﻿Supplementary materials

This package contains exactly 12 tables, numbered S1–S12, matching the Supplementary Materials section of the manuscript. The table titles below are reproduced verbatim from the manuscript and in the CSV filenames.

Table S1: Conventional-benchmark dataset manifest (reconstructed)
File: Table S1 - Conventional-benchmark dataset manifest (reconstructed).csv

Table S2: Supervised-control dataset manifest
File: Table S2 - Supervised-control dataset manifest.csv

Table S3: US restricted-predictor audit
File: Table S3 - US restricted-predictor audit.csv

Table S4: US conventional-protocol predictor documentation
File: Table S4 - US conventional-protocol predictor documentation.csv

Table S5: STATS19 predictor and missingness audit (reconstructed)
File: Table S5 - STATS19 predictor and missingness audit (reconstructed).csv

Table S6: Indian source-field missingness
File: Table S6 - Indian source-field missingness.csv

Table S7: Indian retained and derived predictors
File: Table S7 - Indian retained and derived predictors.csv

Table S8: Indian severity code definitions and counts
File: Table S8 - Indian severity code definitions and counts.csv

Table S9: Dataset-level kNN-DART results
File: Table S9 - Dataset-level kNN-DART results.csv

Table S10: US matched result summaries
File: Table S10 - US matched result summaries.csv

Table S11: STATS19 supervised-control result summary
File: Table S11 - STATS19 supervised-control result summary.csv

Table S12: Protocol-specific table-value comparison
File: Table S12 - Protocol-specific table-value comparison.csv

Scope and provenance
Tables S1 and S5 are reconstructions from the available source data and preparation rules. Exact historical source versions and sampled-record memberships are not independently established.
The Indian severity definitions in Table S8 follow manuscript References [25, 26]. Its count_source field identifies the original audit source; that source path is provenance, not an additional file in this package.
Table S12 compares values from distinct experimental protocols; comparisons between methods should be made within their corresponding protocol.
The CSV files retain their column headers and numerical data. Tables contain no additional numbered supplementary items. README.txt is retained solely as this matching index and scope note.
