Issues Related to Combining Risk Factor Reduction and Clinical Treatment for Eating Disorders in Defined Populations C. Barr Taylor, MD Rebecca P. Cameron, PhD Michelle G. Newman, PhD Juliane Junge, Dipl-Psych Abstract Population-based psychotherapy considers the provision of services to a population at risk for or already affected with a disease or disorder. Using existing data on prevalence, incidence, risk factors, and interventions (both preventive and clinical)for eating disorders (anorexia excluded), this article examines issues related to integrating and providing risk reduction and treatment to a population of female college students. Population-based psychotherapy models have important implications for the provision of services and for future directions in research on eating and other types of mental health disorders, but the assumptions need to be carefully examined. Studies that provide data combining population-based risk factor reduction and clinical treatment are needed to advance this field.

Introduction Population-based psychotherapy considers the provision of services to a population at risk for or already affected with a disease or disorder, l Thus, it incorporates elements of prevention, case identification, and treatment. Population-based psychotherapy uses data from epidemiology, risk factor identification and modification, and clinical treatment outcome research to determine prevenfive/treatment approaches for the targeted population. In this article, the links among risk factors, subclinical and clinical disorders, screening, stepped care, and costs are considered. A model is proposed for heuristic p u r p o s e s - - f o r considering how a prevention/clinical eating disorders intervention might be provided to a defined population such as an insured group or a college population and how different assumptions might affect different cost-benefit models.

Application to a College Population Ideally, in developing a population-based psychotherapy intervention for a particular disorder, the prevalence, incidence, and risk factors of that disorder would be well known. Also, there would be

Address correspondenceto C. Barr Taylor, MD, Professor, Department of Psychiatry and Behavioral Sciences, Stanford University School of Medicine, Stanford, CA 94305-5722; e-mail: [email protected]. Rebecca P. Cameron, PhD, is an Assistant Professor, Department of Psychology,California State University, Sacramento. Michelle G. Newman, PhD, is an Assistant Professor, Department of Psychology,Pennsylvania State University. Juliane Junge. Dipl-Psych, is a Research Associate, Department of Clinical Psychology and Psychotherapy, Dresden University of Technology.

JournalofBehavioralHealth Services&Research,2002, 29(1), 81-90. © 2002 NationalCouncil for CommunityBehavioral Healthcare.

Issues Related to Combining

TAYLORet al.

81

Table 1 Components of population-based psychotherapy 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

Reliable diagnoses Standardized, reliable outcome measures Prevalence and incidence known Potentially modifiable risk factors indentified Demonstration that such risk factors can be modified Demonstration that modification of risk factors leads to a reduction in the incidence of the disorder Effective clinical interventions Methodology for monitoring of outcomes Identification of mediating and moderating factors and subpopulations with special needs Short- and long-term cost data as they pertain to the individual, society, and health care system

evidence that the risk factors can be modified, that modification of risk factors leads to a reduction in the incidence of the disorder, and that effective treatments are available for individuals with the disorder. The development of the model also requires reliable diagnoses and measurable outcomes. These and other features of population-based psychotherapy are summarized in Table 1. The extensive research literature on eating disorders, including treatment studies and recent work on identification of risk factors, allows us to model a population-based approach to these problems.

Incidence and prevalence Extensive epidemiologic work has found that approximately 1% to 3% of the young adult, female population suffers from "full-syndrome eating disorders"; that is, they meet the Diagnostic and Statistical Manual, revised third (DSM-III-R) or fourth edition (DSM-IV) diagnostic . . . .criteria-~ for an eating disorder. 3,4 However, many more young women suffer from "partial syndrome" or subclinical eating disorders; that is, they meet many but not all of the criteria for a full-syndrome eating disorder. For the purposes of this discussion, individuals meeting DSM-IV criteria for partial, subclinical and full-syndrome eating disorders will be considered clinical cases.

Risk factors In recent years, a number of studies using cross-sectional samples and clinical populations have identified various cultural, familial, and personal factors that are associated with eating disorders. 5 More important, several longitudinal studies have shown that in adolescents excessive weight and shape concerns, as well as the drive for thinness, predict the onset of subclinical and clinical eating disorders. 6~7 By combining risk factor, prevalence, and incidence data from older adolescents and college students, it is possible to stratify a college-age female population by both risk and prevalence. For instance, Drewnowski et al4 (see Table 2) undertook a longitudinal survey of 557 college women whom he categorized as no risk (18%), low risk (44%), high risk (25%), subclinical (10%), and clinical (1%-2%). At follow-up, students generally moved from a lesser to a greater symptomatic or risk category.

Evidence That Risk Factors C a n Be Modified Once potential risk factors have been identified, it becomes important to examine interventions that can modify them. As prospective studies identifying potential risk factors for eating disorders

82

The Journal of Behavioral Health Services & Research

29:1 February 2002

Table 2 Distribution of college students by level of risk for eating disorders and clinical status

Category

Percent

No risk Low risk High risk Subclinical Clinical

18 44 25 10 1-2

Data from A. Drewnowski, D.K. Yee, C.L. Kurth, and D.D. Krahn, "Eating Pathology and DSM-III-R Bulimia Nervosa: A Continuum of Behavior," American Journal of Psychiatry, Vol. 151, pp. 1217-1219.

have only been completed recently, it is not surprising that there is relatively little research on reducing risk factors. Nonetheless, cognitive-behavioral and psychoeducational interventions focused on subclinical populations have been shown to reduce eating disorder symptoms and improve attitudes. 8 Interventions to improve body image also have been effective in controlled studies. 9-u Several recent studies have demonstrated that a relatively low-cost, multimedia intervention can significantly reduce body shape concerns in college students at risk for eating disorders. Students who used a computer-delivered psychoeducational eating disorder prevention program (called Student Bodies, Stanford University) were able to improve their body image, as well as adopt healthier eating attitudes and behaviors.12' 13 In one study, a4 14 of 26 women in the intervention group had Body Shape Questionnaire (BSQ) scores at baseline that would put them at risk for an eating disorder (BSQ > 110). At posttreatment, 10 of the 14 students had BSQ scores below 110, 3 reported lower scores but not below 110, and 1 score did not change. It is important to note that the overall effect size of the changes in the intervention actually exceeded those generally reported for controlled studies using face-to-face therapy. Multimedia interventions are certainly not the only approach for reducing risk factors for eating disorders, but they have the advantage of being relatively inexpensive to deliver to large populations (at least those populations with computer access). Student Bodies was recently provided to a group of 1 l0 high school students at a cost of about $25 per student. For purposes of this discussion, it is assumed that Student Bodies could be provided to a group of high-risk students for this amount per student. (Actual delivery costs would depend on a number of factors including the cost of the moderator's time, availability/cost of hardware, Internet connections, and servers.)

Effective interventions Effective psychological therapies for bulimia have been standardized and manualized, permitting widespread dissemination. For instance, a number of controlled trials have found cognitive-behavioral therapy to be effective in reducing symptoms of bulimia, n

Stepped care approaches Recent w o r k 15-17 has focused on stepped care models for eating disorders. For example, Theils et aP 6 compared 3 weeks of a self-help manual followed by eight sessions of cognitive behavior therapy (CBT) for patients still symptomatic with 16 sessions of individual CBT only. The two groups had similar outcomes (40% and 41% were symptom-free at 18 months), with the stepped care approach being considerably less expensive. Agras t5 outlined one such approach for bulimia. Drawing on several studies of treatment effectiveness and cost, he suggested that a cognitive-behavioral self-help

Issues Related to Combining

TAYLOR et al.

83

manual plus periodic clinical supervision might be offered first, which could result in approximately 25% of patients showing clinically significant improvement. Second, antidepressant medications could be added, starting with a tricyclic (TCA) such as desipramine and followed up with a serotonin reuptake inhibitor (SRI), such as fluoxetine, in the event of failure of the TCA. This might result in improvement among 40% of this group (approximately 30% of the original group). A final approach for those who have not responded to the medication(s) would be to use CBT, which might result in a 50% improvement (or about 22% of the original group). Thus, 75% to 80% of patients would be treated successfully. He suggested that the optimal treatment, if cost were not a factor, might be CBT with the option of adding medication if progress is unsatisfactory, which would be a less cost-effective approach to achieving a similar (approximately 78%) rate of improvement. This model combined data from separate investigations, so rates of improvement may not reflect fully the impact of treatment-refractory patients. In the following discussion, this model is extended to risk factor reduction and intervention.

Implementation of the Model with a College Population Step 1 In providing a population-based intervention to a college student population, all female students would take a short screen (eg, the weight concerns instrument developed by Killen et al 7) to determine if they are at high risk for an eating disorder or have an eating disorder. (To increase efficiency and hence reduce cost, the screen would focus on female students only, given the relatively low rate of eating disorders in males.) Students identified as being at high risk could then receive a relatively inexpensive intervention like Student Bodies and those with subclinical/clinical eating disorders could receive a stepped care intervention (except for students with anorexia who would be referred to individual assessment and treatment as needed). The cost-effectiveness of screening varies depending on the prevalence of the disorder and the sensitivity and specificity of the screen. Traditionally, screens are used to identify potential cases; in this model the screen is used to identify a student at risk. Because there is no immediate gold standard as to when a student is at risk (this only becomes evident over time), the use of traditional sensitivity and specificity measures is not entirely applicable. In screening a population for students at risk, one also identifies cases. There are no data in the literature examining the sensitivities and specificities of screens used to stratify a population by risk and caseness at the same time, for instance, to determine in a sample of 100 students that 10 were at risk (needing prevention) and 2 were cases (needing referral and treatment). However, each individual who screened positive would need to be further evaluated and categorized. Table 3 presents the different screening costs/100 identified high-risk students based on different levels of prevalence of high-risk individuals and assuming that the sensitivity and specificity estimates apply to high-risk individuals. "False positives" are those individuals who, on further evaluation, would be considered not at risk or not a case. An on-line or optically scanned screen can be processed very inexpensively ($1.00/screen), and further evaluation could be done cheaply with a paper-and-pencil instrument (such as the Eating Disorder Examination Self-Report Questionnaire 18) for an estimated cost of" $10. The cost of the overall screening jumps dramatically, related to the prevalence of the disorder and the number of false positives that need to be excluded. For instance, it would require $10,125 to identify 100 high-risk cases if the prevalence of high-risk cases in the population was 4% and the screening instrument was 80% specific and sensitive (Table 3). Much of this cost is driven by the suboptimal positive predictive value of only t4%. If the cost of evaluating screen-positive individuals required a clinical interview, which could cost as much as $300 (2 hours of a clinical psychologist's time), then the costs would soar. It is anticipated that the prevalence of high-risk individuals would be about 10% and the instrument would be 90% sensitive and specific. At $10 for a screen, the cost of identifying 100

84

The Journal of Behavioral Health Services & Research

29:1

February 2002

O0

7.

c~

3 3" 3"

Example

25% 25% 10% 10% 10% 10% 4% 4%

Prevalence high risk 80% 90% 90% 70% 80% 90% 80% 90%

Se 80% 90% 70% 90% 80% 90% 80% 90%

Sp 57% 75% 25% 44% 31% 50% 14% 27%

PPV

Screening properties

500 444 1,111 1,429 1,250 1,111 3,125 2,778

Number needed to screen/100 HR 75 33 300 129 225 100 600 267

Number of false positives

Se, sensitivity; Sp, specificity; PPV, positive predictive value; HR, high risk; TP, true positives; FP, false positives

1 2 3 4 5 6 7 8

Table 3

$2,250 $1,774 $5,111 $3,719 $4,500 $3,111 $10,125 $6,448

Cost per 100 H R @ $1/screen and $10 to evaluate TP + FP

Cost of identifying 100 high-risk for eating disorder individuals at different prevalence rates and screening sensitivities and specificities

cases is only $3,111. But if all screen-positive individuals required a $300 clinical interview, then, for Example 4 in the table, the costs for reviewing "false positives" would jump from about $1,000 to $30,000, and the cost of reviewing all screen positive cases would be twice that. Computer-assisted screening procedures that derive from signal-detection-type analyses might make this process more efficient (eg, Smith et all9). Also note that the initial screen costs might very well exceed $1.00 in many settings.

Step 2 Individuals identified as being at high risk (and not being cases) would then be given the Student Bodies 13 multimedia intervention (or some other low-cost psychoeducational program proven to be equally effective and safe). The cost-benefit (cost per case prevented) of providing the intervention would depend on prevalence, incidence, and efficacy of the intervention. As shown in Table 4, these parameters strongly affect the cost-benefit. Under optimistic assumptions (Example 5), the intervention would require $449 per case prevented. However, if the intervention were only 25% effective, a more realistic assumption, the cost would be $898. For low prevalence conditions (eg, 4%) with a relatively low probability of becoming a case (eg, 10%), even a very effective intervention is expensive: $1,790 per case prevented (see Example 11 in Table 4).

Step 3 The students who failed this intervention might benefit from simply being monitored, with more active interventions applied when and if they reached a subclinical level of the disorder. On the other hand, they might then be entered into a more aggressive stepped care approach. The stepped care and intervention numbers presented in Figure 1 are derived from Agras. 15 The 25% of patients not improving with stepped care might require partial day hospital care, inpatient hospitalization, more complicated psychopharmacology, or innovative treatment approaches. Agras 15 estimates that the cost of treating a case in this model is approximately $4,000. In actual practice, many of the high-risk individuals who went on to develop clinical syndromes would neither want nor receive treatment or might be treated in less expensive ways (eg, with group therapy). In addition, changes in any of the parameters listed in Table 4 or the development of new approaches could dramatically change treatment costs. For instance, 12 sessions of CBT/interpersonal group therapy or another group approach 2°'21 may be as effective as Steps 2 and 3 in the above model. All of the examples provided in Table 4 suggest that the prevention cost would be lower than the treatment cost. However, in these models, case identification and referral increase the number of students in a population receiving treatment. An individual is prevented from becoming a case and needing treatment, but that individual may not have actually received treatment had she not been identified through population-based screening. For example, in Table 4, Example 6, the cost of preventing six cases is about $5,610 that would cost $24,000 to treat. But how many of these six individuals would actually seek treatment? If only one of the six cases actually would have wanted/sought treatment without screening/referral, then the cost of prevention is greater than the cost of treatment to the payer of these services. For lower prevalence rates of individuals at high risk for eating disorders, with lower 1-year incidence rates for those high-risk interventions, and with less effective interventions, the cost-benefit of prevention becomes marginal. For instance, in Example 12, in Table 4, the cost of preventing two cases was almost $7,000, not much less than the cost of treating them. It is worth noting that prevention may confer many benefits to those individuals who would not have sought treatment. For instance, participation in the kind of prevention activities described here for high-risk individuals might improve self-image, school, and even work performance that might confer direct personal and perhaps indirect financial benefit.

86

The Journal of Behavioral Health Services & Research

29:1

February 2002

7,

Table 4

0.25 0.25 0.25 0.25 0.1 0.1 0.1 0.1 0.04 0.04 0.04 0.04

Prevalence $1,774 $1,774 $1,774 $1,774 $3,111 $3,111 $3,111 $3,111 $6,448 $6,448 $6,448 $6,448

Cost to screen (90% Se, 90% Sp) .25 .25 .10 .10 .25 .25 .10 .10 .25 .25 .10 .10

1-year incidence/100 HR students without RR 0.5 0.25 0.5 0.25 0.5 0.25 0.5 0.25 0.5 0.25 0.5 0.25

Effectiveness of intervention 12.5 6.25 5 2.5 12.5 6.25 5 2.5 12.5 6.25 5 2.5

Number of cases prevented

$4,274 $4,274 $4,274 $4,274 $5,610 $5,610 $5,610 $5,610 $8,948 $8,948 $8,948 $8,948

Total cost

$342 $684 $855 $1,710 $449 $898 $1,122 $2,244 $716 $1,432 $1,790 $3,579

Cost/prevented case*

Se, sensitivity; Sp, specificity; HR, high risk; RR, risk reduction *Cost to prevent equals: (Screening cost + Cost of prevention)/Number of cases prevented. The intervention is provided at a cost of $252/student = $25 × 100 in these examples.

1 2 3 4 5 6 7 8 9 t0 11 12

Example

Cost per prevented eating disorder case at different prevalence (in HR individuals), incidence (of HR individuals becoming a case), and effectiveness (of the preventive intervention) rates

Figure

1

Projected outcomes from a combined high-risk/clinical intervention

High Risk

I

!

Subclinical/Clinical

i

! High-risk intervention*

S t e p 1 ( P s y c h o e d u c a t i o n plus brief therapist contact)*"

,.-

l/ + 6-month - I ~ 50% outcome

nc 40%

-

10%

I I I +

nc/-

25%

75% I

Step 2 (Medications)----I~ ! I Monitoring

I

+

nc/50%

25% Step 3 (CBT/Interpersonal)

I + 25%

nc/25%

i

Improved treatments.*

Key: +, no longer meets high-risk criteria (if high risk) or subclinical/clinical criteria, as appropriate; nc, no change; - , risk or clinical status is worse; CBT, cognitive behavior therapy *From Winzelberg et al. 14 **From Agrasis

There are many other ways to reduce costs of the preventive intervention. For instance, another approach would be to provide the high-risk intervention only to students who demonstrated that their risk factor profile worsened (a watchful waiting approach) and then provide them a high-risk intervention. An ecologic intervention, designed to reduce the normative acceptance of risk behaviors (eg, bingeing and/or purging among students in the dorm or in sororities, for instance), might also be a cost-effective preventive measure. There are additional costs-benefits to be considered in deciding how to balance prevention and intervention. For instance, step 2, the use of medication, might reduce symptoms but impair sexual function in a high percentage of students. In contrast, high-risk interventions might confer general benefits on students (eg, improving confidence and health) in addition to potentially reducing incidence of eating disorders, although such treatments also might have downsides, such as leading to labeling and stigmatization.22

88

The Journal of Behavioral Health Services & Research

29.'1 February 2002

Implications for Behavioral Health Services Population-based psychotherapy offers an approach to integrating clinical treatments and high-risk interventions in order to offer services to prevent/treat eating disorders within a defined population. This model, presented mostly for heuristic purposes, is based on a number of untested assumptions and focuses on a relatively straightforward problem (subclinical/clinical eating disorders, anorexia excluded). The information necessary to determine how to best treat a population is quite different from the information necessary to treat an individual. The elegant research strategies that have evolved to examine treatment outcome efficacy among individuals in narrowly defined populations of treatment volunteers are only partly relevant to the effectiveness of these interventions as applied to populations with the goal of reducing incidence and/or of providing cost-effective care to everyone in the population who may benefit from it. Furthermore, from the standpoint of linking prevention and intervention, psychopathology often exists on a continuum of disorder not only from clinical to subclinical populations, but also from high-risk to subclinical/clinical groups. The application and monitoring of effects of standardized interventions on defined populations would help elucidate these continua. For instance, are there characteristics of certain individuals with subclinical disorders that predispose them to develop more serious psychopathology? Do successfully treated individuals have higher risk rates than high-risk individuals who have never been cases? The application of standardized and stepped care interventions across populations might be a better way to identify nonresponsive subpopulations of patients requiring different or new treatments and of examining the ecologic validity of narrowly defined population studies. However, stepped care approaches may lead to decreased credibility or increased dropout rates of those who do not benefit from preliminary interventions. Thus, it also might be important for future research to begin to empirically identify individual predictors of nonresponsiveness to particular types of treatment interventions (eg, risk factors for failure of psychoeducation or self-help interventions). Such predictors could be used to identify individuals who would benefit from the immediate application of more costly interventions. The population-based approach offers advantages to researchers interested in the effectiveness or generalizability of their interventions and prevention strategies, or to those interested in understanding the role of economics in whether their research has real-world clinical utility. The cost-benefit models presented in Tables 3 and 4 show that different assumptions about incidence and prevalence, screening and treatment costs, and efficacy, dramatically alter cost-benefit numbers. The model presented focused on treating individuals within a population. Population-based psychotherapy interventions also should consider environmental/ecologic/community/system factors that might positively reduce risk. While primary prevention interventions are much more likely to be effective if they include "environmental or community" interventions, the same also may be true of selected and indicated interventions. For instance, consider a female undergraduate with a body mass index (BMI) >27, who is struggling with poor body image and who attends a fraternity party in which the members are encouraged to carry a stuffed pig around if they date someone in her weight range. An individually directed intervention might help her cope with this humiliating message, but peer pressure on fraternities to discontinue such practices might be a more effective approach. How can the model shown in Figure 1 be used to inform research strategies? Assuming that the end result of a program of research is to develop population-based therapy, then, for any disorder, data on the factors listed in Table 1 need to be considered programmatically. Researchers interested in issues of identifying an at-risk or impaired population might want to consider the cost implications of their measures 22; individuals studying treatment might examine whether treated individuals remain at high risk or fall into some other category (ie, movement back and forth between categories over time can be examined to help guide the judicious use of interventions). Work on moderators and mediators might provide important data on the movement from high-risk to subclinical/clinical cases.

Issues Related to Combining

TAYLOR et al.

89

Acknowledgments This research was partially funded by a grant from the National Institute of Mental Health (NIH 2R01 MH60453-01A1). References 1. Katon W, Von Korff M, Lin E, et al. Population-based care of depression: effective disease management strategies to decrease prevalence. General Hospital Psychiatry, 1997;19:169-178, 2. American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. 4th ed. Washington, DC: American Psychiatric Association; 1994. 3. Fairbum CG, Beglin SJ. Studies of the epidemiology of bulimia nervosa. American Journal of Psychiat~. 1990;147:401-408. 4. Drewnowski A, Yee DK, Kurth CL, Krahn DD. Eating pathology and DSM-III-R bulimia nervosa: a continuum of behavior. American Journal of Psychiatry. 1994;151:1217-1219. 5. Striegel-Moore RG, Silberstein LR. Rodin J, Toward an understanding of risk factors for bulimia. Journal of the American Psychological Association~ 1986;41:246-263, 6. Killen JD, Hayward C, Taylor CB, et at. Factors associated with symptoms of butimia nervosa in a community sample of 6th and 7th grade girls, bzternational Journal of Eating Disorders, 1994;i5(41:357-367. 7. Killen JD, Taylor CB, Hayward CH, et at. Weight concerns influence the development of eating disorders: a fore-year prospective study. Journal of Consulting and Clinical Psychology. 1996;64:936,-940. 8. Carter JC, Fairbum CG. Cognitive-behavioral self-help for binge eating disorder: a controlled effectiveness study. Journal of Consulting and Clinical Psychology. 1998;66(4):616-623, 9. Cash TE Body-Image Therapy- A Programfor Self-Directed Change. New York: The Guilford Press: 1991. 10. Rosen JC. Body image assessment and treatment in controlled studies of eating disorders. International Journal of Eating Disorders. 1996;20:331-343. 11. Lewandowski LM, Gebing TA, Anthony J L O'Brien WH, Meta-analysis of cognitive-behavioral treatment studies or bulimia~ Clinical Psychology Review. 1997:17(7):703-718, 12. Winzelberg AJ, Taylor CB, Altman TM, Eldredge KL, Dev P, Constantinou PS. Evaluation of a computer-mediated eating disorder intervention program. International Journal of Eating Disorders. 1998;24:339--349. 13. Celio AA. Winzelberg AJ, Wilfley DE, et al. Reducing risk factors tbr eating disorders: comparison of an Intemet- and a classroom-delivered psychoeducation program. Journal of Clinical and Consulting Psychology. 2000;68(41:650~557. 14, Winzelberg AJ. Eppstein D, Eldredge KL, et aL Effectiveness of an Internet-based program for reducing risk factors for eating disorders. Journal of Consulting and Clinical Psychology. 2000:68:346-350. 15. Agras WS. The treatment of bulimia nervosa, Drugs of Today. 1997;33(61:405-411. 16. Wilson GT, Vitousek KM, Loeb KL. Stepped care treatment for eating disorders. Journal of" Consulting and Clinical Psychology. 2000;68(41:564-572. 1% Thiels C, Schmidt U, Treasure J, Garthe R, Troop N. Guided self-change for butimia nervosa incorporating use of a self-care manual. American Journal of Psychiatry. 1998:155(71:947-953. 18. Luce KH, Crowther JH. The reliability of the Eating Disorder Examination-Self-Report Questionnaire Version (EDE-Q). International Journal of Eating Disorders, 1999;25:349-351. 19. Smith PM, Kraemer HC. Miller NH, DeBusk RE Taylor CB. In-hospital smoking cessation programs: who responds, who doesn't. Journal of Consulting and Clinical Psychology. 1999:67:19-27. 20. McKisack C, Waller G. Factors influencing the outcome of group psychotherapy for bulimia nervosa. International Journal of Eating Disorders. 1997;22:1-13. 21. Mitchell JE, Pyle RL, Pomeroy C, et al. Cognitive-behavioral group psychotherapy of bulimia nervosa: importance of logistical variables. International Journal of Eating Disorders. 1993;14:277-287. 22. Offord D, Kraemer HC, Kazdin AE, Jensen PS, Harringtun R. Lowering the burden of suffering from child psychiatric disorder: trade-offs among clinical, targeted, and universal interventions. Journal of theAmerican Acadernv of Child and Adolescent Psychiatry, i998:37(7):686694.

90

The Journal o f Behavioral Health Services & Research

29:1

February 2002