Warning: This is an author submission that is not peer-reviewed or edited. Preprints - unless they show as "accepted" - should not be relied on to guide clinical practice or health-related behavior and should not be reported in news media as established information.
Motor Crash Compensation Psychosocial Recovery Journeys: A Comparison of Large Language Model Insights
ABSTRACT
Background:
Artificial intelligence is a potentially useful tool to derive insights from data of both structured and unstructured data types. The psychosocial recovery journey of motor crash injured people in compensable environments is complex, and Large Language Models (LLMs) are tools that could contribute to understanding these challenging experiences.
Objective:
The aim of this study was to model factors important in recovery from a motor crash injury in a compensable environment. Three LLMs were used to derive insights from a data set comprising of psychometric test data and free form notes from telephone coaching interviews and therapy notes of motor crash injured people, by generating recovery personas, intervention personas and personas for feigning spectrum behaviour (FSB).
Methods:
We used Claude 4.5 Sonnet (“thinking”), ChatGPT 5 and Gemini 2.5 Pro (“thinking”) models in a “zero shot” persona generation process on a deidentified data set of 711 motor crash injured people who were supported and treated by the Navigator Group Motor Crash Support Program.
Results:
The LLMs generated eight FSB personas generated with FSB occurring in 13.5% of the sample. There were 12 Recovery Personas and 17 Intervention Personas generated. Term Frequency-Inverse Document Frequency cosine (TF-IDF) assessment of lexical overlap showed the LLMs use distinct vocabularies and terminology (TF-IDF Cosine Mean=.068-.071) but had moderate similarity for semantic patterns (TF-IDF embedded Cosine Mean=.400-.499). Claude 4.5 Sonnet spontaneously generated “Clinical Decision Frameworks” and “Recommendations for Clinical Practice” with many interventions recommended for each persona.
Conclusions:
Systematic modeling with LLMs derived many insights from the data which should be considered dialectically to improve clinical practice. For the tasks of this project, Claude 4.5 Sonnet “Thinking” performed well and spontaneously extended the task in a meaningful way showing adaptation and learning on a “zero shot” prompt.
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.