Development of a Computerized and Automated Cost-Effectiveness Analysis Application to Aid in Glaucoma Surgical Management.

Abstract

Precise: In this article, we describe the development of validated software that automates cost-effectiveness analyses of minimally invasive glaucoma surgeries based on modifications to robust mathematical models of glaucoma progression and management discussed in previous literature.

Purpose: To develop a validated application to streamline use of cost-effectiveness (CE) in clinical management and investigations of minimally invasive glaucoma surgeries (MIGS) in the United States.

Design: Automated cost-utility analysis adapted from Markov models described in prior literature.

Participants: Patients of ages 65 years and greater with mild to moderate primary open-angle glaucoma irrespective of concurrent visually significant cataract.

Methods: Markov models simulating glaucoma progression through 37 states and death were constructed based on previous CE models of minimally invasive trabecular meshwork stents. These states represent combinations of various glaucoma severity (mild, moderate, advanced, and severe/blind) with differences in clinical management, including use of up to 4 medications, selective laser trabeculoplasty, or incisional surgery. These are not mutually exclusive and are based on decisions related to the rate of thinning of the retinal nerve fiber layer. Rather than using fixed sets of transition probabilities for specific surgical interventions, new transition probabilities are dynamically derived based on the expected reduction in intraocular pressure related to visual field mean deviation decline. In addition to the generic MIGS arm, two comparison arms (cataract-surgery or medications-only) are included. Medication reduction, whole costs, and utilities are modifiable inputs in the model. Optimal and worst-case results are determined by uncomplicated or complicated (secondary surgical intervention required/medication nonadherence) intervention outcomes. The model was entirely re-implemented in R and validated by comparing results to TreeAge data.

Main outcome measures: Total costs, quality adjusted life years (QALY), and incremental cost-effectiveness ratio (ICER).

Results: An optimal-case 35-year CE-analysis of the implantation of Hydrus and iStent inject devices provided values of costs and QALY that were similar to prior data (R vs. TreeAge): Hydrus (Cost: $50446.53 vs. $48026.13; QALY: 12.18 vs. 12.26), iStent inject (Cost: $52323.43 vs. $49599.86; QALY: 12.13 vs. 12.21), cataract (Cost: $54150.56 vs. $54409.25; QALY: 12.03 vs. 12.04). Trends of ICER over time were also very similar.

Conclusions: Novel software is available to aid in CE analyses of MIGS with modifiable inputs and outcomes of interest. Such a tool makes CE more accessible for use in clinical management decisions and may guide future investigation.