Under Virginia law, emergency services evaluators employed by 40 community service boards (CSBs) are the gatekeepers to involuntary commitment. This statewide study examined the characteristics of these emergency evaluations and the factors that predict whether involuntary commitment proceedings are initiated.
In Virginia, the process of emergency evaluation usually begins when an individual experiencing a mental health crisis seeks services or is referred by a family member, friend, or health professional in the community for evaluation by a CSB. Individuals in crisis are often unwilling to go to a CSB to be evaluated. If there is concern that the individual is at risk of self-harm or of harming others, the CSB clinician will contact a court magistrate to request an emergency custody order (ECO). An ECO allows a law enforcement officer to take a person into custody and provide transportation to a hospital or other convenient location where a CSB emergency services clinician can provide a face-to-face evaluation. A law enforcement officer, based on personal observation, may also take action to bring an individual experiencing a mental health crisis to the local jail, an emergency department, or a mental health facility.
Individuals in crisis sometimes seek treatment at hospital emergency departments or are taken there directly by rescue squads or family members. In many of these cases, emergency evaluations do not involve a CSB clinician unless the hospital seeks an ECO or a temporary detention order (TDO). In these cases, care is usually provided in the hospital’s psychiatric unit.
When conducting emergency evaluations, CSB clinicians must make a fairly rapid decision whether to pursue involuntary detention, often on the basis of limited information. Yet the decisions can have important consequences both for the individual, who suffers a loss of personal autonomy and liberty, and for the government, which often covers the cost of care. If a CSB clinician believes that the criteria for involuntary commitment are met, the clinician locates an available bed in an acute care facility and requests a TDO from a magistrate authorizing the facility to hold the individual pending a hearing before a judge. A petition for a civil commitment hearing is usually filed simultaneously with the request for the TDO. The hearing must take place within a very short time frame (typically 48 hours in Virginia—the shortest emergency detention period in the country) (1).
In Virginia, as elsewhere, persons are subject to involuntary hospitalization if they are found by a judge to pose a danger to self or others or are unable to care for themselves and if there is no less restrictive alternative to hospitalization. In 2007, when this study was conducted, civil commitment hearings resulted in involuntary commitment to inpatient treatment (for up to 180 days) of approximately 55% of patients, voluntary hospitalization of 25% of patients, and dismissal of the case for about 15% of patients. Although mandatory outpatient treatment in lieu of involuntary hospitalization is authorized, it was rarely used.
A number of studies of predictors of admission for psychiatric hospitalization did not include the possible interplay between the availability of community resources and risk of involuntary commitments (2,3). For example, a study by Rabinowitz and colleagues (4) of 2,072 people evaluated in a public hospital emergency room reported that patient characteristics, such as psychosis and violence, were the most powerful predictors of admission. Olfson and colleagues (5) reported that individuals with schizophrenia, nonadherence to antipsychotic medication, substance use disorders, other positive symptoms, and low social function were at increased likelihood for a hospital admission, but they did not discuss the likelihood of involuntary admission. Flaherty and Fichtner (6) found that over time, emergency psychiatry training appeared to better prepare residents to identify patients who would benefit from outpatient management and community-based services. Although the authors did not study involuntary hospitalization, they found that residents’ hospitalization of psychiatry patients declined by 40% over the training period.
Studies of emergency evaluations of relevance to this study have typically focused on decisions in a particular locality or venue, such as a single emergency department. In general, the studies have shown that the probability that a patient will be hospitalized or held involuntarily may be affected by characteristics of the examining clinician, for example, a heavy caseload; severity of the patient’s condition; availability of hospital beds; setting of the evaluation; and availability of alternatives to hospitalization (7–10). In contrast, a study of emergency evaluations in Ontario, Canada, by George and colleagues (9) found that admission rates were primarily determined by severity of the patient’s illness rather than by bed availability.
The need for involuntary inpatient treatment may be reduced or avoided for many people in crisis if intensive community-based mental health services are in place to engage the person through case management to support treatment adherence (11). In Oregon, a lack of resources was cited as an obstacle to diverting clients from the commitment system (12). Looking specifically at outpatient emergency housing, Slagg (13) found that given the similarity in characteristics of the emergency room and some alternative settings, some hospitalized patients would be appropriate candidates for alternative treatment. Faulkner and colleagues (14) found that availability of an acute care facility for hospitalization lowered the cost and significantly increased the use of the commitment process by local peace officers. Similarly, Turkheimer and Parry (15) argued that the gap between the use of civil commitment procedures in practice and the statutory requirements is largely due to the absence of viable, less restrictive alternatives to inpatient treatment.
Blitz and colleagues (16) highlighted the need to look at the array of options for treating psychiatric emergencies, such as mobile crisis teams, holding beds, access to emergency medications, and medical evaluations rather than a dichotomy of hospitalization versus no hospitalization. More research is needed on the ways in which establishing and utilizing intensive community services and supports can reduce the need for involuntary inpatient treatment (17).
The purpose of this study was to determine predictors of actions to initiate involuntary commitment of individuals experiencing a serious mental health crisis, with a particular focus on indicators of availability of community mental health services and supports. To our knowledge, this study is the first to characterize gateway emergency evaluations for an entire state mental health system.
The characteristics of face-to-face emergency mental health crisis contacts that occurred throughout the Commonwealth of Virginia over approximately one month were obtained as a part of a study on mental health law reform in Virginia. For each contact, a licensed emergency mental health professional sought to determine the level and type of treatment needed by a client experiencing a severe mental health crisis. If the client refused treatment, the professional had the option to seek detention for consideration of involuntary commitment for inpatient treatment.
For this study, mental health clinicians employed by the state’s CSBs provided blind-coded chart review information by using a two-page questionnaire with structured fixed-choice response coding, such as a question about time to locate a bed, as well as options for open-ended responses. The questionnaire included items about client characteristics, circumstances of the evaluation, presenting behavior, clinical status, and evaluator opinion and disposition (particularly whether involuntary action was taken). In addition, the evaluators were asked to indicate whether availability of specific services or resources would have helped address the client’s needs.
An evaluation team at the University of Virginia analyzed the blind-coded data provided by clinicians. This study was reviewed by the University of Virginia Human Investigation Board and was considered exempt because of the use of deidentified data.
A total of 3,808 evaluations were conducted during June 2007. No clients were evaluated more than once during the month of the study. For purposes of the analysis, we excluded all juveniles and adults who were incarcerated at the time of the evaluation because these categories are governed by distinct legal procedures. Previously committed psychiatric inpatients being evaluated for a recommitment after the commitment period expired due to severity of illness were also excluded from analysis. Responses with significant missing data, such as the final disposition or items about needed services, were also excluded from the analyses. The final sample included 2,624 clients.
This study explored the relationships between the outcome of an emergency evaluation and a number of factors identified at the time of assessment. Each client was coded into one of two groups according to the immediate disposition of the evaluation (outcome): involuntary action taken or involuntary action not taken. Sociodemographic variables, services needed, and other characteristics of the outcome groups were compared by t test and chi square test, as appropriate. The correlations between the outcome and the independent variables were analyzed by using the two-tailed parametric method.
A logistic hierarchical multiple regression was conducted, with five blocks (or steps) of variables. The first step included services that the examining clinician believed would have helped address the client’s needs had they been available. The second step included the location of the assessment. The hospital emergency room and the police station were considered proxies for a lack of community mental health resources or services for evaluating the person. The third step added being in police custody. The fourth step included the client’s mental health and behaviors at the time of the evaluation as well as the other variables related to meeting criteria for an involuntary commitment. The last step included the sociodemographic characteristics.
To avoid any collinearity problem, the variance inflation factor was calculated in the model. Data were analyzed by using SPSS 19 statistical software.
The sociodemographic and other characteristics of the sample, including the presence of mental illness and substance abuse and dependence, are described in Table 1. The clients’ ages ranged from 18 to 95 (mean±SD=40.2±15.3); 47% were male, 68% were white, 27% were black, and 6% were from other racial or ethnic groups. An action to initiate involuntary commitment was taken for 1,090 (42%) clients. The other clients were referred to the CSB (N=479, 18%) or a private provider (N=346, 13%), were referred for voluntary hospitalization or voluntary outpatient services (N=581, 22%), or declined services altogether (N=128, 5%).
Table 1Characteristics of clients who were or were not subject to actions to initiate an involuntary commitment
| Add to My POL
|Total (N=2,624)||Subject to actions (N=1,090)||Not subject to actions (N=1,534)|
|Currently in treatment||1,403||54||628||58||775||51||<.01|
| With family||1,382||53||549||50||833||54|
| With nonrelatives||268||10||99||9||169||11|
| Supervised living or with support||135||5||73||7||62||4|
| Other or unknown||139||5||70||6||69||5|
|Place of assessment||<.01|
| Community service board||762||29||147||14||615||40|
| Police station||185||7||101||9||84||6|
| Patient’s home||109||4||37||3||72||5|
| Mental illness and substance abuse and dependence||677||26||318||29||359||23|
| Mental illness and no substance use and dependence||1,618||62||726||67||892||58|
| Substance abuse and dependence and no mental illness||257||10||43||4||214||14|
| Neither mental illness nor substance abuse and dependence||72||3||3||<1||69||5|
|Under the influence of drugs or alcohol||<.01|
|Danger to self||987||38||659||61||328||21||<.01|
|Danger to others||403||15||328||30||75||5||<.01|
|Inability to care for self||893||34||699||64||194||13||<.01|
There were no significant differences in sex or race between clients who were subject to an involuntary action and those who declined services or who were not subject to involuntary action; however, fewer patients who declined services were considered to be a danger to themselves (N=12, 1.2%) or others (N=4, .9%) or were considered unable to care for themselves (N=17, 1.8%).
Significant differences were found in basic characteristics of clients subject to or not subject to involuntary action (Table 1). In summary, involuntary action was more likely if the client was currently in mental health treatment, had a mental illness, showed an overt indication of danger to self or others, showed a personal inability for self-care, had psychotic symptoms, lived alone or with support, had insurance, was assessed at a hospital emergency department or police station, or was in police custody.
Services or resources that would have helped the clinician address the client’s mental health crisis in the community but were not available are presented in Table 2. Overall, two-thirds (N=1,746, 67%) of the clients did not need additional resources or services. The clinicians reported that 614 (23%) needed at least one unavailable service, 172 (7%) needed two, and 92 (3%) needed three or more. Residential crisis stabilization would have helped 134 (12%) clients who were subject to an involuntary action. Short-term crisis intervention would have helped 161 (11%) clients who were not subject to an involuntary action. Immediate medication evaluation would have helped 15% (N=396) of the total sample. Clinicians identified several other services or resources that would have helped 256 clients, but none of the alternatives was mentioned frequently enough to be included in the logistic hierarchical multiple regression.
Table 2Services or resources identified by clinicians that would have helped address the needs of clients who were or were not subject to actions to initiate involuntary commitment
| Add to My POL
|Total (N=2,624)||Subject to actions (N=1,090)||Not subject to actions (N=1,534)|
|Service or resource||N||%||N||%||N||%||pa|
| Medication evaluation||396||15||137||13||259||17||<.01|
| Safe transportation||104||4||33||3||71||5||<.05|
| Temporary housing||136||5||35||3||101||7||<.01|
|Community-based mental health treatment|
| Short-term crisis intervention||249||10||88||8||161||11||<.05|
| Residential crisis stabilization||241||9||134||12||107||7||<.01|
| In-home crisis stabilization||156||6||64||6||92||6||ns|
Table 3 shows the variables included in each step of the logistic hierarchical multiple regression as well as the Pearson correlation between the outcome and independent variables. The model shows that the addition of variables, except sociodemographic variables, in each step explained significantly more of the total variation (Table 4). Sociodemographic variables explained relatively little of the total variation. The variance inflation factor (VIF), which quantifies the severity of multicollinearity by measuring the extent that the variance of a coefficient is increased because of collinearity, was less than 1, indicating low collinearity.
Table 3Variables in the regression model and their association with actions to initiate involuntary commitment
| Add to My POL
|Step 1: needed services|
| Immediate medication evaluation||–.06||<.01|
| Safe transportation||–.04||<.05|
| Temporary housing||–.08||<.01|
| Short-term crisis intervention||–.04||<.05|
| Residential crisis stabilization||.09||<.01|
| In-home crisis stabilization||<.01||ns|
|Step 2: location of assessment|
| Police station||.07||<.01|
|Step 3: police custody||.24||<.01|
|Step 4: mental health conditions and behaviors|
| Mental illness||.21||<.01|
| Under the influence of drugs or alcohola||.10||<.01|
| Psychotic symptoms||.44||<.01|
| Danger to self||.40||<.01|
| Danger to others||.34||<.01|
| Inability to care for self||.54||<.01|
|Step 5: demographic and social characteristics|
| Living arrangementa||.06||<.01|
| Currently receives treatment||.07||<.01|
Table 4Variables contributing to actions to initiate involuntary commitment of clients referred for emergency mental health evaluationa
| Add to My POL
|Step 1||Step 2||Step 3||Step 4||Step 5|
|Needed services (reference: did not need service)|
| Immediate medication evaluation||–.32||.73||<.01||–.20||.82||.10||–.22||.80||.09||–.16||.85||.33||–.15||.86||.39|
| Safe transportation||–.27||.76||.23||–.30||.74||.20||–.24||.79||.33||–.26||.77||.41||–.27||.76||.40|
| Temporary housing||–.71||.49||<.01||–.64||.53||<.01||–.67||.51||<.01||–.38||.68||.17||–.41||.66||.15|
| Short-term crisis intervention||–.28||.76||.06||–.38||.68||.01||–.44||.64||<.01||–.61||.54||<.01||–.62||.54||<.01|
| Residential crisis stabilization||.81||2.24||<.01||.82||2.27||<.01||.85||2.35||<.01||.45||1.56||.02||.41||1.50||.04|
| In-home crisis stabilization||–.12||.89||.52||–.07||.93||.71||–.10||.91||.62||–.28||.76||.24||–.29||.75||.24|
|Location of assessment (reference: not in hospital)|
| Police station||1.35||3.87||<.01||.52||1.69||<.01||.76||2.13||<.01||.74||2.09||<.01|
|Police custody (reference: no)||1.23||3.42||<.01||.84||2.32||<.01||.85||2.35||<.01|
|Mental health conditions and behaviors (reference: absent)|
| Mental illness||.79||2.20||<.01||.83||2.28||<.01|
| Under the influence of drugs or alcohol (reference: not under the influence)|
| Psychotic symptoms||1.67||5.34||<.01||1.73||5.62||<.01|
| Danger to self||1.91||6.75||<.01||1.89||6.65||<.01|
| Danger to others||1.24||3.44||<.01||1.23||3.44||<.01|
| Inability to care for self||1.93||6.89||<.01||1.95||7.01||<.01|
|Demographic, social, and other characteristics|
| Race (reference: African American)|
| Living arrangement (reference: lives with family)|
| Lives alone||–.10||.91||.53|
| Lives with nonrelatives||.07||1.07||.74|
| Lives with support||.23||1.26||.39|
| Other or unknown||–.26||.77||.36|
| Currently receives treatment (reference: no treatment)||–.01||.99||.94|
| Insurance (reference: no insurance)||–.08||.92||.52|
Step 1 included the services or resources that would have helped clinicians address the clients’ need at the time of crisis. The outcome for more than half of the clients (N=1,553, 59%) was predicted correctly by using the independent variables included in step 1. Of the total variance, 3% was explained in this step. For example, an evaluation had twice the likelihood of a voluntary outcome if the clinician had stated the client’s need for residential crisis stabilization.
Step 2, adding the location of the assessment, explained an additional 9% of the variation. An action to initiate involuntary commitment was taken three times more often when clients were assessed in hospitals and police stations versus other community locations, such as a CSB or the client’s home. By the end of step 2, 12% of the total variation was explained by the model.
Step 3 added whether the client was in police custody at the time of evaluation, which contributed significantly to the model. This one variable added 6% to the variation explained, resulting in a total of 18% of the variation explained. The model with three blocks correctly predicted the outcome for 1,705 (65%) clients. Clients who were in police custody were three times more likely to be subject to involuntary commitment.
Step 4 included the most highly correlated clinical predictors of the outcome. Individuals who had a mental illness, who were under the influence of drugs, who had active psychotic symptoms, and who were considered to be a danger to self or others or who displayed an inability for self-care were more than twice as likely to be subject to involuntary commitment as individuals who lacked these clinical characteristics. These variables boosted the results of the model substantially, explaining 44% of the variance, bringing the total variance explained to 62%. By this step, 2,145 (84%) client outcomes were correctly predicted.
Step 5 added the demographic and social characteristics to adjust for differences between the groups.
As expected, the acuity and severity of the individuals’ presenting symptoms accounted for a substantial portion of the variance in predicting whether emergency services clinicians initiated action to initiate involuntary commitment. The study also showed that involuntary commitment was twice as likely if the emergency services clinician reported that voluntary residential crisis stabilization would have been helpful to the client had it been available. Overall, residential crisis stabilization, short-term crisis stabilization, and immediate medication would have helped clinicians provide immediate care to 715 (27%) individuals they evaluated. For over a third of clients (N=878, 33%), the clinician confirmed that at least one of the services listed in Table 2 would have helped to better address the individual’s need.
The study also found that action to initiate involuntary commitment was taken three times more often when clients were assessed in hospital emergency departments and police stations compared with other community locations. It is likely that evaluations were conducted in those locations because there was no suitably secure mental health facility for assessment and stabilization of the client.
Sociodemographic characteristics did not explain much of the variation in outcome in the predictive model, although initiation of involuntary commitment was more likely among older adults. It was noteworthy that the outcome was not related to client’s race-ethnicity or living situation.
This study had limitations. Data were not collected on the specific DSM-IV diagnosis, so we are unable to better define the type of mental illness of the clients and how it might relate to the outcome. The clinicians reported the availability of community mental health services. Although we expect that these reports were reliable, no documentation of all available community services was available for the study. Although the emergency services clinicians were not asked directly whether the availability of the needed services would have avoided action to initiate involuntary commitment, we think this is the most plausible interpretation of the findings. It is possible, however, that the clinicians, knowing that the commitment process was under scrutiny, used their survey responses to advocate for closing gaps in intensive services or to justify a decision to seek involuntary commitment.
Strengths of the study included a prospective design that evaluated all individuals who experienced a serious mental health crisis in every CSB in the state. The study was conducted over a month, resulting in a large sample. In addition, the lack of community resources reported by clinical evaluators permitted a much needed investigation of community mental health resources in a predictive model.
Although the results were not conclusive, this study suggested that the rate of actions to initiate involuntary care (TDOs and eventual commitments) would fall if individuals experiencing a mental health crisis could be directed to intensive community services with trained personnel as an alternative to hospitalization. The findings provide a sound empirical basis for the claim that investing in a continuum of crisis stabilization and other intensive outpatient services would reduce the need for involuntary interventions.
A portion of Dr. McGarvey's salary as director of research for the Commission on Mental Health Law Reform in the Commonwealth of Virginia was paid by state grant funds (Commonwealth of Virginia Interagency Agreement 720C-04079-07M00). The other authors report no competing interests.