Introduction
The Dermatology Life Quality Index (DLQI) [
1] is a simple, practical questionnaire for routine clinical use and has been widely adopted as a patient-reported outcome measure in dermatology. It was developed more than 20 years ago according to the principles of classical test theory. Because of its reliability, validity and ease of use, it benefits both researchers and clinicians. It has been used for over 36 different skin conditions and translated into 55 languages [
2]. However, the psychometric properties of the DLQI seems not to fulfill the requirements of modern test theory, e.g., Item Response Theory (IRT).
IRT is now seen as the new standard for developing and improving questionnaires [
3,
4]. It has several advantages over classical test theory, particularly for measuring variables in the social sciences [
5]. IRT parameters can be visualized by plotting item characteristic curves (ICCs). ICCs depict the probability of a response (endorsing a symptom) given the level of the underlying characteristic or disease state measured by the whole scale (such as severity of squamous cell carcinoma). In this study, ICCs are defined by difficulty and discrimination, that is, discrimination means how well an item can distinguish lower and higher quality of life while difficult means that results in a higher score (i.e., a lower quality of life), as described by Beth and colleagues [
6], which govern the shape and position of S-curves. Recently, Rasch analysis, which is one specific application of IRT, has been applied to assess the DLQI [
5]. The application of the DLQI for psoriasis, atopic dermatitis (AD) and hand eczema (HE) was criticized based on Rasch analysis [
7‐
9]. The objective of the present study was to perform a psychometric test of the DLQI in a sample of neurodermatitis patients using the Rasch analysis.
Neurodermatitis, also called lichen simplex chronicus (LSC), is a disorder commonly encountered by dermatologists and caused by constant rubbing of the skin that clinically induces thickening and lichenification. Neck, elbow, ankles, vulva, eyelid, faces and even conjunctiva [
10] can be affected. Although it is not life-threatening, LSC can result in psychosocial problems, and it has been suggested that patients who get it are more likely to suffer from depression, anxiety and other treatable psychological disorders [
11]. It can also impair quality of life through sleep disturbance and sexual dysfunction [
12]. All these data indicate that it is necessary to pay attention to the negative impact of neurodermatitis on patients’ quality of life (QoL).
In our previous study, an evaluation of the DLQI in LSC was conducted within the framework of classical test theory and demonstrated good feasibility and internal consistency [
13]. However, to date, it appears that no studies available have used a Rasch analysis approach to examine the psychometric property of the DLQI in persons with LSC.
Therefore, the overall aim of our study was to use Rasch analysis to evaluate the DLQI in a sample of Chinese persons with LSC. In this study, we used Rasch analysis to test rating scale function, item fit to the Rasch model, aspects of person-response validity, unidimensionality, person-separation reliability, and differential item function (DIF)..
Discussion
This study is, to the best of our knowledge, the first to use Rasch analysis to examine the psychometric properties of the DLQI in a sample of patients with LSC. Psychometric deficiencies that have previously been reported for the DLQI’s use with other disorders were also found in Chinese patients with LSC.
In previous studies, Rasch analysis was applied to the DLQI data to study its psychometric properties in patients with psoriasis, AD and HE. In psoriasis and AD patients, inadequate measurement properties were observed [
7]. One study revealed disordered thresholds and DIF for data from psoriasis patients across different cultures [
9], and another detected similar problems with data from patients with HE [
8]. Our study documented psychometric problems with LSC patients.
None of the response categories pose a threat to the validity of the DLQI. Good separation ability of the 10-item version (PSI 2.38) indicates that the items may reliably identify persons with higher QoL and well-being. The first latent dimension (general QoL) accounted for 50.8 % of the variance in the 10-item DLQI, but item 9 demonstrated poor fit to the Rasch model. The variance explained by the second dimension (7.1 %) exceeded the criterion of 5 %, suggesting the possibility of a minor second dimension. There were also limitations related to person-response validity, with ≥ 5 % (18.1 %) of cases in the 10-item DLQI demonstrating unacceptable fit to the Rasch model. Figure
1 shows that the DLQI has a poor spread of thresholds, indicating a limitingmeasurement range.
Item 9 demonstrated misfit to the Rasch model. This may indicate a general misfit of the item across sample populations, and not a particular misfit among persons with LSC, given that a study on HE patients found similar problems with this item [
8]. Item misfit essentially indicates that the respondents rated this item inconsistently in relation to their overall response pattern. For item 9 the misfit may be explained by a difference between Asian and Western cultures. Discussing sexual activity is a cultural taboo, and body contact is not considered acceptable in everyday interactions between men and women. Recently in Chinese society, notions of sexual activities have changed considerably; however, it is still considered too private an issue to discuss frankly, even with friends or family. Therefore, the forthrightness of respondents’ answers remains unknown. It is possible, however, that the LSC is not severe enough (especially in stage II) to influence respondents’ sexual lives, or that most of the LSC lesions were not in the anogenital region. Ideally, we would have analyzed the effect of lesion locations on sexual life. Of course, it would be a disservice to eliminate an item that had particular relevance to a segment of the LSC population [
24,
25].
Removing item 9 could also make the total score of the remaining items difficult to compare with previous research results. Therefore, if future studies with participants with LSC use a modified version of the DLQI, the scores should be adjusted to correct for the reduced number of items. However, comparisons with 10-item scores, in particular to shorter scales composed of different items, should be made with caution as there are still controversies regarding comparative outcomes [
9,
26].
Our study also showed lower DLQI levels for our participants than have been reported previously with psoriasis population samples [
7,
27], which is consistent with our previous study [
13]. This makes sense because severe itching is a prominent feature of LSC but is rarely a concern for psoriasis patients. Consistent with our finding, Reich and colleagues recently found itching in 89.2 % of psoriatic patients [
28], and the presence and intensity of itching did not depend on age, sex, type of psoriasis, duration of disease or disease severity [
29]. In addition, LSC patients had more sexual problems than did their counterparts with psoriasis [
24]. Further researches are needed to examine the observations and test them with Rasch analysis.
In contrast, age, education and severity of illness may be related to QoL. In a sample of patients younger than 35 years, QoL was substantially lower (mean DQLI = 8.75) than in a sample of patients over 35 (mean DQLI = 9.95). The difference was discussed in light of psychological factor theory [
12], that is, the younger, the more anxious and therefore the more severe and persistent LSC. For patients with a higher education level, LSC influenced their ability to dress appropriately. Thus, the lower levels of QoL in this group may be partly attributable to their attention to appearance, which might be rooted in unstable situations and the stress they may experience during the process of job change. This comparison may indirectly speak to a larger impact of stress, rather than age, on QoL. Dressing up and having close relationships were both affected in patients with severe LSC; it is to be expected that severe itching and lichenoid skin would affect such aspects of QoL. In addition, there is no evidence of DIF for sex in our study, which is different from previous reports [
25,
30] . When the effect on QoL of a specific entity is being considered, the larger sample, the more representative the result will be. As for sex, its influence on QoL displays no common pattern, although sex effects have been reported previously in patients with HE [
31] and psoriasis [
32]. During the investigation, many patients reported fear of the effects of diet on LSC. This factor is not included in the DLQI because of cross-cultural differences, although it may affect QoL. In addition, many items are ambiguous; that is, one item may include more than one idea. For example, item 5 asks how much the patient’s LSC has affected any social or leisure activities. With such an item, one patient may respond to the “social” aspect of the question while another considers the “leisure” aspect; this becomes problematic for the measure if different parts of the question represent different levels of impact. In addition, there are many manual workers in China, and labor is exercise for them. Therefore, it is difficult for them to evaluate the influence of a skin disease on sports activities, as described in item 6. Item 6 exceeded our expectations for fit to the Rasch model in our analysis.
Strengths and limitations
This is the first study to report on Rasch analysis of the DLQI in LSC. The study was based on a sample of patients with a high participation rate (99.3 %) and relatively little missing data, thereby minimizing the likelihood of bias. The total sample of 149 provided a sufficient number for Rasch parameter estimates as it was large enough to give 99 % confidence that the reported estimates fall into a range of +/− 0.5 logits. Furthermore, evaluating the psychometric properties of the DLQI allows researchers and clinicians to decide whether they can benefit from it. In terms of limitations, the study was conducted in a single hospital and respondents were predominantly of Chinese cultural background. Despite a number of studies that have demonstrated sexual dysfunction associated with LSC, our study found that items about sexual difficulties have poor fit to the Rasch model, possibly because of dimensions of sexuality in the Chinese polutation. Research with other cultural populations may show quite different experiences of sexuality in respondents with LSC. We did not analyse the impact of lesion locations on QoL. Finally, this study only illustrated the use of the DLQI with LSC. The DLQI is also used with several other dermatological conditions, and further research could determine whether the scale is appropriate for use with those conditions.
Conclusions
The DLQI was one of the first disease-specific HRQoL instruments to be developed in dermatology and has contributed greatly to our understanding of the patient’s perspective on LSC. However, this study suggests that the instrument is not well suited to measure disease impact among LSC patients in China. The DLQI is multidimensional, has limitations in person-response validity, and has two items that displayed non-uniform DIF. Also, the generalizability of the total scale and the underlying structure of the subscales could not be confirmed. The psychometric properties of instruments designed to assess the overall impact of a disease and its treatments on patients’ lives should be thoroughly tested in populations that vary culturally, demographically, and in disease severity before they are recognized as valid HRQoL assessment tools.
In sum, for LSC patients, the DLQI seems to have poor fit to the Rasch model in this population. As such, DLQI is not an approriate measure of QoL in Chinese patients with LSC.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
YL and TL were responsible for data collection, participated in the data analysis and drafted the manuscript. JA designed the study and performed the data analysis. WZ participated in the data collection and manuscript writing. SX conceived of the study, participated in its design and coordination and helped to draft the manuscript. All authors read and approved the final manuscript.