Dataset Information

PS3-14: CREX: Utility of a Computerized Methodology to Identify Health Conditions Using the EMR for GWAS, in the Kaiser Permanente Research Program on Genes, Environment, and Health.

ABSTRACT: Background/AimsThe Genetic Epidemiology Research on Adult Health and Aging (GERA) Cohort has genotyped data on over 100,000 participants. In order to characterize many health conditions of interest found among respondents for sample-size estimation and GWAS, we sought to extend methodologies that utilized the Electronic Medical Record (EMR) in an automated way to characterize many diseases without the expense and limitation of numerous, complex algorithms specific to individual diseases and conditions.MethodsWe tested a probabilistic approach that scored clinical decisions recorded in the EMR to capture specific diagnostic and treatment domains. We first considered physician diagnosis alone. We assessed sensitivity and specificity against internal registries at the Kaiser Permanente Northern California Division of Research. We also used the methodology to characterize phenotypes for Types 1 and 2 diabetes for GWAS. We tested a single diagnostic domain in a logistic model based on an ICD-9-CM taxonomy and found high sensitivity and specificity when compared with internal registries for breast, lung, colon and prostate cancers; Barrett’s esophagus; HIV; Crohn’s disease; ulcerative colitis, and diabetes. We then assessed a logistic regression model to distinguish among members with Types 1 and 2 diabetes in the genotyped cohort, utilizing the ICD-9 taxonomy, the earliest age at diagnosis available in the EMR or by self-report, and pharmacy prescription utilization of anti-diabetic drugs.ResultsThe model exhibited a sensitivity of 96.3% and specificity of 99.6% for Type 1 diabetes and a sensitivity of 94.0% and specificity of 98.4% for Type 2 diabetes when compared with the gold standard internal diabetes registry. We identified an additional 60 cases of Type 1 diabetes and conducted comparative GWAS.ConclusionsUtilizing diagnostic information in the EMR as independent domains in probabilistic models to accomplish phenotype creation appears to be a reliable approach to facilitate robust characterization for evaluation, analysis and mapping of numerous disease phenotypes. The method is agnostic to input taxonomies as long as the EMR record contains sufficient and reliable atomic detail. It can also be adapted for machine learning given expert user feedback when gold standard databases are not available.

SUBMITTER: Sciortino S

PROVIDER: S-EPMC3788526 | biostudies-other | 2013 Sep

REPOSITORIES: biostudies-other

ACCESS DATA

Similar Datasets

Project description:Background/AimsThe Kaiser Permanente Northern California (KPNC) Research Program on Genes, Environment, and Health (RPGEH) provides a research resource to support investigations of environmental and genetic factors in the development of a wide variety of conditions. While the resource is still evolving with new data collection, it consists of data from surveys and electronic medical records on over 400,000 adult members of KPNC; biospecimens collected and stored on ~200,000 of these individuals; and data from genome-wide and telomere-length assays on ~110,000 of those who have contributed biospecimens; and linkage of these members to environmental, area-level databases.MethodsThe RPGEH was developed in part with the understanding that it would be made available to the scientific community for appropriate studies. An Access and Collaborations Core has developed procedures for submission of applications for research studies, their review, and decisions on approval and support. Review of proposals by an Applications Review Committee follows a two-step process, with a pre-application to assess feasibility (e.g., adequate numbers of the phenotype: availability of appropriate data, given inclusion criteria) and a full application to assess appropriateness of the study in the RPGEH context. Scientific merit; alignment with RPGEH guiding principles, including ethical, legal, and social implications; consistency with informed consents; potential overlap with prior approvals; and collaboration with a researcher affiliated with the Division of Research are among the criteria for approval. As an alternative for select analyses, genomic and selected phenotypic data will be available in the NIH database of genotypes and phenotypes (dbGaP) for the substantial subset of RPGEH participants who have consented to dbGaP deposition.ResultsAs of October 31, 2012, the RPGEH has received 74 pre-applications and full applications for the use of its resources; only 6 pre-applications were not approved. In 2011–2012, 13 approved applications were funded by NIH and other agencies. Studies currently underway include genome-wide association studies of prostate cancer, bi-polar disorder, multiple sclerosis, and mammographic density.ConclusionsAccess to the unique and outstanding research resources of the RPGEH balances the mission of promoting research with the need to shepherd finite resources and safeguard member confidentiality.

Project description:BackgroundTo inform Danish health care reform efforts, we compared health care system inputs and performance and assessed the usefulness of these comparisons for informing policy.MethodsRetrospective analysis of secondary data in the Danish Health Care System (DHS) with 5.3 million citizens and the Kaiser Permanente integrated delivery system (KP) with 6.1 million members in California. We used secondary data to compare population characteristics, professional staff, delivery structure, utilisation and quality measures, and direct costs. We adjusted the cost data to increase comparability.ResultsA higher percentage of KP patients had chronic conditions than did patients in the DHS: 6.3% vs. 2.8% (diabetes) and 19% vs. 8.5% (hypertension), respectively. KP had fewer total physicians and staff compared to DHS, with 134 physicians/100,000 individuals versus 311 physicians/100,000 individuals. KP physicians are salaried employees; in contrast, DHS primary care physicians own and run their practices, remunerated by a mixture of capitation and fee-for-service payments, while most specialists are employed at largely public hospitals. Hospitalisation rates and lengths of stay (LOS) were lower in KP, with mean acute admission LOS of 3.9 days versus 6.0 days in the DHS, and, for stroke admissions, 4.2 days versus 23 days. Screening rates also differed: 93% of KP members with diabetes received retinal screening; only 46% of patients in the DHS with diabetes did. Per capita operating expenditures were PPP$1,951 (KP) and PPP $1,845 (DHS).ConclusionCompared to the DHS, KP had a population with more documented disease and higher operating costs, while employing fewer physicians and resources like hospital beds. Observed quality measures also appear higher in KP. However, simple comparisons between health care systems may have limited value without detailed information on mechanisms underlying differences or identifying translatable care improvement strategies. We suggest items for more in-depth analyses that could improve the interpretability of findings and help identify lessons that can be transferred.

Dataset Information

PS3-14: CREX: Utility of a Computerized Methodology to Identify Health Conditions Using the EMR for GWAS, in the Kaiser Permanente Research Program on Genes, Environment, and Health.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets