Tornado diagram depicts the range of cost-effectiveness results around the $35 663 per quality-adjusted life-year baseline incremental cost-effectiveness ratio (ICER) (in 2016 US dollars) that result from changing each listed factor along its distribution or assumed range. Three alternative assumption scenarios indicate potential cost savings. VEGF indicates vascular endothelial growth factor.
eAppendix. Economic Analysis of Home AMD Monitoring
eTable 1. Annual Risk of CNV, Calculated From AREDS Report 18
eTable 2. Estimated Current Population by AMD Stage
eTable 3. Visual Acuity Assignment Parameters
eTable 4. Anti-VEGF Treatment Efficacy Parameters
eTable 5. Examination Rate Parameters
eTable 6. Program Cost Parameters
eTable 7. Program Compliance Parameters
eTable 8. Medical Cost Parameters
eTable 9. Annual Cost per Person of Blindness, Aged 65 and Older
eTable 10. Annual Cost per Person of Visual Impairment, Aged 65 and Older
eTable 11. Annual Productivity Losses From Low Vision, Aged 65 and Older
eTable 12. Utility Values Based on Acuity in the Better-Seeing Eye
eFigure. Annual AMD State Transition Diagram (Reverse Transitions Not Shown)
Customize your JAMA Network experience by selecting one or more topics from the list below.
Wittenborn JS, Clemons T, Regillo C, Rayess N, Liffmann Kruger D, Rein D. Economic Evaluation of a Home-Based Age-Related Macular Degeneration Monitoring System. JAMA Ophthalmol. 2017;135(5):452–459. doi:10.1001/jamaophthalmol.2017.0255
What are some economic considerations of home telemonitoring of age-related macular degeneration?
In this economic analysis using a simulation model, home telemonitoring was considered to be cost-effective in developed countries for patients at high risk for the neovascular form of age-related macular degeneration ($35 663 per quality-adjusted life-year gained). Home monitoring for age-related macular degeneration currently would cost society $907 and be cost saving for patients, incurring $1312 in government expenditures during 10 years.
This simluation model suggests that supplementing usual care with home telemonitoring for patients at high risk for developing neovascular age-related macular degeneration not only reduces risk of vision loss but also is cost-effective, although incurring net costs for Medicare.
Medicare recently approved coverage of home telemonitoring for early detection of incident choroidal neovascularization (CNV) among patients with age-related macular degeneration (AMD), but no economic evaluation has yet assessed its cost-effectiveness and budgetary impact.
To evaluate a home-based daily visual-field monitoring system using simulation methods and to apply the findings of the Home Monitoring of the Eye study to the US population at high risk for wet-form AMD.
Design, Setting, and Participants
In this economic analysis, an evaluation of the potential cost, cost-effectiveness, and government budgetary impact of adoption of a home-based daily visual-field monitoring system among eligible Medicare patients was performed. Effectiveness and visual outcomes data from the Age-Related Eye Disease Study 2 Home Monitoring of the Eye study, treatment data from the Wills Eye Hospital Treat & Extend study, and AMD progression data from the Age-Related Eye Disease Study 1 were used to simulate the long-term effects of telemonitoring patients with CNV in one eye or large drusen and/or pigment abnormalities in both eyes. Univariate and probabilistic sensitivity analysis and an alternative scenario using the Treat & Extend study control group outcomes were used to examine uncertainty in these data and assumptions.
Home telemonitoring of patients with AMD for early detection of CNV vs usual care.
Main Outcomes and Measures
Incremental cost-effectiveness ratio, net present value of lifetime societal costs, and 10-year nominal government expenditures.
Telemonitoring of patients with existing unilateral CNV or multiple bilateral risk factors for CNV (large drusen and retinal pigment abnormalities) incurs $907 (95% CI, −$6302 to $2809) in net lifetime societal costs, costs $1312 (95% CI, $222-$2848) per patient during 10 years from the federal government’s perspective, and results in an incremental cost-effectiveness ratio of $35 663 (95% CI, cost savings to $235 613) per quality-adjusted life-year gained.
Conclusions and Relevance
Home telemonitoring of patients with AMD who are at risk for CNV was cost-effective compared with scheduled examinations alone. Monitoring patients with existing CNV in one eye is cost saving, but monitoring is generally not cost-effective among patients with low risk of CNV, including those with no or few risk factors. With Medicare coverage, monitoring incurs budgetary expenditures for the government but is cost-saving for patients at high risk of AMD. Monitoring could be cost saving to society if monitoring reduced the frequency of scheduled examinations or led to a reduction of one or more injections of ranibizumab.
Telemonitoring systems are an emerging technology designed to provide low-cost solutions to increase access to quality health care.1 Home-based telemonitoring technologies in particular have the potential to increase monitoring rates and adherence for patients with diagnosed conditions. In recognition of the expanding role of telemedicine services, including telemonitoring, the 2014 Medicare physician fee schedule greatly expanded telemedicine coverage.2 Telemonitoring may be particularly well suited for application to visual health, such as age-related macular degeneration (AMD).
Age-related macular degeneration is the leading cause of blindness in the United States and currently affects more than 11 million Americans.3,4 Most patients with AMD have the dry form and usually experience slow or limited vision loss. However, an estimated 2.2 million patients have the wet form of AMD or choroidal neovascularization (CNV), which can cause metamorphopsia (the distortion of the central vision) and can rapidly progress to profound blindness.5,6 In the past decade, anti–vascular endothelial growth factor agents have been used to substantially slow the rate of vision loss among patients with CNV and even restore some vision among treated patients.7 However, early detection of CNV is vital to preserving vision.7,8
Choroidal neovascularization is diagnosed through scheduled dilated eye examinations or, more often, patient self-referral after visual symptoms.9,10 However, even patients who self-refer usually only do so after losing substantial vision. Among 6 community-based studies10-15 that reported mean or median acuity at the time of CNV diagnosis, all found acuity of 20/80 or worse, with some patients already blind before diagnosis. Clinical experts have argued that the current system of in-office examinations and self-referral is an inadequate method to ensure timely detection of CNV.16
The Home Monitoring of the Eye (HOME) Study was a National Eye Institute–initiated clinical trial of the ForeseeHome (Notal Vision Ltd) device deployed among a subgroup of the Age-Related Eye Disease Study 2 (AREDS2) study patients.17 ForeseeHome is a home-based daily visual-field monitoring system intended to allow patients with the dry form of AMD to detect minute changes in vision associated with the onset of CNV before this vision loss is apparent to the patient, triggering an alert to the prescribing physician to facilitate an emergency dilated eye examination. The HOME study found that supplementing standard care with this home telemonitoring system resulted in significantly better visual acuity at CNV diagnosis compared with standard care only.18 ForseeHome is currently the only device cleared by the US Food and Drug Administration for home monitoring for CNV, and Medicare approved coverage of the device for patients who met the eligibility criteria in December 2015.
The HOME study was designed to measure acuity outcomes at diagnosis but was not designed to measure the long-term outcomes and costs that may be important for consideration of the overall efficacy and cost-effectiveness of this technology. The objective of the current study was to conduct an economic evaluation of the ForeseeHome system using simulation methods and to apply the findings of the HOME study to the US population at high risk for wet-form AMD. We produced estimates of the long-term effects and outcomes of implementing the ForeseeHome system in the Medicare population, including program costs, visual outcomes, cost-effectiveness, net societal costs and a 10-year Congressional Budget Office–style nominal budgetary impact analysis.
Full details of our approach are included in the eAppendix in the Supplement (which includes eTables 1-12 and an eFigure). In brief, we developed a microsimulation model to assess the effects of ForeseeHome monitoring on long-term epidemiologic, economic, and budgetary outcomes. The model tracks 2 identical populations: one comprising individuals using standard care and the other comprising individuals supplementing standard care with the ForeseeHome system. Values for patients’ bilateral initial visual acuity, acuity at time of CNV diagnosis, CNV diagnosis rate, and false-positive rate are based on those observed in the HOME study. We assumed that all diagnosed CNV was treated with anti–vascular endothelial growth factor therapy. Other costs of AMD are based on Medicare claims costs or Medicare fee schedule reimbursement for procedures that occur at frequencies observed in other published studies17,19-21 or the HOME study. Costs of low vision and productivity losses are based on per-person costs attributable to blindness or visual impairment as reported in the Cost of Vision study. Quality-adjusted life-years (QALYs) were calculated using published utility values based on visual acuity in the better-seeing eye, adjusted by age-specific background utility levels. All data were deidentified, and no primary data were collected as part of this analysis. This analysis was deemed exempt because it was not human subjects research by the institutional review board at NORC at the University of Chicago.
Results are presented by initial AMD stage, using the AREDS simplified severity scale, where stages 0 to 4 represent early AMD defined by the cumulative number of risk factors (presence of large drusen or retinal pigment abnormalities) across both eyes, GA represents geographic atrophy, and CNV represents choroidal neovascularization or wet-form AMD. We denote patients with CNV in 1 eye as having CNV1 and with CNV in 2 eyes as having CNV2. Baseline results include patients with stage 3 AMD and higher (stage 3 AMD, stage 4 AMD, and CNV1), which corresponds with the patient population indicated for Medicare reimbursement. Major analysis variables, including examination rates, acuity distributions, treatment efficacy, AMD natural history, monitoring rates, and costs, are listed in Table 1.
We performed 3 primary analyses: (1) a cost-effectiveness analysis, (2) a societal net cost analysis, and (3) a Congressional Budget Office–style 10-year nominal federal budgetary impact analysis. The cost-effectiveness analysis was determined as incremental cost-effectiveness ratios (ICERs), equal to the incremental change in the current value of medical, monitoring, and low vision costs divided by incremental QALYs. The QALYs were calculated by assigning utility values based on acuity of the better-seeing eye or the presence of any monocular vision loss. Utility decrements were assigned to patients based on their expected background utility. The societal net cost analysis was calculated as the current lifetime net costs, including monitoring, medical, and low vision costs, and productivity losses. The 10-year budgetary impact analysis estimated the cumulative nominal budgetary cost per patient during the initial 10 years of Medicare coverage of home monitoring. All costs were adjusted to 2016 US dollars using the medical or overall Consumer Price Index. Future costs and QALYs in the cost-effectiveness and societal net cost analysis were discounted to the current year by 3% annually. Costs for the government budgetary analysis are reported in nominal terms during 10 years.
All factors in the model were sampled from their prior distributions for each of 1000 iterations. We report the central 95% of results as the 95% credible interval. We conducted a univariate sensitivity analysis by individually varying each factor over its respective 95% CI or a plausible range. We also included 3 alternative assumption scenarios that could potentially affect real-world outcomes if (1) monitoring resulted in 1 fewer scheduled examinations per year, (2) earlier detection through monitoring avoided a single injection of ranibizumab, or (3) home monitoring outcomes were compared with outcomes in a separate control group represented by baseline CNV diagnosis acuity observed in the Wills Eye Hospital Treat & Extend study.
Table 2 reports the expected per-person life-years lived with vision loss for standard care and home monitoring in the baseline population representing patients with stage 3 AMD, stage 4 AMD, and CNV1. Monitoring led to a 10.8% reduction in expected life-years lived with blindness, a 4.4% reduction in life-years with moderate impairment, and a 1.4% reduction in mild impairment, expressed per person eligible and qualified for monitoring.
Cost-effectiveness, net societal cost, and 10-year government budgetary impact per person in the baseline population (AMD stage 3, AMD stage 4, and CNV1) are reported in Table 3. We estimated that monitoring costs totaled $2645 per patient. Monitoring increased other medical costs by $158 to a lifetime total of $78 098 but reduced the lifetime costs of visual impairment and blindness by $1251 to $9225, leading to an increase in net program and medical costs of $1552 per patient over a lifetime. We estimated an increase in QALYs of 0.044 (from 4.317 to 4.361). The incremental cost ($1552) divided by the incremental QALYs (0.044) yields an ICER of $35 663 (95% CI, cost savings to $235 613). Home monitoring reduced lifetime productivity losses by $644 to $6193. Subtracting averted productivity losses from the incremental cost yields a net societal cost (cost benefit) of $907 (95% CI, −$6302 to $2809) per patient monitored. Compared with standard care, Medicare coverage of home monitoring is expected to increase net federal governmental payments by $1312 (95% CI, $222-$2848) per enrollee during the first 10 years of implementation.
Monitoring patients with only CNV1 resulted in cost savings because averted medical and low vision costs exceeded monitoring costs. However, government expenditures would total $851 during 10 years per patient with CNV1 enrolled because government accrues a greater share of costs than benefits from avoided vision loss than patients do, and many potential budgetary savings would not be achieved until after the initial 10 years. Monitoring patients with preadvanced AMD is relatively less cost-effective. Monitoring patients with stage 4 AMD results in an ICER of $73 799, whereas monitoring patients with stage 3 AMD results in an ICER of $127 584 per QALY gained.
The results of the univariate sensitivity analysis are shown in the Figure, which depicts the 8 factors that could cause at least a $1000 change in the ICER and 3 alternative assumption scenarios. Among the variation scenarios, results are most sensitive to an assumed ±50% range in the cost of monthly monitoring, followed by the 95% CIs of the costs of blindness. Results were comparatively insensitive to all other factors because variation in no other factors changed the ICER by more than 10%. Results are highly sensitive to the 3 hypothetical alternative assumption scenarios, where avoiding 1 ophthalmologic examination per year, avoiding a single injection of ranibizumab, and comparing home monitoring outcomes to usual care outcomes represented by the Wills Eye Hospital Treat & Extend study each resulted in cost savings for the baseline population.
Our study’s results suggest that home telemonitoring for patients with AMD at high risk for CNV2, including those with existing CNV1 or with risk factors for CNV2, is likely to be cost-effective relative to commonly cited thresholds for cost-effectiveness. Using conservative estimates of the potential benefits of the system, we estimated that home monitoring among these patients costs $35 663 per QALY gained, led to a lifetime net societal cost of $907 per patient, and would cost the federal government a total of $1312 per patient during the initial 10 years of Medicare coverage. These findings compare favorably to the oft-cited threshold of $50 000 per QALY.28
Cost-effectiveness is highly dependent on patients’ probability of progression to CNV2. Our results found poor cost-effectiveness for patients with early dry-form AMD, who are unlikely to quickly progress to CNV2. However, monitoring appears to be cost saving for patients with existing CNV1. Model results are largely insensitive to most variation; only 50% higher monthly monitoring costs increased the ICER to more than $50 000 per QALY.
This study is subject to several limitations. A major limitation of all model-based studies is the requirement to combine data from multiple sources and disparate populations because no single study fully captures the natural history, diagnosis, treatment, costs, and long-term outcomes of AMD. We based our analyses primarily on the results of the HOME,17,18 AREDS,22 Comparison of AMD Treatment Trials,7 and the Seven-Year Update of Macular Degeneration studies,8 but each is subject to certain limitations and includes a different population; thus, their integration into a single model may introduce bias.
The HOME study18 results demonstrated that monitoring patients at high risk for CNV using the ForeseeHome monitoring system results in lower levels of acuity loss at the time of CNV diagnosis. However, the HOME study did not conclusively demonstrate the full benefit of monitoring during a longer time horizon when presumably the lower diagnosis rates observed in the control group would lead to increasing levels of acuity loss over time. The HOME study also did not track costs and outcomes for patients after diagnosis. For this reason, basing results on the HOME study acuity outcomes may underestimate the long-term potential benefits of the program.
We test the potential effect of some of these limitations in the alternative scenario results, which are intended to capture plausible real-world effects that were not measured in the HOME study. For example, the 728 patients included in the device monitoring arm of the HOME study underwent nearly 2000 scheduled examinations in the 1.4-year study period, but only 14 of these visits resulted in the detection of CNV. Reducing the mean frequency of scheduled examinations from twice a year to once a year could potentially achieve cost savings, assuming visual outcomes were not substantially affected. Results are also sensitive to potential effects on CNV treatment intensity. If higher acuity values attributable to monitoring avoided even a single injection of ranibizumab or aflibercept, the intervention would be cost saving. Finally, results are even more sensitive to real-world observed acuity outcomes. The HOME study revealed unusually good acuity outcomes even in the control group, possibly because of the short 1.4-year duration and the high rates of scheduled ophthalmologic examinations. Community-based studies typically find far worse levels of acuity at the time of CNV diagnosis. For example, Acharya et al,29 Fong et al,10 Olsen et al,30 and the Wills Eye Hospital Treat & Extend study11 data reveal that many, if not most, patients who present with CNV are already moderately impaired or blind. When comparing monitoring outcomes with those in an alternative control group based on any of these studies, monitoring is substantially cost saving.
Finally, another major limitation of this analysis is that the cost-effectiveness results are largely dependent on patient utility measures used to calculate QALYs. Utility measures are subject to uncertainty, and their use is subject to controversy and misunderstanding.31 In particular, limited information exists on the utility loss of monocular impairment, which is far more common than bilateral vision loss in our model. Our model assigns QALY losses of 0.0396 for ages 65 to 74 years and 0.036 for ages 75 years and older from any level of monocular impairment or blindness.25,26,32 Because this is assigned as a single threshold, in the model the effect of monitoring patients who only ever progress to monocular vision loss is essentially limited to 1 or 2 years of avoiding this small QALY differential. Our sensitivity results reveal that results are largely insensitive to a 0% to 200% change in the monocular vision utility loss value. However, monitoring substantially reduces the incidence of monocular vision loss.
Although these limitations inhibit our ability to definitively state what the cost-effectiveness of the intervention will be when deployed in the real world, our results indicate that monitoring is likely to be comparatively cost-effective for patients at high risk of progression to bilateral CNV even under conservative assumptions. If patients will not receive high levels of care, such as those observed in the HOME study control group, schedule fewer examinations, or incur lower treatment intensity because of home monitoring, the intervention could potentially be cost saving for the indicated population and may be cost-effective for some lower-risk patients.
Before Medicare’s decision to provide coverage, use of ForeseeHome was driven by individual patient and family investment decisions to purchase the system based on their health preferences, perceptions of risk, and financial resources. With Medicare coverage, the decision on the part of patients becomes simpler; our results indicate that with 80% Medicare coverage of the device set to begin in 2016, patients and their families may expect to accrue lifetime savings of $586. However, although patients may expect to achieve positive life-term benefits and savings, the program would remain a net cost to government, costing $1312 per enrollee during the initial 10 years of coverage.
Given the large population at risk for CNV and the increased importance of timely CNV diagnosis that has arisen because of the recent adoption of anti–vascular endothelial growth factor therapy for wet-form AMD, the effects of this investment decision are great. In addition, these stakes will only increase in the future as the large increase in the population at risk for AMD in the coming years will rapidly increase the population at risk for blindness from wet-form AMD. The population 90 years and older has the highest prevalence of AMD and is projected to increase 4-fold by 2050.4,33 Likewise, the economic burden of vision loss and eye disorders is projected to increase from $139 billion in 2013 to more than $700 billion by 2050 in nominal dollars, with most of these costs attributable to the indirect costs of vision loss.24,34 Faced with this explosive growth in the burden and cost of vision disorders, including AMD, the cost of not preventing AMD-related vision loss may become staggering. Thus, although Medicare coverage of ForeseeHome remains a near-term cost, it could potentially serve as one avenue of investment to mitigate some of the potentially drastic increases in the future costs of vision loss from AMD.
Corresponding Author: John S. Wittenborn, NORC at the University of Chicago, 55 E Monroe St, 30th Floor, Chicago, IL 60603 (firstname.lastname@example.org).
Accepted for Publication: February 3, 2017.
Published Online: March 30, 2017. doi:10.1001/jamaophthalmol.2017.0255
Author Contributions: Mr Wittenborn and Dr Rein had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Wittenborn, Clemons, Regillo, Rein.
Acquisition, analysis, or interpretation of data: Wittenborn, Clemons, Rayess, Liffmann Kruger, Rein.
Drafting of the manuscript: Wittenborn, Regillo, Liffmann Kruger, Rein.
Critical revision of the manuscript for important intellectual content: Clemons, Regillo, Rayess, Rein.
Statistical analysis: Wittenborn, Rein.
Obtained funding: Wittenborn, Rein.
Administrative, technical, or material support: Wittenborn, Clemons, Liffmann Kruger, Rein.
Study supervision: Regillo, Rayess, Rein.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This study was funded by an unrestricted grant from Notal Vision LLC, Tel Aviv, Israel.
Role of the Funder/Sponsor: The funding source had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and the decision to submit the manuscript for publication.
Additional Contributions: Data were provided by The Emmes Corporation, which conducted the Home Monitoring of the Eye (HOME) Study, and the Mid Atlantic Retina and Wills Eye Hospital Retina Service, which conducted the Treat & Extend Study. All other data were collected from publicly available sources.