Abstract
Purpose::
We previously demonstrated that a predictive model of risk for an outcome of threshold or severe prethreshold ROP warranting surgery (ROPWS) could be used to select low-risk eyes for alternative screening protocols that reduce eye examinations without excessive delay to the detection of ROPWS. This study extends the analysis to a model outcome variable of threshold or type 1 ROP, to account for current recommendations for early treatment of type 1 ROP.
Methods::
This is a retrospective chart review of 299 infants 401-1250 g at birth admitted to a level 3 NICU from 1998 to 2003. We performed logistic regression for outcome variable of threshold or type 1 ROP using predictive variables of birth weight, gestational age, multiple birth, race, gender, and CRIBSUB (Clinical Risk Index for Babies Subscore). Correction for inter-eye correlation within the same infant was obtained with generalized estimating equations. This model has high predictive accuracy by receiver operating characteristics (area under the curve = 0.870). Using a probability of ≥ 0.15 as the cut-off for high risk, the specificity and sensitivity of this model were 78.5% and 85.1%, respectively. For eyes assigned as low-risk, an alternative screening protocol with first examination at 35 wks PMA followed by every 3 wks was simulated while high-risk eyes were screened normally. Reductions in eye examinations and delay to detection of threshold or type 1 ROP due to the alternative screening protocols were calculated by comparison with 1) the actual sequence of eye examinations or 2) an ideal sequence of eye examinations that should have been performed according to an idealized screening protocol.
Results::
Using alternative screening for low-risk eyes and conventional screening for high-risk eyes, the overall reductions in eye examinations was 17.3% to 22% (p<0.05) . There was no delay to detection of threshold ROP. Six eyes (10.9%) had a delay to detection of type 1 ROP of 1 (3 eyes) and 2 (3 eyes) weeks, respectively. Two of the 6 type 1 ROP eyes subsequently developed threshold, were treated, and had favorable structural outcome at ≥ 3 months. The 4 remaining type 1 ROP eyes spontaneously regressed without developing threshold ROP.
Conclusions::
A predictive model of risk for threshold or type 1 ROP outcome combined with alternative screening for low-risk eyes can safely reduce ROP examinations for premature infants ≤ 1250 g. Potential benefits include cost-savings and reduced exposure of premature infants to the risks associated with examinations. Before changing current screening recommendations, these findings should be confirmed in multicenter studies.
Keywords: retinopathy of prematurity • clinical (human) or epidemiologic studies: health care delivery/economics/manpower