Linking databases has value, can raise regulatory questions
Linking databases has value, can raise regulatory questions
Researchers sought help from IRB
Several national forces are driving energy into research that heavily relies on clinical data registries. The federal stimulus money that's available to hospitals that upgrade their electronic database technology and for comparative effectiveness research is a chief driver of this trend.
But the question for clinical trial sites and investigators is: How can we best use the available clinical data registries, and what are the drawbacks and ethical concerns of doing so?
For example, one of the main drawbacks of some of these databases has been a lack of long-term follow-up, says Rachel Dokholyan, MPH, project leader at the Duke Clinical Research Institute (DCRI) of Duke University in Durham, NC.
Dokholyan is the project leader at DCRI for the analysis and data warehouse for the Society of Thoracic Surgeons National Database. This database consists mainly of adult cardiac patients, and it was started in 1989, she notes.
"They needed a larger data management contractor; they needed risk adjustment, and they needed a statistical message, and that's how Duke got involved," Dokholyan says. "I've been the project leader since 2002."
The issues Dokholyan has experienced with this database could serve as learning points for others in clinical research.
For instance, large referral centers often don't know what has happened with their patients beyond 30 days, Dokholyan says.
"It's very hard for some of these hospitals to do long-term follow-up," she explains. "But a crucial question is 'What happens to patients six months down the road, or one year or five years later?'"
Even finding out whether patients have died is problematic.
One potentially valuable database that could provide these answers is an administrative claims database.1
These databases can be helpful because they typically are very large and provide comprehensive coverage of patients' entire health care experience, often with patient identifiers.
This makes it possible to create a longitudinal patient care record.
Since administrative claims databases lack accurate or detailed clinical information, the ideal solution from a research perspective would be to link these databases with registry data.
However, there are obstacles to linking the two types of databases, including the challenge of creating the link, Dokholyan says.
"We usually need patient identifiers to link databases," she says. "So the issue is whether we could link these registries to external information for long-term follow-up."
The Society of Thoracic Surgeons did not have an answer to the follow-up dilemma: "After 20 years of doing this they hadn't come up with a way that the registry could expect surgeons to do one-year or five-year follow-ups."
The process had evolved, and now it was time to find answers.
"We finally got down to the logical conclusion that what we needed was to use existing data sources out there, including claims data that already were being collected, and we needed identifiers to get these data sources," Dokholyan says.
Another issue involved the original purposes of the databases: "These registries were created and designed mainly for quality assessment purposes," Dokholyan says. "Hospitals are sending us data, and what they get back is a detailed report every quarter that shows their data in comparison to the overall population that's submitting data."
There is some analysis that creates general knowledge for improving care, but even if research were not allowed, these registries would exist, she explains.
"They're self-funded, so they'd still serve their primary purpose, which is feedback reports to surgeons to help them change their practice," she adds.
The reports provide aggregate information, such as a statement that 25% of a physician group's patients have hypertension, Dokholyan says.
"The reports show how their patients look in aggregate in the region and how they look overall in the nation," she adds. "No patient level information goes into the report."
Dokholyan and co-investigators met with a chair of a Duke University Health Systems IRB to discuss the database and how they'd like to link it with other registries for research purposes.
They settled on the point involving identifiers.
"The fact that we were moving to directly identifying information was the main thing the IRB wanted to make sure everyone was going to be comfortable with," Dokholyan says.
Researchers received a waiver of informed consent for the database repository.
"There are different research projects we may move forward with from there, including a big class of analyses we can do on data subsets in the repository, and these don't have direct identifiers," Dokholyan says.
"We needed identifiers to connect information like mortality data," she adds. "So we take a HIPPA limited dataset, and for that class of analyses, we went back to the IRB and requested a declaration that it did not involve human subjects, and they gave us that."
So if an investigator wants to do an analysis that requires identifiers, the investigator will have to ask the IRB for a waiver and approval, Dokholyan says.
"This whole conversation with the IRB needed to happen for the single reason we needed patient identifiers, and one of the main reasons we needed the identifiers was to link to these other datasets," she says.
- Dokholyan RS, Muhlbaier LH, Falletta JM, et al. Regulatory and ethical considerations for linking clinical and administrative databases. Am Heart J. 2009;157:971-982.
Subscribe Now for Access
You have reached your article limit for the month. We hope you found our articles both enjoyable and insightful. For information on new subscriptions, product trials, alternative billing arrangements or group and site discounts please call 800-688-2421. We look forward to having you as a long-term member of the Relias Media community.