Latent transition learning maps short-term changes in Parkinson clinical profiles
Source check: Source checked. Not independent scientific confirmation.
Finding: A probabilistically learned latent-transition model followed 250 people with Parkinson disease across two clinical assessments about 2.5 years apart. Most retained their model-defined profile, while a fourth severe multidomain profile emerged at follow-up and included a high proportion with dementia. These are observational clinical groupings and transitions, without established biological subtypes or lifetime convergence trajectories.
Technology: Latent class mixture learning; latent transition analysis in Mplus
Publication: Peer-reviewed paper · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Probabilistic latent-mixture and transition learning estimated multidomain classes and membership transitions from clinical measurements; these learned representations enabled the reported stability and emergent severe-profile finding. Conventional secondary logistic regression is not the enabling AI claim.
- Test and result
- Eleven motor, cognitive and psychiatric indicators defined learned latent classes. Mplus latent-transition models compared three- and four-class follow-up structures using likelihood-based fit, information criteria and entropy; the selected model had three baseline and four follow-up classes. Among 250 analytic participants, 190 retained their class over a mean 2.51 years. The severe multidomain follow-up class had dementia in 64%, versus at most 7% in the other classes. Available autopsy information supported exclusion of nine non-Parkinson diagnoses before the analytic cohort was formed.
- Limitation / still unknown
- Only two assessment times and a predominantly White, non-Hispanic cohort constrain trajectory interpretation. Classes are data-dependent clinical profiles, not validated biological taxonomy. Dementia-related indicators contribute to the learned profiles, so dementia enrichment is not independent diagnostic validation. Same-cohort class transitions and secondary regression associations do not establish causal progression, universal eventual convergence, independent forecasting or treatment benefit.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://pmc.ncbi.nlm.nih.gov/articles/PMC13696777/
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Read original primary latent-class/transition model fitting, measurement-invariance and missing-data Methods, cohort exclusions, class-transition Results, dementia enrichment and observational limitations. Learned non-neural mixture inference is the qualifying model operation; author use of the word AI is not required. Exact canonical online date October25 retained. Exact primary identity/date and fresh primary PubMed correction/retraction relationships checked..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.