Show simple item record

dc.contributor.advisorDiGuiseppi, Carolyn G.
dc.contributor.advisorCrume, Tessa L.
dc.contributor.authorDuca, Lindsey Marie
dc.contributor.committeememberKahn, Michael G.
dc.contributor.committeememberKhanna, Amber D.
dc.contributor.committeememberOng, Toan C.
dc.contributor.committeememberPyle, Laura L.
dc.date.accessioned2019-05-17T20:03:28Z
dc.date.available2021-05-16T19:43:49Z
dc.date.submitted2019
dc.descriptionIncludes bibliographical references.
dc.descriptionSpring
dc.description.abstractPopulation-level electronic health record (EHR)-based surveillance systems are becoming increasingly common resources for clinical research and remain relatively novel in the U.S. for chronic disease surveillance. Although these high-dimensional data present many opportunities for analyzing large numbers of patient records in near real time, one of the primary challenges of using EHR data is all eligible cases in the source population may not be identified. While it has been acknowledged that incomplete case ascertainment may affect prevalence estimates, it is often assumed that it should not impact estimates of association. In this dissertation, capture-recapture methodology was performed to determine the extent to which incomplete case ascertainment biases prevalence estimates of congenital heart defects (CHDs) among adolescents and adults in Colorado. Then the impact of incomplete case ascertainment on an exposure-outcome relationship was assessed. First, predictors of a lapse in cardiac care during the transition from pediatric to adult care among young adults with a CHD were examined. Then conditions under which incomplete case ascertainment lead to selection bias in EHR-based studies of lapses in cardiac care during the critical transition period were illustrated. My findings revealed that after accounting for bias introduced by incomplete case ascertainment, prevalence estimates of CHDs among adolescents and adults in Colorado increased (Aim 1). Over a third of young adults with a CHD experienced more than a 2-year lapse in recommended cardiac care and a combination of both individual- and census-level variables predicted this lapse in care (Aim 2). Based on simulations, capture-recapture methodology was a novel initial indicator of selection bias in an exposure-outcome association and quantitative bias analysis techniques were effective in correcting for this bias (Aim 3). Incomplete case ascertainment in EHR-based surveillance studies poses a continuing challenge for research on rare chronic conditions. Even with multiple sources of ascertainment, chronic disease surveillance programs typically cannot achieve complete case ascertainment when using EHRs in the U.S. due to the fragmented nature of the healthcare system. However, use of methods to quantify and correct for incomplete case ascertainment can improve the accuracy of these estimates.
dc.identifierDuca_ucdenveramc_1639D_10622.pdf
dc.identifier.urihttps://hdl.handle.net/10968/3446
dc.languageEnglish
dc.publisherUniversity of Colorado at Denver, Anschutz Medical Campus. Health Sciences Library
dc.rightsCopyright of the original work is retained by the author.
dc.rights.accessEmbargo Expires: 05/16/2021
dc.subjectSurveillance
dc.subjectBias analysis
dc.subject.meshHeart Defects, Congenital
dc.subject.meshSelection Bias
dc.subject.meshElectronic Health Records
dc.subject.meshData Analysis
dc.subject.meshPrevalence
dc.titleInvestigating bias in a population-level electronic health record surveillance system of individuals with congenital heart disease
dc.typeThesis
dcterms.embargo.expires2021-05-16
thesis.degree.disciplineEpidemiology
thesis.degree.grantorUniversity of Colorado at Denver, Anschutz Medical Campus
thesis.degree.levelDoctoral
thesis.degree.nameDoctor of Philosophy (Ph.D.)


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record