MindHealth Observatory: Depression screening, population inequality and survey-weighted inference

01 · Context

Research question

How does the prevalence of elevated depressive symptoms differ across socioeconomic and demographic strata once the NHANES sample design is respected?

02 · Methods

Approach

• Validate all nine PHQ-9 symptoms and survey eligibility before scoring.
• Use 2-year MEC weights, masked strata and PSU identifiers for design-based estimates.
• Estimate weighted prevalence, survey-weighted GLMs, design-aware confidence intervals and nonlinear income sensitivity.
• Audit complete-case selection and report all weights, denominators and exclusion rules.

03 · Evaluation

Reported results

• An R 4.5.0 execution on CDC NHANES 2017–2018 analyzed 5,068 eligible complete PHQ-9 respondents. Design-weighted PHQ-9 ≥10 prevalence was 8.53% (95% CI 7.52%–9.55%; SE 0.52 percentage points). This is a screening threshold, not a clinical diagnosis.
• The observed execution artifacts, input hashes, package versions and restrictions are documented in the research evidence ledger. Full reproducibility and sensitivity gates are still outstanding.
• Interpretation remains exploratory and no causal, diagnostic or peer-reviewed conclusion is asserted.
• Fresh design-based sensitivity on the same 5,068 respondents found PHQ-9 screen-positive percentages of 24.72% (cutoff 5), 8.53% (cutoff 10), and 2.81% (cutoff 15). These are descriptive screening cutoffs, not diagnoses.