VALUE IN HEALTH 20 (2017) 1034–1040

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/jval

Cost-Utility Analysis of Extending Public Health Insurance

Coverage to Include Diabetic Retinopathy Screening by
Optometrists
Sasha van Katwyk, MSc1, Ya-ping Jin, MD, PhD2,3, Graham E. Trope, MB, PhD, FRCSC3,
Yvonne Buys, MD, FRCSC3, Lisa Masucci, MSc4, Richard Wedge, MD6, John Flanagan, OD, PhD3,7,
Michael H. Brent, MD, FRCSC3, Sherif El-Defrawy, MD, PhD, FRCSC3, Hong Anh Tu, PhD5,
Kednapa Thavorn, MPharm, PhD1,8,9,*
1
Clinical Epidemiology Program, Ottawa Hospital Research Institute, The Ottawa Hospital, Ottawa, Ontario, Canada; 2Dalla Lana
School of Public Health, University of Toronto, Toronto, Ontario, Canada; 3Department of Ophthalmology & Vision Sciences,
University of Toronto, Toronto, Ontario, Canada; 4Institute of Health Policy, Management and Evaluation, University of Toronto,
Ontario, Canada; 5Health Quality Ontario, Toronto, Ontario, Canada; 6Health PEI, Charlottetown, Prince Edwards Island, Canada;
7
School of Optometry and Vision Science Program, University of California Berkeley, California, USA; 8School of Epidemiology, Public
Health and Preventive Medicine, University of Ottawa; Ottawa, Ontario, Canada; 9Institute of Clinical Evaluative Sciences (ICES
UOttawa), Ottawa, Ontario, Canada

AB STR A CT

Background: Diabetic retinopathy (DR) is one of the leading causes of higher costs ($9,908,543.32) and improved QALYs (156,862.44), over
vision loss and blindness in Canada. Eye examinations play an 30 years, resulting in an incremental cost-effectiveness ratio of
important role in early detection. However, DR screening by optom- $1668.43/QALY gained. Sensitivity analysis showed that the most
etrists is not always universally covered by public or private health influential determinants of the results were the cost of optometric
insurance plans. This study assessed whether expanding public screening and selected utility scores. At the commonly used threshold
health coverage to include diabetic eye examinations for retinopathy of $50,000/QALY, the probability that the new policy was cost-effective
by optometrists is cost-effective from the perspective of the health was 99.99%. Conclusions: Extending public health coverage to eye
care system. Methods: We conducted a cost-utility analysis of examinations by optometrists is cost-effective based on a commonly
extended coverage for diabetic eye examinations in Prince Edward used threshold of $50,000/QALY. Findings from this study can inform
Island to include examinations by optometrists, not currently publicly the decision to expand public-insured optometric services for patients
covered. We used a Markov chain to simulate disease burden based on with diabetes.
eye examination rates and DR progression over a 30-year time Keywords: cost-utility analysis, diabetic retinopathy, optometrist,
horizon. Results were presented as an incremental cost per quality- publicly funded eye examination.
adjusted life year (QALY) gained. A series of one-way and probabilistic
sensitivity analyses were performed. Results: Extending public health Copyright & 2017, International Society for Pharmacoeconomics and
coverage to eye examinations by optometrists was associated with Outcomes Research (ISPOR). Published by Elsevier Inc.

proliferation of new blood vessels can be managed with a

Introduction
combination of laser photocoagulation, intravitreal injections of
Diabetic retinopathy (DR) is one of the leading causes of vision steroids, antivascular endothelial growth factor therapy, and/or
loss and blindness [1]. In Canada, approximately 14% of patients vitrectomy [3]. Early detection through eye examination allows
with diabetes (500,000) have some form of DR, and this preva- for timely treatments, which can significantly prevent or delay
lence is expected to rise in conjunction with the increasing vision loss [4].
incidence of diabetes [2]. The treatment for DR depends on the Several clinical practice guidelines highlight the role of pri-
disease state. Background retinopathy, a condition whereby the mary care physicians in detecting DR and facilitating appropriate
eye’s blood vessels are slightly swollen, can be managed with referral to an ophthalmologist [5,6]; however, primary care
strict control of blood sugar levels and careful regulation of blood physicians often have limited access to specialized instruments
pressure and renal function, and more severe vision loss from and training to perform the required testing to accurately

* Address correspondence to: Kednapa Thavorn, Ottawa Hospital Research Institute, The Ottawa Hospital, 501 Smyth Road, PO Box201B,
Ottawa, Ontario, K1H 8L6 Canada.
E-mail: [email protected]
1098-3015$36.00 – see front matter Copyright & 2017, International Society for Pharmacoeconomics and Outcomes Research (ISPOR).
Published by Elsevier Inc.
http://dx.doi.org/10.1016/j.jval.2017.04.015
 VALUE IN HEALTH 20 (2017) 1034–1040 1035

diagnose DR. Despite this, a large proportion of DR screenings are Study Population
performed by primary care physicians [7]. Optometrists have
The eligible population was PEI residents Z45 years of age who
better training in eye care diagnosis and have the necessary
had diabetes. The 45-year-old cutoff was used because it was the
equipment to diagnose eye-related conditions compared with
mean age of patients with diabetes in PEI. Based on PEI’s
primary care physicians. Although lack of government-insured
Statistics Bureau [11], a cohort with 7958 patients with diabetes
optometric services for patients with diabetes was found to
were 444 years of age. Of these 3514 (44.2%) were between 45 and
negatively impact patients’ access to health care services and
65 years of age. We used data from the same source to project the
vision health outcomes [8], optometric services are not covered in
future population in PEI. We categorized our target population to
some provincial health insurance plans in Canada, including
two age groups (45–64 years and 65þ years) to account for
those of New Brunswick, Prince Edward Island (PEI), and New-
observed differences in the mortality rates, DR prevalence rates,
foundland and Labrador.
and screening rates. Given the projected increase in the older age
Health economic evaluation is a systematic approach that can
cohort in PEI, it was important to model changes in disease
inform policymakers whether the added benefits of expanding
burden and related costs over time. Therefore, each year we
publicly funded optometric services justify their costs. To date,
added a new cohort of diabetes cases to the Markov model
no study has assessed the cost-effectiveness of publicly funded
representing the newly 45-year-old population of PEI. The newly
optometric services. Existing studies have shown that a system-
added cases were distributed across the stages of DR according to
atic DR screening program is cost-effective compared with
the prevalence rate at 45 years of age [11].
opportunistic screening, as in North America and Europe [9]
and that DR screening is a cost-effective addition to a national
diabetes treatment policy [10]. Decision Analytic Model
The objective of our study was to assess whether expanding We used a probabilistic decision analytic model to simulate the
public health insurance coverage to include eye examinations for long-term costs and outcomes for a cohort of PEI residents living
patients with diabetes by optometrists is cost-effective from the with diabetes. We simulated the natural history of DR among the
perspective of the health care system. population with diabetes according to the most recent prevalence
and transition probability data by using a Markov chain. The
stages of DR used in the Markov chain were as follows: 1) no
diabetic retinopathy (NDR); 2) background diabetic retinopathy
(BDR); 3) pre-proliferative diabetic retinopathy (PPDR); 4) prolifer-
Methods
ative diabetic retinopathy (PDR); and 5) end-stage disease—loss of
useful vision [12]. Each year individuals may transit through one
Overview of the five disease states or death [6], according to annual
We conducted a cost-utility analysis of a publicly insured optom- transition probabilities.
etric service for patients with diabetes in PEI from the perspective The Markov model estimated the costs and effects over a
of the health care system. The intervention of interest was a new period of 30 years, since that was the duration of time for which
optometric policy, whereby eye examinations for DR by optom- we had population projection data that could be reliably trans-
etrists are publicly insured. The comparator was usual care, lated into rates of diabetes within our age cohorts. The long
whereby eye examinations for patients with diabetes performed follow-up period also allowed us to capture changes in patients’
by primary care physicians or referrals to ophthalmologists are demographics and the incidence rates of diabetes over time. Our
publicly insured. We assumed that patients in the usual care patient cohort was expanded to capture new cases of diabetes for
group were not screened for DR by optometrists at baseline and each year. We used PEI Statistics Bureau’s population projection
that an extended insurance coverage would increase annual DR data to simulate population growth from 2014 to 2044, as the
screening rates. expected change in annual incidence of diabetes [11].
The diagnostic accuracy of DR screening modalities and the Each year, probability rate of individuals with diabetes being
effectiveness of treatment alter the likelihood that patients with screened was equal to the annual screening rates. Recommended
diabetes progress through the stages of DR to ultimate vision loss. regular screening for DR is once every 2 years. Individuals who
This means that changes in the screening rate and screening did not receive any screening would transit through the Markov
accuracy were the main drivers of difference in DR rates between chain according to the baseline transition probabilities. For our
the usual care (government-insured DR by primary care physi- baseline model, we assumed that the 53% of PEI residents who
cians plus ophthalmologists) and the intervention scenarios received DR screening regularly continued going to primary care
(government-insured DR screening by optometrists and primary physicians and ophthalmologists over the 2-year period and that
care physicians plus ophthalmologists). the expanded coverage to optometrists would result in 35%
increase in screening. Of those screened in the usual care group,
we assumed that 80% of DR screening was performed by general
practitioners, who then potentially referred patients to ophthal-
Setting mologists on the basis of test findings. We varied this share of
PEI is an island province in eastern Canada with a total popula- screenings in a scenario analysis that ranged general practitioner
tion of 146,447 (as of 2015), of whom approximately 70,000 are share of screenings from 50% to 100%. For those who were
over the age of 44 years. In 2015, prevalence of diabetes among screened, their risk of being diagnosed with DR was estimated
those 444 years of age was just over 11% [11]. Canadian according to the incidence of DR and the diagnostic accuracy of
provincial governments bear the primary cost burden for provid- the DR screening by an optometrist or an ophthalmologist. The
ing health care to their residents and are therefore the primary sensitivity and specificity of DR screening was based on a
arbiters for determining what will be publicly covered. Although systematic review that included two observational studies in
physician visits (including eye examinations), as well as oph- which the true severity of DR was known and several screening
thalmologic DR screening and most treatment, are covered by the approaches and practitioners were provided with the same
PEI provincial health plan, optometric services are not included in patient [4]. We assumed that individuals who were diagnosed
the plan. with DR (whether true or false positive) received a treatment.
1036 VALUE IN HEALTH 20 (2017) 1034–1040

Fig. 1 – Decision tree model and imbedded Markov chain.

Input Parameters distribution of optometrists in PEI, we estimated that

government-insured DR screening by optometrists would
Transition probabilities increase the screening rate by an additional 35%, resulting in a
We derived transition probabilities from a study by Liu et al. [13],
screening rate of 88% over 2 years [8]. To test the impact of the
which described the probabilities that patients with diabetes
2-year eye examination rate, we performed a scenario analysis
move across the DR pathway from proliferative retinopathy to
that varied the change in screening rate from 10% (5% per year) to
blindness. Mortality rates were derived by applying a relative risk
45% (22.5% per year).
of death resulting from diabetes reported in a 2013 systematic
review [14] to all-cause mortality in the general Canadian pop-
ulation. We used the all-cause annual mortality rates for each age Treatment efficacy and health utility values
group (45–64 and 65þ years) reported by Statistics Canada, which Treatment efficacy was based on a 2004 multisite randomized
were 0.5% and 8.2%, respectively (sex adjusted) [15]. To properly control trial that evaluated clinical efficacy of minor and
account for the effect of age on the incidence of DR, a portion of advanced treatments [18]. Patient cohorts were then placed in
the 45–64 years cohort could move into the 65þ years cohort the Markov chain according to transition probabilities, test
during every reiteration of the Markov chain. The transition accuracy, and treatment effectiveness within each respective
probabilities between two age groups (45–64 and 65þ years) were outcome branch (Fig. 1). Health utility data were taken from a
derived by calculating the number of individuals Z65 years of 2012 utility survey that included two catchment surveys of
age obtained from the PEI Statistics Bureau (accounting for patients with diabetes across all age groups and levels of DR.
death and migration in the same method as calculated for each Using both a health utility index (HUI-3) metric and time trade-off
age group). exercise, the survey calculated the relative utility score for each
level of DR adjusted for all relevant patient factors [19].

Eye examination rate Costs

According to Statistics Canada, in 2005, 53% of PEI residents Z45 All costs were taken from the PEI Schedule of Benefits (Master
years of age who were living with diabetes received an eye Agreement) [20]. Screening costs for the intervention constituted
examination within the previous 2 years [16]. This percentage a consultation, which included a DR examination by an optom-
captured all eye examinations and did not differentiate between etrist. Since there is no standard remuneration for optometrist
public and private payers or consider who administered the eye consultations, we used the remuneration of a general practitioner
examination. Nonetheless, we utilized this as a baseline exami- consultation as a base case. We varied the price in our sensitivity
nation rate, since it was the best available data for PEI. We analysis according to the highest (Manitoba: $89.35) and the
obtained the impact of publicly insured optometric services on lowest (Nova Scotia: $52.00) consultation remuneration found in
the examination rate from a published Canadian study that had all Canadian provinces other than PEI as of 2015. The comparator
evaluated the effect of delisting routine eye examinations on (current) screening protocol consisted of a consultation by a
retinopathy screening among patients with diabetes in Ontario primary care physician followed by referral to an ophthalmolo-
[17]. The Ontario study reported that delisting routine eye gist and included the fees of a consultation for a DR examination.
examinations decreased the eye examination rate by 8.7% among Costs of treatment were estimated according to disease severity;
patients 40 to 65 years of age who were living with diabetes, background DR was assumed to receive an intensification of
although this age group was not the intended delisting target. existing diabetes treatment (e.g., increased insulin schedule), and
Based on the results of this Ontario study, the ratio of optomet- more severe DR was treated with laser photocoagulation.
rists to ophthalmologists in PEI (32:6), and the geographic Although multiple treatments are possible for severe DR,
 VALUE IN HEALTH 20 (2017) 1034–1040 1037

Table 1 – Input parameters for the model.

Type / Practitioner Parameter Base Range or 95% Distribution Source
value confidence
interval (CI)

Transition probabilities
Prevalence BDR 10.18% (9.7–10.6) Normal PEI administrative data, 2012
PPDR 1.20% (1.1–1.6) Normal PEI administrative data, 2012
PDR 1.70% (1.4–2.0) Normal PEI administrative data, 2012
Blind 0.25% (0.2–0.3) Normal PEI administrative data, 2012
Probability NDR to BDR 0.09 (0.08–0.11) Beta Liu et al. (2003)
BDR to PPDR 0.12 (0.11–0.13) Beta Liu et al. (2003)
PPDR to PDR 0.60 (0.49–0.72) Beta Liu et al. (2003)
PDR to Blind 0.46 (0.39–0.53) Beta Liu et al. (2003)
Death (45-65) 0.01 (0.008–0.015) Normal Liu et al. (2003)
Death (65þ) 0.16 (0.14–0.25) Normal Liu et al. (2003)
General Practitioner Sensitivity (%) 46 (35–56) Normal Hutchinson et al. (2000)
Specificity (%) 81 (72–91) Normal Hutchinson et al. (2000)
Ophthalmologist Sensitivity (%) 82 (80–84) Normal Hutchinson et al. (2000)
Specificity (%) 95 (94–96) Normal Hutchinson et al. (2000)
Optometrist Sensitivity (%) 74 (67–81) Normal Hutchinson et al. (2000)
Specificity (%) 84 (73–96) Normal Hutchinson et al. (2000)
Costs
General Practitioner Consultation (fee per visit) $62.00 Fixed Health PEI (2015)
Ophthalmologist Consultation (fee per visit) $87.00 Fixed Health PEI (2015)
Ophthalmologist DR Exam $11.50 Fixed Health PEI (2015)
Treatment Photocoagulation $275.20 Fixed Health PEI (2015)
Optometrist Consultation (fee per visit) $62.00 Fixed Health PEI (2015)
Optometrist DR Exam $11.50 Fixed Health PEI (2015)
Utilities NDR 0.88 (0.82–0.95) Beta Heintz et al. (2012)
BR 0.86 (0.79–0.93) Beta Heintz et al. (2012)
PPDR 0.87 (0.78–0.97) Beta Heintz et al. (2012)
PDR 0.83 (0.72–0.93) Beta Heintz et al. (2012)
Blind 0.48 (0.34–0.61) Beta Heintz et al. (2012)
Dead 0 Fixed
BDR, background diabetic retinopathy; PPDR, pre-proliferative diabetic retinopathy; PDR, proliferative diabetic retinopathy; PEI, Prince Edward
Island.

photocoagulation is the most frequently used and requires fewer Canadian dollars. In accordance with Canadian economic guide-
costing assumptions regarding specific clinical patient character- lines [21], an annual discount rate of 5% was applied for both
istics [3]. costs and outcome. An incremental cost-effectiveness ratio (ICER)
All key input parameters, including transition probabilities, was expressed as an incremental cost per QALY gained Table 2–4.
treatment efficacy, and health utility values, are shown in Table 1. All analyses were performed in Microsoft Excel version 2012
Table 2–4.

Analysis
We calculated the aggregated costs, the number of screening Sensitivity analysis
tests, and quality-adjusted life years (QALYs) of patients with A series of one-way sensitivity analysis was performed on
diabetes over the 30-year period. Costs were expressed in 2015 the following parameters: 1) cost of optometrist consultation;

Table 2 – Results (167,514 screenings over a 30-year Table 3 – Cost-effectiveness results by

period). discount rate.
Baseline Intervention Incremental Incremental ICER
cost QALY
Expected cost (CAD) $7,677,187.94 $9,908,543.32
Expected QALY 155,525.05 156,862.44 0% $ 4,592,523.41 3,021.96 $1,519.72
Incremental cost (CAD) $2,231,355.39 3% $ 2,905,086.84 1,816.83 $1,598.99
Incremental QALY 1,337.39 5% $ 2,231,355.39 1,337.39 $1,668.43
ICER ($/QALY) $1,668.43 10% $ 1,311,767.45 689.91 $1,901.37
CAD, Canadian dollars; ICER, incremental cost-effectiveness ratio; ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted
QALY, quality-adjusted life year. life year.
1038 VALUE IN HEALTH 20 (2017) 1034–1040

Table 4 – Cost-effectiveness results by time triple from 1% in 2015 to 3% in 2030. PDR would rise from 2% to
horizons. 4% while blindness was expected to be prevalent among 4% of
diabetics. Overall, our model predicted that the DR prevalence
Incremental Incremental ICER among diabetics over the age of 44 would increase from 14% to
cost QALY 30% by 2030.
1 year $ 93,684.30 4.95 $18,927.00
3 years $ 279,262.52 26.87 $10,394.02 Sensitivity Analysis
5 years $ 458,556.27 73.45 $6,242.93
The result of the one-way sensitivity analysis is presented as a
10 years $ 881,462.86 323.74 $2,722.74
tornado plot (Fig. 2). The cost of an optometrist examination,
20 years $ 2,231,355.39 1337.39 $1,668.43
utility values, and the transition probabilities from NDR to BDR
ICER, incremental cost-effectiveness ratio; QALY, quality-adjusted were the top three most influential factors. The probability that
life year. transition of patients with diabetes between NDR and BDR also
had substantial impact on the cost-effectiveness findings.
Figure 3 shows the results of the probabilistic sensitivity
analysis. All iterations fell within the northeast quadrant, indi-
2) screening sensitivity; 3) screening specificity; 4) utility values;
cating that extending government-insured screening for DR by
5) treatment effectiveness; 6) DR prevalence; and [7–10] transition
optometrists was more expensive but also delivered better health
probabilities for each disease state. A probabilistic sensitivity
outcomes compared with PEI’s current coverage policy. Com-
analysis was also performed on all parameters according to the
pared with a common threshold of $50,000/QALY in Canada,
reported 95% confidence interval and reported distributions or, in
extending optometric services for DR screening was considered a
the case of optometrist screening rate, based on an estimated
cost-effective option. Similarly, Figure 4 shows the cost-
range and distribution following a review of the literature. The
effectiveness acceptability curve and suggests that the probabil-
sensitivity analysis was performed by using the Monte Carlo
ity of the new PEI optometrist policy is cost-effective increased
method with 20,000 iterations. The results of the probabilistic
with a greater value of willingness to pay. At the common
sensitivity analysis were used to create a cost-effectiveness
threshold of $50,000/QALY, the probability that the new optom-
acceptability curve, which shows the probability of the interven-
etric policy is cost-effective is 99.99%.
tion being cost-effective over a range of willingness to pay
thresholds.
Scenario Analysis
The ICER results depended on the assumed share of the cost of
Results screenings performed by general practitioners, which ranged
Extending publicly insured eye examination for patients with from $1303.47 at the lower end (50%) to $2023.73 at the upper
diabetes to include examinations by optometrists was associated end (100%). The larger the proportion of screenings performed by
with higher health system cost ($9,908,543.32 vs $7,677,187.94 general practitioners, the smaller was the improvement in
over 30 years) and improve health outcomes (1337.39 QALYs patients’ QALYs. There is also a small inverse relationship
gained over 30 years), resulting in the incremental cost- between the share of the cost of screenings performed by general
effectiveness ratio of $1668.43/QALY gained. Extending publicly practitioner and the total costs in the comparator case. The
insured services to include optometrists was estimated to reason is that with fewer DR cases being identified, there are
increase the number of screenings by 2221 per year. slightly fewer patients who receive treatment.
Over the 30-year period, we found that while patient distri- The cost-effectiveness results also varied according to the
bution across health states remained roughly the same in the expected effect of public reimbursement of optometrist services
first ten years, significant increases in projected diabetes preva- on the annual screening rate. The ICER was $430.45, assuming
lence and an absolute rise in diabetes among those over the age only a 5% increase in screening rates. Although the increased
of 65 led to higher proportions of severe DR in the population. We screening rates raised the costs of treatment, they also improved
projected that the prevalence rate of background DR would nearly
double to 19% of diabetics over the age of 44, while PPDR would

Fig. 2 – Tornado plot representing select results of one-way

sensitivity analysis, incremental cost-effectiveness ratio Fig. 3 – Probabilistic sensitivity analysis, average annual cost
(ICER). and quality-adjusted life year (QALY).
 VALUE IN HEALTH 20 (2017) 1034–1040 1039

cohort to the 65 as they aged, using the expected changes in

population size and the prevalence of diabetes in each age group.
Our projected prevalence and mortality rates for diabetes closely
match those reported by Statistics Canada combined with the
Canadian Diabetes Cost Model prevalence projections [15,24,25]
(Fig. 5). Although we would like to extend our analysis timeframe
to 45 years to represent the lifetime of all patients, Statistic
Canada only provides 30 years’ projection data. We consider this
a small limitation of our study design, especially since it is an
open cohort design, in which new younger individuals are
continuously being added and there is no set lifetime duration
that we could cover.
Most added costs were accrued from additional treatments
performed by ophthalmologists after severe DR was detected.
Fig. 4 – Cost-effectiveness acceptability curve. Although these are added costs to the health care system,
researchers have argued that these costs are likely already
expressed in PEI’s health system through other care costs linked
health outcome (QALYs). The proportional change in cost relative to vision loss [26]. For instance, although this intervention will
to change in QALY varied over time as the age distribution of bring about higher rates of laser photocoagulation, intravitreal
patients changed; however, these ratios were within a $2000/ injections and vitrectomy procedures in the short term, a sig-
QALY range. If the effect of expanding public insurance on nificant share of such procedures would have eventually
screening rates were to reach 100%, the ICER was still lower than occurred at a later point in the patient’s lifetime when the patient
$6000/QALY. experiences vision loss at a more severe stage.
This health economic evaluation utilized a decision analytic
model to synthesize data from various sources, leading to some
limitations that merit discussion. First, our study population was
Discussion
Z45 years of age—that is, we could have missed the small
Our study shows that extending government-insured eye exami- portion of DR cases in younger patients and therefore could have
nations of patients with diabetes to optometrists is a cost- underestimated the cost of screening. This matters, since most
effective option from the perspective of the health care system. clinical guidelines recommend that regular screening start 3 to 5
Extending public coverage for optometric screening of DR will add years after diagnosis of type I DM, which is the most prevalent
additional cost to the health care system in the expected amount type of diabetes among those o45 years of age. Second, we
of $74,378.51 per year over 30 years; however, improving access to assumed that the costs of intensified blood glucose and blood
screening by optometrists will significantly delay the progression pressure treatments were comparable in the intervention and
of DR, resulting in an average annual gain of 44.58 QALYs for usual care groups and therefore did not include them in our
patients with diabetes over the same period. analysis. Given the limited variation in the study results because
An extension of publicly funded services to include optomet- of the changes in costs shown in our sensitivity analyses, the
rists in Canada would ease the pressure on ophthalmologists’ inclusion of these drug costs could have had a negligible impact
time and resources and lower the incidence of primary care on our findings. Third, health utility values were based on a
physicians’ performing incomplete screenings that do not adhere single US study providing utility values that matched our cate-
to established guidelines [22]. In the case of PEI, optometric gorization of DR stages. It should be noted that the sociodemo-
services potentially offer better access to care, particularly in graphic and clinical characteristics of the participants in our
rural areas, as all ophthalmologists in PEI live in one of two urban study might have been different from those of our target
centers [23]. population. Although the structural uncertainty introduced by
Our sensitivity analysis showed that the highest level of our reliance on a single source would be best characterized by
model uncertainty is in the utility value of blindness. The reason using health utility values from alternative sources, such evi-
may be that it was estimated from a study with a small sample dence is currently unavailable. Further studies should measure
size [13]. Also high in variance was the cost of optometric
services, which is to be expected, since the cost of services
account for 23% of total cost of the new policy. Interestingly, we
found that the relative share of screening performed by optom-
etrists and ophthalmologists had only a moderate impact on the
cost-effectiveness results. This finding may be, in part, attributed
to the closely comparable screening accuracy of both practi-
tioners. The probabilistic sensitivity analysis showed that the
probability of the intervention being cost-effective increased with
a greater value of willingness to pay. This probability was 499%
after the willingness to pay value was 4$6000/QALY.
Projecting outcomes for 30 years presents some challenges
when accounting for changes in patient demographics and the
prevalence of diabetes. We utilized basic projections from Sta-
tistics Canada and the PEI Bureau of Statistics with regard to
population trends according to age as well as diabetes prevalence
rates by population group. We also attempted to better represent Fig. 5 – Diabetes prevalence among Prince Edward Island
the changes in the prevalence of diabetes as it pertains to DR by (PEI) residents with diabetes 45 years of age or older,
calibrating a Markov model to introduce new patients aged 45 at projected national average compared to our projected
the start of the annual cycle and transition patients from the 45 prevalence in PEI.
1040 VALUE IN HEALTH 20 (2017) 1034–1040

the health utilities of Canadian patients with DR by disease guidelines for the prevention and management of diabetes in Canada.
stages. Finally, our transition probabilities were derived from a Can J Diabetes 2013;37:S1–212.
6 General practice management of type 2 diabetes—2014–15. Melbourne,
Taiwanese study, which had different patient characteristics and
Australia: The Royal Australian College of General Practitioners and
a different health care system affecting disease progression. Diabetes Australia, 2014.
However, this Taiwanese study was the most recent study and 7 Bragge P, Gruen RL, Chau M, et al. Screening for presence or
based on the same type of disease stage progression as that in absence of diabetic retinopathy: a meta-analysis. Arch Ophthalmol
2011;129:435–44.
our study. Moreover, our sensitivity analyses showed that these
8 Jin YP, Wedge R, El-Defrawy S, et al. Lack of government-funded
transition probabilities had minimal impact on the study find- optometric services is associated with reduced utilization of eye care
ings. Future studies should take advantage of the availability of providers and increased utilization of family doctors. CAHSPR
routinely collected administrative databases and derive long- Conference, 2014.
term progression of DR among patients with diabetes in devel- 9 Jones S, Edwards R. Diabetic retinopathy screening: a systematic review
of the economic evidence. Diabet Med 2009;27:249–56.
oped countries. 10 Li R, Zhang P, Barker LE, et al. Cost-effectiveness of interventions to
prevent and control diabetes mellitus: a systematic review. Diabetes
Care. 2010;33:1872–94.
Conclusions 11 Prince Edward Island: population projections 2014–2053. Charlottetown,
PEI: Department of Finance and Municipal Affairs, 2014.
Our study suggests that extending public health coverage to 12 Kohner E. The evolution and natural history of diabetic retinopathy. Int
Ophthalmol Clin 1978;18:1–16.
include DR screening by optometrists would be cost-effective
13 Liu WJ, Lee LT, Yen MF, et al. Assessing progression and efficacy of
from the perspective of the health care system. Our study lends a treatment for diabetic retinopathy following the proliferative pathway to
strong economic case to support the extension of publicly blindness: implication for diabetic retinopathy screening in Taiwan.
insured optometric services to patients with diabetes. Diabet Med 2003;20:727–33.
14 Chukwuemaka N, Cooper H, Bowen-Jones D. Mortality in Type 2
Diabetes mellitus: magnitude of the evidence from a systematic review
and meta-analysis. Br J Diabetes Vasc Dis 2013;13:192–207.
Acknowledgments 15 Statistics Canada. Canadian Vital Statistics, Deaths Database, 2009,
Survey 3233 and Demography Division, demographic estimates. Ottawa,
Authors JP, GET, RW, JF, SED, and KT contributed to the conception ON: Statistics Canada, 2009.
of the study question and design and obtained the funding. SK and 16 Statistics Canada. Most recent eye exam, diabetic population aged 18 or
KT developed the health economic model, performed the data older. Health report: diabetes—prevalence and care practices. 2005
Canadian Community Health Survey. Ottawa, ON: Statistics Canada,
analysis, interpreted the results and drafted the initial version of 2005.
the manuscript. JP, GET, YB, RW, JF, SED, and MB contributed to 17 Kiran T, Kopp A, Moineddin R, et al. Unintended consequences of
framing the details of the clinical questions and provided clinical delisting routine eye exams on retinopathy screening for people with
expertise for the development of the manuscript. LM provided diabetes in Ontario, Canada. CMAJ 2013;185:E167–73.
18 Matthews DR, Stratton IM, Aldington SJ, et al; UK Prospective Diabetes
methodologic support and helped with a model validation. HAT Study Group. Risks of progression of retinopathy and vision loss related
helped with literature review. All authors reviewed and approved to tight blood pressure control in type 2 diabetes mellitus: UKPDS 69.
the final manuscript and can take responsibility for the integrity of Arch Ophthalmol 2004;122:1631–40.
the data and the accuracy of the data analysis. 19 Heintz E. Health economic aspects of diabetic retinopathy. Ottawa, ON:
Centre for Medical Technology Assessment, Canadian Agency for Drugs
The authors have no conflicts of interest to disclose. and Technologies in Health, 2012.
This study was supported by the Canadian Institutes of Health 20 Health PEI. Master agreement between the Medical Society of Prince
Research (CIHR). Edward Island and the Government of Prince Edward Island and
Health PEI. April 1, 2010 – March 31, 2015. Charlottetown, PEI: Health PEI,
2015.
R EF E R EN CE S 21 Guidelines for the Economic Evaluation of Health Technologies: Canada
(3rd ed). Ottawa, ON: Canadian Agency for Drugs and Technologies in
Health, 2006.
22 Anderson D, Lindley S, LeBlanc R, et al. Submission to the Commission
1 Canadian National Institute for the Blind (CNIB). Paying the price: what on the Future of Health Care in Canada. Ottawa, ON: Canadian
vision loss costs Canadians and what we should do about it. Available Ophthalmological Society, June 2002.
from: http://www.cnib.ca/eng/CNIB%20Document%20Library/Research/ 23 Canadian Institute for Health Information (CIHI). Diabetes care gaps and
Paying_the_Price.pdf. disparities in Canada, 2009. Available from: https://secure.cihi.ca/
2 Canadian Diabetes Association. Diabetes in Canada. Available from: free_products/Diabetes_care_gaps_disparities_aib_e.pdf.
https://www.diabetes.ca/getmedia/513a0f6c-b1c9-4e56-a77c-6a492bf 24 Canadian Diabetes Association. An economic tsunami: the cost of
7350f/diabetes-charter-backgrounder-national-english.pdf.aspx. diabetes in Canada. Available from: https://www.diabetes.ca/CDA/
3 Mohamed Q, Gillies MC, Wong TY. Management of diabetic retinopathy: media/documents/publications-and-newsletters/advocacy-reports/
a systematic review. JAMA 2007;298:902–16. economic-tsunami-cost-of-diabetes-in-canada-english.pdf.
4 Hutchinson A, McIntosh A, Peters J, et al. Effectiveness of screening and 25 Statistics Canada. Health status statistics: diabetes. Available from:
monitoring tests for diabetic retinopathy—a systematic review. Diabet http://www.statcan.gc.ca/pub/82-229-x/2009001/status/dia-eng.htm
Med 2000;17:495–506. [Accessed September 5, 2016].
5 Canadian Diabetes Association Clinical Practice Guidelines Expert 26 Cruess AF, Gordon KD, Bellan L, et al. The cost of vision loss in Canada.
Committee. Canadian Diabetes Association 2013 clinical practice Can J Ophthalmol 2011;46:315–8.