The heterogeneity of insomnia presentations has long challenged research and clinical practice, motivating efforts to identify reliable disorder phenotypes. Person-centered, data-driven approaches such as latent class analysis (LCA) have provided new insights, suggesting that insomnia subtypes may differ not only in nocturnal symptoms but also in perceived impact and daytime distress. Despite this progress, LCA solutions often remain confined to the original datasets, limiting replication and applied use. To address this gap, we developed the insomnia-LCA classifier, an open-source web application that assigns new Insomnia Severity Index (ISI) response profiles to one of four subtypes identified in a previously published LCA of Italian university students: no insomnia (NI), subthreshold insomnia (SI), high insomnia risk (HI), and predominant daytime symptoms (DS). Using the original model's class priors and item-level conditional response probabilities, the app computes posterior class probabilities from user-entered ISI responses, individually or in batch mode. Outputs include class probabilities and modal assignment, ISI total and subscale scores, and a visual comparison between the individual profile and subtype mean patterns. Reclassification of the original dataset showed near-perfect agreement with the latent class model (accuracy = 0.999; Cohen's kappa = 0.999), and synthetic profiles behaved as expected. The insomnia-LCA classifier provides a practical, reproducible tool for deploying and testing LCA-derived phenotypes in clinical research.
Carpi et al. (Sat,) studied this question.