Journal of Clinical Psychology in Medical Settings

, Volume 17, Issue 3, pp 258–271

Bibliotherapy as a Treatment for Depression in Primary Care

Authors

    • Department of Family & Community MedicineUniversity of Nevada School of Medicine
  • David O. Antonuccio
    • Department of Psychiatry and Behavioral SciencesUniversity of Nevada School of Medicine
  • Mark Litt
    • Division of Behavioral SciencesUniversity of Connecticut Health Center
  • Gary E. Johnson
    • Department of Family & Community MedicineUniversity of Nevada School of Medicine
  • Daniel R. Spogen
    • Department of Family & Community MedicineUniversity of Nevada School of Medicine
  • Richard Williams
    • Department of Family & Community MedicineUniversity of Nevada School of Medicine
  • Catherine McCarthy
    • Department of Family & Community MedicineUniversity of Nevada School of Medicine
  • Marcia M. Lu
    • Department of Family & Community MedicineUniversity of Nevada School of Medicine
  • David C. Fiore
    • Department of Family & Community MedicineUniversity of Nevada School of Medicine
  • Dianne L. Higgins
    • Tahoe Forest Multispecialty Clinics
Article

DOI: 10.1007/s10880-010-9207-2

Cite this article as:
Naylor, E.V., Antonuccio, D.O., Litt, M. et al. J Clin Psychol Med Settings (2010) 17: 258. doi:10.1007/s10880-010-9207-2

Abstract

This study was designed to determine whether a physician-delivered bibliotherapy prescription would compare favorably with the prevailing usual care treatment for depression in primary care (that often involves medication) and potentially offer an alternative. Six family physicians were trained to write and deliver prescriptions for cognitive-behavioral bibliotherapy. Thirty-eight patients were randomly assigned to receive either usual care or a behavioral prescription to read the self-help book, Feeling Good (Burns, D. D. (1999). Feeling good: The new mood therapy. New York: HarperCollins). The treatment groups did not differ in terms of overall outcome variables. Patients in both treatment groups reported statistically significant decreases in depression symptoms, decreases in dysfunctional attitudes, and increases in quality of life. Although not statistically significant, the mean net medical expenses in the behavioral prescription group were substantially less. This study provided empirical evidence that a behavioral prescription for Feeling Good may be as effective as standard care, which commonly involves an antidepressant prescription.

Keywords

DepressionPrimary careBibliotherapyUsual care

Introduction

Pharmacological interventions prevail as the predominant treatment delivered in primary care medical settings for depression (Olfson et al., 2002). Among adults who are at least 18-years-old, office visits where an antidepressant was mentioned in progress notes increased from 45.1 million in 1998 to 60 million in 2001, i.e. 7.1% of all office visits in outpatient settings (Moore, 2004).

Although psychotropic medications have been demonstrated to be helpful in ameliorating mental and behavioral problems, all medical interventions for depression carry with them certain risks and side effects (Antonuccio, Burns, & Danton, 2002; Antonuccio, Danton, DeNelsky, Greenberg, & Gordon, 1999). The short-term side effects of selective serotonin reuptake inhibitors (SSRIs), a class of antidepressant medications commonly prescribed in primary care, are well established and may include agitation, sleep disruption, gastrointestinal problems, and sexual dysfunction (Antonuccio et al., 1999). There is recent evidence suggesting that sexual dysfunction may be a long-term irreversible effect for some patients after discontinuing antidepressant medications (Bahrick, 2008). In a study that examined the prevalence of cognitive and physical symptoms in a sample of 117 positive responders to antidepressant treatment, Fava et al. (2006) found that cognitive symptoms, including apathy, inattentiveness, forgetfulness, word-finding difficulty, and mental slowing were endorsed by more than 30% of the participants, while the physical symptoms of fatigue and sleepiness or sedation were reported by more than 40% of the sample. Impaired driving ability has also been associated with antidepressant usage (Brunnaur, Laux, Geiger, Soyka, & Moller, 2006). Side effects and medical risks (including death) increase when antidepressants are combined with other medications (Dalfen & Stewart, 2001), as is often the case (Antonuccio et al., 1999). As patients age and develop multiple health problems, the risks associated with combining medications also multiply. In addition, the withdrawal symptoms of antidepressants are substantial for many if not most patients (Coupland, Bell, & Potokar 1996; Fava, 2002; Rosenbaum, Fava, Hood, Ashcroft, & Krebs, 1998). Antidepressant usage has also been associated with an increased risk of obesity, diabetes, and stroke (Patten et al., 2009; Raeder, Bjelland, Emil Vollset, & Steen, 2006; Smoller et al., 2009). Induced manic episodes (Preda, MacLean, Mazure, & Bowers, 2001) and acts of deliberate self-harm have also been associated with antidepressants (e.g., Donovan et al., 2000; Hammad, Laughren, & Racoosin, 2006; Healy, 2002) resulting in FDA warnings (Antonuccio, 2008). Though data are mixed and somewhat controversial, other potential risks that warrant further investigation include the association (though not yet established causal relationship) of antidepressants with breast cancer (e.g., Bahl, Cotterchio, & Kreiger, 2003; Cotterchio, Kreiger, Darlington, & Steingart, 2000; Halbreich, Shen, & Panaro, 1996; Moorman, Grubber, Millikan, & Newman, 2003; Sharpe, Collet, Belzile, Hanley, & Boivin, 2002), and the possibility of irreversible biochemical changes predisposing some susceptible patients to chronic depression (e.g., Baldessarini, 1995; Fava, 1995, 2002).

Given the risks associated with antidepressants, it is relevant to consider whether other non-pharmacological interventions can also be delivered in primary care settings to give physicians effective choices with fewer medical risks. This is especially important given that patients in a primary care setting often do not adhere to antidepressant medications due to side effects, lack of improvement in symptoms, or because they may actually feel worse (Haslam, Brown, Atkinson, & Haslam, 2004). A review of the literature suggests that physicians have positive attitudes about writing non-pharmacological prescriptions for behavior change such as exercise (Swinburn, Walter, Arroll, Tilyard, & Russell, 1997), patients benefit from non-pharmacological prescriptions for behavior change (Swinburn, Walter, Arroll, Tilyard, & Russell, 1998), and physicians are interested in alternatives to pharmacological intervention for depression (Mental Health Foundation, 2006).

In the era of managed care, it is not enough to be effective, treatments must be cost-effective. A number of studies (Guthrie et al., 1999; McLeod, Budd, & McClelland, 1997; Simon et al., 2001) have focused on the reduction of health care utilization cost via the development of psychosocial intervention programs for a variety of psychological conditions such as depression. When medical cost-offset (Hunsley, 2003), relapse, and side effects are considered in a cost-benefit analysis, cognitive-behavioral psychological interventions for depression can be very cost-effective, particularly in a psychoeducational (e.g., therapist-assisted bibliotherapy) or group format (Antonuccio, Thomas, & Danton, 1997). In addition, cost-benefit analyses suggest that cognitive-behavioral interventions are likely to be at least as cost-effective as antidepressant prescriptions (Antonuccio et al., 1997; Dobson et al., 2008). Furthermore, integrating such psychological treatment into primary care settings can result in a reduction in overall medical expenditures (Cummings, Pallik, Dorken, & Henke, 2000).

The threefold rationale behind this study, that behavioral prescriptions could be an effective alternative to drug prescriptions, was that (1) psychological interventions (particularly cognitive therapy, behavioral activation, and interpersonal therapy) compare favorably with medications in the short term, even when the depression is severe (e.g., Antonuccio, Danton, & DeNelsky, 1995; DeRubeis, Gelfand, Tang, & Simons, 1999; DeRubeis et al., 2005), (2) they appear to result in more enduring benefits than medications when long term follow-up is considered (Antonuccio et al., 1995; Hollon, Shelton, & Loosen, 1991; Hollon et al., 2005), and (3) they have minimal apparent medical side effects (e.g. TADS, 2004, 2007). Several meta-analyses, reported in both psychiatry and psychology journals covering multiple studies with thousands of patients, are remarkably consistent in support of the perspective that psychological interventions are at least as effective as medication in the treatment of depression (e.g., Conte, Plutchik, Wild, & Karasu, 1986; Cuijpers, van Straten, Warmerdam, & Andersson, 2008; de Maat, Dekker, Schoevers, & de Jonghe, 2007; DeRubeis et al., 1999; Dobson, 1989; Hollon et al., 1991; Imel, Malterer, McKay, & Wampold, 2008; Robinson, Berman, & Neimeyer, 1990; Steinbrueck, Maxwell, & Howard, 1983; Wexler & Ciccheti, 1992). Additionally, Fournier et al. (2010) in a recent meta-analysis found that antidepressants do not show clinically significant benefits over placebo treatment, except for very severely depressed patients (i.e. only 30% of patients for whom they are actually prescribed).

It is generally impractical for physicians in a primary care setting to deliver psychotherapy for depression. However, it is reasonable to assume that they may be able to deliver components of empirically supported cognitive-behavioral psychological interventions in the form of brief written “behavioral” prescriptions. One such evidence-based psychological intervention for depression is bibliotherapy (Ackerson, Scogin, McKendree-Smith, & Lyman, 1998; Cuijpers, 1997; Gould & Clum, 1993; Gregory, Schwer Canning, Lee, & Wise, 2004; Jamison & Scogin, 1995; Pantalon, Lubetkin, & Fishman, 1995; Scogin, Jamison, & Davis, 1990; Scogin, Jamison, Floyd, & Chaplin, 1998; Smith, Floyd, Jamison, & Scogin, 1997). Of the more impressive findings, Cuijpers’ (1997) meta-analysis of bibliotherapy for depression that examined six studies yielded a large mean effect size of 0.82 (with a 95% confidence interval of 0.50–1.15); and more recently, the meta-analysis of Gregory et al. (2004) bibliotherapy for depression reviewed seventeen studies and demonstrated a large effect size of 0.77 (with a 95% confidence interval of 0.61–0.94).

Cognitive-behavioral bibliotherapy has been effectively delivered in settings outside of the primary care environment. The research conducted by Scogin and his colleagues provides a model for examining the efficacy of bibliotherapy (Jamison & Scogin, 1995; Scogin et al., 1990; Smith et al., 1997). The principal objective of the current study was to build on the results of the Jamison and Scogin (1995) study in a primary care setting.

In this study, a physician written behavioral prescription for the book entitled Feeling Good (Burns, 1999) was compared to a usual care control condition in order to test practicality and feasibility. Patients were randomly assigned to either treatment. In the usual care control group, physicians delivered their treatment of choice. A usual care control condition has been recommended as an ethical and practical control group in clinical settings (Borkovec, 1990; Haaga & Stiles, 2002).

The rationale for using usual care as the control group was that for behavioral prescriptions to be practical they must compare favorably to the prevailing standard treatment. Otherwise, there would be no reason to prescribe them at all. Designs using a usual care control group have been successfully implemented in previous depression studies in a primary care setting (e.g., Bruce et al., 2004; Lin et al., 2003). This design required neither deception nor delay in treatment, thereby providing all patients with some kind of treatment. The behavioral prescription was designed to be delivered in 5 min. Concerns that not prescribing an antidepressant may deprive patients of a preferred treatment at the least, or at the worst might increase suicide risk relative to the usual care condition, are not supported by recent scientific studies (Antonuccio et al., 2002; Khan, Warner, & Brown, 2000). The prevailing literature suggests that most patients prefer a psychotherapeutic intervention when given the choice (Chilvers et al., 2001; Hall & Robertson, 1998; Jorm, 2000; Paykel, Hart, & Priest, 1998; Priest, Vize, Roberts, Roberts, & Tylee, 1996). In addition, it is important to note that there is no evidence that placebo treatments increase the risk of suicide in comparison with antidepressants in randomized controlled trials (Khan et al., 2000; Storosum, van Zwieten, van den Brink, Gersons, & Broekmans, 2001). Furthermore, actively suicidal patients were excluded from the study to minimize the risk of overdose in those patients who received antidepressants and to avoid placing suicidal patients in a limited contact treatment.

The primary hypotheses of this study were (1) patients would adhere to behavioral prescriptions at rates similar to medical prescriptions, (2) behavioral prescriptions would be at least as effective as usual care according to the primary outcome measures, (3) behavioral prescriptions would be at least as cost-effective as usual care based on a direct medical cost comparison, (4) physicians would be at least as satisfied with delivering a behavioral prescription as prescribing an antidepressant for depression treatment, and (5) patients who received a behavioral prescription for depression would be at least as satisfied with their treatment for their depression as those who received usual care treatment.

Methods

Participants

Physicians

Eight staff physicians at the Family Medicine Center, an outpatient primary care clinic run by the Department of Family and Community Medicine affiliated with the Nevada School of Medicine in Reno, NV, were offered the opportunity to participate in the study at a monthly staff meeting after listening to a didactic presentation that provided an overview of the study and participation criteria. Six out of the eight physicians volunteered to participate and there was no physician attrition. Overall, the physician sample included an equal number of men and women, was predominantly Caucasian, and ages ranged from 34 to 54. Number of years practicing family medicine ranged from 5 to 25 and the percentage of depressed patients on each physician’s case load per week ranged from 20 to 50%. No physicians reported any previous specific behavioral or mental health training.

Patient Recruitment

All patient participants in this study were recruited from the patients receiving primary care treatment at the Family Medical Center. Out of a total of 294 flyers that were distributed, a total of 290 patients responded. There were 251 patients who were excluded from the study for the following reasons: 111 were not interested in participating, 86 scored below 4, the clinical cut-off for mild depression on the Beck Depression Inventory-Fast Screen for medical patients (BDI-FS; Beck, Guth, Steer, & Ball, 1997; Beck, Steer, Ball, Ciervo, & Kabat, 1997; Beck, Steer, & Brown, 2000), 49 reported antidepressants use within the 30 days of screening, two endorsed suicidal ideation (i.e., circled a “2” on question number six of the BDI-FS and were referred to other mental health treatment), and three indicated an inability to read. The BDI-FS has demonstrated reliability and validity in adolescent, adult, and geriatric medical patients (Scheinthal, Steer, Giffin, & Beck, 2001; Steer, Cavalieri, Leonard, & Beck, 1999; Winter, Steer, Jones-Hicks, & Beck, 1999).

The sample size for each of the two interventions was calculated to detect a statistically significant (p < .05) difference in depression outcome variables by at least 50% before and after treatment, achieving 5% type I error, 20% Type II error, and a power of 80%. Factoring in a predicted 20% drop-out rate, 19 participants were calculated to be needed in each group respectively.

Patients were recruited at the Family Medicine Center and met the following inclusion criteria: at least 18-years-old, English-speaking, sixth-grade reading level, and reported mild (BDI-FS score between 4 and 6), moderate (BDI-FS score between 7 and 9), or severe depression symptoms (BDI-FS score of 10 or above). Exclusion criteria were limited to the following: suicidality (endorsing a “2” or “3” on question number six of the BDI-FS), death of an immediate family member or a suicide attempt in the prior 6 months, active antidepressant usage within 30 days of screening, or psychosis (as determined by medical chart and physician examination).

Minimal exclusion criteria were used to maximize generalizability and reflect the reality that antidepressant prescription practices in the real world of primary care utilize minimal exclusion criteria. After a patient indicated interest in the study by responding affirmatively to a flyer, a study investigator co-located in a separate office in the primary care environment met with him or her individually before a routine physician’s appointment. During this encounter, the investigator reviewed and obtained a patient’s signature on a screening consent form and determined his or her eligibility for the study using the BDI-FS and a short screening form designed specifically for the study to evaluate suitability based on the inclusion and exclusion criteria. If a patient met inclusion criteria, he or she reviewed and signed a second consent form to participate in the study, and the physician was given a randomized intervention folder that prompted the delivery of either usual care or a bibliotherapy prescription. In order to ensure random assignment, 38 random numbers were generated in a randomization table (Dean et al., 1995) for a total sample of 38 participants who were randomly assigned to either standard care or a prescription for bibliotherapy. Patients recruited for the study were numbered from 1 to 38 consecutively based on natural recruitment. There was only one patient who was randomized to a condition (usual care) and due to physician error received the wrong intervention (a behavioral prescription). This patient was allowed to continue with the study; however, she was removed from the randomization, numbered 39, and not included in data analysis in order to preserve the randomization process. This research study was approved by the Office of Human Research Protection at the University of Nevada, Reno.

Final Patient Sample

Sample mean characteristics are displayed in Table 1. The age of the total patient sample ranged from 22 to 83 years-old, was predominantly female and Caucasian, and was equally represented by employed and unemployed patients. Of the 38 patient participants who were originally recruited and randomized into the study, five participants (14%) were unable to be contacted for posttreatment assessment, and therefore, were considered dropouts for the treatment portion of the study.
Table 1

Patient sample characteristics

Variable

M or % (SD)a

Behavioral prescription (n = 19)

Usual care (n = 19)

Total (n = 38)

Age (years)

48.6 (17.06)

54.3 (11.8)

51.45 (15.1)

Sex

 Female

94.7 (n = 18)

73.7 (n = 14)

84.2 (n = 32)

 Male

5.3 (n = 1)

26.3 (n = 5)

15.8 (n = 6)

Ethnicity

 Caucasian

89.5 (n = 17)

94.7 (n = 18)

92.1 (n = 35)

 African American

5.3 (n = 1)

5.3 (n = 1)

5.3 (n = 2)

 Hispanic Caucasian

5.3 (n = 1)

0 (n = 0)

2.6 (n = 1)

Employment status

   

 Employed

47.4 (n = 9)

52.6 (n = 10)

50 (n = 19)

 Unemployed

52.6 (n = 10)

47.4 (n = 9)

50 (n = 19)

Absenteeismb

4.8 (n = 9) (6.3)

1.5 (n = 10) (3.1)

3.1 (n = 19) (5.0)

BDI-FS

8.3 (3.7)

7.5 (4.0)

7.9 (3.8)

BSI-FS severity (number of patients)

 Minimal

0

0

0

 Mild

8

11

19

 Moderate

3

3

6

 Severe

8

5

13

Life satisfactionc

2.5 (1.0)

2.8 (1.0)

2.7 (1.0)

OAT sensibility

4.5 (1.7)

5.5 (1.6)

4.97 (1.7)

OAT implementation

5.0 (0.5)

5.8 (1.7)

5.4 (1.6)

OAT confidence

4.5 (1.4)

5.2 (1.4)

4.8 (1.4)

OAT total

14 (4.3)

16.5 (4.0)

15.2 (4.3)

TEF

 

1.5 (1.1)

1.9 (1.0)

1.17 (1.0)

BDI-FS Beck Depression Inventory-Fast Screen for medical patients (Scale: 0–3 = minimal symptoms of depression, 4–6 = mild symptoms of depression, 7–9 = moderate symptoms of depression, 10+ = severe symptoms of depression), OAT Opinions About Treatment (Scale: min. = 1, “not at all”, max. = 7, “extremely”), TEF Treatment Expectations Form (Scale: min. = −3 “I expect to manage my depression much worse”, max. = +3 “I expect to manage my depression much better”)

aThere were no statistically significant differences between groups on any of these primary measures

bAbsenteeism = No. of days from work missed in last month for reasons other than a planned vacation

cOverall Life Satisfaction = a stand alone question on the Q-LES-Q SF that asks: How would you rate your overall life satisfaction in the last week; 1 = very poor, 2 = poor, 3 = fair, 4 = good, 5 = very good

Procedures

Nineteen participants in the study received a treatment that consisted of (1) a physician-delivered verbal prompt, along with a physician-signed behavioral prescription note to read the self-help book Feeling Good (Burns, 1999) and (2) instructions to communicate with a study investigator weekly during treatment, and 1 and 3 months later. These calls were limited to 5–10 min, the time it took to complete the standardized assessment questions and address any questions related to study participation (approximately 5–10 min). Patients were not provided with any psychoeducation or advice regarding how to manage their depressive symptoms. Feeling Good has been established as the most rigorously researched self-help book for depression (Jamison & Scogin, 1995; Scogin, Hamblin, & Beutler, 1987; Scogin, Jamison, & Gochneaut, 1989; Scogin et al., 1990; Smith et al., 1997). It is widely available (with more than four million copies in print) and inexpensive (costing $7.95). Participants in this pilot study were encouraged to read the first 423 pages, the section on cognitive strategies, which was roughly equivalent to the reading amount and content in previous studies. Participants assigned to the bibliotherapy condition continued to receive the usual care services (with the exception that no new psychotropic medications were added).

Nineteen participants received a treatment that consisted of (1) usual care and, (2) instructions to communicate with a study investigator weekly during treatment, and 1 and 3 months later. As in the experimental condition, these calls were limited to 5–10 min, the time it took to complete the standardized assessment questions and address any questions related to study participation (approximately 5–10 min). Patients were not provided with any psychoeducation or advice regarding how to manage their depressive symptoms. Standard clinic services were not influenced in any way for the usual care patients throughout the study. Usual care interventions delivered during the study included a prescription for antidepressant medications (11/19), no additional treatment other than communication with study investigator (4/19), regular exercise (1/19), biweekly physician visits (1/19), and referral to psychotherapy (3/19). One out of the 19 participants randomized to usual care was prescribed the combination of psychotherapy and medication, thus she was tallied in both categories. There were no changes in medical care delivered to patients throughout the treatment and follow-up phases of the study.

Prior to patient recruitment, the physicians who participated in the study were trained in approximately 15 min to deliver both treatments. Their training involved (1) a didactic presentation about the study and its rationale, (2) watching a role play of the physician–patient interaction during the delivery of standard care and a behavioral prescription, (3) participating in a question and answer session regarding the study protocol, and, (4) reading prototypes of standard care and behavioral prescription scripts. Comprehension was indicated by a check mark or initial next to every numbered step. Immediately following physician training, physicians’ opinions and expectations about the cognitive-behavioral bibliotherapy intervention were assessed using the self-report measures, Opinions About Treatment (OAT; Borkovec & Mathews, 1988) and the Treatment Expectations Form (TEF; Elkin, Shea, Watkins, & Imber, 1989).

After a physician prescribed either usual care or the behavioral prescription, the patient was directed to the study investigator located within the clinic to complete a battery of brief baseline assessment measures. In addition to the BDI-FS, baseline measures administered to patients included an assessment of patients’ attitudes and expectations about treatment (OAT; Borkovec & Mathews, 1988; Treatment Expectations Form; Elkin et al., 1989), dysfunctional attitudes (abbreviated Dysfunctional Attitude Scale; DAS-A, Weissman & Beck, 1978), and quality of life (Quality of Life Enjoyment and Satisfaction Questionnaire Short Form [Q-LES-Q SF]; Endicott, Nee, Harrison, & Blumenthal, 1993).

During the 6 week intervention phase, all patient participants were telephoned weekly and asked if they had any questions regarding the study and the BDI-FS was re-administered. Reading Feeling Good in six weeks required patients to read an average of 11 pages/day. Adherence to the behavioral prescription condition was measured by asking each participant to state the number of pages read in Feeling Good during the previous week. Adherence to the usual care condition was assessed by patient self-report of the estimated percentage of antidepressant pills taken as prescribed during the previous week, or, by patient self-report of the estimated percentage of completed behaviors recommended by the physician for those patients who were prescribed exercise, biweekly physician appointments, or psychotherapy.

At the end of the 6 week intervention phase, the degree of patient adherence in the behavioral prescription condition was assessed by calculating a percentage of the total (1) number of pages read, and, (2) behavioral schedules completed from the Feeling Good book. Degree of comprehension of reading material was assessed by the Cognitive Bibliotherapy Test, a 23-item true or false test that was designed as a measure of comprehension of Feeling Good for previous studies (Scogin et al., 1998). The BDI-FS, DAS-A, Q-LES-Q SF were again administered. In addition, physician and patient satisfaction were assessed using a two question Likert scale questionnaire specifically designed for this study. One month and three months following the intervention phase, participants were contacted and the BDI-FS, DAS-A, and Q-LES-Q were re-administered. Telephone calls were limited to 10 min.

A cost analysis and comparison between the two treatment conditions was conducted at the completion of the study. Although cost was not a primary focus of this study, some modest efforts to estimate cost were applied. An archived quantification of all direct medical treatment costs for each participant was obtained from the medical billing record during the course of the study. Direct medical costs included all billable procedures performed at the FMC or with a FMC physician at the local inpatient hospital. These included scheduled visits at FMC, referral from FMC to an inpatient hospital emergency room, basic imaging, general primary care procedures, and any other billable medical and/or diagnostic interventions documented in a patients’ medical billing record. The number of completed physician appointments, visits to the emergency room, inpatient days in the hospital, and the total number of medical conditions noted in the medical billing system between baseline and three-month follow-up for each participant were also summed and compared between groups at the conclusion of the study. Although it would have been preferable to audiotape a sampling of physician behaviors with patients enrolled in the study and determine both adherence and time required to deliver both usual care and the behavioral prescription interventions, this monitoring was not feasible at the FMC.

Results

Analyses of mean differences and mixed models were conducted using the statistical package SPSS (SPSS for Windows, Rel. 12.0, 1999; SPSS Graduate Pack for Windows, Rel. 15.0, 2006). Multiple one-way analyses of variance (ANOVAs) and chi-square tests were used to compare the baseline characteristics of the two groups. No significant differences on any variable were found between the two groups at baseline.

Physician Adherence

Physician adherence to treatment delivery was assessed by comparing physician self-report of behaviors to patient report of completed behaviors listed on each treatment delivery script immediately after the physician–patient contact. Percentage of adherence for each physician per patient was calculated by dividing the number of completed items (as reported by both physician and patient) on each script by the total number of items. If a discrepancy was found between the patient and physician report, the item was considered not completed. Mean percentage adherence for each physician was calculated by dividing the sum of a physician’s adherence percentages by the total number of patients seen by that physician. The number of patients per physician in the study ranged from 3 to 12 (M = 6.5, SD = 3.6). There was no discernible pattern in the prescription practices of the physicians in the usual care condition, i.e., no physician seemed to favor one intervention over another. Five out of six physicians achieved 100% adherence to treatment delivery in both conditions. The sixth physician achieved a 92% adherence rating.

Patient Adherence

A one-way ANOVA yielded no significant differences between groups regarding the primary adherence variable. The percentage of total pages read in Feeling Good ranged from 0 to 100 (M = 50.9, SD = 41.2). Patients who were not finished reading the book at posttreatment were encouraged to continue reading at the 1- and 3-month follow-up check-ins. By the final follow-up point at 3 months, the reading adherence rate among those for whom data were available was 62.9%. Time constraints associated with child rearing and employment were cited as the primary reasons by patients who read less than 50% of the assigned reading. There was not a significant correlation between the number of pages read in Feeling Good and final BDI-FS scores, (r = −.42, p = .11).

The percentage of adherence to physicians’ recommendations within the usual care condition (that included antidepressants, psychotherapy, exercise, regular physician contact) ranged from 0 to 100 (M = 73.5, SD = 35.0). Selective serotonin reuptake inhibitors, including Zoloft (sertraline) and Prozac (fluoxetine), were prescribed to seven out of eleven patients who were prescribed medication; Cymbalta (duloxetine), Trazadone (desyrel), Wellbutrin (bupropion), and Buspar (buspirone) were prescribed to the other four. As in the behavioral prescription group, patients in the usual care condition were encouraged to keep following their physicians’ recommendations after treatment when appropriate. At 3 months, the adherence rate to medication (measured by the percentage of days in which medications were taken as prescribed according to patient self-report) among those for whom data were available was 79.2%. Three out of the four patients who were prescribed interventions other than medication all reported 100% adherence to treatment. These included exercise, psychotherapy, and regular visits to a primary care physician. The fourth patient who was prescribed psychotherapy reported adhering to the recommendation 75% of the time.

Comparable findings were revealed by secondary adherence variable analyses. A one-way ANOVA yielded a non-significant difference between the two groups’ participation in weekly telephone check-ins (of which there were six) during the 6 week treatment phase. Similarly, a series of chi-square analyses revealed non-significant differences between groups in terms of attrition rate at posttreatment and 1 and 3 month follow-up time points. Missing data were found to be higher for those participants who were employed, suggesting that the employed patients were less likely to participate in follow-ups. The only observable significant difference detected between the two conditions was performance on the Cognitive Bibliotherapy Test (F(1, 30) = 4.167, p = .05, see Table 2 for sample means), suggesting that the behavioral prescription patients showed a higher level of comprehension of cognitive therapy principles as compared to those within the usual care group. Lastly, the mean percentage of completed book exercises within the behavioral prescription condition ranged from 0 to 100 (M = 50.3, SD = 45.8). Analysis of group means indicated that patients in both groups were at least 50% adherent to their treatment and that adherence rates did not differ significantly by treatment condition. The correlation between the number of treatment contacts with a mental health professional and adherence ratings was not significant (r = .29, p = .12).
Table 2

Treatment adherence

Variable

M or % (SD)

Behavioral prescription (baseline n = 19)

Usual care (baseline n = 19)

Attrition from baseline to

  

 Posttreatment

21.1 (n = 4)

5.2 (n = 1)

 1 Month posttreatment

26.3 (n = 5)

10.5 (n = 2)

 3 Months posttreatment

42.1 (n = 8)

15.8 (n = 3)

Adherence to Txa

50.9 (n = 17) (41.2)

73.5 (n = 15) (35.0)

Missed Tx contacts

1.6 (n = 19) (2.0)

0.6 (n = 19) (1.5)

CBT (% correct)

68.2 (n = 15) (12.1)

59.5 (n = 17) (11.9)

Book exercises (% complete)

50.3 (n = 15) (45.8)

Not applicable

CBT Cognitive Bibliotherapy Test, CBT data not available for one patient in the usual care condition

aAdherence was measured by pt. self report of % of total book pages read in Feeling Good in the bibliotherapy prescription group (note that n is 2 more than at posttest, accounting for 2 patients who were contacted during tx but not at post), and % of medication prescribed or % of physician recommendations followed in usual care (Four patients in the usual care condition were instructed to do nothing other than check-in with co-investigator 1/wk; thus, they were removed from the usual care adherence)

Treatment Efficacy

Linear mixed models (Bryk & Raudenbush, 1992) were primarily used to evaluate the overall effectiveness of the two treatments. The goal of these analyses was to detect any significant differences in the mean predicted values of the BDI-FS, Q-LES-Q SF, and DAS-A, (the primary dependent outcome variables) as a function of treatment condition and time (the independent variables). This mixed modeling procedure was chosen because it uses maximum likelihood estimation to calculate parameter estimates, and thus uses all available data. Additionally, this procedure is better able to take into account unequal time intervals than standard parametric statistics (Singer & Willett, 2003). Main effects of Treatment and Time (from baseline through weeks 6, 10 and 18), and the Treatment × Time interaction were examined on the BDI-FS, Q-LES-Q SF, and DAS-A. Planned contrasts were used to analyze the differences between the two treatments as varying by time, which modeled the BDI-FS, Q-LES-Q SF, and DAS-A as curvilinear (squared) functions. Lastly, simple effects of the two treatments were calculated through a series of independent pairwise comparisons among the estimated marginal means of the BDI-FS, Q-LES-Q SF, and DAS-A at each time interval (see Table 3 for sample means and SDs).
Table 3

Means and standard deviations of outcome variables

Measure

Behavioral prescription

Usual care

Medication

B (n = 19)

PostT (n = 15)

1 M (n = 14)

3 M (n = 11)

B (n = 19)

PostT (n = 18)

1 M (n = 17)

3 M (n = 16)

B (n = 11)

PostT (n = 11)

1 M (n = 11)

3 M (n = 10)

BDI-FS

8.3 (3.7)

4.4 (5.3)

3.3 (3.9)

3.0 (3.5)

7.53 (4.0)

4.9 (5.3)

3.9 (4.4)

4.8 (5.2)

7.91 (4.11)

6.09 (6.20)

4.36 (5.07)

3.80 (4.92)

DAS-A

41.8 (11.6)

47.7 (12.0)

47.6 (11.7)

50.33 (8.0)

39.8 (10.2)

46.1 (12.9)

45.8 (10.2)

45.63 (10.4)

    

Q-LES-Q SF

 

46.6 (18.0)

58.5 (24.7)

59.9 (20.8)

59.8 (21.5)

46.1 (15.9)

54.7 (19.6)

58.1 (16.8)

56.6 (21.9)

    

No statistically significant differences were detected between groups on any of these variables

BDI-FS Beck Depression Inventory-Fast Screen for medical patients (Scale 0–3 = minimal symptoms of depression, 4–6 = mild symptoms of depression, 7–9 = moderate symptoms of depression, 10+ = severe symptoms of depression), DAS-A Dysfunctional Attitude Scale-Abbreviated (Scale: higher scores = increased maladaptive attitudes, lower scores = decreased maladaptive attitudes), Q-LES-Q SF Quality of Life Questionnaire Short Form (Scale: lower scores = lower enjoyment and satisfaction, higher scores = higher enjoyment and satisfaction)

No significant main effects were found for Treatment or for the Treatment × Time interaction on any of the numerical outcome variables. However, a significant Time effect was found between baseline and posttreatment for the BDI-FS [F(3, 56.718) = 8.86, p < .001], DAS-A [F(3, 63.053) = 3.38, p < .05] and, the Q-LES-Q SF [F(3, 56.833) = 4.08, p < .05]. A series of contrast analyses (varying by time) revealed no further significant findings.

Finally, post hoc one-way and repeated measures ANOVAs were used to determine if any significant differences existed between the mean BDI-FS change scores of the behavioral prescription group and those within the usual care group who were prescribed antidepressant medication (n = 11) at posttreatment and at the 1 and 3 month follow up time points. No significant differences were revealed.

These results displayed in Table 3 suggest that patients in both conditions demonstrated significant decreases in reported depression symptoms, decreases in dysfunctional attitudes, and increases in quality of life and enjoyment. These results suggest that the behavioral prescription treatment was at least as efficacious as the usual care treatment, the second hypothesis evaluated by this study.

Clinical significance was assessed using the following two methods: (1) by examining the proportion of subjects who reported, and continued to report, clinically significant symptom levels (i.e., BDI-FS ≥ 4) at each time point; and (2) by using Jacobson and Truax’s (1991) reliable change index to determine what proportion in each condition reported clinically significant improvement from one time period to the next and from baseline to 1 and 3 month follow-up points. A reliable change index value of 2.821 was calculated based on the test–retest reliability of the BDI-FS (r = .75). Evaluation of group means indicated that BDI-FS scores in both treatment conditions reliably improved from baseline to posttreatment. A series of chi-square analyses was used to determine clinically significant differences between the behavioral prescription and usual care groups at all time points (Fig. 1).

Figure 2 displays the percentage of patients who continued to report clinically significant symptom levels of depression (i.e., BDI-FS ≥ 4). Chi-square analyses revealed non-significant differences between groups at all time-points. These results suggest that treatment gains within the behavioral prescription group were similar to those in the usual care group, providing additional support for the second hypothesis of the study.
https://static-content.springer.com/image/art%3A10.1007%2Fs10880-010-9207-2/MediaObjects/10880_2010_9207_Fig1_HTML.gif
Fig. 1

Mean fixed predicted BDI-FS scores

https://static-content.springer.com/image/art%3A10.1007%2Fs10880-010-9207-2/MediaObjects/10880_2010_9207_Fig2_HTML.gif
Fig. 2

Percentage of patients reporting clinically significant depression symptoms

Medical billing records for each participant in both conditions were reviewed between baseline and three-month follow-up points in order to compare medical costs and variables associated with medical expenditures (e.g. number of physician contacts, emergency room visits, inpatient hospitalization days, medical problems). For participants with incomplete data, their medical records were reviewed as if they had completed the study. Thus, all patients originally randomized into the study were included in these analyses.

All variables related to cost were significantly correlated at the p < .001 level. Table 4 displays means and standard deviations of these variables. A multivariate analysis of covariance (MANCOVA) was used to analyze the multiple cost measures as a function of treatment. Gender and age (two variables that correlate with increased depression symptoms and medical costs) were included as covariates. Results showed a non-significant multivariate F for the Treatment effect. Post hoc univariate F tests also showed no significant differences between treatment conditions on any of the individual cost measures. However, the mean difference in medical costs per person was 37.2% less for patients in the behavioral prescription condition as compared to those in usual care. This is equivalent to a total mean savings of $281.32 per patient [F(1, 34) = 3.222, p = .08]. These results provide statistical evidence that the behavioral prescription was at least as cost-effective as usual care, supporting the third hypothesis of the study.
Table 4

Means and standard deviations for variables related to medical cost

Variable

M (SD)

Behavioral prescription (n = 19)

Usual care (n = 19)

Total medical costs

$476.20 (269.2)

$757.52 (726.8)

Physician contacts

3.3 (1.8)

4.4 (4.0)

ER visits

0.1 (0.3)

0.3 (0.9)

Inpatient hospital days

0.4 (0.8)

0.8 (2.4)

Medical problems during study

5.6 (7.1)

7.1 (5.1)

Physician satisfaction was measured through responses to questions that were developed specifically for this study. Table 5 displays the physician mean responses. Results demonstrate a 0% difference between the mean physician satisfaction responses, suggesting physicians are equally satisfied with prescribing a behavioral prescription to a medication, a finding that supports the fourth hypothesis of the study.
Table 5

Means and standard deviations for physician satisfaction

Prescription type

M (SD)

Behavioral prescription (n = 6)

5.8 (1.2)

Medication (n = 6)

5.8 (1.2)

Note: Satisfaction was measured by physicians responding to the followings questions at the end of the study: (1) “How satisfied are you with prescribing medication as a treatment for depression?” and (2) “How satisfied are you with prescribing a behavioral prescription for Feeling Good as a treatment for depression?” Physicians answered these questions using a Likert scale ranging from 1 (very dissatisfied) to 7 (very satisfied)

Satisfaction levels for the patient treatment groups were measured by responses to questions that were developed specifically for this study. One-way ANOVAs were used to determine whether or not differences in group means were statistically significant. No significant differences between the two patient group mean responses on either satisfaction question were detected. These results demonstrate that patient satisfaction within the behavioral prescription condition was equivalent to patient satisfaction within the usual care condition, evidence that supports the fifth hypothesis of the study.

Discussion

The primary outcome findings from this study suggest that a physician-delivered prescription for bibliotherapy was no less effective than usual care in reducing symptoms of depression, decreasing dysfunctional attitudes, and increasing life satisfaction and enjoyment. Results from this study suggest that a bibliotherapy behavioral prescription can be practically and feasibly delivered by physicians as a treatment for depression in real world primary care settings with medical patients. However, no efficacy beyond that of usual care was observed, as no significant differences were found between the two treatment groups on the primary outcome variables. Despite some of the methodological limitations of this study, these results suggest that both physicians and patients responded as favorably to a behavioral prescription for depression as they did to standard medical care. Significant improvements were observed over time on multiple treatment outcome variables. Considering the expense of antidepressant medication, its association with aversive side effects, contraindication for some patients, and its lack of effectiveness for others, these results suggest that physicians may have the option of another empirically supported treatment. The results of this study are promising enough to warrant controlled trials with larger samples.

Behavioral prescriptions were also found to be as cost-effective as usual care when comparing the medical cost-offset associated with the two interventions. Although analyses of direct medical expenditures demonstrated no statistical differences between the medical expenditures of the two groups; the mean net medical expenses in the behavioral prescription group was substantially less, a difference that may have proven to be significant with a larger sample size. In terms of direct costs, the book at $7.95 has clear cost advantages over a prescription for an antidepressant, which could cost around $25 per month (e.g. for a typical dose of a generic form of Paxil), or as much as $100 per month (e.g. for a typical dose of Paxil).

Although side effects were not tracked in this study, prior research (TADS, 2004, 2007) suggests that a cognitive-behavioral psychological intervention has substantially fewer psychiatric and non-psychiatric adverse events in comparison to a psychotropic treatment. Considering the medical complications and contraindications that afflict many primary care patients, offering physicians and patients an alternative empirically supported treatment seems warranted and needed. Other assets of a non-pharmacological bibliotherapy prescription over a medication treatment for depression are that this type of treatment can be revisited time and time again without additional cost, shared with family members and friends, and allows patients to attribute their improvement to their efforts rather than a pill. Anecdotally, several of the patients who reported a positive response to the behavioral prescription intervention indicated that they had lent their Feeling Good books to other friends and family members who they thought might benefit from the intervention.

This study had a number of methodological limitations. The findings must be interpreted with caution. The lack of a no treatment control group and small sample size may have contributed to a lack of main effects found for treatment. Additionally, we are unaware of a test of equivalence of repeated measures in a mixed model, and therefore were not able to state definitively that the treatment conditions produced equivalent results. However, we were able to conduct a test of available power (PASS; Hintze, 2008). Given the number of subjects in each condition, and the repeated measures design, the current study had the power to detect a small-to-moderate difference (d = 0.4) between conditions over time with a power of .82 and alpha set at .05. That is, if a small systematic difference had existed it would most likely have been detected.

Although time appeared to be a significant causal factor of positive treatment outcomes, it is unknown whether the treatment delivered in each group contributed to any of these improvements, given the lack of treatment and treatment × time effects. Thus, the large effects found for time may have had little to do with the physician-delivered interventions. Weekly contact from a mental health professional during the treatment phase was probably more frequent than standard usual care in most primary care environments, which limits the generalizability of the condition and may have contributed to some of the significant improvement found. Another drawback to the study that likely contributed to a decrease in statistical power was the within group variability in the usual care condition. Although medication was tracked and analyzed independently, the small sample may have contributed to a lack of any statistically significant findings. A small physician sample and the lack of physician adherence data collected were other limitations. Physician–patient interaction variables that were not tracked due to an inability to video-tape physician–patient interactions during treatment delivery may have included supportive listening and recommendations to engage in behaviors other than reading Feeling Good (e.g. exercise, a vacation, less work, etc.). These may have contributed to some of the significant improvements in patient outcomes. Patient self-selection, spontaneous remission, and response biases may also have confounded the results.

The lack of a standardized diagnostic interview for depression, like the HAM-D or SCID, prevents a comparison of the study findings to other studies, including randomly controlled antidepressant studies. Thus, the results of this study can only be generalized to patients complaining of depression symptoms as indicated by the BDI-FS, and not those who have been diagnosed with a depressive disorder according to the DSM-IV criteria. Unfortunately, a diagnostic interview required too much time and was deemed not practical or feasible for the fast-paced and real world primary care study setting. In addition, patients in primary care are rarely comprehensively assessed for clinical depression using a standardized diagnostic interview. The non-standardized measures that were used to measure satisfaction and adherence also limit the generalizability of the results. A lack of appropriate standardized measures for these variables accounted for the use of these measures.

Lastly, there are several reasons why a behavioral prescription for depression might not be the best treatment for some patients. First, some patients may not respond particularly well to written materials, due to their learning style or limited language skills. These patients may be more responsive to other auditory, visual, physical, or interactive media. Second, patients with reading disabilities, who are unable to speak or read English at a sixth-grade reading level, or may have significant time constraints, would also likely be poor candidates for a bibliotherapy prescription. It should also be noted that the length of the cognitive-behavioral portion of Feeling Good may be too burdensome for some patients and thus, limit the effectiveness of the intervention. Some participants’ inability to finish the required portion of reading within the 6 week intervention phase is testament to this limitation. However, given that there were no significant differences found between treatment outcome of the two groups and the bibliotherapy condition did show significant improvement between baseline and posttreatment on several outcome variables, a prescription for Feeling Good may be an effective option for many.

Future research on physician-delivered behavioral prescriptions for depression could target efficacy, and therefore, be expanded in a number of ways. Different behavioral prescription conditions (e.g. bibliotherapy, relaxation training, social/assertiveness skills training, aerobic exercise, pleasant activities) could be compared to an antidepressant condition in order to evaluate the efficacy of a multitude of behavioral prescriptions by using a 2 × 2 design (Prescription Type × Amount of Contact). High contact patients could receive weekly phone contact to monitor progress and support their adherence with the behavioral prescription or an antidepressant prescription. Low contact patients could be contacted only at the beginning and at the end of treatment, more accurately simulating the real world. Research targeting physician-delivered behavioral prescriptions that utilize empirically tested technologies other than reading could also be considered in order to match treatments to patient preferences. These may include interactive websites, DVDs, and other media. Furthermore, a prescription for Feeling Good could be tested as an adjunctive treatment to medication for those who are uncomfortable with discontinuing their medication, but would like to enhance the effectiveness of their treatment. Finally, future studies with larger sample sizes could assess the dosage effect of Feeling Good to determine whether or not total pages read is a statistically significant variable related to treatment outcome.

Copyright information

© Springer Science+Business Media, LLC 2010