The Mental Health Provider Self-Assessment of Stigma Scale (MHPSASS): Development, Validation, and Special Efforts to Address Social Desirability Bias

Background: 

This paper describes the development and statistical testing of a measure assessing the presence and magnitude of mental health provider-based stigma.  Also highlighted are efforts to mitigate social desirability bias.  Often unintentional and unknowingly conveyed, the phenomenon of provider-based stigma has been indicated in previous research (e.g. Lauber, Nordt, Braunschweig, & Rössler, 2006; Caldwell & Jorm, 2001; Hugo, 2001; Schulze, 2007) and presents a serious impediment to effective service delivery.  In order to assess the incidence and severity of provider stigma the profession requires a reliable and valid measure.

To circumvent the influence of social desirability bias, a predicted threat to validity, this measure made use of forgiving wording.  In essence, forgiving wording is a language strategy which involves phrasing items in such a way that encourages respondents to answer more truthfully (Sudman & Bradburn, 1982; Groves, Fowler, Couper, Lepkowski, Singer, & Tourangeur, 2004).  For the MHPSASS, items were crafted taking two approaches: 1) ‘everyone does it’ or 2) the attitude or behavior exists for comprehensible reasons. 

Methods:

The MHPSASS’ item pool was generated following Nunnally and Bernstein’s (1994) domain sampling method, in reflection of Charles’ experience-based model of provider stigma (2013).  Next, a series of focus and stakeholder discussion groups served as an expert panel reviewing both model and item pool.  The electronic survey contained 82 items, including a social desirability scale.  A purposive sample of mental health providers was assembled with assistance from Virginia’s DBHDS.  In total, 50 public agencies were invited to participate.  Of the final sample of n=220 individuals, most respondents were female (n=180, 81.8%), commonly identifying with the professions of social work (n=64, 29.1%) and counseling (n=54, 24.5%).  Statistical analyses included univariate, bivariate, and multivariate procedures. 

Results: 

An exploratory factor analysis using principal component extraction with varimax rotation indicates a four factor solution.  Refinement resulted in a 20-item scale demonstrating good internal consistency, Cronbach’s α=0.817.  The four factors were labeled: Irritation & Impatience; Choice & Capacity; Adherence & Dependence; Devalue & Depersonalize.  Maximum possible score was 120 and minimum = 0, the range of MHPSASS scores was from 16 to 80.  Mean score = 50.450, sd = 12.308.  The relationship between MHPSASS score and social desirability bias scale, computed via Pearson’s r = -0.169 (p = 0.015) indicates a significant negative relationship. 

Of the 20 retained items, all but one was crafted using the forgiving language strategy.  When comparing near duplicate items, response variance was greater for items utilizing forgiving language as opposed to more traditionally worded items, as evidenced by higher means and larger standard deviations.  

Conclusion:

Meaningful findings include the promising psychometric properties of this measure.  Additionally, the use of forgiving language proved helpful in this instrument’s development.  Even still, the correlation between the social desirability scale and MHPSASS score indicates the influence of bias.  Further investigation into the usefulness of this language strategy in item development is warranted.  In addition, refinement of the MHPSASS is suggested, as well as replication studies and comparisons to similar measures completed by consumers of services.