Pediatric Gastrointestinal Risk Dataset (PedGI-FRAD)

The PedGI-FRAD 2026 dataset is a large-scale longitudinal pediatric gastrointestinal healthcare dataset containing 485,605 healthcare observations associated with 18,000 pediatric patient trajectories distributed across five healthcare clients. The records span from February 2018 to March 2026 and occur at irregular observation intervals reflecting differences in clinical follow-up, physiological monitoring, laboratory assessment, and digital healthcare interactions. Each patient remains associated with a single healthcare client throughout the observation period. Patient ages range from 1 to 17 years, with a mean age of 9.41 years. Each patient contributes a mean of 26.98 observations, with a median of 24 and an interquartile range of 15--35 observations. The dataset includes pediatric healthcare information covering demographic and anthropometric characteristics, gastrointestinal symptoms, inflammatory laboratory biomarkers, nutritional and lifestyle factors, medication and clinical history, disease progression indicators, wearable physiological measurements, and healthcare context variables. Gastrointestinal symptom attributes include abdominal pain, bloating, nausea, vomiting episodes, diarrhea frequency, constipation duration, stool frequency and consistency, blood and mucus in stool, appetite variation, and nocturnal gastrointestinal discomfort. Laboratory variables include C-reactive protein, erythrocyte sedimentation rate, fecal calprotectin, white blood cell count, platelet count, hemoglobin, albumin, vitamin D, and iron measurements. Anthropometric information includes age, weight, height, body mass index, and weight percentile. A limited processing stage was applied to organize patient observations, maintain consistency across repeated records, verify anthropometric ranges, identify incomplete measurements, and prepare the data for analysis. In addition to the directly available healthcare variables, several derived clinical and progression-related features were calculated from repeated patient observations. These include 7-day symptom progression, pain progression rate, stool abnormality trends, 30-day weight-change progression, 90-day relapse frequency, and 6-month clinical-visit frequency. Physiological monitoring variables include resting heart rate, heart-rate variability, sleep efficiency, stress score, body temperature, sleep duration, and physical activity. Healthcare data sources are represented through electronic health record, wearable, mobile-application, and connected-monitoring categories. Incomplete observations are retained in the dataset because the availability of laboratory, wearable, symptom, lifestyle, and clinical measurements varies across healthcare encounters. Overall, approximately 13.50% of feature values are missing. Additional context variables include healthcare-client identifier, device source, regional category, and record-level missing-value ratio. The primary outcome variable, GI_Risk_Level, represents binary pediatric gastrointestinal risk classification, comprising 66.94% low-risk and 33.06% high-risk observations. The additional GI_Severity_Level variable provides gastrointestinal severity stratification.