Partial identification of suicide outcome disparities under incomplete recording in United States high school students
Background A group difference estimated among students with a recorded response and the corresponding population difference are distinct quantities when sensitive outcomes are incompletely recorded. Methods We analyzed five national Youth Risk Behavior Survey waves from 2015 to 2023; unweighted N = 81,401. Ideation, planning, attempt, and medically treated attempt were analyzed separately among students whose group variable was recorded and classifiable. We report recorded outcome differences and marginal regions allowing unrecorded binary outcomes to vary over their full logical range, with logically constrained joint regions for attempt and treated attempt. A common-scale maximum statistic from 2,000 design-based bootstrap replicates provided simultaneous uncertainty control for the family of 80 comparisons. Threshold, small cell, harmonization, breakdown, and simulation analyses establish how far each conclusion extends. Results Individual outcome recording ranged from 77.7% to 98.8%. Joint constraints reduced average region width by 0.63 percentage points for attempt and 13.27 percentage points for treated attempt. Thirty-one identification regions excluded zero. Cellwise percentile envelopes retained 22 directions, whereas the primary simultaneous band retained nine: all five gay, lesbian, or bisexual minus heterosexual ideation comparisons and four of the five corresponding planning comparisons. The primary procedure achieved 94.75% to 97.25% family-wise coverage across simulation conditions; the studentized diagnostic achieved 55.75% to 76.75%. Among 49 directionally reversible cells, the median fraction of movement from the recorded difference toward its adverse endpoint needed to reach zero was 0.253. Conclusion Recorded outcome estimates, identification regions, and simultaneous sampling uncertainty answer different questions. The nine retained comparisons are hypothesis-generating candidate signals that call for prospective confirmation. Interpreting a precise recorded response estimate requires jointly examining incomplete recording and family-wide uncertainty.
Authors
- Mengmeng Guo (ORCID: https://orcid.org/0000-0001-9410-0253)
- Junjie Zhang (ORCID: https://orcid.org/0000-0002-4299-6112)
Institutions
- Shanghai International Studies University (CN)
- East China Normal University (CN)
Publication Details
- Journal
- Frontiers in Psychology
- Published
- 2026-11-18
- DOI
- https://doi.org/10.3389/fpsyg.2026.1943174
- Primary Topic
- Suicide and Self-Harm Studies
- Type
- article
- Field-Weighted Citation Impact
- 0.00