Rebooting for Whom?

Abstract

We share Kazdin and Blase’s (2011) sense of urgency about finding better ways to reduce the burden of mental illness. Although effective psychosocial treatments exist, they do not often reach the patients who need them most. Kazdin and Blase’s portfolio approach aims to cast a wider net through increased use of technology, media, self-help, nonprofessional providers, and collaborations with other disciplines. It is unclear, however, whether reaching more people would suffice to reduce the burden of mental illness, much less offset the small effect sizes of simplified, scaled-down interventions such a portfolio approach would likely entail. We focus here on an underdeveloped theme in Kazdin and Blase’s essay—that bending the curve of mental illness will require better knowledge of for whom simplified intervention and prevention strategies will suffice and for whom more intensive intervention is necessary. Such “for whom” questions deserve a central place on the national research agenda as we move toward individualized or personalized health care. In the absence of such knowledge, we risk treatment decisions guided by accessibility to resources rather than patient needs—the very problem Kazdin and Blase aim to solve.

Keywords

treatment mental illness personalized intervention

We share Kazdin and Blase’s (2011) sense of urgency about finding better ways to reduce the burden of mental illness. Although effective psychosocial treatments exist, they often do not reach the patients who need them most. Kazdin and Blase’s portfolio approach aims to cast a wider net of empirically supported interventions through increased use of technology, media, self-help, nonprofessional providers, and collaborations with other disciplines. The authors’ off-the-shelf examples provide an intriguing glimpse of what this could look like and usefully bridge the artificial boundary between prevention and treatment. Because efficacious treatments will usually require simplification for portfolio dissemination, we especially appreciate Kazdin and Blase’s cautionary note about the importance of understanding how these treatments work. Without such knowledge, an abbreviated (though more accessible) intervention could unwittingly sacrifice essential mechanisms of change.

Although reaching more people is a laudable aim, it is not clear whether this by itself will reduce the burden of mental illness, much less offset the small effect sizes of simplified, scaled-down interventions such a portfolio approach would likely entail. It is interesting that the same National Comorbidity Study the authors cite for burden statistics demonstrated a 50% increase in treatment between 1991 and 2001 without any decrease in prevalence or morbidity (Kessler, Berglunk, Borges, Nock, &Wang, 2005; Kessler et al., 2005). Apparently, more treatment does not necessarily mean less burden, especially if the treatment is insufficient or inappropriate.

In this commentary, we focus mainly on an underdeveloped theme in Kazdin & Blase’s essay—that bending the curve of mental illness will require better knowledge of for whom simplified intervention and prevention strategies will suffice and for whom more intensive intervention is necessary. In our view such “for whom” questions deserve a more central place on the national research agenda as we move toward individualized or personalized health care. In the absence of such knowledge, we risk treatment decisions guided by accessibility to resources rather than patient needs—the very problem Kazdin and Blase aim to solve.

Which Treatments for Whom?

As the cornerstone of personalized intervention, research on prospective treatment moderators (what works for whom) necessarily cuts across a wide range of case and treatment characteristics. The basic question in this Attribute × Treatment Interaction (ATI) paradigm is which cases characteristics moderate (interact with) which treatment conditions to predict clinical outcomes. Because the most conspicuous case-level moderators—psychiatric diagnosis—have not proved terribly useful for guiding psychosocial intervention, the search for meaningful moderators has recently expanded to include such diverse factors as current and historical problem severity, cognitive processes, and characteristics of the family social environment. There is equal or greater diversity on the treatment side, in which Kazdin and Blase’s big-menu portfolio approach now ups the ante on answering “for whom” questions to guide selection of interventions. And this is not to mention an even-more pressing ATI question where serious mental illness is concerned—namely, for whom will combined psychosocial intervention and pharmacotherapy be more beneficial than either modality alone (Nemeroff et al., 2003).

Psychosocial ATI research is not new, dating back at least to Cronbach and Snow (1977), and its history has included both excitement and disappointments. For example, the hope engendered by early ATI findings in psychotherapy research (e.g., Beutler, 1991; Shoham-Salomon & Hannah, 1991) was dampened by the high-profile failure of Project Match Research Group (1997) to find significant moderators of alcoholism treatments. Factors limiting the yield of ATI research include investigation of post hoc, hard-to-replicate moderators; moderators unrelated to theory-derived, hypothesized mechanisms of change; underpowered tests of moderation; comparisons among similar or overlapping treatments (as in Project MATCH), which limits moderator detection; and unbalanced measurement of A and T variables, in which researchers assess case attributes (As) in precise detail while documenting treatments (and treatment fidelity) only grossly. Shulman (1981) noted years ago this tendency to measure As with micrometers and Ts with divining rods.

Despite this checkered history, recent years have seen a burgeoning interest in treatment moderators, and preliminary findings highlight the importance of pursuing this line of research more vigorously than we have so far. Thus, regarding cognitive treatments for depression, there is good evidence that optimal treatment selection depends on factors such as age of illness onset (Jarrett et al., 2001), current and historical symptom severity (Bockting et al., 2005; Fournier et al., 2011), and patient preference for pharmacological or psychosocial intervention (Kocsis et al., 2009). Similarly, in the schizophrenia domain, a patient’s age at treatment initiation appears to moderate the effects of cognitive rehabilitation (CR) interventions, such that younger patients (< 40) benefit more than older patients from receiving CR (McGurk & Mueser, 2008; Wykes et al., 2009).

A more promising class of potential treatment moderators relates to theory-derived mechanisms of problem formation or problem maintenance. For example, there is evidence that baseline levels of maladaptive cognitions hypothesized to maintain depressive symptoms serve to moderate the success of cognitive-behavior therapies relative to control conditions (Hollon et al., 2005) and that theory-relevant aspects of a patient’s trauma history are associated with better response to a variant of cognitive behavior therapy than to antidepressants (Nemeroff et al., 2003). Similarly, looking beyond the patient, the quality of family or couple communication appears to moderate the success of family- and couple-focused treatments (Miklowitz et al., 2009; Shoham, Rohrbaugh, Stickle, & Jacob, 1998).

Despite these advances, we are still a long way from having a sufficient body of evidence to guide psychosocial variants of personalized medicine based on treatment-matching algorithms. Part of the problem is that current evidence based on statistical significance often lacks clinical significance. Our search for moderators is still very much in a discovery phase, and the clinical value of any discoveries will ultimately need to be tested in clinical trials with patients stratified on putative moderator variables.

Where can we expect to discover the most promising moderators? The horizon includes several promising developments we think are worth mentioning. One is the Research Domain Criteria (RDoC) project at the National Institute of Mental Health (NIMH; Insel & Cuthbert, 2009; Sanislow et al., 2011), which is attempting to ground the patient attribute (A) side of the ATI equation in underlying neurobiological dimensions of psychopathology. Given the high variability in pathophysiology among patients diagnosed with the same disorder (as determined by the Diagnostic and Statistical Manual of Mental Disorders), variability in treatment response among patients similarly classified is not surprising. The science-based, bottom-up RDoC approach to mental disorders aims to establish validity in ways that may ultimately align better with treatment response.

Other promising developments are methodological. For example, new applications of so-called adaptive randomized designs can illuminate the most efficient and effective sequencing of several interventions rather than just one—as when nonresponders to a first-line intervention receive more intensive treatment in a stepped-care framework (Collins, Dziak, & Li, 2009; Collins, Murphy, & Srecher, 2005). These adaptive designs are also well suited for testing treatment moderators, including for whom starting with a “light” version of a given treatment might ultimately prove iatrogenic or counterproductive. One could approach this question either sequentially, as in adaptive design research, or simultaneously, by considering moderators of pared-down interventions in the population more broadly.

A related innovation is the Kraemer, Wilson, Firburn, and Agras (2002) approach to creating moderator profiles. Whereas traditional ATI research employing group factorial designs typically stumbles on the prospect of multiple interacting moderators—Cronbach’s (1975, p. 119) “hall of mirrors that extends to infinity”—the Kraemer et al. method allows for testing multiple moderators and identifying the strongest ones via a probability index of replication (p_rep) instead of the traditional significance level. Thus, a p_rep ≤ .90 indicates that there is at least a 90% chance to replicate the moderator’s effect with a new sample from the same population, provided that the effect size of the moderator is at least as strong as the effect size of the treatment. This then leads to hypothesis-testing studies in which prospective, empirically based moderators serve as stratification variables.

In a notable application of this moderator-profile approach, Vittengl, Clark, and Jarrett (2010) investigated which responders require continuation-phase cognitive therapy (CT) to achieve stable remission of depressive symptoms and which could sustain positive treatment outcomes without further intervention. Based on examining a range of plausible and partially overlapping moderators in a hypothesis-generation framework, Vittengl et al. found that a profile including younger current age and younger age of onset in combination with high social inhibition and emotional detachment served to discriminate the criterion groups. They are now seeking to replicate this profile result in a prospective, hypothesis-testing design that could solidify an empirical basis for providing continuation CT. The investigators speculate that continuation CT may be “too little, too late” for some older patients, which brings us to the crucial role of “for whom” research in prevention.

Prevention for Whom?

Kazdin and Blase note that “the portfolio idea would be beneficial for conceptualizing the task of prevention because it begins with who ought to be reached in the population, what interventions are likely to accomplish that for various groups, and what the effects are” (p. 28). This important point deserves further development: On one hand, the potential payoffs from well-targeted preventive intervention could be enormous. On the other hand, without better understanding of who benefits from which prevention strategies, we risk shooting in the dark and hitting targets indiscriminately, which could be costly and even iatrogenic. Most important, we need to understand risk and resilience at an individual level. Despite some good leads on risk factors from both the nature (genetics) and nurture (experience) sides of the mental illness equation, we do not yet have biomarkers or psychological attributes with high predictability for any individual.

Like physical illnesses, most mental disorders have a clear developmental trajectory. It is disconcerting in this respect that treatment for mental disorders begins on average 11 years after problem onset (Wang et al., 2005). The field of medicine has rarely reduced the burden of any illness when initial intervention takes place so long after onset. Observable symptoms of mental illness, possibly reflecting underlying biological processes, may have a relatively long latency period. Add to this the long delay for treatment, and prospects for reducing the burden of mental illness appear even more daunting. At the risk of medicine envy, it is worth noting that early detection of specific risk factors coupled with “for whom” risk factor reduction interventions has enabled cardiology to realize a 60% reduction in mortality from coronary artery disease. Imagine what just a fraction of that accomplishment could mean for mental health.

A promising line of prevention research involves early stage intervention with major mental illnesses such as schizophrenia and bipolar disorder. For example, a combination of features now allows detecting the prodrome of schizophrenia with more than 80% accuracy in adolescents who have not yet become psychotic (Cannon et al., 2008). Further along in the prevention spectrum, characteristics of the family environment moderate how adolescents with early-stage bipolar disorder respond to psychosocial interventions combined with medication; here a family-focused approach appears to be most beneficial—and perhaps essential—for families showing high “expressed emotion” (criticism, hostility, and emotional overinvolvement) in relation to the patient (Miklowitz et al., 2009).

This approach of personalized and preemptive interventions is a major focus of the NIMH Strategic Plan (www.nimh.nih.gov/about/strategic-planning-reports/index.shtml). Beyond studies of the prodrome of schizophrenia and bipolar disorder, we have launched a broad effort on biomarkers that could serve as moderators or predictors of response. One such study, EMBARC (which stands for Establishing Moderators/Mediators for a Biosignature of Antidepressant Response in Clinical Care), is combining genomics, imaging, quantitative EEG, and cognitive measures to develop a profile or biosignature of antidepressant response. In another effort, the Study to Assess Risk and Resilience in Soldiers, we are looking for predictors of posttraumatic stress disorder and depression in soldiers. And in another, we are following younger siblings of children with autism to identify the earliest signs of this disorder. We hope that the identification of such risk factors will translate into treatment moderators, thus leading to better targeted interventions.

Technology for Whom?

Technology-assisted treatments are surely here to stay, but these too need better targeting to be efficient. Such interventions, sometimes referred to as e-Health (Baker, McFall, & Shoham, 2008), are highly replicable and portable and thus easy to disseminate. Computer-based e-Health interventions have the additional virtue of permitting exposure to diverse realistic contexts achieved via virtual reality capabilities (Bordnick et al., 2008), and they have the potential to reduce utilization of more expensive health care options (Boberg et al., 1995). Because e-Health interventions allow for some tailoring based on a variety of patient characteristics (Strecher et al., 2005), we were surprised that Kazdin and Blase do not emphasize this.

Although technology could prove a game changer, it may also have some unintended consequences. As commentators like Abraham Verghese (2011) have pointed out, the complaints we hear from patients, family, and friends are rarely about the dearth of technology but about its excesses, turning patients into “i-patients.” In the “which treatments for whom” framework of ATI research, we know virtually nothing about the treatments for which, or the patients for whom, a human relationship or therapeutic alliance is essential to productive behavior change. Without such knowledge, even in a best-case scenario, e-Health will to some extent require shooting in the dark. A worst-case scenario is that e-Health interventions could “spend out” some of our most effective techniques, rendering them less amenable to subsequent, face-to-face intervention.

It bears repeating that technology-based intervention portfolios imply simplifying and abbreviating treatments that were empirically, even experimentally, supported in their original format. Yet, by Kazdin’s (2007) own account, the field does not know much (and certainly not enough) about how multicomponent or even simple psychosocial interventions actually work. Apart from the problem of abbreviated (if more accessible) interventions sacrificing essential mechanisms of change, we worry that pared-down portfolio interventions gaining premature adoption in community settings will yield effects no larger than those for “treatment as usual,” which are very small. The e-Health picture may well improve as additional efficacy and effectiveness data accumulate, but in our view the “which treatment for whom” question will not soon go away.

On balance, Kazdin and Blase do the field an important service by highlighting the diverse ways in which technology could enhance the world of psychotherapy. We are hopeful that the increased access and increased flexibility of this approach will deliver improved outcomes. At the same time, we would caution that technology is a tool, not an answer: With a better understanding of how and for whom technology-assisted treatments work (see Amir, Taylor, & Donohue, in press, for a promising example of this), mental health professionals should be in a better position to personalize psychosocial intervention and ultimately reduce the burden of mental illness.

Footnotes

The author(s)declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

The author(s) received no financial support for the research, authorship, and/or publication of this article.

References

Amir

Taylor

C. T.

Donohue

M. C.

(in press). Predictors of response to an attention modification program in generalized social phobia. Journal of Consulting and Clinical Psychology.

Baker

McFall

Shoham

(2008). The current status and future of clinical psychology: Towards a scientifically principled approach. Psychological Science in the Public Interest, 9, 67–103.

Beutler

L. E.

(1991). Have all won and must all have prizes? Revisiting Luborsky et al.’s verdict. Journal of Consulting and Clinical Psychology, 59, 226–232.

Boberg

E. W.

Gustafson

D. H.

Hawkins

R. P.

Peressini

Chan

Bricker

(1995). Development, acceptance and use patterns of a computer-based education and social support system for people living with AIDS/HIV infection. Computers in Human Behavior, 11, 289–312.

Bockting

C. L. H.

Schene

A. H.

Spinhoven

Koeter

M. W.

Wouters

L. F.

Huyser

Kamphuis

J. H.

(2005). Preventing relapse/recurrence in recurrent depression with cognitive therapy: A randomized controlled trial. Journal of Consulting and Clinical Psychology, 73, 647–657.

Bordnick

P. S.

Traylor

Copp

H. L.

Graap

K. M.

Carter

Ferrer

Walton

A. P.

(2008). Assessing reactivity to virtual reality alcohol based cues. Addictive Behaviors, 33, 743–756.

Cannon

T. D.

Cadenhead

Corhblatt

Woods

S. W.

Addington

Walker

. . . Heinssen

(2008). Prediction of psychosis in youth at high clinical risk: A multisite longitudinal study in North America. Archives of General Psychiatry, 65, 28–37.

Collins

L. A.

Dziak

J. J.

(2009). Design of experiments with multiple independent variables: A resource management perspective on complete and reduced factorial designs. Psychological Methods, 14, 202–224.

Collins

L. A.

Murphy

S. A.

Srecher

(2005). The multiphase optimization strategy (MOST) and the sequential multiple assignment randomized trail (SMART): New methods for more potent e-Health interventions. American Journal of Preventive Medicine, 32, S112–S118

10.

Cronbach

L. J.

(1975). Beyond the two disciplines of scientific psychology. American Psychologist, 30, 116–126.

11.

Cronbach

L. J.

Snow

R. E.

(1977). Aptitude and instructional methods: A handbook for research on interactions. New York, NY: Irvington.

12.

Fournier

J. C.

DeRubeis

R. J.

Hollon

S. D.

Dimidjian

Amsterdam

J. D.

Shelton

R. C.

Fawcett

(2011). Antidepressant drug effects and depression severity: A patient-level meta-analysis. Journal of the American Medical Association, 301, 47–53.

13.

Hollon

S. D.

DeRubeis

R. J.

Shelton

R. C.

Amsterdam

J. D.

Salomon

R. M.

O’Reardon

J. P.

. . . Gallop

(2005). Prevention of relapse following cognitive therapy vs. medication in moderate to severe depression. Archives of General Psychiatry, 62, 417–422.

14.

Insel

T. R.

Cuthbert

B. N.

(2009). Endophenotypes: Bridging genomic complexity and disorder heterogeneity. Biological Psychiatry, 66, 988–989.

15.

Jarrett

R. B.

Kraft

Doyle

Foster

B. M.

Eaves

G. G.

Silver

P. C.

(2001). Preventing recurrent depression using cognitive therapy with and without a continuation phase. Archives of General Psychiatry, 58, 381–388.

16.

Kazdin

A. E.

(2007). Mediators and mechanisms of change in psychotherapy research. Annual Review of Clinical Psychology, 3, 1–27.

17.

Kazdin

A. E.

Blase

S. L.

(2011). Rebooting psychotherapy research and practice to reduce the burden of mental illness. Perspectives on Psychological Science, 6, 21–37.

18.

Kessler

R. C.

Berglunk

Borges

Nock

Wang

P. S.

(2005). Trends in suicide ideation, plans, gestures, and attempts in the United States, 1990–1992 to 2001–2003. Journal of the American Medical Association, 293, 2487–2495.

19.

Kessler

R. C.

Demler

Frank

R. G.

Olfson

Pincus

H. A.

Walters

E. E.

. . . Zaslavsky

A. M.

(2005). Prevalence and treatment of mental disorders, 1990 to 2003. New England Journal of Medicine, 352, 2515–2523.

20.

Kocsis

J. H.

Leon

A. C.

Markowitz

J. C.

Manber

Arnow

Klein

D. N.

Thease

M. E.

(2009). Patient preference as a moderator of outcome for chronic forms of major depressive disorder treated with Nefazodone, cognitive behavioral analysis system of psychotherapy, or their combination. Journal of Clinical Psychiatry, 70, 354–361.

21.

Kraemer

H. C.

Wilson

G. T.

Firburn

C. G.

Agras

W. S.

(2002). Mediators and moderators of treatment effects in randomized clinical trials. Archives of General Psychiatry, 59, 877–884.

22.

McGurk

S. R.

Mueser

(2008). Response to cognitive rehabilitation in older versus younger persons with severe mental illness. American Journal of Psychiatric Rehabilitation, 11, 90–105.

23.

Miklowitz

D. J.

Axelson

D. A.

George

E. L.

Taylor

D. O.

Schneck

C. D.

Sullivan

. . . Birmaher

(2009). Expressed emotion moderates the effects of family-focused treatment for bipolar adolescents. Journal of the American Academy of Child and Adolescent Psychiatry, 48, 643–651.

24.

Nemeroff

C. G.

Heim

C. M.

Thase

M. E.

Klein

D. N.

Rush

A. J.

Schatzberg

A. F.

. . . Keller

M. B.

(2003). Differential responses to psychotherapy versus pharmacotherapy in patients with chronic forms of major depression and childhood trauma. Proceedings of the National Academy of Sciences, USA, 100, 14293–14296.

25.

Project Match Research Group. (1997). Matching alcoholism treatments to client heterogeneity: Project MATCH posttreatment drinking outcomes. Journal of Studies on Alcohol, 58, 7–29.

26.

Sanislow

C. A.

Pine

D. S.

Quinn

K. J.

Kozak

M. J.

Garvey

M. A.

Heinssen

R. K.

. . . Cuthbert

B. N.

(2011). Developing constructs for psychopathology research: Research domain criteria. Journal of Abnormal Psychology, 119, 631–639.

27.

Shoham

Rohrbaugh

M. J.

Stickle

T. R.

Jacob

(1998). Demand-withdraw couple interaction moderates retention in cognitive-behavioral vs. family-systems treatments for alcoholism. Journal of Family Psychology, 12, 557–577.

28.

Shoham-Salomon

Hannah

M. T.

(1991). Client-treatment interactions in the study of differential change processes. Journal of Consulting and Clinical Psychology, 59, 217–225.

29.

Shulman

L. S.

, (1981, August). Educational psychology returns to school ( Stanley Hall

Series). Paper presented at the 89th Annual Convention of the American Psychological Association, Los Angeles, CA.

30.

Strecher

V. J.

Marcus

Bishop

Fleisher

Stengle

Levinson

. . . Nowak

(2005). A randomized controlled trial of multiple tailored messages for smoking cessation among callers to the cancer information service. Journal of Health Communication, 10, 105–118.

31.

Verghese

(2011, February 26). Treat the patient, not the CT scan. The New York Times. Retrieved from http://www.nytimes.com/2011/02/27/opinion/27verghese.html?pagewanted=all

32.

Vittengl

J. R.

Clark

L. A.

Jarrett

R. B.

(2010). Moderators of continuation phase cognitive therapy’s effect on relapse, recurrence, remission, and recovery from depression. Behaviour Research and Therapy, 48, 449–458.

33.

Wang

P. S.

Berglund

Olfson

Pincus

H. A.

Wells

K. B.

Kessler

R. C.

(2005). Failure and delay in initial treatment contact after first onset of mental disorders in the National Comorbidity Survey Replication. Archives of General Psychiatry, 62, 603–613.

34.

Wykes

Reeder

Landau

Matthiasson

Haworth

Hutchinson

(2009). Does age matter? Effects of cognitive rehabilitation across the age span. Schizophrenia Research, 113, 252–258.