Clinical Pharmacoepidemiology Linking Electronic Health Records With Claims Data
MetadataShow full item record
CitationLin, Joshua. 2017. Clinical Pharmacoepidemiology Linking Electronic Health Records With Claims Data. Doctoral dissertation, Harvard T.H. Chan School of Public Health.
AbstractElectronic health records (EHR) and insurance claims databases have been considered cost-effective data sources for conducting comparative effectiveness research (CER) in routine healthcare settings. However, these secondary databases have some major limitations. Claims data lack in some clinical information essential for CER. On the other hand, most US EHR systems fail to capture substantial amount of medical information due to care-discontinuity in the study EHR (i.e., receiving care outside of the EHR system). Linking EHR to claims data can supplement the limitations of the individual databases and help combat various biases.
In chapter 1, we demonstrated the added value of linking EHR with claims data by developing a prediction score for anticoagulation quality in patients receiving Vitamin K antagonist (VKA) therapy. Anticoagulation quality was shown to have large impact on the effectiveness and safety of VKA therapy and a prediction model for having good anticoagulation quality can be critical for safe and optimal prescribing of VKA. Data in EHR is necessary for ascertaining anticoagulation quality outcomes and some key predictors and the linkage to claims data helps to avoid information bias due to care-discontinuity.
In chapter 2, we sought to quantify care-discontinuity in an EHR by comparing encounter records in an EHR with the claims data. Within levels of care-discontinuity, we quantified information bias by comparing classification of CER-relevant variables based on EHR data alone with that based on EHR linked with claims data. We also evaluated whether high care-continuity patients would have representative comorbidity profiles when compared to the remaining population in the EHR.
In chapter 3, we aimed to develop and validate algorithms to identify patients with high care-continuity with the predictors available in a typical EHR system. This is important because researchers often need to address information bias caused by care-discontinuity in an EHR without linkage to additional data sources due to privacy and compliance concerns. Our approach may help researchers answer rapidly emerging drug safety and effectiveness questions in a timely fashion with the exponentially growing EHR resources, leveraging rich clinical data while minimizing information bias due to care-discontinuity.
Citable link to this pagehttp://nrs.harvard.edu/urn-3:HUL.InstRepos:42066817
Contact administrator regarding this item (to report mistakes or request changes)