Telehealth autism evaluation under suboptimal network conditions: Technical parameters and clinical implications

Telehealth expands access to autism diagnostics, but scoring reliability depends on connection stability. Using an optical-fiber testbed, we imposed controlled packet loss and Wi-Fi attenuation to derive four freeze patterns (1-second freezes every 16.7, 11.1, 8.3, or 5.0 seconds), then applied these to 10 children's BOSA videos (40 clips) scored by four blinded psychologists. Analyses included Fisher's exact tests with Benjamini-Hochberg correction, Spearman correlation, and per-item logistic regression with Bonferroni correction. Performance degraded in a dose-dependent manner rather than collapsing: below a Received Signal Strength Indicator (RSSI) of -70 dBm or above ~0.04% packet loss, freeze frequency increased from ~1/16.7 s to ~1/5.0 s. Time-sensitive items-Spontaneous Joint Attention (B10), Integration of Gaze and Behavior (B4), Immediate Echolalia (A4), and Stereotyped/Idiosyncratic Language (A5)-were 4-9 × more likely to be coded unscorable, and poorer stream quality correlated with more unscored items (ρ ≈ -0.58). Importantly, overall BOSA scoring remained stable, with no individual item reaching significance after correction and substantial inter-rater consistency under mild degradation (ICC = 0.93, Cohen's κ = 0.76), indicating robustness to moderate degradation. In this pilot study, these results support BOSA as a first-line option when packet loss remains below ~0.04% and RSSI stays above -70 dBm, pending validation in larger, more diverse cohorts. Where possible, using wired connections and awareness of item-level sensitivities can further preserve diagnostic confidence.