DIABETES TECHNOLOGY & THERAPEUTICSVolume 1, Number 4, 1999Mary Ann Liebert, Inc.

The Role of Digital Fundus Photography in DiabeticRetinopathy Screening

DANNY Y. LIN, M.D.,1 MARK S. BLUMENKRANZ, M.D.,1 ROSE BROTHERS, M.D.,2

and the Digital Diabetic Screening Group (DDSG)3*

ABSTRACT

Purpose: The role of digital monochromatic nonmydriatic fundus photography as an adjunctin the diagnosis of diabetic retinopathy is evaluated. Methods: 197 patients were sequentiallyevaluated by three different techniques: dilated ophthalmoscopy by an experienced ophthal-mologist, performance of 7 standard color mydriatic stereo fields, and a single digital mono-chromatic nonmydriatic image incorporating the disc and macula. Stereo color photographsserved as the reference standard and were compared to either ophthalmoscopy performed bya physician, or a single digital photograph transmitted electronically to a reading site andevaluated by a trained non-physician grader. Sensitivity and specificity of the three methodswere compared. The decision as to whether or not to refer to an ophthalmologist for poten-tial treatment (Kaiser modified ETDRS level . 21) was then chosen for analysis. Results: Asingle nonmydriatic monochromatic digital photograph appeared equivalent to standard colorphotography and more sensitive than mydriatic ophthalmoscopy in the detection of diabeticretinopathy in this patient population. Sensitivity of digital photography compared with colorphotography was 78%, and the specificity 86% constrasted with comparable ratios of 34% and100% for ophthalmoscopy versus color photography. No patient identified by opthalmoscopyalone for referral based on retinopathy level of . 21 would have been missed by a single dig-ital monochromatic photographic image. Conclusion: A single nonmydriatic monochromaticwide field digital photograph of the disc and macula in diabetic patients is a sensitive andcost-effective means for detecting diabetic retinopathy in high-risk populations.

477

THE SCOPE OF THE PROBLEM

W ITH AS MANY AS 11,000,000 DIABETICS esti-mated in the United States,1 diabetic

retinopathy remains the leading cause of visualimpairment and blindness in the working in-dividual today.2,3 Data from the Wisconsin Epi-demiologic Study of diabetic retinopathy

Review Article

1Department of Opthalmology, Stanford University Medical Center, Stanford, California.2Opthalmic Imaging Systems, Sacramento, California.3Kaiser Permanente, Northern California.*See the Acknowledgments for a complete list of DDSG participants.

(WESDR) suggests that between 47% and 59%of retinopathy-free insulin users will developsome evidence of retinopathy, and between 2%and 11% proliferative diabetic retinopathy over4 years. Twenty-five percent to 41% of thosewith preexisting retinopathy will worsen, indi-cating the critical need for retinopathy assess-ment in one form or other in a timely and on-going basis.4

The results of the Diabetic Control and Com-plications Trial (DCCT), Diabetic RetinopathyStudy (DRS), and Early Treatment DiabeticRetinopathy Study (ETDRS) clearly define thebenefits of tight metabolic control, as well astimely and accurate laser photocoagulation inthe slowing of visual loss in this condition.5–8

The initial report of the ETDRS demonstratedthat patients with clinically significant macu-lar edema receiving prompt focal photocoag-ulation were much less likely (12%) to lose 15or more letters on the ETDRS chart comparedwith those not receiving photocoagulation(24%) at a similar time point.8

Recent computerized modeling studies in-corporating the above data suggest that if ap-propriate eye screening and laser treatmentregimens are employed for patients with type2 diabetes mellitus, annual U.S. health care ex-penditures could be reduced by $247.9 to$472.1 million per year on an average of $975per patient. It is estimated this would result inbetween 53,986 and 94,340 person-years of ad-ditional sight. These numbers would be furtherincreased by inclusion of treatment of patientswith proliferative and other more severe formsof retinopathy, which were not included in thisanalysis.9,10

In addition to ethical and clinical issues re-lated to improved screening and treatment ofdiabetic retinopathy, patient governmental ini-tiatives, and other forces in the medical marketplace have increased the need for practice effi-ciency, quality assurance, and quality improve-ment regimens. Recently published HEDISguidelines promulgated by the National Com-mittee for Quality Assurance (NCQA) includediabetic retinopathy screening as a criticalmeasure for grading of quality of health caredelivered by managed health care organiza-tion.11

FUNDUS PHOTOGRAPHY AS ANALTERNATIVE TO PHYSICIAN

PERFORMED OPHTHALMOSCOPY

A number of studies have been performedconfirming the feasibility of fundus photogra-phy as an alternative to physician performedophthalmoscopy in the detection of diabeticretinopathy (Tables 1 and 2). Seven high-qual-ity stereo fields of the fundus taken by an ex-perienced, certified ophthalmic photographerare considered the gold standard for photo-graphic detection of diabetic retinopathy. Eval-uation of seven stereo fields by an experiencedand certified grader is acknowledged to becomparable to the sensitivity and specificity ofophthalmoscopy by an ophthalmic specialist inthe detection of diabetic retinopathy with85.7% agreement.12 However, it is not estab-lished that the cost and logistical concerns in-volved in obtaining seven stereo high qualityfields would make this a practical method forwidespread, cost-effective screening.13

On the other hand, fundus photography witha nonmydriatic camera with fewer fields andwider viewing-angle (45° vs. 30°) presents amuch more efficient and cost-effective screeningalternative. A large number of published reportsindicate the equivalence or superiority of modi-fied fundus photography performed by a trainedtechnician when compared with ophthal-moscopy performed by a physician or trained as-sistant.14–29 Reported sensitivities ranged fromapproximately 60%–90%, and specificities from85%–100% in published studies, comparable todilated fundus ophthalmoscopy by ophthalmol-ogists. Agreement between ophthalmoscopy and

LIN ET AL.478

TABLE 1. TYPES OF FUNDUS PHOTOGRAPHY USED

IN DIABETIC RETINOPATHY SCREENING

A. Conventional (film-based) color photography1. Seven standard 30° stereo fields (mydriatic;

slides)2. Less than seven standard or nonstandard fields

(30° or 45°)3. Nonmydriatic color transparencies4. Nonmydriatic color polaroids

B. Digital photography1. Stereo color images2. Nonmydriatic monochromatic images

fundus photog-raphy by various methods rangefrom a low of 61.9% to high of 86.3%. Associatedkappa values, nonparametric measurements ofassociation, range from a low of 0.38 to a high of0.74 (Table 3).

In a marketplace that demands cost-reduc-tion, practice efficiency, and effective diseasescreening, fundus photography offers manyadvantages over physician-performed ophthal-moscopy. However, the current use of a film-based photographic screening protocol is not

without flaws, because it is often complicatedwith problems such as delayed identification ofungradeable photographs, lack of qualifiedgraders, difficult archiving of photographs,and limited communication of images amongclinicians. With the advent of digital imagingtechnology, it is now possible to address theproblems inherent in the current film-basedscreening protocols and also to add many at-tractive features to the existing screening meth-ods (Table 4).

DIGITAL FUNDUS PHOTOGRAPHY 479

TABLE 2. COMPARISON OF OPHTHALMOSCOPY VS FUNDUS PHOTOGRAPHY FOR DIABETIC RETINOPATHY SCREEN ING

Ophthalmosc opy Fundus photography

Advantages Direct physician contact Improved quality assuranceEvaluation of far retinal periphery Improved data mangement and reportMaximal stereoscopic imaging generationComplex decision making Improved complianceEvaluation of other potentially Patient comfort (nonmydriatic)

treatable disease Lower manpower costPractice efficiencyArchiving

Disadvantages Practice inefficiency Higher initial equipment costHigh physician cost Reduced physician contactLack of archiving capability Potential technology problemsPatient compliance, scheduling, Storage format

examination cooperation Data transmissionLimited quality assurance Missed other diseasesPatient discomfort

TABLE 3. RESULTS OF PUBLISHED STUDIES COMPARING OPHTHALMOSCOPY VS FUNDUS PHOTOGRAPHY

AgreementRetinopathy between

Investigator Examination type prevalence Sensitivity Specificity methods kappa

Moss et al.12 MSSSF 51.9% 73% 95.1% 85.7% 0.749[1984] Ophthalmoscopy 44.9%n 5 2059 patients

Schachat et al.17 MTSSF 38.7% 76% 99.0% 75.7% NA[1993] (of diabeticn 5 1168 total population)

Patients Ophthalmoscopy 33.8%n 5 266 diabetic

PatientsLee et al.19 MOWNF 59.0% NA 86.3% 0.74n 5 795 eyes Ophthalmoscopy 63.0% No refn 5 410 patients standard

Pugh et al.23 MSSSF 51.1%[1993] MThWSF 39.2% 81% 97.0% 76.7% 0.74n 5 352 patients NOWNF 30.4% 61% 85.0% 61.9% 0.62

Ophthalmoscopy 22.4% 33% 99.0% 55.7% 0.38

MSSF, mydriatic seven stereo standard field (30°); MTSSF, mydriatic two standard stereo fields (30°); MOWNF,mydriatic one wide-field nonstereo field (45°); NOWNF, nonmydriatic one wide-field nonstereo field (45°); MThWSF,mydriatic three wide-field stereo fields (45°).

FUNDUS PHOTOGRAPHY:CONVENTIONAL (FILM-BASED) vs.

DIGITAL PHOTOGRAPHY

Limited data are available on the compari-son between digital photography and conven-tional (film-based) color fundus photographyfor the detection of diabetic retinopathy (Table5). A single study30 compared color trans-

parencies with color digital images both shoton a nonmydriatic 45° camera with pupillarydilation on 75 eyes of 40 patients. The authorsconcluded that color digital images viewed ona monitor were equally sensitive to color trans-parencies with regard to both any degree ofretinopathy, as well as that they termed sight-threatening diabetic retinopathy, which corre-sponds roughly to the description of clinically

LIN ET AL.480

TABLE 5. RESULTS OF PUBLISH ED STUDIES COMPARING CONVENTIONAL (FILM -BASED ) VS DIG ITAL PHOTOGRAPHY FOR DIABETIC RETINOPATHY SCREENING

AgreementRetinopathy between

Investigator Examination type prevalence Sensitivity Specificity methods kappa

George et al.30 MTWNF standard 100.0%[1998] Color transparency 93.0% NA NA 93% 0.92n 5 40 patients Color digital 93.0%75 eyes withpreexistingRetinopathy)

Kerr et al.32 Color transparency[1998] standardn 5 142 patients Ophthalmoscopy 62% 84% 0.62

by optometristDigital 90% 97% 0.90

DDSG** MSSSF 58.4%Abstracted standard

with permission Ophthalmoscopy 25.4% 34% 100% 55% 0.26n 5 197 patients Monochrome 57.4% 78% 86% 65% 0.49(population- digitalbased survey) NOWNF

MTWNF, mydriatic two wide nonstereo fields (45°); MSSSF, mydriatic seven stereo standard field (30°); MOWNF,mydriatic one wide-field nonstereo field (45°); NOWNF, nonmydriatric/one wide-field nonstereo field (45°).

TABLE 4. COMPARISON OF CONVENTIONAL (FILM -BASED ) VS DIGITAL

PHOTOGRAPHY FOR DIABETIC RETINOPATHY SCREENING

Ophthalmoscopy Fundus photography

Advantages Gold standard Requires less skilled photographerLarge existing database Session faster and more comfortableHighest resolution for patientsBest color Lower and fewer flashesLower equipment cost Efficient archiving

Less reduplicationGrading more efficientImage enhancementLow consumable costsEasier report generation

Disadvantages Requires skilled photographer Higher equipment costPhotography session lengthy and more Lower resolution than Kodachrome

uncomfortable for patient Reduced stereo capabilityHigh consumable costs (film, Smaller existing database

transparency, slides) Potential storage and transmissionArchiving inefficient incompatibilitiesPotential data loss (loss of slides

and copying)Grading time consuming

significant macular edema by ETDRS criteria.There was exact agreement between the twomethods in 92.3% of patients and a kappa of0.92. In a related study31 the authors concludedthat digitally acquired images viewed on amonitor were superior to, although more ex-pensive than images acquired on the same non-mydriatic wide-angle fundus camera capturedon Polaroid film. A most recent study (n 5 142patients) reported a remarkable sensitivity of90%, specificity of 97% and kappa of 0.90 whendigital imaging was compared to conventional35-mm color transparencies.32

A DIGITAL DIABETIC SCREENINGGROUP STUDY EVALUATING

OPHTHALMOSCOPY, CONVENTIONALPHOTOGRAPHY AND DIGITALPHOTOGRAPHY WITH REMOTE

INTERPRETATION

In the current Digital Diabetic ScreeningGroup (DDSG) study (unpublished data ab-stracted with permission), all subjects under-went three different fundus examinationmethodologies: (1) dilated ophthalmoscopy byphysician; (2) seven-standard field photogra-phy (the gold standard); and (3) single-field,45°, nonmydriatic, nonstereoscopic, monochro-matic, digital photography incorporating discand macula. The digital images were transmit-ted to a remote reading center via a telephoneline to be reviewed by trained graders. The re-sults from the three screening protocols werecompared to each other for their sensitivity,specificity and percentage of exact agreement.

In this study population of 197 patients, oph-thalmoscopy by a physician detected 25.4% ofpatients as having any degree of diabeticretinopathy and 12.8% appropriate for referraland possible treatment ( . ETDRS level 21). Incomparison, a single nonmydriatic digitalmonochromatic image acquired by a technicianand read by a trained and experienced graderdetected any retinopathy in 57.4% of the popu-lation at risk, with severity sufficient for warrantreferral in 29.4%. This was roughly equal to theprevalence rate of 58.4% for any retinopathy,and 36.5% appropriate for referral as deter-mined by the gold standard, seven mydriaticstandard stereo fields. The specificity, sensitiv-ity, and exact agreement rates for ophthal-moscopy were 34%, 100% and 55% respectively,contrasted with comparable rates of 78%, 86%,and 65% for a single, nonmydriatic monochro-matic, nonstereo image using seven stereo fieldsas a comparative gold standard (Table 6A).These results demonstrate the potential utility ofdigital photography protocol compared withophthalmoscopy, and the near-equivalence toseven standard field photography.

None of 26 patients (0%) identified by oph-thalmoscopy alone for referral based onretinopathy level of greater than 21 would havebeen missed by a single digital monochromaticphotographic image. Sixteen of 72 patients(22%) identified for referral by color photogra-phy would have been missed by a single digi-tal monochromatic image. Eighteen of 74 pa-tients (24%) identified by a single digitalmonochromatic image would have beenmissed by color photography (and may reflecteither false digital positives or, conversely,

DIGITAL FUNDUS PHOTOGRAPHY 481

TABLE 6A. DIGITAL DIABETIC SCREEN ING GROUP (DDSG) STUDY: SENSIT IVITY AND SPECIFICITY VALUES

FOR DIG ITAL DIABET IC RET INOPATHY SCREENING : DECISION TO REFER FOR FURTHER EVALUATION*

ExactComparison Sensitivity Specificity Undercalls Overcalls agreement

Ophthalmoscopy vs. color 34% 100% 45% 0% 55%n 5 192 (MSSSF)

Digital vs color 78% 86% 15% 12% 65%n 5 192(NOWNF vs. MSSSF)

Digital vs. Ophthalmoscopy 100% 69% 0% 43%** 49%n 5 197 (NOWNF)

Gold standard of mydriatic standard stereo fields.*DDSG study (abstracted with permission).**When adjudicated by color photography, digital overcalls primarily reflect reduced sensitivity of ophthalmoscopy.

false color negatives depending on whetherdigital, or conventional photography is used asthe gold standard. Thirty-six of 74 patients(48.6%) identified by digital photography asappropriate for referral would have beenmissed by ophthalmoscopy alone. When com-pared to ophthalmoscopy, the sensitivity ofdigital photography was 1.0 while the speci-ficity was 0.69 (Table 6B). This profile for ascreening protocol closely matches the sensi-tivity of 90% reported by Kerr et al.32

The DDSG study suggests that nonmydriaticwide field digital nonstereoscopic photogra-phy appears to be a sensitive and specificmeans for the detection of diabetic retinopathyin high-risk populations. The technique appearto be at least comparable to and possibly su-perior to ophthalmoscopy as performed in atypical outpatient setting in accurately gradinglower levels of diabetic retinopathy (ETDRSLevel 35 or less). The method of image acquisi-tion appears to be comfortable (nonmydriatic),safe and amenable to remote diagnostic inter-pretation by a trained nonphysician grader.

While a single, nonmydriatic digital, non-stereo wide angle image may be slightly lesssensitive than seven standardized stereo colorimages, the former approach appears to be rel-atively less costly, and possibly preferable forthe purposes of screening, assuming appropri-ate strategies can be employed to separate outhigh and low-risk populations. One such strat-

LIN ET AL.482

egy may involve the use of microaneurysmcounts of greater than 4 (abstracted from theDDSG study with permission) as a means forstratifying patients identified as lower risk bydigital means, who may be classified as higherrisk using a more extensive color field.

APPLICATION OF TELEMEDICINE TODIGITAL DIABETIC RETINOPATHY

SCREENING

The rationale and ultimate goal for digitalphotographic screening is the conversion of acomplex medical examination to a highly sen-sitive and specific laboratory test capable ofpredicting patients from within a given diseasepopulation at highest risk for vision loss. Thislaboratory test needs to be cost effective for thehealth care system, and readily accessible to theat-risk population. Digital imaging technologyand computer networking is now maturing tothe point where both goals can be easilyreached in diabetic retinopathy screening.

The costs associated with incorporating adigital approach into diabetic retinopathyscreening include: (1) equipment and facilitycost; (2) software and telecommunication costs;(3) networking access costs; and (4) profes-sional staff. Image acquisition hardware, in-cluding cameras and associated computerizedcomponents, is already available in many in-

TABLE 6B. DDSG STUDY: COMPARISON OF SCREEN ING NONMYDRIATIC DIG ITAL

MONOCHROMATIC PHOTOS TO DILATED FUNDUS OPHTHALMOSCOPY: DECISION TO REFER*

Digital Images

ETDRS Level , 20 ETDRS Level . 21No Referral Referral

Ophthalmoscopy: 123 54 177ETDRS Level , 20 (87.2%)

No ReferralOphthalmoscopy: 0 26 26

ETDRS Level . 21 (12.8%)Referral

123 (60.6%) 80 (39.4%) 203**

Sensitivity of digital Photography: 1.0.Specificity of Digital Photography: 0.69.Assuming gold standard to be ophthalmoscopy.*DDSG study (abstracted with permission).**Total of 203 eyes is greater than 197 patients due to some patients having more than

one eye eligible for referral.

stitutional settings as well as in ophthalmolo-gists’ offices and may be easily modified fortelemedicine communication for the purpose ofdata interpretation and report transmission.

Current software is available for image com-pression,33 enhancement, and transmission viaexisting telephone, ISDN, T-1, DSL, cable mo-dem, or ATM transmission with minor en-hancements. The application of high-speed

data transmission allows remote acquisition ofretinal images in primary care physician of-fices, endocrinology specialty clinics, commu-nity locations, and mobile units in addition toophthalmology offices. A recent study foundthat equipping a diabetic clinic specially forscreening with conventional photographywould not be more expensive than delegatingeye care to external parties.34 Remote image

DIGITAL FUNDUS PHOTOGRAPHY 483

A

B

FIG. 1. A: Microaneurysm only. In a red-free digital image, microaneurysms appear as dark gray, round dots withsmooth margins. Several microaneurysms are clearly seen in the inferior temporal quadrant of the original image.These monochrome images allow side-to-side comparison with fluorescein angiography images. B: The digitally en-hanced image increases the detection of subtle microaneurysms in the macular area and along the inferior temporalperiphery. ETDRS Level: 20.

acquisition should dramatically increase theproportion of diabetic patients screened whilereducing the need to schedule a separate oph-thalmic appointment. The lack of pharmacolog-ical dilation and lower illumination digital cap-ture imaging systems make this telemetricscreening protocol very user-friendly to the pa-tients, thus further improving patient compli-ance and convenience.

Manpower costs are relatively low for the useof a skilled ophthalmic technician and trainedimage readers on an hourly or yearly basis. Ad-

ditional savings can be realized once regionalreading centers are established to review im-ages submitted via the computer network.Given the digital format of the fundus images,computerized reading such as microaneurysmcounts using artificial neural network technol-ogy can be utilized to stratify patients into high-risk vs. low-risk populations.35,36 The benefitsof immediate patient feedback, instant identifi-cation of ungradeable images, electronic archiv-ing, report generation, and cross-referencing ofdigital images are self-evident, compared with

LIN ET AL.484

A

B

FIG. 2. A: Severe nonproliferative diabetic retinopathy. This monochrome image exhibits retinal hemorrhages, in-traretinal microvascular abnormalities (IRMA), venous beading and cotton wool spots (soft exudates). B: Digitally en-hanced image demonstrating enhanced diabetic retinal changes. ETDRS Level: 53.

the methods available for conventional pho-tographs. Additionally, the availability of im-age enhancement features within existing andfuture software, including both contrast en-hancement and digital sharpening filters, may

improve the gradeability of photographs witheither borderline or poor clarity due to mediaopacities or other technical factors (Figure 1A,1B, 2A, 2B, 3, 4).

All the basic tools to apply telemedicine to

DIGITAL FUNDUS PHOTOGRAPHY 485

FIG. 3. Enhanced severe proliferative diabetic retinopathy. This red-free digital image exhibits new vessels on theoptic disc and fibrous proliferation. ETDRS Level: 70.

FIG. 4. Normal fundus.

diabetic retinopathy screening are availablenow. Teleophthalmic screening studies bygroups such as the DDSG and Marcus et al.37

are beginning to show promising results andefficacy. Preliminary economic analysis is un-der review by the DDSG. Additional obstaclesto implementation of telemedicine applicationof digital diabetic screening will include avail-ability of reliable low-cost telecommunicationsoftware, clarification of third party reimburse-ment, resolution of legal/regulatory issues in-volved in image reading, and physician accep-tance of novel or alternative methods forpatient evaluation and management.

SUMMARY

In conclusion, numerous publications docu-ment that photographic diabetic retinopathyscreening methods meet or exceed the sensi-tivity and specificity of ophthalmologists per-forming ophthalmoscopy. The digital imagingtechnology adapts these protocols in place ofconventional, film-based photography in theophthalmic screening of patients with diabetesmellitus. When combined with telemedicineapplications, digital fundus imaging is a prac-tical, cost-effective adjunct to the ophthalmo-scopic evaluation of diabetic patients for thepurpose of determining the presence or ab-sence of retinal features associated with dia-betic retinopathy. Most importantly, this newtechnology will extend the sight-saving bene-fits of prompt ophthalmic attention to manymore patients with diabetes mellitus.

ACKNOWLEDGEMENTS

Members of the Digital Diabetic ScreeningGroup (DDSG) include: Stanley Azen, PhMD,University of Southern California; Mark S. Blu-menkranz, M.D., Stanford University; Rose-mary Brothers, Ophthalmic Imaging Systems;Rick Brown, M.D., Kaiser Permanente, Red-wood City; Danny Lin, M.D., Stanford Univer-sity; Barry Lindler, M.D., Kaiser Permanente,Oakland; Steven Verdooner, Ophthalmic Im-aging Systems. DIGITAL FUNDUS PHOTO-GRAPHS (nonmydriatic, single field, mono-chrome, 45° wide-angle images acquired by

the Canon CR6-NM nonmydriatic retinal cam-era; courtesy of Ophthalmic Imaging Systems,Sacramento, CA)

REFERENCES

1. Ruben R, Oltman W, Mendelson D: Healthcare ex-penditures for people with diabetes mellitus in 1992.J Clin Endocrinol Metab 1994:78:809a, 809f.

2. National Society to Prevent Blindness: OperationalResearch Dept. Vision Problems in the U.S.A.: Statis-tical Analysis. [New York]: The Society, 1980.

3. Klein R, Klein BEK: Vision disorders in diabetes. In:National Diabetes Data Group. Diabetes in America:Diabetes Data Compiled 1984. [Bethesda, MD]: USDept. of Health and Human Services, 1985; Chap. XIII.

4. Klein R, Klein BE, Moss SE, Davis MD, DeMets DL:The Wisconsin Epidemiologic Study of DiabeticRetinopathy. X. Four-year incidence and progressionof diabetic retinopathy when age at diagnosis is 30years or more. Arch Ophthalmol 1989;107:244– 249.

5. Diabetes Control and Complications Trial (DCCT): NEngl J Med Sept. 1993.

6. Diabetic Retinopathy Study Group: Photocoagulationtreatment for proliferative diabetic retinopathy, thesecond report of Diabetic Retinopathy Study Find-ings. Opthalmology 1978;85:82– 106.

7. Blankenship GW: Fifteen-year argon laser and xenonphotocoagulation results of Bascom Palmer Eye In-stitute’s patients participating in the Diabetic Retinopa-thy Study. Ophthalmology 1991;98:125– 128.

8. Early Treatment Diabetic Retinopathy Study Group.Photocoagulation for diabetic macular edema, ETDRSReport #1. Arch Ophthalmol 1985;103:1796– 1806.

9. Javvit JC, Aiello L, Chiang Y, Ferris F, Canner J,Greenfield: Preventive eye care in people with dia-betes as cost savings to the federal government, theimplications for healthcare reform. Diabetes Care1994;17:909– 917.

10. Javitt JC, Aiello L, Bassi OJ: Detecting and treatingretinopathy in patients with type I diabetes mellitus,savings associated with improved implementation ofcurrent guidelines. Ophthalmology 1991;98:1165–1173.

11. National Committee for Quality Assurance: HealthPlan and Employer Data Information Set. Version 3.0.Washington, D.C., National Committee for QualityAssurance, 1998.

12. Moss SE, Klein R, Kessler SD, Richie KA: Compari-son between ophthalmoscopy and fundus photogra-phy in determining severity of diabetic retinopathy.Ophthalmology 1985;92:62– 67.

13. Moss SE, Meuer SM, Klein R, Hubbard LD, BrothersRJ, Klein BE: Are seven standard photographic fieldsnecessary for classification of diabetic retinopathy?Invest Ophthalmol Vis Sci 1989;30:823– 828.

14. The Diabetes Control and Complications Trial Re-search Group: Color photography vs fluorescein an-

LIN ET AL.486

giography in the detection of diabetic retinopathy inthe Diabetes Control and Complications Trial. ArchOphthalmol 1987;105:1344– 1351.

15. Lewis JM, Jovanovich-Peterson L, Ahmadizadeh I,Bevier W, Peterson CM, Williams B: The Santa Bar-bara County Diabetic retinopathy screening feasibil-ity study: Significance of diabetes duration in systolicblood pressure. J Diabetes Complications 1994; Janu-ary: 51–54.

16. Kristinsson JK, Stefansson E, Jonasson F, et al: Sys-tematic screening for diabetic eye disease in insulindependent diabetes. Acta Ophthalmol 1994;72:72– 78.

17. Schachat A, Hyman L, Leske M: Comparison of dia-betic retinopathy detection by clinical examinationsand photographic gradings. Arch Ophthalmol 1993:111:1064– 1070.

18. Peters A, Davidson M, Ziel F: Cost effectivenessscreening for diabetic retinopathy using a non-my-driatic retinal camera in prepaid health care setting.Diabetes Care 1993;16:1193– 1195.

19. Lee VS, Kingsley RM, Lee ET, Lu M, Russell D, AsalNR, Bradford RH Jr, Wilkinson CP: The diagnosis ofdiabetic retinopathy: Ophthalmoscopy versus fundusphotography. Ophthalmology 1993;100:1504– 1512.

20. Heaven C, Cansfield J, Shaw K: The quality of pho-tographs produced by the non-mydriatic fundus cam-era in a screening program for diabetic retinopathy:A one year prospective study. Eye 1993;7:787– 790.

21. Jacobs J, Stad J, Sykes J, et al: A report on the use oftechnician ophthalmoscopy combined with the use ofthe Canon non-mydriatic camera in screening for di-abetic retinopathy in the community. Diabetes Med1995;12:410– 425.

22. Phillipov G, Alimat A: Screening for diabeticretinopathy: Ophthalmoscopy versus fundus photog-raphy. Ophthalmology 1993;100:1504– 1512.

23. Pugh J, Jacobson JM, Van Heuven WA, Watters JA,Tuley MR, Lairson DR, Lorimor RJ, Kapadia AS,Velez R: Screening for diabetic retinopathy: The wide-angle retinal camera. Diabetes Care 1993;16:889– 895.

24. Griffith S, Freeman WL, Shaw CJ, Mitchell WH, OldenCR, Figgs LD, Kinyoun JL, Underwood DL, Will JC:Screening for diabetic retinopathy in a clinical setting:A comparison of direct ophthalmoscopy by primarycare physicians with fundus photography. J FamPract 1993;37:49– 56.

25. Kalm H: Non-stereo photographic screening and longterm follow-up for detection of proliferative diabeticretinopathy. Acta Ophthalmol 1992;70:228– 234.

26. Sculpher MJ, Buxton MJ, Ferguson BA, SpiegelhalterDJ, Kirby AJ: Screening for diabetic retinopathy: A rel-ative cost-effectiveness analysis of alternative modal-ities and strategies. Health Econ 1992;1:39– 51.

27. Diamond JP, McKinnon M, Barry C, Geary D, McAl-lister IL, House P, Constable IJ: Non-mydriatic fun-dus photography: A viable alternative to fundoscopy

for identification of diabetic retinopathy in an abo-riginal population in rural Western Australia? AustN Z J Ophthalmol 1998;26:109– 115.

28. Hernaez-Ortega MC, Soto-Pedre E, Vazquez A,Gutierrez MA, Asua J: Study of the efficiency of a non-mydriatic retinal camera in the diagnosis of diabeticretinopathy. Rev Clin Esp 1998;198:194– 199.

29. Ryder RE, Close CF, Krentz AJ, Gray MD, Souten H,Taylor KG, Gibson JM, Kritzinger EE: A ‘fail-safe’screening programme for diabetic retinopathy. J RColl Physicians Lond 1998;32:134– 137.

30. George L, Halliwel M, Hill R: A comparison of digi-tal retinal images and 35 mm colour transparencies indetecting and grading diabetic retinopathy. DiabetMed 1998;15:250– 253.

31. Ryder RE, Kong N, Bates AS, Sim J, Welch J,Kritzinger EE: Instant electronic imaging systems aresuperior to Polaroid at detecting sight-threatening di-abetic retinopathy. Diabet Med 1998;15:254– 258.

32. Kerr D, Cavan DA, Jennings B, Dunnington C, GoldD, Crick M: Beyond retinal screening: digital imagingin the assessment and follow-up of patients with di-abetic retinopathy. Diabet Med 1998;15:878– 882.

33. Yogesan K, Constable IJ, Eikelboom RH, van SaarloosPP: Tele-ophthalmic screening using digital imagingdevices. Aust N Z J Ophthalmol 1998;26(supp1):S9–11.

34. Porta M, Rizzitiello A, Tomalino M, Trento M, PasseraP, Minonne A, Pomero F, Gamba S, Castellazzi R,Montanaro M, Sivieri R, Orsi R, Bondonio P, Moli-natti GM: Comparison of the cost-effectiveness ofthree approaches to screening for and treating sight-threatening diabetic retinopathy. Diabet Metab1999;25:44– 53.

35. Cree MJ, Olson JA, McHardy KC, Sharp PF, ForresterJV: A fully automated comparative microaneurysmdigital detection system. Eye 1997;11(pt 5):622–628.

36. Gardner GG, Keating D, Williamson TH, Elliot AT:Automatic detection of diabetic retinopathy using anartificial neural network: a screening tool. Br J Oph-thalmol 1996;80:940– 944.

37. Marcus DM, Brooks SE, Ulrich LD, Bassi FH, LairdM, Johnson M, Newman C: Telemedicine diagnosisof eye disorders by direct ophthalmoscopy. A pilotstudy. Ophthalmology 1998;105:1907– 1914.

Address reprint requests to:Mark S. Blumenkranz, M.D.Department of Ophthalmology

Stanford University Medical Center300 Pasteur Drive

Stanford, CA 94305

E-mail: mu.msb@forsythe.stanford.edu

DIGITAL FUNDUS PHOTOGRAPHY 487