NURS 8740 Week 7 Peer Critique Post Example

Week 7 Discussion: Peer Critique of Priya's Findings

Where the Sentence Gets Ahead of the Data: A Critique of a Pre-Post Adherence Finding From a Heart Failure Text Reminder Project

Priya, your draft conclusion reads: "The text reminder program improved medication adherence among heart failure clinic patients, increasing the proportion of days covered from 68% to 81%." I want to separate what your data hold from what that sentence claims, because I think the data support more than half of it.

What the data support. You measured proportion of days covered from pharmacy fill records for the 42 patients enrolled, over the 90 days before enrollment and the 90 days after. The rise from 68% to 81% is real in your records, the measure is a standard one and the fill data are not self-reported, which removes the usual worry that patients tell nurses what they want to hear. You can state with confidence that these patients filled their heart failure medications more consistently in the three months after enrollment than in the three months before.

Where the sentence overtakes them. The word improved claims that the program caused the change, and the design cannot carry that. This is a single group measured before and after, with nothing running alongside it. In that design, at least three other explanations produce the same numbers (Shadish et al., 2002). The one I would worry about most is regression to the mean. Your methods say patients were enrolled at their first clinic visit after a heart failure admission. A hospital admission is often preceded by a period of poor adherence, so the 90 days before enrollment may have caught many patients at their lowest point. Measured again, a group selected at a low point will tend to move back toward its usual level whether or not anything is done (Barnett et al., 2005). Second, a hospital admission is itself an intervention: patients leave with reconciled medication lists, new prescriptions and often a scare. Third, proportion of days covered measures fills, not doses taken, so a change in pharmacy auto-refill enrollment during the same months would raise the measure without any change in behavior. Your clinic's pharmacy partner launched auto-refill in the spring, if I read your setting description correctly.

What the design can license. A before-and-after design in improvement work can show that a change happened in a group over a period; it needs a comparison, a time series with several points before the change or both to support an attribution (Portela et al., 2015). You may already have some of this. If the clinic had heart failure patients discharged in the same months who did not enroll, their fill data would make a rough comparison group. If not, monthly points for the enrolled patients over the year before enrollment would show whether 68% was typical or a dip.

A sentence the data would support now: "Among 42 heart failure clinic patients enrolled after a hospital admission, the proportion of days covered for heart failure medications was 81% in the 90 days after enrollment, compared with 68% in the 90 days before. Without a comparison group, this change cannot be attributed to the text reminder program, since adherence measured after an admission may rise for other reasons."

My point: your finding is worth reporting; the verb is the problem. Could the clinic list give you a comparison group?

Follow-Up Reply to Priya

Priya, thank you for checking. You found 31 patients discharged from the same hospital with heart failure in the same months who did not enroll, mainly because they were followed by an outside cardiologist. Their proportion of days covered went from 70% to 77% over the same windows. That is useful and humbling: a rise of seven points without the program, against thirteen with it. The groups are not equivalent, since patients with outside cardiologists may differ in ways that matter, so I would report both rises side by side and describe the difference cautiously rather than subtracting one from the other. I think your revised discussion is much stronger for including them, and it shows the program may have added something beyond the post-admission rebound. I would still keep the auto-refill question in your limitations, because it applies to both groups and could explain part of both rises.

References

Barnett, A. G., van der Pols, J. C., & Dobson, A. J. (2005). Regression to the mean: What it is and how to deal with it. International Journal of Epidemiology, 34(1), 215-220. https://doi.org/10.1093/ije/dyh299

Portela, M. C., Pronovost, P. J., Woodcock, T., Carter, P., & Dixon-Woods, M. (2015). How to study improvement interventions: A brief overview of possible study types. BMJ Quality & Safety, 24(5), 325-336. https://doi.org/10.1136/bmjqs-2014-003620

Shadish, W. R., Cook, T. D., & Campbell, D. T. (2002). Experimental and quasi-experimental designs for generalized causal inference. Houghton Mifflin.