When the California Supreme Court vacated its 1974 Tarasoff decision (3) and redecided the case in 1976 (1), it replaced the phrase "duty to warn" with "duty to protect." Much has been made of this. In fact, the earlier phrase was accurate, the later one rhetorical and misleading.
The core innovation of Tarasoff was the creation of a new exception to psychotherapist-patient confidentiality. Naturally, if justification for the exception lapses in a particular case, the duty (and authority) to warn vanishes. Death of the patient would be an extreme example. It goes without saying, literally, that incapacitation by virtue of civil commitment, already available to clinicians in California and elsewhere when Tarasoff was decided, similarly vitiates any duty to warn.
This is why most current Tarasoff statutes—including California’s, which supersedes the Tarasoff decision—couch the duty exclusively in terms of warning, making no mention of commitment or any other alternative: Alabama, California, Connecticut, the District of Columbia, Florida, Idaho, Illinois, Kansas, Louisiana, Minnesota, Mississippi, Montana, Nebraska, New York, Oregon, Rhode Island, Texas, Utah, Washington, and West Virginia. Although other states list civil commitment as an option, nowhere is it a duty. Warning alone always discharges the duty. If anything, Tarasoff thus weakens the case for a duty to protect by providing a potential defense (if a warning was given) against a suit for negligent noncommitment.
Further, axiomatically, a court cannot decide matters that are not before it—cannot impose liability for not doing what was done. In Tarasoff, the clinicians had done everything but warn the victim: they had tried to persuade the patient to continue with therapy, they had notified the police (and through them the victim’s family), and they had tried to commit the patient (Ref. 1, p 341). Therefore, "duty to protect," insofar as it connotes anything beyond a duty to warn in the Tarasoff opinion, is dictum, not law.
Finally, protecting an unwarned third party by lesser remedial measures, such as referral to a "program for continuing care, referral to a V.A. hospital or other outpatient clinic, and implementation of a program to monitor [the patient’s] medication" (4), will, after a tragedy, be unlikely to impress a jury that knows more could have been done and hears expert testimony that more should have been done (5).
Tarasoff and its statutory and case law progeny therefore, as a practical matter, distill down to a duty to warn, in essentially two situations. One is where the therapist believes the patient is not a danger to himself (or herself or others) or is not mentally ill—hence, not committable—but he (or she) has made a threat to harm another (or, in some jurisdictions, a suicide threat). This would occur either in discharging an inpatient or in electing not to hospitalize an outpatient—that is, in the case of a decision not to contain the patient. The second situation arises from inability to contain the patient, such as when an outpatient phones in a threat or an inpatient elopes.
Privilege versus confidentiality
The psychotherapist-patient privilege, some version of which all jurisdictions now have (6), is distinct from psychotherapist-patient confidentiality (7).
Privilege, a modern concept (6), is the right of the patient not to have a psychotherapist disclose in judicial proceedings "any confidential communication between patient and psychotherapist" (8).
Confidentiality is the ancient obligation of the psychotherapist not to divulge in any setting the identity of the patient as a patient or any information about the patient known to the therapist by virtue of the therapeutic relationship. As Hippocrates enjoined, "Whatsoever things I see or hear concerning the life of men, in my attendance on the sick or even apart therefrom, which ought not to be noised abroad, I will keep silence thereon, counting such things to be as sacred secrets" (9).
One of the exceptions to privilege recognized in some jurisdictions is the "dangerous patient exception" designed to untie the hands of courts (e.g., in a commitment hearing) with respect to future violence. Some courts have melded the dangerous patient exception to privilege with the Tarasoff exception to confidentiality (also conceived to prevent future violence), to make psychotherapists into obligatory witnesses for the prosecution against their patients with respect to purely past acts, reasoning that confidentiality is permanently lost once a Tarasoff duty arises, even if no warning is given (10). Some commentators have denounced this "criminalization" of Tarasoff (11), and one court has agreed (7).
Ethics codes are unlikely to be determinative. After all, no disciplinary sanctions could ensue from following the law, and no clinician (one hopes) would subordinate his or her moral compass to a generic code of ethics on a matter, literally, of life and death. The psychiatry code of ethics comes down squarely on the fence: "Psychiatrists at times may find it necessary, in order to protect the patient or the community from imminent danger, to reveal confidential information disclosed by the patient" (12).
In Tarasoff, the court declared that "once a therapist does in fact determine, or under applicable professional standards reasonably should have determined, that a patient poses a serious danger of violence to others, he bears a duty to exercise reasonable care to protect the foreseeable victim of that danger" (Ref. 1, p 345). Because it was a decision by a state appellate court, it was law only in that state, California. Soon, however, the notion of a duty to warn crept across most of the nation, by court decisions or legislative enactment. In the process, important variations have emerged.
Common law versus statute
New law can be made or existing law changed by appellate courts (common law, or precedent) or by legislatures (statutory law). Tarasoff was a judicial innovation in (or extension of) common law. Many courts have considered a duty to warn since Tarasoff was decided—and invariably cite Tarasoff in doing so—but most of the duty to warn law as it exists today is statutory.
California is illustrative. Tarasoff itself no longer defines the duty to warn in that state. In 1985, the state legislature superseded the case by enacting a statute that currently provides:
. . . no cause of action shall arise against . . . any . . . psychotherapist in failing to warn of and protect from a patient’s threatened violent behavior or failing to predict and warn of and protect from a patient’s violent behavior except where the patient has communicated to the psychotherapist a serious threat of physical violence against a reasonably identified victim. . . . If there is a duty to warn and protect under the limited circumstances specified above, the duty shall be discharged by the psychotherapist making reasonable efforts to communicate the threat to the victim . . . and to a law enforcement agency (13).
This statute appears to curtail Tarasoff slightly in requiring an explicit threat from the patient. At the same time, it potentially expands Tarasoff slightly, in that it expressly requires notification of both law enforcement and the victim (not just the victim, as Tarasoff implied). Further, it makes plain, notwithstanding the incantation "and protect," that the only duty is to warn: " . . . the duty shall be discharged by . . . reasonable efforts to communicate the threat to the victim . . . and to a law enforcement agency" (13).
As in California, the basic duty-to-warn law currently in force in most of the jurisdictions that have such a law is statutory. This, however, does not end or negate the role of courts in sculpting such law.
Again, California is illustrative. The California Court of Appeal was called on to apply that state’s duty-to-warn statute (just quoted) in Barry v. Turek (14). A brain-injured patient on a locked psychiatric ward habitually engaged in grabbing and fondling improprieties toward female nurses and had to be continuously redirected from such activities. However, as the court makes clear, he "never made verbal threats of violence within the hearing of [defendant, the ward’s chief psychiatrist]"(Ref. 14, p 554). Under the statute, the absence of an explicit threat should seemingly end the case, as the court itself acknowledged (Ref. 14, p 554).
Eventually, a hospital employee was assaulted more seriously than those in the previous incidents, and she sued the patient’s psychiatrist for failure to warn. Despite the statute’s clear language requiring that the patient communicate to the psychotherapist a serious threat of physical violence, the court asked "whether [the plaintiff has sufficiently shown that [the defendant] ought to have been aware that [the patient] presented a serious threat of physical violence" (Ref. 14, p 555).
The court concluded that the assault was not reasonably foreseeable. The point, however, is that statutes, no matter how clearly written, may be malleable in the hands of courts called on to interpret them. Thus, although most duty-to-warn law now is statutory, psychotherapists can never rest fully assured that a court decision will not abruptly alter their obligations in this area, as occurred in Tarasoff itself (15).
The 52 jurisdictions in the United States (the 50 states and the District of Columbia, plus federal law governing tribal lands, Puerto Rico, Guam, the Virgin Islands, and other territories) currently break down into four basic categories.
About half of the states (27), following Tarasoff, impose a mandatory duty to warn, although the precise contours of the duty vary considerably. Another 10 jurisdictions (9 states plus the District of Columbia) accord psychotherapists permission to warn (viz., an exception to psychotherapist-patient confidentiality) without explicitly imposing a duty to warn. One state, Virginia, flatly rejects Tarasoff, at least in the outpatient context. The remaining 14 jurisdictions (13 states, plus federal law) have no definitive law on the issue.
Twenty-seven states impose a duty to breach psychotherapist-patient confidentiality and warn of potential violence against a third party: Arizona (16), California (13), Colorado (17), Delaware (18), Idaho (19), Indiana (20), Kentucky (21), Louisiana (22), Maryland (23), Massachusetts (24), Michigan (25), Minnesota (26), Mississippi (27), Missouri (28), Montana (29), Nebraska (30), New Hampshire (31), New Jersey (32), Ohio (33), Oklahoma (34), Pennsylvania (35), South Carolina (36), Tennessee (37), Utah (38), Vermont (39), Washington (40), and Wisconsin (41).
The contours (and clarity) of this duty differ widely from jurisdiction to jurisdiction—a function, in part, of variability in the quality of legislative craftsmanship and, in part, presumably, owing to quite disparate levels of basic enthusiasm for the duty-to-warn principle. The general formulation is that a mental health worker is obligated promptly to notify either the potential victim or the police when a patient makes an explicit threat of serious physical harm against a readily identifiable third party (or optionally, in some states, to hospitalize the patient).
An important variation among the duty states is whether the duty is nondiscretionary—that is, an essentially ministerial function of simply transmitting a threat of violence versus subject to a predicate judgment by the psychotherapist as to the patient’s factual ability to fulfill the threat. Twelve states—California, Colorado, Indiana, Kentucky, Minnesota, Mississippi, Montana, Nebraska, New Hampshire, South Carolina, Utah, and Washington—couch the duty as relatively nondiscretionary. Montana’s statute, for example, provides:
A mental health professional has a duty to warn of or take reasonable precautions to provide protection from violent behavior only if the patient has communicated to the mental health professional an actual threat of physical violence by specific means against a clearly identified or reasonably identifiable victim (29).
Another 14 duty states—Arizona, Delaware, Idaho, Louisiana, Maryland, Massachusetts, Michigan, Missouri, New Jersey, Ohio, Oklahoma, Pennsylvania, Tennessee, and Vermont—explicitly incorporate the therapist’s judgment into the duty equation. Idaho’s statute is typical:
A mental health professional has a duty to warn a victim if a patient has communicated to the mental health professional an explicit threat of imminent serious physical harm or death to a clearly identified or identifiable victim or victims, and the patient has the apparent intent and ability to carry out such a threat" (19).
The law of Wisconsin, the remaining duty state, is unclear on this, as it is on many other particulars. The Wisconsin Supreme Court, in a four-to-three vote, established Tarasoff liability by dictum in a case not involving threats or any other evidence of violent intention (41).
Many additional variations exist among the duty states. Only a few can be highlighted.
Indiana, Maryland, Massachusetts, and New Jersey all require a psychotherapist not only to warn of explicit threats by the patient but also to discern in his or her actions or the circumstances any threat of violence and to warn of this as well. The first three of these states specifically mandate that the psychotherapist incorporate into this calculus the patient’s past acts of or propensity for violence.
In Minnesota and Ohio, a therapist is explicitly obligated to warn not only of threats to which he or she is personally privy but also threats reported by a third party.
Under the New Jersey law, threats of suicide also trigger a duty to warn. In Minnesota and Oklahoma, disclosure of suicide threats is discretionary.
In Minnesota, not only are suicide threats subject to the Tarasoff scheme, but provision is made for "[o]ptional disclosure . . . to third parties in a good faith effort to warn against or take precautions against a client’s violent behavior . . . for which a duty to warn does not arise" (26). This potentially creates a large hole in the veil of confidentiality. Are ruminations that are not quite threats no longer confidential? What about threats that are not necessarily violent, threats that are vague as to the identity of any targets, threats of minor (not serious) violence, or threats against property?
New Hampshire’s Tarasoff statute mandates disclosure of threats against real property (as well as those against persons). Vermont’s supreme court decision establishing Tarasoff liability also appears to extend to threats against real property (39). However, the plurality opinion is most reasonably read as requiring a warning only when the threat against property also jeopardizes life.
The definition of covered mental health workers varies widely. In most states, it includes both psychiatrists and psychologists. Oklahoma imposes specific statutory Tarasoff liability only on psychologists, leaving the status of psychiatrists unclear (34). Most states include licensed social workers and some other categories of licensed mental health workers, often including psychiatric nurses and licensed marriage and family therapists. Michigan includes music therapists.
Half of the duty states—Arizona, California, Colorado, Delaware, Idaho, Kentucky, Louisiana, Maryland, Michigan, Montana, Nebraska, Ohio, Utah, and Washington—require both prompt warning to the police and reasonable attempts to warn the potential victim. The other 13 purport to countenance less thorough warnings, but a prudent practitioner should abjure such an invitation into legal peril. Pennsylvania’s, South Carolina’s, Vermont’s, and Wisconsin’s Tarasoff case law all provide that the potential victim must be warned and leave unaddressed the question of notification of the police. Indiana, Massachusetts, Mississippi, Missouri, New Hampshire, New Jersey, and Oklahoma by statute all permit a warning to be made to the police or to the potential victim. Tennessee implies the same either/or approach. Minnesota requires notification of the police only if the potential victim cannot be reached. Clearly the safest (actually and legally) and simplest course is both to notify law enforcement and to document reasonable and prompt attempts to warn the potential victim.
The permission jurisdictions
Nine states and the District of Columbia have split the baby by authorizing but not requiring a breach of psychotherapist-patient confidentiality, ostensibly leaving it to the discretion of the therapist whether to warn a third party of a patient’s threat of violence: Alaska (42), Connecticut (43), the District of Columbia (44), Florida (45), Illinois (46), New York (47), Oregon (48), Rhode Island (49), Texas (50), and West Virginia (51).
The important distinction among these jurisdictions is that in four of them—Illinois, New York, Oregon, and Texas—the therapist ostensibly has true discretion whether to disclose or not. The Illinois statute contains the unique phrase, "in the therapist’s sole discretion":
. . . communications may be disclosed . . . when, and to the extent, in the therapist’s sole discretion, disclosure is necessary to warn or protect a specific individual against whom a recipient has made a specific threat of violence . . . . (46)
How absolutely an Illinois court will construe "sole discretion" remains to be seen.
New York’s statute provides for permissive disclosure "to an endangered individual and a law enforcement agency when a treating psychiatrist or psychologist has determined that a patient or client presents a serious and imminent danger to that individual . . . and then emphasizes: "Nothing in this paragraph shall be construed to impose an obligation upon a treating psychiatrist or psychologist to release information pursuant to this paragraph" (47). As in Illinois, this limiting language has not yet been judicially construed.
Oregon’s statute, like New York’s, uses permissive language—"a clear and immediate danger to others or to society may be reported to the appropriate authority"—and then declares: "A decision not to disclose information under this subsection shall not subject the provider to any civil liability" (48). Oregon’s highest court has examined this language and has unanimously accepted its plain meaning—"may" means may, not must (7). One problem with Oregon’s statute is that it applies only to "providers," defined as public institutional mental health agencies and their staffs, leaving the law unclear for private practitioners (7).
The Texas statute unambiguously permits but does not require disclosure to law enforcement of a patient’s threats but does not require or permit disclosure to anyone else (including the victim). The Texas Supreme Court in Thapar v. Zezulka (52) emphatically interpreted the statute to mean exactly what it says, flatly rejecting any Tarasoff duty in Texas. Thus in Texas a therapist has, by dint of the law at least, no dilemma, morally or legally. If the therapist wants to warn, he or she may notify the police (not the victim). If the therapist wants to maintain the sanctity of confidentiality (or is unconvinced the threat is serious), he or she may remain silent, incurring no legal exposure.
All of the other permission jurisdictions—Alaska, Connecticut, the District of Columbia, Florida, Rhode Island, and West Virginia—leave troublesomely open the possibility that a court may engraft a duty onto permission (all the more likely, because the statutory grant of permission carries with it ipso facto immunity from liability for breach of confidentiality). Thus, prudence militates for treating these as duty states.
The anti-tarasoff jurisdiction
In Nasser v. Parker (53) the Virginia Supreme Court rejected a Tarasoff duty, where a voluntary psychiatric inpatient, under treatment for anger and depression over a romantic rejection, was allowed to leave the hospital and went to his erstwhile lover’s home and killed her (then killed himself). Although the admitting psychiatrist had treated the patient for 17 years and well knew his "history of violence toward women who rejected him and . . . that [he] recently had threatened [the victim]," the court rejected liability for failure to warn. Terming Tarasoff "unpersuasive," the court ruled:
[W]e disagree with the holding of Tarasoff that a doctor-patient relationship or a hospital-patient relationship alone is sufficient, as a matter of law, to establish a "special relation" under Restatement [of Torts] § 315(a) . . . . [T]here must be added . . . the factor . . . of taking charge of the patient . . . meaning that the doctor or hospital must be vested with a higher degree of control over the patient than exists in an ordinary doctor-patient or hospital-patient relationship before a duty arises concerning the patient’s conduct [Ref. 53, pp 505—506].
Given the facts of the case, this is a strong statement, by a unanimous court. The psychotherapist-patient relationship had gone on for 17 years; the therapist was aware of many acts (not just threats) of violence by the patient (including recently holding a gun to the victim’s head); the patient was hospitalized; there were actual, specific threats; and the victim had come out of hiding in reliance on the patient’s hospitalization. Still, there was no duty to keep the patient in the hospital or to warn the victim when the patient signed out.
In this light, it is fair to surmise that in Virginia containment (or "tak[ing] charge") on the order of involuntary hospitalization would be necessary to trigger a Tarasoff duty (e.g., in the event of elopement or discharge). Indeed, involuntary hospitalization was the predicate for liability in a recent Virginia Supreme Court case in which a female patient was sexually assaulted by a male patient known to hospital staff to be dangerous and HIV positive (54).
The no-tarasoff-law jurisdictions
Thirteen states appear to have no law, statutory or by court decision, specifically addressing a mental health clinician’s responsibilities in the event of a patient’s threat to harm a third party: Alabama (55), Arkansas, Georgia (56), Hawaii (57), Iowa (58), Kansas, Maine, Nevada, New Mexico (59), North Carolina (60), North Dakota, South Dakota, and Wyoming. Nor is there law in Puerto Rico or substantive federal law on the issue (apart from the District of Columbia statute). Thus, all U.S. territory beyond the 50 states and the District of Columbia, such as foreign military installations, can be viewed collectively as also a no-law jurisdiction. (Diversity jurisdiction cases in federal court (61) and, by virtue of the Federal Tort Claims Act, suits against federal employees for actions within state boundaries (60) are governed by state law.) The problem, of course, with a legal vacuum on so ubiquitous an issue is that the clinician is continuously in jeopardy: warn, and face breach-of-confidentiality exposure; keep silent, and risk a Tarasoff suit.