Epidemiology Personality Thomas Myrna

A. Widiger, M. Weissman,

of Borderline Disorder Ph.D. Ph.D.

The limited epidemiological data available on borderline personality disorder suggest that the prevalence of the disorder is between .2 and 1.8 percent in the general community, 15 percent among psychiatric inpatients, and 50 percent among psychiatric inpatients with a diagnosis of personality disorder. No data on the incidence-tbe rate of new cases-of the disorder have been reported, and inferences about incidence based on prevaknce rates are complicated by differences in the formal designation ofpersonality disorders before and afterDSM-IlI was issued. Current findings suggest that about 76 percent of borderline patients are female. Epidemiologicalstudy of borderline personality disorder has been hindered by the la#{128}k ofa brief semistructured interview that can be used with large population samples and that does not require substantial dinical expertise. The authors discuss alternative research methods, including use of lay interviewers, receding of existing data, telephone interviews, and self-report inventories.

Dr. Widiger is professor of psychology at the University of Kentucky, 1 1 5 Kastle Hall, Lexington, Kentucky 40506. Dr. Weissman is professor of epidemiology in psychiatry at Columbia UniversityNew York State Psychiatric Institute in New York City. This paper is part of a special section on borderline personality disorder.

Hospital

and

Community

Psychiatry

Epidemiology is the study ofthe distribution and determinants of disorders ( 1 ,2). Epidemiologic studies yield data on prevalence rate (the number ofcases ofa given condition at a particular time) and incidence rate (the number of new cases in a population), as well as on correlates (conditions and variables associated with occurrence of the disorder) and factors associated with risk of developing a disorder. Borderline personality disorder has been of considerable clinical and theoretical interest, and a review of its epidemiology is certainly warranted. However, epidemiologic research requires systematic, reliable sampling of large numbers of representative cases within explicitly defined community and clinical populations (3); it is therefore very cxpensive. Epidemiologic research in psychiatry has benefited substantially from the development of specific, explicit diagnostic criteria and structured interviews that can be administered by lay interviewers (4). Epidemiological research on the personality disorders, however, has been hindered by difficulty in defining a disorder ofpersonality using a brief list of specific symptoms and difficulty in assessing such a disorder by a concise set of simple and direct questions that require little or no inference or judgment on the part of the rater (5-7). Antisocial personality disorder was the only personality disorder to be included in the Epidemiologic Catchment Area (ECA) study sponsored by the National Institute of Mental Health (8). The DSM-III-R diagnosis of antisocial personality disorder has the most specific and cx-

October

1991

VoL 42

No.

10

plicit diagnostic criteria of the personality disorders, but the validity of these criteria has been controversial precisely because of their specificity (9,10). Usc of a brief list of specific questions to assess an equally concise set ofcriteria to diagnose borderline personality disorder would be even more difficult and controversial (1 1). The most popular semistructured interview for assessing borderline personality disorder is the Diagnostic Interview for Borderlines-Revised (DIB-R), which takes about an hour to administer and requires “substantial clinical experience” (12, p. 12). Because ofthese limitations, there arc few data on the epidemiology of borderline personality disorder in large community samples. In this paper, we review existing studies and provide estimates of epidemiologic parameters

based

on data

reported

in

these studies. Our discussion is necessarily confined to questions of prevalence, incidence, and sex ratio, for which there still remains substantial controversy and dispute. We condude our review with a discussion of alternative approaches to obtaining epidemiologic data on borderline personality disorder. Prevalence Based on epidemiologic studies conducted before DSM-Ill was issued, Merikangas and Wcissman (1 3) estimated the prevalence of borderline personality disorder within the community to be between .2 percent (14) and 1.7 percent (15). The latter figure is remarkably close to more recent estimates. Swartz and associates (16) recoded data collected in Durham, North Carolina, using the Diagnostic Interview Schedule

1015

(DIS) as part of the ECA study (8). The prevalence rate of borderline personality disorder among the 1,541 community residents in the sample, who were between ages 19 and 55, was 1.8 percent. Zimmerman and Coryell (17) interviewed 797 relatives of normal control subjects and of schizophrenic and depressed patients using the Structured Interview for DSM-III Personality Disorders (SIDP) (18). They reported a prevalence rate of 1 .6 percent for borderline personality disorder. One of the more curious findings was that borderline personality disorder was not the most preyalent personality disorder diagnosed in the sample. Six personality disorders were diagnosed more frequently than borderline personality disorder. A total of3.3 percent ofthe subjects received a diagnosis of passive-aggressive personality disorder. Zimmerman and Coryell(19) also administered the Personality Diagnostic Questionnaire (PDQ) (20), a self-report instrument, to 697 relatives ofpsychiatric patients and normal control subjects. The prevalence rate of borderline personality disorder based on the PDQ was 4.6 percent. Reich and colleagues (21) sent the PDQ to 379 adults randomly selected from a university community in Iowa with a population of more than 36,000. The prevalence rate of borderline personality disorder in the 235 respondents was 1 .3 percent. Widiger and Trull (22) identified 30 studies published between 1975 and 1988 that used a clinical interview and provided sufficient data to estimate prevalence of borderline personality disorder in specific populations of psychiatric patients. The average prevalence across eight outpatient studies that had no inclusion or exclusion sampling criteria was 8 percent; for the 14 inpatient studies with no exclusion or inclusion critena the average prevalence was 15 percent. In four outpatient studies that were confined to patients with a personality disorder, the prevalence of borderline personality disorder was 27 percent, and in four inpatient studies that were confined to patients with apersonality disorder, the prevalence rate was 5 1 percent.

Widiger and Trull (22) also summarized the results from 14 studies that examined the prevalence rate for all 1 1 personality disorders. Borderline personality disorder was typically the most prevalent, and the differences between the rate of that disorder and the rates ofother personality disorders were substantial in some studies. In a study by Zanarmni and colleagues (23) of inpatients admitted with a probable axis II diagnosis, borderline personality disorder was diagnosed in 60 percent of the cases. The next most frequently diagnosed personality disorder was histrionic personality disorder, with a prevalence rate of 42 percent (subjects in this study and in those discussed below could receive more than one diagnosis). Standage and Ladha (24) reported that 70 percent ofthe cases in a samplc drawn from a general hospital psychiatric inpatient unit were diagnosed with borderline personality disorder; the next highest prevalence was 5 5 percent for dependent personality disorder. In a study by Skodol and colleagues (25) of applicants for admission to a long-term treatment program for severe personality disorders, 62 percent of the subjects received a diagnosis of borderline personality disorder; 49 percent received a diagnosis of avoidant personality disorder. Among a sample of outpatients studied by Morey (26), most of whom had a diagnosis of personality disorder, 32 percent had a diagnosis of borderline personality disorder and 22 percent had a diagnosis of histrionic personality disorder. Hyler and Lyons (27) found a 21 percent prevalence rate ofborderline personality disorder in a sample composed mostly of outpatients referred by psychiatrists who treated a large number of patients with personality disorder; the next highest prevalence rate in the sample was 1 1 percent for obsessive-compulsive personality disorder. The prevalence rates for most personality disorder diagnoses tend to be much lower in data sets drawn from hospital and clinic charts than would be suggested by results of studies using semistructured inter-

views (28,29), but the diagnosis of borderline personality disorder may provide one exception to this trend. Koenigsberg and others (30) obrained the medical chart diagnoses for 2,462 patients. Borderline personality disorder was diagnosed in 1 2 percent of the cases. The next most frequent personality disorder diagnosis was histrionic personality disorder, diagnosed in only 3 percent of the cases. Loranger (3 1) studied the prevalence of personality disorder diagnoses in a sample ofpatients treated in a university-affiliated psychiatric hospital between 1981 and 1985, the first five years after DSM-III was issued. Of a total of 5 ,77 1 patients, 2,840 received a diagnosis of personality disorder. The most common diagnosis was atypical, mixed, or other personality disorder, diagnosed in 32.6 percent of2,916 cases among the 5 ,77 1 patients. Borderline personality disorder was diagnosed in 26.7 percent ofthe cases, and the distant third was dependent personality disorder, diagnosed in 9. 1 percent of the cases.

1016

October1991

Hospital

Vol.42

No.10

Incidence The high prevalence rates that have been obtained for borderline personality disorder call into question the validity of the diagnosis. Borderline personality disorder was not formally recognized as a diagnosis until 1980, but the disorder is now diagnosed more frequently than any other personality disorder. About 1 5 percent of all inpatients and half of all inpatients with a personality disorder receive this diagnosis (32). Examination of the incidence of borderline personality disorder-the rate ofnew cases in the population at risk during a specified period of time (3)-could help clarify findings about the prevalence of the disorder. However, no data on the incidence of borderline personality disorder has been reported. Loranger’s study (31) ofthe prevalence of the disorder did include a comparison of clinical diagnoses made before and after DSM-III was issued. In that study, the most prevalent personality disorder diagnoses made between 1975 and 1979, before DSM-II1 became available, were

and

Community

Psychiatry

“other” (54.3 percent of984 cases diagnosed with a personality disorder), hysterical personality (9. 5 percent), antisocial personality (9.3 percent), schizoid personality (5 .8 percent), and passive-aggressive personality (5.5 percent). From 1981 to 1985, after DSM-III was issued, the most prevalent diagnoses were atypical, mixed, or other (32.6 percent of 2 ,9 1 6 cases of personality disorder), borderline personality disorder (26.7 percent), dependent personality disorder (9. 1 percent), narcissistic personality disorder (5 .9 percent), and antisocial personality disorder (4.6 percent). One could not, however, infer an increased incidence of borderline personality disorder from Loranger’s data. These data show that the prevalence rate for any personality disorder diagnosis increased dramatically, from 19.1 percent ofall cases before DSM-III to 49.2 percent of all cases after DSM-IIl. The increase appears to reflect the conversion to a multiaxial diagnostic system rather than an actual increased incidence of personality disorders. In the development of DSM-iI, the principal reason personality disorders were placed on a separate axis was to encourage clinicians to overcome the tendency to miss or ignore the presence ofa personality disorder in the context

ofa

more

florid

or immediate

clinical syndrome (3 3). Loranger’s data suggest that this educational aim has been accomplished. Millon (34) suggested that the current

prevalence

of borderline

per-

sonality disorder does in fact reflect increased incidence. Millon wrote, “Our contemporary epidemic of borderline personality disorder can be best attributed to two broad sociocultural trends that have come to characterize much of Western life this past quarter century, namely the emergence of social customs that cxacerbate rather than remedy early, errant parent-child relationships and, second, the diminished power of formerly reparative institutions” (34, p. 355). Examples ofdivisive social customs include rapid industrialization, changing sex roles and increased divorce, poor role models in the media, and increased availability

Hospital

and

Community

Psychiatry

of illegal drugs. Examples of diminished reparative institutions include declining schools and religious institutions and the absence of nurturing surrogates due to the scattering ofthe nuclear and extended ftmily. It could also be the case that 15 percent ofall inpatients had borderline personality disorder before DSM-Ill but that they received instead the nonspecific diagnosis of “other” or a diagnosis ofhysterical or passive-aggressive personality. Vaillain (35), however, has argued that the current prevalence rates may be somewhat iatrogenic. He wrote, “These disorders are still usually ohserved only in American cities that have opera houses and psychoanalytic institutes” (35, p. 543). There may be some truth in Vaillant’s suggestion that the prevalence of borderline personality disorder reflects local interest. The high rates of borderline personality disorder reported by Koenigsberg and others (30) and Loranger (3 1) were found in studies conducted at Cornell University Medical College, where considerable attention has been given to the borderline diagnosis. The excessive prevalence, however, is consistent with Kernberg’s formulation (36) of borderline personality disorder as a borderline level ofpersonality organization. Borderline personality organization does not reftr to a distinct personality disorder but rather a level or degree of personality dysfunction. Borderline personality organization may be compared with other types of personality organization rather than with various types ofpersonality disorders such as histrionic, antisocial, or schizotypal personality disorder. Borderline personality organization is distinguished from neurotic personality organization by a predominance of primitive defensive operations centering on the mechanism of splitting; it is distinguished from psychotic personality organization by its maintenance ofa capacity for reality testing. Many if not most inpatients with a personality disorder would likely be functioning at a borderline level of personality organization. The excessive prevalence of bor-

October

1991

Vol.

42

No.

10

derline personality disorder is also consistent with an interpretation of the disorder as representing excessive neuroticism (37). A general trait of personality dysfunction, excessive neuroticism includes elements of impulsivity, hostility, anxiety, depression, self-consciousness, and vulnerability and would likely be characteristic of a substantial proportion ofmentally ill patients.

Sex ratio DSM-III-R

states that borderline personality disorder is more commonly diagnosed in females than in males (38). Akhtar and others (39) verified this clinical impression in a meta-analysis of 23 studies. Across the 23 studies, the average percentage ofcases ofborderline personality disorder in which the patient was female was 77 percent. However, they included results from studies with overlapping data sets and biased samples. For example, they included the results of a study by Loranger and others (40) although the sample in that study was intentionally and explicitly limited to females. Widiger and Trull (22), however, identified 75 studies that provided nonoverlapping and unbiased estimates of the sex ratio for borderline personality disorder. In those studies the average percentage of cases involving female patients was 76 percent, with a 95 percent confidence interval of 73 to 80 percent, a finding consistent with that ofAkhtar and associates. Other researchers have argued, however, that the increased rate of borderline personality disorder in females reflects sex bias by diagnosing clinicians (41). Henry and Cohen (42) administered a 35-item questionnaire

borderline 277

college

based

on

the

personality students

(80

features

of

disorder

to

males

and

197 females). Males scored significantly higher on items concerned with impulsivity, unstable relationships, identity disturbance, and feelings of emptiness. Henry and Cohen indicated that because normal males acknowledge more borderline personality disorder symptoms than normal females, it was inconsistent that females would be more likely than males to receive

1017

the diagnosis. They suggested that clinicians might consider borderline personality disorder traits to be congruent with the masculine sex role and thus find them more acceptable in males, whereas “in women these traits may be seen as less appropriate to [their] sex role, and therefore women may be more likely to be labeled as having borderline personality disorder” (42, p. 1529). Henry and Cohen’s hypothesis suggests that a self-report assessment of borderline personality disorder would provide a different sex ratio than diagnoses based on clinical interviews. However, this has not been the case. Females have typically scored higher than males on the Personality Diagnostic Questionnaire and on the scales for borderline personality disorder in the Minnesota Multiphasic Personality Inventory (MMPI) (43). Kass and associates(44)and Reich (43) reported that the proportion of women who received a diagnosis of borderline personality disorder was not significantly higher than the proportion of women seen at the respective clinics where the data were collected. Both authors suggested that these findings are inconsistent with the sex bias hypothesis, since clinicians did not tend to give the diagnosis to more women than would be expected, given the number ofwomen who sought treatment. However, there is no reason to presume that the proportion of women with borderline personality disorder should be equivalent to the proportion of women treated in a given setting. Simply because there are more females than males at a clinic does not suggest that there will be more females than males with every disorder including borderline personality disorder) that is diagnosed at that clinic (45). In their comparison of clinicians’ diagnoses based on clinical impression versus clinicians’ diagnoses based on a systematic assessment of each DSM-Ill criterion for borderline personality disorder, Morey and Ochoa (46) found that clinicians tended to give a diagnosis of borderline personality disorder to patients who met the criteria of suicidal ges-

tures,

self-damaging impulsivity, affective instability or any cornbination of these criteria, even though the patient lacked additional features of borderline personality disorder. Less experienced clinicians, psychodynamically oriented clinicians, and female clinicians were also more likely to give a diagnosis of borderline personality disorder, and white patients with a low income were more likely to receive this diagnosis. Females were only marginally more likely to be given a diagnosis of borderline personality disorder, and the effect ofgender was insignificant compared with the other effects noted. Widiger and Tail! (22) compared the sex ratios obtained in studies of borderline personality disorder that used a semistructured interview (N=34) to those obtained in studies that used an unstructured clinical interview (N=41). Ifsex biases affect clinicians’ diagnoses, one would cxpect the proportion of women with the diagnosis to be higher in studies that did not use a structured interview. However, the opposite result was found. The average proportion of women with the diagnosis of borderline personality disorder was 80 percent in studies that used a semistructured interview and 73 percent in studies that used an unstructured interview (z=3.05, p< .01).

1018

October

and

disorder could be adequately cxplained using more traditional concepts ofpersonality disorder, such as dependent and histrionic personality disorder, or by more traditional concepts of personality, such as neuroticism. The inclusion of additional measures of personality and personality disorder would be very important to the internal validity of the study design but would escalate the cost considerably. However, a variety of alternative approaches

to assessment

that

might

Methodological innovations A reasonable epidemiologic study of borderline personality disorder might require the administration of a semistructured interview by experienced clinicians to more than a thousand subjects chosen by a probability sampling. The cost of such a community study would be substantial and perhaps prohibitive. One may then ask whether an epidemiologic study ofborderline personality disorder at this time would be worth the cost. A large-scale epidemiologic study would likely provide useful esti-

reduce the cost have been proposed. These strategies include the use of lay interviewers, use ofdata that have been recoded from earlier studies, telephone interviews, and self-report inventories. Each of these options will be discussed in turn. Lay interviewers. The success of the ECA study was due in part to the development of a structured interview that could be administered by persons with little or no professional or clinical experience (4). Antisocial personality disorder could be included in the Diagnostic Interview Schedule form used in the ECA study because the criteria for this diagnosis, such as unemployment and illegal activity, require relatively little clinical inference and judgment. It is questionable whether a lay interviewer could assess the more subtle criteria for borderline personality disorder, such as identity disturbance or chronic feelings of emptiness and boredom (7). However, lay interviewers could be trained to assess the criteria for borderline personality disorder; in fact, some studies of the disorder have successfully used them (47). One advantage is that lay interviewers lack the clinical assumptions and expectations that tend to bias findings (7). On the other hand, training and supervising them can be expensive. Widiger and associates (47) found that these activities required more time and effort than

mates

would

.

ofprevalence,

risk

ftctors,

and

correlates, but these data would still be ambiguous because the validity of the diagnosis is still in question. It would be especially important to assess whether findings about the epidemiology of borderline personality

1991

VoL 42

No.

10

have

been

required

if experi-

enced clinicians had been employed. However, only 84 subjects were interviewed in their study. The benefits of using lay interviewers may outweigh the cost when sample sizes exceed 5(X).

Hospital

and

Community

Psychiatry

Recoding existing data sets. Swartz and associates (1 6) derived epidemiologic data on personality disorder by recoding data collected using the DIS in the ECA study. However, the DIS obtained axis I symptoms

that

had

occurred

during

the past year such as suicide attempts, anxiety attacks, anxiety for one month or more, tenseness or jumpiness, and three or more depressive symptoms. These symptoms are likely to be seen in borderline patients during a one-year period, but inferring the presence of borderline personality disorder on the basis of axis I symptoms that are known to have occurred only during the last year risks confusing a personality disorder with a mood disorder. It is not surprising that Swartz and associates found substantial comorbidity of borderline personality disorder with mood and anxiety disorders, given that the respective diagnoses were based on the same diagnostic criteria. In a separate report by Swam and associates (48), diagnoses of borderline personality disorder using DIS data showed significant agreement with diagnoses using the DIB, but these findings could also raise questions about the validity of the DIB borderline personality disorder diagnosis. To the extent that borderline personality disorder can be diagnosed on the basis of axis I symptoms that have occurred during the past year, borderline personality disorder may be no more than another name for a nonspecific assortment ofaxis I symptomatology. The recoding ofexisting data sets saves the expense of collecting new data, but one does need to ensure that the algorithm provides a valid ftcsimile ofthe diagnostic criteria for borderline personality disorder. It is particularly important to ensure that the symptoms are lifelong and chronic, given the ease with which axis I disorders are confused with axis II disorders (28,49). Telephone interviews. Telephone interviews can provide a substantial savings in time and effort. The epidemiologic interviews of 797 subjects by Zimmerman and Corycli (17) were conducted for the most part by telephone. Only about 27

Hospital

and

Community

Psychiatry

percent ofthose interviews were conducted in person. Kendler (50) questioned the validity ofthese telephone assessments, which used the Structured Interview for DSM-III Personality Disorders. He noted, for example, that the SIDP requires observation of nonverbal responses to assess the schizotypal personality

disorder

criterion

of made-

quate rapport in fitce-to-face interactions (19). There is also a risk that a telephone interview will provide an overly simplistic assessment. Zimmerman and Coryell (17) did not indicate the average length ofeach interview, nor did they provide any control for confidentiality (that is, the subjects may not have been alone during the interview). They mdicated that there were no significant differences between the prevalence rates obtained using flice-to-face interviews and those obtained using telephone interviews, but the actual data were not provided. Zimmerman and Coryell subsequently provided the correlation ofthe PDQ and SIDP scores for 697 of the subjects (19). Agreement about the presence of a diagnosis of borderline personality disorder was not substantial, but it was statistically significant (k=.30, p< .05); agreement in subjects’ total scores on items assessing borderline personality disorder was also small, but again it was significant (r=.39, p

Epidemiology of borderline personality disorder.

The limited epidemiological data available on borderline personality disorder suggest that the prevalence of the disorder is between .2 and 1.8 percen...
2MB Sizes 0 Downloads 0 Views