Discussion: Mini Scenarios
1. You are creating a database for a research study that is investigating the effectiveness of trampoline jumping for improving cognitive functioning in alzheimer's patients. Enrolled patients are being randomized into either intervention or control groups. Patients in the intervention group will receive a total of 24 sessions over 3 months, which are supposed to occur twice a week. The Principal Investigator (PI) wants a tally of the number of sessions that each patient participated in, in order to confirm that patients received the correct "dose" of the intervention.
In your database you create a table to record the specific date of each trampoline session for every patient. The PI wants to simplify the database and suggests only recording the total number of sessions that each patient participated in. Explain to the PI why collecting more "granular" data could be worthwhile in this instance.
2. You are using an Electronic Medical Record report to collect data on patients who have received a specific procedure. Midway through the study you realize that the specific procedure code that the report is using relates to more than one procedure, not only the one you are studying meaning that when that code appears in a patient's record, they may or may not have actually had the procedure you are interested in. The only way to be sure is to review the notes entered in each patient's EMR.
Unfortunately that will take too long for you to do alone. Luckily there are 2 interns in your office looking for something to work on! How would you divide the work of reviewing the patient records between the 3 of you in a way that would ensure that the data is being recorded accurately and consistently? How would you avoid this in the future?
3. The health and human services department in your state has just compiled a massive "all claims database" that includes all public and private insurance payments for every procedure provided at all hospitals in the state over the past decade a total of 4.3 billion claims.
A couple of your colleagues at state university are exploring the data and discovers a very small but significant increase in the number of hearing tests provided in months beginning with the letter J. They begin to hypothesize about why people may be experiencing increased hearing problems during those months. How would you help them understand the need for caution in interpreting patterns found in "big data"?
4. You are working on a study related to exposure to secondhand smoke using secondary selfreported survey data that was collected by someone else for an earlier study. One of the things you are interested in doing is comparing the number of subjects who reported living with someone who smokes. You notice that that particular variable is coded as "Yes" or is blank. There are no notes in the codebook, so you are not sure if a blank cell should be interpreted as a "No" or as missing data. You contact the original
investigators, and they say that they accidently forgot to include those responses in the final dataset. They send you a list of the subjects who answered "No" to the question, and let you know that data is missing for anyone who isn't a "Yes" already, or isn't on this "No" list. Explain the general steps you would take to go about updating your dataset without losing or corrupting any of the original data.