Abstract
BackgroundConverting electronic health record (EHR) entries to useful clinical inferences requires one to address the poor scalability of existing implementations of Generalized Linear Mixed Models (GLMM) for repeated measures. The major computational bottleneck concerns the numerical evaluation of multivariable integrals, which even for the simplest EHR analyses may involve millions of dimensions (one for each patient). The hierarchical likelihood (h-lik) approach to GLMMs is a methodologically rigorous framework for the estimation of GLMMs that is based on the Laplace Approximation (LA), which replaces integration with numerical optimization, and thus scales very well with dimensionality.MethodsWe present a high-performance, direct implementation of the h-lik for GLMMs in the R package TMB. Using this approach, we examined the relation of repeated serum potassium measurements and survival in the Cerner Real World Data (CRWD) EHR database. Analyzing this data requires the evaluation of an integral in over 3 million dimensions, putting this problem beyond the reach of conventional approaches. We also assessed the scalability and accuracy of LA in smaller samples of 1 and 10% size of the full dataset that were analyzed via the a) original, interconnected Generalized Linear Models (iGLM), approach to h-lik, b) Adaptive Gaussian Hermite (AGH) and c) the gold standard for multivariate integration Markov Chain Monte Carlo (MCMC).ResultsRandom effects estimates generated by the LA were within 10% of the values obtained by the iGLMs, AGH and MCMC techniques. The H-lik approach was 4–30 times faster than AGH and nearly 800 times faster than MCMC. The major clinical inferences in this problem are the establishment of the non-linear relationship between the potassium level and the risk of mortality, as well as estimates of the individual and health care facility sources of variations for mortality risk in CRWD.ConclusionsWe found that the direct implementation of the h-lik offers a computationally efficient, numerically accurate approach for the analysis of extremely large, real world repeated measures data via the h-lik approach to GLMMs. The clinical inference from our analysis may guide choices of treatment thresholds for treating potassium disorders in the clinic.
Highlights
Converting electronic health record (EHR) entries to useful clinical inferences requires one to address the poor scalability of existing implementations of Generalized Linear Mixed Models (GLMM) for repeated measures
We found that the direct implementation of the h-lik offers a computationally efficient, numerically accurate approach for the analysis of extremely large, real world repeated measures data via the h-lik approach to GLMMs
Using simulations and real world big data, we show that the results obtained by our implementation are very similar to those obtained via theoretically more accurate techniques for GLMM estimation, i.e. Adaptive Gaussian Hermite (AGH) or Markov Chain Monte Carlo (MCMC) or even the original interconnected Gamma Model implementation of the h-lik [12]
Summary
Electronic Health Records (EHR) have been adopted near universally in medical practices across the United States They contain information about vital signs, lab results, medical procedures, diagnoses, medications, admissions, healthcare facility features (e.g., type, location and practice) and individual subject outcomes (e.g., death or hospitalizations). Existing approaches to big data such as deep learning [5] overlook this feature of EHR data, missing on a unique opportunity to mine variation at the level of IP/, HCPs or HCFs. Subject or healthcare facility focused inference requires the deployment of Generalized Linear Mixed Models (GLMM) for the analyses of repeated measures at the group (IP/HCP/HCF) level. GLMMs can tackle the entire spectrum of questions that a clinical researcher would want to examine using EHR data Such questions typically involve the analyses of continuous (e.g., biomarker values, vital signs), discrete (e.g., development of specific diagnoses), time to event (e.g., survival) or joint biomarker-discrete/time-event data
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
