Insights, Advantages, and Barriers of Teledermatology vs. Face-to-Face Dermatology for the Diagnosis and Follow-Up of Non-Melanoma Skin Cancer: A Systematic Review

Simple Summary This systematic review investigates the diagnostic concordance, advantages, and barriers of teledermatology in comparison to face-to-face dermatology for the diagnosis, management plan, and follow-up of non-melanoma skin cancer. Factors of increasing its sensitivity (teledermoscopy, quality of images, artificial intelligence, experience in generating clinical and teledermoscopy images) and its role as a tool for non-melanoma skin cancer triage for both underserved and high-risk populations are discussed. Abstract Background: Teledermatology is employed in the diagnosis and follow-up of skin cancer and its use was intensified during and after the COVID-19 pandemic. At the same time, demographic changes result in an overall increase in non-melanoma skin cancer and skin precancerous lesions. The aim of this study was to elucidate the role of teledermatology in comparison to conventional face-to-face dermatology for such lesions and determine the advantages and limitations of this workflow for patients and physicians. Methods: Research was performed using relevant keywords in MEDLINE and CENTRAL. Relevant articles were chosen following a predetermined standardized extraction form. Results: Diagnostic accuracy and interrater/intrarater agreement can be considered comparable—although lower—than in-person consultation. Improvement of particular features such as image quality, medical history availability, and teledermoscopy can further increase accuracy. Further aspects of limitations and advantages (mean time-to-assessment, time-to-treatment, cost-effectiveness) are discussed. Conclusions: Teledermatology has comparable diagnostic accuracy with face-to-face dermatology and can be utilized both for the effective triage of non-melanocytic epithelial tumors and precancerous lesions, as well as the follow-up. Easy access to dermatologic consultation with shorter mean times to diagnostic biopsy and/or treatment coupled with cost-effectiveness could compensate for the lower sensitivity of teledermatology and offer easier access to medical care to the affected populations.


Introduction
Teledermatology (TD) utilizes telecommunication means to allow the exchange of medical information for diagnosis, consultation, treatment, and teaching [1].The use of personal computers and smartphones allowed the unprecedented evolution of digital photography over the last decades and their implications also facilitated a revolution in telemedicine, especially during times when personal consultation was significantly restricted, such as during the COVID-19 pandemic.TD allowed the continuation of highquality medical care while protecting more vulnerable populations and medical staff and many countries adopted a more relaxed policy concerning the exchange of medical data [1,2].Dermatology has the advantage of being a mostly visual medical specialty and this unique characteristic makes it excellent for its implication in telemedicine [3].
Two forms of TD are currently used: the store-and-forward (SAF) technology, where clinical images are assessed asynchronously at different times and locations, and the synchronous method, where diagnosis and consultation take place simultaneously via video conferencing.The assessment can take place between physicians of different specialties (teleconsult) or directly with the patients (teleconsultation) [1].
The low density of dermatologists both in Europe and the USA in combination with their uneven distribution, the demographic alterations due to the aging populations, thus creating subpopulations with mobility restrictions but increased medical-including dermatological-needs is likely to lead to an expansion of TD use in the future in order to address relevant needs.Non-melanoma skin cancer (NMSC) is traditionally a diagnosis, which makes a high percentage of primary care providers uncomfortable with its management and might often result in direct referrals to a dermatologist [4].Furthermore, the rates of NMSC are on a continuous rise of 2-4% [5].Consequently, there is greater pressure on dermatologists for a timely and accurate diagnosis since early diagnosis might considerably improve patient prognosis.In this review, the barriers and advantages of TD were compared with conventional dermatologic care.

Materials and Methods
The initial search was performed on 2 October 2023 on the electronic databases CEN-TRAL and MEDLINE using a set of relevant search terms: "teledermatology", "diagnosis", "skin cancer", "basal cell carcinoma", "squamous cell carcinoma", "cutaneous lymphoma", "actinic keratosis", "fibroxanthoma", "dermatofibrosarcoma", and "Merkel cell carcinoma" in order to identify all articles that included the aforementioned terms in the title or abstract of the study.The exact term combination can be found in the Supplementary Material.Melanoma and other non-cutaneous epithelial tumors were omitted from the retrieved results.Randomized clinical trials, case series, retrospective studies, prospective comparative studies, cross-sectional studies, cohort studies, and letters to the editor that were published after 2000 and included a comparison of TD with conventional face-to-face dermatologic consultation (FTF) for NMSC were included, as well as studies that presented certain advantages or limitations of the one method versus the other.The abstracts were scanned for eligibility by two independent reviewers (G.N. and A.V.) using a standardized and predetermined data extraction form.In cases of discrepancy, a senior author (G.G. or C.C.Z.) would decide on the eligibility of the manuscript.Only manuscripts in English and German were considered eligible.Case reports and descriptive or systematic reviews were not included in the study, although the references of the latter were screened for eligibility.A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) chart, which summarizes the methodological approach, is demonstrated in Figure 1.From a total of 56 articles retrieved after extraction of duplicates, 33 articles were deemed eligible and were included in the study.The results were organized into subcategories according to the barrier, limitation, or future perspective discussed and are presented below.

Results
It is crucial for the TD-FTF comparison to set the gold standard for the diagnosis of an NMSC or a non-malignant skin lesion.For the former, the reference method is histopathology, while for the latter, FTF evaluation from a dermatologist or agreed evaluation of more than one dermatologist suffices [6].Interestingly, in the histopathological diagnosis of NMSC, a reported interrater disagreement in 2-7% of the cases is recorded, which is not taken into consideration when comparative studies between the two workflows are being assessed [7].

Teledermatology for Non-Melanoma Skin Cancer: Triage in Primary Care
One key benefit of TD is its potential as a cost-effective triage system in primary care, providing timely patient assessments (see Table 1).In a UK study, 58% of lesions were managed in primary care, saving GBP 12460 in unnecessary referrals over three years.Over 90% of patients were satisfied with the service [8].A retrospective analysis in Sao Paulo, Brazil, with 6633 individuals over 60 years and 12,770 skin lesions showed that TD reduced the waiting time for face-to-face dermatologist assessment from 6.7 to 1.5 months, with 66% of lesions referred back to primary centers [9].A UK comparative study revealed a significant reduction in waiting times for squamous cell carcinoma (SCC) and basal cell carcinoma (BCC) from 50 and 58 days to 28 and 35 days, respectively [10].Similar results were found in a US retrospective chart review, where TD decreased the time to surgery from 125 to 104 days, with initial consultations completed in 4 days compared to 48 days for face-to-face referrals [11].
An Austrian observational study focused on TD's effectiveness in diagnosing suspicious tumors preselected by general practitioners.Only 2% of lesions were non-melanocytic, and a low percentage of patients were referred for face-to-face evaluation [12].
Despite some controversial studies on efficacy, TD enhances accessibility to teledermatologists, particularly benefiting underserved populations.A retrospective cohort study showed that TD increased access to dermatologists from 11% to 44%, reducing the median waiting time to 28 days [13].Interestingly, 11% of the cases were diagnosed as NMSC.

Diagnostic Concordance of TD vs. FTF: Not Such a Simple Matter
There is a considerable number of studies that address the utility of TD in the diagnosis of skin cancer, including NMSC, and evaluate the respective diagnostic accuracy (see Table 2).A Swiss control trial randomized 30 of 78 general practitioners to an intervention, which was the TD assessment on digital images of skin lesions.Four categories were formed according to the feedback received [14], namely (1) no further investigation; (2) clinical observation; (3) biopsy; and (4) other.Diagnostic accuracy was assessed as the concordance with the 236 histological examinations, which were performed for the first three groups.One BCC was the only NMSC in the first category from TD, while no SCC, Bowen's disease, or actinic keratosis were categorized as lesions with no further steps needed.Notably, the percentage of overall lesions in group 1 was 21.1%.The authors argue that missed malignancies in group 1, where only 10/197 histology examinations were performed, might be higher and this raises questions about the sensitivity of the method.
A prospective study from Kroemer et al. [15] found strong intrarater agreement in 104 lesions, with TD showing high specificity for 56% of NMSC cases.Another Swedish prospective multicenter study observed increased reliability with the introduction of teledermoscopy for NMSC and confirmed that a TD approach significantly reduces the time required for surgery, while 40% of the cases could have avoided a direct dermatologist appointment, thus creating slots for other patients [16].
A Hungarian review assessed TD's diagnostic accuracy for skin cancer, achieving 85.3% concordance with FTF for primary diagnosis and 87.9% for aggregated diagnoses.Regarding the concordance of each lesion separately, the kappa coefficient was moderate for SCC (κ = 0.627), while it was higher for BCC (κ = 0.714) and actinic keratosis (κ = 0.739) [17].A US study compared dermatologist and teledermatologist diagnostic efficacy, revealing lower agreement for non-pigmented lesions (Kappa 0.32-0.86)[18].Lamel et al. [19] also confirmed low discordance (82% agreement) between TD and in-person dermatology, with actinic keratosis and BCC as common diagnoses.Another Brazilian study [20] focused on the diagnostic accuracy of TD in comparison to FTF dermatology and looked for the 10 most common ICD-10 diagnoses that teledermatologists have referred for a biopsy or an FTF consultation.The overall agreement with histopathology was 54%, with diagnosis of BCC showing complete agreement with histopathology, while SCC and AK showed moderate agreement.The concordance was higher if the assessment would only distinguish between malignant and benign lesions.

Some Help along the Way: Introducing Teledermoscopy (TDS) to Facilitate NMSC TD Diagnosis
TDS enhances the sensitivity of TD, using digital images taken regularly either after conventional dermoscopy from dermatologists, experienced nurses, or general practitioners.In some cases, images were made by using smartphone-adapted cameras, which allow digital self-dermoscopy (See Table 3).Zink et al. [21] compared the results of a clinical FTF examination including dermoscopy with TDS consultations including a TDS image, showing similar results for actinic keratosis and BCC, with a 92.3% agreement.The use of a derma smartphone microscope for TD diagnosis of non-melanocytic lesions had moderate interrater agreement, but weekly TD team meetings improved patient safety and reduced biopsy rates [22,23].A retrospective study of 59,729 primary care patients revealed that a TD workflow using a dermatoscope-fitted camera, image archiving, and retrieval on a large monitor had a 9% higher probability of cancer detection compared to face-to-face referral.However, it was more time-consuming [24].
In a prospective observational study, TD with app-assisted teledermatologic dermoscopy expedited treatment for NMSC (BCC, SCC, or SCC in situ), reducing the median time to diagnosis and treatment to 36 days compared to 85 days in the FTF group.A direct comparison of SAF TD with and without TDS followed by FTF evaluation showed a modest (68%) diagnostic concordance for skin cancer [25].
A randomized clinical trial in Spain [26] demonstrated that adding TDS to TD improved concordance with FTF consultation from 79.20% to 94.30%, making it a cost-effective strategy for routine skin cancer screening.Incorporating digital microscopy into TD increased diagnostic accuracy by over 10%, especially for BCCs and SCCs [27] with similar sensitivity and specificity to FTF dermatologic evaluation [28,29].However, studies suggested challenges in TDS's diagnostic concordance, with varying agreement levels for different dermatologists and lesions.Interrater agreement was adequate for BCC (κ = 0.55-0.67)but relatively poor for invasive SCC (κ = 0.05-0.15),actinic keratosis, and SCC in situ [30].A UK study on SAF teledermatology triage for melanoma and SCC with digital photography demonstrated a decrease in the median waiting time to clinic examination for SCC assessed with TD to 13.5 days compared to 24 days without photographs [31].

Teledermatology and Occupational Dermatology: Screening and Follow-Up of High-Risk Populations
TD was also applied in occupational dermatology [1].Experience has shown that outdoor workers make scarce use of skin cancer screening programs covered by health insurance funds (see Table 3).Moreover, follow-up visits can be time-consuming for elderly patients and dermatologists.TD might improve the financial burden of such follow-ups and increase the flexibility both for the patient and the physician, thus increasing adherence.
One of the most important lesions for occupational dermatology is actinic keratosis.A large single-center comparative study was conducted in Barcelona, Spain [32] comparing TD with TD including TDS and FTF consultation for lesions such as actinic keratosis and field cancerization.A total of 1000 lesions from 636 patients were assessed by primary care physicians, followed by TD assessment with or without TDS and FTF and subsequently FTF diagnosis.It was agreed that FTF diagnosis was considered the gold standard, and if consensus was lacking or an epithelial tumor was suspected, a histopathologic evaluation was performed.TD diagnostic concordance for actinic keratosis and field cancerization were high and superior to diagnosis made by primary care physicians (92.4% vs. 62.4% and 96.7% vs. 51.8%,p < 0.001).TDS has significantly further increased diagnostic concordance and identification of the specific actinic keratosis subtype.The kappa coefficient for intraobserver and interobserver agreement was over 0.83, indicating the importance of such technologies for the detection of early epithelial lesions as part of primary prevention for outdoor worker screening or screening of the elderly population, which are populations expected to develop such lesions [32].
Apart from outdoor workers, a high-risk population for developing NMSC are immunosuppressed patients, such as solid organ transplant recipients or patients undergoing hemodialysis.A retrospective review of organ transplant recipients demonstrated no significant disruption in dermatologic care with the implementation of TD before and during the pandemic, showing similar rates of new NMSC diagnoses during these two periods [33].

Image Quality and Digital Health Innovations as a Tool for TD Improvement for Diagnosis of NMSC
Since TD is based on visual images, it is expected that the quality and standardization of the images play a major role in the results provided by TD workflows (see Table 3).An interesting prospective comparative study [34] underlined the importance of photos taken by experienced, skilled personnel, such as dermatologists or specialized nurses, for the diagnostic accuracy of TD.When clinical pictures were taken from GPs and were forwarded to teledermatologists, the diagnostic accuracy was moderate to low with a kappa coefficient of TD and histopathology diagnosis of 0.41-0.63,while FTF agreement with histopathology diagnosis was 0.55-0.73.
Another UK study independently assessed the diagnostic accuracy of 163 SAF referrals suspicious for skin cancer through an experienced dermatologist and a three-year trainee dermatologist.Interestingly, the diagnosis was identical in less than 50% of the cases and the result was independent of the experience of the doctor.The authors mention the poor quality of images or poor technique of the images taken with inadequate lighting prior to generating such images and the lack of important details of the medical history, which suggests that technical aspects and medical history details included in the referral play an equal or more important role than the experience of the clinician [35].A retrospective comparative study from the United Kingdom [36] revealed that for 10% of the TD-assessed patients, no diagnosis was made.The authors suggested that a lack of diagnostic features, the possibility of malignancy, and technical factors might be responsible for this outcome.
In 2015, a pivotal pilot study on 20 volunteers showed that automated photography analysis based on color space transforms and morphological features achieved a modest correlation in comparison to live dermatologic consultation, with a correlation of 0.62 on the face and 0.51 on the arms [37].The assessment of lesions suspicious for NMSC showed high interrater agreement between a mobile prevention unit physician who assessed the lesions live and two oncologists who assessed the digital images made by the former, remotely.The observers should only distinguish between malignant and benign lesions [38].
Artificial intelligence in TD was also assessed as an assisting tool in primary care for the diagnosis of skin diseases [39].The data on skin lesions were assessed FTF in primary care and then images were taken and assessed by teledermatologists and an AI system, providing the five most probable diagnoses per image.The gold standard was the consensus of two dermatologists and if this was not the case, a third dermatologist was asked to evaluate the image separately.Mean sensitivity concerning malignant tumors was lower for the AI system both in comparison to GP and TD evaluation.Interestingly, this sensitivity was the second lowest after infectious diseases, suggesting that in the case of NMSC, AI sensitivity might not be equal to or higher than clinical evaluation, as is often reported in the literature.

Discussion
The effort of offering satisfying results on the diagnostic accuracy of TD in comparison to FTF evaluation for NMSC appears an arduous and complex procedure and the currently available data do not support solid conclusions.The available studies are particularly heterogeneous in what is established as "hit or a miss", whether it is a conventional clinical diagnosis after FTF dermatology consultation or histology.A recent Cochrane meta-analysis [40] showed that less than 7% of malignant skin lesions were missed by TD.For lesions considered malignant by TD, usually, histology confirms the final diagnosis and is a reliable method to confirm true positive results.In contrast, negative results were usually confirmed by FTF clinical diagnosis including dermoscopy.Although biopsies for presumptive benign lesions might be considered unethical, their lack affects the reliability of excluding skin cancer.Moreover, these lesions should be followed up prospectively, to reliably exclude the development of malignancy clinically.Once again, this was not clearly stated in many studies.
From the point of view of the patient and the dermatologist, TD and TDS should ideally be characterized by high sensitivity, so that malignant lesions of patients who are not seen via a personal clinical consultation will not be missed.In contrast, regulatory authorities, which fund primary and tertiary skin cancer prevention programs, concentrate on the specificity of the method.The ability of the method to accurately assess non-malignant cases effectively reduces FTF visits, thus reducing costs for the health system and increasing slots for other patients.The meta-analysis demonstrateda sensitivity of 93.5% and specificity of 95.8% for BCC, with a surprisingly lower sensitivity despite the implementation of dermoscopic images in one study, while the FTF approach demonstrated better sensitivity and specificity [15].For the cutaneous SCC, the sensitivity was variable, between 57-69% (decision for excision or follow-up) and 67-85% for the decision to refer a lesion.The current study confirms that newer studies and the diagnostic concordance with FTF dermatologic evaluation and histopathology remain moderate.Despite the developments and the digital innovations of recent years, which in some cases demonstrated an increase in diagnostic accuracy even in comparison to FTF, most recent studies confirmed a moderate or even low diagnostic concordance for NMSC [17,20,22,24,28] (primarily SCC, and secondarily for BCC, AK, and SCC in situ), raising concerns about the use of this workflow for this type of disease.
The visual assessment of skin lesions Is inherent to the dermatologic diagnostic process.For this reason, digital health innovations, including TD, not only facilitate the diagnostic procedure but can also be incorporated into other aspects of patient care, such as selfmonitoring of skin diseases, follow-up, reminders for the application of treatment, and a more patient-oriented engagement in health choices [41].The increasing incidence of skin cancer in our currently aging population and the presently available resource-consuming FTF screening strategies underscore the need to efficiently facilitate speedy diagnosis and/or referral procedures.TD is a rapidly growing field in this direction and its use demonstrates high patient satisfaction and confidence in the service [42], reduced waiting times, and diagnostic accuracy, which in many cases was shown to be comparable with the standard of care.The somehow moderate diagnostic concordance of TD/TDS with the reference standard might be explained by the fact that NMSC does not have the same well-established and widely accepted dermoscopic criteria for pigmented lesions.The uncertainty might lead to both misdiagnosis and easier FTF referral, in comparison to a melanocytic lesion, especially if an SCC is suspected.On the other hand, the available data show that this diagnostic inconsistency can be improved with simple measures such as the inclusion of targeted history details or the improvement of technical aspects in the acquisition of photographs.Likewise, one of the most important drawbacks in the use of TD for skin cancer screening or follow-up is the difficulty of performing a total body examination.A more lesion-based approach has similar detection rates to FTF dermatology consultation and is more time efficient.
Good clinical practice of SAF dermatology of such skin lesions includes information on the medical history of the lesion such as the site, the age of the patient, family history of skin cancer, estimated duration of the lesion, and symptoms such as bleeding, itching, and/or pain.Other risk factors such as the type of occupation, occupational or leisure time sun exposure, and immunosuppression can provide additional relevant information to support diagnostics accuracy.A concise and complete medical history form should always be an integral part of a TD/TDS consultation and to be stated in the methods section of the manuscript.
An efficient TD consultation for NMSC is shown to decrease waiting times for biopsy or diagnostic excision, offer a quick consultation for the population of rural areas, and cover the high demand for diagnosis and lesion-specific follow-up for non-melanocytic epithelial tumors and their precursors.Additionally, TD can be used to ameliorate the three most important barriers to accessing specialty dermatology care, i.e., being uninsured continuously, living in an underserved county, or being under the poverty level.For these cases, TD might be an affordable alternative for efficient NMSC screening [43].
This systematic review has limitations, primarily the difficulty of a direct head-to-head comparison of TD (with or without TDS) and FTF evaluation of NMSC because of the heterogeneity of the primary endpoints, the reference standard used, and the variability of the follow-ups, especially for cases recognized as "benign" and their documentation.Moreover, qualitative studies referring to skin cancer in general with no possibility to extract any data referring specifically to NMSC (at least as a category) might have been excluded.The databases searched for this study were MEDLINE and CENTRAL and there is a possibility that some articles in other databases (for example SCOPUS, EMBASE) could have been missed.The references of the articles retrieved were also searched, in order to expand the search and include articles not retrieved from the primary search.

Conclusions
TDS has comparable accuracy with TD for the assessment of NMSC lesions and can be a valuable tool for NMSC triage, diagnosis, and follow-up for specific lesions.Despite the controversies of a potential lower sensitivity to an FTF workflow, it is cost-effective, allows a significant rapid FTF consultation and treatment for suspicious lesions, and allows access to dermatology care for underserved and high-risk populations in a more comfortable setting.Digital innovations, tools, and technological advancements can reduce the limitations of its application and further improve the sensitivity of this workflow, thus improving diagnostic concordance with FTF consultation.

Figure 1 .
Figure 1.PRISMA flow chart of the study.Precancerous lesions such as Bowen's disease and actinic keratosis were included in those articles.

Table 1 .
Studies concerning the implementation of TD as triage in primary care.AK: actinic keratosis, FTF: face-to-face, GP: general practitioner, SAF: store and forward, TD: teledermatology.