Covering vascular surgery? Watch for selection bias in this database

Tara Haelle

About Tara Haelle

Tara Haelle (@TaraHaelle) is AHCJ's medical studies core topic leader, guiding journalists through the jargon-filled shorthand of science and research and enabling them to translate the evidence into accurate information.

If you cover medical research related to vascular procedures and conditions, you’ve likely come across studies using data from the Society for Vascular Surgery Vascular Quality Initiative (SVS VQI).

As a database designed to improve patient safety, the SVS VQI can be very useful for analyzing outcomes and associated variable for 12 major vascular procedures as long as researchers (and journalists) are aware of the limitations of the data set.

Studies looking at quality improvement initiatives, safety outcomes or comparative effusiveness of these procedures may be likely to use SVS VQI data, so the Medical Studies Core Topic Data section now includes a primer on what this database includes and what caveats and considerations researchers — and therefore journalists — need to be aware of for this type of research. This new edition is part of the ongoing series of articles from JAMA Surgery on the uses, considerations and limitations of various databases frequently used in surgery-related research.

The biggest limitation for studies using SVS VQI data is selection bias: it’s a self-reported database (from physicians and hospitals), and it’s optional. Physicians might report data for only a couple of the vascular procedures if those are the only ones they do. By the same token, though, hospitals might only report data for certain vascular procedures for … whatever reason, and those using the database would not know what the hospitals don’t report (or why). If only the best outcomes or the outcomes associated with certain procedures, patients or other conditions are reported, they can’t be relied on as representative.

The database also includes hundreds of possible variables, which means p-hacking is a risk if the researchers haven’t clearly laid out their hypothesis, their planned subgroups, their included variables, their reasons for excluding patients and any statistical calculations needed to adjust for issues in any of these areas — all of which are ripe for asking outside sources or biostatisticians about.

Leave a Reply