time-series-consistency

name	time-series-consistency
description	Detect temporal distribution shifts (change points) in clinical data across years using GradientBoosting classifier + SHAP. For each year, trains a model to distinguish year-1 vs year data — AUROC ≥ 0.8 indicates a change point. No API key required. Use for LYDUS data quality assessment of temporal consistency across data collection periods.
tier	community
category	lydus
parameters	{"quiq_path":{"description":"Path to QUIQ-format CSV file. Must contain data spanning multiple years with non-null Event_date.","type":"string"},"save_path":{"description":"Directory path to save output files and plots.","type":"string"}}

Time Series Consistency

Detects temporal distribution shifts (change points) in clinical data by training a GradientBoosting classifier to distinguish consecutive year pairs. If AUROC ≥ 0.8, the distribution between year-1 and year differs significantly — indicating a change point (e.g., coding practice change, dataset collection change).

When to Use This Skill

• After QUIQ conversion, to detect whether clinical variable distributions have shifted over time
• To identify years with sudden changes in documentation or coding patterns
• To assess longitudinal stability of data collection
• As part of LYDUS quality management assessment