Telemedicine evaluation of cutaneous diseases: A blinded comparative study Jack L. Lesher, Jr., MD, a Loretta S. Davis, MD, a Frederick W. Gourdin, MD, a Debra English, MD, b and William O. Thompson, PhD c Augusta, Georgia Background: Numerous telemedicine programs have been created in the United States, but studies documenting the fidelity and effectiveness of telemedicine for evaluation of skin diseases are lacking. Objective: We attempted to determine the percentage of encounters in which two different dermatologists, one using telemedicine and one on-site, could independently arrive at the same primary diagnosis. Methods: Two clinical telemedicine sites linked through the Georgia Statewide Telemedicine Program were used in this study of 60 patients with skin problems. One dermatologist evaluated the patients on telemedicine (interactive television) and a second then took the patients into a separate examination room and evaluated them on-site. Each investigator recorded their diagnoses with no discussion with each other. As a control group, the investigators independently and in a blinded fashion (to each other's diagnoses) recorded diagnoses for a group of patients from a third dermatologist's clinic. Raw data were evaluated and classified by this third dermatologist who assigned diagnoses to categories of complete agreement, partial agreement, or disagreement. Results: There were no significant differences with regard to disagreement. However, there was a higher probability of complete agreement between the two dermatologists when each examined the patient on-site and in person than when one evaluated the patient on telemedicine and one examined the patient on-site and in person. Conclusion: Our results suggest that telemedicine is an effective means of diagnosing cutaneous diseases. However, because partial interobserver agreement on diagnoses was greater for the telemedicine group than for the control group (p < 0.05), it is likely that optimum use of medical assistants at the remote site will be necessary to increase the likelihood of complete agreement on diagnoses among dermatologists using interactive television. (J Am Acad Dermatol 1998;38:27-31.) From the Section of Dermatology, Department of Medicine, a the Department of Family Medicine, b and the Office of Biostatistics, c Medical College of Georgia. Accepted for publication Sept. 29, 1997. Supported by the Telemedicine Center of the Medical College of Georgia and the Georgia Statewide Telemedicine Program. Reprint requests: Jack L. Lesher, Jr., MD, Section of Dermatology, Medical College of Georgia, 1004 Chafee Ave., Augusta, GA 30912-3190. Copyright 1998 by the American Academy of Dermatology, Inc. 0190-9622/98/$5.00 + 0 16/1/86467 Telemedicine uses computers and telecommunications technology to provide medical information and services. This form of remote electronic clinical consultation is increasingly accepted in Georgia as a means for primary care physicians to obtain consultations. Telemedicine programs are used in more than 40 states in the United States. 1 There is debate whether this tool to provide specialty care to underserved areas represents more efficient use of existing medical resources, or a misallocation of scarce health care dollars. Studies documenting the fidelity and effectiveness of telemedicine in the evaluation of cutaneous diseases are lacking (see Addendum). In a previously published comparative unblinded observational study, one of us examined a group of 22 patients via telemedicine and then evaluated the patient in person within 3 days of the telemedicine evaluation. 2 The same diagnosis or differential diagnosis was obtained in 21 of the 22 patients. The purpose of the present study was to expand the pilot study to 60 patients to determine the percentage of encounters in which two different dermatologists (one using telemedicine and one on-site) could independently arrive at the same primary diagnosis. 27
28 Lesher et al. Journal of the American Academy of Dermatology January 1998 Table I. Investigators' diagnoses and agreement status for telemedicine/local cases 1 Dermatofibroma Dermatofibroma CA 2 Folliculitis Keratosis pilaris/folliculitis CA 3 Eczema Eczema/Lichen simplex chronicus CA 4 Tinea versicolor Tinea versicolor CA 5 Acne keloidalis Acne keloidalis CA 6 Postinflammatory hyperpigmentation Postinflammatory hyperpigmentation CA 7 Dermatofibroma Dermatofibroma CA 8 Actinic keratosis Actinic keratosis CA 9 Cutaneous horn Hypertrophic actinic keratosis CA 10 Actinic keratosis Actinic keratosis CA 11 Seborrheic dermatitis Seborrheic dermatitis CA 12 Furuncle Furuncle CA 13 Basal cell carcinoma Basal cell carcinoma CA 14 Actinic keratosis Actinic keratosis CA 15 Dermatofibroma Dermatofibroma CA 16 Nummular eczema Psoriasis PA 17 Bites Bites CA 18 Contact dermatitis/lichen simplex chronicus Eczema/Tinea PA 19 Tinea/Eczema Eczema/Tinea PA 20 Psoriasis Psoriasis CA 21 Urticaria Urticaria CA 22 Postinflammatory hyperpigmentation Postinflammatory hyperpigmentation CA 23 Atopic dermatitis Atopic dermatitis CA 24 Actinic keratoses/factitious Rosacea/Sarcoid D 25 Acne Acne CA 26 Stasis/Eczema Eczema PA 27 Vitiligo Vitiligo CA 28 Furunculosis/Hidradenitis Hidradenitis CA 29 Scabies Scabies CA 30 Seborrheic dermatitis Seborrheic dermatitis CA 31 Onychomycosis Onychomycosis CA 32 Postinflammatory hyperpigmentation Postinflammatory PA caused by lupus hyperpigmentation/photosensitivity 33 Eczema/Atopic Xerosis/Ichthyosis PA 34 Basal cell carcinoma Actinic keratosis PA 35 Basal cell carcinoma Hypertrophic actinic keratosis/ PA Basal cell carcinoma 36 Epidermal inclusion cyst Epidermal inclusion cyst CA 37 Bites/Factitious Excoriations/Bites CA 38 Tinea versicolor Tinea versicolor CA 39 Healed candidiasis Postinflammatory hyperpigmentation CA 40 Rosacea Rosacea CA 41 Acne Acne CA 42 Paronychia/Eczema Eczema/Tinea PA 43 Seborrheic keratosis Seborrheic keratosis CA 44 Punctate keratoderma Punctate keratoderma CA 45 Nummular eczema Eczema CA 46 Postinflammatory hyperpigmentation/lentigo Actinic keratosis/seborrheic PA keratosis/lentigo 47 Psoriasis/Eczema Eczema/Psoriasis PA 48 Basal cell carcinoma Sebaceous hyperplasia/ PA Basal cell carcinoma 49 Nevus Nevus CA 50 Telangiectasia Telangiectasia CA 51 Stucco keratosis Seborrheic keratosis CA 52 Acne Acne CA
Journal of the American Academy of Dermatology Volume 38, Number 1 Lesher et al. 29 Table I. Cont d 53 Tinea versicolor Tinea versicolor CA 54 Seborrheic dermatitis Seborrheic dermatitis CA 55 Bite Folliculitis/Bite CA 56 Seborrheic dermatitis/eczema Eczema CA 57 Scabies Scabies CA 58 Scabies Scabies CA 59 Acne Acne CA 60 Nummular Eczema Eczema CA 61 Basal cell carcinoma Basal cell carcinoma CA 62 Skin tags/seborrheic keratoses Seborrheic keratoses/skin tags CA 63 Dermatofibroma Dermatofibroma CA 64 Seborrheic keratoses Seborrheic keratoses CA 65 Scabies Xerosis/Scabies PA 66 Seborrheic dermatitis Seborrheic dermatitis CA 67 Hand eczema Hand eczema CA 68 Psoriasis Trauma/Psoriasis PA CA, Complete agreement; D, Disagreement; PA, Partial agreement. METHOD AND DESIGN The telemedicine sites used for this study included the Tri-County Health System in Warrenton, Georgia and the Medical College of Georgia in Augusta. These two sites are linked through the Georgia Statewide Telemedicine Program by means of a dedicated T-1 line (1.5 megabytes per second or the equivalent of 24 telephone channels); the full T-1 line is used except when communication between computers is established (i.e., for image capturing and annotation). The system uses a Video Telecom Mediamax CODEC digital communication video coder-decoder device. The video uses 30 frames per second. The device also provides full duplex audio. The evaluating physicians and patient are able to view one another on Sony Trinitron (PVM-2030) color monitors with a standard resolution of 560 TV lines. The cameras used included a single chip JVC TK-1280U remote-controlled room camera with a Computar TV zoom lens H1020812 MP with focal length of 8 to 80 mm, and a three-chip Panasonic WV-E550 remote-controlled patient camera with a Fujinon TV zoom lens S16 6.7 BMD-D24 with focal length of 6.7 to 107 mm. The amount of lighting in each telemedicine room is dictated by the size of the room, and the anticipated clinical uses of the room. The goal is to create a "daylight" atmosphere while eliminating any external light source from windows. This is necessary to ensure the integrity of the color balance set for each camera and the computer during system validation. A ceiling-mounted, daylight, 4-bulb fluorescent fixture is required. Close-up lighting is not generally used. Sixty patients at least 18 years of age with a skin problem were recruited randomly from among patients at the Tri-County Health System. Informed consent was obtained from each participant. One dermatologist (J. L. L.) evaluated patients via interactive television from the Medical College of Georgia site. Immediately afterward, a second dermatologist (L. S. D.) on-site in Warrenton evaluated the patient in person. Each investigator independently recorded their observations and diagnosis or differential diagnosis for each patient. After 30 patients were evaluated, the investigators switched locations. Case notes were not discussed through the data collection phase of the study. No written historical or demographic material was provided to investigators. The dermatologists were permitted to question patients randomly. In live encounters, palpation was permitted but no other methods for augmenting the examination were used. To provide a control group at the conclusion of the telemedicine portion of this study, the two investigators independently evaluated 36 on-site patients whom they had never seen before, again recording their primary diagnosis or differential diagnosis in a blinded fashion (i.e., with no discussion of findings). Again, palpation of lesions was permitted. For study and control groups there was no time limitation. Investigators made an effort to record time for each encounter, but this was not uniformly accomplished; therefore these data were not statistically evaluated. In the telemedicine and control groups, if a participant had more than one cutaneous problem or lesion that were clearly unrelated, these were treated as separate pieces of data, and diagnoses were recorded separately. The raw data were evaluated and classified by an unbiased, independent third dermatologist (F. W. G.),
30 Lesher et al. Journal of the American Academy of Dermatology January 1998 Table II. Investigators' diagnoses and agreement status for local/local cases 1 Bowen's disease Bowen's disease CA 2 Ruptured epidermal inclusion cyst Cyst/Epidermal inclusion cyst CA 3 Cellulitis/Stasis dermatitis Stasis dermatitis CA 4 Nummular eczema Nummular eczema CA 5 Actinic keratosis Actinic keratosis CA 6 Basal cell carcinoma vs squamous Basal cell carcinoma/actinic keratosis/ CA cell carcinoma Squamous cell carcinoma 7 Venous lake Venous lake CA 8 Tinea cruris Tinea cruris CA 9 Actinic keratosis Actinic keratosis CA 10 Seborrheic dermatitis Seborrheic dermatitis CA 11 Stasis dermatitis Stasis dermatitis CA 12 Seborrheic dermatitis Seborrheic dermatitis CA 13 Seborrheic keratosis Seborrheic keratosis CA 14 Basal cell carcinoma Basal cell carcinoma CA 15 Sebaceous hyperplasia Sebaceous hyperplasia CA 16 Tinea manuum Tinea manuum CA 17 Nummular eczema Lichen simplex chronicus/eczema CA 18 Basal cell carcinoma/irritated actinic keratosis Actinic keratosis/basal cell carcinoma PA 19 Onychomycosis Tinea unguium CA 20 Ichthyosis Ichthyosis CA 21 Basal cell carcinoma Basal cell carcinoma CA 22 Lichen planus Lichen planus CA 23 Contact/Eczema Eczema PA 24 Diabetic dermopathy Diabetic dermopathy CA 25 Hemangioma Venous lake/hemangioma CA 26 Folliculitis Folliculitis CA 27 Venous dermatitis Stasis dermatitis CA 28 Folliculitis Folliculitis CA 29 Postinflammatory hyperpigmentation Postinflammatory hyperpigmentation CA 30 Rosacea Rosacea CA 31 Nummular eczema Nummular eczema CA 32 Folliculitis/Acne Acne/Folliculitis CA 33 Epidermal inclusion cyst Epidermal inclusion cyst CA 34 Tinea Tinea CA 35 Tinea pedis Tinea pedis CA 36 Psoriasis Sebopsoriasis/Psoriasis CA 37 Paronychia Paronychia CA 38 Preauricular sinus Congenital anomaly CA 39 Vitiligo Vitiligo CA 40 Postinflammatory hyperpigmentation Postinflammatory hyperpigmentation CA 41 Actinic elastosis Solar elastosis CA 42 Nummular eczema Eczema CA 43 Bowen's disease/eczema Seborrheic keratosis/bowen's disease PA 44 Tinea versicolor Tinea versicolor CA 45 Ecthyma Traumatic ulcer CA 46 Irritant dermatitis Irritant dermatitis CA 47 Favre-Racouchot Milia CA CA, Complete agreement; PA, partial agreement. ential diagnosis for lesions but no definite agreement on a single primary diagnosis, diagnoses were placed in the "partial agreement" category; and (3) if there was no agreement or overlap in primary diagnoses or differential diagnoses, diagnoses were placed in "disagreement" category. There was only one case of diswho assigned the diagnoses to categories of "complete agreement," "partial agreement," or "disagreement" (Tables I and II). Criteria for assignment into these categories were as follows: (1) If the primary diagnosis was identical, diagnoses were placed in the "complete agreement" category; (2) if there was a similar differ-
Journal of the American Academy of Dermatology Volume 38, Number 1 Lesher et al. 31 agreement, so patient outcomes in cases of partial agreement or disagreement were combined before statistical analysis, testing the hypothesis that the probability of examiners arriving at the same diagnosis group would be equal regardless of type of examination. Fisher's exact test was used to determine statistical significance of the diagnosis outcomes, with significant differences defined as p < 0.05. RESULTS Diagnosis outcomes are listed in Table III. The visit labeled "telemedicine/local" represents the group examined by one dermatologist on interactive television and one dermatologist on-site, with a total of 68 cutaneous problems evaluated. The visit labeled "local/local" represents the control group, in which both dermatologists examined patients on-site with a total of 47 problems evaluated. If the assumption is made that diagnosis outcome does not depend on which type of visits enter into the comparison of diagnosis, then the results of our trial with ratios of 53:15 in the telemedicine group versus 44:3 in the control group (complete agreement: partial agreement or disagreement) have to be considered a rare outcome (Fisher's exact test, p < 0.05). More likely, the results of our trial indicate that the type of visit influences the probability of similar diagnosis outcomes. In our table of results, there is no significant difference in comparative visits with frank disagreement. Rather, the local/local comparison gives rise to a higher probability of complete agreement of diagnosis on the first visit (94% of diagnoses) than the telemedicine/local comparison (78% of diagnoses). With the telemedicine/ local comparison, there was a higher probability of partial agreement on diagnoses (21% of diagnoses) among examiners than with the local/local (control) group (6% of diagnoses). Of the 41 study group encounters for which investigators recorded evaluation time, telemedicine required a longer time for evaluation in 37 of 41 cases (90%). DISCUSSION Some early work with telemedicine in dermatology was published in 1972 and documented the usefulness of this technique. 3 Dermatology, as a visual specialty, lends itself Table III. Comparison of investigator agreement for telemedicine and control groups Complete Partial Type of visit agreement agreement Disagreement Telemedicine/ 53 (78%) 14 (21%) 1 (1.0%) Local Local/Local 44 (94%) 3 (6%) 0 (0.0%) well to use of telemedicine. To our knowledge, ours is the first study to compare, in a blinded fashion, telemedicine diagnoses to local, on-site (in person) diagnoses, also using a local/local control group for comparison. We found that examination of patients by both dermatologists in a local, in-person setting gave rise to a higher probability of complete agreement of diagnosis on the first visit compared with the telemedicine/local comparison group. A greater proportion of patients fell into the "partial agreement" category when examined with telemedicine; specifically, with telemedicine evaluation, 21% of the cutaneous problems or lesions fell into a partial agreement category, compared with 6% of lesions examined on-site by both investigators in the control group. There may be several explanations for this (e.g., the investigators believed there were instances when palpation enabled them to make a more secure primary diagnosis). However, the finding of complete agreement in 78% of the diagnoses comparing telemedicine with local, in-person diagnosis is encouraging. We suspect that this percentage could be increased with optimum use of medical assistants at the remote site. ADDENDUM: Since submission of our article, a teledermatology study appeared in the September 1997 issue of this Journal (Phillips CM, Burke WA, Shechter A, Stone D, Balch D, Gustke S. 1997;37:398-402) that also compared diagnosis by a live dermatologist versus a teledermatologist. The authors found a reasonable degree of agreement between the two examining physicians and observed that despite a high concordance rate on diagnosis, the teledermatologist had a lower degree of confidence in his diagnosis. REFERENCES 1. Perednia DA, Allen A. Telemedicine technology and clinical applications. JAMA 1995;273:483-7. 2. Warren FM, Lesher JL Jr., Hall JH Jr., Ward DF, Sanders JH, Tison J. Telemedicine. J Fam Pract 1995;41:17,20. 3. Murphy RLH, Fitzpatrick TB, Haynes HA, Bird KT. Accuracy of dermatologic diagnosis by television. Arch Dermatol 1972;105:833-5.