Diversion to the Mental Health System: Emergency Psychiatric Evaluations

Methods

This study was a retrospective chart review of patients referred for a psychiatric evaluation via an emergency petition (EP) to the Johns Hopkins Hospital Emergency Room (ER) from January 1, 2002 to April 21, 2003. The Johns Hopkins Hospital is a major urban referral center in Baltimore. The emergency room provides 24‐hour emergency psychiatric consultation services by psychiatric residents, nurses, and attending physicians. A log of adult patients (age 16 or older) who had been brought to the ER by police officers serving an EP is maintained in the ER by the Department of Psychiatry. The Institutional Review Board of Johns Hopkins Hospital approved this study and authorized a waiver of the usual requirement for informed consent.

We reviewed the available medical and psychiatric records of each patient brought to the emergency room because of an EP during the study period. Some evaluees had more than one visit during the study period, in which case each visit was considered a separate case. A case was excluded from the study if no EP was found in the chart, even if the existence of an EP was suggested in the emergency room log or record. There were several potential reasons for missing EPs. Police sometimes assist a patient to the emergency room without completing an EP. Clerical errors, with the clerk's confusing a police referral without an EP with a police referral with an EP, could also account for this discrepancy.

Each record was assigned a unique identifying case number. Study data were recorded on a standardized research form. Every reasonable effort was made to redact potentially identifying information from the EP. Demographic and historical data listed on the EP form were collected and transcribed into a database. The information contained in the narratives describing the reasons for the petition was collected verbatim from the EP. Clinical data were also collected from the emergency room record, including risk assessment, diagnosis, and disposition.

One reviewer assessed dangerousness based solely on the data contained on the EP form. The standard for assessing EP dangerousness was the risk posed to the life or safety of the evaluee or others, using reasonable citizen criteria. If the evaluee was thought to be dangerous, the reviewer also commented on the type and imminence of the risk. Imminent dangerousness reflected the language of the emergency petition statute in 2002, and was defined as posing a risk to the life or safety of the evaluee or others in the near future. The type of risk was identified and classified as a verbal or physical threat to self, a verbal or physical threat to others, or failure to perform adequately the activities of daily living (ADLs). If more than one type of risk was present in an individual case, each risk was recorded independently. When the EP reflected that the individual made a threat to others or a physical threat toward property, that person was considered to be eligible for arrest.

Another reviewer made an assessment of clinical dangerousness based on the entire emergency room record, including the petition for emergency evaluation, emergency room physician and nursing notes, psychiatric consultation notes, and physicians' involuntary certification forms, if present. If no risk assessment (either by a psychiatric consultant or an emergency room attending physician) was found in the record, “insufficient information” was coded. Dangerousness in this context was a clinician‐based standard that followed the language of Maryland's civil commitment statute defined as danger to the life or safety of the evaluee or others when no less restrictive intervention than psychiatric hospitalization is appropriate.

Analyses of the data were conducted with SPSS version 12.01. Categorical variables were assessed for significant differences with two‐tailed Fisher's exact tests whenever possible. A two‐tailed chi‐square test was used when noted. The significance of differences between continuous variables was assessed with a t test or a one‐way analysis of variance (ANOVA), as noted. The correlation between the reviewers' assessments of dangerous was tested with the Pearson's R and Fisher's exact test for significance. The level of statistical significance was set at p < .05.

Results

We found 339 emergency room records corresponding to the cases listed in the police referral record. Among those records, 39 (11%) were excluded because no EP was located in the psychiatric or emergency room files. Therefore, 300 cases were used in the analysis. No significant differences in age, gender, or referral sources were identified between the included and excluded groups. Cases in the excluded group were more likely to have no information recorded for race (Fisher's exact test, p = .008). No significant difference in race was identified between these two groups when we controlled for the missing data.

The evaluees in the study group ranged in age from 16 to 85 years. The mean age of the study sample was 38 years (SD = 14). Fifty‐seven percent of the patients were male. The study sample was predominately African‐American (69%) and white (30%). In 68 (23%) cases, a citizen had initiated and a judge had endorsed the EP. A law enforcement officer (59%) or a health care provider (18%) initiated the remainder.

Among the EPs endorsed by the court, a parent had requested the evaluation in 28 percent of the cases. Other petitioners included adult children (18%), spouses (15%), other relatives (19%), and friends (6%). Neighbors, landlords, concerned citizens, and health care providers (who were either unwilling to initiate an EP without court involvement or ignorant of the process) constituted the remainder of the cases. The petitioner lived with the evaluee in 16 percent of the cases. When only judge‐ordered petitions were considered, 65 percent of the petitioners lived with the evaluee.

The diagnoses and medication types most frequently listed on the EP form are presented in Tables 1 and 2. In most of the cases, the petitioner had not recorded information about the evaluee's treatment history. EPs initiated by law enforcement officers were more likely to be incomplete when compared with petitions initiated by health care providers or the court. Police officers documented prior hospitalizations in 28 percent, previous diagnoses in 14 percent, and prescribed medications in 27 percent of cases. In contrast, judge‐initiated petitions included information about hospitalizations on 88 percent (χ² = 112.4, df = 10, p < .001), previous diagnoses on 77 percent (Fisher's exact test, p < .001), and prescribed medications on 53 percent (Fisher's exact test, p < .001) of the forms. Health care providers were the most likely petitioners to be aware of the evaluee's current medications, as these were listed on 62 percent of the forms (Fisher's exact test, p < .001).

The EP form asks the petitioner whether the evaluee has access to firearms or weapons. In 63 percent of all cases, and on 67 percent of petitions initiated by law enforcement officers, this item was left blank. In an additional 11 percent of the cases, the petitioners stated that they did not know. When answered in the affirmative, a knife (9%) or another household item (2%) was usually listed as the weapon. Access to a firearm was reported in one (0.3%) case in the sample.

On review of the EP narrative information, the reviewer determined that sufficient evidence to support the diagnosis of a mental disorder was present in 74 percent of the cases. Petitions initiated by law enforcement officers were significantly less likely to articulate the presence of a mental disorder (62%; Fisher's exact test, p < .001) than court‐ordered (91%) or health care provider–initiated (95%) petitions. The reviewer found sufficient evidence for documenting dangerousness on the EP form in 92 percent of the total sample. Of these, 80 percent were found to have sufficient evidence for documenting imminent dangerousness (the standard required for initiation of an EP at the time of the study). The frequency with which individual types of dangerousness were found in the EP narratives is listed in Table 3. There was no significant difference between the three referral groups for support of the finding of dangerousness (Fisher's exact test, p = .231). However, support for the finding of imminent dangerousness was more likely to be present in petitions initiated by a judge (88%; Fisher's exact test, p = .005) or a health care provider (82%) than in those initiated by a law enforcement officer (70%). A total of 433 risks were identified among these cases, for an average of 1.56 types of risks per case identified as dangerous. Threats to self (verbal: 42%; physical: 29%) and others (verbal: 26%, physical: 27%) were the most frequently observed types of risk, regardless of the type of referral. Sixty‐one percent of petitions described individuals who made verbal or physical threats of self‐harm.

We analyzed dangerousness types by referral source and show the results in Table 3. When compared with all other referral sources, police‐initiated petitions were more likely to describe verbal or physical threats of evaluee self‐harm. However, police‐initiated petitions were less likely to document threats to others, threats to property, or problems with activities of daily living. In contrast, court‐ordered evaluations were more likely to describe threats to others, threats to property, and problems with activities of daily living. These petitions were significantly less likely to involve self‐directed verbal threats. Petitions initiated by health care providers were not readily distinguished from the other groups by risk type, except that these petitions were less likely to involve a physical threat against the evaluee's self.

We found that 140 (47%) petitions described at least one behavior that could have led to an arrest (i.e., danger to others or property). Such behavior was described on 77 percent of court‐initiated petitions, 55 percent of health care provider petitions, and 33 percent of police‐initiated petitions (Fisher's exact test, p < .001). These individuals were significantly more likely to have diagnoses of schizophrenia (18% versus 5%; Fisher's exact test, p < .001) or bipolar disorder (22% versus 9%; Fisher's exact test, p = .002) listed on the EP. They were more likely to have prior hospitalizations (60% versus 29%; Fisher's exact test, p < .001), more likely to be currently receiving treatment (36% versus 20%; Fisher's exact test, p = .019), and more likely to be noncompliant with prescribed medication (43% versus 26%; Fisher's exact test, p = .01).

Among EPs describing possible behaviors that could have lead to arrest, the reviewer found the presence of a mental disorder at a significantly higher frequency compared with other EPs (89% versus 62%; Fisher's exact test, p < .001). The rates of dangerousness (99% versus 86%; Fisher's exact test, p < .001) and imminent dangerousness (90% versus 64%) found by the reviewer were also higher.

After reviewing the complete emergency room record, we found evidence of clinical dangerousness (the same level of dangerousness that would be required for civil commitment) in 71 percent of the cases (Table 4). In five percent of the cases, there was insufficient information to make that determination. In each of those cases, a psychiatric consultation was not requested or was missing from the record.

A comparison of our findings of dangerousness as reflected in the EPs versus clinical dangerousness as reflected in the emergency room record is presented in Table 4. A small but statistically significant correlation was found between the two reviewers' assessment of imminent dangerousness based on EP review and clinical dangerousness determined by reviewing the entire ER record (Pearson's r = .178, exact significance = .004). Evaluees whose emergency petitions were initiated by law enforcement officers were significantly less likely to be found clinically dangerous (61%; Fisher's exact test, p < .001) than were those with court‐ordered (87%) or health care provider‐initiated (86%) petitions.

The frequencies of clinical diagnoses made in the psychiatric emergency room are presented in Table 5. More than one psychiatric diagnosis was applicable per case. Major mental illness and substance abuse were the most commonly observed comorbid diagnoses. The most commonly diagnosed condition was a substance‐related disorder (29% of all cases). Among the major psychiatric illnesses, bipolar disorder (15%) and schizophrenia (13%) were the most frequently made diagnoses. A substantial number (22%) of nonpsychotic disorders, such as personality disorders, adjustment disorders, substance‐induced mood disorders, and depressive disorder, not otherwise specified, were also diagnosed. Among individuals who could have been arrested, rates of bipolar disorder (25% versus 14%; Fisher's exact test, p = .027) and schizophrenia (24% versus 11%; Fisher's exact test, p = .004) were made more frequently than in other evaluees.

Sixty‐three percent of evaluees brought to the ER on EPs were admitted to a hospital. Four percent of all evaluees were returned to some other custodial situation (such as a nursing home or police custody). Only four cases were released to police custody, and only two of these cases had documented behavior that could have resulted in an arrest noted on the EP. Of the evaluees admitted to a hospital, 34 percent were involuntary psychiatric admissions, 62 percent were voluntary psychiatric admissions, and 4 percent were medical admissions.

The mean age of evaluees admitted to a hospital was 40. Evaluees who were discharged were significantly younger (mean age = 33; ANOVA, p = .001). Evaluees who were medically admitted tended to be older (mean age = 44, ANOVA, p = .004). The evaluee's race and gender had no statistically significant effect on the dispositions of the cases.

Evaluees were discharged from the emergency room in 37 percent of the cases. Evaluees who were brought in on a police‐initiated petition were significantly more likely to be discharged from the emergency room (45%; Fisher's exact test, p = .001) than were those brought on petitions initiated by a health care provider (24%) or endorsed by a judge (27%). Petitions initiated by a health care provider were significantly less likely to result in a discharge from the emergency room (Fisher's exact test, p = .03) than were those from all other referral sources. When discharged, 65 percent of the patients who were released received a referral for outpatient psychiatric treatment.

Persons with evidence of a mental disorder documented on the EP form had a higher rate (66%; Fisher's exact test, p = .004) of psychiatric admission than did those without a documented mental disorder (47%). Those with evidence of dangerousness documented on the EP form did not have a statistically significant higher rate of psychiatric admission (62% versus 48%; Fisher's exact test, p = .20), but those with documented evidence of imminent dangerousness were more likely to be psychiatrically admitted (65% versus 48%; Fisher's exact test, p = .012). Persons with documented behavior that could result in arrest were significantly more likely to be psychiatrically admitted (68% versus 55%; Fisher's exact test, p = .025) and significantly less likely to be discharged from the emergency room (29% versus 43%; Fisher's exact test, p = .016). Of the cases that resulted in an involuntary psychiatric admission, 68 percent involved behavior documented on the EP that could have resulted in an arrest (Fisher's exact test, p < .001).

Cases with evidence of clinical dangerousness based on a review of the complete emergency room record were significantly associated with a higher rate of admission (82% versus 9%; Fisher's exact test, p < .001). The reviewer agreed with the disposition in 91 percent of the cases with sufficient documentation to make that determination. Evaluees with a substance‐related diagnosis in the emergency room were significantly less likely to be admitted psychiatrically (52% versus 67%; Fisher's exact test, p = .011). Evaluees who were not evaluated by psychiatric residents were also significantly less likely to be admitted psychiatrically (16% versus 84%; Fisher's exact test, p < .001).

Discussion

These data support the EP process in Maryland as an effective means of delivering psychiatric services to individuals who are unwilling or unable to seek them. Sixty‐three percent of evaluees presenting to the Johns Hopkins Hospital emergency room during the study period were psychiatrically admitted. Sixty‐five percent of the patients discharged from the emergency room (65%) were referred for outpatient psychiatric treatment. Less than two percent of the cohort was released into police custody. Only 34 percent of the psychiatric admissions that resulted from an EP were involuntary in nature. For many patients, the emergency room visit itself provides an important opportunity to intervene when an individual's symptoms or behavior reach a crisis point. During the often lengthy evaluation process, patients may receive a medical evaluation, emergency psychiatric medications, supportive counseling, and referrals for further treatment. It is therefore not surprising that civil commitment, or even voluntary hospitalization, was avoidable in many cases.

Sixty‐eight percent of evaluees admitted involuntarily to the psychiatric hospital had emergency petitions that documented dangerous behavior toward others or property that could have resulted in arrest. These evaluees were more likely to be diagnosed with a major mental illness, more likely to have previous hospitalizations, more likely to be in treatment at the time of admission, and more likely to be noncompliant with medications. These data suggest that behavior that could be grounds for arrest is common in this cohort and, when present, is associated with a higher burden of mental illness.

Law enforcement officers initiated the majority (59%) of EP evaluations, but were least likely to describe behavior that could result in arrest. The police‐initiated petitions were more likely to describe threats to self, rather than threats to others or property. It is possible that these latter types of threats were underrepresented because the police were more likely to interpret such behavior as criminal in nature and instead of initiating an EP, to arrest the person, and take him or her to jail rather than to an emergency room. Many of these individuals, when less serious crimes are at issue, might be more appropriately and efficiently managed in a mental health care setting.

Although we did not specifically compare the accuracy of EP data to the data that we gleaned from the complete emergency room record, the information supplied on the petition for emergency evaluation was frequently incomplete. We observed this most often with police‐initiated EP forms, and least often with court‐endorsed petitions. At the time of the study, Maryland law required that when a law enforcement officer initiated an EP, it was because of evidence of a mental disorder and dangerousness that occurred in their presence. Unlike concerned family members requesting an emergency evaluation from a judge in an ex parte hearing, potential evaluees who are experiencing a crisis in the presence of a police officer may be unwilling or unable to provide historical information helpful to the petitioner. However, when data were missing, a field was usually left blank, suggesting that an inquiry into that subject was not made. We were most concerned that the question about access to weapons was left blank on two‐thirds of the forms completed by law enforcement officers.

Although not promoted as a jail diversion program, the EP process in practice diverts citizens who police officers, community members, and health care providers believe may be mentally ill and dangerous to the person and property of others from the criminal justice system into the mental health system. Diversion occurs before any criminal court involvement and therefore can best be described as a prebooking diversion process. Emergency rooms functioning as EP receiving sites meet many of the basic criteria for pretrial diversion programs described by Steadman et al.,⁷ including being identifiable, having a central drop‐off, having no‐refusal policies, being linked to community programs either through inpatient admission or discharge planning, and having a legal foundation for police referrals.

At the time of the study by Jayaram et al.¹⁵ and the current study, the law in Maryland required that a petitioner “[have] reason to believe that the individual has a mental disorder and that there is clear and imminent danger of the individual's doing bodily harm to the individual or another.''¹¹ In October 2003, the Maryland legislature altered the emergency room statute so that it no longer required that a potential evaluee's dangerousness be “imminent.''¹⁷ Patients' rights groups such as NAMI‐Maryland had lobbied for a change in the language of the statute because of a concern that the requirement of “imminent” dangerousness might prevent some families from obtaining appropriate and necessary interventions.¹⁸ Although we found that the documentation of “dangerousness” alone on the EP had no significant effect on psychiatric admission, over half (52%) of the evaluees who were “dangerous” but not imminently so were assessed as clinically dangerous when the entire emergency room record was considered. This suggests that the change in the language of the statute was appropriate. Whether or not the petitioner is able to articulate the “imminence” of the danger should not necessarily be a barrier to a complete evaluation by a mental health professional.

There were several limitations to this study. The data were collected through a retrospective chart review. Both reviewers were active clinicians at the Johns Hopkins Hospital during the study period and may have had preexisting knowledge about some of the evaluees. This study may not be generalizable to states with dissimilar procedures for emergency evaluations, or to hospitals with a different referral base. The reviewers assessed different sets of information, and inter‐rater reliability was not formally evaluated. Each reviewer assessed the documentation of the presence of mental illness and level of dangerousness based on clinical judgment alone and not on a specific instrument or protocol.

Based on the results of this study, we have specific recommendations for the emergency evaluation process that we hope will also be adopted by the legislature in Maryland and in other states that have similar provisions. First, we suggest that the need for complete information on the Petition for Emergency Evaluation be strongly reinforced. Emergency room physicians should insist that police officers who initiate an EP complete this form to the best of their ability before leaving the hospital. Although it is the duty of the evaluating physician to ensure that all pertinent data and collateral information are collected, to ensure a proper psychiatric assessment, evaluees facing even this brief intrusion into their liberty are entitled to a fair inquiry into the need for such an evaluation before they are forced to undergo psychiatric evaluation. In addition, missing information may place the safety of law enforcement officers at risk, particularly if they are unaware that the evaluee had access to a weapon. A streamlined EP form, making liberal use of checkboxes, may increase the likelihood that it would be appropriately completed without unduly delaying the officers.

We also suggest that there be a more specific inquiry into the presence and severity of substance‐related disorders. Substance abuse was markedly underreported in this sample of emergency petitions. Only 1.6 percent of the EPs documented the evaluee's history of a substance‐related disorder. However, after emergency room evaluation, substance abuse was found to be a common comorbidity and was the most frequently made clinical diagnosis overall (29%). Considering that in 2002 Baltimore ranked third nationally for emergency room visits for heroin, cocaine, and opiate containing pain relievers,¹⁹ as well as the serious potential for medical and psychological complications of substance‐related comorbidities, the importance of relating this information to the emergency room staff cannot be understated. Our findings also suggest that receiving facilities should have the ability to divert evaluees into substance abuse treatment facilities based on the patient's needs.

Finally, we believe that the results of this study identify an opportunity to improve the EP process by increasing the involvement of mental health professionals at all stages. Considerable recent interest has been generated in the provision of mental health training to law enforcement officers. In a survey of 150 police officers asked about the importance of mental‐health training, 89.6 percent rated “dangerousness” as a “very important” topic.²⁰ Our findings support the need for training in this area. Qualified mental health professionals could provide training to police officers about the emergency petition process, identification of psychiatric emergencies, and dangerousness. While it would be inappropriate to expect a police officer to perform a skilled psychiatric assessment of every citizen suspected to have a mental illness, awareness of common psychiatric emergencies may improve their ability to relay critical information to health care providers, to divert mentally ill persons from incarceration when appropriate, and to reduce the risk of harm to themselves and the general public. The involvement of a trained mental health professional before arrest, either as a consultant or as part of a crisis response team in the field, may help to improve the identification of mentally ill individuals before booking and to facilitate emergency psychiatric evaluation and treatment as an alternative to incarceration.²¹