Electronic health records (EHR) are rich heterogeneous collections of patient health information, whose broad adoption provides clinicians and researchers unprecedented opportunities for health informatics, disease-risk prediction, actionable clinical recommendations, and precision medicine. However, EHRs present several modeling challenges, including highly sparse data matrices, noisy irregular clinical notes, arbitrary biases in billing code assignment, diagnosis-driven lab tests, and heterogeneous data types. To address these challenges, we present MixEHR, a multi-view Bayesian topic model. We demonstrate MixEHR on MIMIC-III, Mayo Clinic Bipolar Disorder, and Quebec Congenital Heart Disease EHR datasets. Qualitatively, MixEHR disease topics reveal meaningful combinations of clinical features across heterogeneous data types. Quantitatively, we observe superior prediction accuracy of diagnostic codes and lab test imputations compared to the state-of-art methods. We leverage the inferred patient topic mixtures to classify target diseases and predict mortality of patients in critical conditions. In all comparison, MixEHR confers competitive performance and reveals meaningful disease-related topics.

Purpose: Despite potential clinical benefits, implementation of pharmacogenomics (PGx) faces many technical and clinical challenges. These challenges can be overcome with a comprehensive and systematic implementation model. Methods: The development and implementation of PGx were organized into eight interdependent components addressing resources, governance, clinical practice, education, testing, knowledge translation, clinical decision support (CDS), and maintenance. Several aspects of implementation were assessed, including adherence to the model, production of PGx-CDS interventions, and access to educational resources. Results: Between August 2012 and June 2015, 21 specific drug–gene interactions were reviewed and 18 of them were implemented in the electronic medical record as PGx-CDS interventions. There was complete adherence to the model with variable production time (98–392 days) and delay time (0–148 days). The implementation impacted approximately 1,247 unique providers and 3,788 unique patients. A total of 11 educational resources complementary to the drug–gene interactions and 5 modules specific for pharmacists were developed and implemented. Conclusion: A comprehensive operational model can support PGx implementation in routine prescribing. Institutions can use this model as a roadmap to support similar efforts. However, we also identified challenges that will require major multidisciplinary and multi-institutional efforts to make PGx a universal reality. Genet Med 19 4, 421–429.

Introduction: Several common breast cancer genetic susceptibility variants have recently been identified. We aimed to determine how these variants combine with a subset of other known risk factors to influence breast cancer risk in white women of European ancestry using case-control studies participating in the Breast Cancer Association Consortium. Methods: We evaluated two-way interactions between each of age at menarche, ever having had a live birth, number of live births, age at first birth and body mass index (BMI) and each of 12 single nucleotide polymorphisms (SNPs) (10q26-rs2981582 (FGFR2), 8q24-rs13281615, 11p15-rs3817198 (LSP1), 5q11-rs889312 (MAP3K1), 16q12-rs3803662 (TOX3), 2q35-rs13387042, 5p12-rs10941679 (MRPS30), 17q23-rs6504950 (COX11), 3p24-rs4973768 (SLC4A7), CASP8-rs17468277, TGFB1-rs1982073 and ESR1-rs3020314). Interactions were tested for by fitting logistic regression models including per-allele and linear trend main effects for SNPs and risk factors, respectively, and single-parameter interaction terms for linear departure from independent multiplicative effects. Results: These analyses were applied to data for up to 26,349 invasive breast cancer cases and up to 32,208 controls from 21 case-control studies. No statistical evidence of interaction was observed beyond that expected by chance. Analyses were repeated using data from 11 population-based studies, and results were very similar. Conclusions: The relative risks for breast cancer associated with the common susceptibility variants identified to date do not appear to vary across women with different reproductive histories or body mass index (BMI). The assumption of multiplicative combined effects for these established genetic and other risk factors in risk prediction models appears justified.

We provide a mini‐review of how biobanks can support clinical genetics in the era of personalized medicine. We discuss types of biobanks, including disease specific and general biobanks not focused on one disease. We present considerations in setting up a biobank, including consenting and governance, biospecimens, risk factor and related data, informatics, and linkage to electronic health records for phenotyping. We also discuss the uses of biobanks and ongoing considerations, including genotype‐driven recruitment, investigations of gene–environment associations, and the re‐use of data generated from studies. Finally, we present a brief discussion of some of the unresolved issues, such as return of research results and sustaining biobanks over time. In summary, carefully designed biobanks can provide critical research and infrastructure support for clinical genetics in the era of personalized medicine.

Purpose: CHEK2*1100delC is a founder variant in European populations that confers a two- to threefold increased risk of breast cancer (BC). Epidemiologic and family studies have suggested that the risk associated with CHEK2*1100delC is modified by other genetic factors in a multiplicative fashion. We have investigated this empirically using data from the Breast Cancer Association Consortium (BCAC). Methods: Using genotype data from 39,139 (624 1100delC carriers) BC patients and 40,063 (224) healthy controls from 32 BCAC studies, we analyzed the combined risk effects of CHEK2*1100delC and 77 common variants in terms of a polygenic risk score (PRS) and pairwise interaction. Results: The PRS conferred odds ratios (OR) of 1.59 (95% CI: 1.21–2.09) per standard deviation for BC for CHEK2*1100delC carriers and 1.58 (1.55–1.62) for noncarriers. No evidence of deviation from the multiplicative model was found. The OR for the highest quintile of the PRS was 2.03 (0.86–4.78) for CHEK2*1100delC carriers, placing them in the high risk category according to UK NICE guidelines. The OR for the lowest quintile was 0.52 (0.16–1.74), indicating a lifetime risk close to the population average. Conclusion: Our results confirm the multiplicative nature of risk effects conferred by CHEK2*1100delC and the common susceptibility variants. Furthermore, the PRS could identify carriers at a high lifetime risk for clinical actions. Genet Med advance online publication 06 October 2016