Published September 3, 2026
| Version v1
Dataset
Open
dlm12138 pneumonia-k4-physiological-phenotyping-v1.2
Authors/Creators
Description
This record contains the reproducibility code, aggregated results, and methodological documentation accompanying the study “Early physiological phenotypes and mortality within 24 hours after a 12-hour landmark in pneumonia-associated critical illness: a three-database cohort study.”
Version 1.1 corrects the provenance of SpO2-derived variables used in selected severity-adjustment, clinical-characterization, and incremental predictive-performance analyses. The corrected analyses use verified 0–11-hour SpO2 source data.
The frozen K4 model structure, phenotype definitions, phenotype assignments, cohort definitions, observation window, 12-hour landmark, primary outcome definition, and primary P3-versus-P2 effect estimates are unchanged.
The primary P3-versus-P2 odds ratios remain:
- MIMIC-IV: OR 1.7337
- MIMIC-III: OR 4.3928
- eICU: OR 3.6389
The corrected extended severity-adjusted P3-versus-P2 estimates are:
- MIMIC-IV: OR 1.6323 (95% CI 1.1396–2.3379)
- MIMIC-III: OR 3.9317 (95% CI 1.6844–9.1775)
- eICU: OR 2.8597 (95% CI 1.9741–4.1426)
The incremental predictive-performance analyses were also updated using the corrected SpO2 source. Addition of K4 to conventional physiological summaries did not demonstrate a consistent or clinically substantial improvement in discrimination across the three databases.
A textual variance-estimator label was additionally corrected from “HC1 robust” to “HC0 sandwich robust” to match the actual implementation. This labeling correction did not change any numerical estimates.
The archived materials include code for feature preparation, locked K4 phenotype assignment, primary and sensitivity analyses, severity-adjustment analyses, comparison with a fixed heart-rate/respiratory-rate rule, incremental predictive-performance analyses, domain-shift diagnostics, time-order falsification analyses, hospital-level transportability analyses, and aggregated study results.
The K4 framework should be interpreted as a multitime-point physiological stratification model rather than as a disease-specific biological subtype or a clinically validated treatment-selection tool. The P3 pattern is most appropriately interpreted as an early cardiorespiratory physiological stress pattern and residual illness severity.
No patient-level MIMIC-III, MIMIC-IV, or eICU data are included or redistributed in this archive. Frozen K4 model parameter artifacts derived from credentialed-access clinical data are not publicly redistributed and are handled separately in accordance with the applicable data-access and data-use requirements.
This version supersedes v1.0 for the analyses affected by the corrected SpO2 source. The historical v1.0 record remains unchanged as part of the permanent Zenodo version history.
The materials are provided to support transparency and reproducibility of the corresponding BMC Pulmonary Medicine manuscript.
Repository:
https://github.com/dlm12138/pneumonia-k4-physiological-phenotyping
Files
coverage_sensitivity.csv
Files
(141.9 kB)
| Name | Size | |
|---|---|---|
|
md5:c2ba68186299fc3a8947cce1935c54ad
|
3.0 kB | Download |
|
md5:3d4becd4b9dd0b74fe2687de866dae95
|
6.4 kB | Download |
|
md5:79fd191b2c1448f63575e7abe05d7e79
|
1.6 kB | Download |
|
md5:f8e15e54e06f23eb3d00a480c61617e5
|
3.3 kB | Download |
|
md5:492712b3a78cec89f8dcfdd21ab1d8c1
|
2.3 kB | Download |
|
md5:6d60b860a3b49967bb8ecc14cb576b83
|
5.7 kB | Download |
|
md5:62f28018f565331dd87d1d489d105cd6
|
9.6 kB | Download |
|
md5:9e8cc2b7179bebe4e40233afb0dc2272
|
6.2 kB | Download |
|
md5:64f0cb7f02c0694d0065ba079bb950a1
|
7.1 kB | Download |
|
md5:76835f2c5982b11850fc2ab6c1ce59ac
|
10.8 kB | Download |
|
md5:2285aa19d5ddba039549f9cb8d3040e2
|
14.5 kB | Preview Download |
|
md5:88768aaad26f7587477184527cb273f2
|
1.6 kB | Preview Download |
|
md5:018ccd8fe5628f44da763dafe029ecd3
|
1.2 kB | Preview Download |
|
md5:904f88880b2c091b47f43b54c4a79229
|
10.3 kB | Preview Download |
|
md5:6eae7a3d2d3607d21975da0d306dbb76
|
601 Bytes | Preview Download |
|
md5:930bd6f2a310f9a200965f823471cfb4
|
7.5 kB | Preview Download |
|
md5:722b4031757153a0ef8dedfe6605f12f
|
12.0 kB | Preview Download |
|
md5:e7ae581e32caf5fc58bb374604784a5f
|
2.1 kB | Preview Download |
|
md5:29c53d7fc9b9a827aa38f3c425926474
|
1.2 kB | Preview Download |
|
md5:f28124490061e086dd6443cdea4a1da9
|
5.7 kB | Preview Download |
|
md5:ddc76c5bb819720faaf78c534ae2e842
|
4.9 kB | Preview Download |
|
md5:f752c1c927d40aa61b87133c1cfec4b5
|
3.6 kB | Preview Download |
|
md5:35d63908cf53644528add4a473c3ace9
|
97 Bytes | Preview Download |
|
md5:11cfd83d58ec7e83ed6ce36e1e1f2342
|
13.2 kB | Preview Download |
|
md5:19aea8cee9d63b1a1062a3a8ee1dadcf
|
4.0 kB | Preview Download |
|
md5:97c2bf27f3d319a0062fa4022b4f6473
|
3.5 kB | Preview Download |