Establishing the Feasibility and Research Utility of a Linked Community-VA Database

OBJECTIVE

Many veterans receiving VA health care seek care outside VA. In the Chicago area for example, of the approximately 208,553 veterans who were enrolled for VA care in Federal Fiscal Year (FY) 2012, just 57% used VA care in that year. These numbers raise multiple questions relevant to veterans’ healthcare and measurement of VA outcomes. But, despite VA’s longstanding electronic health record, relevant data is limited and privacy concerns prevail. Medicare and Medicaid fee-for-service claims data are helpful but many covered veterans are enrolled in managed care plans which do not generate billing claims. Further, at least half of VA enrollees (mostly veterans under 65 years) are not enrolled in either of those programs. The Chicago Health Atlas (CHA) database is a unique repository of linked healthcare data for 5.4 million area residents. We examined the feasibility and potential research utility of a CHA-VA record linkage.

 

METHODS

We studied VA enrollees in FY 2012 who lived in one of two Chicago-area VA service areas. CHA data are sourced from federally qualified health centers, practice consortiums, private practices, academic medical centers, and the county public hospital. Information on VA enrollment, demographics, and inpatient, outpatient, and laboratory utilization were obtained from VA administrative data sources. Using a software application developed by creators of the CHA, we generated an irreversible anonymous (“hash”) code from each veterans’ identifying information (SSN, first name, last name, and date of birth) and used that code as our CHA-VA linking variable. Descriptive statistics evaluated the record linkage and characterized veterans’ VA and non-VA healthcare use and diagnoses.

 

RESULTS

The linkage yielded 71,641 matches among the 208,553 Chicago-area VA enrollees and 43,945 (21%) had a non-VA provider encounter during FY 2011-2012. Among those, 28,882 (65.7%) had used both VA and non-VA health care. Nearly half (47%) of the enrollees using Non-VA care were under 65 years old. The most common chronic condition diagnoses among VA/Non-VA users were hypertension (58%/71.7%), hyperlipidemia (57%/53.8%) and diabetes (25%/42.6%).

CONCLUSIONS/IMPACT

We established that a linkage of Chicago-area veterans’ VA and non-VA records can be accomplished without the disclosure of patient identifiers. Further, we demonstrated the linkage’s utility for research as we obtained additional important clinical information on over 20,000 younger veterans, a group about whom we know little regarding use of Non-VA services.