The study reported here demonstrated the consistency of sensitivity and specificity scores across raciallethnic groups of African American, Latino, and White prisoners entering prison substance abuse treatment programs, which extends prior research (Sacks et al., 2007a
) that validated the CODSI-MD as a screening instrument for any
mental disorder and the CODSI-SMD as a screening instrument for severe
mental disorders among substance-abusing prisoners. Although the main focus of this supplementary study was on sensitivity and specificity, other measures typically used to determine the performance of screening instruments were considered. These additional measures (i.e. negative likelihood ratio, positive predictive value, positive likelihood ratio) also showed no differences between the raciallethnic groups for either the CODSI-MD or the CODSI-SMD, given the precision of these values as indicated by the overlapping confidence limits. Thus, the CODSI-MD and SMD were shown to produce equivalent accuracy across all three of the major racial1 ethnic groups.
The absolute values of sensitivity and specificity scores of the CODSI, containing few items, compared favorably with other screening instruments that Sacks and colleagues tested (Sacks et al., 2007a
). In general, the CODSI-MD produced acceptable sensitivity and specificity. Furthermore, the sensitivity was somewhat higher than the specificity, which indicates more efficiency in the identification of individuals with a mental disorder than in the identification of those without a mental disorder; given the high lifetime prevalence of mental disorders (over 70%) in the populations tested, the emphasis on identifying individuals with the disorder is appropriate. The CODSI-SMD (for severe mental disorders) also produced acceptable sensitivity and specificity, with the specificity being considerably higher than the sensitivity. This is an appropriate balance for the populations tested, because the lifetime prevalence for severe mental disorders drops to less than 30%; at this prevalence, the emphasis shifts to excluding individuals who do not have the disorder, as there are many more offenders without a severe mental disorder.
In addition to demonstrating good psychometric properties on the primary measures of sensitivity and specificity, it is notable that both the CODSI-MD and SMD meet acceptable standards for clinical efficiency, expressed as the percent of individuals referred to assessment who actually have the condition. Efficiency gains importance when assessment resources are scarce, optimizing the number of individuals in need of services who actually receive those services, and minimizing the assessment resources that are spent on individuals who do not have the condition. In this instance, of the individuals with positive scores on the CODSI-MD (for any mental disorder), 84% of African Americans, 87% of Latinos, and 85% of Whites had a corresponding SCID diagnosis of a mental disorder, showing that the instrument is similarly efficient across raceiethnic groups. Consequently, relatively few who did not have a mental disorder would have been referred for additional assessment. Of those with positive scores for a severe mental disorder on the CODSI-SMD, 55% of African Americans, 53% of Latinos, and 78% of Whites had a corresponding SCID diagnosis. The efficiency of the CODSI-SMD increases considerably when individuals with any mental disorder are included; in this instance, 89.7% of African Americans, 97.0% of Latinos, and 95.2% of Whites who had a positive score on the CODSI-SMD had a SCID diagnosis of a mental disorder, again demonstrating similar characteristics across raceiethnic groups. The disadvantage of using only the CODSI-SMD to identify those with any type of mental disorder is that the sensitivity (true positive) will be low, and many individuals who do have a mental disorder will be missed. Where assessment resources are in very short supply, using a combination of the CODSI-MD and the CODSI-SMD will assure the efficient use of those scarce resources; using either or both of the instruments will assure a culturally equitable means of identifying individuals in need of further assessment.
Although the results of the current study are encouraging, readers should be cautious when generalizing these findings. Half the sample was drawn from four CJDATS sites (NDRI Rocky Mountain, Lifespan-Brown University, University of California at Los Angles, Texas Christian University); the additional African American and Latino offenders were sampled only from the NDRI Rocky Mountain site in Colorado. Furthermore, although the study increased the sample of African Americans and Latinos, after disaggregating by race/ethnicity, the high prevalence of mental disorder meant that each subgroup contained few individuals without some sort of mental disorder. Also, given the low prevalence of severe mental disorders overall, relatively few individuals within each subgroup had a severe mental disorder. As a consequence, the point estimates of sensitivity for severe mental disorders and specificity for any mental disorder were less precise. In other words, a larger sample capable of more precise estimates could detect clinically meaningful differences in performance across racial/ethnic groups. Another limitation is the one-month maximum between the Intake Interview and screening battery, and the admission of the SCID; it would be preferable to eliminate the interval between the two interview sessions to avoid any influence that substance abuse treatment might have on mental disorder symptoms in the interim. Finally, the CODSI and SCID diagnoses were compared using lifetime disorders only; for those interested in more recent evidence of mental disorders, the data presented in this article will be overly inclusive of psychiatric diagnoses.