Vous êtes sur la page 1sur 5

Australasian Journal of Dermatology (2011) 52, 259263

doi: 10.1111/j.1440-0960.2011.00800.x

ORIGINAL RESEARCH

Tele-assessment of Psoriasis Area and Severity Index:


A study of the accuracy of digital image capture
ajd_800

259..263

Philip Singh,1 H. Peter Soyer,1,2 Jason Wu,1,2 Wolfgang Salmhofer3 and Stephen Gilmore1
1

Dermatology Research Centre, The University of Queensland, School of Medicine, and 2Department of
Dermatology, Princess Alexandra Hospital, Brisbane, Australia; and 3Department of Dermatology,
Medical University of Graz, Graz, Austria

ABSTRACT
Background: The implementation of remote Psoriasis Area and Severity Index (PASI) determinations
would greatly enhance the delivery of specialist dermatological care to patients with severe psoriasis
unable to attend face-to-face dermatological consultations. Here we investigate the feasibility of
the remote determination of PASI scores by comparing the results of face-to-face with digital image
assessment.
Methods: Twelve patients with confirmed psoriasis
were recruited for the study. Initially, two dermatologists scored the PASI at the patients usual scheduled
face-to-face visits, at which time standardized digital
images were obtained. PASI scoring based on digital
images was then performed on two separate occasions by three dermatologists with a time-interval
period between assessments, facilitating an assessment of score reproducibility. Linear weighted kappa
statistics were applied to the PASI scores to ascertain
agreement between sets of observations.
Results: While we found a moderate (k = 0.51)
agreement between the face-to-face scores, there was
very good (k = 0.83) agreement between the first
round of telescores and moderate (k = 0.60) agreement between the second round of telescores. Comparison between the face-to-face and telescores
revealed good (k = 0.67 and 0.63) agreement for the
scorers respectively.
Conclusion: We demonstrate that PASI scores can
be determined with moderate to good accuracy by

Correspondence: Dr Philip Dilip Gian Singh, Dermatology


Research Centre, The University of Queensland, School of Medicine, Princess Alexandra Hospital, Brisbane, 4102, Queensland,
Australia. Email: p.singh1@uq.edu.au
Philip Singh, MBBS. H. Peter Soyer, MD. Jason Wu, MBBS.
Wolfgang Salmhofer, MD. Stephen Gilmore, MBBS.
Submitted 7 June 2011; accepted 7 June 2011.

dermatologists using standardized digital images.


Our results imply the implementation of a tele-PASI
service may be a practical and effective adjunct to the
dermatological care of patients with severe psoriasis
where incapacity or distance prevent the realisation
of face-to-face consultations.
Key words: dermatology, PASI, psoriasis, teledermatology, telemedicine.

INTRODUCTION
Psoriasis is a common, chronic, papulosquamous inflammatory skin condition of varying severity, affecting up to 2%
of Western populations.1 Management demands individualized treatment approaches, and while most patients will
only ever require topical therapy, a significant number will
require systemic intervention.1,2 These patients undergoing
systemic therapy require ongoing clinical evaluation and
monitoring for response as well as any development of drug
toxicities.2 With the introduction of government-funded
potent immunomodulating agents, otherwise known as
biologics, the determination of Psoriasis Area and Severity
Index (PASI) scores has become a routine part of dermatologists workload.25 The PASI score is used as the gold
standard for assessing efficacy of systemic agents in the
majority of recently published data on the treatment of psoriasis. Indeed, for patients with severe psoriasis requiring
ongoing treatment with biological therapy, the determination of PASI scores is mandatory in order to objectively
document efficacy.6
Given the documented successes of teledermatology in
general,7 we are motivated to investigate the feasibility of
tele-PASI the remote determination of PASI scores in
patients with psoriasis. For incapacitated patients with
severe psoriasis, and for patients with severe psoriasis where
distances are prohibitive, the implementation of a tele-PASI
service would greatly augment their dermatological care by
permitting the remote introduction or continuation of biological therapy. In addition, such a service would lessen the

2011 The Authors


Australasian Journal of Dermatology 2011 The Australasian College of Dermatologists

260

P Singh et al.

strain on government health budgets since the costs of transport and accommodation would be greatly reduced.8,9

MATERIALS AND METHODS


Twelve patients aged between 22 and 63 years with psoriasis
were recruited consecutively from the Queensland Institute

Table 1

Patient characteristic summary table

Patient

Age

Sex

Site

Psoriasis Type

Mean
PASI score

1
2
3
4
5
6
7
8
9
10
11
12

55
63
27
46
39
22
33
36
29
43
33
58

Male
Male
Male
Male
Female
Male
Male
Female
Female
Male
Male
Male

PAH
QID
QID
QID
PAH
QID
PAH
PAH
QID
QID
QID
QID

Chronic
Chronic
Chronic
Chronic
Chronic
Chronic
Chronic
Chronic
Chronic
Chronic
Chronic
Chronic

12.5
27.1
5.2
11.1
5.2
13.9
36.8
21.4
20.2
24.9
11.9
15.5

Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type
Plaque-Type

PAH, Dermatology Outpatients, Princess Alexandra Hospital,


Brisbane; PASI, Psoriasis Area and Severity Index; QID, Queensland
Institute of Dermatology, Brisbane.

Figure 2 Inter-observer Psoriasis Area and Severity Index (PASI )


scores for the face-to-face examination between clinicians 1 and 2.
Here k = 0.51, indicating moderate agreement. The red disk corresponds to the patient shown in Figure 1. The dashed line represents
perfect agreement.

Figure 1 Example of the digital images used in the calculation of Psoriasis Area and Severity Index (PASI) scores. For this patient, the PASI
scores are represented by the red disk in each panel of Figures 2 and 3, 4 and 5.
2011 The Authors
Australasian Journal of Dermatology 2011 The Australasian College of Dermatologists

Tele-PASI accuracy study

261

(a)

(b)

(c)

(d)

(e)

(f)

Figure 3 Interobserver Psoriasis Area and Severity Index (PASI) scores for the first and second photo evaluations for clinicians 1 and 2 (a,
d), clinicians 1 and 3 (b, e), and clinicians 2 and 3 (c, f ). The red disk in each panel corresponds to the patient shown in Figure 1. Here the
respective agreements are given by k = 0.87 (a), k = 0.88 (b), k = 0.74 (c), k = 0.56 (d), k = 0.72 (e) and k = 0.51 (f ).

of Dermatology and Dermatology Outpatients at the Princess


Alexandra Hospital, both in Brisbane, Australia (see Table 1).
While the diagnosis of psoriasis was required to be established either by dermatologists assessment or by histopathology, sex, severity of disease, and treatment modality
were not exclusion parameters. All patients consented to
the documentation and collation of PASI scores, and the
capture and storage of digital images.
Initially, face-to-face PASI assessments were performed
independently on the same day by two dermatologists, with
the results documented on de-identified standard proformas. A printout of the guidelines for PASI scoring (those
used for submission of biological applications to Medicare
Australia) was made available as reference material at the
time of scoring. Digital images were then taken on this
same initial visit, either by a digital camera alone (Nikon
Coolpix 4500, 4MegaPixel resolution, 2272 1704, Nikon,
Tokyo, Japan ) or by both digital camera and phone camera
(Nokia n95 in-built camera application, 5MegaPixel resolution, 2582 1944, Nokia, Hong Kong) , using the standard-

ized Halpern imaging method.10 The images were then


collated using PDF-creating software so that all images
taken of the one patient (from either the digital camera
alone or both digital and phone camera) were merged into
a single PDF file and treated as one individual image set
(Fig. 1).
At 6 and 14 weeks post face-to-face assessment, the
digital image sets were examined by three dermatologists
(the same two that performed the initial face-to-face assessment and one additional independent clinician) using computer monitors in order to determine PASI scores, again
using the standard proformas. For all dermatologists, the
order of presentation of the image sets in the two timeinterval separated observations was randomized to minimize recall bias.
For our statistical analyses, we performed a linear
weighted kappa analysis using data obtained by reducing
the raw PASI scores from all 96 observations (ranging from
2.7 to 41.9) to integer scores between 1 and 6. With this
coarse graining we thus defined variation in PASI scores up

2011 The Authors


Australasian Journal of Dermatology 2011 The Australasian College of Dermatologists

262

P Singh et al.

to ~6 as equivalent. Following standard protocol, we interpreted kappa agreement scores as follows:11,12 0.0 k < 0.2
(poor), 0.2 k < 0.4 (fair), 0.4 k < 0.6 (moderate),
0.6 k < 0.8 (good), 0.8 k < 1.0 (very good).

(a)

(b)

(c)

(d)

RESULTS
We first examined interobserver agreement using the faceto-face PASI scores (Fig. 2), followed by the first and second
(Fig. 3) digital image PASI scores. We found a moderate
agreement between clinician 1 and clinician 2 who were
involved in the face-to-face assessments (k = 0.51). While
the first round of digital PASI scores revealed an average
kappa value of 0.83, indicating a very good agreement, the
agreement on the second round was moderate (k = 0.60).
Averaging all statistics revealed good overall agreement of
PASI scores between all three dermatologists (k = 0.68).
We next examined the intra-observer agreement of PASI
scores, first by comparing the face-to-face scores with the
first and second (Fig. 4) rounds of digital image PASI scores,
and then by comparing the results of the first and second
rounds of digital image PASI scores (Fig. 5). We found moderate and good agreement between the face-to-face and the
first round of digital image PASI scores (k = 0.41 and 0.64 for
clinician 1 and clinician 2, respectively), and moderate and
good agreement between the face-to-face and the second
round of digital image PASI scores (k = 0.55 and 0.71 for
clinician 1 and clinician 2, respectively). Finally, we found
good, very good and good agreement between the first and
second rounds of digital image PASI scores (k = 0.72, 0.93
and 0.65) for clinicians 1, 2 and 3, respectively).

DISCUSSION

Figure 4 Intra-observer Psoriasis Area and Severity Index (PASI)


scores for the face-to-face and first photo evaluation (a, b) and
face-to-face and second photo evaluation (c, d) for clinicians 1 and
2. The red disk in each panel corresponds to the patient shown in
Figure 1. Here the respective agreements are given by k = 0.41 (a),
k = 0.64 (b), k = 0.55 (c) and k = 0.71 (d).

(a)

The PASI as an assessment tool has been widely studied and


reported with evidence of substantial reproducible reliability for both interrater and intrarater assessments.4,5 Here we
present data showing that PASI scores can be determined
with reasonable accuracy in comparison with face-to-face
assessment by dermatologists using digital images. In addition, we provide data that shows good agreement in the
determination of PASI scores between different dermatolo-

(b)

(c)

Figure 5 Intra-observer Psoriasis Area and Severity Index (PASI) scores for the first and second photo evaluations for clinician 1 (a),
clinician 2 (b), and clinician 3 (c). The red disk in each panel corresponds to the patient shown in Figure 1. Here the respective agreements
are given by k = 0.72 (a), k = 0.93 (b) and k = 0.65 (c).
2011 The Authors
Australasian Journal of Dermatology 2011 The Australasian College of Dermatologists

Tele-PASI accuracy study


gists (both face-to-face and via digital image assessment),
and good agreement in the digital image assessment of the
PASI between the same dermatologist at different times,
findings that are in keeping with those by Furhauf et al. in
their pilot study on teledermatology for high-needs patients
with psoriasis.13
Although our results are encouraging, there are a number
of issues that need to be addressed before tele-PASI could be
implemented as part of mainstream dermatological care.
First, we see quality control and the problems associated
with the standardization of image sets as a potential barrier
to its successful implementation, particularly if patients
photographs are taken by relatives or friends. Here problems associated with inexperience or the use of poorly
equipped cameras for example, under or over-exposure,
poor focus, or inappropriate framing may severely impact
the dermatologists ability to provide an accurate PASI
assessment. We therefore suggest that tele-PASI imaging
should be performed in the context of a general practitioner
consultation, where a trained nurse or assistant is primarily
responsible for the image capture process. Second, the variability of agreement in our results suggests that there are a
number of degrees of freedom of inaccuracy that can play a
role in mitigating against accurate PASI assessments,
including the variation in PASI scores when the same dermatologist looked at the same sets of digital images twice
(Fig. 5a,c), and interobserver variability (Fig. 4). One potential solution to the problem of intra-observer variability is to
explore the possibility of computerized determination of
PASI scores using a suitable artificial intelligence
algorithm.1416 Figure 4 reveals an interesting feature: while
the patterns of PASI scores between face-to-face and the
digital image assessments were similar with respect to the
modality of scoring (compare Fig. 4a with Fig. 4c, and
Fig. 4b with Fig. 4d), they differed with respect to the given
observer (compare Fig. 4a with Fig. 4b, and Fig. 4c with
Fig. 4d). This finding suggests that intra-observer PASI
variation is much less of an issue than interobserver variation (with the caveat that intra-observer variation has been
noted above). It has been suggested in previous research
that the area component of the PASI is responsible for interrater variability,4,5 and this certainly seems a sound explanation for the results seen here in our study.
Although PASI scoring protocols are designed to yield
objective results, our findings indicate that subjectivity will
influence scoring. Whether such limitations can be overcome is an issue for PASI determinations in general. Second,
we note that when PASI scores were compared between the
face-to-face and either the first or second digital image
assessments (Fig. 4), the majority of points found in the four
panels were lying to the left of the perfect agreement diagonal, indicating that tele-PASI assessments consistently
underestimated the results obtained in the clinic, independent of the observer. Although this finding may be due to
foreshortening effects inherent in the two dimensional projection of a surface existing in three dimensions, or a consequence of the difficulties associated with capturing all of
the skin surface in a small number of images, the result
raises the possibility that tele-PASI determinations, in order

263

to agree with those obtained in the clinic, may require


calibrated adjustments. Finally, it is clear that further
studies are required; in particular, it needs to be shown that
tele-PASI can be implemented successfully in the context of
day-to-day general practice, not only from a logistics viewpoint, but also in terms of remote diagnostic accuracy.
To conclude, our results suggest tele-PASI may be a practical method for allowing patients with severe psoriasis to
be managed remotely. Further work is required to validate
its utility in the field.

REFERENCES
1.

2.

3.

4.

5.

6.

7.
8.

9.

10.

11.
12.
13.

14.

15.

16.

Griffiths CEM, Camp RDR, Barker JNWN. Psoriasis. In: Burns


T, Breathnach S, Cox N, Griffiths C (eds.). Rooks Textbook of
Dermatology, 7th edn. Turin: Blackwell Publishing, 2004; 35.1
69. Vol. 2, Chapter 35.
Menter A, Gottlieb A, Feldman SR et al. Guidelines of the care
for the management of psoriasis and psoriatic arthritis. J. Am.
Acad. Dermatol. 2008; 58: 82650.
Schmitt J, Wozel G. The psoriasis area and severity index is the
adequate criterion to define severity in chronic plaque-type
psoriasis. Dermatology 2005; 210: 1949.
Berth-Jones J, Grotzinger K, Rainville C et al. A study examining inter- and intrarater reliability of three scales for measuring severity of psoriasis: psoriasis Area and Severity Index,
Physicians Global Assessment and Lattice System Physicians
Global Assessment. Br. J. Dermatol. 2006; 155: 70713.
Puzenat E, Bronsard V, Prey S et al. What are the best outcome
measures for assessing plaque psoriasis severity? A systematic
review of the literature. J. Eur. Acad. Dermatol. Venereol. 2010;
24 (Suppl 2): 1016.
Australian Government Medicare Australia. 2010, Severe
Chronic Plaque Psoriasis. Sydney New South Wales, Available
from URL: http://www.medicareaustralia.gov.au/provider/
pbs/drugs2/psoriasis.jsp#N10067 (Accessed 30 Oct 2010.).
Burg G (ed.). Telemedicine and Teledermatology. Basel: Karger
Publishing, 2003.
Radtke MA, Augustin J, Blome C et al. How do regional factors
influence psoriasis patient care in Germany. J. Dtsch. Dermatol. Ges. 2010; 8: 51624.
Berger K, Ehlken B, Kugland B et al. Cost-of-illness in patients
with moderate and severe chronic psoriasis vulgaris in
Germany. J. Dtsch. Dermatol. Ges. 2005; 3: 51118.
Halpern AC, Marghoob AA, Bialoglow TW et al. Standardized
positioning of patients (poses) for whole body cutaneous photography. J. Am. Acad. Dermatol. 2003; 49: 5938.
Altman DG. Practical Statistics for Medical Research. London:
Chapman and Hall, 1991.
Landis JR, Koch GG. The measurement of observer agreement
for categorical data. Biometrics 1977; 33: 15974.
Furhauf J, Schwantzer G, Ambros-Rudolph CM et al. Pilot
study using teledermatology to manage high-need patients
with psoriasis. Arch. Dermatol. 2010; 146: 2001.
Savolainen L, Kontinen J, Rning J et al. Application of
machine vision to assess involved surface in patients with
psoriasis. Br. J. Dermatol. 1997; 137: 395400.
Savolainen L, Kontinen J, Alatalo E et al. Comparison of actual
psoriasis surface area and the psoriasis area and severity index
by the human eye and machine vision methods in following
the treatment of psoriasis. Acta Derm. Venereol. 1998; 78:
4667.
Setaro M, Sparavigna A. Quantification of erythema using
digital camera and computer-based colour image analysis: a
multicentre study. Skin Res. Technol. 2002; 8: 848.

2011 The Authors


Australasian Journal of Dermatology 2011 The Australasian College of Dermatologists

Vous aimerez peut-être aussi