Investigate NHANES vs. meta-analytic MSM prevalence: source comparison

[smjenness]

Context

We have used the Grey et al. 2016 estimate (~102K MSM for the Atlanta CBSA, anchored to Purcell 2012's meta-analytic 3.9%) for years. The new Islek et al. 2026 paper uses NHANES 2015 to March 2020 directly with a 12-month-recall prevalence of 1.75%, producing ~53K for the same CBSA. The recall-window difference (5-year vs 12-month) is the dominant numerical driver of the gap, but a methodological-bias component is suspected as well (NHANES in-home administration may under-report male-male behavior relative to anonymous modes).

This issue scopes the literature work needed to choose between (or combine) these prevalence sources.

Investigation tasks

• Catalog the surveys feeding Purcell et al. 2012's meta-analytic 3.9%: GSS, NSFG, NHSLS, NHANES. Record administration mode (in-home vs telephone vs web vs ACASI within clinical visit) for each.
• Document NHANES's administration mode (in-home with ACASI). Identify any published evidence for under-reporting of male-male behavior in in-home settings vs anonymous modes.
• Find any head-to-head comparisons of MSM prevalence across BRFSS, NSFG, GSS, NHANES, and Add Health for comparable cohorts.
• Pull Bennett et al. 2024 (cited as Islek ref 6, reports 3.3%) and document its methodology, recall window, and source surveys.
• Tabulate prevalence estimates by recall window and administration mode side-by-side; assess whether the Grey/Islek gap is fully explained by recall-window definition or whether mode bias is also implicated.

Priority

High. Directly informs the decision on which N to use.

Source

Plan: inst/popsize/msm_population_size_discussion.md, §Q1.