Health-Related Quality of Life Assessment in Dermatology: Interpretation of Skindex-29 Scores Using Patient-Based Anchors
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
ORIGINAL ARTICLE
Health-Related Quality of Life Assessment in
Dermatology: Interpretation of Skindex-29 Scores
Using Patient-Based Anchors
Cecilia A. C. Prinsen1, Robert Lindeboom2, Mirjam A. G. Sprangers3, Catharina M. Legierse1
and John de Korte1
In dermatology, the clinical use of health-related quality of life (HRQL) scores is impeded by lack of empirically
and clinically based interpretation of these scores. We aimed to facilitate the interpretation of Skindex-29
domain and overall scores by identifying clinically meaningful cut-off scores, using patient-based anchors.
Consecutively included dermatology outpatients completed the Skindex-29 and four sets of anchor-based
questions, such as questions on the impact of skin disease on HRQL, on global disease severity, and on
psychiatric morbidity. Pearson’s correlations and receiver operating characteristic analysis were used to identify
the optimal Skindex-29 cut-off scores corresponding to severely impaired HRQL. A total of 339/434 patients
completed the questionnaires (response rate 78%), of which 322 could be used for data analysis. Cut-off scores
associated with the patient-based anchors on the impact of skin disease on HRQL showed the highest accuracy
(area under the curve ranged from 0.83 to 0.91). The corresponding Skindex-29 cut-off scores for severely
impaired HRQL were as follows: X52 points on symptoms, X39 on emotions, X37 on functioning, and X44 on
the overall score. The estimated cut-off scores can be used in clinical practice to identify patients with (very)
severely impaired HRQL.
Journal of Investigative Dermatology advance online publication, 24 December 2009; doi:10.1038/jid.2009.404
INTRODUCTION emotions, and functioning. The domain scores and an overall
Health-related quality of life (HRQL) reflects patients’ score are expressed on a 100-point scale, with higher scores
evaluation of the impact of disease and treatment on their indicating lower levels of quality of life (Chren et al., 1996,
physical, psychological, and social functioning and well- 1997a, b; De Korte et al., 2002). However, a score in itself
being (Essink-Bot and Haes de, 1996; Testa and Simonson, has little or no direct meaning and cannot be interpreted in a
1996). In clinical practice, HRQL is considered to be an aid straightforward manner.
for clinical decision making, monitoring the therapeutic Two types of methods to establish a clinically meaningful
process, communicating with the patient, and evaluating interpretation of HRQL scores exist: distribution-based and
treatment outcome (Guyatt et al., 1993, 2002). A good anchor-based methods. Distribution-based methods rely on
understanding of the concept of HRQL and a correct the score distributions of clinically distinct subgroups of
interpretation of HRQL scores are essential. patients (Guyatt et al., 2002). A categorization of Skindex-29
The well-established Skindex-29 is a three-dimensional, scores using this method was recently published (Nijsten
dermatology-specific HRQL questionnaire. Twenty-nine et al., 2009). However, in this study, patient-based anchors
items are combined to form three domains: symptoms, were not used and further research was suggested. Anchor-
based methods examine the relationship between scores on
an HRQL instrument and an independent measure or anchor
1
Department of Dermatology, Academic Medical Center, Amsterdam, (Guyatt et al., 2002). This method was developed to estimate
The Netherlands; 2Department of Clinical Epidemiology, Biostatistics and
clinically meaningful cut-off scores for HRQL instruments to
Bioinformatics, Academic Medical Center, Amsterdam, The Netherlands and
3
Department of Medical Psychology, Academic Medical Center, Amsterdam, allow clinicians to interpret scores more straightforward (e.g.,
The Netherlands a score of X44 indicates severely impaired HRQL). An
Correspondence: Cecilia Anna Catharina Prinsen, Department of anchor should be itself interpretable and should at least
Dermatology, Academic Medical Center, PO Box 22660, Amsterdam 1100 moderately correlate with the HRQL instrument under study
DD, The Netherlands. E-mail: c.a.prinsen@amc.uva.nl
(Guyatt et al., 1993; Norman et al., 2001). With respect to
Abbreviations: AUC, area under the curve; GDS, global disease severity;
dermatology-specific questionnaires, the interpretation of
GHQ, General Health Questionnaire; GQ, general question;
HRQL, health-related quality of life; ROC, receiver operating characteristic scores by using a patient-based anchor was previously
Received 17 April 2009; revised 12 November 2009; accepted 13 November studied for the Dermatology Life Quality Index (Finlay and
2009 Khan, 1994; Hongbo et al., 2005).
& 2009 The Society for Investigative Dermatology www.jidonline.org 1CAC Prinsen et al.
HRQL Assessment in Dermatology
In this study, we aim to determine Skindex-29 domain and patients were overrepresented (440%), we also examined
overall cut-off scores using different patient-based anchors the cut-off scores of this subgroup and the cut-off scores of the
(Andersen and Newman, 1973). entire study population without this subgroup. The cut-off
scores of the subgroup analyses did not significantly differ
RESULTS from the presented cut-off scores of the entire study
Study population population. The resulting cut-off scores for eczema patients
At nine outpatient dermatology clinics in the Netherlands, on the anchors relating to the impact of disease on HRQL and
434 patients were asked to complete the questionnaires after on disease severity are higher than the study results
informed consent was obtained. A total of 339 patients presented, but those for psychiatric morbidity were similar
returned the questionnaires (response rate 78%). Seventeen (data not shown).
patients were excluded from data analysis as X25% of the
Skindex-29 items were missing, leaving 322 patients for DISCUSSION
analysis. In these patients, only 0.4% of the Skindex-29 items We aimed to facilitate the interpretation of Skindex-29 scores
had to be imputed. by determining clinically important cut-off scores. We found
The 95 non-respondents did not significantly differ from robust Skindex-29 cut-off scores with all three patient-based
respondents with regard to gender, but they were younger anchors expected to indicate severely impaired HRQL: GQs
(45.2 vs 49.5 years). Table 1 shows the characteristics of the on the impact of the skin disease on HRQL, patients’
study population, their Skindex-29 scores, their scores on assessment of disease severity, and the presence of psychia-
disease severity, and their scores on the 12-item General tric morbidity as measured with the GHQ-12. Except for the
Health Questionnaire (GHQ-12) at baseline. In this study Skindex-29 functioning domain using the GHQ anchor, the
population, the prevalence of psychiatric morbidity was cut-off scores were highly comparable.
24.4%. In case of more than one dermatological condition, In clinical practice, the primary focus should be on the
the diagnosis that had bothered the patient the most during profile of the three domain scores, as these scores will
the past week was taken as the diagnosis. provide clinicians with information on which domain of
HRQL bothered the patient the most. The overall score of the
Patient-based anchors Skindex-29 should be interpreted with some caution, as the
Correlations were calculated for the Skindex-29 domain and validity of the overall score as such is debatable. Patients with
overall scores versus four sets of anchor-based questions. scores equal to or above the presented cut-off scores in at
Five anchor-based questions (the four general questions least one of the three domains are significantly affected by
(GQs), the question on global disease severity (GDS)), and their skin disease. These scores may signal a need for
the GHQ-12 had a correlation of X0.40 (range 0.42–0.79) (adjustment of current) treatment and/or for additional care or
with the relevant Skindex-29 domain and overall scores, support. However, they do not automatically indicate what
and thus met the requirements for a patient-based anchor kind of treatment, care or support is appropriate: the specific
(see Supplementary Table S1 online). Low correlations were needs of an individual patient should be explored in direct
found for seven anchor-based questions with regard to contact with the patient. HRQL scores may also facilitate
patients’ treatment needs. Therefore, these questions were doctor–patient communication and mutual decision making
excluded from further analysis. (Velikova et al., 2004). With the formal external anchors on
disease severity and psychiatric morbidity, we were able to
Skindex-29 cut-off scores evaluate and confirm the robustness of the given cut-off
We established cut-off scores for ‘‘severe to very severe’’ scores on impaired HRQL.
impact of disease on HRQL (further referred to as ‘‘severe’’). The low correlation between the Skindex-29 domain and
Table 2 shows the estimated cut-off scores associated with overall scores and the patient-based anchors with regard to
severely impaired HRQL, severe disease severity, and patients’ treatment needs indicates that HRQL and treatment
psychiatric morbidity for the Skindex-29 domain and overall needs are likely to be two different constructs.
scores. The Skindex-29 domain and overall cut-off scores Interestingly, our results with respect to psychiatric
associated with the patient-based anchors relating to the morbidity are consistent with the results of a previous study,
impact of disease on HRQL showed the highest accuracy: the thereby giving further evidence for a relatively high pre-
area under the curve (AUC) ranged from 0.83 to 0.91. The valence of psychiatric morbidity among dermatological
AUC for the anchor on disease severity ranged from 0.69 to patients (Sampogna et al., 2004).
0.76, and for psychiatric morbidity from 0.73 to 0.83. Earlier results on the categorization of Skindex-29 scores
The optimal and, according to the AUC statistic, most by Nijsten et al. (2009), using a distribution-based method to
accurate Skindex-29 cut-off scores for severely impaired establish a clinically meaningful interpretation of Skindex-29
HRQL were as follows: symptoms X52, emotions X39, scores, were similar with respect to the results of our study for
functioning X37, and for the overall score X44 points. the functioning domain and the overall score, but different for
the symptoms and emotions domains. This may, in part, be
Subgroup analyses the result of differences in the distribution of diagnoses and
We have performed subgroup analysis for psoriasis patients disease severities of the samples and also of the statistical
(N ¼ 138) and patients with eczema (N ¼ 76). As psoriasis methods used to derive cut-off scores.
2 Journal of Investigative DermatologyCAC Prinsen et al.
HRQL Assessment in Dermatology
Table 1. Baseline characteristics of the study Table 1. Continued
population (N=322) SD
n % Mean (minimum–maximum)
Global disease severity2 (N=312)3 3.2 NA
Male gender 146 45.3
GHQ-124 score (N=303)3 2.5 NA
Mean age in years (SD, range) 49.5 (17.4, 18–85)
Psychiatric morbidity (N=74) 8.2 NA
Abbreviation: NA, not applicable.
1
Diagnoses, n (%) The domain scores and the overall score are expressed on a 100-point
scale, with higher scores indicating a lower level of quality of life.
Acne, other disorders 19 5.9 2
Global Disease Severity (GDS): one question on patients’ perception of
of sebaceous, apocrine, the degree of global severity of the skin disease.
or eccrine glands 3
Different sample sizes are because of missing values.
4
Autoimmune disorders 5 1.6 GHQ-12: the 12-item General Health Questionnaire designed to
measure psychiatric morbidity.
Benign pigmented lesions 1 0.3
and naevi
Benign skin and 2 0.6
vascular tumors Table 2. Skindex-291 cut-off scores for severely
Decubitus 1 0.3 impaired health-related quality of life
Eczematous lesions 76 23.6 Patient-based anchors
Genetic disorders 4 1.2 Cut-off score Sensitivity Specificity AUC
Genital skin disorders 3 0.9 Impact on HRQL
Granuloma annulare 1 0.3 Symptoms (r=0.54) X52 0.67 0.82 0.83
Hair and scalp disorders 3 0.9 Emotions (r=0.73) X39 0.72 0.92 0.88
Infection of skin transplant 1 0.3 Functioning (r=0.79) X37 0.83 0.88 0.91
after trauma
Overall (r=0.75) X44 0.82 0.85 0.90
Jessner–Kanoff lymphocytic 1 0.3
infiltrate
Lichen sclerosus 9 2.8 Disease severity
Non-melanoma skin cancers 17 5.3 Symptoms (r=0.42) X52 0.70 0.63 0.69
and premalignant lesions Emotions (r=0.46) X35 0.70 0.67 0.74
Pigmentary disorders 6 1.9 Functioning (r=0.44) X42 0.93 0.45 0.74
Pityriasis lichenoides 1 0.3 Overall (r=0.51) X39 0.81 0.62 0.76
chronica
Pruritus 4 1.2
Psychiatric morbidity2
Psoriasis 138 42.9
Symptoms (r=0.42) X55 0.64 0.71 0.73
Reactive skin disorders and 4 1.2
drug reactions Emotions (r=0.55) X39 0.78 0.71 0.81
Skin malignancies not 3 0.9 Functioning (r=0.57) X28 0.80 0.72 0.81
otherwise specified Overall (r=0.60) X42 0.74 0.81 0.83
Superficial fungal infections 1 0.3 Abbreviation: AUC, area under the curve.
1
Ulcers 4 1.2 The domain scores and the overall score are expressed on a 100-point
scale, with higher scores indicating a lower level of quality of life.
Urticarial disorders 9 2.8 2
Five or more points on the 12-item General Health Questionnaire.
Varicose veins 1 0.3
Viral skin lesions 8 2.5
Two limitations of this study merit attention. First, an
SD unexpectedly high number of psoriasis patients (440%) were
Mean (minimum–maximum) included in the sample. However, the results of the subgroup
Skindex-291 score analyses did not significantly differ from the presented cut-off
Symptoms 46.7 22.3 (0–96.4) scores. Nevertheless, further research using similar techni-
ques in other dermatoses is recommended.
Emotions 37.6 22.0 (0–100.0)
Second, owing to the relatively small sample sizes per
Functioning 26.4 21.2 (0–97.9) diagnostic category, subgroup analysis could only be mean-
Overall 35.2 19.0 (0–97.9) ingfully performed for psoriasis patients (N ¼ 138) and
www.jidonline.org 3CAC Prinsen et al.
HRQL Assessment in Dermatology
patients with eczema (N ¼ 76). Further research on the of correlation: rX0.40), Pearson’s correlations were calculated
generalizability of the established cut-off scores, particularly between Skindex-29 domain and overall scores and the anchor-
in specific diagnostic categories, is recommended. based questions: GQ1, GQ2, and GQ3 were related to the Skindex-
We conclude that the estimated cut-off scores of the 29 domain scores for symptoms, emotions, and functioning,
Skindex-29 can be used in clinical practice to identify respectively, whereas GQ4 was related to the overall Skindex-29
patients with (very) severely impaired HRQL. score. The score on GDS, the scores on the seven questions with
regard to patients’ treatment needs, and the GHQ-12 score were
MATERIALS AND METHODS related to both the Skindex-29 domain and overall scores. All
Setting and study population questions referred to the past week.
We conducted a multi-center, cross-sectional study in dermatology ROC-curve analysis was then used to determine optimal Skindex-
outpatients with unselected chronic skin disease. Patients were 29 cut-off scores for the selected anchors (Turner et al., 2009). The
consecutively recruited at nine dermatology outpatient clinics during ROC–AUC indicates the overall accuracy of the Skindex-29 cut-off
a predetermined period of 4 weeks (14 April to 9 May 2008). Patients scores; a higher value indicates a better discriminating capacity of a
eligible for this study had a chronic skin disease and were 18 years or given Skindex-29 cut-off score to distinguish patients, for instance,
older. Excluded were patients who were mentally and/or physically with and without impaired HRQL (Streiner and Norman, 2003). For
unable to complete the questionnaires and patients with insufficient the construction of the ROC-curves, the five-category anchor variables
mastery of the Dutch language. Patients who gave their written of the Skindex-29, namely, (1) never, (2) seldom, (3) sometimes, (4)
informed consent during their visit at the dermatology outpatient often, and (5) all the time, were dichotomized using ratings 1–3 vs 4–5
clinics were asked to complete the questionnaires independently for severe and very severe impairment of HRQL. For the GHQ-12, the
and to return the completed package by using a stamped return presence of psychiatric morbidity was indicated by a score of five
envelope. The central Ethics Committee AMC (EC AMC) exempted points or more (Picardi et al., 2000; Sampogna et al., 2004).
this study for ethical approval. For non-interventional questionnaire Cut-off scores were rounded to zero decimal places. The Youden
research, this is common policy in the Netherlands. A written Index was used to determine the optimal balance between sensitivity
confirmation of this policy was given by the EC AMC. The study was (true positive rate) and specificity (true negative rate) in the
conducted according to the Declaration of Helsinki Principles. estimation of the Skindex-29 cut-off scores (Fluss et al., 2005). All
analyses were run under SPSS, (Chicago, IL), version 16.0.
Measurements
As it is strongly recommended to use multiple independent anchors to CONFLICT OF INTEREST
examine cut-off scores (Guyatt et al., 2002), the questionnaires The authors state no conflict of interest.
comprised of the Skindex-29 and four sets of anchor-based questions:
(i) four GQs, evaluating the impact of the skin disease on the three ACKNOWLEDGMENTS
We thank all dermatologists whose collaboration made the study possible:
domains of the Skindex-29 (GQ1-3) and on overall impairment of HRQL
M.T.W. Gaastra, MD, Flebologisch Centrum Oosterwal, Alkmaar; D.B. de
(GQ4); (ii) one question on patients’ perception of the degree of global Geer, MD, Diakonessenhuis, Zeist; A.Y. Goedkoop, MD, PhD, St Antonius
severity of the skin disease; (iii) seven questions on patients’ treatment Hospital, Nieuwegein; C.L.M. van Hees, MD, Reinier de Graaf Group,
needs; and (iv) the GHQ-12 consisting of 12 items and designed to Voorburg; W.J.A. de Kort, MD, Amphia Hospital, Breda; M.C.G. van Praag,
MD, PhD, St Franciscus Gasthuis, Rotterdam; M.L.A. Schuttelaar, MD,
measure psychiatric morbidity, usually depressive or anxiety disorders University Medical Center Groningen, Groningen; A.M.E. Visser-Van Andel,
(Goldberg, 1972; Koeter and Ormel, 1991; Picardi et al., 2001). MD, Gelderse Vallei Hospital, Ede, The Netherlands. Furthermore, we
A pilot study among seven patients of the Academic Medical acknowledge the contributions of F.J. Oort, PhD, and B. King-Kallimanis, MSc,
Center was performed to test whether there was any difficulty or from the Academic Medical Center, Department of Medical Psychology,
Amsterdam, for their contribution to this study.
ambiguity in the wording of the anchor-based questions, with the
exception of the standardized GHQ-12.
SUPPLEMENTARY MATERIAL
Statistical analysis Supplementary material is linked to the online version of the paper at http://
Sample size calculations were based on the precision of the www.nature.com/jid
estimates of the receiver operating characteristic – area under the
curve (ROC–AUC) statistic. The ROC–AUC statistic expresses the REFERENCES
strength of the relation between the Skindex-29 domain and overall Andersen R, Newman JF (1973) Societal and individual determinants of
medical care utilization in the United States. Milbank Mem Fund Quart
scores, and the anchor-based question dichotomizations (see below) 51:95–124
to indicate severe impairment of HRQL. A total of 300 patients were
Chren MM, Lasek RJ, Quinn LM et al. (1996) Skindex, a quality-of-life
needed to have ±10% error margins around the validity coefficients measure for patients with skin disease: reliability, validity, and
under conventional study reliability requirements (5% type I error responsiveness. J Invest Dermatol 107:707–13
rate, 80% power), assuming a 3:1 ratio between the test negative and Chren MM, Lasek RJ, Flocke SA et al. (1997a) Improved discriminative and
test positive groups and a correlation of 0.60 between tests. Patients evaluative capability of a refined version of Skindex, a quality-of-life
instrument for patients with skin diseases. Arch Dermatol 133:1433–40
were excluded from data analysis in case X25% of the items of the
Chren MM, Lasek RJ, Quinn LM et al. (1997b) Convergent and discriminant
Skindex-29 were missing. In case of o 25%, the missing values were
validity of a generic and a disease-specific instrument to measure quality
imputed with the mean of the domain scores per patient. of life in patients with skin disease. J Invest Dermatol 108:103–7
To assess whether the anchor-based questions met the formal Essink-Bot ML, Haes de JCJM (1996) Kwaliteit van Leven in Medisch
characteristics of a patient-based anchor (at least a moderate degree Onderzoek. Amsterdam: University Press, The Netherlands
4 Journal of Investigative DermatologyCAC Prinsen et al.
HRQL Assessment in Dermatology
Finlay AY, Khan GK (1994) Dermatology Life Quality Index (DLQI): Norman GR, Sridhar FG, Guyatt GH et al. (2001) Relation of distribution- and
a simple practical measure for routine clinical use. Clin Exp Dermatol anchor-based approaches in interpretation of changes in health-related
19:210–6 quality of life. Med Care 39:1039–47
Fluss R, Faraggi D, Reiser B (2005) Estimation of the Youden Index and its Picardi A, Abeni D, Pasquini P (2001) Assessing psychological distress in
associated cutoff point. Biom J 47:458–72 patients with skin diseases: reliability, validity and factor structure of the
Goldberg DP (1972) The Detection of Psychiatric Illness by Questionniare. GHQ-12. J Eur Acad Dermatol Venereol 15:410–7
Oxford: University Press, London Picardi A, Abeni D, Melchi CF et al. (2000) Psychiatric morbidity in dermatological
Guyatt GH, Feeny DH, Patrick DL (1993) Measuring health-related quality of outpatients: an issue to be recognized. Br J Dermatol 143:983–91
life. Ann Intern Med 118:622–9 Sampogna F, Picardi A, Chren MM et al. (2004) Association between poorer
Guyatt GH, Osoba D, Wu AW et al. (2002) Methods to explain the quality of life and psychiatric morbidity in patients with different
clinical significance of health status measures. Mayo Clin Proc dermatological conditions. Psychosom Med 66:620–4
77:371–83 Streiner DL, Norman GR (2003) Health measurement scales: a practical guide
Hongbo Y, Thomas CL, Harrison MA et al. (2005) Translating the science of to their development and use. In: From Items to Scales. Oxford
quality of life into practice: what do dermatology life quality index scores University Press: New York. pp 102–25
mean? J Invest Dermatol 125:659–64 Testa MA, Simonson DC (1996) Assesment of quality-of-life outcomes. N Engl
Korte de J, Mombers FM, Sprangers MA et al. (2002) The suitability of quality- J Med 334:835–40
of-life questionnaires for psoriasis research: a systematic literature Turner D, Schünemann HJ, Griffith LE et al. (2009) Using the entire cohort in
review. Arch Dermatol 138:1221–7 the receiver operating characteristic analysis maximizes precision of the
Koeter MWJ, Ormel J (1991) General Health Questionnaire - Instruction minimal important difference. J Clin Epidemiol 62:374–9
Manual. Dutch version, Lisse: Swets and Zeitinger BV, the Netherlands Velikova G, Booth L, Smith AB et al. (2004) Measuring quality of life in
Nijsten T, Sampogna F, Abeni D (2009) Categorization of Skindex-29 scores routine oncology practice improves communication and patient well-
using mixture Analysis. Dermatology 218:151–4 being: a randomized controlled trial. J Clin Oncol 22:714–24
www.jidonline.org 5You can also read
