamating interventions with other features. One rule across all interventions may not capture differential effects. Also, interventions change with time as innovations are made, requiring prediction models to evolve over time. To address these gaps, we propose a prediction framework that explicitly models interventions by extracting a set of latent intervention groups through a Hierarchical Dirichlet Process (HDP) mixture. Data are split in temporal windows and for each window, a separate distribution over the intervention groups is learnt. This ensures that the model evolves with changing interventions. The outcome is modeled as conditional, on both the latent grouping and the patients' condition, through a Bayesian logistic regression. Learning distributions for each time-window result in an over-complex model when interventions do not change in every time-window. We show that by replacing HDP with a dynamic HDP prior, a more compact set of distributions can be learnt. Experiments performed on two hospital datasets demonstrate the superiority of our framework over many existing clinical and traditional prediction frameworks.

The role of middle class in democratic diffusion

The modernization hypothesis and the democratic domino theory have been at the forefront in explaining the democratization around the globe. This paper empirically investigates the ‘middle class-driven modernization’ hypothesis and the ‘middle class-driven democratic domino’ effect in a panel of 145 countries over the period 1985 to 2013. Using several middle class measures and a dynamic panel estimator, we show that the ‘middle class-driven modernization’ hypothesis finds strong empirical support in the sample of developing countries excluding Eastern Europe and Central Asia, while the ‘middle class-driven democratic domino’ effect finds support in the sample of developing countries excluding East Asia and the Pacific

Data clustering using side information dependent Chinese restaurant processes

Side information, or auxiliary information associated with documents or image content, provides hints for clustering. We propose a new model, side information dependent Chinese restaurant process, which exploits side information in a Bayesian nonparametric model to improve data clustering. We introduce side information into the framework of distance dependent Chinese restaurant process using a robust decay function to handle noisy side information. The threshold parameter of the decay function is updated automatically in the Gibbs sampling process. A fast inference algorithm is proposed. We evaluate our approach on four datasets: Cora, 20 Newsgroups, NUS-WIDE and one medical dataset. Types of side information explored in this paper include citations, authors, tags, keywords and auxiliary clinical information. The comparison with the state-of-the-art approaches based on standard performance measures (NMI, F1) clearly shows the superiority of our approach.

Differentially private random forest with high utility

Privacy-preserving data mining has become an active focus of the research community in the domains where data are sensitive and personal in nature. For example, highly sensitive digital repositories of medical or financial records offer enormous values for risk prediction and decision making. However, prediction models derived from such repositories should maintain strict privacy of individuals. We propose a novel random forest algorithm under the framework of differential privacy. Unlike previous works that strictly follow differential privacy and keep the complete data distribution approximately invariant to change in one data instance, we only keep the necessary statistics (e.g. variance of the estimate) invariant. This relaxation results in significantly higher utility. To realize our approach, we propose a novel differentially private decision tree induction algorithm and use them to create an ensemble of decision trees. We also propose feasible adversary models to infer about the attribute and class label of unknown data in presence of the knowledge of all other data. Under these adversary models, we derive bounds on the maximum number of trees that are allowed in the ensemble while maintaining privacy. We focus on binary classification problem and demonstrate our approach on four real-world datasets. Compared to the existing privacy preserving approaches we achieve significantly higher utility.