PUBH 8502 Week 8 Peer Critique Example

Week 8 Discussion: Critiquing a Classmate's Appraisal

The Reading the Paper Also Allows: A Peer Critique Building the Opposite Interpretation of Two Classmates' Appraisals, a Reminder Trial With a Hidden Second Kit and a Home Visit Cohort Shaped by Who Said Yes

Reply to Hannah

Hannah, your appraisal concludes that "text reminders doubled colorectal screening kit return in this trial." I agree the trial was well randomized, and your reading is the one the authors give.

The methods section allows a different reading, though. The reminder arm received the texts and also a second mailed kit three weeks later, while the control arm received one kit only. The gain could belong to the second kit rather than the texts. Reviews in my own topic, chlamydia retesting, have found mailed kits among the more effective ways to bring people back (Guy et al., 2012), so a second kit is a plausible cause on its own. The trial cannot separate the two because they were delivered together. What would decide between the readings? A trial arm with a second kit and no texts. If one exists in the literature, it would settle the matter. If not, I would suggest writing that the trial shows the combined package worked, without crediting texts alone.

Reply to Owen

Owen, you concluded that "home visits reduced children's emergency visits for asthma by a third." The cohort is large and the follow-up long. The paper also shows that families who accepted visits differed at the start: more had a regular doctor, fewer had moved in the past year. The adjusted model includes some of these differences but not housing stability, which the authors mention as unmeasured. Families who agree to visits may be better placed to manage asthma whether or not anyone visits. That is a selection explanation, and observational designs are especially open to it (Shadish et al., 2002). Fewer emergency visits might reflect who agreed rather than what was delivered. Would a comparison of families offered visits, whether or not they accepted, be possible from the data? That comparison would be less vulnerable to who said yes.

Turning It on My Own Appraisal

I applied the same test to my Week 6 appraisal, which concluded that a before-and-after study overstated the effect of reminders. The opposite reading is that I may have credited too much to the express testing service. Reminded clients could have chosen the express service because it was quicker, so some tests I attributed to express testing may have been prompted by the reminders. I had followed the CASP questions closely, yet considered only biases pointing one way. Even the real Dutch study I used for comparison relied on historical controls (Kampman et al., 2015), and I had not asked which direction that bias ran either. I'll add a sentence noting that the true effect is uncertain in both directions.

References

Guy, R., Hocking, J., Low, N., Ali, H., Bauer, H. M., Walker, J., Klausner, J. D., Donovan, B., & Kaldor, J. M. (2012). Interventions to increase rescreening for repeat chlamydial infection. Sexually Transmitted Diseases, 39(2), 136-146. https://doi.org/10.1097/OLQ.0b013e31823ed4ec

Kampman, C. J. G., Koedijk, F. D. H., Driessen-Hulshof, H. C. M., Hautvast, J. L. A., & van den Broek, I. V. F. (2015). Retesting young STI clinic visitors with urogenital Chlamydia trachomatis infection in the Netherlands; response to a text message reminder and reinfection rates: A prospective study with historical controls. Sexually Transmitted Infections, 92(2), 124-129. https://doi.org/10.1136/sextrans-2015-052115

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