Simulation and Model Inputs

Each hypothetical infant was modeled by using a program developed in Visual Basic 2012 (Microsoft Corporation, Redmond, WA). We began by assigning GA probabilistically and then modeled the weekly status of the retina from 32 through 40 weeks’ PMA and the outcomes of digital imaging using data from the e-ROP study.^8,16 To assure that the analysis reflected the underlying probability distributions related to the likelihood of developing ROP and the accuracy of digital imaging, a second-order simulation was conducted by replicating the cohort 10 000 times. The initial cohort simulation was based on the point estimate for each probability regarding the likelihood of developing ROP and the accuracy of the digital imaging. Each subsequent cohort simulation was based on sampling from the underlying probability distributions (Supplemental Information).

Statistical analyses were conducted by using Stata version 13 (StataCorp, College Station, TX). All analyses took into account the clustering of probabilities within each replication to generate SEs for the estimates. The Duke University School of Medicine and The Children’s Hospital of Philadelphia institutional review boards approved this study.

Each infant was assigned GA based on the distribution of prematurity in the United States in 2012 as follows: 22 to 25 weeks, 24.1%; 26 to 27 weeks, 20.6%; 28 to 29 weeks, 31.7%; and 30 weeks, 23.6%.¹⁷ The average PMA and SD at the time of discharge or transfer^18,19 decreased with increasing GA: 22 to 25 weeks’ GA had an average discharge at 40 weeks’ PMA (SD: 3); 26 to 27 weeks’ GA had an average discharge at 38 weeks’ PMA (SD: 2); 28 to 29 weeks’ GA had an average discharge at 37 weeks’ PMA (SD: 1.5); and 30 weeks’ GA had an average discharge at 36 weeks’ PMA (SD: 1). For each hypothetical infant, ROP development after discharge or transfer or >40 weeks’ PMA, whichever comes first, was not evaluated.

Cost estimates were based on expected allowed charges for an initial inpatient consultation and subsequent examinations (Supplemental Information).²⁰ The baseline estimated cost of an ROP examination was $104 for the initial examination and $74 for subsequent examinations, with a range of $79 to $312 for the first consultation and $56 to $222 for subsequent ROP examinations. The baseline estimated cost of digital imaging with remote interpretation was $52, with a range of $39 to $156. The cost analysis is from the health care system perspective, reflecting those costs that could accrue in the NICU, and do not reflect the cost of ROP treatment or any services provided after discharge, transfer, or 40 weeks’ PMA. For each ROP detection strategy, we considered the total cost across the hypothetical NICU and the total costs stratified according to GA. As a primary measure of cost-effectiveness, the average cost per infant for each case of type 1 ROP detected was evaluated. Most infants will not develop ROP that requires treatment. Because follow-up after NICU discharge or transfer can be logistically challenging, the average cost per infant for each case in which ROP monitoring has been completed was also evaluated. ROP monitoring is completed for an infant when follow-up is no longer required (eg, mature retina according to ROP examination or considered low-risk according to the specific digital imaging rule). In the analysis, an infant could develop type 1 ROP, but it could have been missed because the infant was classified as no longer requiring monitoring.