One hundred fifty-nine psychiatric outpatients were examined to determine which of the DSM-III Borderline Personality Disorder (BPD) criteria were most valid in terms of sensitivity, specificity, predictive power positive, and predictive power negative. Combinations of two criteria predicted better than individual ones and combinations of three criteria predicted better still. It is possible that in many instances BPD can be efficiently diagnosed by three criteria. These results were compared to previous findings. The exact criteria that best discriminate depends to a large extent on the comparison groups used. In general it appears the current criteria for BPD are capable of discriminating diverse subject groups. The possibility of searching for "core criteria" that broadly discriminate between differing subject groups is discussed.
Any psychiatric diagnosis in development requires a period of emprical testing before its criteria are definitively established. Initially a set of prospective criteria is established from literature, theory, factor analytic studies, or skilled clinicians. These criteria must then be empirically evaluated. The personality disorders are at such a stage of development now. A question then arises as to how to evaluate them. There are many different approaches from many different disciplines. Baldessarini et al. and subsequently Widiger et al. have described a technique taken from psychology for formally proceeding with this process. This technique calculates sensitivity, specificity, predictive power positive (PPP), and predictive power negative (PPN) for proposed criteria and combinations of proposed criteria (with special emphasis on PPP and PPN). Criteria that appear important across multiple studies and populations may be "core criteria." (Core criteria are criteria which are necessary or sufficient factors for the diagnosis of the disorder.) Criteria can vary in their degree of importance in defining a disorder. For example, although transient depression is often associated with many Axis I and Axis II disorders (i.e., high sensitivity, prevalence), it is seldom sufficiently pathognomonic enough to be included in defining criteria. This report presents a set of results from a new population of patients with borderline personality disorders (BPD) using sensitivity, specificity, PPP, and PPN. The discussion places current results in the context of existing findings. The following sections briefly review previous findings. Unless otherwise specified, DSM-III criteria are used. In order to prevent unnecessary repetition of the criteria they will be referred to by one word abbreviation as follows: Impulsivity, Impulsive acts; Interpersonal, Unstable-Intense relationships; Anger, Intense-Uncontrolled anger; Identity, Identity Disturbance; Affective, Affective instability; Alone, Intolerance of being alone; Self-damaging, Physically self-damaging acts; and Boredom, chronic emptiness and boredom. (For easy reference these are shown in Table 1.) In order to make the descriptions of the following studies easier to follow I have summarized the key findings in Table 2.
Table 1. DSM-III Criteria for Borderline Personality Disorder (BDP
|| Impulsive acts
|| Unstable-intense relationships
|| Affective instability
||Intolerance of being alone
||Physically self-damaging acts
||Feelings of chronic emptiness and boredom
In a sample of 76 outpatients with a primary Axis II disorder (BPD base rate = 0.34), Widiger et al. found the PPP for the individual DSM III criteria for BPD ranged from 0.56 to 0.73. They felt that PPP was the measure best reflecting what clinicians are trying to achieve by diagnosis. (PPP indicates how many of those who are given the diagnosis actually have it. Because it is influenced by the prevalence of a disorder a direct comparison of PPP cannot be made across studies.) More interesting, however, was their examination of combinations of two symptoms. Of 28 pairs, 19 had a PPP of greater than 0.80. They identified some especially valuable pairs (Interpersonal-Self-damaging, Selfdamaging-Identity, Interpersonal-Identity, Impulsivity-Interpersonal, Impulsivity-Identity, Affective-Boredom) with PPP > 0.90. Widiger et al. concluded that in their population the five symptoms required by DSM-III to make the diagnosis of BPD were not necessary. Their population was particularly sick, as inpatient populations go, so it may not be representative and easily generalizable.
Nurnberg et al. carefully examined 17 BPD-related symptoms on a population of 17 hospitalized BPD patients and 20 controls. Their study basically confirmed the value of DSM-III criteria. However, when their four best criteria [roughly, DSM-III Impulsivity, Interpersonal (Alone and/or Boredom) and “acting out"] were combined into any combination of two symptoms, their PPP was 0.94. They reported no appreciable gain using a combination of three criteria and concluded that five symptoms were not necessary to makea BPDdiagnosisin their population. Nurnberg et al. also published a further analysis of their data, in which they report their best results for two criteria as Impulsivity-Identity, Impulsivity-Boredom, and Interpersonal-Identity. They also report that when using their five best criteria there was a slight increase in the total error rate when five criteria were used as the cutoff rate instead of four. (This is due to an increased false-negative rate.)
McGlashan examined combinations of the BPD criteria with sensitivities greater than 0.50 in their Chestnut Lodge follow-up study. He found that the combination of Anger and Self-damaging was the best predictor for populations that excluded psychotic patients. He found, as did Nurnberg et al., that there was no gain from using more than two criteria. A weakness of this study was that criteria were assessed retrospectively by chart review.
Hurt et al., in a study using advanced statistical methods not commonly used in this area of research, analyzed four previous studies. They felt that the use of explicit criteria allowed the decision rules for making a diagnosis to be examined and derived statistically. Their analyses indicated that there were three core dimensions to BPD—Identity disturbance, affective disturbance and impulse disturbance. (They feel the latter two would be especially effective for identifying BPD.) They conclude that a rule requiring three of the following four criteria would be close to optimum—Impulsivity, Interpersonal, Affective, and Identity. Limitations of the Hurt report are its particularly ill populations, many of whom were diagnosed by chart review, and also the fact that the report included some of the studies reported above, allowing the danger of circularity.
The current study reports on a new population of psychiatric outpatients. The present contribution differs from previous studies in that it reports on general psychiatric outpatients. The current results will be integrated with previous findings.
Two proband populations were used for the present study. The first was a group of panic disorder patients recruited to take part in a treatment trial (N = 80) by advertising. Axis I disorders in this group were diagnosed by the Structured Clinical Interview for DSM-III Disorders administered by a board-certified psychiatrist. All patients were required to meet criteria for panic disorder and to be having at least one panic disorder per week . Patients were excluded if they were schizophrenic, mentally retarded, had an organic brain syndrome, mania, obsessive-compulsive disorder, drug or alcohol abuse in the last year, or major depression that dominated or preceded their panic disorder symptoms. The second group was a sample of randomly selected psychiatric outpatients drawn from new intakes to the psychiatry outpatient clinic ( N = 79) . Patients with psychotic symptoms, an organic brain syndrome, or mental retardation were excluded. Here the patients' Axis I diagnoses were determined for this group by a masters' level interviewer using the Schedule for Affective Disorder and Schizophrenia Lifetime Version (SADS-L), which uses Research Diagnostic Criteria (RDC).
Although different procedures were used to diagnose the different populations, in both groups most diagnoses had been excluded except for the anxiety and depressive disorders. DSM-III and RDC criteria do have some differences in these diagnoses, but they are not major and are not the focus of this study.
Personality disorders in probands were determined by the Personality Diagnostic Questionnaire (PDQ). The PDQ is a 152-item, self-administered, true-false instrument measuring all 11 DSM-III, Axis II personality disorders. Test-retest reliability for psychiatric outpatients at 1 month is 0.56 or above for paranoid, schizotypal, antisocial, BPD, avoidant, and compulsive disorders. (The specific kappa for BPD is 0.63.) Our own work with the PDQ for 8- week test-retest gave a kappa for BPD of 0.50 . A comparison of the PDQ with two DSM-III Axis II interview instruments [Personality Disorder Examination (PDE), version1and Structured Clinical Interview for DSM-III, Axis II (SCIDII)] indicated that, in general, the PDQ agreed with the interview instruments as well as they agreed with each other. In that study when the PDQ was compared to an all data personality "L.E.A.D. standard" BPD had a PPP of 0.63 and an PPN of 0.88. These are acceptable values given the state of the art of DSM-III personality measurement....
Download PDF to continue reading article, footnotes, tables, and references.