Article Text

Download PDFPDF
Validation of a model for the prediction of retinopathy in persons with type 1 diabetes
  1. Vivian Schreur1,
  2. Heijan Ng1,
  3. Giels Nijpels2,
  4. Einar Stefánsson3,
  5. Cees J Tack4,
  6. B Jeroen Klevering1,
  7. Eiko K de Jong1,
  8. Carel B Hoyng1,
  9. Jan E E Keunen1,
  10. Amber A van der Heijden2
  1. 1 Department of Ophthalmology, Donders Institute for Brain, Cognition and Behaviour, Radboud University Medical Center, Nijmegen, The Netherlands
  2. 2 Department of General Practice and Elderly Care Medicine, Amsterdam Public Health Research Institute, VU Medical Centre, Amsterdam, The Netherlands
  3. 3 Department of Ophthalmology, University of Iceland, Reykjavík, Iceland
  4. 4 Department of Internal Medicine, Radboud University Medical Center, Nijmegen, The Netherlands
  1. Correspondence to Dr Vivian Schreur, Ophthalmology, Radboud university medical center, Donders Institute for Brain, Cognition and Behaviour, Nijmegen 6525 EX, The Netherlands; Vivian.Schreur{at}


Background/Aim To validate a previously developed model for prediction of diabetic retinopathy (DR) for personalised retinopathy screening in persons with type 1 diabetes.

Methods Retrospective medical data of persons with type 1 diabetes treated in an academic hospital setting were used for analysis. Sight-threatening retinopathy (STR) was defined as the presence of severe non-proliferative DR, proliferative DR or macular oedema. The presence and grade of retinopathy, onset of diabetes, systolic blood pressure, and levels of haemoglobin A1c were used to calculate an individual risk estimate and personalised screening interval. In persons with STR, the occurrence was compared with the calculated date of screening. The model’s predictive performance was measured using calibration and discrimination techniques.

Results Of the 268 persons included in our study, 24 (9.0%) developed STR during a mean follow-up of 4.6 years. All incidences of STR occurred after the calculated screening date. By applying the model, the mean calculated screening interval was 30.5 months, which is a reduction in screening frequency of 61% compared with annual screening and 21% compared with biennial screening. The discriminatory ability was good (Harrell’s C-statistic=0.82, 95% CI 0.74 to 0.90), and calibration showed an overestimation of risk in persons who were assigned to a higher risk for STR.

Conclusion This validation study suggests that a screening programme based on the previously developed prediction model is safe and efficient. The use of a personalised screening frequency could improve cost-effectiveness of diabetic eye care.

  • diagnostic tests
  • investigation
  • retina
  • epidemiology

Data availability statement

Deidentified participant data and statistical code are available upon request.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Data availability statement

Deidentified participant data and statistical code are available upon request.

View Full Text


  • VS and HN contributed equally.

  • Contributors VS and HN contributed equally to this work and share first authorship of this manuscript. VS and HN collected the data, HN, VS and AAvdH analysed the data and wrote the manuscript.

  • Funding Support from BBMRI grant (enrichment of cohort with ophthalmological phenotyping): CJT. Support from Stichting Blindenhulp: EKdJ.

  • Competing interests None declared.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Linked Articles

  • At a glance
    Frank Larkin