Validation of the Colorado Retinopathy of Prematurity Screening Model
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
Importance: The Colorado Retinopathy of Prematurity (CO-ROP) model uses birth weight, gestational age, and weight gain at the first month of life (WG-28) to predict risk of severe retinopathy of prematurity (ROP). In previous validation studies, the model performed very well, predicting virtually all cases of severe ROP and potentially reducing the number of infants who need ROP examinations, warranting validation in a larger, more diverse population. Objective: To validate the performance of the CO-ROP model in a large multicenter cohort. Design, Setting, Participants: This study is a secondary analysis of data from the Postnatal Growth and Retinopathy of Prematurity (G-ROP) Study, a retrospective multicenter cohort study conducted in 29 hospitals in the United States and Canada between January 2006 and June 2012 of 6351 premature infants who received ROP examinations. Main Outcomes and Measures: Sensitivity and specificity for severe (early treatment of ROP [ETROP] type 1 or 2) ROP, and reduction in infants receiving examinations. The CO-ROP model was applied to the infants in the G-ROP data set with all 3 data points (infants would have received examinations if they met all 3 criteria: birth weight, <1501 g; gestational age, <30 weeks; and WG-28, <650 g). Infants missing WG-28 information were included in a secondary analysis in which WG-28 was considered fewer than 650 g. Results: Of 7438 infants in the G-ROP study, 3575 (48.1%) were girls, and maternal race/ethnicity was 2310 (31.1%) African American, 3615 (48.6%) white, 233 (3.1%) Asian, 40 (0.52%) American Indian/Alaskan Native, and 93 (1.3%) Pacific Islander. In the study cohort, 747 infants (11.8%) had type 1 or 2 ROP, 2068 (32.6%) had lower-grade ROP, and 3536 (55.6%) had no ROP. The CO-ROP model had a sensitivity of 96.9% (95% CI, 95.4%-97.9%) and a specificity of 40.9% (95% CI, 39.3%-42.5%). It missed 23 (3.1%) infants who developed severe ROP. The CO-ROP model would have reduced the number of infants who received examinations by 26.1% (95% CI, 25.0%-27.2%). Conclusions and Relevance: The CO-ROP model demonstrated high but not 100% sensitivity for severe ROP and missed infants who might require treatment in this large validation cohort. The model requires all 3 criteria to be met to signal a need for examinations, but some infants with a birth weight or gestational age above the thresholds developed severe ROP. Most of these infants who were not detected by the CO-ROP model had obvious deviation in expected weight trajectories or nonphysiologic weight gain. These findings suggest that the CO-ROP model needs to be revised before considering implementation into clinical practice.
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 | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.001 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| 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