Psychological Assessment As A Tool For Screening And Diagnosis

One of the most apparent ways in which psychological assessment can contribute to the development of an economical and efficient behavioral health care delivery system is by using it to screen potential patients for need for behavioral health care services and to determine the likelihood that the problem identified is a particular disorder or problem of interest. Probably the most concise, informative treatment of the topic of the use of psychological tests in screening for behavioral health care disorders is provided by Derogatis and Lynn (1999). They clarify the nature and the use of screening procedures, stating that the screening process represents a relatively unrefined sieve that is designed to segregate the cohort under assessment into "positives," who presumably have the condition, and "negatives," who are ostensibly free of the disorder. Screening is not a diagnostic procedure per se. Rather, it represents a preliminary filtering operation that identifies those individuals with the highest probability of having the disorder in question for subsequent specific diagnostic evaluation. Individuals found negative by the screening process are not evaluated further (p. 42).

The most important aspect of any screening procedure is the efficiency with which it can provide information useful to clinical decision-making. In the area of clinical psychology, the most efficient and thoroughly investigated screening procedures involve the use of psychological assessment instruments. As implied by the foregoing, the power or utility of a psychological screener lies in its ability to determine, with a high level of probability, whether the respondent is or is not a member of a group with clearly defined characteristics. In daily clinical practice, the most commonly used screeners are those designed specifically to identify some aspect of psychological functioning or disturbance or provide a broad overview of the respondent's point-in-time mental status. Examples of screeners include the Beck Depression Inventory-II (BDI-II; Beck, Steer, & Brown, 1996) and the Brief Symptom Inventory (BSI; Derogatis, 1992).

The establishment of a system for screening for a particular disorder or condition involves determining what it is one wants to screen in or screen out, at what level of probability one feels comfortable about making that decision, and how many incorrect classifications or what percentage of errors one is willing to tolerate. Once one decides what one wishes to screen for, one must then turn to the instrument's classification efficiency statistics—sensitivity, specificity, positive predictive power (PPP), negative predictive power (NPP), and receiver operating characteristic (ROC) curves—for the information necessary to determine if a given instrument is suitable for the intended purpose(s). These statistics are discussed in detail in the chapter by Wasserman and Bracken in this volume.

A note of caution is warranted when evaluating sensitivity, specificity, and the two predictive powers of a test. First, the cutoff score, index value, or other criterion used for classification can be adjusted to maximize either sensitivity or specificity. However, maximization of one will necessarily result in a decrease in the other, thus increasing the percentage of false positives (with maximized sensitivity) or false negatives (with maximized specificity). Second, unlike sensitivity and specificity, both PPP and NPP are affected and change according to the prevalence or base rate at which the condition or characteristic of interest (i.e., that which is being screened by the test) occurs within a given setting. As Elwood (1993) reports, the lowering of base rates results in lower PPPs, whereas increasing base rates results in higher PPPs. The opposite trend is true for NPPs. He notes that this is an important consideration because clinical tests are frequently validated using samples in which the prevalence rate is .50, or 50%. Thus, it is not surprising to see a test's PPP drop in real-life applications where the prevalence is lower.

Was this article helpful?

0 0

Post a comment