Predictors of Recurrent Use of Psychiatric Emergency Services

Previous studies have indicated that 2% to 18% of adult patients who repeatedly visited psychiatric emergency services generated 21% to 65% (1–5) of the total clinical activity in these settings. Recurrent use of emergency services is a significant cost burden (6,7). Prevalence rates of recurrent visits depend on how researchers define recurrence (8); however, frequent visits raise the question of which modalities are best adapted to psychiatric emergency care. Although deinstitutionalization is an important trend in psychiatric care (9), an increasing number of patients seek psychiatric care from emergency services (10).

Authors of past studies have defined frequent visits arbitrarily, with no consensus about the definition. Some have defined frequent visits as more than two above the mean standard deviation (1,2), and some have used the upper tenth percentile of the sample (4). Others have defined frequent visits as one visit (3) or six visits (2,7) per year, four visits in a quarter (2), or at least six visits in six years (11). Moreover, sample characteristics differ between studies, and some studies have limited the sample to patients with specific diseases, such as psychotic disorders (11), or to “incident” patients (that is, those newly detected as recurrent visitors) with no previous visits to psychiatric services (4). Unfortunately, previous investigations have focused mainly on recurrent visits over the short term, usually within a single year. To date, only a few studies have extended their follow-up to longer periods (1–4,11).

Several studies have attempted to identify clinical and sociodemographic characteristics of frequent users of psychiatric emergency services. Frequent visitors were mostly men (1–4,7,12) and lived in a more precarious socioeconomic environment (1,2,4,5,7,8,11). Some studies have established that frequent visitors were younger (3,11,12), contrary to other studies (1,4). Frequent visitors were more likely to have a diagnosis of a personality disorder (1–3,10) or of schizophrenia or another psychotic disorder (1,3,11,12); they were also found to have more severe psychiatric symptoms (4). Several authors found an association between frequent visits and having a substance use disorder (8,11,12), whereas others did not (7) or considered this relationship to be dependent on the severity of patients' clinical status (5). Finally, a history of previous psychiatric hospitalizations, greater use of various types of psychiatric and nonpsychiatric health services, and more intense utilization of regional mental health services have been found to be related to frequent visits (2,7).

Recurrent visits to emergency services by patients whose condition is not urgent can result in negative attitudes toward patients and avoidance reactions among the staff of emergency services, which some have attributed to emergency staff members’ lack of confidence in their ability to manage patients who make frequent visits (10). Development of specific care models for these patients requires identification of their characteristics. Thus we assessed sociodemographic and diagnostic characteristics that predicted recurrent visits. To address the lack of longer follow-up in previous studies, we conducted a three-year retrospective study of community-dwelling patients who visited psychiatric emergency services. We compared patients who stopped making frequent visits (three or more per year) after one year with those who made recurrent visits for more than one year. In addition, we included both patients who were newly detected as recurrent visitors (“incident” group) and those who were known to be recurrent visitors at study entry (“prevalent” group). We hypothesized that frequent use of psychiatric emergency services in itself would predict recurrent use over the next three years.

Methods

Sample Selection

The psychiatric emergency inpatient and outpatient ward—Unité d’Accueil des Urgences Psychiatriques (UAUP)—is part of the emergency division of the University Hospitals of Geneva. Its catchment area (about 700,000 individuals) is Geneva County and the French zone located close to the Swiss border. A multidisciplinary team of psychiatrists, nurses, and administrative staff offers evaluations and interventions 24 hours a day, seven days a week. Depending on the nature and severity of patients’ distress, they may be hospitalized or referred to outpatient services, private psychiatrists, or general practitioners. A primary care nurse performs an initial screening via clinical interview to determine whether to refer patients for further in-depth general medical or psychiatric assessment. Patients who need a psychiatric assessment receive a detailed evaluation, supervised by a senior psychiatrist, that includes assessment of psychiatric diagnoses according to DSM-IV criteria (13). These evaluations are part of the routine clinical assessment in the UAUP. Our use of anonymous data from these assessments was approved by the Ethics Committee of the University Hospitals of Geneva.

This retrospective cohort study analyzed data from computerized psychiatric administrative and medical records of the 4,322 patients age 18 years or older who visited the UAUP in 2008. To determine which patients in this cohort were frequent visitors, no exclusion criteria were used; patients were included without regard to their clinical status, and the study included patients who were newly detected as recurrent visitors (“incident” group) and those known to be recurrent visitors in 2008 (“prevalent” group). Frequent visitors were defined as patients who made at least three visits between January and December 2008, which replicates the arbitrary frequency criterion adopted by Saarento and colleagues (4) in their three-year follow-up. Three frequent visitors were excluded from our study because of missing data, and the final sample consisted of 210 patients who made at least three visits to the UAUP in 2008. The number of annual visits was extracted from the computerized medical files. Among the 210 patients, recurrence was defined as at least three visits in any of the following years (2009–2011) (46 patients), whereas nonrecurrence was defined as no further visits in any of the following years (2009–2011) (164 patients). [A figure illustrating the composition of these groups is available in an online data supplement to this article.]

Sociodemographic and Diagnostic Characteristics

Sociodemographic data gathered at the time of the first visit in 2008 were extracted, including age, gender, marital status (single or divorced versus separated, married, or widowed), professional activity (unemployed, active worker or student, or retired or homeworker), and citizenship (Swiss or non-Swiss). DSM-IV psychiatric diagnoses were grouped into five main categories, as in previous studies (8,14): mood disorders, anxiety disorders, psychotic disorders, substance use disorders, and personality disorders.

Data Analysis

Descriptive statistics were used to explore sample characteristics. To predict the probability of recurrence of visits to the UAUP, a logistic regression model was built with recurrence in any of the follow-up years as the dependent variable. This variable was coded as a binary variable—recurrence over three years (coded 1) versus nonrecurrence (coded 0). Three sets of independent variables were included. The first focused on sociodemographic characteristics in 2008: age, gender, marital status, professional activity, and citizenship. The second focused on diagnosis at the first visit in 2008: number of diagnoses and type of disorder (that is, using the five categories). The third set included recent use of the UAUP: number of visits in 2008 and visits before 2008. All the analyses were conducted with SPSS, version 20.0.

Results

Recurrent Visits

Of the 210 patients in the sample, 164 (78%) were in the nonrecurrent group: they visited the UAUP three or more times in 2008 but did not return in the following years. A second group of 46 patients (22%) were in the recurrent group: 26 visited the UAUP again three or more times in 2009, 11 visited three or more times in both 2009 and 2010, and nine patients visited three or more times in 2009, 2010, and 2011 [see figure in the online supplement].

Sociodemographic and Diagnostic Characteristics

Tables 1 and 2 present data on sociodemographic and diagnostic characteristics of the sample, by recurrence group. The mean±SD age of the 210 patients was 38.7±13.5, and 51% were female. Most patients were single, divorced, or separated (N=156, 74%) and unemployed (N=141, 67%). A third (N=74, 35%) were not Swiss citizens. Most had two or more psychiatric diagnoses (mean=2.1±1.2). About half the patients had a mood disorder (N=103, 49%) or a substance use disorder (N=109, 52%). Smaller proportions of the sample had personality (N=84, 40%), psychotic (N=43, 21%), or anxiety (N=29, 14%) disorders. Of the 84 patients with a personality disorder, 75% (N=63) had an emotionally unstable personality disorder (60 with borderline type), 14% (N=12) had an unspecified personality disorder, 5% (N=4) had antisocial personality disorder, 2% (N=2) had dependent personality disorder, and the remaining three patients had paranoid, schizoid, and narcissistic personality disorder, respectively.

Prevalent and Incident Groups

A total of 111 patients were in the prevalent group (known to be frequent visitors at the start of the study because they had three or more visits per year to the UAUP before 2008). A total of 99 patients were in the incident group (that is, newly detected as recurrent visitors in 2008, with no UAUP visit in the 18 months before January 1, 2008). The cutoff of 18 months was used by Saarento and colleagues (4). Of interest, of the 99 patients in the incident group, close to two-thirds (65%, N=64) had no recorded psychiatric history. Of the 111 patients in the prevalent group, 23% (N=25) had also sought previous treatment in psychiatric facilities other than the UAUP. The study results did not change when these 25 patients were removed from the logistic regression analyses.

Number of Visits

The number of UAUP visits during 2008 varied from three to 26 per patient. About half of the 210 patients (54%, N=113) had three visits to the UAUP in 2008, 20% (N=42) had four, 11% (N=22) had five, and 16% (N=26) had six or more visits. The mean time between two visits was 59.8±41.7 days (range 2–195 days).

Determinants of Recurrent Visits

The logistic regression model analyzing determinants of recurrence showed that no sociodemographic characteristics distinguished patients with recurrent visits from those in the nonrecurrent group (Table 3). Patients with a personality disorder were four times more likely than patients with mood, anxiety, substance use, or psychotic disorders to have recurrent visits over the three-year period. The number of visits before 2008 was a significant predictor of having recurrent visits in the following years. However, the number of visits in 2008 did not predict recurrent visits.

A final model included only the two significant predictors (personality disorder diagnosis and visits prior to 2008) as independent variables, with recurrence or nonrecurrence as the dependent variable. The model explained 14% of the variance in recurrent visits. Significant predictors of recurrent visits in this model were the same: a personality disorder diagnosis (odds ratio [OR]=2.48, 95% confidence interval [CI]=1.24–4.95, p=.010) and visits prior to 2008 (OR=3.80, CI=1.78–8.11, p=.001).

Discussion

To our knowledge, this is the first study to focus on the prevalence rate and sociodemographic and clinical determinants of frequent visits to psychiatric emergency services in Switzerland. A unique feature is its focus on a highly urban area in which extensive outpatient psychiatric care is available in both public and private settings. The study traced the evolution of recurrent use of psychiatric emergency services over a three-year period. In the absence of a consensus regarding the definition of frequent visits, we decided to follow the arbitrary criterion of three or more visits per year used by Saarento and colleagues (4) because of its clinical relevance in the context of our study.

We found that among patients who had three or more visits in the first year, 13% also had three or more visits in the second year. However, this rate dropped to 5% after two years and remained stable in the third year (4%). These findings stress the importance of extending the duration of longitudinal studies beyond one year to identify the subpopulation with chronically high use and to develop appropriate care alternatives for this vulnerable group.

In contrast with previous findings (1–4,11), sociodemographic factors, such as age, gender, marital status, professional activity, and citizenship, did not predict recurrent visits over three years. As in the study by Schmoll and colleagues (1), which found no gender or age effect, our findings suggest that specific sociodemographic profiles are insufficient to distinguish frequent visitors to emergency psychiatric services. In addition, being foreign born (not a Swiss citizen) was not a significant predictor, contrary to previous findings (15). Therefore, lack of formal social support, financial status, and race-ethnicity may not be the only determinants of long-term recurrent use of psychiatric emergency services, despite evidence from previous studies (1,2,4,5,7,8,11). However, the generalizability of our results depends on country-specific health insurance systems. Indeed, the mandatory insurance system in Geneva covers emergency psychiatric services to the same level as treatment in other specialized psychiatric settings, even for patients who lack financial resources. Patients are generally free from financial restraints in their treatment options. This might explain the difference between our findings and those of Arfken and colleagues (7), who found an association between frequency of emergency service visits and low financial resources. Unlike our study, their study was conducted in a disadvantaged area, in an emergency service facing the challenge of providing high-quality psychiatric care in a resource-poor environment.

A personality disorder diagnosis predicted recurrent visits, which strengthens previous evidence (1–3,11). Notably, a diagnosis of borderline personality disorder was a significant predictor. The prevalence of this disorder may be underestimated because it is frequently misdiagnosed and underdiagnosed even by highly skilled psychiatrists and because it is often comorbid with other mental disorders, such as bipolar disorder (16,17). Borderline personality disorder makes the therapeutic or helping alliance hard to establish and maintain (18), which suggests that having this diagnosis may be related to interpersonal difficulties in general and may explain why it was a significant predictor of recurrent visits. In support of this idea, Arfken and colleagues (7) reported that frequent visitors to an emergency psychiatric service were less likely than other patients to be able to provide the name of a family member or friend.

As noted, in Geneva patients can be admitted to a psychiatric inpatient unit or referred to specialized outpatient crisis services by their psychiatrist or general practitioner. Therefore, frequent visitors do not have to rely on psychiatric emergency services of the general hospital. However, not all frequent visitors to emergency services seek care in traditional outpatient settings, because some may have difficulties with the organizational constraints of such settings. Indeed, previous studies reported that frequent visitors used psychiatric emergency services because of their convenient location and because no appointment was required (7). Further research on features of psychiatric emergency services that facilitate their access and use is warranted to explore this issue.

Our finding of a significant difference between patients in the prevalent and incident groups over the three-year follow-up confirmed the necessity of taking into account a patient’s pattern of visits before study entry. Our hypothesis that the repeated use of psychiatric emergency services in itself may predict recurrence was partly confirmed—the prevalent nature of the visits, but not their number, predicted their recurrence. Patients who had visited the UAUP three or more times per year in the 18 months before study entry were three times more likely than other patients to have three or more repeat visits per year in the following three years. These findings complement those of Saarento and colleagues (4), who concluded that an initial visit does not predict future visits; our study underscored that past repetition predicts future repetition. Seeking psychiatric treatment in nonemergency settings was not associated with recurrent visits to the UAUP.

From a clinical point of view, a promising alternative to simply reducing the number of emergency visits is the creation within the psychiatric emergency service setting of specialized programs specifically adapted to the needs of frequent visitors. Frequent visitors establish therapeutic alliances with many different caregivers in the emergency service, which is open 24 hours a day, seven days per week, with no requirement to make an appointment or other commitments. Others have pointed out that a significant number of frequent users do not seek treatment from outpatient services but regularly visit psychiatric emergency services on their own initiative (8). Psychiatric emergency services typically offer clinical assessment, crisis counseling, and orientation; such services should also consider offering patients the option of engaging in follow-up treatment on site, such as flexible individual sessions or once-a-week open group sessions focusing on crisis work. Ensuring that frequent visitors comply with recommended aftercare is also important, because noncompliance with aftercare is associated with shorter intervals between repeat visits (8).

This study had methodological limitations. First, the size of the sample did not allow us to define a more complex regression model to integrate variables such as precariousness (homelessness and receipt of financial assistance) or occurrence of stressful live events. Similarly, the small sample did not allow for more nuanced analyses of demographic characteristics of patients with recurrent visits in 2010 and 2011. Without replicating the results in a larger sample, we cannot know whether the percentage fluctuations in gender and citizenship status merely reflect a lack of stability in time or reveal actual differences between the nonrecurrent and the recurrent groups. Second, at the initial screening, 17% of the patients who were found to need both general medical and psychiatric care (for example, for abuse of drugs, care of wounds caused by self-injury, or binge drinking) were referred for a general medical assessment and were not included in the study. This may have led to an underestimation of the number of patients with recurrent visits. Finally, even though all psychiatric diagnoses were made under the supervision of a senior psychiatrist, the additional use of traditional, semistructured interviews would have been preferable.

Conclusions

A diagnosis of a personality disorder was found to be a significant predictor of recurrent visits to psychiatric emergency services, whereas sociodemographic characteristics were not. Thus recurrent visits appeared to be determined by patients’ mental health status rather than their underprivileged social status. Patients seemed to select this health care setting as their primary treatment source, as indicated by their recurrent visits. Psychiatric emergency services may benefit from establishing an on-site follow-up treatment program for this clinical population. Because patients who are frequent visitors to emergency services have difficulty establishing therapeutic alliances within traditional outpatient mental health settings, treatment programs created within the emergency setting may ensure better access to care for this population.