When data goes bad: Even small errors matter

Concerned that a long-time patient's record showed her name with two middle initials, Fred N. Pelzman, M.D., set to find out how this error happened.

Neither the front-desk staff, practice supervisors or IT administrators could explain how a second middle initial has appeared in patient records now five times, he writes in an article at MedPage Today. Yet he worries that somewhere down the line, lab or test results will not be matched up with that patient's record because of this computer problem.

"As we've created these massive databases, electronic systems that house our data, the risk of errors getting incorporated into it has multiplied, growing exponentially," he writes. "When we try to use our databases to create systems of quality, the quality of the data going in matters."

In the case of the multiple middle initials, however, the data apparently was correct when it was entered. What happened to it after that, it seems no one can explain.

Pelzman notes other instances when patients have been "attributed" to him--especially in insurance company records--that he's never heard of.

He says of the efforts to digitize healthcare: "Getting bad data like this leads us to feel like it may not be worth the effort. If none of this is relevant to me, is this really going make a difference? If we got this much wrong, can any of it really be right?"

"More data, more problems" is becoming a new refrain, one that the National Quality Forum is tackling while advocating for meaningful quality frameworks for the information.

"Everybody needs and deserves to have consistent and accurate information about healthcare," NQF President and CEO Christine Cassel, M.D., recently told HealthITAnalytics. "[B]ut in a world where data sources are multiplying by leaps and bounds ... there's a lot of complexity when it comes to actually implementing systematic improvements."

UPMC Health System has been trying to standardize data across its organization through a data governance council, starting with mundane issues such as data definitions, where data is sent and how to send data to different places.

To learn more:
- read the article