Introduction

Teledermatology (TD) represents a significant advancement in telehealth, utilizing a store-and-forward model to remotely interpret dermatological data1. As many studies have shown, TD is a good alternative to in person consultations with a dermatologist this approach has proven to be an effective alternative to in-person consultations2,3,4. A variety of health care settings can be utilized the TD, including nursing homes and primary health centers in rural areas without access to a dermatologist. In addition, several studies reported that over two-thirds of medical residents incorporating mobile phones into clinical care. On the other hand, mobile TD has emerged as a vital solution in regions with limited internet access or where dermatologists are scarce5. Wherever access to computers with the internet is unreliable or insufficient, the mobile apps provide a feasible and reliable technology to increase access to dermatologists in poorer regions of the world6.

The implementation models of TD are categorized into two main types, the asynchronous or store-and-forward (S&F) and the synchronous or live interaction (LI). The S&F model is the most utilized method in dermatology due to the inherent characteristics of skin lesions images, and the lower administrative requirements in comparison to LI7,8,9.

Despite its widespread adoption, TD faces numerous challenges, including internet speed limitations, user satisfaction, the doctor-patient relationship, training requirements, diagnostic accuracy, privacy concerns, data confidentiality, reimbursement issues, and technical difficulties, with the latter being the most reported obstacle10.

Globally, the ratio of dermatologists to the population is alarmingly low, a situation that is even more dire in developing countries like Iran. Here, the scarcity of dermatologists, who are predominantly based in urban centers, exacerbates the challenge of providing specialist care. The burgeoning telecommunication technologies have revolutionized the management of skin disorders, prompting us to leverage TD to enhance dermatological access. This study aims to address this gap by developing a mobile app based on the store-and-forward model, tailored to the unique needs of Iran's patient population.

Materials and methods

From July 2022 to July 2023, we developed an Android-based mobile application utilizing the store-and-forward model to assess its efficacy in remote dermatological diagnoses. The study lasts one year for the obtaining the ethical approval, constructing the dataset, developing a mobile app, recruiting the patients and analyzing the data. The in-person diagnoses made by a seasoned dermatologist with 15 years of experience served as the gold standard.

Patient enrollment and data collection

A total of 89 patients were enrolled from a specialized dermatology clinic named Athar, as well as from Imam Khomeini hospital in Ahvaz, Iran. The inclusion criteria encompassed all patients regardless of age, sex, or type of skin disorder. For patients under 18, parental consent was obtained. To select patients for the study, a convenience sampling method was used. According to the inclusion criteria, patients were non-randomly selected until the required sample size was achieved. Using the sample size formula, with a significance level (α) of 0.05, an estimated proportion (p) of 87% for concordance, a confidence level (Z) of 1.96, a margin of error (d) of 0.07, and (q = 1−p), the desired sample size (n) was estimated to be 89 patients.

Standardized photographs of skin lesions (right, left, and oblique views) were captured using a smartphone camera. Alongside the images, demographic data, clinical history, and the gold standard diagnosis were recorded within the app.

$$n = \frac{{Z^{2}_{1} - \frac{\alpha }{2}pq}}{{d^{2} }}$$

Reliability assessment

To evaluate intra-observer reliability, the same dermatologist reviewed the photographs via the app two months later, recalling only two patients. In addition, to evaluate inter-observer reliability another two dermatologists, totally three dermatologists, assessed the recorded information through smartphone app.

App development

The Minimum Data Set (MDS) for the app was established through a structured questionnaire distributed among ten dermatologists and five medical informatics specialists. The app was developed using C# and the Xamarin platform and installed on the smartphones of the participating dermatologists (three dermatologists). A trained researcher captured lesion images using a Xiaomi Redmi Note 8T, 48/8/2/2 megapixel camera. The app’s language was Persian, which is the official and most wildly spoken language in Iran. However, the app can be translated to English using built-in feature that allows users to switch the language settings.

Ethical considerations

The study was conducted in accordance with the Declaration of Helsinki and approved by the Research Ethics Committee (IR.AJUMS.REC.1401.175) of Jundishapur University of Medical Sciences, Ahvaz, Iran. Informed consent was obtained from all participants, and parental consent was secured for minors. The study adhered to the principles of voluntariness, anonymity, and confidentiality.

Statistical analysis

Data analysis was performed using SPSS version 22. Continuous variables were presented as means ± SD, and categorical variables as frequencies and percentages. Chen’s kappa coefficient was used to report inter- and intra-observer reliability, with values interpreted as per Landis et al.'s concordance criteria: poor (ƙ ≤ 0.2), fair (ƙ = 0.21–0.4), good (ƙ = 0.41–0.6), very good (ƙ = 0.61–0.8), and excellent (ƙ ≥ 0.81)11.

Results

In our study, 89 patients and three dermatologists participated, with none of the dermatologists having prior experience in teledermatology. The gender distribution was 35% male (n = 31) and 65% female (n = 58). The majority of patients were aged between 6 and 60 years, with only 14% outside this age range (Table 1). Lesion distribution was predominantly on the whole body (47%), followed by the face (26%), hands (17%), legs (7%), and scalp (5%).

Table 1 Demographic characteristics of participants.
Full size table

Lesion distribution was predominantly on the whole body (47%), followed by the face (26%), hands (17%), legs (7%), and scalp (5%) (Fig. 1). The most common diagnoses were psoriasis (19.1%), acne vulgaris (11.2%), and eczema (10%) No cancer diagnoses were reported.

Fig. 1
figure 1

Pattern of skin diseases seen in the clinic (face-to-face) by dermatologist in 2022–2023: 89 cases.

Full size image

Table 2 shows the number and the frequency of diagnoses made by in-person dermatologist and three TD dermatologists. After two months, the clinical dermatologist reassessed the cases via the teledermatology app, recalling only two patients. Excluding these two cases, the intra-observer agreement remained excellent, with a kappa statistic of 0.887. An in-person diagnosis of eczema was made in nine patients. However on the app, this diagnosis was made in 14 patients. The opposite applies to psoriasis 17 versus 14. After two months, the clinical dermatologist reassessed the cases via the teledermatology app, recalling only two patients. Excluding these two cases, the intra-observer agreement remained excellent, with a kappa statistic of 0.887.

Table 2 The frequency and percentage of diagnoses made by in-person and on mobile by three dermatologists.
Full size table

The inter-observer reliability between the in-person diagnosis and teledermatology assessments was also high, with a kappa statistic of 0.865, indicating near-perfect agreement. (Table 3).

Table 3 The efficiency comparisons between in-person examination as the gold standard and teledermatolgy.
Full size table

Screenshots of the mobile app interface are presented in Fig. 2, illustrating the app's functionality and design.

Fig. 2
figure 2

Screenshots of designed mobile app. Left-above: The preview screen of clinical infrmation in the mobile app. Middle-above: a screen from the left view of a lesion. Right-above: the right view of lesion. Left-below: the mobile app’s screen designed for recording clinical information. Middle-below: the mobile apps screen designed for uploading photographs of three views from the lesion. Right-Below: A photograph taken by smartphone Xiaomi Redmi Note 8T, 48/8/2/2 megapixel camera.

Full size image

Discussion

This prospective study evaluated the concordance between teledermatology (TD) and in-person dermatological diagnoses among 89 cases. TD diagnoses were made by three dermatologists using a mobile app 2 months after in-person visits. The high agreement rates, ranging from 77.5 to 89.9%, underscore the potential of TD in clinical settings, particularly when in-person consultations are not feasible. It is worth mentioning that after 2 months, the clinical dermatologist who participated in the teledermatology group, which assessed the same patients, did not remember any patients except two of them. Skilled doctors in Iran have extremely high workload to do because there is lack of specialists. On average, they visit 30 patients daily, so it is difficult to recall. They usually remember patients who have very rare and specific diseases, but the patients in this study had common diagnoses. The dermatologist who did the in-person exam in this study was also skilled and visit at least 30 patients daily on average, so it is doubtful that he could remember the same patient after 2 months. This time interval for physicians with low workload may be short and can impact recalling the patients.

Compared to previous studies, our findings align with the reported kappa values, indicating excellent diagnostic agreement. However, variations in methodologies, such as the use of teledermoscopy and different imaging devices, make direct comparisons challenging. Despite these differences, the literature consistently suggests that TD is a reliable method for dermatological diagnosis. Two previous studies which implemented a mobile app reported the inter-observer kappa values of excellent (0.958)12, and good (0.6)13.

Similar to the current study, most research on teledermatology’s concordance has compared diagnoses made by dermatologists using the gold standard of in-person evaluations14,15. Previous surveys have yielded mixed results regarding the accuracy of teledermatology for pigmented lesions. Some studies have demonstrated good diagnostic concordance with in-person dermatology, with kappa values ranging from 0.4 to 0.92,13,15,16,17,18,19. In a study by Nami et al., the concordance between in-person and store-and-forward diagnoses was 91.05%, supporting the validity of the SAF model15. Another study reported a high concordance rate of 93% between teledermatology and in-person diagnoses20. A study from the USA found moderate concordance for non-infective disorders (66%) and low concordance for infective disorders (39%)21. A recent systematic review reported a pooled diagnostic agreement rate of 68.9% between teledermatologists and in-person dermatologists, with a kappa value of 0.6722. However, another systematic review of 78 studies found that approximately two-thirds reported better diagnostic accuracy with in-person dermatology23. These variations make the comparisons between the findings of this study and those of the former studies and literature difficult, although the existing literature suggests that TD is accurate. It should be noted that there were significant methodological variations in previous studies19,20,24,25.

In the current study, it seems that the mobile app over-diagnosed the eczema (9 in in-person vs 14 on app) and underdiagnosed the psoriasis (17 vs. 14). It seems that mobile app may have tendency toward confusing eczema with psoriasis. There are several factors that may explain this discrepancy, such as: the difficulty of distinguishing eczema from other skin conditions, especially psoriasis, based on images alone; the variability of eczema presentation and its severity among different patients; and the lack of a gold standard or a definitive test for eczema diagnosis.

However, it may also reflect the inverse relationship between eczema and psoriasis, which are both inflammatory skin diseases but have different immune mechanisms. Some studies have suggested that patients with eczema may have a lower risk of developing psoriasis, and vice versa. Therefore, it is possible that the app and the dermatologists may have misdiagnosed some cases of psoriasis as eczema, or vice versa.

Some studies use teledermoscopy for capturing the images, while we did not utilize it in the current study due to the fact that it is not used in real-world applications of TD and is not available in many primary health care settings. Indeed, we captured macroscopic images in the clinic using the high quality camera of a smartphone, similar to previous studies20,21,24,26 as compared to other studies which utilized a digital camera2,16,17,19,27. Although these two devices may produce poor-quality images, eliminating them can cause selection bias. Similar to a previous study2, in the current study, one of the three TD dermatologists was the clinical dermatologist who made in-person encounters and decisions. However, the TD interpretations were performed 2 months later to reduce the recall bias by the dermatologist.

In general, trained photographer is an important factor in producing high quality photographs that can enhance the diagnostic accuracy and clinical outcomes. In current study, the implementation of TD is dependent on the preparation of images by a trained person. In other words, the findings of the study may be dependent on this factor. Given that the aim of this study, which is a part of telemedicine, is to develop an app for the empowerment of the medical staff. Therefore, patient empowerment is not our objective. Patient empowerment belongs to the domain of personal health records. In the other word, in current study, the patient wouldn’t use mobile in his/her possession at least for image acquisition.

The common infectious diseases can be handled. However, more challenging skin disorders such as dermatological emergencies, those requiring immunosuppressive treatment, and targeted therapies may not be diagnosed by this app. Further research is needed to evaluate the effectiveness of the SAF mobile app method for these conditions. Cuurent app was developed to provide convenient way for patients with mild to moderate skin conditions to consult with dermatologists remotely, without the need to physical visits. However, this app with current specifications is not intended to replace the standard care for patients with severe to complex skin disorders that require urgent or specialized treatment.

Store-and-forward TD helped clinicians avoid unnecessary physical contact with patients in the out-patient and in-patient settings during the COVID-19 pandemic. Asynchronous teledermatology will likely play a greater role in the future as the store-and-forward images become integrated into synchronous teledermatology workflows23,28.

Some important aspects which need further attention during the setting up of a mobile app for TD programs, especially in the developing countries, are include privacy and ethical issues as well as medical responsibility of diagnosis and medication.

Our study's strengths include the use of board-certified dermatologists and the reduction of bias through independent reviews. The use of high-quality smartphone cameras for image capture also supports the generalizability of our results.

Nevertheless, the study’s limited sample size and the narrow range of skin lesion types examined are notable limitations. These factors may impact the utility of TD for specific lesion groups. Additionally, while our protocol was ideal and standardized, real-world applications of TD may vary, potentially affecting the applicability of our findings.

Future research should focus on expanding the sample size and lesion types to enhance the robustness of TD as a diagnostic tool. Attention must also be given to privacy, ethical considerations, and medical responsibility when implementing TD programs, especially in developing countries.

Conclusion

Our study demonstrates that mobile app-based teledermatology is a valuable and reliable adjunctive tool for diagnosing skin lesions, particularly in countries like Iran where access to dermatologists is severely limited. To facilitate the broader adoption of TD, ongoing efforts are needed to further validate its reliability and expand its utility in diverse clinical settings.