PUBH 6036 Week 5 Experimental Design Memo Example

Memorandum

To: Study Team and Director of Property Management, Cedar Hollow Housing Authority

From: Graduate Research Assistant

Date: Month Day, Year

Re: Which bed bug interventions could be assigned at random, and how the chosen one would be rolled out

Purpose

The cohort tells us whether infestation and absence go together; it cannot tell us whether reducing infestation would reduce absence. Random assignment can. This memo considers three trials the housing authority could support, sorts each as impossible, wrong or right, and lays out the one worth running.

Candidate A: Treat or Do Not Treat Infested Apartments

The cleanest trial on paper would assign infested apartments at random to treatment or to no treatment and compare children's absences. It is wrong. The authority has a duty under its leases and local housing code to treat reported infestations, and withholding treatment from a family known to be living with bed bugs, in order to observe harm to their children, could never be justified. No review board would approve it, and the authority would not permit it.

Candidate B: Mattress Encasements, Apartment by Apartment

A second trial would give encasements for every bed to a random half of households with children. This is ethical, since encasements are an extra on top of normal treatment. It is close to impossible to interpret, however. Bed bugs move between adjoining units through wall voids and along pipes, so an encased apartment's neighbors share its insects. Assignment at the household level invites contamination, where the comparison group is changed by what the treated group receives; Torgerson (2001) describes how this dilutes a trial's estimate toward no effect. Families also share information and supplies across hallways.

Candidate C: Building-Wide Monitoring, Building by Building

The third trial assigns the timing of a proactive building-wide monitoring program. Under the program, trained staff place interceptor cups under beds and sofas in every apartment, return a week later to read them, and send confirmed infestations straight to treatment. A survey of 43 New Jersey buildings that used a similar protocol found that interceptors detected 89% of infestations and that the protocol cost about $12 per apartment (Wang et al., 2016). Every building will eventually receive it, so no family is denied a benefit; only the order is randomized. This is the right call.

Table 1

Sorting the Three Candidates

Note. Prepared for this course from the authority's current pest management plan.

The Design: A Stepped Wedge

A stepped-wedge cluster trial moves clusters from control to intervention one group at a time, in an order set by chance, until all have crossed (Hemming et al., 2015). With nine family buildings, three steps of three buildings each fit the school year:

Period 1 (September to October): all nine buildings follow usual practice, which is treatment on complaint.

Period 2 (November to December): three randomly chosen buildings begin monitoring.

Period 3 (January to February): three more buildings begin.

Period 4 (March to April): the last three begin, and all nine continue through June.

The primary outcome is the rate of school days missed per child per month, from district records linked with caregiver permission. A secondary outcome is the share of apartments with a confirmed infestation at each reading.

Why a Stepped Wedge Rather Than a Parallel Design

The property management director has said the authority cannot inspect all nine buildings at once, since the pest contractor's crew is limited. The stepped wedge turns that staffing limit into a design feature. It also answers the fairness concern families will raise: every building gets the program by spring. Hemming et al. (2015) caution that the design needs careful analysis because time trends, such as the winter rise in illness absences, are tangled with the rollout. Our analysis will therefore adjust for calendar period, and the randomization of step order protects against the authority choosing its best-run buildings first.

Limits

Nine clusters is a small number, so the trial will estimate the effect with wide uncertainty. Monitoring finds infestations sooner but does not guarantee they are cleared; in one small trial, intensive treatment eradicated bed bugs from only half of treated apartments within 10 weeks (Wang et al., 2009). Any effect on absence would therefore run through faster treatment and would likely be modest. The memo recommends that the trial be described as a test of earlier detection, not of eradication.

Recommendation

Approve Candidate C as a stepped-wedge trial across the nine buildings, with step order drawn by an analyst outside the authority and recorded before the school year begins.

References

Hemming, K., Haines, T. P., Chilton, P. J., Girling, A. J., & Lilford, R. J. (2015). The stepped wedge cluster randomised trial: Rationale, design, analysis, and reporting. BMJ, 350, Article h391. https://doi.org/10.1136/bmj.h391

Torgerson, D. J. (2001). Contamination in trials: Is cluster randomisation the answer? BMJ, 322(7282), 355-357. https://doi.org/10.1136/bmj.322.7282.355

Wang, C., Gibb, T., & Bennett, G. W. (2009). Evaluation of two least toxic integrated pest management programs for managing bed bugs (Heteroptera: Cimicidae) with discussion of a bed bug intercepting device. Journal of Medical Entomology, 46(3), 566-571. https://doi.org/10.1603/033.046.0322

Wang, C., Singh, N., Zha, C., & Cooper, R. (2016). Bed bugs: Prevalence in low-income communities, resident's reactions, and implementation of a low-cost inspection protocol. Journal of Medical Entomology, 53(3), 639-646. https://doi.org/10.1093/jme/tjw018