A 20-year follow-up survey of police officers' experience with Tarasoff warnings: How law enforcement reacts to clinicians' duty to protect.

Since the Tarasoff case of 1976, mental health professionals are recognized to have a "duty to protect" third-party targets from violence-threatening patients, but little is known about what happens after clinicians warn law enforcement. In 2000, Huber et al. published a study that surveyed Michigan police about "Tarasoff warnings." We conducted a 20-year follow-up study, inviting all Michigan police and sheriff departments to participate. There were no significant differences between studies about knowledge of Tarasoff-related policies, which was low in both surveys. We found significant decreases in the number of officers who had ever intervened due to warning calls. Of the survey respondents, 83% supported documenting warning calls. For those who received warnings, 96% followed up with at least one intervention. In both studies, notifying other officers was the most common action taken. 56% said they would take action to remove a firearm. We identified opportunities for training law enforcement.

INTRODUCTION

In 1969, Prosenjit Poddar, a University of California (UC) Berkeley student, told a campus psychologist that he intended to kill Ms. Tatiana Tarasoff, a student who had rebuffed his romantic advances. The psychologist informed campus police, recommending civil commitment for dangerousness. Police briefly detained Mr. Poddar but released him because he “appeared rational,” and he stopped seeing the psychologist (Tarasoff v. Regents, 1976, p. 430). A few months later, Mr. Poddar stabbed Ms. Tarasoff to death. Mr. Poddar was convicted of second‐degree murder, which was overturned by an appellate court on the condition that he return to his home country of India. Ms. Tarasoff's parents sued UC and the psychologist on the basis that no one ever directly notified her of threats. The trial court and appeals court ruled that the psychologist could not be held liable for failing to successfully hospitalize Mr. Poddar, but the California Supreme Court ruled that clinicians have a common law “duty to warn” individuals threatened by patients in People v. Poddar (1974), also known as Tarasoff I. Protests led the California Supreme Court to re‐hear the case. The California Supreme Court—in a 5‐2 decision—further refined the previous ruling in Tarasoff v. Regents (1976), also known as Tarasoff II. The Court ruled that the therapeutic relationship created a “duty to protect” others from identifiably dangerous patients and “exercise reasonable care” such as directly warning potential victims or others likely to apprise the victim of the danger, notifying police, and/or taking other reasonable steps under the circumstances. Regarding confidentiality, the Court stated, “protective privilege ends where the public peril begins” (Tarasoff v. Regents, 1976, p. 442). The Court also noted that “the professional inaccuracy of predicting violence cannot negate the therapist's duty to protect” (Tarasoff v. Regents, 1976). This was landmark case because it established that clinicians have a duty to take reasonable steps to protect third parties from patients on the basis of patients' confidential statements, and went against the prevailing practices of that time that made confidentiality of patient information revealed during therapy a higher priority.

The murder of Ms. Tarasoff and the resultant landmark California Supreme Court case changed how the mental health system, legal system and public viewed the responsibility of mental health professionals to potential victims of violence at the hands of persons in their care. Tarasoff established the “duty to protect,” that set the stage for jurisdictions to not only allow but also in some states to obligate certain mental health clinicians to breach patient confidentiality when there is a perceived danger to an identifiable victim revealed in the context of a therapeutic relationship. Although Tarasoff technically only applies to California, its principles are now largely encapsulated in law throughout the US, via case law and/or statutory law. Since the landmark case in California, most other states have enacted laws that address either requirements or permissions for clinician action when patients are considered a threat and limit liability to narrow circumstances, and case law continues to define the contours of these expectations. These standards vary in language and scope, with some conforming with duty to warn (i.e., reasonable attempts to communicate the threat to the third person) and/or duty to protect standards (e.g., initiating proceedings to hospitalize the patient, communicating threats to law enforcement in addition to or instead of communicating the threat to the third person, etc.), and/or defining limits on clinician liability (e.g., only reasonably identifiable targets, direct communication, apparent intent, ability to carry out the threat in the foreseeable future).

Michigan enacted a statute to address the duty to protect (Michigan Compiled Laws § 330.1946), which was enacted in 1989 and last amended in 1996. The duty is only triggered by the following circumstances:

If a patient communicates to a mental health professional who is treating the patient a threat of physical violence against a reasonably identifiable third person and the recipient has the apparent intent and ability to carry out that threat in the foreseeable future, the mental health professional has a duty to take action as prescribed….

The duty is discharged when the mental health professional does one or more of the following actions in a “timely manner”:

Initiate hospitalization proceedings

Notify the potential victim's local or state law enforcement, and make “a reasonable attempt” to notify the potential victim

If the potential victim is a minor or incompetent, notify individuals parent or guardian and notify their local social services, in addition to (b)

Tarasoff progeny often include the option of making a warning to LEOs, but little is known about what happens after a warning is made. Despite the statute, there are still cases that have resulted in confusion about the intersection of statutory, common law and professional ethical duties (e.g., Hartwig et al., 2011). Although the Michigan statute helps clarify actions that both fulfill a mental health clinician's duties following patient threats of physical violence against a third person and limit liability for certain confidentiality breaches, the law does not cover subsequent responsibilities and actions of law enforcement when they receive calls. This is important because inaction following a warning call could be a missed opportunity to save lives. For example, James Holmes' psychiatrist warned police that he was dangerous and had homicidal ideations before he murdered 12 and injured 70 (USATODAY, 2013).

This study evaluates perceptions by law enforcement of police policy and procedure after a clinician makes a “Tarasoff call” to a law enforcement officer (LEO). The “Tarasoff” case only applies in California, but it has become a common term used to reference the duty to protect third parties, and thus will be used throughout this paper. To our knowledge, only one other study has evaluated this topic (Huber et al., 2000). The original study revealed low rates of knowledge among LEOs of the Tarasoff case and the related law. This study set out to conduct a 20‐year follow up to Huber et al. (2000) in Michigan as an exploratory analysis of any evolving patterns. In addition, we aimed to specifically assess how warning calls are documented and, given the high‐profile attention to firearms and mental illness over the last 20 years, how the involvement of firearms in warnings influences interventions. This study aims to improve understanding of policies and procedures, and inform system improvements related to managing risk among individuals with mental illness who make a credible threat of violence.

METHODS

This study was approved by the Institutional Review Boards of the Michigan Department of Health and Human Services, and University of Michigan. All police departments and sheriff departments in Michigan were invited to participate. In 2019, online surveys were distributed though the state law enforcement organizations, the Michigan Sheriff's Association and the Michigan Association of Chiefs of Police. A cover sheet to the survey explained the objectives, and indicated that completion of the survey implied consent. No names or other personal identifying information were collected. We used survey anonymizing processes through Qualtrics (umichumhs.qualtrics.com.).

We contacted an author from Huber et al. (2000). The authors did not have the list of police stations they surveyed, but indicated that they were selected randomly. They also did not have their original questionnaire, but provided examples of questions they asked based on recall and review of the article (these questions match the seven variables assessed in the original article), and gave us permission to use these questions:

Do you know what the Tarasoff case is? Have you heard about it?

Does your station have a policy on such warnings?

Have you ever received a warning from a clinical site or clinician regarding someone threatening to kill/harm someone?

If you have, have you notified a potential victim?

Have you monitored a potential victim after such notification?

Have you documented a warning call from a clinical source?

Have you notified other officers about the warning?

Regarding questions two to seven, each question was prefaced with, “Regarding a warning from a clinical site or clinician about someone threatening to kill or harm someone else (aka Tarasoff warning)....” We elected to use these questions (with some modification to use language also to convey Michigan law) as well as additional investigator‐designed queries:

How are warning calls documented, if at all?

What is the procedure for an individual with mental illness who may have access to a firearm and for whom a warning has been received?

Answers were compared between the current study and the Huber et al. (2000) study using t tests, which were two‐tailed and presented at the α = 0.05 level of significance.

RESULTS

Eighty‐eight officers returned completed surveys, compared to 48 from the original study. Because the survey could be forwarded, it is unclear how many of the over 17,000 LEOs in Michigan (Michigan Joint Task Force, 2019) received one or the number of those who chose not to complete it. Figure 1 provides a pictorial summary of all results. Table 1 compares the current study's results with those of Huber et al. (2000). Compared to the original study, only knowledge of the Tarasoff case increased (by 6%), but the change was not statistically significant. All other percentages decreased, significantly so for notifying other officers about a warning (by 37%), monitoring a potential victim (34%), notifying a potential victim (33%), and documenting a warning call (32%). In both studies, notifying other officers was the most common action taken in response to a warning call, followed by documenting the call, notifying a potential victim and monitoring a potential victim. Among those who received a warning call, 96% responded with one or more of these interventions. In both our study and Huber et al. (2000), there were several LEOs who had taken action without having personally received the warning call.

FIGURE 1

Summary of results regarding duty to protect warning

TABLE 1

Comparison of law enforcement survey responses in current and previous study

	Current study N = 88	Huber et al. (2000) N = 48
	N (%)	N (%)	p
Station has a policy on warnings	27 (31%)	19 (40%)	0.2979
Officer has received a warning	44 (50%)	31 (65%)	0.1037
Officer has notified a potential victim	50 (57%)	43 (90%)	<0.0001
Officer has monitored a potential victim	34 (39%)	35 (73%)	<0.0001
Officer has documented a warning call	51 (58%)	43 (90%)	<0.0001
Officer has notified other officers about a warning	52 (59%)	46 (96%)	<0.0001
Officer has knowledge of the Tarasoff case	9 (10%)	2 (4%)	0.2184

When asked if warning calls are documented, 73 (83%) responded affirmatively, and Table 2 summarizes how they responded in an open‐ended format about the process. LEOs used a variety of terms when referring to reports and listed a myriad of different methods and systems—some electronic and some written. Respondents used the following terms: “police report,” “criminal report,” incident report,” “complaint,” “threat report,” “caution summary,” “daily log,” and “brief report.” It is likely that many of these are synonymous with one another, but it is also clear based on the responses that different types of reports are used for different levels of officer‐perceived severity of the information.

TABLE 2

Law enforcement responses about how warning calls should be documented

	N (%)
Formal report	49 (67%)
Dispatch/call for service	12 (16%)
Recorded lines	1 (1.3%)
Memo to all officers/potential victims	1 (1.3%)
Notice to chief of police and email to all officers	1 (1.3%)
Mental health petition	1 (1.3%)

Respondents were asked to describe the procedure for an individual with mental illness who may have access to a firearm and for whom a warning has been received (Table 3). Many responded with multiple options, and many stated that situations would be handled on a case‐by‐case basis. The most common response was to try to remove a firearm for safekeeping, followed by initiating mental health assessment, law enforcement investigation, or trying to prevent purchasing firearms. With regard to preventing firearm purchases, some said they would contact the prosecutor, court, state police, and/or seek a mental health petition. With regard to removing firearms, most did not describe a specific mechanism for doing so. Some simply said they would seize the weapon “if possible,” “if circumstances warrant,” “if legal grounds to do so are established,” or “when legally able to,” without describing what would determine their ability to do so. Among those that did, most said they would seek a court order to seize them, while many said they would request voluntary surrender of firearms, and/or would contact family members or friends to remove them. Many responses indicated an acute awareness of the law and were very mindful of acting in accordance with the law and respecting rights when attempting to limit access to firearms (e.g., “fourth amend,” “search and seize”). However, some stated they would seize the weapon without mention of a court order and only return the weapon with a court order, and one respondent stated that they would seize firearms “until a doctor signs off on the return of the weapons.”

TABLE 3

Law enforcement responses about how warning calls involving firearms should be handled

	N (%)
Take action to remove firearm	49 (56%)
Via prosecutor/court order	17 (35%)
Via voluntary surrender	8 (16%)
Via family members or friends	7 (14%)
Via probable cause	2 (4%)
Only returned via court order	2 (4%)
Via state police	1 (2%)
Only returned via physician order	1 (2%)
Via confiscation by nondescript legal mechanism	5 (10%)
No legal or practical mechanism described	23 (57%)
Initiate mental health evaluation	13 (15%)
Investigate/assess/interview/observe	11 (13%)
Take action to prevent purchase of firearms	3 (3.4%)
Monitor/“heightened awareness”	2 (2.3%)
Notify area law enforcement	1 (1.1%)
Notify the potential victim	1 (1.1%)
Unknown/no policy	2 (2.3%)

DISCUSSION

Our study indicates that, in the last 20 years, Michigan LEOs have become somewhat more familiar with the elements of the Tarasoff case (4% vs. 10%) but are less aware of station policies about the handling of warning calls from clinicians about potentially violent patients (40% vs. 31%). The finding about increased LEO familiarity is consistent with a 20‐year follow up study (Anfang & Appelbaum, 1996) that reported that clinicians seemed to have gained familiarity with the Tarasoff ruling and how it might impact treatment, as well as how to work with patients to effectuate a duty to protect while balancing a treatment alliance. Though neither of our differences in general knowledge were statistically significant, the fact that the majority of LEOs were not aware of either the Tarasoff law (90%) or their station's policy (59%) indicate an important training and/or knowledge gap. Though the landmark case that resulted from her killing only applies to California, Ms. Tarasoff's tragic story is powerful and memorable, which may help LEOs remember the legal and ethical principles that resulted (Tarasoff v. Regents, 1976). We acknowledge that the national evolution in various states of the duty to warn or protect may be less relevant to local law enforcement than training on the requirements and intent of the local law, but we believe a brief overview of Tarasoff and case history is a compelling introduction. Although the story of Tatiana Tarasoff is fascinating, the most relevant focus of training should be the specifics of local law (e.g., statutory and/or “Tarasoff progeny” cases) and departmental policies. If no such policies exist, law enforcement agencies may benefit from partnering with mental health professionals to help develop specific trainings and policies on how to handle these warnings.

Compared to Huber et al. (2000), respondents to our study were less likely to have received warnings and significantly less likely to have taken action in response to these warnings. Although the Huber et al. (2000) study does not break down responses for just those who received warning calls, in this study, among those who had received warning calls, 96% of LEOs responded with one or more interventions. In fact, 91% responded with three or more interventions. Even if the current study showed lower rates of action, notifying other officers was the most common intervention in both studies. This is somewhat surprising since it was more common than the seemingly easiest action—documentation—which was a close second. Conversely, direct interactions with potential victims—via notification or monitoring—were the least likely interventions. This may speak to the discomfort involved in notifying someone of a potential threat or perhaps indicates that, because Michigan law suggests that clinicians can warn potential victims, LEOs consider this option less frequently. However, frequency of notification was comparable to the other most prevalent interventions, with only monitoring being much less prevalent. This may indicate a lack of time or resources to provide indefinite monitoring or a sense of futility when there is not a known timeframe of a potential threat. Overall, our findings suggest that LEOs are less likely to receive warnings than 20 years ago, but those who do are likely to take them seriously based on almost always responding with multiple interventions.

With regard to warning calls involving firearms, there was even more variability in the responses that LEOs described. Our findings indicate that, while most LEOs are aware of legal mechanisms and civil rights and would not automatically seize firearms based on a warning call, some LEOs are not aware. For most of those who responded that they favored seizing firearms, it is unclear if the LEO was aware of legal mechanisms for seizing firearms. However, it is clear that a few LEOs did not know or would not utilize a legal means to confiscate firearms (e.g., via probable cause and/or via state police). Some stated that the firearms would only be returned via court order or physician, and all of these did not describe a legal mechanism for seizure in the first place. Furthermore, it is notable that few LEOs would utilize the mechanism for civil commitment (Michigan Compiled Laws § 330.1401) that could lead to treatment of individuals with mental illness who might be at risk of suicide or violence related to a firearm. In the authors' experience, many LEOs are unaware that they may take an individual into protective custody and to a preadmission screening unit (e.g., hospital, community mental health crisis service) without a petition or certificate by a clinician if the “officer observes an individual conducting himself or herself in a manner that causes the peace officer to reasonably believe that the individual is a person requiring treatment” (Michigan Compiled Laws § 330.1427). The need for formal policies and/or training on existing policies is clearly needed. Given our findings, LEOs may benefit from training on their ability (or lack thereof) to seize firearms from persons with mental illness based on local gun restriction laws and/or other court orders (e.g., criminal, civil), and limitations based on local and federal laws and civil rights (e.g., second and fourth amendments). Despite individuals with serious mental illness accounting for 0.02%–1% of gun‐related homicides, many firearm policies tend to focus on individuals with mental illness despite their low violence risk (Knoll & Annas, 2016; Lankford, 2015; Thornicroft, 2020). Though there have been mixed results regarding connecting the effects from general gun access bills on violence (e.g., Fleegler et al., 2013 vs. RAND, 2018, de Jager et al., 2018 vs. Sarani et al., 2019), some jurisdictions have risk protection orders (e.g., red flag laws), which have been found to save one life for every 10–20 removals, though the majority of these lives saved may be from suicide (Swanson et al., 2017; National Council, 2019).

There are real challenges for determining what actions might be protective and preventive in a given case, while also respecting rights (e.g., typically an individual may have not yet committed any crimes). The nature of the responses to our survey indicates that LEOs often make determinations based more on subjectivity than official policy. For example, regardless of whether a survey respondent had ever received a call, 83% of those we surveyed reported that warning calls should be documented. Of those, most said they would file a report (67%), but the type of report varied (perhaps based on subjective determination by the officer or on interdepartmental practice variation), and a call for service was a distant second (16%). On the one hand, this indicates the need for formal policies outlining reasonable interventions and proper documentation, and trainings about such policies. On the other hand, one could argue that the reason there are not clear policies and guidelines related to warnings is that these are incredibly complicated cases that require balancing the rights of individuals in treatment and often lack clarity about what actions might be helpful for preventing violence from a LEO perspective. For example, in the Tarasoff case, LEOs were notified and interviewed Mr. Poddar, but assessed him as not posing a threat and released him (Tarasoff v. Regents, 1976). Our findings point to a recommendation that LEO trainings include discussion of their options based on state and local laws and policies, potential risks and benefits of each, legal and ethical considerations (e.g., rights of potential victims and rights of individuals who make threats), and the potential negative outcomes for victims and law enforcement agencies if there is no action after receipt of a warning. It is still important that LEOs are aware of the dangers of lives lost from inaction following a warning (e.g., the Holmes Case [USATODAY, 2013]) and that courts with Common law traditions have extended the duty to protect to law enforcement (e.g., Doe v. Metropolitan Toronto Commissioners of Police [1998] in Canada). Case examples and interactive discussions could greatly benefit all involved in preparing LEOs to more thoroughly consider the legal, ethical and practical implications of their options.

Limitations of our study include cross‐sectional survey methodology and generalizability. Although we were fortunate to have the support of statewide police and sheriff organizations that allowed us to reach out to all law enforcement agencies in Michigan, our anonymous surveys do not allow us to know which departments, geographical locations, or types of LEOs responded. Therefore, we are not able to account for overall selection bias or ensure generalizability. Given how our surveys were widely distributed with an email link and anonymously collected online, there is no data on how many of the over 17,000 LEOs in Michigan (Michigan Joint Task Force, 2019) received one or the number of those who chose not to complete it. Although in theory perhaps only those with some vested interest and bias would complete the story, there were no clear patterns among respondents that would suggest that this was the case. Additionally, our n was higher than the Huber et al. (2000) study which, to our knowledge, is the only other study of this kind. Surveys themselves are subject to self‐report bias, but our anonymous surveys likely allow for more candidness and veracity than an interview might, and increase the breadth of LEO experiences that we were able to capture across the whole state. Still, the survey data available does not provide information about the respondents. That some LEOs reported taking action without receiving a warning call could potentially mean that they misunderstood the questions, but we think it is more likely that the LEO who intervened was not always necessarily the LEO that received the original warning call (e.g., LEOs answering phones initiate a report or dispatch that leads to action by another LEO), especially since this was a consistent finding in both our study and Huber et al. (2000). Because most of our survey questions used the exact wording from Huber et al. (2000) we were bound by some limitations of that wording. However, because it is possible that a LEO may not have heard of “Tarasoff” but was familiar with local policies, we added language (see Section 2) in most survey questions to prevent underestimating knowledge. Although we believe this change removes that limitation, it adds a limitation still of making it more difficult for direct comparisons with the previous study. Finally, the author of Huber et al. (2000) no longer had all the original data and therefore we were unable to include a comparison of those who did and did not receive warning calls between the two studies among all the other comparative analyses we performed. Nevertheless, we believe that the strengths of our study outweigh the limits, given the difficulty of assessing LEO experiences and the paucity of studies on this important topic.

In conclusion, this study uniquely advances the field by conducting a 20‐year follow up study of the knowledge and responses police officers would have to receiving a Tarasoff warning call from someone in their community. Awareness of the Tarasoff case was low and LEOs responded in a variety of ways to warnings, including with regard to firearm removal, sometimes even proposing interventions without legal foundation. We identified many opportunities for targeted education of law enforcement, especially about local policies and local, state and federal laws related to duty to protect, civil commitment, and firearm restrictions. Although this study was focused on knowledge of LEOs, from this study, there is the potential for further collaboration and cross‐training between law enforcement and clinicians to achieve the best responses.

CONFLICT OF INTEREST

None declared.