Advancing data collection of hospital-related harms: Results from hospital discharges dually coded with ICD-10 and ICD-11
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
IntroductionHospital safety performance is difficult to monitor when under-coding of hospital harms is common. The beta version of ICD-11 includes a 3-part model for coding harms to enhance adverse event descriptions. This method includes code clusters to detail each condition/event (e.g. bleed), cause (e.g. anticoagulant drug), and mode (over-dose).
 Objectives and ApproachThe study objective was to compare the proportion of adverse events captured using different classification systems. A large field trial of inpatient charts, previously coded in ICD-10 were coded with ICD-11. Coding training for the new ICD-11 focused on new codes, code clustering, and extension codes for cause and mode of the harm. Sensitivity, Specificity, NPV and PPV were reported for ICD-10 compared to ICD-11.
 ResultsOf the 1,009 records reviewed and coded using ICD-11 to date, 128 cases were coded as a harm in ICD-10 using our previously published PSI work. Coders identified 88 cases with the new ICD-11. Sensitivity and specificity were as follows: 31.3% and 94.6%. ICD-11 had NPV and PPV of 45.5% and 90.5% respectively compared to ICD-10. Detailed clinical comparison of mismatched codes will be completed. Study case examples will demonstrate advanced features of ICD-11, the coding rules being collaboratively developed by our team, CIHI, and WHO representatives, and potential analytic challenges.
 Conclusion/ImplicationsThe new ICD-11 found 8% of hospital admission were associated with a harm. Although the sensitivity was modest, specificity is quite high and correctly Identifies those cases without a harm. Clinical review of mismatched codes will provide further detailed code comparisons.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.004 | 0.006 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.000 | 0.004 |
| Open science | 0.001 | 0.001 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it