recommend referral for severe nonproliferative retinopathy, proliferative retinopathy, diabetic maculopathy, presence of photocoagulation scars, and presence of nondiabetic pathology discovered coincidentally. The term diabetic maculopathy includes exudates within 1 disk diameter of the fovea, any circinate or group of exudates within the macula, and any microaneurysm or hemorrhage within 1 disk diameter of the fovea.61 At the most liberal end of the spectrum for referral criteria, some groups mandate referral for presence of any retinopathy more severe than having microaneurysms alone.5 Others set the bar for referral at a more severe level of retinopathy. Some guidelines

place all ungradable photographs in the referral category.5,26,63 The criteria for referral will mark-

edly impact the cost of screening programs. For example, changing the threshold for referral from very mild DR (ETDRS level 20) to mild DR (ETDRS level 35) changed rates of referral from 36.1 to 26.1% of those undergoing photographic screening in one study.11

14.14Obstacles to the Use of Teleophthalmic Screening Methods

Several obstacles hinder the acceptance of teleophthalmic screening for DR. Perhaps the greatest is the concern among ophthalmologists that patients and primary care doctors will mistakenly interpret screening photographs as a substitute for a dilated eye examination by an ophthalmologist. Although these systems are implemented with explicit acknowledgment that they serve as a net to catch those non-compliant patients who will not be seen for this recommended process, in actual practice the concern seems valid.21 Reimbursement concerns are another obstacle slowing acceptance of these methods. In countries without nationalized health-care systems, primary care physicians are reluctant to invest in the camera, although prices are falling and retinopathy screening companies have developed proformas that identify the volumes of patients required in a given time interval at prevailing rates of reimbursement to ensure no financial loss. Worry about legal liability is another issue for

photographic graders. In countries with nationalized health care, photographic screening does not encounter these obstacles and implementation of these programs has progressed further.

There is evidence that poor rates of screening for DR represent an active avoidance of patients who realize the importance of a dilated eye examination for the detection of a sight-threatening condition, yet choose not to comply.11 There is also evidence that a proportion of these patients actively prefer a nonmydriatic, teleophthalmic method for screening as an alternative to the recommended gold standard for detection of retinopathy.11 Thus, eye care professionals may need to accept that photographic screening is the only practical way to access a fraction of the diabetic population at risk for eye disease.21

14.15Combination Methods of Screening

Single-field mydriatic fundus photography interpreted by a retina specialist added to mydriatic ophthalmoscopy increased the sensitivity of general practitioners and ophthalmic opticians for detecting any DR.45 For general practitioners, the sensitivity increased from 55 to 71%. For ophthalmic opticians, the sensitivity increased from 73 to 88%.45 Similarly, adding a macular stereoscopic pair of fundus photographs to mydriatic direct ophthalmo-

scopy improved sensitivity for detecting DR from 65 to 93%.48

14.16 Case Yield Rates

Rates of referable retinopathy have been consistent

across many studies and locales, ranging from 4.4 to 5.1%.68,21,78 Although screening for DR is worth-

while in many venues, as adequate detection of retinopathy is widely lacking, some venues have higher yield rates for disease needing treatment. In a study from Scotland, rural screening yielded higher rates of detected advanced retinopathy than urban screening.68

14.17Compliance with Recommendation to Be Seen by an Ophthalmologist

Regardless of the success of screening for detection of referable retinopathy, if advice to be seen for examination is not heeded, the screening program has failed. Rates of compliance with a recommen-

dation to be seen for examination have ranged from 48.0 to 90.5%.27,64,71 In one study the rate depended on reason for referral.27,64,71 Patients referred for

PDR were more likely to comply than patients referred for NPDR, possibly because the risk of visual loss was higher with PDR and this was effectively communicated to the patient.27 Even when screening is free to the patient (although not to some alternate payor), efforts are frequently unsuccessful. In a demonstration project among medically underserved residents of rural Colorado, the screening rate increased from <10% per year to 37% over a 3-year period when free-to-the-patient screening was introduced.21 In a study from a prepaid Health Maintenance Organization in which

there was no additional cost to the patient for the ophthalmologic examination, the compliance rate was 74%.71 If the additional examination costs the patient, the compliance rate would be expected to be less.71 In comparing screening rates for diabetic retinopathy across managed care versus fee-for-ser- vice settings, the managed care setting did not outperform the fee-for-service setting as might have been expected based on its cost structure for patients.82 Factors other than simple cost of screening may explain this situation, including such problems as population transience, effectiveness of communication, and variable enthusiasm among different screening sites for the screening mission.

14.18Intravenous Fluorescein Angiography and Oral Fluorescein Angioscopy

Adding fluorescein angiography to fundus photography does not add sufficient sensitivity for the detection of early NPDR to make it worth the

expense.83 Oral fluorescein angioscopy is too time consuming and expensive to be practical for largescale screening purposes.36

14.19Automated Fundus Image Interpretation

A drawback of all operational photographic screening methods is the requirement of a human grader, who may range from a trained technician to a fully trained retina specialist. This adds to the cost of screening. Although computerized automated interpretation is a

research concept only, it promises to reduce the cost of screening in the future.84,85 One early prototype of

such a system accurately classified 14 of 22 (63.6%) images of PDR and 46 of 58 (79.3%) images of NPDR compared to a human grader taken as the gold standard.86 Another system based on an artificial neural network reading color fundus photographs achieved sensitivity and specificity of 88.4 and 83.5%, respectively, for DR detection compared to the results of an ophthalmologist’s reading of the photographs.85

14.20Subgroups Needing Enhanced Screening Efforts

Certain subgroups of patients with diabetes have lower rates of usage of eye care services. In a study of Medicare claims data, males, blacks, and persons living in areas with higher poverty and fewer ophthalmologists had lower rates of having an eye examination within the past year.16 State blindness prevention programs that target persons with diabetes, who have not had eye care previously, have reported that users of these programs differ from other diabetics by a higher probability of being female, on insulin, obese, and poor.27

14.21 Screening in Pregnancy

Pregnant patients with preexisting diabetes should have a dilated eye examination in the first trimester and close follow-up for the remainder of pregnancy if sight-threatening retinopathy is present. Women who develop gestational diabetes do not require such screening.35

14.22 Economic Considerations

The cost to society of DR is high and the cost-effective- ness of screening and treatment of DR has been firmly established.87 Economic simulations in the United States and the United Kingdom based on epidemiologic data from WESDR, treatment benefit data from the Diabetic Retinopathy Study (DRS) and ETDRS, and costs based on country-specific data show that screening is cost effective for both types 1 and 2 diabetes, but more so for the former because of the higher incidence of advanced retinopathy and the longer survival of type 1 diabetics after screening.35 Annual screening was most cost effective, and screening by photographic methods was more cost effective than screening by ophthalmoscopy.35 For the type 1 diabetics in the United States, the net savings was $3,300 per person screened. For the type 2 diabetics not taking insulin, annual screening with ophthalmoscopy added years of sight at a cost of $1,500 per year of sight saved.88

There is controversy, however, on which screening methodology and interval are preferable on economic grounds. The cost of photographic screening depends on the initial capital outlay for the camera, van, and other equipment; the salary expense of those performing the functions of the screening process; the cost of training image graders; and the ongoing cost of goods associated with screening, such as film, fuel, and disposables.21 Indirect costs associated with the decision to screen and the interval of screening include lost productivity that may result from degrees of visual impairment.6 When denominated per screening event and referable event, the efficiency and volume of cases screened will affect the results. Given all these variables, the reported economic costs of screening vary widely. In the United Kingdom, the cost per screening event has been reported to be £60.30 in a nursing home environment and £10–12.75 with a mobile van model among ambulatory patients.45,68 The cost per treatable event was £3,859 in the nursing home environment compared to £1,000 with the mobile van model.68 For the nursing home model, these costs were judged

to be unsustainable, but they were judged worthwhile in the mobile van model.42,68 Using a different

method of cost assessment, the cost per true positive by photoscreening was £209 in the Liverpool Diabetic Eye Study and the cost per laser-treated patient

was £1,000.32,68 The cost per case of blindness prevented was estimated to be £1,095.45 In Canada, the cost per screening event was 100 Canadian dollars using a mobile camera unit set up in local pharmacies.26 In the United States, the cost per case of retinopathy detected has been estimated at $295 for mydriatic photography and $390 for dilated eye examination by an ophthalmologist.89

Comparing costs of systematic photographic screening to the costs of the current model of opportunistic screening revealed that systematic photographic screening is more cost effective using cost per case detected as the outcome variable. Sensitivity analyses show that this is true for all practically achievable levels of sensitivity (sensitivity < 95%) for opportunistic screening. Systematic screening is more cost effective than opportunistic screening for all levels of compliance with systematic screening >54%. Although cost-effectiveness is better with systematic screening, the total cost is slightly higher with systematic screening, because it identifies more cases.

The single most important variable in the cost of screening is the cost of the screening examination, but the costs associated with the choice of screening interval currently are also important and have not factored into the choice of the recommended interval.7 As available economic resources increasingly constrain choices, this may change. When QALYs were used as the primary outcome variable of a modeling study based on a US population-based study of retinopathy prevalence, the marginal cost-effectiveness of various screening frequencies varied according to HbA1c levels at baseline. For the US population overall, the marginal cost-effective- ness of annual and biennial screening was $107,510 and $49,760, respectively.7 In a study from Taiwan which examined cost per sight year gained, the cost of screening type 2 diabetics annually was 20,962 New Taiwan dollars.83 Using this as a base, the comparative costs of screening every 2 years, 3 years, and not screening at all were 1.19, 1.47, and 3.98 times as much, respectively.83 Modeling studies depend on assumptions such as the value of a level of vision impairment in QALYs. The values used have ranged from 0.36 to 0.69 QALYs for blindness, for example, yet sensitivity analyses of the modeling results to these assumptions have not been consistently done. Other flaws in modeling studies include failure to stratify retinopathy progression rates by ethnicity, failure to allocate value to non-screening functions of the annual visit (e.g., detection of

co-morbidity and education), and using retinopathy progression rates applicable to newly diagnosed diabetics for all diabetics, which leads to an underestimate of progression.90 For these reasons, recommendations to depart from annual examinations have been met with resistance.90 Still, published value judgments on screening vary from unequivocal statements that the benefits of screening outweigh the costs to statements that costs

may be too high for certain currently employed annual screening strategies.6,7,65,90,91 Actually implemented

screening programs vary from annual to biennial.78 Unlike the situation regarding cost–benefit analyses, there is unanimity that screening yields medical benefits.

Screening yields patients who must be seen for management, which implies an increased workload for eye care providers as a result of screening. On the other hand, diagnosis of DR early may reduce the work required per person diagnosed, thus potentially decreasing ophthalmic workloads. When actual impact of a screening program on ophthalmic workload has been studied, the workload for eye examinations but not laser treatment increases.78 The increase in number of patients to be seen reflects the increase in population and the increase in type 2 diabetes prevalence in recent years together with the epidemic of obesity.78

A separate economic consideration in screening relates to physician efforts at implementing screening guidelines. An unknown fraction of noncompliance in screening may relate to lack of efforts by primary care physicians to educate patients to have annual eye examinations. For example, McCarty et al. found that preferred practice guidelines had been success-

fully disseminated to physicians but had not made a difference in management practices.39,92 Audit bench-

marks with linked financial rewards or penalties related to performance have been instituted in some health-care systems to address this problem.93 Communication about eye examination findings from eye care professionals to primary care physicians may be lacking as well, but has been little examined.

14.23Comparisons of the Screening Methods

The contemporary ideal in screening in the United States continues to be the dilated eye examination by a professional competent to recognize and treat

or appropriately refer for treatment DR.8,55,94 Except in Iceland, this ideal seems currently beyond reach.26 As the burden of DR is projected to outstrip the growth in ophthalmic manpower to perform examinations, more pragmatic methods of screening must be evaluated. In the United Kingdom, screening by ophthalmologists has been deemed impractical. Instead, screening by optometrists and systematic photographic screening have been judged acceptable.95

Modeling studies have been done in which the sensitivity and specificity of different screening methods and various screening intervals have been tested for their effects on outcomes such as years of

sight saved, eye clinic workload produced, and economic cost savings.46,96 All the screening methods

examined (screening by general practitioner, endocrinologist, optometrist, and ophthalmologist) saved sight compared to a strategy of no screening, but there was little difference between the four strategies of screening in effects on outcomes (Fig. 14.1). With screening by general practitioners on a yearly basis, 3.5 years of sight were saved on average compared to 4.1 years of sight saved with screening by ophthalmologists.46 An independent study in type 1 diabetics likewise concluded that

the cost savings generated from screening vary little over the range of sensitivities of the competing methods.96 Thus, the evidence suggests little imperative to mandating ophthalmologic screening rather than other forms of screening for DR. A more cost-effective approach that makes fewer demands on scarce ophthalmologic manpower and produces roughly equal outcomes is to allocate lower cost screening methods with slightly lower sensitivity for universal screening with use of ophthalmologists to manage cases where DR is detected.

When all of the performance characteristics of the various screening methods are compared, the

trend in recommendations for a practical method is toward digital photographic screening.10,50,95 It

appears to be more sensitive and less costly than screening by examination, reduces present inequal-

ities in service provision, and unlike the latter, produces a permanent record.10,95 Because it is not

widely available in the United States, reliance on examination-based screening methods is necessary there for now, but acceptance of photographic screening seems to be growing.10

Noncompliance is the major flaw in all methods of screening.10 Suggestions for overcoming

Fig. 14.1 Plot of the average number of patients out of a cohort of 1,000 patients who would develop severe vision loss from proliferative diabetic retinopathy or central visual loss from diabetic macular edema in age groups by decade by screening strategy. The diamonds represent outcomes with no screening, the circles represent screening by general

practitioners every 2 years until retinopathy is detected with annual screening thereafter until referral to an ophthalmologist for treatment, and the triangles represent screening by ophthalmologists every 2 years until treatment is detected with annual screening thereafter. Reproduced with permission from Davies46