Identifying dyslexia-consistent reading profiles in mild intellectual disability: cluster-derived severity gradients and severity-calibrated classification rules.

Introduction: Translating dyslexia constructs to mild intellectual disability (MID) is challenging because population-based cut-offs may label broadly expected low attainment as 'specific'. This study combined person-centered profiling with explicit, ICD-11-guided, severity-calibrated classification rules to quantify when and why over-identification occurs in MID. Methods: Nationwide diagnostic data from 446 students with MID were analyzed using age-normed sten scores from the SB6/18 battery (decoding, reading fluency, reading comprehension, phonetic coding, rapid automatized naming [RAN]). We conducted hierarchical clustering with bootstrap stability and compared six rule-based groupings that systematically varied severity thresholds and inclusion of phonological/RAN indicators. Results: A robust two-cluster solution reflected a dominant severity gradient across reading and related markers. Broad criteria classified 88% of the sample, including many higher-performing cases. Introducing an extreme-severity anchor reduced prevalence (37%) and concentrated classified cases within the lower-performing profile. Notably, when very low decoding (sten = 1) was required, adding phonological/RAN criteria yielded little further classification gain under this rule structure. Conclusions: Dyslexia-consistent profiles can be delineated within mild ID, but prevalence and subgroup coherence depend critically on severity operationalization. Word-level reading-especially decoding and fluency-provides the most robust basis for subgroup identification; phonological/RAN indicators showed limited incremental classification value within the present rule structure. Clinical trial registration: https://clinicaltrials.gov/, identifier NCT06215092.