Psychology meets mammography: Quasi-experimental design practice

[sue_frantz]

After covering experiments, ask your students to work in small groups to design an experiment that would test this hypothesis:

If patients are able to schedule their own mammograms, they are more likely to actually get a mammogram.

For their design, ask students to identity the independent variable (experimental and control conditions) and dependent variable. They should be sure to include operational definitions. Once discussion seems to have died down, invite a volunteer from each group to share their design.

In this freely available article, researchers conducted such a study using archival data (Waddell et al., 2023). Consider sharing this study with your students. Because the study was archival, it was quasi-experimental—participants were not randomly assigned to conditions. In this particular healthcare system, patients were not able to self-refer for a mammogram; a physician had to order it, say, during an office visit. The healthcare system implemented a new electronic health record portal that gave patients the ability to schedule their mammograms without having to call the clinic. Some patients were not active in the portal so they were considered the control group. Patients who were active in the portal (operationally defined as having logged into the system at least once in the twelve months prior to the visit with their physician who ordered the mammogram)—and therefore could electronically schedule their appointments—were in the experimental group.

The dependent variable was an easy one to measure. Did the patient get a mammogram within six months of the doctor’s appointment when the mammogram order was issued? Approximately 73% of the experimental group got a mammogram within six months as compared to approximately 54% of the control group.

There was one big confounding variable, however. Those who were active in the portal (experimental group) received a reminder email after their doctor’s visit to schedule a mammogram. Those who did not use the portal (control group) did not receive a reminder. There is no way for us to know, then, if the differences seen in the dependent variable were due to being able to electronically schedule a mammogram or due to receiving an email reminding them to schedule.

If you share this study with your students, ask students to consider what other confounding variables there may be. The researchers identify a few others in the “Limitations” section of the article’s discussion. For example, might there be a difference between those who logged into the electronic health record portal and those who hadn't?

Would the experiments your students designed address these confounds? Conclude this activity by reminding students that while each type of research method has its limitations, the more types of methods we use to address a hypothesis, the more confidence we have in the overall results.

As an out-of-class assignment, consider asking your students to generate other hypothesis about patients and healthcare behavior. How would they design an experiment or quasi-experiment using archival data to test one of those hypotheses?

Reference

Waddell, K. J., Goel, K., Park, S.-H., Linn, K. A., Navathe, A. S., Liao, J. M., McDonald, C., Reitz, C., Moore, J., Hyland, S., & Mehta, S. J. (2023). Association of electronic self-scheduling and screening mammogram completion. American Journal of Preventive Medicine, S0749379723004440. https://doi.org/10.1016/j.amepre.2023.11.002