Validation of the CHOP Model of ROP
JAMA Ophthalmology, August 2017
The retinopathy of prematurity (ROP) model of the Children’s Hospital of Philadelphia (CHOP) uses factors such as gestational age, birth weight, and rate of weight gain to predict ROP severity. Binenbaum et al. tested the model’s utility in a population of sufficient size to determine sensitivity for detecting ROP that requires treatment. The model proved valid for clinical use and may help clinicians refine ROP screening guidelines.
The investigation was a secondary analysis of longitudinal data from the Postnatal Growth and ROP Study, which involved 30 hospitals throughout the United States and Canada. The study cohort comprised 7,483 premature infants at risk for ROP with a known ROP outcome. Median gestational age was 28 weeks (range, 22-35 weeks), and median birth weight was 1,070 g (range, 310-3,000 g). The CHOP ROP model was applied weekly to assess its ability to predict ROP risk. If the risk was deemed high, examinations were provided. If the risk was considered low, examinations were not performed. In another test scenario, low-risk infants received fewer exams rather than none.
The model accurately identified 452 of the 459 infants found to have type 1 ROP (98.5% sensitivity). If exams were given only to infants with high ROP risk, the proportion of infants requiring exams would be reduced by 34.3%. Lowering the cutpoint to capture all cases of type 1 ROP (100% sensitivity) would result in examining all but 6.8% of infants. However, if low-risk infants were evaluated at 37 weeks’ postmenstrual age and monitored only if ROP was present at that time, all cases of type 1 ROP would be captured, and the number of exams among infants born after 27 weeks’ gestation would be reduced by 28.4%.
In conclusion, the sensitivity of the CHOP ROP model was high but shy of 100%. However, the model reduces the frequency of examinations and may guide development of improved ROP screening guidelines, allowing resources to be directed to the infants in greatest need of them.
The original article can be found here.