Test-Retest Reliability and Minimal Detectable Change of a Brief Autism Screening Instrument: Evidence from 5,456 Brazilian Adults in a Naturalistic Online Setting

# Test-Retest Reliability and Minimal Detectable Change of a Brief Autism Screening Instrument: Evidence from 5,456 Brazilian Adults in a Naturalistic Online Setting ## Authors - **Unknown** — ORCID: [0000-0001-8815-7309](https://orcid.org/0000-0001-8815-7309) - **Lenifer Siqueira Landim** — ORCID: [0009-0000-9305-1250](https://orcid.org/0009-0000-9305-1250) - **Charlene Coleman** — ORCID: [0009-0004-5829-4854](https://orcid.org/0009-0004-5829-4854) - **Ana Maria Luna Neri Benevides** — ORCID: [0000-0002-7723-8127](https://orcid.org/0000-0002-7723-8127) ## Abstract **Background:** Temporal stability is a fundamental psychometric property of screening instruments, yet large-scale test-retest data for brief autism screening tools in naturalistic settings remain scarce. **Methods:** We analyzed 5,456 Brazilian adults who completed TEAD, a 10-item autism screening instrument, at two time points (median interval 139 days, range 30–1,292). Demographic data were available for a 31.6% subsample (n = 1,723; 74.1% women, mean age 32.4 years). We computed ICC(A,1), SEM, MDC₉₅, MDC₉₀, Cronbach's alpha, McDonald's omega, and Bland-Altman limits of agreement. Time-window sensitivity analyses assessed reliability across five pre-defined interval bands. An exploratory structural analysis of T1 items examined dimensionality. **Results:** TEAD showed moderate test-retest reliability: ICC = 0.574 (95% CI 0.552–0.595). SEM = 2.66; MDC₉₅ = 7.36 points (36.8% of 0–20 scale). Bland-Altman 95% LoA: [−6.65, +7.92], spanning 73% of the scale range. Mean score change: +0.64 (d = 0.17, p < .001). ICC remained moderate across all interval bands (0.52–0.62). Exploratory structural analysis of T1 items revealed moderate multidimensionality (eigenvalues 2.64, 1.12, 1.10) consistent with the intended two-domain structure. **Conclusions:** TEAD demonstrates moderate temporal stability suitable for population-level screening. The MDC₉₅ of 7.36 points (37% of scale) limits individual-level change monitoring. This is, to our knowledge, the largest test-retest study of a brief autism screening instrument reported to date. ## Contents | File | Description | |------|-------------| | `COVER_LETTER.md` | | | `F1_study_flow_intervals.png` | | | `F2_icc_by_interval.png` | | | `F3_sem_mdc_bland_altman.png` | | | `FIGURE_LIST.md` | | | `MANUSCRIPT.md` | | | `MANUSCRIPT.pdf` | | | `README_OSF.md` | | | `T1_sample_characteristics.csv` | | | `T2_icc_estimates.csv` | | | `T3_sem_mdc.csv` | | | `T4_interval_stratified.csv` | | | `results_summary.json` | | | `run_tead_analysis.py` | | ## License CC-BY 4.0 ## Preprint Registered on psyarxiv: [DOI to be assigned] --- *Repository created by neuratip-preprint-agent on 2026-07-21* --- ## File Correction Log — 2026-08-18 The manuscript PDFs previously deposited on this node stated an incorrect institutional affiliation for co-author Charlene Coleman and an incorrect competing-interests declaration. Charlene Coleman is affiliated with the **NeurAtip, Dublin, Ireland**, and is a **partner and equity holder in NeurAtip**. The affected files have been removed and will be replaced with corrected versions. The dataset, figures and tables on this node are unaffected.