Journal of Plastic, Reconstructive & Aesthetic Surgery (2009) 62, 1048e1053

Digital image referral for suspected skin malignancyA pilot study of 300 patients
A. Tadros*, R. Murdoch, J.H. Stevenson
Department of Plastic Surgery, Ninewells Hospital, Tayside, Dundee DD1 7HY, UK Received 19 September 2007; accepted 2 February 2008

KEYWORDS
Electronic referral system; Malignant skin lesion; Reduction outpatient waiting time

Summary Referral of suspect skin cancers as well as non malignant symptomatic skin lesions using high quality digital images transferred via a secure electronic referral system (ERS) potentially offers signicant advantages regarding speed of referral, diagnosis and subsequent treatment over conventional pathways. However concerns over safety of the diagnostic process have been raised. This prospective study looks at 300 patients referred by ERS. A comparison of the diagnoses made from digital images with the diagnoses conrmed on pathology reports for lesions excised is described using a random selection of patients images and referrals. Intra observer analysis of was also assessed. A sample group of patients with lesions deemed as benign, not requiring surgery or other treatment and therefore not seen in secondary care were revisited at a special clinic to determine the safety of the referral system. In this series of 300 patients the study concludes that digital image referral for skin malignancy and other cutaneous lesions reduced the interval between referral and diagnosis by 81% and referral to commencement of treatment in suspect lesions by 30%. Diagnostic accuracy in a random sample of 30 patients was comparable to that reported for patients seen in face to face consultations. High levels of GP and patient satisfaction were recorded. In conclusion digital image referral for skin malignancy and other cutaneous lesions is a safe and cost effective referral pathway, signicantly reducing the interval between referral diagnosis and onset of treatment for skin malignancy. 2008 British Association of Plastic, Reconstructive and Aesthetic Surgeons. Published by Elsevier Ltd. All rights reserved.

There is little doubt that applying information technology to healthcare has the potential to deliver qualitative benets to patient care. Approximately 15% of GP referrals are for skin conditions, of these 4% need secondary referral.1
* Corresponding author. Tel.: 44 7711810786. E-mail address: amirtadros@hotmail.com (A. Tadros).

It has been shown however that most of the suspected urgent skin cancer referrals eventually prove to be benign.2 Effectively this means a delay in true skin malignancy assessment identication rate, as they have to compete with false positive referrals. This delay is compounded even further by the unpredictable period of time from generating the referral letter in the primary care centre

1748-6815/$ - see front matter 2008 British Association of Plastic, Reconstructive and Aesthetic Surgeons. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.bjps.2008.02.005

Digital image referral for suspected skin malignancy until the point it is seen and vetted by the consultant in the secondary care setting. Further delays to the patient journey are encountered in assigning patients to a plastic surgery clinic until nally an operation date is set if surgery is needed. Furthermore due to the wide geographical spread of patients in Tayside, the elderly, who form the highest percentage of patients with skin malignancy, have to make often long journeys to hospital to be seen by a specialist, the consultation often being brief to make a diagnosis. If any of these steps can be bypassed safely, major reductions in waiting times for suspected skin malignancies maybe achieved. A pilot trial conducted by Singh et al3 supported the view that malignant skin lesions can be safely detected using high quality photographic imaging in association with an accurate referral letter. An electronic referral system (ERS) was set up to streamline the skin malignancy referral process between primary and secondary care.3 The aim was to enable true skin malignancies to be diagnosed earlier and subsequently treated sooner, as well as to reduce the pressure on outpatient clinics.

1049 a full letter including history, relevant past medical history and current medication.

Evaluation
In order to accurately evaluate the results of this project, the following parameters were considered:  The outcome (destination) of referrals was assessed.  Diagnostic accuracy (a comparison of diagnosis made from images, with pathology report following excision).  A comparison of the interval between referral to diagnosis, between referral to treatment and between conventional and digital referral pathways.  Patient and GP assessment of the new pathway.

Outcome of referrals
All referrals were initially assessed by one of the four consultants on the unit. The options available to the consultant responsible for the triage of referrals are outlined in Figure 2. The number of referrals per month as well as the destination of these referrals was recorded.

Methods
Design and project objectives
A digital image referral service was established where GPs use the existing electronic referral service to send digital images to the plastic surgery department at Ninewells Hospital of skin lesions including suspected skin cancers. Consultants were given access to the ERS where referrals and images were screened. The consultant on call for a particular day was responsible for screening and vetting the referrals received enabling all patients to be assessed within 24 h. In conjunction with the university medical illustration department, the Cannon powershot camera model A95was selected as the best digital camera for use in this project. Financial support to set up the project (including a project manager) was obtained from the Scottish Executive Centre for Innovation and Change with funding for the camera from primary care CHP. An onsite-training package was established for GPs in participating practices on the use of the digital cameras and the attachment of images to the referral letter electronically (Figure 1). A helpline formed part of the package with feedback to the participants on quality of images submitted. GPs were given a precise protocol for taking digital images. All images referred to the plastic surgery department were accompanied by

Diagnostic accuracy (Inter/Intra observer reliability)

This was assessed by two methods. a) A random sample of 30 patients was selected by an independent observer; a le was created including the GP referral letter and digital image. This patient number was chosen following discussion with a statistician. Each of the four consultants involved with the triaging of images reviewed this sample on three separate occasions with a minimum of 2 weeks apart. On average each review lasted for 5 min. Each consultant was asked to state the main diagnosis, up to two possible differential diagnoses and the level of urgency for treatment (urgent- within 2 weeks, soon- up to 6 weeks and routine- over 6 weeks). Observer consistency (comparing the diagnosis made on each of the three occasions) as well as interobserver consistency was assessed statistically. Retrieval of patient notes was only carried out after COREC ethical approval was granted for this section of the study. The level of intra and inter observer agreement and reliability was analysed using Kappa statistics (SPSS output, for Windows v.6). The approximate significance of the kappa statistics is provided as standard in our prism output. Kappa value is an index which compares the agreement against that which might be expected by chance. Kappa can be thought of as the chance-corrected proportional agreement and possible value range from 1 (perfect agreement) via 0 (no agreement above that expected by chance) to 1 (complete disagreement). This gives a more meaningful indication of agreement than the calculation of simple percentage agreement, as the latter takes no account of the level of agreement that might arise by chance alone.

Each practice issued with Digital camera Media card reader Batteries and charger Patient consent forms Patient information leaflet Scale labels 2 photographs required for each referral - One close up - One wide shot Scale labels must include patient unit number and name One hour training session with GPs and administrative staff on site -

Figure 1

Established training packages for active GP sites.

1050
GP referral Dundee Angus

A. Tadros et al.

Perth & Kinross

NE Fife

On call plastic surgery consultant

Diagnosis/ management

Reassure Repeat photo after interval OPD appointment Redirect referral

Straight onto OPLA list

Surgery by GP Surgery by plastic surgeon ( trainee, consultant) Urgency of treatment Duration of procedure Venue (outpatient clinic, day surgery unit, main Theatre)

Electronic reply to GP

Communication with patient

Audit (safety of system)

Medical records for statistics

Figure 2 Electronic referral pathway showing the sequence of events following the electronic patient journey once the referral is viewed by the consultant.

b) A random sample of 20 patients deemed by the triage process to have lesions that were benign and not requiring any treatment were seen at a special clinic, where a diagnosis was made, and comments from patients recorded. A letter was sent to each explaining the project and the reason for them being sent to attend the clinic. A copy of this was sent to the GP in case the patient had anxieties about the diagnostic process.

the ERS pathway and then progressed to either a one stop or day surgery appointment.

Results
Outcome of referrals
The rst 300 patients were included in the study. The majority of digital image referrals were seen in the 1-stop clinic (41%) for further assessment and immediate excision if needed. A signicant number of patients however were diagnosed with benign lesions and electronic reply being sent to the GP outlining the diagnosis, with either no treatment being advised or management by the GP where appropriate e.g. solaraze gel or cryotherapy for a simple keratosis (Figure 3).

Assessment of the interval between referral to diagnosis and referral to treatment

The mean interval between initial referral from the GP, vetting the referral letter and wait until treatment was assessed using the conventional pathway (standard letter, public mail, etc) using medical records data. Comparative gures using the digital image referral pathway were recorded and a comparison carried out.

Diagnostic accuracy
The mean overall correct diagnosis between the four consultants was 83.25%. A total of 90.25% of malignant lesion (true positive) and 76.58% of benign lesions were correctly identied by the four clinicians (Table 1). No malignant lesions were allocated to the no treatment group. One Malignant lesion was incorrectly diagnosed by

Referring practitioner and patient satisfaction

A questionnaire was sent out to a random selection of 34 referring GPs who had referred a patient through the ERS pathway. A separate questionnaire was also sent to a random sample of 31 patients who were referred through

Digital image referral for suspected skin malignancy

Total number of referrals

1051
Table 2 (p < 0.05) Kappa indices, interobserver agreement

Clinician 1 Clinician 2 Clinician 3 Clinician 4

1-stop La daycase Inpatient GP Outpatient

Clinician 1 e Clinician 2 e Clinician 3 e

0.59 e e

0.62 0.47 e

0.64 0.56 0.52

Figure 3 The total number of patients using ERS. Out of 300 patients, 122 patients attended 1-stop clinic, 80 patients were added to the local anaesthetic day surgery list, 40 patients were discharged with a letter to the primary care practitioner, 57 patients attended the outpatient department and one patient was admitted directly for surgery as an inpatient.

reduction in time from referral of patients with a suspected skin malignancy to vetting, 10.9 days to 2 days (81%) Figure 4. Following the initial vetting the average waiting time in the patient pathway from clinic until treatment was similar in both groups (Table 3); however when the additional waiting time from vetting to a clinic appointment is added a signicant (30%) reduction from initial referral to treatment is observed (Figure 5).

Referring practitioner and patient satisfaction

GP questionnaire- Overall 25/34 (74%) questionnaires were returned. (88%) felt that the feedback received from the plastic surgeons following electronic referral has been educationally valuable and has enhanced the patient journey (84%). A few examples of the comments raised are: Still struggling a little re good close-up shots! Otherwise excellent service and Overall a very good service, a real improvement in patient care with a very palpable improvement in communication. Patient questionnaire- Overall 21/31 (68%) of patients responded. All patients were happy to have photographs taken in GP surgery and sent electronically to the hospital. None of the patients were unhappy about the overall treatment. Examples of the comments are: I received excellent service throughout this procedure.

three out of four clinicians; this was a cutaneous metastatic deposit from a primary renal clear cell carcinoma. It was commented that the images of this particular referral were of poor quality. The patient was still listed as urgent or soon by the clinicians as there was doubt in the diagnosis. Interobserver agreement between the four clinicians was signicant with kappa index ranging between 0.47 and 0.64 (p< 0.05) Table 2.

Re-evaluation of patients discharged by ERS

20 patients were randomly selected from a list of those vetted as not requiring either surgical excision or treatment in secondary care. A total of 18 patients attended the special clinic. No patients discharged via ERS triaging were found to have skin malignancy. A number of patients attending the clinic were anxious about other skin lesions were keen for reassurance and a number of patients were requesting excision of the benign lesions (despite reassurance) for cosmetic purposes.

Discussion
Digital image referral triage and diagnosis potentially offers signicant advantages to primary and secondary care practitioners, as well as, patients involved in the management of skin malignancy and lesions causing concern to

Interval between referral and diagnosis/ treatment

Assessment of the patient journey using conventional compared to the electronic pathway revealed a signicant
Table 1 Accuracy of diagnosis; Average correct diagnosis rates between all 4 clinicians, true positive values are measured as sensitivities, true negative values are measured as specicity Clinician 1 Clinician 2 Clinician 3 Clinician 4 Sensitivity 0.827 Specicity 0.78 PPV 0.67 NPV 0.82 Correct 87.3 diagnosis 0.819 0.68 0.78 0.77 78.9 0.842 0.84 0.64 0.81 83.1 0.804 0.76 0.75 0.8 80.2

30

20

No of days
10

*
0 Digital letter

referral

Figure 4

Average waiting time from referral to vetting.

1052
Table 3 Patient average waiting times following clinician vetting of referral Urgent Letter Referral/clinic Clinic/procedure Total ERS Referral/procedure Total 27 32 59 Soon 74 52 126 Routine 223 80 303

A. Tadros et al. particularly important in malignant melanoma. In our series, the melanomas diagnosed demonstrated key features enabling rapid diagnosis and treatment from the images alone. Where these features are not present, then urgent referral-where malignancy is suspected-to a specialist will allow closer examination using these modalities. Again, the digital referral pathway is likely to reduce the interval between referral, diagnosis and treatment. In the special clinic set up to review a random selection of patients designated as having benign lesions, no skin malignancy was found. As a safety net GPs can be advised, if there is any uncertainty in the diagnosis to refer the patient with a repeat image after an appropriate interval eg 3 months, for comparison with the original picture. Unlike other investigators,9 we feel that the safeguards introduced into the study, and the continuing audit of the pathway, should ensure safety in diagnosis and management pathways. Standardisation of cameras, images and views taken, as well as accuracy in the accompanying documentation should reduce the risk of poor digital images or inappropriate information, which has been highlighted as a problem in other studies.16 Access to a helpdesk has proven invaluable in maintaining the high quality of images submitted, and resolving any technical problems arising between primary and secondary care. In summary, this pilot has highlighted that digital image referral for skin cancer and other cutaneous lesions can be safe, cost effective and offers considerable advantages to both GPs and patients with high satisfaction rates demonstrated in both groups.

28 28

52 52

65 65

patients.4e6 For malignant melanoma early diagnosis and treatment is crucial as the treatment options in advanced melanoma are still very limited. The signicantly shortened interval between referral, diagnosis and treatment highlighted in this paper is an illustration of how the early management of patients with suspected and actual melanomas can be achieved, and examples are given in this paper. Basal and squamous cell carcinomas, the other common skin malignancies affect predominantly an older age group who often nd travel to and from hospital difcult for what is often a brief consultation to make a diagnosis. Digital image referral can reduce the number of visits to hospital for this group of patients.8 Legitimate concerns have been raised, however, regarding the safety of this pathway as well as theoretical advantages in saving on patient journeys to hospital.9 Several studies have looked at observer diagnostic accuracy for skin lesions/malignancies.7,10e14 The diagnostic accuracy in this series in a random selection of patients is as good as (and in some cases exceeds) that of face to face consultation. Importantly, the ability to differentiate between malignant and benign lesions has been clearly demonstrated. In one case only of a skin malignancy was the diagnosis uncertain, but each of the clinicians still allocated this for urgent or soon excision. Enhanced diagnostic accuracy using dermatoscopy and more recently siascopy has been demonstrated and is particularly valuable in those lesions where there is doubt about the diagnosis.14,15 This is
400 ERS Letter 300

References
1. Leggett P, Gilliland AE, Cupples ME, et al. A randomized controlled trial using instant photography to diagnose and manage dermatology referrals. Fam Pract 2004;21:54e5. 2. Kerr AC, Leonard S, Gupta G. A prospective survey of skin cancer referrals to a Scottish dermatology department. Br J Dermatol 2005;152:1065e6. 3. Singh S, Stevenson JH, McGurty D. An evaluation of Polaroid photographic imaging for cutaneous-lesion referrals to an outpatient clinic: a pilot study. Br J Plas Surg 2001;54:140e3. 4. Phillips RR. Photography as an aid to dermatology. Med Biol Illus 1976;26:161e6. 5. Murray CK, et al. The remote diagnosis of malaria using telemedicine or e-mailed images. Mil Med 2006 Dec;171: 1167e71. 6. Levy JL, Trelles MA, Levy A, et al. Photography in dermatology: comparison between slides and digital imaging. J Cosmet Dermatol 2003 Jul;2:131e4. 7. Herrmann FE, Sonnichsen K, Blum A. Teledermatology versus consultationsea comparative study of 120 consultations. Hauttarzt; 2005 Mar::942e8. 8. Braeunling F, Jones M, Lister RK, et al. Digital photography enhances the prioritization of suspected skin cancer referrals. Br J Dermatol 2004;151:27. 9. Bowns I, Collins K, Walters S, et al. Telemedicine in dermatology: a randomised controlled trial. Health Technol Assess 2006; 10:iiieiv. 10. Hallock G, Lutzg D. Prospective study of the accuracy of the surgeons diagnosis in 2000 excised skin tumours. Plast Reconstr Surg 1988 Apr;101:1255e61. 11. Ew E, Giorlando F, Su S, et al. Clinical diagnosis of skin tumours: How good are we? ANZ J Surg 2005 Jun;75:415e20.

No of days

200

100

urgent

urgent

soon

soon

routine

routine

Vetting type

Figure 5

Average waiting time from referral to procedure.

Digital image referral for suspected skin malignancy

12. Har-Shai Y, Hai N, Taran A, et al. Sensitivity and positive predictive value of presurgical clinical diagnosis of excised benign and malignant skin tumours: a prospective study of 835 lesions in 778 patients. Plast Reconstr Surg 2001 Dec;108:1982e9. 13. Green A, Leslie D, Weedon D. Diagnosis of skin cancer in the general population: clinical accuracy in the Nambour survery. Med J Aust 1988 May 2;148:447e50. 14. Tehrani H, Wall J, Price G, et al. A prospective comparison of spectrophotometric intracutaneous analysis to clinical

1053
judgment in the diagnosis of nonmelanoma skin cancer. Ann Plast Surg 2007 Feb;58:209e11. 15. Moncrieff M, Cotton S, Claridge E, et al. Spectrophotometric intracutaneous analysis: a new technique for imaging pigmented skin lesions. Br J Dermatol 2002 Mar;146:448e57. 16. Mahendran R, Goodeld M, Sheehan-Dare RA. An evaluation of the role of a store-and-forward teledermatology system in skin cancer diagnosis and management. Clin Exp Dermatol 2005 May;30:209e14.