The Cost–Utility of Telemedicine to Screen
for Diabetic Retinopathy in India
Sudhir Rachapelle, MD, MPH,1 Rosa Legood, DPhil,2 Yasmene Alavi, PhD,3
Robert Lindfield, MRCOphth, MFPH,3 Tarun Sharma, MD,1 Hannah Kuper, ScD,3,4 Sarah Polack, PhD3,4
Purpose: To assess the cost-effectiveness of a telemedicine diabetic retinopathy (DR) screening program in
rural Southern India that conducts 1-off screening camps (i.e., screening offered once) in villages and to assess
the incremental cost-effectiveness ratios of different screening intervals.
Design: A cost– utility analysis using a Markov model.
Participants: A hypothetical cohort of 1000 rural diabetic patients aged 40 years who had not been
previously screened for DR and who were followed over a 25-year period in Chennai, India.
Methods: We interviewed 249 people with diabetes using the time trade-off method to estimate utility values
associated with DR. Patient and provider costs of telemedicine screening and hospital-based DR treatment were
estimated through interviews with 100 diabetic patients, sampled when attending screening in rural camps (n ⫽
50) or treatment at the base hospital in Chennai (n ⫽ 50), and with program and hospital managers. The sensitivity
and specificity of the DR screening test were assessed in comparison with diagnosis using a gold standard
method for 346 diabetic patients. Other model parameters were derived from the literature. A Markov model was
developed in TreeAge Pro 2009 (TreeAge Software Inc, Williamstown, MA) using these data.
Main Outcome Measures: Cost per quality-adjusted life-year (QALY) gained from the current teleophthalmology program of 1-off screening in comparison with no screening program and the cost– utility of this program
at different screening intervals.
Results: By using the World Health Organization threshold of cost-effectiveness, the current rural teleophthalmology program was cost-effective ($1320 per QALY) compared with no screening from a health provider
perspective. Screening intervals of up to a frequency of screening every 2 years also were cost-effective, but
annual screening was not (⬎$3183 per QALY). From a societal perspective, telescreening up to a frequency of
once every 5 years was cost-effective, but not more frequently.
Conclusions: From a health provider perspective, a 1-off DR telescreening program is cost-effective
compared with no screening in this rural Indian setting. Increasing the frequency of screening up to 2 years also
is cost-effective. The results are dependent on the administrative costs of establishing and maintaining screening
at regular intervals and on achieving sufficient coverage.
Financial Disclosure(s): The author(s) have no proprietary or commercial interest in any materials discussed
in this article. Ophthalmology 2013;120:566 –573 © 2013 by the American Academy of Ophthalmology.
The prevalence of diabetes is increasing rapidly worldwide.1 There are approximately 35 million people with
diabetes in India, the largest number of diabetic patients
in any given country, and this is predicted to increase to
80 million by 2030.2 Increases in the prevalence of
diabetes have been observed in both urban and rural
settings in India.2,3 Diabetic retinopathy (DR) is a potentially blinding complication of diabetes. The condition
progresses through different stages from mild to moderate nonproliferative stages (nonproliferative diabetic retinopathy [NPDR]), which are largely asymptomatic, to
severe NPDR and proliferative diabetic retinopathy
(PDR), which can lead to blindness if untreated. Vision
loss also can be caused by diabetic macular edema, which
can occur at any stage of DR. Early detection and timely
treatment are effective at reducing the risk of vision loss
from DR.4,5 This requires regular retinal examinations to
detect the sight-threatening stages of the condition before
sight loss occurs.
566
© 2013 by the American Academy of Ophthalmology
Published by Elsevier Inc.
In India, the majority of the population (72%) live in
rural areas.6 However, the majority of ophthalmologists are
based in urban centers.6 Ensuring regular diabetic ophthalmic examinations for rural populations therefore poses a
considerable challenge.7 To address this gap in access to
ophthalmic staff, a rural mobile telemedicine screening program has been established in Tamil Nadu, which aims to
provide examination services for populations who may not
otherwise have access to such a program. The program uses
a digital retinal camera that is linked via satellite to a base
hospital in urban Chennai. Ophthalmologists at the base
hospital can examine retinal photographs taken in the village and provide DR diagnoses to the patient in real time.
Studies in other countries suggest that community-based
telemedicine DR screening increases the eye examination
rates of diabetic patients.8
Regular screening for DR has been shown to be costeffective in different high-income settings.9 –12 There is also
evidence that telemedicine screening for DR is cost-effective
ISSN 0161-6420/13/$–see front matter
http://dx.doi.org/10.1016/j.ophtha.2012.09.002
Rachapelle et al 䡠 Cost–Utility of DR Screening in India
in some populations, including residents of a remote municipality in Norway13 and a prison population in the United
States.14 In South India, Rachapelle et al15 found that the
cost per case of DR detected for rural diabetic patients
screened through rural teleophthalmology compared with
screening of rural patients at an urban hospital was higher
from the health provider perspective, but lower once patient
costs of attending screening (i.e., travel and income loss)
were included. They further commented that because of the
large travel distances involved, attendance at hospital
screening was likely to be poor, suggesting this may not be
a viable option for rural populations. To the best of our
knowledge, a comprehensive cost–utility analysis of telemedicine screening programs for DR in rural India or other
low-income settings has not been undertaken. In addition, it
is unclear what the optimum screening frequency of diabetic
persons for DR should be in this setting. Although some
studies in high-income countries indicate that annual
screening is optimal,12 Vijan et al10 suggested that screening
every 2 years might be sufficient for the majority of diabetic
patients.
This study aimed to assess the cost-effectiveness of the
Sankara Nethralaya Medical Research Foundation teleophthalmology program in Southern India. We estimated the
provider and societal costs of their rural telemedicine
screening and of hospital-based DR treatment, the sensitivity and specificity of the DR examination method used, and
the utility values associated with different stages of DR in
this setting. This information was used to develop a model
to estimate the cost per quality-adjusted life-year (QALY)
gained from this screening approach and the incremental
cost–utility ratios associated with different screening
intervals.
Materials and Methods
Study Setting and Program Description
The study was undertaken in rural Tamil Nadu, Southern India. A
recent survey in rural Tamil Nadu estimated the diabetes prevalence at 7.8% among people aged ⱖ20 years.16
The Sankara Nethralaya Medical Research Foundation teleophthalmology program conducts DR screening camps in villages in
the rural districts neighboring Chennai. In the current program,
1-off screening (i.e., screening offered once) is conducted in each
village rather than diabetic persons being invited for screening
repeatedly at regular intervals. People with known diabetes (i.e.,
those with a doctor diagnosis of diabetes or receiving drug, insulin,
or diet control for diabetes) living within a 25-km radius of the
screening camp are informed of the screening day via their general
practitioners and via door-to-door visits by trained village-level
volunteers, local newspapers, and posters. All attendees are enumerated and undergo a visual acuity (VA) examination at a central
place (e.g., school or community center). The program uses a
customized mobile van with an in-built ophthalmic unit, where the
patients are examined by an optometrist using a handheld slit-lamp
and have 4 dilated stereoscopic 45-degree fields digital retinal
photographs taken (posterior pole covering disc and macula, nasal,
superior, and inferior) using a nonmydriatic camera (Orion fundus
camera, Nidek Technologies Srl, Albignasego, Padova, Italy). The
retinal images are transferred by satellite to the base hospital in
urban Chennai in real time where they are reviewed by a vitreo-
retinal surgeon and graded using the international DR classification system.17 Diagnosis and management advice are given directly to the patient by the surgeon via the satellite link-up. All
people identified with severe NPDR, PDR, or clinically significant
macular edema (grouped as sight-threatening diabetic retinopathy
[SDTR] for the purposes of this program) are referred to the base
hospital, Sankara Nethralaya. Patients with no signs of DR or with
mild or moderate nonproliferative retinopathy (non-STDR) are not
referred.
At the base hospital, referred patients undergo reexamination
by an ophthalmologist using a slit-lamp and indirect ophthalmoscope. Some patients also will undergo further investigations as
necessary (e.g., fluorescein angiography). Laser photocoagulation
treatment is provided as appropriate.
Eye examination at the screening camps is provided free of
charge to the patients. Examination and treatment costs at the
hospital are typically paid by a charity for patients referred through
this telemedicine program.
Model Structure
We used a Markov model to estimate the cost–utility of (1) the
current teleophthalmology program, which involves 1-off screening in comparison with no screening program; and (2) the cost–
utility of this program at different screening intervals. The model
was developed in TreeAge Pro 2009 (TreeAge Software Inc,
Williamstown, MA). The model was based on a hypothetical
cohort of 1000 rural diabetic patients 40 years of age who had not
been previously screened for DR. Each cycle of the model was 1
year, and the simulation costs and outcomes were estimated over
a 25-year period on the basis of average life expectancy in India.
Simulated patients were classified into 1 of 5 different health
states: no DR, non-STDR (includes mild and moderate NPDR),
STDR (includes severe NPDR and PDR), clinically significant
macula edema, and blind from DR (bilateral best-corrected VA
⬍6/60).
These groupings are based on the international classification
system and modified to reflect the referral system in the study
setting whereby all patients with severe NPDR, PDR, and clinically significant macula edema are considered to have STDR and
referred for further examination and treatment. We used the definition of blindness most commonly used in India: VA ⬍6/60. In
the model, patients were screened after the teleophthalmology
program protocol described earlier. Those identified with STDR
were referred to the base hospital for further ophthalmic examination. Patients with confirmed STDR received laser photocoagulation treatment and were subsequently referred for annual examination at the hospital.
Screening Strategies Evaluated
We compared 6 different screening interval strategies: no screening, once-in-a-lifetime screening (current approach) and twice-ina-lifetime screening, screening every 5 years, screening every 3
years, screening every 2 years, and annual screening. Incremental
cost-effectiveness ratios (ICERs) were calculated for each successive strategy comparing the most with the next least costly (e.g.,
comparing annual screening with screening every 2 years). The
ICERs were calculated as the difference in cost of the 2 strategies
divided by the difference in QALYs gained. The optimal strategy
was assessed by comparing ICERs with a threshold for costeffectiveness. In the absence of an established Indian-recommended
threshold, we used the approach recommended by the World
Health Organization,18,19 which is based on the gross domestic
product (GDP) for India ($1061 in 2009) as follows: Interventions
less than the GDP (⬍US $1061 per QALY) are considered very
567
Ophthalmology Volume 120, Number 3, March 2013
Table 1. Health Provider, Household, and Societal Costs Per
Person for Diabetic Retinopathy Screening and Treatment in
US Dollars
cost-effective, interventions between 1 and 3 times the GDP
($1061–$3183 per QALY) are considered cost-effective, and interventions more than 3 times the GDP (⬎$3183 per QALY) are
considered not cost-effective.
advertising the camp, mobile van, and fuel; equipment for ophthalmic examination and satellite connection; consumables; and
the community building. Detailed information on the resources
used and their associated quantity and value were estimated from
interviews with the program manager, as well as screening camp
log books and payroll records. We derived per patient cost using
the average number of people screened collected from program
records over the past 2 years.
Information on hospital retinal examination and treatment costs
was provided by Sankara Nethralaya Hospital Chennai and included personnel, equipment, consumables, and buildings. Average staff time required per patient for retinal examination and
treatment was estimated through interviews.
Household Costs. To estimate costs to the household, we
interviewed 25 consecutive people attending 2 different rural
screening camps (total of 50 interviews) and 50 consecutive patients attending the base hospital in Chennai City for laser
photocoagulation.
Patients were interviewed using structured questionnaires about
direct costs (travel, food, accommodation, hospital fees, and medicines) and indirect costs (income lost) incurred by the patient and
his or her carer(s) as a result of attending screening/treatment.
Indirect costs included paid work only. Treatment costs included
expenses associated with the retinal examination at the hospital
and the first laser photocoagulation treatment.
“Societal” costs for telescreening were calculated as the sum of
the health provider and household costs. For hospital costs, this
approach would introduce double counting because hospital fees
charged to patients (or the charity covering the fees) include a
profit margin and are in excess of the actual costs incurred by the
hospital in providing the services. To avoid double counting,
hospital societal costs were calculated as the sum of household
direct and indirect costs minus the provider costs.
Model Parameters
Effectiveness
Primary data were collected on the costs of screening and treatment, utility values associated with different stages of DR (effectiveness), and sensitivity and specificity of the retinal examination
method. Other model parameters were derived from the published
literature (specified below).4,9,10,20 –23
We estimated utility values associated with different stages of DR.
The methodology and results will be presented in detail in a future
publication. Briefly, 249 people with diabetes aged ⱖ40 years
from urban Chennai were recruited from an ongoing DR epidemiology study and a laser clinic at Sankara Nethralya hospital. Of the
249 participants, 30 had no DR, 73 had NPDR, 114 had STDR,
and 32 were blind from DR. Utility values were elicited during
face-to-face interviews in Tamil using the time trade-off method,
whereby the maximum number of years a patient was willing to
trade for perfect vision was elicited using a series of choice-based
questions.25
Teleophthalmology
Health provider costs
Personnel
Consumables
Buildings
Equipment
Utilities and
administration
Transport
Promotion/advertising
Totals
Household costs
Transport
Accommodation
Food
Hospital fees
Medication
Income loss
Totals
Societal costs
Hospital
Screening
Dilated
Retinal
Examination
Laser
Photocoagulation
1.94
0.50
0.91
1.60
0.09
1.91
0.56
0.60
0.65
0.55
2.09
0.35
0.41
2.38
0.78
0.65
0.43
7.36
0
0
5.84
0
0
7.51
0.30
0.00
0.54
0.00
0.00
1.17
2.02
9.38
5.32
0.37
1.79
7.48
0.04
4.72
20.57
14.72
4.33
0.19
1.17
44.96
0.03
2.02
52.70
45.19
Costs
Health Provider Costs. We estimated the per person provider
cost of (1) telemedicine screening and (2) retinal examinations and
laser photocoagulation treatment at Sankara Nethralaya hospital,
Chennai (Table 1). We used an ingredients costing approach
whereby information was collected on the quantities of the different resources used and the value (or prices) of each resource, and
the total cost was calculated as the product of quantity and the
value. Costs were collected in Indian rupees and converted into US
dollars on the basis of the average exchange rate for 2009 of US
$ ⫽ 46.8 rupees. We used the economic definition of cost whereby
the costs of all inputs were included whether or not they incurred
a direct financial cost to the program. For example, costs of
donated items (e.g., the satellite dish) were included on the basis of
the principal of opportunity cost: the value forgone by not using
this resource for the next best alternative use. Capital item costs
were calculated using the equivalent annual cost method with a
discount rate of 3%.24 Equipment was assumed to have a 5-year
life span for computer hardware and digital cameras and 10-year
life span for medical/transport equipment.
Telescreening costs included the personnel time (for screening
camp staff and the vitreoretinal surgeon at the base hospital);
568
Accuracy of Diagnostic Tool for Diabetic
Retinopathy
We assessed the sensitivity and specificity of the screening test
used in the teleophthalmology program by comparing it with a
gold standard of dilated 7 stereoscopic 45-degree field retinal
images taken using a Carl Zeiss fundus mydriatic camera (Visucamlite, Jena, Germany). For this evaluation, 346 participants were
recruited from a diabetic clinic in Chennai using quota sampling so
that there were at least 50 people in each group: no DR, mild
NPDR, moderate NPDR, severe NPDR, and PDR. Retinal photographs taken using the 2 different cameras were graded by separate
ophthalmologists. These data were used to estimate the sensitivity
and specificity of teleophthalmology screening tool and the proportion of patients in whom the stage of DR was misdiagnosed
using this method.
Rachapelle et al 䡠 Cost–Utility of DR Screening in India
Other Model Parameters
The baseline distribution of the population among the different
health states was estimated from a population-based survey of DR
among people aged ⱖ40 years in Chennai.20 We applied the
prevalence estimates for the survey participants aged 40 to 50
years. Annual progression rates between the different DR stages
were estimated from previous studies,9,10,21 with calibration to DR
prevalence estimates derived from the recent population-based
survey in Chennai.20 The effectiveness of photocoagulation treatment on reducing the rate of progression to blindness was estimated from the results of the Early Treatment Diabetic Retinopathy Study trials.4 Estimates of the mortality rates were taken from
Indian life tables, with a mortality multiplier for blindness and
diabetes based on observational studies.22,23 The proportion of
patients identified as having STDR at telescreening who actually
attended their referral to the hospital was estimated to be 50%
according to telescreening records.
We made the following assumptions in the model:
●
●
●
All patients with confirmed STDR at the base hospital would
undergo the recommended laser photocoagulation treatment.
Three laser treatments were given to patients with PDR and
1 treatment was given to patients with macular edema, as is
typical practice in this setting.
After treatment, patients would be examined annually at the
base hospital and would no longer be included in the telescreening program.
Sensitivity Analysis
We undertook a 1-way sensitivity analysis in which the parameter
values were varied one at a time using the estimate ranges shown
in Table 2. The minimum and maximum values were from 95%
confidence intervals (CIs) for the prevalence estimates, utility
values, diagnosis accuracy, and mortality multiplier. For costs, a
range of ⫾30% was applied. Hospital attendance after referral was
estimated to range from 30% to 80% according to hospital records
and discussion with program managers.
We also carried out a probabilistic sensitivity analysis that
enables uncertainty across all the variables to be varied simultaneously. This approach takes repeated (10 000) samples from
across the ranges of all the parameters. Cost-effectiveness acceptability curves were used to summarize the uncertainty on costeffectiveness estimates at different willingness to pay thresholds.
Ethical Considerations
Ethical approval for this study was obtained from the ethics
committees of the London School of Hygiene & Tropical Medicine and Vision Research Foundation Chennai, India. Written
informed consent was obtained from all participants in the interviews and the assessment of the accuracy of the diagnosis methods. The research adhered to the tenets of the Declaration of
Helsinki.
Results
Table 2. Markov Model Parameter Estimates and Assumptions
Parameter
Utility values
No DR
Non-STDR DR
STDR
Bilateral blindness from DR
(VA ⬍6/60)
Annual transition probabilities9,10,14,21
No DR to mild/moderate NPDR
Non-STDR to STDR
Macula edema to blind without
laser
Macula edema to blind with laser
PDR to blind without laser
PDR to blind with laser
Prevalence of DR among diabetic
persons20
No DR
Non-STDR
STDR
Proportion of patients misdiagnosed
by retinal camera
No DR graded as mild/moderate
NPDR
No DR graded as STDR
Non-STDR graded as no DR
Non-STDR graded as STDR
STDR graded as no DR
STDR graded as non-STDR
Mortality multipliers7,8
Diabetes
Blind
Probability of attending for treatment
after referral
Baseline
Case
Sensitivity
Analysis
Range
0.87
0.79
0.70
0.55
0.78–0.97
0.72–0.86
0.64–0.77
0.46–0.64
0.06
0.06
0.05
0.01–0.10
0.01–0.11
0.03–0.07
0.03
0.09
0.02
0.01–0.05
0.05–0.11
0.002–0.03
0.89
0.11
0.004
0.84–0.93
0.06–0.16
0.0–0.016
0.05
0.01–0.12
0.00
0.29
0.13
0.09
0.23
0.00
0.21–0.39
0.07–0.21
0.05–0.16
0.16–0.32
1.90
2.34
0.50
1.04–2.7
2.22–2.46
0.30–0.80
DR ⫽ diabetic retinopathy; NPDR ⫽ nonproliferative diabetic retinopathy; PDR ⫽ proliferative diabetic retinopathy; STDR ⫽ sight-threatening
diabetic retinopathy; VA ⫽ visual acuity.
were $9.38. Costs for retinal examination and a single laser photocoagulation treatment at the hospital were $5.84 and $7.51 per
person, respectively, from the health provider perspective. From
the societal perspective, these costs increased to $14.72 per person
for retinal examination and $45.19 for laser photocoagulation.
Hospital fees were the largest cost incurred by the household (85%
of the total cost).
Table 2 shows the other Markov model parameter estimates.
Mean utility values worsened with increasing severity of DR, and
the utility value for bilateral blindness from DR was 0.55 (95% CI,
0.46 – 0.64). The proportion of patients misdiagnosed by teleophthalmology examination according to the sensitivity and specificity
analysis was generally low. The mortality rate was approximately
doubled for diabetic or blind people.
Model Parameters
Cost-effectiveness
Table 1 shows the per-person costs of telescreening, hospital
retinal examinations, and laser photocoagulation treatment from
the health provider and societal perspective. The health provider
costs of telescreening were $7.36 per person screened, and the
societal costs (i.e., including direct and indirect household costs)
Table 3 shows the costs, QALYs, and ICERs for each strategy. The
incremental QALYs gained from increased frequency of screening
are relatively small, which concurs with the relatively small
proportion of people who progress to blindness over the model
duration.
569
Ophthalmology Volume 120, Number 3, March 2013
Table 3. Cost-effectiveness of Different Telemedicine Diabetic Retinopathy Screening Intervals from the Health Provider and
Societal Perspective
Strategy
Health provider perspective
No screening
Screening once in a lifetime
Screening twice in a lifetime
Screening every 5 yrs
Screening every 3 yrs
Screening every 2 yrs
Annual screening
Societal perspective
No screening
Screening once in a lifetime
Screening twice in a lifetime
Screening every 5 yrs
Screening every 3 yrs
Screening every 2 yrs
Annual screening
Cost $
Incremental
Cost
Quality-Adjusted
Life-Year
Incremental
Quality-Adjusted Life-Year
Incremental
Cost-Effectiveness Ratio*
0.0
6.5
11.7
31.4
48.8
67.2
118.5
—
6.5
5.3
19.6
17.4
18.4
51.4
12.6722
12.6771
12.6810
12.6907
12.6992
12.7067
12.7195
—
0.0049
0.0039
0.0097
0.0084
0.0075
0.0127
—
1320
1343
2027
2034
2435
4029
0.0
13.2
22.9
53.2
81.7
109.4
181.7
13.2
9.7
30.3
28.4
27.7
72.4
12.6722
12.6771
12.6810
12.6907
12.6992
12.7067
12.7195
0.0049
0.0039
0.0097
0.0084
0.0075
0.0127
2692
2475
3134
3365
3669
5677
*Incremental cost-effectiveness ratio calculated for each successive alternative comparing the least costly with the most costly.
The cheapest though least-effective strategy is once in a lifetime screening. From a health provider perspective and compared
with no screening, once-in-a-lifetime screening has a cost–utility
ratio of $1320 per QALY gained, which is within the range
considered “cost-effective” in this setting ($1061–3183). Increasing the frequency of screening to twice in a lifetime and every 5,
3, or 2 years, respectively, increases the costs but also increases the
QALYs gained. The ICER of all these strategies were within the
cost-effective range ($1061–$3183/QALY). Increasing to annual
screening doubled the QALYs gained, but the additional cost
placed the ICER ($4029/QALY) outside of the range considered
cost-effective in this setting.
If the societal perspective is considered, screening once in a
lifetime, twice in a lifetime, and every 5 years is cost-effective
($1061–3183/QALY), whereas the ICER for screening strategies of every 3 to 1 years is no longer below the cost-effective
threshold.
Sensitivity Analysis
The most influential parameters in determining the ICERs were
utility values, transition probabilities, and costs of telescreening.
By considering the health provider perspective, if the utility value
for blindness is reduced to 0.46 (lower CI), the ICERs for onceand twice-in-a-lifetime screening seems to be highly cost-effective
(⬍$1048/QALY) and the ICER for annual screening improves to
$2639 (i.e., within the threshold of cost-effectiveness). If the utility
value for blindness is increased to 0.64 (upper CI), the ICERs are
⬎$3144/QALY for all screening frequencies greater than every 5
years. A 30% reduction in the cost of screening moved the ICER
for annual screening into the cost-effective range ($2905). Finally,
applying the highest of the range estimates for progression rates
from no DR to NPDR and from NPDR to STDR increases the
cost-effectiveness for all ICERS and moved annual screening to
within the range considered cost-effective. Applying the lower end
of these estimates reduces the ICERs so that any strategy more
frequent than every 5 years would not be considered cost-effective.
Varying other parameters (i.e., costs of hospital examination and
treatment, sensitivity and specificity of diagnostic test for DR,
hospital attendance rate) had negligible influence on the ICER
estimates.
570
The results of the probabilistic sensitivity analysis are shown in
Figure 1. This suggests that within the ranges considered costeffective on the basis of the GDP, there is considerable uncertainty
as to the optimal strategy. Considering the cost-effective threshold
of 3 times the GDP ($3144), screening every 2 years has the
highest probability of being cost-effective (28% simulations). If
the threshold is increased to ⬎$3144, annual screening has the
highest probability of being most cost-effective.
Discussion
Our cost–utility model suggests that from a health provider
perspective, the current Sankara Nethralaya rural teleophthalmology screening program is cost-effective compared
with no DR screening for rural diabetic patients if the World
Health Organization–suggested threshold (1–3 times the
Indian GDP) is used to define cost-effectiveness.18,19 Increasing the screening frequency to provide screening at
regular intervals would increase the costs of the program;
however, the increased QALYs gained (because of reduction in progression to sight loss among treated patients)
would suggest that screening up to a frequency of once
every 2 years would also be considered cost-effective in this
setting. Annual screening fell outside of the cost-effective
range. If household costs are included to provide estimates
from a societal perspective, the program becomes more
expensive, although the majority (85%) of the additional
cost is from the hospital fees charged to patients rather than
travel costs or productivity losses. From the societal perspective, screening once or twice per lifetime remains costeffective, but systematic screening at intervals of 5 years or
more is no longer within the cost-effectiveness range.
There are few studies from low- and middle-income
patients with which to compare our findings. Further, caution on comparison with other studies is warranted because
of variation in cost-effectiveness analysis methods used.
Our findings concur with previous studies in the United
Rachapelle et al 䡠 Cost–Utility of DR Screening in India
No Screen
Every 3 years
Once in a life me
Every 2 years
Twice in a lifeme
Annual
Every 5 years
1
Cost-effecve Probability
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
600
1200
1800
2400
3000
3600
4200
4800
5400
6000
Willingness to pay ($/Quality adjusted Life Year)
Figure 1. Cost-effectiveness acceptability curve generated through probabilistic sensitivity analysis: probability that screening interval is cost-effective
given willingness to pay.
Kingdom that show that systematic screening compared
with opportunistic screening for DR was cost-effective.26,27
Studies in Norway (remote populations)13 and the United
States (prison populations)14 also concluded that telemedicine was a cost-effective screening strategy. In terms of
screening interval, annual screening is recommended in
many countries, and a study in the United Kingdom has
found evidence for the cost-effectiveness of this approach.12
In contrast, a cost–utility analysis of DR screening in the
United States indicated that for the majority of diabetic
patients, screening every 2 years was sufficient and that
annual screening added considerable extra costs for marginal benefit when compared with screening every other
year.10 The authors recommended that screening frequency
could be adapted according to patient risk factors (e.g., age
and glycemic control) so that those at greater risk were
examined more frequently. These risk factors were not
included in our model because it is not considered feasible
to implement different screening protocols for different
patients in this setting.
Our sensitivity analysis showed that the cost of screening
influenced the cost-effectiveness of the program. Reducing
per person costs by 30% resulted in improved incremental
cost-effectiveness of all strategies and moved the ICER for
annual screening to within 1 to 3 times the GDP. Given that
equipment contributes one of the largest costs, it may be
possible to reduce per person screening costs by increasing
the number of people screened (i.e., by increasing number
of screening camps per year). Retinal examinations undertaken at the camp through ophthalmoscopy instead of retinal
photographs sent via satellite to the hospital also may reduce costs. However, this would be dependent on the availability and willingness of trained ophthalmic staff to attend
the screening camps. In accordance with other DR screening
economic evaluations, our model was sensitive to the utility
value used for blindness.9,10 The lower utility value parameter estimate (0.46) that increased the effectiveness and
ICERs of all screening strategies is similar to that used in a
previous US study that found DR screening to be highly
cost-effective.9 In contrast, applying the higher estimate
(0.64) increased the ICERs of screening frequencies greater
than every 5 years to ⬎$3144/QALY. This value is similar
to that applied by Vijan et al,10 who suggest that for the
majority of diabetic patients in the United States, annual
screening may not be warranted on the basis of costeffectiveness.10 A strength of our study was that we derived
our utility value (0.55) from a sample of diabetic patients
from the program area. Transition probabilities among no
DR, NPDR, and STDR also were important determinants of
cost-effectiveness. In the absence of data on rates of transition for the Indian population, we relied on estimates from
previous studies in non-Asian populations calibrated to reflect the distribution of DR states in the recent populationbased survey in Chennai.20 Calibration led to slightly lower
estimates of transition between no DR to NPDR and STDR
than in some previous economic evaluation studies.9,28 Accordingly, population-based surveys of DR in India report a
lower prevalence of DR and STDR among diabetic patients
compared with other countries, which may suggest differing
epidemiology of the disease in this setting.20,29 Epidemiologic data on rates of DR progression in India are needed to
further inform screening program evaluation. The results of
the probabilistic sensitivity analysis suggest that there is a
great deal of uncertainty about the optimal strategy. It
suggests that no screening would be the best option at the
threshold for interventions to be cost-effective. However, at
571
Ophthalmology Volume 120, Number 3, March 2013
the higher threshold for cost-effective screening, screening
every 2 years was the best option.
An important strength of this study is that we collected
primary data on costs, effectiveness, and diagnostic tool
accuracy from the program setting rather than relying on
published studies undertaken in other countries (mainly in
the United Kingdom and United States). In addition, data on
baseline DR prevalence were estimated from a recent population-based study conducted in Chennai.20 We also undertook a probabilistic sensitivity analysis that allowed exploration of uncertainty around the cost and effectiveness.
Study Limitations
This study had a number of limitations. Our model assumed
100% attendance of the initial cohort at each subsequent DR
screening. Although there is evidence that DR screening in
rural communities may increase compliance compared with
hospital-based screening,8,30 complete attendance may not
be realistic. Compliance influences cost-effectiveness when
comparing screening intervals. Davies et al12 found that
lower compliance necessitates more frequent screening in
their analysis of DR screening in the United Kingdom.
Compliance rates in this setting need further study to inform
decisions about implementing systematic telescreening programs. We did not include the costs of establishing and
maintaining an administrative system for regular (i.e., every
1 to 5 years) screening (e.g., setting up screening databases,
contacting screening participants, monitoring attendance)
because empirical data on these costs are not available.
However, from the model, we calculated that if implementing such a system increased the telescreening costs to more
than $12.40 per person, then screening at any of the regular
screening intervals (i.e., every 5 years and more) would no
longer be within the cost-effective range. Assessment of the
cost and logistic feasibility of regular screening is needed.
We did not include potential societal cost-savings of averting costs associated with blindness, such as productivity
gains, and this may underestimate cost-effectiveness. In
addition, we applied a single average utility value for blindness (defined as VA ⬍6/60) rather than different estimates
for progressively lower levels of vision loss. The estimated
prevalence of DR among diabetic patients was obtained
from an urban Chennai sample,20 and if the prevalence of
DR is lower in the rural population, then this may reduce the
cost-effectiveness of telemedicine screening.
Our analysis has highlighted the need for further research
into areas that are likely to influence estimates of costs and
effectiveness of DR screening programs. First, further data
are needed on diabetes and DR epidemiology in rural India,
and in particular on DR progression rates. Second, investigation into the costs and logistic feasibility of setting up and
maintaining systematic rural telescreening programs is needed,
as well as data on compliance of diabetic patients in such
programs. Finally, a clear definition of cost-effectiveness for
India would be of value for facilitating interpretation of
economic evaluations.
In conclusion, diabetic populations in rural communities
in India typically have limited access to ophthalmic care. A
recent study in rural Tamil Nadu found that the majority of
572
people with diabetes (63%) had not previously had a retinal
examination.31 Our analysis suggests that a 1-off telescreening program for DR is cost-effective compared with no
screening in this rural Indian setting from a health provider
perspective. Increasing the frequency of screening up to 2
years also is cost-effective, although this is dependent on the
administrative costs of establishing and maintaining screening at regular intervals and on achieving sufficient coverage.
References
1. International Diabetes Federation. IDF Diabetes Atlas. 4th ed.
Brussels, Belgium: IDF Executive Office; 2009:25.
2. Wild S, Roglic G, Green A, et al. Global prevalence of
diabetes: estimates for the year 2000 and projections for 2030.
Diabetes Care 2004;27:1047–53.
3. Misra P, Upadhyay RP, Misra A, Anand K. A review of the
epidemiology of diabetes in rural India. Diabetes Res Clin
Pract 2011;92:303–11.
4. Early Treatment Diabetic Retinopathy Study Research Group.
Early photocoagulation for diabetic retinopathy. ETDRS report number 9. Ophthalmology 1991;98(Suppl):766 – 85.
5. Diabetic Retinopathy Study Research Group. Photocoagulation treatment of proliferative diabetic retinopathy. Clinical
application of Diabetic Retinopathy Study (DRS) findings,
DRS report number 8. Ophthalmology 1981;88:583– 600.
6. Prathiba V, Rema M. Teleophthalmology: a model for eye
care delivery in rural and underserved areas of India. Int J
Family Med [serial online] 2011;2011:683267. Available at:
http://www.hindawi.com/journals/ijfm/2011/683267/. Accessed August 18, 2012.
7. Sharma M, Chakrabarty AS, Pavan R, et al. An integrated,
mobile service for diabetic retinopathy in rural India. Community Eye Health 2011;24:17– 8.
8. Davis RM, Fowler S, Bellis K, et al. Telemedicine improves
eye examination rates in individuals with diabetes: a model for
eye-care delivery in underserved communities [letter]. Diabetes Care 2003;26:2476.
9. Javitt JC, Aiello LP. Cost-effectiveness of detecting and treating diabetic retinopathy. Ann Intern Med 1996;124:164 –9.
10. Vijan S, Hofer TP, Hayward RA. Cost-utility analysis of
screening intervals for diabetic retinopathy in patients with
type 2 diabetes mellitus. JAMA 2000;283:889 –96.
11. Jones S, Edwards RT. Diabetic retinopathy screening: a systematic review of the economic evidence. Diabet Med 2010;
27:249 –56.
12. Davies R, Roderick P, Canning C, Brailsford S. The evaluation of screening policies for diabetic retinopathy using simulation. Diabet Med 2002;19:762–70.
13. Bjorvig S, Johansen MA, Fossen K. An economic analysis of
screening for diabetic retinopathy. J Telemed Telecare 2002;
8:32–5.
14. Aoki N, Dunn K, Fukui T, et al. Cost-effectiveness analysis of
telemedicine to evaluate diabetic retinopathy in a prison population. Diabetes Care 2004;27:1095–101.
15. Rachapelle SR, Frick KD, Raman R, et al. Mobile teleophthalmology: a cost effective screening tool for diabetic
retinopathy in rural South India. eHealth Int 2005;2:2– 8.
16. Anjana RM, Pradeepa R, Deepa M, et al, ICMR–INDIAB
Collaborative Study Group. Prevalence of diabetes and prediabetes (impaired fasting glucose and/or impaired glucose tolerance) in urban and rural India: phase I results of the Indian
Council of Medical Research-INdia DIABetes (ICMRINDIAB) study. Diabetologia 2011;54:3022–7.
Rachapelle et al 䡠 Cost–Utility of DR Screening in India
17. Wilkinson CP, Ferris FL III, Klein RE, et al, Global Diabetic
Retinopathy Project Group. Proposed international clinical
diabetic retinopathy and diabetic macular edema disease severity scales. Ophthalmology 2003;110:1677– 82.
18. World Health Organization Commission on Macroeconomics
and Health. Macroeconomics and Health: Investing in
Health for Economic Development. Report of the Commission
on Macroeconomics and Health. Geneva: World Health
Organization; 2001:103. Available at: http://whqlibdoc.who.int/
publications/2001/924154550x.pdf. Accessed August 18, 2012.
19. Hutubessy RCW, Baltussen RMPM, Edejer TTT, Evans DB.
Generalized cost-effectiveness analysis: an aid to decision
making in health. In: Tan-Torres Edejer T, Baltussen R, Adam
T, et al, eds. Making Choices in Health: WHO Guide to CostEffectiveness Analysis. Geneva; World Health Organization; 2003:277– 88. Available at: http://www.who.int/choice/
publications/p_2003_generalised_cea.pdf. Accessed August
18, 2012.
20. Raman R, Rani PK, Reddi Rachepalle S, et al. Prevalence of
diabetic retinopathy in India: Sankara Nethralaya Diabetic
Retinopathy Epidemiology and Molecular Genetics Study report 2. Ophthalmology 2009;116:311– 8.
21. Tung TH, Chen SJ, Shih HC, et al. Assessing the natural
course of diabetic retinopathy: a population-based study in
Kinmen, Taiwan. Ophthalmic Epidemiol 2006;13:327–33.
22. Klein R, Moss SE, Klein BE, DeMets DL. Relation of ocular
and systemic factors to survival in diabetes. Arch Intern Med
1989;149:266 –72.
23. Mohan V, Shanthirani CS, Deepa M, et al. Mortality rates due
to diabetes in a selected urban south Indian population—the
Chennai Urban Population Study [CUPS—16]. J Assoc Physicians India 2006;54:113–7.
24. Johns B, Baltussen R, Hutubessy R. Programme costs in the
economic evaluation of health interventions. Cost Eff Resour
Alloc [serial online] 2003;1:1. Available at: http://www.
resource-allocation.com/content/1/1/1. Accessed August 18,
2012.
25. Gudex C, ed. Time Trade-off User Manual: Props and Selfcompletion Methods. York, UK: Centre for Health Economics, University of York; 1994:1–54. Occasional Paper Series.
Available at: http://www.york.ac.uk/media/che/documents/
papers/occasionalpapers/CHE%20Occasional%20Paper%2020.
pdf. Accessed August 18, 2012.
26. Facey K, Cummins E, Macpherson K, et al. Health Technology
Assessment Report 1: Organisation of services for diabetic retinopathy screening. Glasgow, Scotland: Health Technology
Board for Scotland; 2002:88 –121. Available at: http://www.
ndrs.scot.nhs.uk/Links/Docs/hta1.pdf. Accessed August 18,
2012.
27. James M, Turner DA, Broadbent DM, et al. Cost effectiveness
analysis of screening for sight threatening diabetic eye disease. BMJ 2000;320:1627–31.
28. Tung TH, Chen SJ, Lee FL, et al. A community-based study
for the utility values associated with diabetic retinopathy
among type 2 diabetics in Kinmen, Taiwan. Diabetes Res Clin
Pract 2005;68:265–73.
29. Rema M, Premkumar S, Anitha B, et al. Prevalence of diabetic
retinopathy in urban India: the Chennai Urban Rural Epidemiology Study (CURES) eye study, I. Invest Ophthalmol Vis
Sci 2005;46:2328 –33.
30. Au A, Gupta O. The economics of telemedicine for vitreoretinal diseases. Curr Opin Ophthalmol 2011;22:194 – 8.
31. Namperumalsamy P, Nirmalan PK, Ramasamy K. Developing
a screening program to detect sight-threatening diabetic retinopathy in South India. Diabetes Care 2003;26:1831–5.
Footnotes and Financial Disclosures
Originally received: May 2, 2012.
Final revision: September 4, 2012.
Accepted: September 4, 2012.
Available online: December 1, 2012.
4
Manuscript no. 2012-634.
1
Department of Preventive Ophthalmology, Sankara Nethralaya, Vision
Research Foundation, Chennai Tamil Nadu, India.
2
Department of Health Services Research and Policy, London School of
Hygiene & Tropical Medicine, London, United Kingdom.
3
International Centre for Eye Health, Clinical Research Department, London School of Hygiene & Tropical Medicine, London, United Kingdom.
International Centre for Evidence in Disability, London School of Hygiene & Tropical Medicine, London, United Kingdom.
Financial Disclosure(s):
The author(s) have no proprietary or commercial interest in any materials
discussed in this article.
This study was funded by a grant from Sightsavers. The funding organization had no role in the design or conduct of this research.
Correspondence:
Hannah Kuper, ScD, London School of Hygiene & Tropical Medicine,
Keppel Street, London WC1E 7HT, UK. E-mail: Hannah.Kuper@
lshtm.ac.uk.
573