Reliability and validity of the Sheehan disability scale modified for pathological gambling

The Sheehan Disability Scale (SDS) was originally designed to assess functional impairment associated with anxiety disorder diagnoses [1]. Functional impairment or disability is related to, but conceptually distinct from, symptom severity and distress. The SDS is a three-item self-completion scale measuring the impact of symptomatology on work, social and family functioning. Because of its simplicity, brevity, and high face validity, and because of the lack of competing brief measures of functional impairment, the SDS has become a widely used outcome measure in clinical trials of a variety of mental health disorders. The SDS has been used for obsessive compulsive disorder, post-traumatic stress disorder, major depressive disorder, panic disorder, social anxiety disorder, schizophrenia [2] and, most recently, pathological gambling [3‐6]. As of 2007, the SDS had been translated from English into 48 languages [2].

Despite its popularity, limited data on the psychometric properties of the SDS have been published. Leon, Shear, Portera and Klerman [7] analyzed clinical trial data obtained from patients with panic disorder. The authors reported that the scale had adequate internal reliability (α = .56 to .86) and a stable one factor structure at baseline and post-treatment follow-up. They also demonstrated that the measure is sensitive to treatment-related change and that the scores, as expected, were moderately associated with panic symptoms. Leon and colleagues [8] also examined the SDS in a sample of primary care outpatients concurrently assessed for mental health disorders. Internal reliability was excellent (α = .89) and a one factor structure was also supported. In terms of validity, patients who met the criteria for a mental disorder had more impairment than those who did not. The authors used a cut-point of five or greater on the 30-point scale, and they found that the SDS had good sensitivity and specificity in identifying the mental disorder patients.

Sheehan and Sheehan [2] reviewed the use of the scale in published clinical trials in order to establish interpretation guidelines for patient improvement. They provide preliminary recommendations that a total score of 6 or less indicates remission and a score of 12 or less indicates treatment response. A reduction of approximately 4 points indicates improvement [2].

In the gambling arena, a number of well validated interview and self-rated measures of symptom severity are available [9] including: the South Oak Gambling Screen (SOGS) [10], the Problem Gambling Severity Index (PGSI) of the Canadian Problem Gambling Index [11, 12], the National Opinion Research Centre DSM Screen for Gambling Problems (NODS), [13], the Gambling Symptom Assessment Scale [14], and the Gambling Treatment Outcome Monitoring System [15]. Additionally, interview measures of self-reported gambling behaviours (frequency, money and time spent) show good psychometric properties [16, 17]. Measures of impairment include the SDS as well as a gambling adaptation of the Addiction Severity Index (ASI-G), which is an interview measure that assesses functioning and need for treatment in a variety of domains including gambling problems, physical health, employment, social relationships, psychiatric functioning, alcohol and other drug use and legal problems. The gambling subscale has been well validated in treatment populations [18‐20] whereas the other subscales have been mostly validated with substance abuse populations. The SDS has not been assessed psychometrically despite its use.

The purpose of this report is to provide data on the reliability and validity of an interview version of the Sheehan Disability Scale modified for gambling (SDS-G). Data from a trial of a distance relapse prevention program were used to assessment the psychometric properties. One of the unique features of the SDS is that the scale items are presented visually as a horizontal line marked with both numbers (0 to 10) and verbal anchors (0 = Not at all, 1–3 = mildly, 4–6 = moderately, 7–9 = markedly, and 10 = extremely). The numbers are presented in a larger font than the verbal anchors and respondents are required to choose a specific number to indicate their response. This design is easy for patients to complete, but, unfortunately, limits the scale from being used in a face-to-face or telephone interview format. Because telephone follow-up is often used in gambling treatment studies, the first step in this investigation was to solicit feedback from interviewers and respondents with pathological gambling about the clarity of different response options in an interview format. The response options were modified from the instructions provided for the self-completion version.

Ethical approval for this study was provided by the Conjoint Faculties Ethic Review Board at the University of Calgary and all participants provided written informed consent.

These analyses provide evidence in support of a gambling modified SDS to assess gambling treatment outcomes via interview. Because of the nature of the relapse prevention study, this sample of pathological gamblers includes a range of gambling involvements from abstinent to heavy gambling, which is optimal in terms of validating a scale. The results supported a one factor solution and the scale had good internal reliability for a three item scale. The SDS-G appears sensitive to change and it appears to correlate more strongly with other measures of current functioning than lifetime measures.

Future research should assess of retest reliability of the scale and it would also be useful to assess concurrent validity of the very brief scale against more lengthy measures of functional impairment such as the Social Adjustment Scale [30] and the Medical Outcome Study SF-36 [31]; however, these scales have not been validated for pathological gambling samples. Validation of a self-completion version would also be valuable. Moreover, some clinical trial studies report the individual item scores in addition to the total scores. The reliability and validity of individual items needs further examination. In the structural equation model, the work variables loaded less strongly on the latent disability construct than the family and social variables and, similarly, the work variable correlated less strongly with the SDS-G total (see Table 1). The work variable also correlated less strongly with the concurrent validity variables than did the social, family and total scores (see Table 2). This difference may be related to the fact that gambling problems have less impact on work functioning or that preservation of work functioning is important to finance continued gambling (see mean scores on Table 1), or it may indicate a weaker measure of the work disability construct.

Although these analyses support the psychometric strength of this brief scale, they do not provide validated interpretation guidelines for improvement or adequate social functioning. The existing guidelines are not disorder specific and are meant to be used for a variety of disorders. Nonetheless, validation for each disorder individually is crucial. Interpretation guidelines would increase the practical utility of using this scale in outcome studies as well as use by clinicians monitoring patent outcome.

In conclusion, the SDS-G total score should work well as an outcome measure that is correlated with both gambling behaviour and gambling problem severity, but also conceptually and statistically distinct. Despite its brevity, it shows good internal reliability, good evidence of factorial validity and it is sensitive to patient changes.