TELEDERMATOSCOPY
Master’s Thesis
Femke van Sinderen, BSc.
Master Medical Informatics - University of Amsterdam
August 2017
Evidence on its accuracy and reliability in skin cancer detection and quality indicators
Supervisors
Dr. J.P. van der Heijden Prof. Dr. M.W.M. Jaspers E. Tensen, MSc.2
Teledermatoscopy
Evidence on its accuracy and reliability in skin cancer detection and quality indicators Student
F. van Sinderen, BSc Student number: 11162953
f.vansinderen@amc.uva.nl
Mentor
Dr. J.P. van der Heijden
Head of Research and Development KSYOS TeleMedical Centre
j.vanderheijden@ksyos.org
E. Tensen, MSc., PhD Candidate Department of Medical Informatics University of Amsterdam
e.tensen@amc.uva.nl
Tutor
Prof. Dr. M.W.M. Jaspers
Adjunct head of Medical Informatics department University of Amsterdam
m.w.jaspers@amc.uva.nl
SRP coordinator Prof. Dr. A. Abu-Hanna
Head of Medical Informatics Department University of Amsterdam
a.abu-hanna@amc.uva.nl
Place of the SRP Project KSYOS TeleMedical Center Professor J.H. Bavincklaan 2-4 1183 AT Amstelveen
Academic Medical Center (AMC) Department of Medical Informatics Meibergdreef 9
1105 AZ Amsterdam
Practice teaching period
3
Table of Contents
Preface ... 5 Acknowledgements ... 5 Abstract ... 6 Samenvatting ... 7 Chapter 1 – Introduction ... 8Chapter 2 – Teledermatoscopy for detecting skin cancer ...10
Burden of primary care ...10
Teledermatology ...11
The teledermatoscopy consultation ...12
Chapter 3 – Teledermatoscopy: a systematic literature review ...13
Introduction ...13
Materials and Methods ...14
Search strategy ...14
Selection process ...15
Data extraction ...17
Methodological quality assessment ...19
Results...20 Search strategy ...20 Selection process ...20 Data Extraction ...22 Discussion ...55 Conclusion ...57
Chapter 4 – A retrospective data analysis on quality and efficiency of teledermatoscopy consultations performed in Dutch primary care ...58
Introduction ...58
Materials and Methods ...60
Teleconsultations ...60
The teledermatoscopy process ...60
Quality of care ...61 Efficiency of care ...62 Results...63 Distribution ...63 Quality of care ...65 Efficiency of care ...70 Discussion ...71 Conclusion ...73 References ...74 Appendix ...78
5
Preface
I have conducted my Scientific Research Project at KSYOS TeleMedical Center and at the Department of Medical Informatics at the Amsterdam Medical Center. This was the last part of the Master Medical Informatics from the University of Amsterdam and has resulted in this Master’s Thesis.
Chapter 1 and 2 of this thesis includes an introduction on this thesis, and background information on the burden on primary care and teledermatology.
Chapter 3 includes the preliminary results of a systematic literature review on the diagnostic/management accuracy and reliability of teledermatoscopy (TDsc) in detecting skin cancer and are part of the PhD study of Esmée Tensen. Regarding her PhD project, I have also assisted with a study design for a future TDsc study, although not included in this thesis.
Chapter 4 of this thesis provides a retrospective data analysis of TDsc consultations performed in The Netherlands. This data analysis describes several quality indicators used to analyze the efficiency and quality of the TDsc consultations.
Acknowledgements
First of all, I would like to thank my supervisors Job van der Heijden, Monique Jaspers and Esmée Tensen for their supervision and feedback during my scientific research project. Second, I would like to thank my family, friends and fellow students from the room J1b-123 for their support and encouragement. Without all of them, this scientific research project would not exist. Lastly, I would like to thank KSYOS TeleMedical Center for the opportunity to perform my Scientific Research Project and all colleagues for their support.
6
Abstract
Introduction: Skin cancer incidences of the most common skin cancer types have tripled in The Netherlands the last twenty years (2) and are expected to increase even more the coming
years (3). TDsc is used in Dutch practice and could increase the number of accurately
detected skin cancers by supporting GPs in contacting remote dermatologists. However, consistent evidence on the accuracy and reliability of skin cancer diagnoses obtained with TDsc is not available.
Aim: To gather evidence on the value of TDsc on the diagnostic/management accuracy and reliability in detecting skin cancer. In addition, the TDsc consultations performed in The Netherlands were assessed concerning its quality and efficiency.
Methods: Three databases (PubMed (MEDLINE), Embase (OVID), Cochrane Skin Review Group) were searched by using multiple search queries. Data about diagnostic/management accuracy/reliability, and implementation barriers/facilitators were extracted from full papers with a focus on skin cancer detection by healthcare providers using TDsc with dermatoscopic images (10-30x magnification).
In addition, TDsc consultations from routine clinical practice performed between general practitioners (GPs) and dermatologists using KSYOS TeleDermatoscopy services were analyzed. Quality of care was measured according to six quality indicators (request advice, extra found skin cancer cases resulting from TDsc consultations, learning effect of GPs regarding diagnostic ability, GPs ratings of the dermatologist response as helpful, and agreement and interrater reliability of GPs and dermatologists on ICD10 diagnoses and referrals. Efficiency of care was defined as the number of prevented physical referrals by using TDsc.
Results: 31 papers were included. A diagnostic accuracy of K = 0.179 – 0.91 for face-to-face (FTF) consultations was reported, whereas a K = 0.35 – 0.94 (50.7 – 96%) was obtained for TDsc consultations. However, some studies did report improvements on diagnostic and management outcomes by TDsc compared to FTF consultations.
In addition, 10,184 TDsc consultations were analyzed. Quality of care indicators obtained moderate to good results. A significant reduction in physical referrals by TDsc was obtained.
Conclusion: There is still no convincing evidence on the diagnostic/management accuracy/reliability of TDsc compared to FTF dermatology. Many limitations need to be addressed to secure the safety of skin cancer patients and successful implementation of TDsc. FTF dermatology is still the method of choice based on the systematic review.
Based on the retrospective analysis, we concluded that TDsc could support GPs in obtaining a diagnosis, treatment, and management plan for patients suspected of skin cancer, and decide whether a referral of a patient is necessary by consulting a remote dermatologist.
Keywords: teledermatoscopy, skin cancer, retrospective data analysis, systematic review, quality and efficiency, diagnostic/management accuracy/reliability
7
Samenvatting
Introductie: Incidentie van de meest voorkomende huidkanker soorten zijn verdrievoudigd in Nederland de laatste twintig jaar(2), de verwachting is dat dit de komende jaren verder stijgt(3).
Teledermatoscopie (TDsc) wordt gebruikt in de Nederlandse huisartsenpraktijk en zou kunnen bijdragen aan het aantal nauwkeurig gediagnosticeerde huidkankers waarbij huisartsen contact hebben met dermatologen op afstand. Desondanks is consistent bewijs omtrent de diagnostische nauwkeurigheid en betrouwbaarheid van huidkanker diagnoses met TDsc nog niet beschikbaar.
Doel: Bewijs verzamelen omtrent de nauwkeurigheid en betrouwbaarheid van
diagnostische/management uitkomsten van TDsc in het diagnosticeren van huidkanker. Daarnaast worden TDsc consulten in Nederland beoordeeld op kwaliteit en efficiëntie.
Methoden: Drie databases (PubMed, Embase, Cochrane Skin Review Group) werden doorzocht met behulp van meerdere zoekopdrachten. Data met betrekking tot
diagnostische/management nauwkeurigheid/betrouwbaarheid, en implementatie barrières/stimulansen werd gehaald uit papers gericht op huidkanker detectie door zorgverleners door middel van TDsc met dermatoscopische foto’s (10-30x vergroot).
Daarnaast werden TDsc consulten uitgevoerd tussen huisartsen en dermatologen door middel van KSYOS TeleDermatoscopie geanalyseerd. Kwaliteit van zorg werd gemeten aan de hand van zes kwaliteitsindicatoren (advies vragen, extra huidkanker patiënten door middel van TDsc, leereffect van de huisartsen omtrent diagnosticeren, de beoordeling omtrent een waardevol TDsc consult, overeenkomst en interrater betrouwbaarheid van de huisarts en dermatoloog op basis van ICD10 diagnosen en verwijzingen). Efficiëntie van zorg werd gedefinieerd als het aantal voorkomen fysieke verwijzingen door het gebruik van TDsc.
Resultaten: 31 papers werden geïncludeerd. Een diagnostische betrouwbaarheid van K = 0.179 – 0.91 voor FTF (face-to-face) consulten werd gerapporteerd, waarbij een K = 0.35 – 0.94 (50.7 – 96%) werd verkregen voor TDsc consulten. Enkele studies rapporteerden verbeteringen omtrent de diagnostische/management uitkomsten bij het gebruik van TDsc vergeleken met FTF consulten.
Daarnaast werden 10,184 TDsc consulten geanalyseerd. De zes kwaliteitsindicatoren behaalden gemiddelde tot goede resultaten. Een significantie reductie in voorkomen fysieke verwijzingen bij het gebruik van TDsc werd behaald.
Conclusie: Er is nog geen overtuigend bewijs omtrent de diagnostische/management
nauwkeurigheid/betrouwbaarheid van TDsc in vergelijking met FTF dermatologie. Vele studie limitaties moeten worden aangepakt om te resulteren in betere study designs. Op basis van het systematische literatuur onderzoek heeft FTF dermatologie nog de voorkeur.
Op basis van de retrospectieve analyse kunnen we zeggen dat TDsc gebruikt kan worden in het ondersteunen van huisartsen bij het verkrijgen van een diagnose, behandeling en management plan voor de patiënten met een verdenking op huidkanker, en beslissen wanneer een patiënt verwezen moet worden naar een dermatoloog.
Trefwoorden: teledermatoscopie, huidkanker, retrospectieve data analyse, systematisch literatuur onderzoek, kwaliteit en efficiëntie, diagnostische/management
8
Chapter 1 – Introduction
KSYOS TeleMedical Center is a Dutch virtual hospital offering various telemedicine and e-health services including teledermatology and TDsc. By using telemedicine services, the aim is to improve efficiency, effectivity, and quality of healthcare by providing healthcare faster, cheaper and closer to the patient instead of referring patients to hospitals. TDsc, a
combination of teledermatology and dermatoscopy, is a non-invasive technique to examine suspicious skin lesions. With this technique, the general practitioner (GP) takes a
dermatoscopic image (10 – 30x magnification) of the lesion which is sent online to a
dermatologist (4). As a result, a diagnosis from a remote dermatologist is obtained without the
need to refer the patient for a physical consult (teledermatology). Since the incidence of malignant skin lesions has increased dramatically the last years and is expected to increase even more the coming years (5), TDsc could contribute to increase faster detection of
malignant skin lesions and decrease the burden on primary and secondary healthcare by preventing unnecessary referrals and therewith costs.
According to the Dutch Cancer Society (Dutch: KWF Kanker Bestrijding), 55,836 patients got the diagnosis of skin cancer in 2016 in The Netherlands of which 12% was diagnosed with melanoma (N=6,787). Skin cancer accounts for 15% of all newly diagnosed cancers and is, therefore, the most common type of cancer in The Netherlands (6, 7).
Furthermore, The Netherlands have the highest melanoma incidence of Europe (1), indicating
that the burden on GPs is increasing regarding consultations concerning suspicious skin lesions (8). Further background information regarding the burden of primary care and TDsc
can be found in Chapter 2.
TDsc is used in Dutch practice and could increase the number of accurately detected skin cancers by supporting GPs in contacting remote dermatologists. However, consistent evidence on the accuracy and reliability of skin cancer diagnoses obtained with TDsc is not available. Outcomes regarding the sensitivity and specificity of skin cancer detection obtained via TDsc remain contradictory. Results between K = 0.35 – 0.94 (50.7 – 96%) on diagnostic accuracy have been reported for TDsc, indicating variability on the diagnostic accuracy of skin cancer detections by TDsc. Besides those variable outcomes,
implementation barriers of TDsc have been reported as well. Among others, patients are not physically examined by the dermatologist, by which malignant lesions on other body parts could be revealed. Second, it has been reported that skin images of high quality should be obtained and diagnosis should be made by an experienced dermatologist in order to obtain a reliable diagnosis with TDsc (9-11). As a result, whether TDsc would be a useful tool for
detecting malignant skin lesions is still under debate (12, 13).
Therefore, the uptake of TDsc in general practice has been limited and deeper insight into the added value of TDsc for detecting malignant skin lesions is required. The aim of this thesis is to gather evidence on the value of TDsc by conducting a systematic literature review, more specific on the diagnostic accuracy and reliability of TDsc in detecting skin cancer. Also, insight in whether unnecessary excisions and referrals are prevented by TDsc is required.
Second, we investigated the limitations and implementation barriers of TDsc. Besides known implementation barriers (e.g. lack of skin lesion palpation), more barriers might be revealed by the systematic literature review. Further, we will investigate the implementation facilitators of TDsc (e.g. the use of TDsc leads to a faster diagnosis, preventing unnecessary skin excisions (14) and physical referrals). Also, insight is needed whether the expertise and
9
conventional dermatology. However, the limitations and implementation barriers must be known first in order to optimize the use of TDsc to experience the potential benefits (e.g. faster diagnosis).
Seventy percent of Dutch GPs have indicated that they lack knowledge and skills concerning correctly diagnosing skin cancer (2, 15). Third, we would therefore like to know
whether TDsc could be a valuable tool to educate GPs in recognizing skin cancer.
The preliminary results of the systematic review are presented in chapter 3. The following two primary research questions are answered in chapter 3:
- Is the diagnostic accuracy and reliability of teledermatoscopy concerning various types of skin cancer comparable with the diagnostic accuracy and reliability found in conventional dermatology?
o If not, what is the difference in diagnostic accuracy and reliability between these two methods for the detection of skin cancers?
o Which factors may explain this difference in diagnostic accuracy and reliability?
- Is the management plan following teledermatoscopy concerning various types of skin cancer comparable to conventional dermatology care pathways?
o If not, what is the difference in the management plan for these skin cancers? In answering these questions we will focus on whether patients will be physically referred more or less after the use of TDsc, and whether patients are correctly referred, and whether the treatment plan is correct.
o Which factors may explain this difference?
Further, one secondary research question is:
- What barriers and facilitators impact the implementation and use of teledermatoscopy by GPs and dermatologists?
o What are known barriers and facilitators?
o What in- or exclusion criteria should be applied to patients when offering TDsc ?
In addition, the TDsc consultations performed in The Netherlands will be analyzed
retrospectively to assess the quality and efficiency of TDsc. This retrospective analysis will be performed in addition to the systematic literature review to gain more insight in quality
indicators concerning the learning effect of GPs and the agreement between GPs and dermatologists on ICD10 diagnoses and referrals. Second, efficiency of care was assessed regarding the number of prevented physical referrals. The KSYOS TeleMedical Center database contains 10,184 TDsc consultations from February 2009 – March 2017 in The Netherlands. Chapter 4 describes the retrospective data analysis that was conducted. The following primary research question is answered in chapter 4:
- What is the effect of teledermatoscopy applied in general practice regarding the quality and efficiency of dermatology healthcare?
The retrospective data analysis will provide an overview of the quality and efficiency of TDsc as used by Dutch GPs.
10
Chapter 2 – Teledermatoscopy for detecting skin cancer
Burden of primary care
Various skin conditions are presented at the consultations to the GP and include among others, skin infections, benign tumors, moles, and various types of eczema (16). According to
a study by Verhoeven et al., the prevalence of consultations for skin diseases presented to the GP was 12.4% in 2007 (17), indicating that a major part of all consultations concerns skin
disorders. Besides those usual skin conditions, potential malignancies (skin cancer) presented to the GP become more common. It is known that skin cancer is a major health concern in The Netherlands. Moreover, The Netherlands had the highest melanoma
incidences of all European countries in 2012 (1) (Figure 1). A study by Koelink et al. reported
that the number of consultations for suspicious skin lesions between 2001 and 2010 increased by 7.3% each year (Groningen, The Netherlands) (18). This might be caused by
increased skin cancer incidences and awareness among the Dutch population. Moreover, most skin cancers are caused by uv-radiation by the sun (19). Almost half of those patients will
have a small surgery (excision) or will be referred to a dermatologist. However, malignancies were only reported in one in eleven patients by which the majority of excisions and
corresponding healthcare costs might be unnecessary. Unfortunately, it has been reported that Dutch GPs lack knowledge about recognizing melanoma, (2, 15) as a result skin lesions
might be incorrectly diagnosed or patients are referred too late to the dermatologist causing a suboptimal treatment. Koelink suggested that improved dermatology training for GPs will result in a better recognition of malignancies thereby saving costs(18).
0 5 10 15 20 25 30 Finland Germany Czech Republic Belgium Ireland United Kingdom Slovenia Sweden Denmark The Netherlands
Number per 100,000 inhabitants
Melanoma incidences in EU-countries in 2012
Females Males
11
Teledermatology
In order to assist GPs in diagnosing skin conditions, teledermatology could be applied. Due to the visual character of dermatology, telemedicine is well-suited for dermatology
healthcare. Several modalities of teledermatology are available, including primary, secondary and tertiary teledermatology. Primary teledermatology concerns the direct communication between the patient and the healthcare provider (e.g. GP). More common is secondary teledermatology by which the GP exchanges medical information with the dermatologists to obtain information concerning the skin disorder. Finally, tertiary dermatology concerns the communication between dermatologists, whereby mostly one dermatologist from an academic hospital is involved (20).
Teledermatology can be performed using three different techniques: store-and-forward (SF), real time (RT) and hybrid (21, 22). SF is the most common technique and uses
still digital images and as a result is place and time independent. However, the GP has to wait for the dermatologists’ response. On the other hand, RT teledermatology uses video conferencing technology, thereby facilitating a live connection between the GP and dermatologist. As a result, RT teledermatology is place independent, however, RT teledermatology is time dependent and could, therefore, interrupt the workflow of the GP and/or dermatologist. Finally, hybrid teledermatology combines SF and RT teledermatology by which video conferencing and digital still images are used. Again, users of this technique are separated by space, however, the hybrid technique is still time dependent.
TDsc is mostly applied in a secondary modality using an SF technique and is used for a variety of skin disorders, among others psoriasis, eczema and suspicious skin lesions.
12
The teledermatoscopy consultation
Nowadays, 3,849 GPs and 251 dermatologists distributed across The Netherlands make use of the teledermatology services by KSYOS TeleMedical Center, and 730 GPs and 95
dermatologists for TDsc. The teleconsult (Figure 2) is started by the GP directly via the GP Information System (Dutch: Huisartsen Informatie Systeem, HIS) that automatically shows the patient demographics (e.g. name, date of birth). Then, the GP adds the macro and dermatoscopic pictures he took earlier and relevant patient information or medical history. The GP fills in his question or anamneses in a free text field. Finally, the GP sends the teleconsult to the dermatologist, including the pictures and information. The dermatologist provides the diagnosis according to the ICD-10 (International Classification of Diseases) code. Further, the dermatologist prescribes medication, gives advice or requests a physical referral of the patient. In the case of any uncertainties or additional questions, the GP can start a second round of the teleconsult by requesting more or specific information from the dermatologist.
Contracted healthcare insurance companies reimburse the costs of the
teleconsultations for healthcare providers. The healthcare insurance companies reimburse KSYOS, that in turn pays the dermatologists. In The Netherlands, teledermatology and TDsc consults are fully reimbursed for the patient since 2005, although the consultations will be charged from their own risk from their health insurance. The charged amount will depend on the Diagnosis Treatment Combination (DTC). KSYOS has established an agreement with the GP to perform at least four TDsc consultations, to assure the quality of performing
teleconsultations. The GPs can reimburse the teleconsults via a MI (modernization and innovation) transaction at the health care insurances (23). KSYOS takes responsibility in case
of any missed melanoma by the dermatologist. The flow diagram for a secondary teledermatology or TDsc consult can be seen in Figure 2.
Figure 2. The process of a secondary Teledermatology (TD) or Teledermatoscopy (TDsc) consult. Blue arrow = information flow. Red arrow = financial flow
13
Chapter 3 – Teledermatoscopy: a systematic literature review
Femke van Sinderen (FS), Esmée Tensen (ET), Job P. van der Heijden (JH), Monique W. M. Jaspers (MJ)Introduction
Incidences of malignant skin lesions have increased dramatically the last years and are expected to increase even more the coming years (3). According to the Dutch Cancer Society,
55,836 patients got the diagnosis of skin cancer in The Netherlands in 2016 of which 12% got the diagnosis of melanoma (N = 6,787). Skin cancer accounts for 15% of all newly diagnosed cancers and is, therefore, the most common type of cancer in The Netherlands (6, 7). Also, it is known that melanoma is the most aggressive form and the leading cause of skin
cancer deaths with a mortality of 89% in The Netherlands (24). Further, the burden on GPs is
increasing regarding consultations concerning skin cancer (8). Unfortunately, it has been
reported in The Netherlands that GPs are under-educated regarding recognizing and diagnosing skin cancer (2, 15).
Telemedicine is applied in many medical areas, due to the visual character of dermatology it is ideally suited for telemedicine. TDsc could be valuable to support GPs in skin cancer detection by providing dermatoscopic pictures to a remote dermatologist for a consultation. Studies on TDsc have shown reductions in physical referrals of patients and the time to obtain a diagnosis, thereby offering opportunities to increase the efficiency and
quality of care, respectively (25, 26). However, TDsc remains debated due to disagreement on
diagnostic accuracy and detection of malignant lesions (12, 13). Furthermore, a full physical
examination is not performed by the dermatologist by which malignant lesions on other body parts might be missed.
TDsc is used, among others in Dutch practice using a store-and-forward technique. It is since 2005 fully reimbursed and integrated into the Dutch healthcare system. Since then, over 10,000 TDsc consultations were provided by KSYOS TeleMedical Center. TDsc could improve the diagnostic accuracy of skin lesions by supporting GPs. However, no meta-analysis concerning evidence on diagnostic accuracy of skin lesion detection by TDsc has been made yet. It has yet been reported in the literature that TDsc is comparable to conventional dermatology regarding diagnostic accuracy and outcomes (21, 27), but it is still
unclear what the evidence base is concerning diagnostic accuracy and reliability. As a result of this, the uptake of TDsc in general practice has been limited.
Therefore, the aim of this systematic literature review is to obtain an overview of the reported diagnostic accuracy and reliability of TDsc in detecting skin cancer. Second, insight is required regarding the limitations and implementation barriers of TDsc. Therefore, the following primary research questions have been formulated:
- Is the diagnostic accuracy and reliability of TDsc concerning various types of skin cancer comparable with the diagnostic accuracy found in conventional dermatology?
- Is the management plan following TDsc concerning various types of skin cancer comparable to conventional dermatology care pathways?
We aim to answer one secondary research question:
- What barriers and facilitators impact the implementation and use of teledermatoscopy by GPs and dermatologists?
14
Materials and Methods
Search strategy
The PRISMA guidelines and a systematic review guideline served as a basis for the search strategy method (28, 29). It was established whether systematic reviews on TDsc were
available or registered by searching the electronic databases PubMed (MEDLINE), Embase (OVID), The Cochrane Library, and the PROSPERO database (30) by using the search query
“teledermatoscopy OR teledermoscopy”.
The systematic literature search was conducted at 30 January 2017 by searching the earlier mentioned databases and the Skin Review Group from Cochrane Library. Cochrane Central consists only of Medline articles not indexed with MeSH (Medical Subject Headings) which would not yield additional articles compared to PubMed and was therefore not
searched. An information specialist from the medical library assisted on the search methodology.
Multiple search queries were used and adjusted to search the databases of PubMed and Embase (Table 1). Nesting was applied by constructing two clusters with key terms either related to teledermatoscopy or skin cancer. Key terms in both clusters were combined with the operator “OR”. The first and second cluster were combined with the operator “AND” to result in the final search queries to review the literature for studies on TDsc describing original data on the detection of skin cancer. Titles of the Cochrane Skin Group were scanned. Multiple search queries were applied to result in a broad search strategy by which we aimed to include as many papers as possible on TDsc.
One author (FS) extracted the databases supervised by author ET. Limits were set for publication date until 1st January 2017, allowing an easy update for 2017 and onwards. No
restrictions were applied in advance on article type. Only English and Dutch articles were included by applying a language restriction in PubMed. This was not applied in Embase by which other languages were removed manually after removing the duplicates. Figure 3 shows the search flow to identify relevant articles (phase 1).
Table 1. Search queries PubMed; MeSH = Medical Subject Heading, [tiab] = title and abstract. Search queries Endnote; .ab,ti = abstract, title (equivalent to [tiab], exp = explode (equivalent to MeSH).
Nr. Search query PubMed Search query Embase
1 teledermatoscopy [tiab] OR teledermoscopy [tiab] teledermatoscopy.ab,ti OR teledermoscopy.ab,ti
2 (telemedicine [MeSH] OR teledermatology [tiab] OR
teledermatoscopy [tiab] OR teledermoscopy[tiab]) AND dermoscopy [MeSH]
(exp telemedicine/ OR teledermatology.ab,ti OR teledermatoscopy.ab,ti OR teledermoscopy.ab,ti) AND exp dermoscopy/
3 (teledermatoscopy [tiab] OR teledermoscopy[tiab]) AND
telemedicine [MeSH]
(teledermatoscopy.ab,ti OR teledermoscopy.ab,ti) AND exp telemedicine/
4 (teleconsult* [tiab] OR telederm*[tiab]) AND dermoscopy
[MeSH]
(teleconsult*.ab,ti OR telederm*.ab,ti) AND exp dermoscopy/
5 (teledermoscopy[tiab] OR teledermatoscopy[tiab] OR
telemedicine [MeSH] OR teleconsult* [tiab]) AND (dermatol* [tiab] OR skin*[tiab] OR neoplasm [MeSH])
(teledermoscopy.ab,ti OR teledermatoscopy.ab,ti OR exp telemedicine/ OR teleconsult*.ab,ti) AND (dermatol*.ab,ti OR skin*.ab,ti OR exp neoplasm/)
6 (teledermoscopy [tiab] OR teledermatoscopy [tiab] OR
teledermatology [tiab]) AND neoplasms [MeSH]
(teledermoscopy.ab,ti OR teledermatoscopy.ab,ti OR teledermatology.ab,ti) AND exp neoplasm/
7 (teledermoscopy [tiab] OR teledermatoscopy[tiab] OR
teledermatology[tiab] OR teleconsult*[tiab] OR
telemedicine [MeSH]) AND (dermatol*[tiab] OR skin*[tiab] OR neoplasm [MeSH])
(teledermoscopy.ab,ti OR teledermatoscopy.ab,ti OR teledermatology.ab,ti OR teleconsult*.ab,ti OR exp telemedicine/) AND (dermatol*.ab,ti OR skin*.ab,ti OR exp neoplasm/)
8 (teledermoscopy[tiab] OR teledermatoscopy[tiab] OR
teledermatology[tiab] OR telemedicine[MeSH] OR teleconsult*[tiab] OR telederm*[tiab] ) AND (dermatol*[tiab] OR skin*[tiab] OR neoplasms [Mesh] OR dermoscopy [MeSH])
(teledermoscopy.ab,ti OR teledermatoscopy.ab,ti OR teledermatology.ab,ti OR epx telemedicine/ OR teleconsult*.ab,ti OR telederm*.ab,ti) AND
(dermatol*.ab,ti OR skin*.ab,ti OR exp neoplasm/ OR exp dermoscopy/)
15
Selection process
Exclusion criteria of publications for the title, abstract and full paper scan were determined a priori (Table 2 – 4) by all authors. Two reviewers (ET and FS) independently performed the title, abstract and full paper scan of the included studies against predefined exclusion criteria. Consensus between the two reviewers (ET and FS) was obtained when both agreed whether a study was in- or excluded. A third reviewer (JH) assisted when consensus on exclusion was not obtained in any phase of the search flow.
Title scan
Due to our extensive search strategy and a large amount of studies, a title scan was performed first according to the exclusion criteria mentioned in Table 2.
Table 2. Title scan exclusion criteria
Title scan exclusion criteria
Other domain than telemedicine/e-health
Other domain than teledermatology/teledermatoscopy Other domain than dermatology
Abstract scan
A title scan was followed by an abstract scan. Papers extracted from Embase were indexed with their article type (e.g. editorial). A verification was performed for available papers indexed with an article type. If papers were not available, the article type was used as a reason for exclusion. Otherwise, abstracts were scanned and the last three named exclusion criteria from Table 3 were applied to decide on exclusion of a paper. Those exclusion criteria were also applied to conference papers after they were read.
Table 3. Abstract scan exclusion criteria
Abstract scan exclusion criteria
(Systematic) review
Letter (to the editor)/correspondence letter Note/opinion/commentary/short report/column Case report/study
Editorial Erratum Book
Only survey/questionnaire/interview was used as a study method Conference abstract/proceeding
No healthcare provider involved
No dermatoscopic picture; picture is not made with a dermatoscope or mobile phone with dermatoscope attachment
Other domain than teledermatoscopy for detection and management of skin lesions (e.g. psoriasis, teledermatopathology, the goal of the study is education or cost-benefit analysis).
16
Full paper scan
Only full papers were included with a focus on skin cancer detection by healthcare providers using TDsc with dermatoscopic images by applying the exclusion criteria from Table 4. We included studies whereby dermatoscopic pictures were taken by a healthcare professional or patient with a digital dermatoscope or mobile dermatoscope attachment. Of importance was that dermatoscopic pictures must always be assessed by a health care provider. Studies that reported on training or education in the field of telemedicine were excluded unless TDsc was used as an educational method for healthcare providers.
Table 4. Full paper scan exclusion criteria
Full paper scan exclusion criteria
No dermatoscopic picture; picture is not made with a dermatoscope or mobile phone with dermatoscope attachment
Outside the domain for the detection of skin cancer/lesions Outside the domain of telemedicine/teledermatoscopy No healthcare provider involved
No original study (see Table 3)
Studies were not excluded based on the title or abstract scan when the aim of the study was unclear thereby preventing false exclusion of studies. In those cases, a full paper review of the articles was performed. Studies resulting from our selection process were validated by comparing the included studies from the systematic reviews of Finnane et al.,(31) and Bruce et
al.,(32) that were published after we performed our search strategy. Next, outside the scope of
this thesis, not freely available papers will be requested at the medical library (Academic Medical Center, Amsterdam, The Netherlands). And snowballing of the reference lists of (systematic) reviews captured by our search strategy will be performed to include additional publications.
Inter-rater reliability
Inter-rater reliability was assessed using the Cohen’s Kappa statistics for each of the phases according to the flow diagram (Figure 3). A Kappa value of 0.01-0.20 was seen as none to slight agreement, 0.21-0.40 as fair, 0.41-0.60 as moderate, 0.61-0.80 as substantial, and 0.81-1.00 as almost perfect agreement (33).
17
Data extraction
Subsequently, data was extracted for the included studies using four data extraction formats (Table 5-8). Two reviewers (ET, FS) extracted the data. The data extraction format included study characteristics and study outcomes related to the research questions. The study outcomes are defined below for each research question:
- Is the diagnostic accuracy and reliability of teledermatoscopy concerning various types of skin cancer comparable with the diagnostic accuracy found in conventional dermatology? The outcomes regarding diagnostic accuracy and reliability, sensitivity, specificity, and interrater reliability were extracted. Where the diagnostic accuracy is defined as the
agreement between the teledermatologist or face-to-face dermatologists and the used golden standard, either histopathology (excised lesions) or face-to-face dermatologists (non-excised lesions). So, the test diagnosis versus final true diagnosis is assessed. Diagnostic reliability (concordance) assess whether a patient is diagnosed with the same skin condition by all raters. Sensitivity describes the proportion of patients correctly identified with the skin condition. Whereas specificity describes the proportion of patients correctly identified as not having the skin condition. Finally, interrater reliability, assessed by Cohen’s Kappa Statistic, concerns the extent of agreement among raters where chance agreement is taken into account.
- Is the management plan following teledermatoscopy concerning various types of skin cancer comparable to conventional dermatology care pathways?
The outcomes regarding management concordance and training for health care
professionals were extracted to answer the second primary research question. This included concordance on management decisions relating to the referral or treatment plan of the patient between teledermatologists and face-to-face dermatologists.
- What barriers and facilitators impact the implementation and use of teledermatoscopy by GPs and dermatologists?
The outcomes regarding reported limitations, barriers, and facilitators of the study were extracted to answer the secondary research question.
A third reviewer (JH) was asked when consensus on in- or exclusion of papers was not obtained between the two reviewers (ET and FS). It was investigated whether the meta-analysis on TDsc studies concerning diagnostic accuracy and reliability was possible regarding the homogeneity of included studies.
18 Table 5. Data extraction format study characteristics
Author, study aim Study Design Country, Study setting Country: Population In/exclusion on patients Timeframe Outcome measurements
Table 6. Data extraction format diagnostic accuracy and reliability (primary research question)
Author Skin cancer Modality: Camera Magnification dermatoscope Additional clinical information Intervention test Golden standard (Statistical) analysis Outcomes
Table 7. Data extraction format management outcomes (secondary research question)
Author Training/Education/Experience Management decisions Management concordance
Table 8. Data extraction format barriers and facilitators (secondary research question)
19
Methodological quality assessment
After the data extraction, a quality assessment according to QUADAS-2 (Quality Assessment of Diagnostic Accuracy Studies, 2nd edition) (34) will be performed. However, this will be
performed outside the scope of this thesis and is therefore not discussed in the result section. A quality assessment will be performed in order to assess the bias and quality of included studies and to rank the studies based on their quality. The included studies will be ranked based on their quality by answering the signaling questions of QUADAS-2 (Table 9). Second, strength of evidence to the research questions will be assessed. Studies with a higher ranking (e.g. less bias) are seen as studies of higher quality and are therefore seen as studies who have more strength to answer our research questions. The ranking of the
studies based on their bias and quality will influence the interpretation of study outcomes answering our research questions.
Table 9. Signaling questions QUADAS-2. Adapted from Whiting et al. (34)
Domain 1: Patient Selection
Risk of Bias: Could the Selection of Patients Have Introduced Bias?
Signaling question 1: Was a consecutive or random sample of patients enrolled? Signaling question 2: Was a case–control design avoided?
Signaling question 3: Did the study avoid inappropriate exclusions?
Applicability: Are There Concerns That the Included Patients and Setting Do Not Match the Review Question?
Domain 2: Index Test
Risk of Bias: Could the Interpretation of the Index Test Have Introduced Bias?
Signaling question 1: Were the index test results interpreted without knowledge of the results of the reference standard?
Signaling question 2: If a threshold was used, was it prespecified?
Applicability: Are There Concerns That the Index Test, Its Conduct, or Its Interpretation Differ From the Review Question?
Domain 3: Reference Standard
Risk of Bias: Could the Reference Standard, Its Conduct, or Its Interpretation Have Introduced Bias? Signaling question 1: Is the reference standard likely to correctly classify the target condition?
Signaling question 2: Were the reference standard results interpreted without knowledge of the results of the index test?
Applicability: Are There Concerns That the Target Condition as Defined by the Reference Standard Does Not Match the Question?
Domain 4: Flow and Timing
Risk of Bias: Could the Patient Flow Have Introduced Bias?
Signaling question 1: Was there an appropriate interval between the index test and reference standard? Signaling question 2: Did all patients receive the same reference standard?
20
Results
Search strategy
No published systematic reviews focusing specifically on diagnostic accuracy and reliability of teledermatoscopy were found prior to our systematic literature search. The search in PROSPERO resulted in one record of a systematic review on teledermatoscopy (31) which
was published after we started our search strategy. Subsequently, one more systematic review was published during our search strategy (32). Those two systematic reviews were
used to validate the inclusion of our studies.
The search queries (Table 1) resulted in a total of 4,436 references from PubMed and 7,716 references from Embase. The Cochrane Skin Group consisted of 126 titles (30
January 2017) which were all scanned, none of them were included. A total of 4,408
duplicates were removed resulting from the multiple search queries applied in PubMed and Embase. Articles not meeting Dutch or English language resulting from Embase were removed (N = 390). And another 851 duplicates were removed after combining the
references from PubMed and Embase. A total of 3,167 titles remained and were subject to further screening.
Selection process
2,262 titles were excluded after performing the title scan (Table 2). Another 750 abstracts were excluded after the abstract scan (Table 3) of which 104 abstracts were excluded since consensus on in- or exclusion was not obtained by the two reviewers. Abstracts were mainly excluded if the article was not an original study (e.g. reviews, letters, notes, editorials).
In total, 155 full papers were assessed for eligibility of which 41 were not available and copies will be requested at the Medical Library. Those were therefore not further assessed for this thesis. Remaining papers (N = 114) were scanned for eligibility. Based on the full paper scan, 83 studies were excluded, among them, there were 66 studies where no dermatoscopic image was used for the skin cancer detection. And 3 studies were excluded since no consensus was obtained by the two reviewers. These papers will be reviewed by the third reviewer outside the scope of this thesis together with the 104 abstracts where a third review was necessary.
Finally, this selection process resulted in 31 papers (9, 12, 13, 27, 35-61) published between
1998 and 2016. The validation by the systematic reviews of Bruce et al (32)., and Finnane et
al (31) obtained the following results. Two studies (62, 63) were not included in our selection
since those will be reviewed by the third reviewer, one study (64) was not included since it was
not an original study, two studies (65, 66) were not included in the full paper scan since, to our
understanding, no dermatoscopic images were used, one study (67) was not included in the
abstract scan since it was a conference abstract, and two studies (68, 69) were not freely
available. Only the included studies concerning the diagnostic accuracy, diagnostic
concordance, treatment accuracy and treatment concordance from the systematic review of Finnane et al., were used for validation. Thus, eight out of the 25 assessed studies resulting from those two systematic reviews were not included. Further, 15 from our 31 included studies were included by Finnane et al., and Bruce et al. Elaborated results for excluding full papers can be seen in Figure 3.
21
Interrater reliability
Interrater reliability was assessed by using the Cohen’s Kappa Statistic for the title, abstract and full paper scan. A Kappa of 0.68 was found for both the title and abstract scan, indicating a substantial agreement. An almost perfect agreement was found for the full paper scan with a Kappa of 0.83.
22
Data Extraction
Study characteristicsData was extracted for the included studies resulting from the selection process (N = 31)
(9, 12, 13, 27, 35-61). Study characteristics can be seen in Table 10 where each study is described
by information on study design, the country where the study was performed, included study population, in/exclusion criteria applied on the study population, timeframe of the study, and the aim and outcomes of the study. Our results show that the included studies assessed among others, the diagnostic accuracy/reliability, the management accuracy, the interrater reliability, and the sensitivity and specificity. Most studies were performed in a prospective quantitative manner. The majority of studies were performed in real practice settings at a GP practice or at a dermatology department. Further, these studies were performed across different countries, with most studies originating from Austria, Italy and the USA. If in- and exclusion criteria on patients were applied, the included patients were mostly patients who presented or were referred to a dermatology (outpatient) clinic and suffered from multiple lesions. Second, most patients were excluded due to skin lesions on body sites that did not allow photography (e.g. scalp, genital area). All studies were performed in an SF modality, except for the study of Grimaldi et al. (41) which was performed in a RT modality. Second, four
studies (35, 43, 47, 48) assessed the mobile TDsc, of which two studies (35, 43) assessed the
patient-performed mobile TDsc. Dermatoscopic images were mostly taken by a GP, dermatologist or melanographer (medical photographer) and assessed by a
(tele)dermatologist. Reported timeframes of the studies were between 1996 and 2016 and publications originated from 1999 to 2016.
A meta-analysis was not possible due to the heterogeneity of included studies. Data extraction tables can be found below concerning each research question.
23 Table 10. Study characteristics of included studies
* only outcomes regarding dermatoscopy have been taken into account. FTF = face-to-face, TDsc = teledermatoscopy, TD = teledermatologist, F = female, M = male, GP = general practitioner, NA = not applicable, NM = not mentioned
Author Study Design Country, Study setting
Country: Population
In/exclusion on patients Timeframe Outcomes Arzberger et al., 2016 (27).
To compare the medical recommendations of a FTF with clinical /dermoscopic images obtained by melanographer and assessed by 4 TDs (New Zealand, Germany, Austria)
Retrospective chart review Pigmented Skin Lesion Clinic, Graz, Austria Austria: 70 patients, 35F, 35M, median 39yr (range 11-81), 1680 body sector photographs (n=24 per patient), 1922 detail-images of single lesions
Inclusion: (i) personal or first-degree relative history of melanoma; (ii) history of dysplastic naevi; (iii) > 5 atypical naevi; (iv) > 100 naevi; (v) lesion suspicious for melanoma. May - Oct 2009 (patient enrollment). June 2013: medical chart review interobserver agreement, management recommendations/concordance Blum et al., 2004 (60). To
determine whether the diagnostic accuracy of an inexperienced, average, and highly experienced investigator in dermoscopy can be significantly improved by incorporating information about
morphological changes, location of the lesion, and the age and sex of the patients.
Retrospective study Pigmented Lesion Clinic, Department of Dermatology, University of Tuebingen, Germany Germany: 157 patients (86F, 71M), median age 38.9yr (SD 16.8 yr - 2-87). Exclusion: Malignant epithelial tumours (basal cell carcinoma, squamous cell carcinoma)
September 1998 - March 1999
diagnostic accuracy, sensitivity, specificity
Borve et al., 2013 (48).
images taken by FTF dermatologist and assessed by 2 TDs. Prospective diagnostic Accuracy, interobserver concordance, management adequacy Department of Dermatology as Sahlgrenska University Hospital, Sweden Sweden: 62 patients, 69 lesions, (24F, 38M), mean age 64 years (range 25-94) (6 patients had multiple lesions) Inclusion: ≥ 1 suspicious skin lesions deemed to need a biopsy/excision. Exclusion: ≤18 years, no knowledge of the Swedish language, skin lesions located on a part of the skin that did not allow for digital photography with the smartphone and customised dermoscope used in the study and skin lesions in which histopathological 16-week period diagnostic accuracy, interobserver concordance, adequacy management decisions
24
examination was not performed.
Braun et al., 2000 (56).
Images were obtained by dermatologists from private practice and assessed by a physician from the PSL.
Prospective diagnostic accuracy Pigmented Skin Lesion Unit of the University Hospital Geneva, Switzerland 51 patients (55 lesions)
NA 6 months diagnostic accuracy
Carli et al., 2002 (49).
Assess the diagnostic performance of in vivo dermosocpy (before surgical excision, clinical + dermoscopic examination) with that of dermoscopy (photography).
Case series general dermatological clinic, University hospital, Italy Italy: 256 pigmented skin lesions
Inclusion: skin lesions consecutively identified as suspicious or equivocal during examination in a general dermatological clinic (by dermatologsists)
Jun 1997 - Dec 1998
(inclusion)
diagnostic accuracy, true positive (TP), false positive (FP), true negative (TN) and false negative (FN) diagnosis, sensitivity, specificity
Congalton et al., 2015 (51).
Dermoscopic picture obtained by specialist nurse and assessed by experienced TDs Prospective diagnostic accuracy Virtual lesion clinic, Auckland, New Zealand New Zealand/European (78%): 310 patients, 613 skin lesions, 168F, 142M. Median age 58y (15-92yr). Inclusion: patients referred from PC with suspected melanoma. Exclusion: lesions on scalp/genitals/not clearly identified in the referral
1 April 2012 - 31 March 2013 (patient referral) sensitivity, specificity, PPV, NPV de Giorgi et al., 2016 (9).
Assess the diagnostic concordance between FTF and TDsc of images assessed by 10 dermatologists Retrospective diagnostic/interobserver concordance Dermatology Department, University of Florence, italy 10 patients (10 lesions), 6F, 4M; mean age 51,3yr (range 26-69)
NA NM diagnostic concordance, interobserver concordance
Di Stefani et al., 2007 (44).
Clinical and dermoscopic images were sent to remote experts to assess FTF and remote management. Selected lesions for further dermoscopic examination were marked by the remote consultants directly on the clinical image. In the second step, dermoscopic images of selected lesions were independently evaluated by remote experts. Prospective diagnostic accuracy Department of Dermatology, Italy. Italy: 18 patients (11M, 7F), mean age 28.4year (range 10-55). 465 PSL, mean number of 26 (10-67) PSL per patient inclusion: consecutive patients with multiple PSL, exhibiting ≥ 3 clinically atypical melanocytic nevi on their back, defined as: flat or slightly elevated, acquired melanocytic nevi showing at least one of the following criteria: (1) diameter of more than 5 mm, (2) color
variegation,(3)
NM interobserver agreement, diagnostic concordance
25
asymmetry, and/or (4) irregular borders.
Fabbrocini et al., 2008 (58).
determine the diagnostic reliability, according to interobserver agreement, between clinical and dermoscopic
diagnosis of 'pink' lesions comparing FTF
diagnosis and telediagnosis by two dermatologists.
Cross-sectional, non-experimental
Italy (setting not mentioned) Italy: 44 lesions (39 melanocytic, 5 non-melanocytic) Recruitment of patients is unkown NM interobserver agreement, diagnostic reliability Grimaldi et al., 2009 (41). 13 PCPs not expert in detecting pigmented skin lesions took images which were sent to a group of physicians in the reference centre. Prospective diagnostic accuracy Peripheral centers, reference centre (unit of plastic surgery, Italy) Italy: 197 consecutive patients, 235 lesions
Exclusion: lesions whose removal had been explicitly demanded by the patients for aesthetic reasons, as well as those irritated or subjected to trauma. Inclusion: 1 Oct 2005 - 31 Mar 2006 (completed follow up 30 Sept 2006) diagnostic agreement, management agreement Kroemer et al., 2011 (47).
Assess the diagnostic accuracy of clinical image tele-evaluation and TDs for mobile skin tumour screening of images obtained by the clinician and assessed by the board-certified dermatologist. Prospective diagnostic accuracy General outpatient clinic at the Department of Dermatology, Medical University of Graz, Austria. Austria: 80 patients (104 tumours)
NA 3 month period diagnostic accuracy, sensitivity, specificity, concordant vs discordant cases
Livingstone et al., 2015
(61). evaluate the
cost-effectiveness, safety of referral and patient satisfaction of TD in a general practice setting. Images obtained by trained member of staff and assessed by dermatologists. Retrospective case record analysis/service evaluation General practice (6500 patients), suburban area of Greater London, UK.
UK: 248 patients Inclusion: 18+ years, benign looking skin lesion, normally referred to secondary care. Exclusion: if patients met the 2-week wait criteria, dermatology condition affecting the genital area, extraordinary
circumstances (e.g. post-transplant
patients).
Nov 2010-nov2013
26
Ludzik et al., 2016 (45)*
Dermoscopic images were sent to a 1 dermoscopy reader to determine diagnostic accuracy and management Observational retrospective analysis Department of Dermatology, Italy. Italy: 316 patients, 316 lesions.
Inclusion: (i) lesion detected by clinical naked-eye examination with
absent pigmentation or containing less than 10% pigment; (ii) absence of pigment network; (iii) all lesions excised with matching
histopathology report; (iv) availability of digital dermoscopy
images; (v) availability of a complete standard set of RCM images. Jan 2009 - Jan 2012 (patient inclusion), Jan - March 2016 (evaluation) sensitivity, specificity, management accuracy Manahan et al., 2015 (35). Patient-performed mobile TDsc. Photos were sent to blinded TDs. All patients received CSE.
Pilot RCT Patient's home. University of Queensland, Brisbane, Australia. Australia: 49 participants (25 intervention, 24 control). Age: 50-64yr. 49% male NA May - Nov 2013, SSE completed within 3 weeks diagnostic concordance (TDsc vs CSE), sensitivity (SSE + mobile TDsc vs. In-person CSE) using either patients or lesions as denominator. Massone et al., 2007 (46). clinical (1st) and dermoscopic images (2nd) were assessed by 2 teleconsultants independently Prospective diagnostic agreement Pigmented Skin Lesions Clinic of the Department of Dermatology, Medical University of Graz (Austria), images acquired during routine conditions. Austria: 18 patients, 12M, 16F, mean age 43, median age: 45, range 14-78 NA Patient recruited: 2 working days diagnostic agreement Massone et al., 2014(53).
Dermoscopic images taken by GPs and assessed by two dermatologists Observational 3 preventive health care centres in Austria in 3 hotels/wellness centres in the setting of a general preventive medicine screening 690 patients, 642 M, 48 F, mean/median age 47 (18-84 years) Exclusion: inflammatory lesion Feb 2008- Feb 2010 (patients were selected)
diagnostic accuracy, sensitivity, specificity, management recommendation
27 programme offered by a private Austrian insurance company. Moreno-Ramirez et al., 2006 (55). Dermoscopic pictures obtained by GP in primary care centers and assessed by dermatologists Prospective diagnostic/management agreement Pigmented lesion and teledermatology clinic, 1 primary care center that use the TD facility, Spain Spain: 61 teleconsultations, 18M, 43 F, mean age 38.8 (range 1-73). Inclusion: pigmented circumscribed lesions fulfilling ≥ 1 of the following criteria: changing lesion (ABCD changes), recent lesion (≤ 3 years old), multiple lesions (> 20 melanocytic naevi counted by the GP)10 symptomatic lesion (pain, itching or bleeding), or patient concern about moles. Exclusion: Patients who did not attend the face-to-face clinic were excluded (N = 2) Start Enrolment Sept 2004 diagnostic/management agreement, sensitivity, specificity Piccoli et al., 2015 (54).
Dermoscopic images taken by technician and assessed by a dermatologist. Observational cross-sectional study Center for Telemedicine and Telehealth of the State Secretariat of Health of Santa Catarina, Brazilië 184 evaluations 135 F, average age 54.74yr, 83.7% white persons (phototypes I, II and III)
Inclusion: who sought a Basic Healthcare Units in Santa Catarina with visible dermatological complaints. Exclusion: no visible/insufficient quality of lesions, undergoing dermatological treatment, only clinical symptoms without visible dermatological symptoms Jan 2012-Jul 2012 sensitivity, specificity, PPV, NPV, accuracy of diagnosis Piccolo et al., 1999 (42). FTF diagnosis vs telediagnosis of pigmented skin tumors was assessed between two centers of dermatologists.
Prospective diagnostic concordance
Skin lesion clinic Italy, university dermatology department Austria. Italy: 66 lesions/patients (34F, 32M), mean age: 41.2 yr (range 8-82). NA NM diagnostic concordance
28
Piccolo et al., 2000 (57).
Evaluate the accuracy of telediagnosis of 43 pigmented skin lesions by 11 colleagues with different degrees of experience in the diagnosis of pigmented skin lesions in general and dermoscopy in particular. Cross-sectional; non-experimental, comparative Department of Dermatology, Graz, Austria (FTF) Austria: 40 patients (21M, 19F), median age 39.5 years (range 3–91), 43 pigmented skin lesions. NA Patients were studied over 3 months. diagnostic concordance/accuracy Piccolo et al., 2004 (40).
Dermoscopic images taken at 2 Dermatology
Departments and assessed by 11 dermatologists in remote centers with different degrees of experience. Prospective virtual multicentre study Dermatology Departments: University of Graz, University of L'Aquila. Remote centers; Japan, Italy, Slovenia, Austria 73 patients (39F, 34M) aged 4-77yr (mean 28y).
NA NM sensitivity, specificity, inter-observer agreement
Pizzichetta et al., 2004 (39).
Images taken in 5 centers and sent to single blinded observer from oncology center in Italy. Retrospective clinical study 4 Clinics in Italy, 1 in USA USA + Italy: 151 patients, 48% male, mean age 47 years (SD 17.5)
Inclusion: (i) confirmation of clinical and ⁄ or dermoscopic amelanotic ⁄ hypomelanotic
lesions with extent of pigmentation ≤30%, and (ii) an image quality sufficiently good for the evaluation of the dermoscopic criteria, particularly vascular patterns. Jan 1996 - Dec 2001 sensitivity, specificity Senel et al., 2013 (59). To
compare the TD and FTF reliability of non-melanocytic skin tumours after the addition of dermoscopic images taken by technicians and assessed by 2 dermatologists. Prospective diagnostic reliability/accuracy Dermatology outpatient clinic, Ankara hospital, Turkey Turkey: 150 patients, 51% male, mean age 55 yr.
Inclusion: presented to the dermatology outpatient clinic, a non-melanocytic skin tumour on clinical examination April 2009 - Sept 2009 (patient inclusion). 2 month interval between with/without dermatoscopic image evaluation.
diagnostic reliability, diagnostic accuracy
29
Senel et al., 2014 (36).
Retrospective evaluation of microscopic images obtained by dermatologist. Clinical images evaluated by two dermatologists and management plan made. Re-evaluation diagnosis and managment plan after 2 months. Retrospective diagnostic accuracy and reliability Department of Dermatology, University of Hitit, Turkey. Outpatient dermatology clinics Turkey: 120 consecutive cases, 68M (57%). Mean age 63 yr (SD 16). NA Re-evaluation after 2 months diagnostic accuracy (N = 38) /reliability (N = 120), management concordance Tan et al., 2010 (12). Dermoscopic images obtained by melanographer were assessed by 5 independent dermatologists (USA, NZ, Australia) to assess interobserver variability. Prospective interobserver diagnostic variability Hospital specialist skin lesion clinic New Zealand: 206 patients, 979 lesions NA NM interobserver variability Tan et al., 2010 (52). Dermoscopic images obtained by melanographer and assessed by dermatologists. Prospective diagnostic accuracy Waikato Hospital, Department Dermatology, Hospital skin lesion clinic, routine outpatient setting, New Zealand. 94% European ancestry: 200 patients, 74M, 126F, 491 lesions. Age: 11-94yr. Most patients Fitzpatrick skin phototype 2.
Inclusion: referred by GP to skin lesion clinic for diagnosis/management of ≥1 skin lesions.
Exclusion: not competent to give informed consent (n=2), did not see FTF dermatologist (n=4), refused participation . March-Sept 2008 inter/intraobserver agreement, diagnosis/management concordance
van der Heijden et al., 2013(37). Images taken by a GP (N=13) compared to FTF dermatological examination by dermatologist (N=4). Prospective diagnostic accuracy and reliability
GP regular practice, The Netherlands. The Netherlands: 105 patients, median age of 47 yr (range 6–84) , 55% F. 108 lesions/TDsc consultations. Inclusion: consecutive patients who presented with a (pigmented) skin lesion. Exclusions: urgent cases
Feb 2010-May 2011
diagnostic accuracy,
management agreement, inter-observer reliability (diagnosis, management)
30
Warshaw et al., 2008(38).
Lesions examined by clinic staff dermatologists and followed by TDsc. Cross-sectional repeated-measures trial Minneapolis Veterans Affairs Medical Center dermatology clinic, USA USA: 712 participants (veterans), male (97,8%) and caucasian (98,9%)
Exclusion criteria included (1) individuals requesting or referred for skin tag removal only; (2) individuals presenting for excision or treatment of a neoplasm previously biopsied; (3) individuals requiring biopsy for papulosquamous or eczematous conditions (non-neoplastic); and (4) inability to comprehend and give informed consent.
NM diagnostic accuracy, management accuracy
Warshaw et al., 2009(13).
comparing SAF TD with traditional, in-person clinical encounters for pigmented lesions. Cross-sectional repeated-measures trial Minneapolis Veterans Affairs Medical Center dermatology clinic, USA 96% Caucasian: 542 patients, 96% male, mean age 66yr (23-94). Inclusion: in need of evaluation of a pigmented skin neoplasm subsequently biopsied, enrolled in dermatology clinic who were undergoing removal/biopsy. NM diagnostic accuracy, management accuracy Warshaw et al., 2015(50). dermatoscopic images taken by research
assistants and assessed by TDs to compare
conventional dermatology to TD for skin neoplasms.
Cross-sectional repeated-measures trial Minneapolis Veterans Affairs Medical Center Dermatology Clinic, USA USA: 2152 patients, 3021 lesions. 96.8% male, mean age 68 y (19-94). 97.5% Caucasian.
Inclusion:
referred/enrolled in dermatology clinic who were undergoing a biopsy. Exclusion: lesions did not meet study criteria (skin tag, rash, lesion treated, lesion resolved), declined participation, unable to provide
consent, in another study, patient died/moved, other
NM diagnostic/management agreement
Wu et al., 2015(43).
Dermoscopic images were obtained in the office-based setting by a dermatologist and evaluated by office-based- and
prospective cohort study
Institutional referral center, New York, USA (patients recruited). Office-based USA: 29 patients, 33 lesions. Inclusion: Consecutive patients ≥18yr, with ≥1 clinically atypical nevi that required short-term monitoring and were accessible by a mobile imaging device. Feb 2013 - Feb 2014 (recruitment) + 3-4 months follow-up. diagnostic concordance
31
teledermatologist. Follow-up visit by dermatologist
setting (images obtained).
Exclusion: could not fully understand/ participate in informed consent process, physical limitations to use mobile device, had lesions on body locations that were difficult to self-monitor and did not have a family member willing to take the image.
32
Diagnostic accuracy and reliability
- Is the diagnostic accuracy and reliability of teledermatoscopy concerning various types of skin cancer comparable with the diagnostic accuracy and reliability found in conventional dermatology?
Elaborated outcomes regarding the diagnostic accuracy and reliability of TDsc are shown in Table 11. Thirteen studies reported the diagnostic reliability of TDsc with a FTF dermatology consultation as the golden standard (35-37, 41, 43, 44, 46, 50, 52, 55-57, 59). No histopathology
confirmation was performed. Three of those studies reported the outcomes in percentage agreement (FTF vs TDsc) and one study reported the Odds Ratio for comparing the first (non-dermoscopic) judgement with the definitive dermatoscopic diagnosis. Further, nine studies reported the outcomes using Kappa values, of which one study assessed the diagnostic reliability of teledermatology versus teledermatology plus the addition of telemicroscopic images which is assessed equal to the magnification of dermatoscopic images, and one study assessed the diagnostic reliability of teledermatology versus TDsc.
Three studies (46, 52, 57) mentioned obtained agreement between TDsc and FTF
ranging from 74% to 94%. Grimaldi et al.(41) reported an OR = 0.179 (P<0.001) for diagnostic
agreement between clinical and TDsc diagnosis.
Nine studies (35-37, 43, 44, 50, 55, 56, 59) obtained Kappa values ranging from 0.37 to 0.91
showing a fair to almost perfect agreement between TDsc and FTF consultations. Three studies explicitly mentioned the obtained diagnostic agreement by clinical (macro) images. Massone et al (46). obtained 89% agreement on clinical images compared to
89% and 94% by dermatoscopic images. Braun et al(56). obtained K = 0.565 on clinical
diagnosis compared to K = 0.742 by TDsc. Warshaw et al (50). obtained a primary diagnostic
agreement of 45.7-75.7% (K=0.32-0.56) on macro images compared to 50.1%-75.3%, K=0.37-0.56 (P = 0.0046) on polarized light dermatoscopic (PLD) images, and 60-80.1%, K=0.52-0.62 (P < 0.022) on contact immersion oil dermatoscopic (CID) images. Thereby, those three studies showed improvements on diagnostic reliability by TDsc. The highest agreement was obtained by the study of Senel et al (36)., where the addition of
telemicroscopic images to teledermatology improved the diagnostic reliability. Livingstone et al (61). mentioned that three lesions physically referred after TDsc thought to be malignant,
which was not confirmed by a secondary care dermatologist.
Fifteen studies reported the diagnostic accuracy of TDsc versus histopathology as the golden standard (13, 36-38, 47-49, 53-60). Studies reported the outcomes either in Kappa values or
percentages. Six studies reported the outcomes in Kappa values, while twelve studies reported the outcomes in percentages.
Studies that reported on the diagnostic accuracy in percentages ranged between 50.7% and 96% (13, 38, 48, 53, 54, 56), irrespective of the specific type of skin cancer. Diagnostic
accuracy specifically assessed for malignant lesions ranged between 60% and 100% (13, 38, 48, 53, 56, 57), while for benign lesions it ranged between 44.44% and 68% (13, 38, 56). Further, Piccoli
et al (54). mentioned that 39.1% of the basal cell carcinomas, 92.4% of the squamous cell
carcinomas, and 77.7% of the malignant melanomas were accurately diagnosed. Piccolo et al (57). mentioned that 95%, 77% and 52% of the melanomas were accurately diagnosed at
low, medium and high diagnostic difficulty. Finally, Kroemer et al (47). noted that 85% (clinical)
vs. 79% (TDsc) were concordant cases, while 15% (clinical) vs. 21% (TDsc) were discordant cases. Overall, these studies show that TDsc is inferior to clinical diagnosis regarding the amount of (dis)concordant cases. Other studies did not report any comparison with FTF
