The International Journal of Psychosocial Rehabilitation
The Influence of Collateral Informants on Psychiatric Emergency Service Disposition Decisions and Access to Inpatient Psychiatric Care

Alisa Lincoln MPH, PhD
Social and Behavioral Sciences Departement, Boston University School of Public Health, Social and Behavioral Sciences Department and the Division of Psychiatry, Boston University School of Medicine, Psychiatry Department
At the time of this work, Dr. Lincoln was a Pre-doctoral Fellow in the
Psychiatric Epidemiology Training Program at Columbia University

Michael H. Allen, MD
Colorado Psychiatric Hospital and the University of Colorado School of Medicine
At the time of this work, Dr. Allen was Director of the Comprehensive Psychiatric Emergency Program
at Bellevue Hospital Center in New York.

Lincoln, A.L., Allen, M.  (2002)   The influence of collateral information on access to inpatient
psychiatric services.  International Journal of Psychosocial Rehabilitation. 6, 99-108

* This research was supported by N.I.M.H training grant #5T32MH13043 and the Sociomedical Sciences Division of the Columbia University School of Public Health. Address correspondence to: Alisa Lincoln, Ph.D., Social and Behavioral Sciences Department, Boston University School of Public Health, 715 Albany Street, TW2, Boston, MA 02118.
Objective: Psychiatric Emergency Services (PES) are the entry point to the mental health system for many and the presentations frequently involve coercion. In these cases, the societal value of access to appropriate care may conflict with the value attached to confidentiality and procedural justice. Many influences on these decisions have been studied previously but financial pressures affecting access and public policy changes affecting patientsí rights have significantly altered the context in which PES assessment and disposition decisions occur This necessitates a re-examination of the factors which influence both voluntary and involuntary hospitalization. This paper addresses the important role of collateral sources of information on the admission process. Methods: A retrospective chart review of 287 people presenting at the Bellevue Comprehensive Psychiatric Emergency Program was conducted. Medico-legal and socio-demographic data were collected on each visit and multiple outcomes, primarily involuntary hospitalization and hospitalization of any type were examined. Step-wise logistic regression models were used to test the importance of multiple influences on these outcomes. Results: The presence of additional sources of information provided by collaterals increases the likelihood of both involuntary and voluntary hospitalization even when diagnosis, level of dangerousness and other socio-demographic variables are controlled. Conclusion: Additional information appears to facilitate access to inpatient care. In the current climate within which psychiatrists must make admission decisions, information from collateral sources takes on increased importance.

The Influence of Collateral Informants on Psychiatric Emergency Service Disposition Decisions and Access to Inpatient Psychiatric Care

Access to inpatient psychiatric care, and particularly the role of involuntary hospitalization in accessing care, is a topic of new importance. Historically the study of psychiatric emergency service (PES) decision making has focused on the array of medical/ legal and social factors that influence involuntary hospitalization. This approach has often assumed that involuntary hospitalization is undesirable and searched for inappropriate influences, such as race, on the involuntary hospitalization process. However, recent changes related to reimbursement have created a situation in which inpatient hospitalization, even through civil commitment, may be a desired resource for some people, their families and communities. If this is the case, the influences on PES decision making must be re-examined using new models.

Medical and legal characteristics such as diagnosis, symptoms and a personís potential for dangerous behavior (1,2,3,4) as well as social factors such as race (2,3,5,6,7), gender (2,3,8), and source to the emergency room (2,3,5,9) have all been examined previously for their impact on PES decisions. As expected, diagnosis, symptoms and level of dangerousness all predict involuntary hospitalization (1,2,3,4).

Social factors have been examined primarily from within a social control framework. According to this model, people who are marginalized and powerless will be more likely to receive undesired and coercive treatment (10). Support for the social control perspective has been found in the disproportionately high rate of involuntary commitment among black patients or patients brought to the PES by police officers (5,11).

However, the current setting in which involuntary hospitalization occurs is dramatically different from that of a decade ago. The numbers of inpatient beds have been greatly restricted while the number of visits to PESs has increased (12, 13). Reforms in civil commitment law beginning with the Lanterman Petris Short Act in California [Lanterman-Petris-Short Act (Cal Welf & Inst Code, div 5 pt 1) have improved procedural justice in the area of involuntary hospitalization by requiring evidence of mental illness and danger. Some have argued that the dangerousness standard permits patients to "die with their rights". Some states allow civil commitment on the basis of grave disability but even the more paternalistic grave disability standard may link disability to evidence of associated danger. Thus, in New York State, an individual in the same mental state may be detained in the winter due to the risk of freezing but released in the summer. Management of hospital utilization has simultaneously resulted in near universal application of a similar dangerousness standard for reimbursement. Under these conditions, clinicians must show evidence that most hospitalized patients satisfy involuntary hospitalization criteria to withstand utilization review if not judicial review. Evidence concerning the presence of mental illness and dangerousness have hence become the sine qua non of hospitalization.

Attention has recently broadened to a number of other factors influencing the quality of PES assessment such as staffing (15) and interrater reliability (16). Collateral information is another factor that has an important bearing on outcome

Collateral sources of information include family members, other medical providers, police officers, friends or prior records. Collateral informants may supply objective data and improve assessment or advocate a particular outcome and add to bias.

Developing collateral information also requires additional time and effort of patients, staff and others; training including cultural competency; and resources such as phones and computers. Consumers may view the additional time as delay and contacting third parties as intrusive. On the other hand, if associated with a poor outcome, failure to take the necessary time and gather additional data may be viewed by some as malpractice.

In this study, we report on the influence of sources of information other than the patient on the decision to admit patients to the hospital from the Bellevue Comprehensive Psychiatric Emergency Program (CPEP).

The study was conducted in one of the busiest urban, public, teaching hospitals in the country. The CPEP receives over 6000 visits per year and, at that time, served as the admission service for 350 inpatient beds. These beds were at 100%+ capacity throughout the entire course of this study. Cases were identified from the CPEP logbook from January, March and June of 1995. There were a total of 1546 visits during these months. Stratified sampling was used in order to guarantee adequate data on police involved cases. Every fifth non-police case (N=228) and every police case (N=315) was included in the sample. This included both Emotionally Disturbed Persons (EDPs) for whom no criminal charges were pending and pre-arraignment cases. EDPs are said to be "aided" by the police in presenting to the hospital and are not charged with a crime. Prearraignment cases are in the process of arrest and will be formally charged with a crime. Prearraignment cases have been excluded from this analysis as the assessment in these cases is limited in scope in an effort to avoid delays in the due process protections afforded by arraignment in the criminal justice system. Such individuals may return to a hospital subsequently either voluntarily or at the direction of the arraignment judge for a more typical examination. Therefore the process of evaluation and disposition of these cases is not similar to those of the rest of the sample. Medical records were available for 362 of the sampled visits. This represented 67% of the non-police involved cases and 66% of the police involved visits. This provides evidence that Ddifferential loss to follow-up related to police involvement did not bias these results. The possibility that other characteristics related to missing charts (ie. admitted vs. not admitted) could not be ruled out.

For each sampled visit the patientís socio-demographic characteristics and medical and legal characteristics were collected from the medical records. Dummy variables for diagnosis were: affective disorders including bipolar disorder, major depression or depressive episode and dysthymia; psychotic disorders including schizophrenia, psychosis nos, brief reactive psychosis, and schizoaffective disorder; substance use disorders; personality disorders; and anxiety disorders. In addition variables were created reflecting the comorbidity among diagnoses. The inclusion of the comorbidity variables in the logistic regression models did not change the results so in order to create parsimonious models only the individual diagnostic variables were included in the final models.

The Ďlevel of dangerousnessí variable, was created from data collected on the precipitating events from the patientís chart. Four levels of dangerousness were created based on the personís most acute level of dangerousness. A person was coded as Ďdangerousí if they had attacked someone or something with or without a weapon or made a suicide attempt. A person was coded as Ďpotentially dangerousí if they were described as being potentially dangerous to themselves or others. The third level of dangerousness included people with noted suicidal ideation and the final level of dangerousness includes people with bizarre behavior who do not fall into any of the above categories. Multiple informant reports of the precipitating events were used whenever available to try to address potential bias in the recording of the precipitating events in the chart.

To check coding reliability of the descriptions of the precipitating events, a random sub-sample of forty cases were coded by three independent coders. All gender and race language as well as any references to police involvement were removed. Cohenís Kappas for the various combinations of the three raters ranged from .34 for bizarre behavior which was a very subjective rating to 1 for suicide attempt.

Data on patient race/ethnicity, sex, age, insurance status and education level were obtained from the CPEP visit sheet. Unfortunately education and insurance data were frequently missing and could not be used in these analyses. Employment data were assessed from the CPEP visit sheet as well as the social work report when available. Homelessness was determined from physician notes in the CPEP visit sheet and variables were created for where the person lived (apartment/house, shelter, SRO, street etc) and with whom (parents, spouse/partner, alone, other family member). This served as a cross-check of the personís actual living situation.

Data on individuals who arrived at the CPEP with the identified patient were collected from the medical record, and confirmed with police reports. Data on the sources of information about the patient available to the clinician in the CPEP were collected from the medical chart and the police reports. Possible dispositions for cases at the CPEP include emergency hold, involuntary commitment, voluntary admissions, transfer to other hospitals, transfer to state hospital or long term care, and treated and released or discharged.

The stratified sampling used in this study necessitates the use of SUDAAN in the final analyses. SUDAAN uses Taylor Series methods to calculate appropriate standard errors when such complex sampling designs are used. Throughout the paper all frequencies are unweighted (N) and all proportions have been computed using the weighted data.

Table 1 contains data describing the sample characteristics. The sample is predominantly male (77%) and the mean age is 38. Thirty-two percent (N=96) of the sample is white, 43% Black (N=115), 20% Hispanic (N=54) and 5% of other ethnic origin (N=22). In addition the hospitalization outcomes were found to be as follows: Forty-six percent of the sample was hospitalized: 10% percent voluntarily, 36% involuntarily. Police reports were present for 12.5% (N=105) of the sample. In 13% (N=31) of the visits prior records were available, family members provided information in 22% (N=74) of the visits, a medical professional outside of the CPEP provided information for 30% (68) of the cases, 6% (n=14) of the sample was noted to be known to the staff, and for 8 cases friends provided information. Nine percent of the sample was ultimately transferred to long-term care, most frequently the state hospital.
Table 1: Frequency Table for the Medico-legal and Socio-demographic Variables 
  Weighted Proportion (%) Unweighted N
Psychotic Diagnosis 53.6 156
Affective diagnosis 34 94
Substance abuse disorder 55.2 155
Personality disorder 4.4 12
Anxiety disorder 5.3 12
Speech irregularity 21.9 73
Suicidal/homicidal ideation 30.4 83
Thought content irregularity 45.5 134
Thought process irregularity 39.8 126
Attacks on people or objects 5.2 31
Threats of attacks on people or objects 17.1 60
Potentially threatening behavior to others or self 24.1 83
Suicidal ideation 29.4 68
Suicide attempt 5.9 22
Bizarre behavior 42.4 122
Seeking social services 7.6 17
Seeking medication 29.8 60
Sex Male






Race White
Other race/ethnicity
Living arrangement
No family involvement 61.7 117
Living alone 64.7 174
Unemployed 77.4 217
Source to the CPEP-Alone 

Initial univariate analyses were conducted on all source of information variables and several were significant univariate predictors of involuntary hospitalization and total hospitalization, ie, either voluntary or involuntary. A police report, information provided by a family member and information provided by a medical professional outside of CPEP all significantly predicted hospitalization outcomes. The source of information variables were then collapsed, due to small numbers in some cells, into one variable indicating the availability of any outside source of information to the clinician in the CPEP.

Step-wise logistic regression models were developed for the role of any outside source of information on the two primary outcomes: involuntary hospitalization (Model 1) and total hospitalization (Model 2). In Model 1, the presence of any outside source of information significantly predicts involuntary hospitalization with diagnoses, age, level of dangerousness, sex, source to CPEP and race controlled. People for whom collateral information was available were 2.9 times more likely to be involuntarily hospitalized than those for whom no collateral information was available. In addition in Model 2 a similar relationship is seen. Controlling for the same medical and social variables, people for whom collateral information was available were 2.5 times more likely to be hospitalized than people for whom no collateral information was available. Parallel analyses were conducted among the subsample of cases without police involvement and the same relationships were found.
Table 2: Univariate Analyses of Source of Information and Hospitalization 
  Frequency % (N) Odds Ratio (95% C.I.)
Source of Information      
Police Report 12.5 (105) 1.98*  (1.07, 3.61)
Medical Chart 13.1 (31) 1.54 (.68, 3.50)
Family Member 22.2 (74) 2.06* (1.04, 4.22)
Friend  2.1 (8) .38 (.04, 3.39)
Known to Staff 6.3 (14) .70 (.22, 2.34)
Medical Professional 30 (68) 2.94** (1.55, 5.61)

Table 3: Final Logistic Regression Models on Involuntary Hospitalization and Hospitalization

Exp B (Odds Ratio)


Exp B (odds ratio)

  Involuntarily Hospitalized Hospitalized
Informed 2.891** 2.469 *
Danger 1 (dangerous) 3.471 2.961 *
Danger 2 (potentially dangerous to self or other) 7.865 4.336 *
Danger 3 (suicidal ideation) 1.069 2.226 *
Danger 4 (bizarre behavior) .047 * * .655
Age 1.052 1.042 **
Dxad (anxiety) 1.448 .816
Dxaff 9.119 ** 10.976**
Dxpd (personality) .957 .4
Dxpsych 23.927 ** 8.843 **
Dxsubs 4.425 ** 1.880
Sex (1=male) .133 ** .647
Source= Alone    
Source1 (family) 7.195 ** 2.090
Source2 (police) 3.163 5.233 *
Source3 (transfer from other hosp) 2.556 3.262
Source4 (other) 185 ** 19.478 **
Black .349 * 1.081
Hispanic 2.965 * 3.847 **
Other race/eth 4.055 5.004
* p<.10    ** p<.05

Other significant predictors of both involuntary hospitalization and total hospitalization include a diagnosis of an affective, psychotic or substance use disorder. Further analyses conducted on the influence of diagnoses found that substance abuse not comorbid with psychotic disorder did not significantly predict hospitalization by any mechanism but in fact decreased the likelihood of inpatient admission (15).

Regarding race and gender, Hispanic patients were more likely than white patients to be admitted. Black patients were a third as likely as white patients to be admitted involuntarily but there was no overall admission difference. Finally female patients were more likely than male patients to be involuntarily hospitalized.

It is fairly common for family or community providers to arrange for a patient to go to an emergency room with the expectation that the patient will be admitted to hospital only to have the patient return to them often with no direct communication from the ER. This is a source of great concern and frustration to the community.

This results of this study seems to suggest that, as admission decisions are increasingly constrained by formal legal and utilization criteria, data from multiple sources improve the chances of clearing the legal and financial barriers to hospital care. If that is the desired outcome, then communication by concerned parties with responsible PES personnel is advisable. This is understood among physicians and is the subject of continuity of care agreements, EMTALA regulations, etc. It may be less well understood by nonmedical providers, police, families and advocates. While confidentiality may limit the extent to which hospital personnel may contact others regarding the patient, those limits do not apply to individuals contacting the hospital to provide data. The credibility of informants and the potential for abuse must be considered in weighing the resulting data but in civil commitment decisions, such reports may be legitimately considered and need not be treated as "hearsay" as they might in criminal proceedings. This study can also be viewed as lending support to the practice of providing a structured means of qualifying informants and communicating relevant data such as a form that may then be entered into the medical record. This is a common practice with police informants.

Segal et al (15) have pointed out this association of informants with admission may reflect bias in favor of third party advocates for admission. However, those authors reported that data from advocates contributed to the assessment of dangerousness and hence facilitated application of civil commitment criteria rather than social bias. In that study, conducted under California law, psychosis and dangerousness were the main contributors to coercive retention modified by availability of alternatives and information from other sources. Our findings are consistent with those of Segal et al. In the authors experience, PES clinicians are receptive to data but less so to opinions or directions. Data may include evidence of the failure of alternatives to hospitalization including the inability of community providers to render care under the circumstances. Alternatively, sending the patient to the ER with a request for admission but little data to support that recommendation may be viewed as "dumping".

Our findings with regard to diagnosis are also consistent with those previously reported. In a different sample in the same facility, one of the authors has reported that cocaine use was inversely correlated with admission (18).

Our finding of a strong relationship between admission and Hispanic origin is interesting. More work is needed to better understand this relationship which has also been reported by others (5). It may be that family were more involved and influential in Hispanic cases. Specific attention needs to be focused on the role of language and culturally specific understandings of family and family involvement in help seeking and admission decision making. Additionally the influence of black race on involuntary admissions must be further explored since our study clearly does not find black patients more likely than white patients to be involuntarily hospitalized but does suggest equal access to hospital care overall. The impact of race and ethnicity will be the subject of a subsequent report.

This study has the limitations inherent in retrospective reviews of routine data collection instruments. However, the design of this study avoids the observer effect that may have been present in the prospective studies.y by Segal et al.

This study was not able to ascertain the reliability or validity of assessments and decisions. We did not attempt to determine whether the disposition was "correct" in light of subsequent examinations or outcomes such as future contacts with ERs, criminal justice system or, in the case of suicides, the coroner or medical examiner. As suggested by the low level of agreement about factors such as bizarre behavior in this study and impulse control in a related study, there is a need to improve the reliability of key constructs in PES decision making(16).

As hospitalization rates decline, Lamb and Weinberger have reviewed evidence suggesting that a stable number of individuals will be confined in either hospitals or jails and prisons and that the trend in the US has been toward the latter, incarceration of the mentally ill in jails and prisons (14). Hence the question may not be if but where and under what conditions incarceration will occur: deprivation of liberty with or without deprivation of care. We must continue to weigh the importance of respecting confidentiality, third-party bias and potential abuse with this question. Increasing the information available to decision makers concerning an individuals behavior in the community may contribute to improving these critical decisions.

1. McNiel DE, Myers RS, Zeiner HK et al:The Role of Violence in Decisions About Hospitalization from the Psychiatric Emergency Room. American Journal of Psychiatry 149:2, 207-212, 1992.

2. Gerson S, Bassuk E: Psychiatric Emergencies: An Overview. American Journal of Psychiatry 137:1-11, 1980.

3. Marson DC, McGovern MP, Pomp HC: Psychiatric Decision Making in the Emergency Room: a Research Overview. American Journal of Psychiatry 145, 918- 925, 1988.

4. Rabinowitz J, Massad A, Fennig S: Factors Influencing Disposition Decisions for Patients Seen in a Psychiatric Emergency Service. Psychiatric Services 46:7, 712-18, 1995.

5. Rosenfield S: Race Differences in Involuntary Hospitalization:Psychiatric v. Labeling Perspectives. Journal of Health and Social Behavior 25,14-23, 1984.

6. Lindsey K, Paul G: Involuntary Commitments to Public Mental Institutions: Issues Involving the Overrepresentation of Blacks and Assessment of Relevant Functioning. Psychological Bulletin 106: 171-183, 1989.

7. Strakowski SM, Lonczak HS, Sax KW et al. The Effect of Race on Diagnosis and Disposition From a Psychiatric Emergency Service. Journal of Clinical Psychiatry 56:3, 101-107, 1995.

8. Rosenfield S. Sex Roles and Societal Reactions to Mental Illness: The Labeling of "Deviant" Deviance. Journal of Health and Social Behavior, 23, 18-24, 1982.

9. Way B, Evans M, Banks S. Factors Predicting Referral to Inpatient or Outpatient Treatment from Psychiatric Emergency Services. Hospital and Community Psychiatry 43:7, 703-708, 1993.

10. Horwitz A: The Logic of Social Control. New York: Plenum Press, 1990.

11. Reinish LW, Ciccone R: Involuntary Hospitalization and Police Referrals to a Psychiatric Emergency Department. Bulletin of the American Academy of Psychiatry and Law 23:2, 289-298, 1995.

12. Fitzgerald M: Structuring Psychiatric Emergency Services for Smaller Communities in Response to Managed Care. Psychiatric Services 47, 233-234, 1996.

13. Cohen, N. L. and L. R. Marcos. Law, Policy, and Involuntary Emergency Room Visits. Psychiatric Quarterly 61: 197-204, 1990.

14. Lamb HR, Weinberger LE: Persons with severe mental illness in jails and prisons: A review. Psychiatric Services 49:483-492, 1998.

15. Segal SP, Laurie TA, Segal MJ: Factors in the Use of Coercive Retention in Civil Commitment Evaluations in Psychiatric Emergency Services. Psychiatric Services 52: 514-520, 2001.

16. Way BB, Allen MH, Mumpower JL, Stewart TR and Banks SM. Interrater agreement among psychiatrists in psychiatric emergency assessments. American Journal of Psychiatry 155:1423-1428, 1998.

17. Breslow RE, Klinger BI, Erickson BJ: Acute Intoxication and Substance Abuse Among Patients Presenting to a Psychiatric Emergency Service. General Hospital Psychiatry 18:3, 183-191, 1996.

18. Currier GW, Allen MH. Cocaine toxicology testing at psychiatric emergency department presentation. American Journal of Forensic Psychiatry 19(3):55-65, 1998.

Copyright © 2002, Hampstead Psychological Associates, ltd.  All Rights Reserved.
A Private Non-Profit Agency for the good of all, published in the UK & Honduras