“Knowing that patients frequently deny suicidal thoughts before suicide attempt and death, clinicians should remain appropriately circumspect regarding declarations of safety when a patient recently expressed suicidal ideation, feelings of hopeless, desperation, and/or affective flooding. This does not mean we should adopt a suspicious or adversarial stance—on the contrary, curiosity, concern, and calm acceptance of the patient’s emotional and cognitive states may serve to enhance the therapeu- tic alliance, encourage the patient to directly explore her or his current distress, and aid in the accurate evaluation of current functioning.”

“Before conducting a formal suicide assessment, clinicians should conduct an introspective review of recent stressful life events facing the patient, including recent ruptures in the thera- peutic alliance, and disturbances in social relationships (Truscott, Evans, & Knish, 1999). Maintaining a therapeutic stance of curiosity and concern (while simultaneously remaining open to the possibility that an alliance rupture may be a precipitant to the crisis) is difficult to sustain when anxieties are running high; however, communicating genuine curiosity and concern about the causes for their unbearable suffering is critical.”

“When clinicians face a potential suicide crisis, they are multi- tasking and are usually in a state of heightened alert and anxiety. Under such stressful circumstances, it is easy to get swept up in personal emotional reactions and lose sight of the patient’s suffering and their efforts to communicate distress.”

My only critique involves the absence of family system context. I would have liked to see some attention to the positive role that family members can play in the assessment process. The article does mention that the quality of family relationships is an important modifiable risk factor, which should be considered and may be the focus of clinical intervention. But most of the article, and the clinical approach advocated, has a decidedly individualistic bent. The article doesn’t address how to involve friends and family members in the interview process, how to build their participation into the decision-making process, how the relational context influences decisions about how to respond to identified risk. For me this is an important gap, and a contribution that is still needed.

This gap notwithstanding, this article will instantly join the Bryan et al. means restriction article on my Top 10 Most-recommended List. In fact, the Fowler article could become my primary go-to reference when clinicians ask for reading on the subject.

Fowler, J. C. (2012). Suicide risk assessment in clinical practice: Pragmatic guidelines for imperfect assessments. Psychotherapy, 49(1), 81–90. doi:10.1037/a0026148

The authors reported unexpected findings from three separate studies that compared the efficacy of a family and non-family treatment. In brief, they found that family-level outcomes measured after applying non-family treatments didn’t just remain static (as they had expected), they actually declined. This relationship is correlational and does not necessarily mean that the treatments in question caused the decline, but the authors argue that the findings are striking enough to raise the question about whether unintended side effects psychosocial treatments should be subject to “safety monitoring” along the lines that biomedical products are. Something like a black box label: “Warning: This treatment manual may be hazardous to your family.”

In the discussion section, Szapocznik and Prado hypothesize about the systemic mechanism for the results they found:

“The family is a system that must be viewed as composed of interdependent or interrelated members…. Family members tend to develop habitual patterns of behavior over time such that each individual in the family is accustomed to act in a certain way that in turn elicits specific predictable behaviors from others. One possible hypothesis is that if an individual is changed by an intervention that is design to change individual and not help the family adjust to these changes….the family may be negatively affected….”

Nothing in these studies relate directly to suicide. But I think there are implications for how we think about intervention, especially in light of what I’ve been reflecting on lately about suicide as a family issue (see posts related to family therapy)

• Need for more systemic work on suicide. With respect to suicide, this article emphasized to me the need for greater conceptual clarity among systems thinkers about suicide in the context of the family system. We need to articulate in what ways suicidality might be a property of the system in which it resides, and what are the mechanisms by which family relationships might reduce the likelihood of suicide.

• The complexity of defining “evidence-based practice.” I’ve posted before (vis-à-vis the ambulatory redesign aspirations in our department) about my concerns that “evidence-based” can get too narrowly defined. What is evidence-based depends a lot on what evidence you look at, and, more to the point here, on what outcomes are measured in the studies that provide supporting evidence for an intervention. Given the documented importance of family functioning for long-term outcomes of many kinds, perhaps one of the criteria we should consider in evaluating the utility of a given treatment approach is its ability to promote family functioning.

• This relates to suicide because of the ways in which I have heard distressed individuals conceptualize their presenting problem. When people seek help it is usually with a functional outcome in mind, often one that has to do with their relationships. Research studies measure symptom reduction, people care about love, work, and play. In delivering a human service, we should organize ourselves in congruence with human concerns. If we organize ourselves around “reducing depression” we run the risk that our language will become reified in our practice-the result of which could be a less connected stance toward a suicidal individual who sees his relationships, finances, or health as the primary problem, not his “symptoms.” As one person I worked with paradoxically stated, “I don’t care about feeling better, I just want all of these problems to go away.”

Ideas around evidence-based practice are evolving. In our department, a vibrant conversation is underway. Simplistic views of what is evidence-based seem to be disappearing, as everyone realizes that “evidence-based” is a much broader and trickier term than we might like. Ultimately, I suspect that the way out of the dilemmas inherent in the term is for clinicians to collect evidence (in informal and formal ways) about change in their own cases. This kind of internal monitoring process will probably promote effectiveness more than selecting the right branded treatment, which may have aggregate data that allows it to be certified as “evidence based,” but which may or may not be helping the particular individual and family we’re working with.

Reference

Szapocznik, J. & Prado, G. (2007) Negative effects on family functioning from psychosocial treatments: A recommendation for expanded safety monitoring. Journal of Family Psychology. Vol 21, p. 468-478.

Related Posts:

Where’s the Family?

At the crossroads of family therapy and suicide prevention

For those who are not aware of SAD PERSONS, it is a 10-item scale to purports to screen for suicide risk. An individual is given one point for each item for which he or she screens positive:

• Sex (male)
• Age less than 19 or greater than 45 years
• Depression (patient admits to depression or decreased concentration, sleep, appetite and/or libido
• Previous suicide attempt or psychiatric care
• Excessive alcohol or drug use
• Rational thinking loss: psychosis, organic brain syndrome
• Separated, divorced, or widowed
• Organized plan or serious attempt
• No social support
• Sickness, chronic disease

The word “simple” in headline of this Psychiatric Times article linked above captures what makes the tool sound appealing, especially for the thousands of health care systems that need a quick way to respond to the JCAHO patient safety goal 15 and 15A: “The organization identifies safety risk inherent in its client populations” and “The organization identifies clients at risk for suicide” (see this .pdf for explication of these goals).

From one perspective, there is nothing wrong with using acronym like this. It can remind clinicians (assuming they can remember what all the letters stand for!) of some of the risk factors and warning signs of suicide. Who can argue with that? However, from a training and clinical perspective, there are a few problems with this approach, especially when the screen is put forward as a scored scale. Let me summarize a few of these. Note that my thinking about some of these concerns is strongly influenced by concerns articulated by my senior (and very brilliant) colleagues in email exchanges we have had about this. I don’t claim originality here, just summary:

1. The “scale” assigns risk level on the basis of a point system: A score of 1 or 2 points indicates low risk, 3-5 points indicates moderate risk, and 7-10 indicates high risk. This approach works under the assumption that these factors are equally weighted. A separated, 46-year old male with diabetes with no depression would have a higher risk level (score=4, moderate), than 40 year-old married woman with chronic depression, current hopelessness who was just released from a psychiatric hospital after a near-hanging. (score=2, low risk).
2. Having a risk “score” creates conditions for clinicians to rely on a number instead of developing an informed clinical formulation of risk.
3. The suggestion that risk for suicide can be boiled down to a single number–even for screening purposes–presents a misleading picture of the complexity phenomenon and how to think about it as a clinician.
4. The evidence that the linked article gathered does not correspond with the alluring headline, “Simple Screen Improves Suicide Risk Assessment.” Evidence reported by those who conducted the study was that, after using the computerized screen, nurses tested showed more knowledge about risk factors for suicide. Of course, knowledge about factors is a long way from demonstrating improved assessment. Obviously, the physicians who reported their study at APA the study did not write the headline. The semantic overreach of the headline speaks to the understandable desire to find easy ways of doing hard things.
5. Finally, from a training perspective, I find acronyms longer that 3 letters almost impossible to remember! SAD PERSONS particularly clumsy, and, IMHO a bit forced. “O” stands for “Organized plan or serious attempt” whereas I would probably make plan a “P” if I were trying to remember it, but of course that’s already taken by “P” for “Previous.” That often ends up being the problem with trying to make these things fit into an acronym. In a way, this gets back to the theme I’ve been harping on lately in my posts about teaching and training about needing a basic-science base about how clinicians learn, remember, and use principles or practices we learn. I’d imagine an expert in human memory could graph the inverse relationship between recall rate and number of letters in an acronym–add to that the need to recall these letters that signify words or concepts with high emotional impact.

In summary, while SAD PERSONS may be helpful to some people as a tool for remembering risk factors, it has some serious limitations as risk assessment “scale” and probably as a mnemonic.

Really, there is a “basic science” set of questions about learning and the clinician mind that gets skipped over when we do the necessary and important work of evaluating educational interventions.

Thanks to a post on PsychNews, I came across this interesting article in Cognitive Daily that attempts to provide some explanations for why visual memories are often so vivid.  One of the take-home points of the study cited in the article is that the vividness of visual memory is directly related to the duration of viewing.    This is unsurprising in some ways, but it supports the educational strategy of using one or two maps or other graphics (rather than a multitude of Powerpoint slides or text handouts) to teach about a clinical concept like risk assessment.   Participants in my trainings, for example, view one map (whose branches I dynamically hid and show) for nearly the entire presentation.

These little bits of basic science evidence remind me, once again, that we pay too little attention to the evidence base of our teaching techniques.   It is well and good to decideto pursue evidence-based interventions and therapuetics (EBIT, as we call it around here), but what is often missing (besides a coherent notion of what constitutes evidence–a topic for another day) is an evidence-based way of disseminating evidence-based practice to clinicians.

Related Posts:

Visual maps and guides in high stress situations

Mindmapping coping strategies

Evidence for visually different presentation format

Tech tools for clinical thinking and training

Researchers at Harvard/Mass General have contributed some interesting data to the conversation about suicidality and antidepressant treatment. Roy Perlis and colleagues examined available DNA info on patients who had new onset suicidal thoughts after starting drug therapy, and found an interaction effect (sex x genetic variation) that suggested that the men with the genetic variation were at greater risk of suicidal thoughts. The article by Roy Perlis and colleagues appeared in the most recent issue of Archives of General Psychiatry. For a lay description of the study see this article in Medical News Today.

Because of the nature of the sample and the narrow scope of the findings, this study contributes most to the understanding about a possible link between suicidality and antidepressant use, about which I have posted before. But my main interest in it here relates to what new findings (especially those with a strong biomedical basis) mean to clinicians and how we think about risk assessment.

If replicated and expanded findings like these might lead to more targeted approaches to suicidality (and probably psychopharmacology for depression). However, for the clinician faced with an at-risk person in treatment, each new discovery will be merely one factor to synthesize into an overall, well-constructed risk formulation. It is doubtful we’ll ever get to the point where a single data point (genetic or otherwise) will be strong enough to predict risk by itself. We’ll always need good, old fashioned, clear-headed, complexity-embracing clinical judgment to discern risk. In fact, new findings related to risk (especially complex ones like those in the Perlis study) point out the importance of having a sturdy framework for thinking through risk assessment. A systematic framework allows the clinician to incorporate new findings into thinking and practice.

The need for a framework for thinking through risk may seem too obvious to mention. But consider what most training in suicide risk assessment consists of. Nearly every clinician I’ve spoken to (including myself) learned to assess suicide by asking a few basic questions about suicidal thinking, plan, intent, and ability to agree to stay safe. Most of us were not taught a “framework at all.” We did not learn to assess for suicide risk the way we do other clinical issues–via systematic assessment and synthesis of multiple data points. In my opinion, that is why many clinicians I talk to feel so unsatisfied with the experience of working with people who voice suicidality.

All that to say…we’re likely to see more genetic links with suicide risk. My goal is to be prepared to assimilate new findings within:

• a compassionate and autonomy-respecting approach to gathering data (of all kinds) and intervening
• a systematic way of thinking through multiple risk factors to arrive at a formulation
• a coherent and predictable format for documenting and responding to risk

The study does not study who is not getting medication that might have in the past–just that fewer kids are getting the medicine.  Whether that is ultimately good or bad remains to be seen.

Kelly Posner, who headed the FDA/Columbia Reclassification study, spoke at our Department Grand Rounds last month and lead a seminar with a smaller group of us about the reclassification scheme.  (The reclassification project looked in detail at reported adverse events that started this controversy.)   She clearly thought the consequences of the warning would be mostly negative.   She shared her concern (which has been stated by others as well) that the unintended consequence of the black box warning will be fewer youth treated for depression and more deaths by suicide, as a result.  From what I have read, that remains an empirical question and one that will require careful interpretation of data before inferring causality. What is clear is that the black box warning was probably based more on political and emotional concerns than on science (see also Marshall, Posner, and Greenhill, 2006), and that the risk of untreated depression is probably greater than the risk of adverse events from treatment.

[mounting soapbox]  That said…when this discussion comes up I think it is important to add psychotherapy and counseling to the landscape.  Untreated depression can be deadly, but that doesn’t mean the treatment has to be medication.  [coming down off soapbox]

What are the implications for risk assessment?   I’m still trying to understand that.  How I am guiding our clinicians at this point is this:   When conducting a risk assessment of an adolescent or young adult in the first weeks following initiation of antidepressants, we need to note that antidepressants have been recently started.  But rather than name this as a “risk factor” we would do better to note how the medication response will be monitored and indicate the risk-related symptoms that are being targeted by the medication.  That is, connect the medication to the risk factors it is designed to reduce, more than to the risk it might carry.

References

1. Marshall, R. D., Posner, K., & Greenhill, L. (2006). Risk Perception Research and the Black Box Warning for SSRIs in Children. Journal of the American Academy of Child & Adolescent Psychiatry, 45(7), 765.
2.  Nemeroff, C.B. et al. (2007. Impact of Publicity Concerning Pediatric Suicidality Data on Physician Practice Patterns in the United States.  Archives of General Psychiatry, 64:466-472.

In KM world, what mental health clinicians call a “treatment team” could be considered a Community of Practice. There are many definitions of this term and treatment teams fit some more than others. But Dave Snowden is clear in the videotaped discussion I pointed to yesterday that one of the failures of contemporary knowledge management is the inability to promote fruitful communities of practice. Snowden argues that organizations make the mistake of trying to organize communities of learning and practice using language, structures, and concepts that are not “naturalistic.” That is, we ignore the processes by which people naturally come together to form knowledge-sharing communities and either over-organize (imposing a hierarchy and structure that we think will promote good functioning, but ultimately stifle innovation) or wrongly organize (bringing people together around a concept or structure that does not promote natural affinity).

Our organization is considering a redesign of our ambulatory service into diagnostic-based treatment teams (e.g. “Comorbid depression team”). The aim is to have well-functioning teams that promote evidenced-based practice. As I listened to Dave Snowden talk about communities of practice, I coudn’t help but think that organizing this way has hints of the kinds of non-naturalistic grouping that Snowden warns against. The intent is good and the organizational principle makes sense on the surface, but grouping clinicans by the DSM diagnosis of their patients has the potential to be structure-rich, story-poor, and human-factor-ignoring.

One of the principal reasons stated for considering organizing by diagnosis is that the research literature about effective treatments is organized by diagnosis. It is an “evidence-based” decision. However, evidence associated with a particular epistemology (categorical psychopathology) is privileged over that of other epistemologies like cognitive science, organizational behavior, human factors research, and systems theory are ignored. Just because treatment studies organize patients into diagnostic groups, doesn’t mean that human clinicians will work most effectively with the human problems and stories we see by grouping ourselves by our patients’ Axis I diagnosis.

What would be a more narrative-rich way or organizing ourselves? Well, if we think of theoretical paradigm as a narrative, than perhaps that could be starting point. Or perhaps provide freedom for people to organize themselves into natural groupings. Or maybe there’s a way of listening to clinicians and patients’ stories about themselves and seeing trends and themes that we don’t now see. It would take time and a new set of methods, those more akin to what Snowden promotes, to discover these themes. But we’re taking time and energy either way. I’m not sure, to be honest, but I think the principle Snowden promotes is a good one: don’t impose a community of practice based on a predetermined epistemology, especially one that is reductionist and devoid of narrative…rather, look at how productive human networks form naturally and spend your time and energy discerning the conditions in which these can develop.

For some time, I have been following the work and blog of Dave Snowden, founder of Cognitive Edge. Snowden is an scientist, theorist, and organizational consultant at the cutting edge of the Knowledge Management (KM) field. Or perhaps it would be more accurate to say that Snowden is a pioneer and visionary who is try to push KM to an entirely different dimension (call it KM 2.0). I must admit that I am still trying to get a handle on Snowden’s thinking (it’s broader and more complex than I can yet grasp), but one of the most interesting things to me about his work is that he emphasizes narrative (versus purely numerical) approach to “sensemaking.” Snowden and others of his ilk argue that you can learn more useful information, detect more weak signals, capture trends earlier through gathering stories than you can by gathering numbers. Stories show emerging trends. Numbers tell you what has already happened.   (For a popular version of this argument see Lori Silverman’s provocatively titled book “Wake me up when the data are over: How Organizations Use Stories to Drive Results“)

Snowden and another KM guru, Gary Klein, were recently videotaped discussing the methodology (and software) that the Government of Singapore has developed to help them detect terrorist risk, the Risk Assessment and Horizon Scanning (RAHS) system. I found their videotaped discussion fascinating, especially Snowden’s critique on the failures of knowledge management (2nd clip on the page). I don’t know enough to understand the differences between the perspectives Klein and Snowden offer (and, can’t in fact follow all of what either one says), but I listened with great interest to their perspective on how one approaches information-gathering, sensemaking, and decision-making in an uncertain, unpredictable, and unstable environment.

Obviously, clinical sensemaking and decision-making is quite different from government counter-terrorism operations. But I could not help but think of parallels, especially for assessment of suicide risk. Here are a couple of developing (and somewhat random) reflections I had:

1. We know about statistical risk factors, but how do we do sensemaking with a particular person’s set of stories. Clinicians have access to rich narratives, but we generally lack methodologies and technologies for sensemaking that retains complexity and guides decision making.
2. Traditional documentation (the principal knowledge management system for clinical care), including the diagnostic evaluation reports, usually flatten the richness of stories (by design) into a language that is more technical, linear, and sterile than real life. We usually don’t capture stories on their own or track raw data, but rather we move quickly to interpretation and synthesis.
3. I noted in a previous post that I use mindmapping to teach about suicide risk. In that post, I suggested one benefit might be “it helps to be able to visualize connections between concepts on a map because it makes complex material more accessible.” In light of what I’m learning from Snowden and KM, I wonder if mindmapping also facilitates sensemaking from narratives better because it is nonlinear and attempts to replicate connections in human thought patterns.
4. Apropos of my previous post, Where’s the family?…family therapy offers an opportunity for gathering anecdotes from multiple perspectives. Snowden has a KM exercise called “Anecdote Circles,” which he uses to help organizations gather information through story. The techniques he uses would be interesting to apply to a family, and to gathering information from family members about suicide risk. This kind of raw data is not available without family members.
5. Our models and language around risk assessment needs to better reflect how fluid and unstable the phenomena of risk and suicidality really are. The act of suicide is a momentary coalescing of a multitude of snippets and anecdotes and narratives. Reading retrospective case studies of people who died by suicide makes that really clear–all of what we categorize as “risk” comes together in a certain way at a certain point in time. As one of my mentors pointed out to me last week, we can “predict” suicide retrospectively, but it is almost impossible to detect prospectively.  As clinicians we want to be sensitive to the snippets, so that we can scan the horizon (a la RAHS) and sense emerging trends, far before the data ever catches up.

As I warned in the beginning, these thoughts are pretty raw, but I’m interested in exploring this intersection more.