June 2015
Volume 56, Issue 7
ARVO Annual Meeting Abstract  |   June 2015
A comparison of strategies for the detection of retinopathy of prematurity: a microsimulation study
Author Affiliations & Notes
  • Alex R. Kemper
    Department of Pediatrics, Duke University, Durham, NC
  • Lisa A. Prosser
    University of Michigan, Child Health Evaluation and Research Unit, Ann Arbor, MI
  • Kelly Wade
    Department of Pediatrics, Children's Hospital of Philadelphia, Philadelphia, PA
  • Gui-Shuang Ying
    Department of Ophthalmology, Perelman School of Medicine, Philadelphia, PA
  • Agnieshka Baumritter
    Division of Pediatric Ophthalmology, University of Pennsylvania, Philadelphia, PA
  • Graham E Quinn
    Division of Pediatric Ophthalmology, University of Pennsylvania, Philadelphia, PA
  • Footnotes
    Commercial Relationships Alex Kemper, None; Lisa Prosser, None; Kelly Wade, None; Gui-Shuang Ying, None; Agnieshka Baumritter, None; Graham Quinn, None
  • Footnotes
    Support None
Investigative Ophthalmology & Visual Science June 2015, Vol.56, 945. doi:
  • Views
  • Share
  • Tools
    • Alerts
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      Alex R. Kemper, Lisa A. Prosser, Kelly Wade, Gui-Shuang Ying, Agnieshka Baumritter, Graham E Quinn, e-ROP Cooperative Group; A comparison of strategies for the detection of retinopathy of prematurity: a microsimulation study. Invest. Ophthalmol. Vis. Sci. 2015;56(7 ):945.

      Download citation file:

      © ARVO (1962-2015); The Authors (2016-present)

  • Supplements

The detection of Type 1 retinopathy of prematurity (ROP) usually relies on repeated eye exams. A potentially more efficient strategy would be to use telemedicine, with remotely evaluated retinal digital images. Although this approach has been adopted by some neonatal intensive care units (NICUs), little is known about the relative comparative effectiveness. The objective is to compare expected outcomes from different Type 1 ROP detection approaches.


We used a microsimulation model with weekly cycles to simulate ROP status and detection strategy outcomes. We simulated a cohort of 650 infants, the typical number of annual admissions to NICUs affiliated with children's hospitals, using probabilities taken from the e-ROP study (n=1,257 from Level III NICUs). We assumed that evaluation for ROP would begin at post-menstrual age (PMA) 32 weeks for those with GA < 30 weeks and at 34 weeks for GA of 30 weeks. Exams were considered to be the gold standard and that evaluation would stop at the time of discharge or after 40 weeks PMA, or upon the diagnosis of Type 1 ROP or mature retinae. Telemedicine findings suggestive of Type 2 or Type 1 ROP would lead to an exam. Telemedicine cannot identify mature retinae because of imaging limitations.


The table summarizes the expected number of exams, digital imaging sessions, Type 1 ROP cases, and the number that need follow-up after discharge among the cohort of 650 infants from three strategies: usual care (biweekly exams after immature retinae but otherwise weekly), telemedicine alone, and telemedicine with an exam at discharge. The number of exams and screens required for each scenario varied significantly by GA because those with greater GA had a lower overall risk of Type 1 ROP and would be discharged earlier. For example, usual care among those with GA≤25 weeks requires an average of 5.4 exams/infant and telemedicine requires 2.6 exams/infant and 4.9 screens/infant compared to, among those with GA=30 weeks, 1.4 exam/infant for usual care and 0.5 exams and 1.5 screens/infant for telemedicine. All results were robust to sensitivity analyses.


The optimal strategy for ROP detection depends upon access to eye care services, the NICU case- mix, and the ability to follow-up infants after discharge.  

Table. Expected Outcomes
Table. Expected Outcomes


This PDF is available to Subscribers Only

Sign in or purchase a subscription to access this content. ×

You must be signed into an individual account to use this feature.