Patient Weight Gain Narratives for Obesity Phenotyping and Outcome Prediction

Miksa M. Henkrich's dataset contains 2,463 patient-reported weight gain narratives labeled by a GPT4.1 model into 12 thematic categories. The data includes patient demographics, treatment outcomes, and unsupervised clustering results. It was last updated on April 14, 2026.

Use Cases

• Automated thematic labeling of patient narratives based on the 12 predefined categories.
• Predicting weight loss treatment outcomes based on reported cause patterns and patient demographics.
• Identifying patient phenotypes via unsupervised clustering of narrative themes, age, sex, and BMI.
• Analyzing associations between specific weight gain causes, such as mental health or disrupted schedules, and unhealthy eating habits.
• Modeling co-occurrence patterns of weight gain causes using direct association networks.

Strengths

• Contains 2,463 patient records with open-ended narratives.
• Automated labeling achieved precision of 0.906 and recall of 0.897 against a reference sample.
• Mean weight loss outcome of 9.2 ± 6.8% over 108.6 ± 111.6 days is documented.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal or source bias inherent to the specific clinical study.

Provenance

Source

figshare

Collection Method

Narratives collected from patients prior to starting a multidisciplinary medical-nutritional weight loss treatment.

Freshness

Last updated 2026-04-14 16:01:34