PUBH 6036 Week 7 Sampling Plan Example

Drawing Children From a Lease File That Does Not List All of Them: A Stratified Sampling Plan for a Composite Bed Bug and School Attendance Cohort Across Nine Family Buildings

Student Name

Walden University

PUBH 6036: Public Health Research Methods

Instructor Name

Month Day, Year

Drawing Children From a Lease File That Does Not List All of Them: A Stratified Sampling Plan for a Composite Bed Bug and School Attendance Cohort Across Nine Family Buildings

The Target Population and the Frame

The study's target population is every child aged 6 to 12 who sleeps most school nights in an apartment in any of the nine family buildings run by the Cedar Hollow Housing Authority. No list of those children exists. The closest thing is the authority's lease file, which records each household's head, the members declared at the last annual recertification and their birth dates. That file is the sampling frame. Sampling guides separate the target population a study wants to describe from the accessible population it can actually reach (Martínez-Mesa et al., 2016), and the distance between the two is where a sampling plan most often goes wrong. This plan measures that distance before drawing anyone.

Who the File Lists and Who It Misses

At the last extract, the file listed 2,140 occupied apartments, of which 860 declared at least one child who will be aged 6 to 12 on September 1. The file is accurate for most households, but it cannot offer up four groups of children:

1. Children living in a unit without being declared, often because a relative moved in after a job loss or eviction and the household fears that reporting an extra occupant would raise the rent or breach the lease.

2. Children who split their week between parents, declared at the other parent's address.

3. Children who joined a household after the last recertification, through birth of a sibling's child, kinship care or a family moving in together.

4. Children in households that recertified late or whose paperwork is in dispute.

These are not random omissions. Undeclared and doubled-up children are likely to live in crowded apartments, where bed bugs spread more easily and sleep is already disrupted. Missing them would bias the study toward the more stable households and could understate any link between infestation and absence.

Reaching Children the File Misses

The screening visit asks the caregiver, "Who usually sleeps here, including anyone staying for now?" and enrolls any child aged 6 to 12 who sleeps in the unit at least four nights a week, declared or not. Because the authority never sees household-level data, this question carries no lease risk, and the permission form says so. Children who split their week are enrolled at the address where they sleep most school nights. These steps cannot reach households who decline the visit, so the plan records how many children are found at screening who were not on the file. That count becomes a measure of the frame's coverage and is reported with the results.

Why Stratify, and by What

Infestation is unevenly spread across buildings. In a survey of 43 low-income New Jersey buildings, rates ranged from 3.8% to 29.5% by building (Wang et al., 2016). A simple random sample of households might, by chance, draw few apartments from the buildings where most infested homes are, leaving the study with too few exposed children. Stratified sampling divides the frame into groups that differ on something important and samples within each group (Martínez-Mesa et al., 2016). Here the strata are formed from the authority's own records of pest work orders in the past 12 months:

Table 1

Strata and Allocation

Note. Infestation assumptions sit inside the building range reported by Wang et al. (2016); the 70% visit rate is an assumption to be checked in the first month.

Allocation and the Exposed Count

The allocation is deliberately disproportionate. Every household in the high stratum is invited, while about 48% of the low stratum is drawn at random within each of its six buildings. Under the stated assumptions, the plan expects about 52 infested households in the high stratum and about 13 in the low stratum, roughly 65 in all. At an average of 1.5 eligible children per household, that yields close to 100 exposed children, the number the team judged necessary for a useful comparison of absence rates.

Weights

Because households in the two strata are selected at different rates, each responding household carries a weight equal to the number of frame households it represents: about 1.43 in the high stratum (300 divided by 210) and about 2.96 in the low stratum (560 divided by 189). Weights will be recalculated from actual visit counts. Estimates for the authority as a whole, such as the share of children living with an infestation, use the weights; the comparison of absences between exposed and unexposed children is analyzed with stratum included in the model.

Nonresponse

Households that decline the visit cannot be described in detail without breaching their privacy, but the frame records building, household size and the age of declared children for everyone. The analysis will compare responders and decliners on those three fields within each stratum and adjust the weights if they differ.

Drawing the Sample in Practice

The draw will be done once, by an analyst at the university, before any household is contacted. The low-history stratum is sorted by building and then by unit number, and a random number generator with a recorded seed selects 270 households, spread across the six buildings in proportion to their size. The seed, the sorted list and the selected units are saved so that anyone can repeat the draw. Selected households who cannot be reached after three attempts on different days and times are counted as nonrespondents; they are not replaced by the next household on the list. Their building and household size still enter the nonresponse comparison described below. Replacement would quietly turn a random sample into a sample of the easiest families to find, who are likely to be the most settled ones.

Checking the Assumptions Early

The plan rests on two guesses: that 70% of invited households will complete a visit, and that infestation will be about three times as common in the high-history buildings. Both will be checked after the first 100 visits. If the visit rate falls below 60%, the team will add evening and weekend visits before inviting more households, and will ask the resident council why families are saying no. If the infestation rate in the high stratum is much lower than assumed, the team will report the expected number of exposed children to the study sponsor, since a smaller exposed group widens every estimate the cohort will produce.

What the Plan Cannot Do

Hulley et al. (2013) note that the findings of a study apply with confidence only to the accessible population from which the sample was drawn. These results will describe children found in nine family buildings of one housing authority, including those the screening question reaches, and will say nothing directly about privately owned rentals, where landlords' treatment practices differ.

References

Hulley, S. B., Cummings, S. R., Browner, W. S., Grady, D. G., & Newman, T. B. (2013). Designing clinical research (4th ed.). Lippincott Williams & Wilkins.

Martínez-Mesa, J., González-Chica, D. A., Duquia, R. P., Bonamigo, R. R., & Bastos, J. L. (2016). Sampling: How to select participants in my research study? Anais Brasileiros de Dermatologia, 91(3), 326-330. https://doi.org/10.1590/abd1806-4841.20165254

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