Abstract
Purpose :
The rising prevalence of neovascular age-related macular degeneration (nAMD) and diabetic macular edema (DME) is expected to burden patients and health care in the United States. Patients with nAMD or DME in the real world receive fewer injections and experience poorer visual outcomes than patients in clinical trials. This analysis clustered patients treated similarly for nAMD and DME and described their characteristics.
Methods :
An observational study was conducted using MarketScan claims databases of patients with incident unilateral nAMD or DME treated with ≥1 anti–vascular endothelial growth factor (VEGF) intravitreal injection from January 1, 2017– July 1, 2018 and followed for 2 years. To group patients with similar treatment patterns, machine learning algorithms clustered patients by treatment characteristics including the number and time between anti-VEGF injections and monitoring visits; corticosteroid usage was included for DME only. For each identified cluster, demographics and baseline patient characteristics such as insurance type were described.
Results :
Three clusters were identified each for nAMD and DME. For nAMD, over 2 years, Clusters 1 (n=1140), 2 (n=604), and 3 (n=294) had mean anti-VEGF injection frequencies of 11.2, 8.8, and 2.1, respectively, and gaps of ≥6 months in treatment/monitoring in 33%, 68%, and 93% of patients, respectively. Cluster 1 had 84% Medicare patients, Cluster 2 had 62% Medicaid patients, and Cluster 3 had 45% Medicare/46% Medicaid patients (p<0.001). Clusters 1, 2, and 3 had 67%, 5%, and 33% Preferred Provider Organization (PPO) insurance, respectively (p<0.001).
For DME, over 2 years, Clusters 1 (n=79), 2 (n=101), and 3 (n=217) had mean anti-VEGF injection frequencies of 7.5, 6.3, and 2.1, respectively, and gaps of ≥12 months in treatment/monitoring in 5%, 11%, and 12% of patients, respectively. Cluster 1 had 70% commercial patients, Cluster 2 had 88% Medicaid patients, and Cluster 3 had 62% Medicaid patients (p<0.001). Clusters 1, 2, and 3 had 63%, 7%, and 21% PPO insurance, respectively (p<0.001).
Conclusions :
Cluster analysis identified three groups within each disease that were separated by treatment characteristics. Payer and insurance type differed by cluster and may be associated with treatment patterns. The impact of these differences in vision and anatomic outcomes remains to be assessed.
This abstract was presented at the 2023 ARVO Annual Meeting, held in New Orleans, LA, April 23-27, 2023.