Diagnostic Concordance in Tertiary (Dermatologists-to-Experts) Teledermoscopy: A Final Diagnosis-Based Study on 290 Cases

Background Teledermoscopy (TDS) improves diagnostic accuracy and decreases the number of unnecessary consultations. Objectives To determine the diagnostic concordance in tertiary (dermatologist-to-experts) TDS with histopathology/follow-up–based diagnosis. Methods A descriptive retrospective cohort study including 290 requests. Results Perfect diagnostic concordance was found in 202 (69.7%) cases and partial agreement in 29 (10%). Disagreement was found in 59 (20.3%) cases. Perfect concordance on the benign/malignant nature of the lesion was found in 227 (78.3%) cases and disagreement in 63 (21.7%). In onychology, diagnostic concordance was perfect in 43 (76.8%) cases, partial in 7 (12.5%), and there was disagreement in 6 (10.7%). Final concordance on the benign/malignant nature of the lesion was perfect in 48 (85.7%) and there was disagreement in 8 (14.3%) nail cases. For pediatric requests, diagnostic concordance was perfect in 29 (65.9%) cases, partial in 5 (11.4%), and there was disagreement in 10 (22.7%). Final concordance on the benign/malignant nature of the lesion was observed in 34 (77.3%) cases, disagreement in 10 (22.7%). Conclusions This study confirms that tertiary TDS improves diagnostic accuracy of pigmented skin lesions. Moreover, it shows encouraging results in unusual conditions such as ungual and pediatric skin tumors. The main limitation was the retrospective nature and the “real-life” setting of our study that could have created a selection bias toward inclusion of the most difficult cases.

management. We restricted this study to dermoscopy containing pictures requests made by dermatologists or skin cancer hyperspecialized GPs. In cases of multiple TD consultations for the same lesion, only the first was considered, whereas in requests for multiple lesions, each lesion was analyzed as a single statistical event. In this "real-life" study, no standardization of picturing mode was used. These pictures were analyzed by 1 among the 7 experts (at least 10 years of practice of dermoscopy and 10 publications on the field of dermoscopy) in TDS in the department; response was sent to the referring clinician and saved for analysis. Experts responded about diagnosis, possible differential diagnoses, the benign/malignant nature of the lesion, and management (excision, follow-up, biopsy, confocal microscopy, or picturing).

Case Revisions
All pictures and answers have been retrospectively analyzed The gold standard was either histopathology or reasonable-delay (at least 1 year) benign evolution in all cases.
Indeed, the level of evidence is weaker in the second case.
However, systematic surgical excision of all cases would have been unethical. Moreover, surgical recommendation for benign conditions to an expert's eyes would have resulted in greater bias since our series could not then be considered a "real-life" one. Cases with neither histopathological nor follow-up information were excluded.
The large number of pediatric and onychology cases allowed us to perform a subgroup analysis.

Definitions of Concordance/Disagreement
The definitions of concordance and disagreement are indicated in Table 1.

Introduction
Diagnosis of skin cancer is challenging for solitary practicing dermatologists. Teledermatology (TD) exists in 2 modes: store-and-forward and live interactive [1]; in store-andforward, the most used, information is sent to an electronic platform for delayed analysis. In 2 systematic reviews, TD was found inferior to face-to-face (FTF) dermatology but the accuracy was deemed acceptable by the authors [2,3]. However, Coates et al pointed out some limitations of TD such as the lack of total-body skin examinations [4].
Teledermatology is classified as primary, secondary, and tertiary TD. Primary TD involves communication between patients and a general practitioner (GP). In secondary TD, GPs communicate with dermatologists. In tertiary TD, dermatologists receive an expert opinion [1]. According to Finnane et al, the main limitation of all TD studies published since 2009 was that tele-expert diagnosis was compared to that of a primary physician, not to the final histopathology/follow-up-based diagnosis [2]. Teledermoscopy (TDS) is based on transmission of a dermoscopy picture.
TDS is known to improve diagnostic accuracy and decrease the rate of unnecessary consultations in dermatology compared with TD alone [5][6][7], yet most of the published studies were performed in a secondary telemedicine setting.
By contrast, the aim of this study, performed in our unit dedicated to private practice dermatologists with special extra-competence in difficult-to-diagnose skin lesions encompassing many digital dermoscopy follow-ups, nail tumors, and pediatric lesions, was to (1) determine the final diagnostic concordance between the diagnosis made by the tele-expert and the final diagnosis and (2) evaluate the efficiency of tertiary TDS.

Patients
We conducted an unselected consecutive cohort study between January 1, 2016, andDecember 31, 2016. Referring clinicians sent TD requests on an encrypted, firewall-protected storeand-forward server of the Hospices Civils de Lyon (https:// myhclpro.sante-ra.fr/). Clinicians provided age, sex, location, personal and family medical history, and macroscopic and dermoscopic pictures to experts and questions on diagnosis and Conclusions: This study confirms that tertiary TDS improves diagnostic accuracy of pigmented skin lesions. Moreover, it shows encouraging results in unusual conditions such as ungual and pediatric skin tumors. The main limitation was the retrospective nature and the "real-life" setting of our study that could have created a selection bias toward inclusion of the most difficult cases.

Requests
Requests included a median number of 2 (range 1-6) dermoscopy pictures in a total number of 3 (range 1-24). In 32 cases (11%) we did not receive accompanying close-up or wide-angle standard pictures. These standard images were variables; and percentages and effectives for discontinuous variables. Variables were compared with chi-square or Fischer exact test when necessary.
The Ethical Committee of Lyons University Hospital approved the study protocol on May 17, 2018. Figure 1 represents the flow chart of the study. Two hundred ninety teledermoscopic requests with known final diagnosis were included on a total of 2,528 tertiary TD requests sent between January 1, 2016, and December 31, 2016. One hundred seventy-seven (61%) patients were female and 113

Main Diagnoses
All diagnoses are reported in

Expert Answers
Experts submitted their answers in a mean time of 2.21 days (median 1 day; range 1-14 days). In 11 ( Table 3. For example, the referring clinician and expert totally agreed on the diagnosis of the histopathology-confirmed melanoma shown in Figure 3. However, the referring clinician and expert disagreed on diagnosis and management of the lesion shown in Figure 4 but histopathology confirmed a dermatofibroma, as proposed by the expert. Benign/malignant concordance was found in 227 (78.3%) cases, discordance in 63 (21.7%) cases ( Figure 5). Prediagnostic benign/malignant concordance and management concordance results are presented in Tables 4 and 5, respectively.

Discussion
We report herein the first robust, final diagnosis-based, reallife concordance study in tertiary (specialists-to-experts) store-and-forward TDS.
Tertiary TD is used in order to seek expert opinion/second opinion, but also to obtain an expert FTF consultation (e-referral). It may also be used for resident training and continuous medical education of specialists [8]. TDS is a specialized   lesion is also suggested by a 45.2% prediagnostic benign/ malignant disagreement. Moreover, analysis of diagnostic disagreements between expert and referring clinician showed that the expert was correct in the majority of cases (72.7%).
Our results in tertiary TDS are similar to previously published results in secondary settings (51%-94% for diagnostic accuracy between TD and histopathology [excised lesion]) and FTF diagnosis (nonexcised lesions) when dermoscopy is performed [2,[9][10][11][12][13][14][15][16][17]. Literature data combined with ours suggest an interesting improvement of diagnosis by transmission of dermoscopy pictures in cases of doubtful PSL. However, one author reported that the addition of TDS did not significantly improve diagnostic accuracy compared to transmission of standard pictures alone for malignant PSL [18] and suggested that TDS was useful only for malignant amelanotic skin tumors (aggregated accuracy, P = 0.0017; primary accuracy, P = 0.0382) [19]. This discrepancy might be explained by the tertiary setting of our study, referred cases being found difficult-to-diagnose by dermatology specialists concerned less with basal cell carcinomas and more with unusual (acral, pediatric) pigmented skin tumors.
The only available report on histopathology-based tertiary TD, including 33 cases, was centered on inflammatory skin diseases; it also demonstrated a high (78.8%) level of concordance [20]. Van der Heijden et al showed that in 81% of cases, dermatologists would have referred the patient to a tertiary center without the help of TD [21]. Our study also approach within TD known to improve diagnostic accuracy and to decrease the rate of unnecessary consultations in dermatology compared with TD without dermoscopy [5][6][7], yet all published studies to date were performed in a secondary telemedicine setting.
In most published TD studies, the main methodological limitation was the absence of correlation study with final diagnosis established either on histopathology or follow-up [2]. Moreover, no secondary-setting TDS published studies included follow-up information for nonexcised lesions [2]. In our study, the gold standard was histopathology in 57.6% of the cases or follow-up in 42.4%. We report a high (79.7%) diagnostic concordance between TDS experts and final diagnosis except for diagnosis of AT, for which the diagnostic discordance was significantly higher (51%) than for PSL (28.6%), NL (25%), and NP (   good-quality pictures [22]. Dermoscopy picture quality was good in 89.7% of cases in our study, similar to Massone et al (88%) [14]. In another secondary-setting TDS study, the quality of transmitted pictures was relatively lower, with only 36% good and 28% fair [12] and a lower accuracy and reliability compared with FTF consultation ( 0.66 on diagnosis and 0.42 on management), supporting the concept that TDS is highly dependent on picture quality [12].
Our study assessed the concordance between TDS and final diagnosis in pediatric skin tumors. Consultation offers in pediatric dermatology are scarce and are by far exceeded by the demand. TD is used to facilitate access to highly specialized pediatric dermatology opinion and to select cases for whom an FTF consultation is needed [23]. In the literature, diagnostic concordance between pediatricians and teledermatologists ranges from 16.7% to 82% [23][24][25][26]. These studies concern inflammatory skin diseases and do not include dermoscopy pictures. Diagnostic concordance was reasonably high (77.3%). Management concordance level between pediatricians and teledermatologists ranged from 25% to 44% vs 55% to 76% in adult patients. Our study concerned exclusively pediatric skin tumors, including dermoscopy pictures, and our results (59%) are very similar.
Our study encompassed a large number (19.3%) of nail teleconsultations. We found a high diagnostic concordance between TDS expert and final diagnosis in 50 (89.3%) and a high number of diagnostic (50%) and management (42.9%) changes after TDS, suggesting a high impact on patient's outcome.
The main limitation of our work was its cohort nature and the "real-life" setting evaluating our routine telemedicine platform with no previously agreed-upon way to report on outcome. Many cases (89.5%) were excluded because of unavailable histopathology/outcome information. This could have created a selection bias toward inclusion of the most difficult cases, the referring clinicians being less willing to retrieve patient information in case of indolent conditions or unchanged diagnosis and management after TDS.

Conclusions
In addition to previously published reports, this study confirms that tertiary TDS improves diagnostic accuracy of PSL