Home | Business News | Browse by Publication | J | Journal of Mental Health Counseling

Reducing diagnostic bias. (Practice).

Article Excerpt
Skill using the American Psychiatric Association's (2000) Diagnostic and Statistical Manual of Mental Disorders, Text Revision is essential for the increased professional credibility, career marketability, and third-party reimbursement of professional counselors. This article focuses on how to improve counselors' skill with DSM-IV-TR, by providing definitions, empirical demonstrations, and strategies for reducing three forms of bias that can lead to misdiagnosis.

**********

Many counselor educators see greater diagnostic skill with the fourth edition of the Diagnostic and Statistical Manual of the American Psychiatric Association, Text Revised (American Psychiatric Association, 2000) as essential for enhanced professional credibility, career marketability, and third-party reimbursement (Fong, 1990, 1993; Foos, Ottens, & Hill, 1991; Geroski, Rogers, & Breen, 1997; Hohenshil, 1993, 1996). Most counselor education literature on the DSM has been focused on instructional approaches, clinical applications, and reconciliation with the traditional counseling focus on normal development (Hershenson, 1992, 1993; Hershenson & Strein, 1991; Ivey & Ivey, 1998). Along with learning about the DSM, counselors also need to learn how to achieve skill in its use.

Achieving skill with the DSM means learning how to diagnose accurately and reduce diagnostic inaccuracy or misdiagnosis (Hohenshil, 1993; Seligman, 1996). Reducing diagnostic bias is one way to reduce misdiagnosis (Cook, Warnke, & Dupuy, 1993; Furlong & Hayden, 1993). Diagnostic bias is defined by Sinecore-Guinn (1995) as an "error in judgment that counselors make when they collect and interpret information" (p. 18), and by Widiger and Spitzer (1991) as "a differential prevalence of either false-positive diagnoses ... and/or false-negative diagnoses" (p. 3). Greater skill with the DSM requires that counselors learn ways of overcoming diagnostic bias.

Bias is defined statistically as measurement error (Mertens, 1998). Widiger and Sptizer (1991) used a statistical definition of diagnostic bias in suggesting it is "deviation from an expected value" (p. 3). For instance, in 100 coin tosses, bias is expected if the number of heads or tails greatly exceeds a 50-50 ratio, the expected value. Widiger and Spitzer identified sampling, assessment, and criterion bias as risks to accurate diagnosis. The purpose of this article is to define, demonstrate, and discuss ways of reducing sampling, assessment, and criterion bias. First, definitions are presented of each form of diagnostic bias. Second, empirical demonstrations are provided of each form of diagnostic bias. A final section lists ways of reducing each of these forms of diagnostic bias, and implications for counselor training, research, and practice.

DIAGNOSTIC SAMPLING BIAS

Definition

Diagnostic sampling bias occurs when there are significant differences between a particular diagnostic sample and the population it is taken to represent (Garb, 1998; Gilovich, 1991). An example of diagnostic sampling bias is drawing conclusions about the gender of individuals with Post-Traumatic Stress Disorder (PTSD) on the basis of a sample taken from a veteran's hospital (Widiger & Spitzer, 1991). Because of the disproportionate number of men in that setting, conclusions about the gender of those with PTSD will be biased.

Demonstrations of Diagnostic Sampling Bias

DeGrandpre (1999) demonstrated diagnostic sampling bias in an examination of the ability of physicians to diagnose Attention Deficit Hyperactivity Disorder (ADHD) solely from observation of children in their offices. His results indicated that more than three of four children described as hyperactive by parents and teachers showed "exemplary behavior and no sign of hyperactivity in the [doctor's] office" (p. 133). This demonstrates how a non-representative sample of observations (i. e., only in a physician's office) can lead to misdiagnosis.

Researchers also demonstrated sampling bias in another study comparing clinical and community samples of children with a diagnosis of ADHD (Sharp, Walter, & Marsh, 1999). In referred samples, four to nine times more boys than girls received an ADHD diagnosis. However, in community samples, ratios as low as two to one have been found. This gender discrepancy between clinical and community samples raises questions about the representativeness of many clinical samples.

Wilke (1994) also demonstrated diagnostic sampling bias in research with women and alcohol abuse. She observed that, because the most research on alcohol abuse and treatment has focused on men, conclusions about treatment are inappropriate when applied to women. Because women with alcohol problems are less likely than men to drink in public, drink with others, become violent or aggressive, or to come into contact with the law, many of their alcohol problems go undiagnosed and untreated. This research demonstrates how over-generalizing from one group to another can lead to misdiagnosis.

The dramatic rise in the number of Multiple Personality Disorder (MPD) diagnoses made during the 1970s in the United States can also be seen as a demonstration of diagnostic sampling bias (Hacking, 1995; Ofshe & Watters, 1994; Spanos, 1994). The increased number of MPD diagnoses given to individuals in the United States did not occur elsewhere (Kutchins & Kirk, 1997). Therefore, some observers have concluded that the increase in the United States constituted a biased sample instead of reflecting an actual increase in the disorder (Hacking; Ofshe & Watters).

DIAGNOSTIC ASSESSMENT BIAS

Definition

Diagnostic assessment bias occurs when flaws in gathering or processing clinical information...