Assembling and validating data from multiple sources to study care for Veterans with bladder cancer

Abstract

BackgroundDespite the high prevalence of bladder cancer, research on optimal bladder cancer care is limited. One way to advance observational research on care is to use linked data from multiple sources. Such big data research can provide real-world details of care and outcomes across a large number of patients. We assembled and validated such data including (1) administrative data from the Department of Veterans Affairs (VA), (2) Medicare claims, (3) data abstracted by tumor registrars, (4) data abstracted via chart review from the national electronic health record, and (5) full text pathology reports.MethodsBased on these combined data, we used administrative data to identify patients with newly diagnosed bladder cancer who received care in the VA. To validate these data, we first compared the diagnosis date from the administrative data to that from the tumor registry. Second, we measured accuracy of identifying bladder cancer care in VA administrative data, using a random chart review (n = 100) as gold standard. Lastly, we compared the proportion of patients who received bladder cancer care among those who did versus did not have full text bladder pathology reports available, expecting that those with reports are significantly more likely to receive care in VA.ResultsOut of 26,675 patients, 11,323 (42%) had tumor registry data available. 90% of these patients had a difference of 90 days or less between the diagnosis dates from administrative and registry data. Among 100 patients selected for chart review, 59 received bladder cancer care in VA, 58 of which were correctly identified using administrative data (sensitivity 98%, specificity 90%). Receipt of bladder cancer care was substantially more common among those who did versus did not have bladder pathology available (96% vs. 43%, p < 0.001).ConclusionMerging administrative with electronic health record and pathology data offers new possibilities to validate the use of administrative data in bladder cancer research.Electronic supplementary materialThe online version of this article (10.1186/s12894-017-0271-x) contains supplementary material, which is available to authorized users.

The impact of low- versus high-intensity surveillance cystoscopy on surgical care and cancer outcomes in patients with high-risk non-muscle-invasive bladder cancer (NMIBC)

“…[8] We identified cystoscopy procedures using Current Procedural Terminology (CPT) codes as previously described [6] and categorized patients into those that received low-versus high-intensity surveillance based on current consensus guideline recommendations and the length of the surveillance window ( Fig 1). [9,10,11] The surveillance window was evaluated by the following intervals: up to 5.5 months, 5.5 up to 9.5, 9.5 up to 13.5, 13.5 up to 17.5, 17.5 up to 21.5, and over 21.5 months. The rationale for these intervals was that surveillance cystoscopy at 4, 8, 12, 16, 20, and 24 months is recommended for high-risk bladder cancer by the AUA.…”

The impact of frequent cystoscopy on surgical care and cancer outcomes among patients with low‐risk, non–muscle‐invasive bladder cancer

“…As previously described, 10 we used procedure codes (see Supporting Table 1) to enumerate the frequency of cystoscopic surveillance during the first 2 years after diagnosis. 11,12 The time period during which we assessed cystoscopic surveillance (the surveillance window) started with the diagnosis date (ascertained using validated algorithms 11 ) and ended 2 years after diagnosis or at the time of cystectomy, radiotherapy, or cancer recurrence, whichever occurred first. We enumerated cystoscopic surveillance procedures only until patients had a cancer recurrence because a recurrence increases the risk for further recurrences.…”

Linking electronic health data in pharmacoepidemiology: Appropriateness and feasibility

“…Data linkage may allow for a more accurate or complete definition of an exposure, outcome, or confounder of interest. Linking EHR (which provides clinical details often absent in other datasets) to administrative claims databases (which contain extensive data on diagnoses and medications, often lacking clinical details) can improve the capture of medication exposure and outcome ascertainment as evidenced by numerous health data linkages across clinical data entities (Table ) …”