Learning probabilistic phenotypes from heterogeneous EHR data

Objective To outline new design directions for informatics solutions that facilitate personal discovery with self-monitoring data. We investigate this question in the context of chronic disease self-management with the focus on type 2 diabetes. Materials and methods We conducted an observational qualitative study of discovery with personal data among adults attending a diabetes self-management education (DSME) program that utilized a discovery-based curriculum. The study included observations of class sessions, and interviews and focus groups with the educator and attendees of the program (n = 14). Results The main discovery in diabetes self-management evolved around discovering patterns of association between characteristics of individuals’ activities and changes in their blood glucose levels that the participants referred to as “cause and effect”. This discovery empowered individuals to actively engage in self-management and provided a desired flexibility in selection of personalized self-management strategies. We show that discovery of cause and effect involves four essential phases: (1) feature selection, (2) hypothesis generation, (3) feature evaluation, and (4) goal specification. Further, we identify opportunities to support discovery at each stage with informatics and data visualization solutions by providing assistance with: (1) active manipulation of collected data (e.g., grouping, filtering and side-by-side inspection), (2) hypotheses formulation (e.g., using natural language statements or constructing visual queries), (3) inference evaluation (e.g., through aggregation and visual comparison, and statistical analysis of associations), and (4) translation of discoveries into actionable goals (e.g., tailored selection from computable knowledge sources of effective diabetes self-management behaviors). Discussion The study suggests that discovery of cause and effect in diabetes can be a powerful approach to helping individuals to improve their self-management strategies, and that self-monitoring data can serve as a driving engine for personal discovery that may lead to sustainable behavior changes. Conclusions Enabling personal discovery is a promising new approach to enhancing chronic disease self-management with informatics interventions.

Section: Background and Significancementioning

confidence: 99%

Personal discovery in diabetes self-management: Discovering cause and effect using self-monitoring data

Mamykina

Heitkemper

Smaldone

et al. 2017

“…One of the points of the PopKLD algorithm to create a reasonable, interpretable, stable, automatable, ordinal summary of laboratory data that can be used in high-throughput situations where machine learning [17,16,15] is used to categorize humans. Based on the results from the clinical evaluation, we believe that PopKLD will be very useful in phenotyping studies.…”

Section: Discussionmentioning

confidence: 99%

Estimating summary statistics for electronic health record laboratory data for use in high-throughput phenotyping algorithms

Albers

Elhadad

Claassen

et al. 2018

We study the question of how to represent or summarize raw laboratory data taken from an electronic health record (EHR) using parametric model selection to reduce or cope with biases induced through clinical care. It has been previously demonstrated that the health care process (Hripcsak and Albers, 2012, 2013), as defined by measurement context (Hripcsak and Albers, 2013; Albers et al., 2012) and measurement patterns (Albers and Hripcsak, 2010, 2012), can influence how EHR data are distributed statistically (Kohane and Weber, 2013; Pivovarov et al., 2014). We construct an algorithm, PopKLD, which is based on information criterion model selection (Burnham and Anderson, 2002; Claeskens and Hjort, 2008), is intended to reduce and cope with health care process biases and to produce an intuitively understandable continuous summary. The PopKLD algorithm can be automated and is designed to be applicable in high-throughput settings; for example, the output of the PopKLD algorithm can be used as input for phenotyping algorithms. Moreover, we develop the PopKLD-CAT algorithm that transforms the continuous PopKLD summary into a categorical summary useful for applications that require categorical data such as topic modeling. We evaluate our methodology in two ways. First, we apply the method to laboratory data collected in two different health care contexts, primary versus intensive care. We show that the PopKLD preserves known physiologic features in the data that are lost when summarizing the data using more common laboratory data summaries such as mean and standard deviation. Second, for three disease-laboratory measurement pairs, we perform a phenotyping task: we use the PopKLD and PopKLD-CAT algorithms to define high and low values of the laboratory variable that are used for defining a disease state. We then compare the relationship between the PopKLD-CAT summary disease predictions and the same predictions using empirically estimated mean and standard deviation to a gold standard generated by clinical review of patient records. We find that the PopKLD laboratory data summary is substantially better at predicting disease state. The PopKLD or PopKLD-CAT algorithms are not meant to be used as phenotyping algorithms, but we use the phenotyping task to show what information can be gained when using a more informative laboratory data summary. In the process of evaluation our method we show that the different clinical contexts and laboratory measurements necessitate different statistical summaries. Similarly, leveraging the principle of maximum entropy we argue that while some laboratory data only have sufficient information to estimate a mean and standard deviation, other laboratory data captured in an EHR contain substantially more information than can be captured in higher-parameter models.

“…modeling phenotypes from anchor variables [44,45] and silver-standard training data [46]. We also note that from the perspective of learning shared representations of diseases (such as the abstraction feature representation in this study), contemporary phenotyping effort has led to a growing body of work that learns phenotypes from population-scale clinical data using the methodology of representation learning [47,48] including i) spectral learning such as non-negative tensor factorization [5], ii) probabilistic mixture models [8], and additionally, when temporal phenotypic patterns are considered, iii) unsupervised feature learning using autoencoders [32] and latent medical concepts [49], etc., and iv) deep learning [6]. …”

Section: Discussionmentioning

confidence: 99%

“…In the general landscape of computational phenotyping research, many endeavors have been made to progressively replace the predominant use of rule-based phenotyping algorithms, comprising the majority in eMERGE network [4], by the predictive analytical approach based on machine learning and natural language processing (NLP) [5–8]. While this data-driven analytic methodology alleviates the tedious process of manually selecting features and their logical combinations that match phenotype definitions in an ad hoc fashion, predictive analytics is not without its own outstanding challenges.…”

Section: Introductionmentioning

confidence: 99%

EHR-based phenotyping: Bulk learning and evaluation

Chiu

Hripcsak

2017

In data-driven phenotyping, a core computational task is to identify medical concepts and their variations from sources of electronic health records (EHR) to stratify phenotypic cohorts. A conventional analytic framework for phenotyping largely uses a manual knowledge engineering approach or a supervised learning approach where clinical cases are represented by variables encompassing diagnoses, medicinal treatments and laboratory tests, among others. In such a framework, tasks associated with feature engineering and data annotation remain a tedious and expensive exercise, resulting in poor scalability. In addition, certain clinical conditions, such as those that are rare and acute in nature, may never accumulate sufficient data over time, which poses a challenge to establishing accurate and informative statistical models. In this paper, we use infectious diseases as the domain of study to demonstrate a hierarchical learning method based on ensemble learning that attempts to address these issues through feature abstraction. We use a sparse annotation set to train and evaluate many phenotypes at once, which we call bulk learning. In this batch-phenotyping framework, disease cohort definitions can be learned from within the abstract feature space established by using multiple diseases as a substrate and diagnostic codes as surrogates. In particular, using surrogate labels for model training renders possible its subsequent evaluation using only a sparse annotated sample. Moreover, statistical models can be trained and evaluated, using the same sparse annotation, from within the abstract feature space of low dimensionality that encapsulates the shared clinical traits of these target diseases, collectively referred to as the bulk learning set.