a b s t r a c t

Objective: The objective was to examine the relationship between violence screening items, suicide risk stratifi- cation, and disposition in a psychiatric emergency service setting.

Methods: A retrospective review of electronic health record data for 286 patient encounters was performed. Results: Four of the 6 violence risk screening items were significantly associated with both involuntary presenta- tion to the psychiatric emergency service and high-risk stratification. These 4 items were also associated with psychiatric hospital disposition in bivariate analysis, however, only indirectly through their association with high-risk stratification, which in turn was directly associated with psychiatric hospital disposition.

Conclusion: Violence screening items inform disposition but only through the use of risk stratification, supporting the need for additional research into the predictive value of standardized suicide risk stratification definitions to inform clinical practice.

(C) 2015

Introduction

Suicide and nonfatal intentional self-directed injuries are serious problems worldwide. The World Health Organization estimates that more than 800,000 people die each year from suicide [1]. According to the most recent data available from the Centers for Disease Control and Prevention, suicide is the 10th leading cause of death for American adults with N 41,000 fatalities attributed to self-inflicted vio- lence reported in 2013 [2]. Suicide is the second leading cause of death for Americans between the ages of 15 and 34, second only to un- intentional injury. The number of suicides per 100,000 people increased from 14.1 in 2003 to 16.4 in 2013, despite greater national and local at- tention on the problem of self-directed violence during this decade. The numbers of nonfatal intentional self-inflicted injuries are far greater, with an estimated N 494,000 such injuries occurring in the United States in 2013 [2].

The global numbers of suicides and intentional self-inflicted injuries have led to national and international efforts to identify and treat indi-

? This work was funded by The University of Texas System Patient Safety Committee Research Grant Award #150271.

?? Disclosures: The authors do not have any relevant disclosures to make for this manu-

script submission.

* Corresponding author at: Department of Psychiatry, University of Texas Southwestern Medical Center, 5323 Harry Hines Blvd, Dallas, TX 75390-8898. Tel.: +1 214 648 8726;

fax: +1 214 648 9627.

E-mail address: [email protected] (K. Roaten).

viduals at risk for suicide, beginning with a call from the World Health Organization in 1996 to address this problem [3]. In the next few years, the US Surgeon General and the US Department of Health and Human Services responded with a Call to Action [4] and a National Strat- egy for suicide prevention [5], the latter of which was recently revised and expanded [6]. In the National Patient Safety Goals for Behavioral Health Care and Hospitals Effective January 1, 2015, The Joint Commission specifically stated that health care systems must identify patients at risk for suicide [7]. Despite these initiatives, neither national nor interna- tional suicide rates have declined. It has been determined that available evidence for recommending suicide risk Screening procedures for pri- mary care providers–the gatekeepers for the majority of mental health care services in the United States–is insufficient [8-11], but existing models of suicide risk assessment and the tools needed for primary care providers to address the problem lack prospective data demon- strating adequate reliability and validity [12].

Studies in Primary care settings have demonstrated that the major- ity of people who die by suicide use some form of health care services before the time of their death, including nearly one-half in the preced- ing month; however, few had contact with mental health providers in this time frame [13,14]. Studies in the United Kingdom have found that nearly 40% of patients who died by suicide visited an emergency department (ED) in the year before their death, although seldom for sui- cidal behavior [15,16]. A research study of computer-based screenings for suicidal ideation among patients presenting to a large, US urban ED for nonpsychiatric complaints found that 11% screened positive for ac- knowledging current suicidal ideation and 2% reported suicidal intent

http://dx.doi.org/10.1016/j.ajem.2015.12.014

0735-6757/(C) 2015

with plan; however, retrospective medical record reviews of these en- counters revealed that the providers (who did not see the screening re- sults) did not detect self-directed harm issues in 80% of the encounters with positive screens [17].

Estimation of suicide risk in the ED setting is a particularly daunting endeavor. In the challenging setting of today’s ED, patients may feel frustrated with wait times, overwhelmed by comorbid medical and psy- chiatric issues, and uneasy about the prospect of disclosing sensitive in- formation to providers they do not know. Providers face time constraints, high patient volumes, and limited access to collateral infor- mation. Even experienced psychiatric providers acknowledge that available knowledge does not sufficiently inform efforts to determine potential for suicide risk in the ED setting. Despite these seemingly in- surmountable difficulties, the ED is a logical setting for developing and testing evidence-based systems of suicide risk stratification for more ef- fective detection of patients at imminent risk for suicide [18,19]. The current study used data from retrospective review of encounters in an electronic health record to improve suicide risk estimation and related treatment planning for patients presenting to the psychiatric emergency service (PES) at a large, urban, safety-net hospital with an extraordinarily socioeconomically and ethnically diverse patient popu- lation. Specifically, the study was designed to explore the process of clinical suicide risk assessment and its relationship with patient disposition.

Methods

Violence risk screening and documentation of the findings in the EHR are mandatory for every patient PES encounter at Parkland Memo- rial Hospital. The violence risk screening consists of 6 questions requir- ing a yes or no response from the providers (Table 1). These items were originally developed by a group of experienced clinicians in Leadership roles for the Parkland PES based on review of relevant literature on acute risk factors for violence directed to self or others and clinical expe- rience with the Parkland PES population. Four of the items pertained to self-directed (suicidal) violence only, one to other-directed violence only, and one to both self- and other-directed violence.

For this study, 3 categories of suicide risk were established to repre- sent low, moderate, and high risk for rating of patient encounters. Def- initions of low-, moderate-, and high-risk encounters (Table 2) were established by group discussion and consensus of 6 PES psychiatrists, and additional suggestions were informally solicited from other PES providers until no further recommendations for adjustment of the def- initions were offered.

The institutional review boards of the participating institutions ap- proved the study in advance. Over a period of 18 months (from February 2013 to August 2014), a weekly list of all Parkland PES encounters was used to select 5 encounter EHRs per week with unduplicated service providers (approximately 30 providers, 90% physicians and 10% nurse practitioners, at any given point) in the encounters selected during any given month, yielding a total of 286 encounters for review. From these clinical records, data were gathered on the providers’ responses to the 6 violence risk screening items and the providers’ ratings of sui- cide risk stratification level for the encounters. An auditor rated the

Table 1

Risk screening constructs and provider screening questions

Risk construct Risk screening question

Want to harm self Self-report of wish to inflict harm to self? Want to harm others Self-report of wish to inflict harm to others? Collateral self/other danger Collateral report of danger to self or others? Prominent death wish Report of prominent wish to no longer live?^a Significant loss Evidence of acute/significant psychologically/

psychiatrically destabilizing loss?^a Hospitalized for risk last week Hospitalization for risk of harm in past week?

^a Includes both patient self-report and/or collateral information.

Table 2

Definitions of low-, moderate-, and high-risk encounters Characteristics of low-risk encounters

Patient provides reassurance of his/her safety

No evidence of intent to engage in self- or other-directed violence

Credible evidence of secondary gain (eg, seeking medication, place to stay, escape from weather)

Credible evidence of primary gain (eg, seeking attention, validation, contact with providers)

Evidence of >= 2 protective factors Expression of future orientation

Solid social/family support evident, with remission of suicidal ideation during encounter No evidence of self-directed violence, with remission of suicidal ideation once sober Collateral suicide risk information precipitating PES encounter is determined to be

unsubstantiated or invalid

Characteristics of moderate-risk encounters

Self-directed violence while intoxicated Suicidal ideation with adequate social support

Suicidal ideation with motivation for appropriate mental health treatment Suicidal plan with credible intent to participate in higher level of mental health

treatment

Psychosis or paranoia without suicidal or homicidal ideation Characteristics of high-risk encounters

Suicidal ideation with highly lethal method and access to identified method

(eg, firearm)

Evidence of suicidal plan with intent

Suicidal ideation with report or endorsement of hopelessness Suicidal ideation with lack of future orientation

Suicidal ideation with b2 protective factors

Acute psychosocial stressor that prompted decompensation Suicide attempt with low likelihood of rescue

Highly lethal suicide attempt (with or without intoxication)

Patient report of intent to die during self-directed violent incident leading to PES encounter

Collateral information provides evidence of high risk

patient reports regret for surviving suicide attempt leading to PES encounter Evidence of preparation for death (eg, letters, giving things away, checking

insurance, preparing will)

encounters blinded to the provider ratings and recorded independent responses to the 6 violence risk screening items and a determination of suicide risk stratification level based on material available in the en- counter record. The auditor (KR) is a clinical psychologist who has more than 5 years of experience with suicide risk assessment and relat- ed research and quality improvement activities. The auditor participat- ed in the formulation of the 6 screening items and the risk stratification categories. The auditor also rated each encounter (yes/no) as to wheth- er the provider used additional elements of the patient presentation and history beyond the 6 required violence screening items in arriving at the suicide risk determination, including acute vs chronic risk estimation, a formal narrative description of presenting issues and risk estimation, and additional elements of the risk assessment (eg, history of Depressive disorder). The auditor ratings for each encounter were entered into the database for comparison with the provider data on the same encounters.

Interrater reliability was assessed with ? statistics, substituting the Yule Y statistic in instances of base prevalence b 20% to address potential for confounding of low base rate with sensitivity and specificity in the calculation of Interrater agreement [20]. The level of interrater reliabili- ty between the auditor and the provider on the 6 violence screening items was excellent, ranging between .84 and .91 (Table 4). To achieve interrater reliability on suicide risk stratification, 2 auditors indepen- dently rated encounter EHRs using the definitions of suicide risk strati- fication developed for this study, with achievement of excellent interrater reliability: ? = .95 for low risk, ? = .80 for moderate risk, ? = .91 for high risk, and overall ? = .84. Once this interrater reliability was obtained for suicide risk stratification, provider ratings were com- pared with independent ratings of 1 auditor on this variable.

The auditor’s and providers’ ratings of the 6 screening questions and level of risk for the encounters were entered into a Microsoft Access da- tabase and then imported into SAS 9.3 (SAS Institute, Cary, NC) for anal- ysis. Descriptive data are presented as raw counts and proportions. Sensitivity and specificity of screening items to high-risk ratings by the auditor were calculated. Comparisons between dichotomous vari- ables were made using ?² tests. Multiple logistic regression models were used (PROC LOGISITIC in SAS) to predict psychiatric hospital dis- position (dependent variable) from 2 independent covariates (risk de- termination with low = 1, moderate = 2, and high = 3; and risk item response) entered simultaneously into the model.

Results

Overall, 286 unique patient encounters were included in the study. Table 3 provides summary statistics for the admission and disposition status of the encounters. For nearly three-fourths of the encounters, the patient had presented to the ED involuntarily. The most common disposition after evaluation by the PES was discharge to home. Approx- imately one-third of the patient encounters culminated in transfer to a psychiatric hospital for a higher level of care. A fraction of the encoun- ters resulted in release of the patient to a homeless shelter, and small numbers of dispositions were to psychiatric residential programs, sub- stance rehabilitation facilities, medical hospitals, or jail.

In 32% of the patient encounter records reviewed, the documenta- tion confirmed that collateral information was collected from the pa- tient’s friends, family, or other health care providers to complete the risk screening and plan of care. More than a quarter (26%) of the records included documentation from the provider specifically addressing acute vs chronic risk estimation for the identified encounter. A formal narra- tive description of the patient’s presenting issues and risk estimation was included in two-thirds (67%) of the records. It was further deter- mined that, in 43% of the records, providers completed additional ele- ments of the risk assessment beyond the 6 required screening items.

Table 4 provides ratings of screening items and acute risk stratifica- tion by auditors and providers. Auditor/provider agreement on re- sponses to the screening items ranged from near excellent to excellent. Four of the 6 auditor-rated screening items (want to harm self, want to harm others, collateral self/other danger, and death wish) were significantly associated with auditor determination of high-risk level. Auditor response to collateral self/other danger was highly sensi- tive for auditor high-risk determination, indicating that a “no” response on the item was likely to indicate a low-risk encounter. Auditor re- sponses to want to harm others, prominent death wish, and significant loss were very specific to high-risk auditor determination, indicating that a “yes” response on any of these items was a strong indicator of a high-risk encounter. Only 16%-17% of the encounters were determined to represent high risk, and auditor/provider high-risk agreement on this rating was excellent. Approximately half or more of the encounters were determined to represent low risk, also with excellent auditor/pro- vider agreement. Approximately one-fifth to one-third of the

encounters were determined to represent moderate risk, also with good auditor/provider agreement.

Table 5 provides proportions of encounters with involuntary admis- sion to ED and psychiatric hospital disposition in association with vio- lence screening item ratings and suicide risk stratification by the auditor. Four of the 6 screening items (want to harm self, want to harm others, collateral self/other danger, and death wish) were significantly as- sociated with both involuntary ED admission and psychiatric hospital disposition. The other 2 items (significant loss and hospitalized for risk in last week) were not associated with either involuntary admission or hospital disposition.

Most encounters with high-risk suicide stratification ratings had an involuntary ED admission status and were discharged to a psychiatric hospital (Table 5). More than three-quarters of encounters with moder- ate auditor risk ratings and more than one-half of encounters with a low-risk auditor rating also had an involuntary ED admission status. In- voluntary ED admission was equally prevalent among high- and moderate-risk encounters. Involuntary ED admission was significantly less prevalent among low-risk encounters compared with moderate- and high-risk encounters. Of 211 patient encounters with involuntary ED presentation status, 41% were rated in the encounter EHR as low risk (not shown in table).

Less than half of encounters with moderate auditor risk ratings were discharged to a psychiatric hospital. Very few encounters with a low- risk auditor rating were discharged to a psychiatric hospital. Psychiatric hospital dispositions were significantly more prevalent for high-risk en- counters compared with moderate-risk encounters, which in turn had a significantly higher prevalence of psychiatric hospital dispositions than did low-risk encounters. Of 5 low-risk encounters with psychiatric hos- pital disposition, one was discharged to a psychiatric hospital for rea- sons other than suicide risk (mania), 3 were reassessed later in the same encounters by different providers and were determined to be high risk, and 1 was discharged to a psychiatric hospital to satisfy payor requirements. Only 32% of those with involuntary presentation required psychiatric hospital disposition, although the proportion was significantly higher (?² = 4.98, df = 1, P = .026) than for those with voluntary presentation (19%).

Two multiple logistic regression models were used to predict psychi-

atric hospital disposition (independent variable) from positive ratings on risk items and risk stratification (Table 6). The first model used a high-specificity screening item variable created by combining re- sponses from 2 violence screening items (wish to harm others or death wish: either or both yes vs both no) and entered as a dependent covar- iate along with suicide risk stratification level. The second model used a high-sensitivity item variable reflecting a negative response to the vio- lence screening item referring to collateral self/other danger and was en- tered as a dependent covariate along with risk stratification level. In both of these models, suicide risk stratification level strongly predicted psychiatric hospital disposition independent of the violence screening item variables, but the violence screening item variables were not asso- ciated with disposition independent of suicide risk stratification level.

Table 3

ED admission status and disposition

ED admission status

% n/N

Discussion

This study of PES encounter records in a large urban safety-net hos- pital examined provider documentation of 6 violence risk screening items, explored the utility of this information in stratification of suicide

Disposition
Home or boarding home	48	137
Psychiatric hospital	29	82
Psychiatric residential program	5	13
Substance (alcohol/other drugs) rehabilitation program	3	8
Medical hospital	1	2
Homeless shelter	15	42
Jail	1	2

Involuntary 74 211

Voluntary 26 75

risk for patient encounters, and then compared suicide risk stratification determinations with final dispositions in the encounters. An important contribution of this study was the development and systematic imple- mentation of an operationalized definition of suicide risk stratification. Definitions of suicide risk stratification levels were operationalized through team members’ expert consensus, and excellent interrater reli- ability was established for both suicide risk stratification and rating of the 6 violence screening items. All 3 categories of the suicide risk strat- ification performed well in informing psychiatric inpatient disposition

Table 4

Responses to violence screening items and suicide risk ratings and by auditor and provider

Positive violence screening itemsa	Auditor n = 286	Provider n = 258b	Auditor/ provider interrater agreement: ? (Yule Yc)	Sensitivity/ specificity of auditor assessment of screening item to auditor high risk rating
	% (n/N)	% (n/N)	n = 258
Want to harm self	34 (95/279)	32 (83/263)	.85	.64, .72???
Want to harm others	18 (50/285)	15 (38/262)	.85 (.90)	.50, .89???
Collateral self/other danger	74 (212/285)	69 (173/251)	.85	.96, .30???
Prominent death wish	11 (31/277)	10 (23/242)	.86 (.91)	.48, .96???
Significant loss	19 (53/274)	14 (35/256)	.73 (.84)	.29, .82
Hospitalized for risk last week	4 (12/270)	4 (9/251)	.73 (.91)	.02, .95
Suicide risk stratification	% (n)			% (n)
High	17 (49)			16 (41)	.91 (.99)
Moderate	34 (96)			21 (55)	.65
Low	49 (141)			63 (162)	.76

^a Some additional auditor and provider ratings of screening items were missing.

^b Thirty-seven provider ratings of acute risk were missing.

^c Yule Y was substituted for ? calculation in instances of base rate b20%.

??? Associated with acute high-risk auditor rating (?² comparisons, P b .001).

status, as demonstrated by differentiation of both low- and high-risk categories from the moderate-risk category in predicting disposition.

Figure presents a model that illustrates the relationship between the violence risk screening and the suicide risk stratification determination within a temporal trajectory of the patient and clinician assessment/dis- position processes through the ED. It begins at the left of the figure with presentation of the patient to the ED, then proceeds through assessment (violence risk screening and suicide risk stratification), decision making, disposition, to the ultimate patient outcome at the right of the figure. The analyses conducted in this study refer to processes that occur along this trajectory and their relationships to one another, as repre- sented by the boxes connected by solid arrows in the figure.

The 6 violence screening items were analyzed for their ability to pre- dict suicide risk stratification and Patient dispositions. A positive re- sponse to 1 violence screening item (collateral self/other danger) had 96% sensitivity for high suicide risk stratification, indicating that if this screening item is negative, it may reassure the clinician that the encoun- ter is likely to be low risk. Positive responses on 3 violence screening items (want to harm others, prominent death wish, significant loss) had 82%-96% specificity for high suicide risk stratification. Violence screen- ing items with very high specificity can serve as clinical Red flags, with a positive indication alerting the clinician to further investigate poten- tial suicide risk and address it in the clinical disposition. Of note, 1 addi- tional violence screening item (hospitalization for risk in the last week)

Table 5

Involuntary admission to ED and psychiatric hospital disposition by ratings of violence screening items and suicide risk stratification by auditor

also had high specificity (95%) for suicide risk stratification in the cur- rent study, but the sensitivity was quite low (2%). These statistics are consistent with the evidence in the existing literature that the postdischarge period from a psychiatric hospital is a particularly high- risk time [18,21,22], yet they also demonstrate how rarely this critical item will be informative. Positive responses on 4 risk screening items (want to harm self, want to harm others, collateral self/other danger, and prominent death wish) significantly predicted psychiatric inpatient disposition.

Together, the bivariate and multiple regression analyses helped to clarify the interrelationships among the violence screening items, sui- cide risk stratification, and disposition. These analyses demonstrated that the risk screening items were associated with the risk stratification level and the risk stratification level in turn was associated with the final disposition, but the relationship between the violence risk screening items and the disposition was only indirect through the apparent effects of the risk screening items on the suicide risk stratification level. The multivariate models made it clear that it was suicide risk stratification that ultimately determined the disposition. In other words, even though 4 of the violence screening items were significantly associated with both high suicide risk stratification and hospital disposition, and high suicide risk stratification in turn was significantly associated with hospi- tal disposition, these 4 violence screening items did not contribute to prediction of disposition independently of the contribution of suicide risk stratification to prediction of disposition, which ultimately deter- mines the patient outcome.

The above analyses clarified that information from the violence screening informs the suicide risk stratification determination, but sui-

Involuntary admission to ED

% (n)

Psychiatric hospital disposition

% (n)

cide risk stratification reflects more than just this information. The pro- viders often went beyond the 6 mandatory violence screening items to use other information from various sources to substantiate their overall

Violence screening items: Positive

item

Want to harm self	93 (88)	64 (117)???		38 (36)	24 (45)?
Want to harm others	94 (47)	69 (163)???		52 (26)	23 (55)???
Collateral self/other	95 (202)	12 (9)???		33 (70)	15 (11)??
danger
Prominent death wish	94 (29)	70 (173)	??	55 (17)	24 (58)	???
Significant loss	81 (43)	71 (157)		32 (17)	26 (57)
Hospitalized for risk last	75 (9)	73 (188)		42 (5)	26 (68)

week

Suicide risk stratification

Negative item

Positive item

???

Negative item

suicide risk stratification, consistent with known behavior of seasoned clinicians who do not move directly from single risk factors (ie, a posi- tive screening item) to judgments about disposition. Operationally, sea- soned clinicians know that not all patients with positive risk screening items will necessarily benefit from psychiatric hospitalization. Thus, the clinicians proceeded to gather all information about a variety of po- tential risk factors and weighed them in determining overall suicide risk stratification before they moved forward to clinical decision making about disposition.

The finding in this study that the 6 violence screening items by

High (vs moderate) 92 (45) 80 (39)

Moderate 83 (80) 40 (38)

Low (vs moderate) 61 (86)??? 4 (5)???

? =p b .05,

?? =p b .01,

??? =p b .001

themselves proved insufficient for clinicians to arrive at their final dis- positions was predicted by Berman and Silverman [12]. They concluded that although screening questions can be of value by combining them with other clinical factors such as the patient’s affective presentation and the evaluation of both chronic and acute factors, screening

Table 6

Two multiple regression models predicting psychiatric hospital disposition (dependent variable) from high-yield violence screening items and suicide risk stratification by auditor (inde- pendent covariates) entered simultaneously into the models

Significantly predicting independent covariates	df	Parameter estimate	Standard error	t value	P value	Odds ratio 95% Confidence
							limits
High-specificity violence screening items model Positive on one or both: wish to harm others, death wish	1	-0.52	.44	1.34	.248	0.60	0.25	1.43
Suicide risk stratificationa	1	2.45	.33	54.20	b.001	11.55	6.02	22.16
High-sensitivity violence screening item model
Negative on collateral self/other danger	1	0.49	.47	1.10	.294	1.63	0.66	4.06
Suicide risk stratificationa	1	2.40	.30	65.16	b.001	11.03	6.16	19.77
a Low = 1, moderate = 2, high = 3.

questions alone will not suffice for informing Clinical decisions regard- ing patient dispositions. Part of what screening questions fail to capture is essential nonverbal and affective clinical cues used by clinicians in assessing patient risk [12]. To address these limitations, Berman and Silverman [12] proposed a suicide risk formulation model that incorpo- rates observable clinical cues along with well-established historical risk factors such as hopelessness, personal history of depressive episodes, and family history of depression and/or suicide to predict imminent risk for self-directed violence.

Most patients in this study presented to the PES under involuntary status. After being assessed with the 6 violence risk screening items, nearly half of those who had presented involuntarily received a low- risk determination, and more than two-thirds did not require disposi- tion to a higher level of care such as psychiatric hospitalization. Thus, al- though the majority arrived involuntarily, the majority were discharged without recommendation for inpatient care. This suggests the possibil- ity that various elements of the system contributing to the identification of risk in the community (such as police or other collateral sources of in- formation) may be functioning with high levels of sensitivity in re- sponse, casting a very wide net in situations of perceived high potential harm. This situation instinctively leads to very conservative re- sponses because missing a case leading to harm could be catastrophic and the consequences of emergency detention of patients who turn out not to be high risk are small in comparison. Once the patient is detained in the ED, the clinicians with the expertise to assess and deter- mine appropriate disposition have an opportunity to make a more pre- cise and definitive determination of risk. At the same time, the risk level of some patients may diminish during the course of the ED visit through various combinations of clinical attention and intervention, adequate connection with follow-up care, and the simple passage of time.

A primary strength of the current study was the systematized, thor- ough EHR Review process. The auditor was blinded to provider risk as- sessment and ultimate patient disposition. Study-specific definitions of suicide risk-stratification terms were developed by the consensus of many clinicians. Excellent interrater reliability was achieved for the in- dependent auditor and provider suicide risk stratification ratings.

Figure. Trajectory of patients through the ED with violence risk screening and suicide risk stratification processes informing clinical management and disposition.

Finally, the study sample was systematically derived from a large ED with a diverse patient population.

A limitation of the study was the retrospective cross-sectional design using data from clinical records of single encounters. This analysis was confined to data obtained from 6 violence risk screening items, overall suicide risk stratification, and eventual disposition. Additional factors that may have played a role in risk assessment and decision making were certainly not part of the data used in these analyses. The findings in this setting may not generalize to other ED settings such as rural and smaller hospital environments.

This study developed a risk assessment framework through genera- tion and analysis of consensus definitions for suicide risk stratification. Utilization of a consistent definition for suicide risk stratification by be- havioral health providers in varioUS settings can move the field toward standardization of the assessment and decision-making process for pa- tients at potential risk of harm. Research has shown that training specif- ically focused on evidence-based suicide risk assessment improves clinical practice and documentation, and has the potential to help pro- viders manage the ethical and legal aspects of caring for suicidal pa- tients [19,23-25].

Although imminent suicide cannot be predicted with 100% certainty for every patient in the ED setting, reliable determination of the level of suicide risk according to established stratification levels can inform clin- ical management. A subsequent suicide of a patient evaluated in the ED is not necessarily grounds for litigation; however, omission of established procedures to determine the suicide risk level in a patient who then dies by suicide may well provide a legitimate basis for litiga- tion. This project has generated a set of procedures to operationalize sui- cide risk assessment in the ED that can represent due diligence in the treatment of patients with risk for suicide.

Additional research is needed to validate this study’s model incorpo- rating violence risk screening with determination of suicide risk stratifi- cation in support of patient management and disposition. This may be accomplished through examination of relevant patient outcomes (eg, ED recidivism, inPatient psychiatric admissions, morbidity, and mortal- ity) in relation to the violence risk screening and suicide risk stratifica- tion elements of the model.

Acknowledgments

The authors would like to acknowledge Mr Dana Downs and Mr James Foster for their assistance with data management and adminis- trative matters.

References

1. World Health Organization. Suicide fact sheet. Retrieved from http://www.who.int/ mediacentre/factsheets/fs398/en/; 2014.
2. Centers for Disease Control and Prevention. Fatal injury data. Retrieved from http:// www.cdc.gov/injury/wisqars/.
3. World Health Organization. Prevention of suicide: guidelines for the formulation and implementation of national strategies. Geneva: World Health Organization; 1996.
4. United States Public Health Service. The Surgeon General’s call to action to prevent suicide; 1999[Washington DC].
5. United States Department of Health and Human Services. National strategy for sui- cide prevention: goals and objectives for action; 2001[Rockville MD].
6. United States Department of Health and Human Services. 2012 National strategy for suicide prevention: goals and objectives for action. Washington, DC: HHS; 2012.
7. The Joint Commission. Behavioral health care: national patient safety goals effective January 1, 2015. Retrieved from http://www.jointcommission.org/assets/1/6/2015_

   NPSG_BHC.pdf; 2015.

   Gaynes BN, West SL, Ford CA, Frame P, Klein J, Lohr KN. Screening for suicide risk in adults: a summary of the evidence for the U.S. Preventive services Task Force. Ann Intern Med 2004;140(10):822-35.

8. LeFevre ML, Force USPST. Screening for suicide risk in adolescents, adults, and older adults in primary care: U.S. Preventive Services Task Force recommendation state- ment. Ann Intern Med 2014;160(10):719-26.
9. O’Connor E, Gaynes BN, Burda BU, Soh C, Whitlock EP. Screening for and treatment of suicide risk relevant to primary care: a systematic review for the U.S. Preventive Services Task Force. Ann Intern Med 2013;158(10):741-54.
10. United States Preventive Services Task Force. Screening for suicide risk: recommen- dation and rationale. Ann Intern Med 2004;140(10):820-1.
11. Berman AL, Silverman MM. Suicide risk assessment and risk formulation part II: sui- cide risk formulation and the determination of levels of risk. BehaviorSuicide Life Threat Behav 2014;44(4):432-43.
12. Luoma JB, Martin CE, Pearson JL. Contact with mental health and primary care pro- viders before suicide: a review of the evidence. Am J Psychiatr 2002;159(6):909-16.
13. Ahmedani BK, Simon GE, Stewart C, Beck A, Waitzfelder BE, Rossom R, et al. Health care contacts in the year before suicide death. J Gen Intern Med 2014;29(6):870-7.
14. Gairin I, House A, Owens D. Attendance at the accident and emergency department in the year before suicide: retrospective study. Br J Psychiatry 2003;183:28-33.
15. Da Cruz D, Pearson A, Saini P, Miles C, While D, Swinson N, et al. Emergency depart- ment contact prior to suicide in mental health patients. Emerg Med J 2011;28(6): 467-71.
16. Claassen CA, Larkin GL. Occult suicidality in an emergency department population. Br J Psychiatry 2005;186:352-3.
17. Appleby L, Shaw J, Amos T, McDonnell R, Harris C, McCann K, et al. Suicide within 12 months of contact with mental health services: national clinical survey. BMJ 1999; 318(7193):1235-9.
18. Simon RI. Improving suicide risk assessment with evidence-based psychiatry. In: M. T., Pompili R, editors. evidence-based practice in suicidology: a sourcebook. Cam- bridge, MA: Hogrefe; 2011.
19. Spitznagel EL, Helzer JE. A proposed solution to the base rate problem in the kappa statistic. Arch Gen Psychiatry 1985;42(7):725-8.
20. Goldacre M, Seagroatt V, Hawton K. Suicide after discharge from psychiatric inpa- tient care. Lancet 1993;342(8866):283-6.
21. Qin P, Nordentoft M. Suicide risk in relation to psychiatric hospitalization: evidence based on longitudinal registers. Arch Gen Psychiatry 2005;62(4):427-32.
22. Koopersmith EG, Tarpey CM. Do no harm: physicians’ duties toward suicidal pa- tients. Med Econ 2014;91(9):30-1.
23. McNiel DE, Fordwood SR, Weaver CM, Chamberlain JR, Hall SE, Binder RL. Effects of training on suicide risk assessment. Psychiatr Serv 2008;59(12):1462-5.
24. Simon RI. Suicide risk assessment: what is the standard of care? J Am Acad Psychi- atry Law 2002;30(3):340-4.