June 2023
Volume 64, Issue 8
Open Access
ARVO Annual Meeting Abstract  |   June 2023
Predicting demand for anti-VEGF injections among neovascular AMD patients using retinal fluid measurements and machine learning; a real-world data study using electronic medical records.
Author Affiliations & Notes
  • David M Wright
    Centre for Public Health, Queen's University Belfast, Belfast, Belfast, United Kingdom
  • Dinah Zur
    Tel Aviv Sourasky Medical Center, Tel Aviv, Tel Aviv, Israel
    Tel Aviv University Sackler Faculty of Medicine, Tel Aviv, Tel Aviv, Israel
  • QING WEN
    Centre for Public Health, Queen's University Belfast, Belfast, Belfast, United Kingdom
  • Marganit Gonen-Shahar
    Tel Aviv Sourasky Medical Center, Tel Aviv, Tel Aviv, Israel
    Tel Aviv University Sackler Faculty of Medicine, Tel Aviv, Tel Aviv, Israel
  • Reut Shor
    Tel Aviv Sourasky Medical Center, Tel Aviv, Tel Aviv, Israel
    Tel Aviv University Sackler Faculty of Medicine, Tel Aviv, Tel Aviv, Israel
  • Gidi Benyamini
    Notal Vision Ltd, Tel Aviv, Tel Aviv, Israel
  • Mor Ben-Nun
    Notal Vision Ltd, Tel Aviv, Tel Aviv, Israel
  • Moshe Havilio
    Notal Vision Ltd, Tel Aviv, Tel Aviv, Israel
  • Omer Dor
    Tel Aviv University Sackler Faculty of Medicine, Tel Aviv, Tel Aviv, Israel
    Tel Aviv Sourasky Medical Center, Tel Aviv, Tel Aviv, Israel
  • Anat Loewenstein
    Tel Aviv University Sackler Faculty of Medicine, Tel Aviv, Tel Aviv, Israel
    Tel Aviv Sourasky Medical Center, Tel Aviv, Tel Aviv, Israel
  • Tunde Peto
    Centre for Public Health, Queen's University Belfast, Belfast, Belfast, United Kingdom
  • Footnotes
    Commercial Relationships   David Wright None; Dinah Zur None; QING WEN None; Marganit Gonen-Shahar None; Reut Shor None; Gidi Benyamini Notal Vision, Code E (Employment); Mor Ben-Nun Notal Vision, Code E (Employment); Moshe Havilio Notal Vision, Code E (Employment); Omer Dor None; Anat Loewenstein Notal Vision, Code C (Consultant/Contractor); Tunde Peto None
  • Footnotes
    Support  British Council - British-Israel Academic Exchange (BIRAX)
Investigative Ophthalmology & Visual Science June 2023, Vol.64, 1759. doi:
  • Views
  • Share
  • Tools
    • Alerts
      ×
      This feature is available to authenticated users only.
      Sign In or Create an Account ×
    • Get Citation

      David M Wright, Dinah Zur, QING WEN, Marganit Gonen-Shahar, Reut Shor, Gidi Benyamini, Mor Ben-Nun, Moshe Havilio, Omer Dor, Anat Loewenstein, Tunde Peto; Predicting demand for anti-VEGF injections among neovascular AMD patients using retinal fluid measurements and machine learning; a real-world data study using electronic medical records.. Invest. Ophthalmol. Vis. Sci. 2023;64(8):1759.

      Download citation file:


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

      ×
  • Supplements
Abstract

Purpose : To evaluate the performance of machine learning models of retinal fluid predicting the number of anti-VEGF injections required in the first three years of neovascular AMD treatment.

Methods : In this retrospective, observational clinical study we determined timing of anti-VEGF injections from anonymised electronic medical records of AMD patients at a tertiary treatment centre in Belfast, United Kingdom between 2008 and 2020. Volumes of intra- and sub-retinal fluid and retinal pigment epithelium irregularities (if present) were quantified from OCT images using the Notal OCT Analyzer (NOA) software. Fluid measurements at the date of first injection and at six months were inputs to a 10-fold cross-validated ensemble machine learning model that predicted the total number of injections received in the first three years of treatment.

Results : Predictions were made for 2309 eyes from 1949 patients. Mean baseline age was 79 and 33% of eyes belonged to males. A total of 27,341 injections were received during the observation period. The machine learning model accurately predicted number of injections in the first three years of treatment. Predicted numbers of injections (median=11, inter-quartile range-IQR:9,15) were similar to the numbers observed (median=11, IQR:6,16). Correlation between predicted and observed counts was high (r=0.95, P<0.001) and mean absolute deviation was 2.4. In clinical terms, predictions for 73% of eyes (1742) were within three injections of the observed count.

Conclusions : We conclude that the number of anti-VEGF injections required during the first three years of treatment for neovascular AMD can be predicted using retinal fluid measurements derived from OCT images taken during the first six months of treatment. Machine learning models have the potential to provide clinically relevant risk stratification among AMD patients for resource planning. Further modelling is required to determine the patterns of fluid changes that drive these predictions.

This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.

×
×

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.

×