Andy Woster explains how to know if your data is missing or unknown

Whew! What a crazy spring! For me, at least, it involved learning an entirely new job and making a transition from a training focused on communicable disease to a professional role in violence and injury prevention. That was wild! I’m glad it’s over and glad to have done it! 

Now it’s camping season for me and, if you’re a member of one of the CFPS review teams, it’s the start of crunch time for you! Remembering that the deadline for the completion of your reviews for 2015 cases is January 1st, 2017, I thought I’d take a moment to share a message with you regarding data entry. (If you haven’t figured it out, you’ll be doing this a lot in the next few months! We still have about 1/3 of all assigned cases that don’t yet have a case number in the National Center Case Reporting System.)

For this edition of the data corner, I wanted to reflect on a piece of information that piqued my interest from the recent data quality webinar the National Center hosted. Missing and unknown information is a distinction epidemiologists and data nerds alike can spend hours bickering like old friends over and still fail to reach consensus on. During the data quality webinar, Dr. Schnitzer indicated that unknown data is information that teams discussed and for which the correct response to the question was not known or not available. Missing information, on the other hand, is information which occurs because the question was skipped or not discussed during the meeting. A simpler guiding principle is “if you looked for it and can’t find the answer then the response is unknown.” If you didn’t look for it or skipped the question, then the response should be left blank.
The purpose for this distinction is straightforward when you are viewing the question from a data nerd’s point of view. For example, for a given question that we should be able to answer, such as was the “Firearm stored loaded?”, 14.6 percent of responses were missing for firearms-related deaths. This information was unknown for 49.1 percent of cases. While this information is certainly something that may be difficult to collect depending on the circumstances of the investigation and scene, it is always something that we should look for in our records and should ask law enforcement about during our reviews. So, for a question like this we would really want all of those responses to be unknown, i.e., indicating that the coordinator or team searched for the information but that it wasn’t available in the information provided. A question of this central importance should never be overlooked or skipped in the review of a case.

So, when you’re navigating the landscape of case abstraction this review season, I hope you find yourself clicking your ruby slippers together and repeating to yourself three times (or more!), “If I looked for it and can’t find the answer, the response is unknown.” This simple trick will keep you flying along the Yellow Brick Road to data entry success.