Determinants of Attrition From Cessation Treatment in Smokers With a History of Major Depressive Disorder

Abstract

Attrition from smoking cessation treatment by individuals with a history of major depression was investigated. An investigation of preinclusion attrition examined differences between eligible smokers who did (n = 258) and did not (n = 100) attend an initial assessment session. Postinclusion attrition was investigated by comparing early dropouts (n = 33), lale dropouts (n = 27), and treatment completers (n = 117). Those who failed to attend the assessment session were more likely to be female, to smoke cigarettes with higher nicotine content, and to have a history of psychotropic medication use. Early-treatment dropouts reported a higher smoking rate than late-treatment dropouts and endorsed more symptoms of depression than late dropouts and treatment completers. Results are compared with previous investigations of smoking cessation attrition, and implications for treatment and attrition prevention are discussed.

Attrition from treatment poses a considerable problem for both researchers evaluating interventions (Howard, Cox, & Saunders, 1990) and clinicians delivering treatment (Stark, 1992). In studying attrition, investigators have found it useful to distinguish between preinclusion attrition—which occurs prior to entering a study, during screening or during intake evaluations—and postinclusion attrition, which occurs during treatment or posttreatment follow-ups (Howard et al., 1990). For example, evaluations of smoking cessation programs have noted preinclusion attrition rates of 30%–50% (Curry, Marlatt, Gordon, & Baer, 1988; Kviz, Crittenden, Madura, & Warnecke, 1994; Hall, Muñoz, & Reus, 1994) and postinclusion attrition rates that range from less than 10% (Curry, Thompson, Sexton, & Omenn, 1989; Zelman, Brandon, Jorenby, & Baker, 1992) to approximately 50% (Curry et al., 1988; Klesges et al., 1988).

Although they are significant, these rates of attrition may be even higher in selected high-risk groups of smokers, such as those with psychiatric comorbidity. Disproportionately high rates of psychiatric comorbidity with cigarette smoking have been found in both adolescent (Brown, Lewinsohn, Seeley, & Wagner, 1996) and adult (Glassman et al., 1990) community samples. In adult smoking cessation programs, rates of past major depressive disorder (MDD) have ranged from 31% (Hall et al., 1994) to 61% (Glassman et al., 1988). In cessation treatment, smokers with a history of major depression have been shown to have elevated negative mood at pretreatment (Ginsberg, Hall, Reus, & Muñoz, 1995; Hall et al., 1994), to experience mood disturbance following cessation (Covey, Glassman, & Stetner, 1990; Ginsberg et al., 1995), and to relapse at higher rates than smokers without past MDD (Glassman et al., 1988; Glassman, 1993).

As the overall prevalence of cigarette smoking decreases (Centers for Disease Control, 1994), those remaining smokers are likely to have more difficulty quitting because of factors such as psychiatric comorbidity and high nicotine dependence (Coambes, Kozlowski, & Ferrence, 1989; Hughes, 1993). Smoking cessation programs are increasingly faced with the prospect of delivering treatment to these subpopulations of high-risk smokers, and thus it will be critically important to understand the factors that contribute to their active participation in smoking cessation treatment.

In the present study we examined the determinants of attrition hi the context of a treatment outcome study for smoking cessation in individuals with a history of MDD. We examined preinclusion attrition in smokers who appeared to be eligible for participation on the basis of a telephone screening. We compared demographic and smoking-related characteristics for individuals who attended an initial diagnostic interview to determine eligibility versus those who did not attend. We examined postinclusion attrition by comparing early-treatment dropouts, late-treatment dropouts and treatment completers on baseline measures of demographics, smoking behavior, depressive symptoms, and social–cognitive variables.

Method

Participants

Participants were recruited over a 3-year period from the community by means of newspaper, radio, and television advertisements promoting a medication-free program to quit smoking as well as by flyers distributed to local health care professionals. The 1,244 participants who requested information about the program were first screened with a brief telephone interview to assess smoking behavior, current and past depression, alcohol and drug use, participation in psychotherapy, and use of psychotropic medication. Participants who were potentially eligible on the basis of this screening (N = 358; 29%) were invited to the study center for a diagnostic interview to further assess eligibility. The mean age of these 358 individuals was 44.5 years (SD = 10.0), and 58.9% were female.

Of the 358 individuals invited, 258 (72%) attended the diagnostic interview. We administered the Structured Clinical Interview for DSM–III–R—Non-Patient edition (SCID-NP, Version 1.0; Spitzer, Williams, Gibbon, & First, 1990) to confirm history of MDD according to the Diagnostic and Statistical Manual of Mental Disorders (3rd ed., rev. [DSM–III–R]; American Psychiatric Association, 1987) and to verify inclusion and exclusion criteria. To qualify, participants met the following inclusion criteria: (a) between 18 and 70 years of age, (b) regular cigarette smoker for at least 1 year, (c) currently smoking at least 10 cigarettes per day, (d) past history of MDD, (e) no current weekly individual psychotherapy or use of psychotropic medication, (f) use of no other tobacco products, and (g) not planning to use nicotine replacement or other pharmacological aids to smoking cessation. Participants were not eligible if the following exclusion criteria applied: (a) current MDD, dysthymia, or other Axis I disorder and (b) current psychoactive substance abuse or dependence (other than nicotine) within the past 6 months. Of the 258 participants who took part in the diagnostic interview, 74 participants did not meet inclusion-exclusion criteria (59% did not have a history of MDD, 19% met criteria for current MDD, 22% met criteria for other exclusionary diagnoses), and 5 withdrew prior to being assigned to a treatment condition.

The remaining 179 participants participated in the smoking cessation treatment. Of these 179, 2 participants could not be classified as early dropouts, late dropouts, or treatment completers (see postinclusion attrition results for classification rules), leaving 177 (69% of those who attended the in-person assessment) participants for inclusion in the postinclusion attrition analyses. On average, these 177 participants were 45.2 (SD = 9.3) years of age and smoked 27.2 (SD = 11.2) cigarettes per day. The median number of previous quit attempts lasting at least 12 hr was 4.0. The majority were female (59.3%), married (41.2%; 28.8% divorced, 10.2% single), and Caucasian (97.7%).

Procedure

After participating in both the telephone screening and the diagnostic interview, eligible and consenting participants were assessed on a variety of interview, self-report, and biochemical measures. The interview and self-report measures included demographics, smoking behavior and history, depressive symptoms, and social–cognitive measures. As part of a treatment outcome study investigating the efficacy of adding cognitive–behavioral treatment for depression to standard smoking cessation treatment (Brown et al., 1997), participants were then randomly assigned to one of two group treatment conditions: standard smoking cessation treatment (ST) or cognitive–behavioral treatment for depression plus standard smoking cessation (CBT-D). Both treatments consisted of eight 2-hr sessions over 6 weeks. The quit date occurred on the fifth session. Groups met weekly, except for the sixth session, which took place 3 days after the quit date.

The ST condition included self-monitoring, self-management, nicotine fading, a designated quit date, identifying and using social support, and relapse prevention training components. The CBT-D condition included all the components of the ST condition integrated with depression coping skills training using a modified version of the Coping With Depression Course (CWD; Brown & Lewinsohn, 1984). The depression skills component involved treatment based on a social learning model of depression and included daily mood monitoring, increasing pleasant activities, changing negative and nonconstructive ways of thinking, and social skills/assertiveness components. The depression coping skills were incorporated into every session of the CBT-D condition and were presented as viable alternatives to smoking that could control feelings of depression following the loss of smoking as a reinforcing activity.

Measures

Telephone screening measures assessed demographic variables, such as age and gender, as well as smoking behavior (number of cigarettes smoked daily, number of minutes to first cigarette after waking in the morning, number of years smoking, and cigarette nicotine level). Past and present alcohol and drug use and abuse and history of psychotropic medication use and past counseling or psychotherapy were based on self-report responses to yes–no questions (e.g., “Do you currently drink alcohol?”).

Baseline measures were administered at pretreatment. A battery of questionnaires assessed demographic and smoking behavior/history variables, including age, gender, years of education, income, number of cigarettes smoked daily, number of previous quit attempts lasting at least 12 hr, and number of years smoking. We assessed severity of nicotine dependence using the Fagerstrom Test of Nicotine Dependence (FTND; Heatherton, Kozlowski, Frecker, & Fagerstrom, 1991), which is derived from the Fagerstrom Tolerance Questionnaire. The FTND is a brief, 6-item measure. Scores range from 0 to 10, with higher scores indicating higher levels of nicotine dependence. Saliva samples for cotinine analysis were collected as a biochemical measure.

The Beck Depression Inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) is a reliable and valid 21-item self-report measure of depressive symptoms. Total scores range from 0 to 63; higher scores represent greater depression. The Hamilton Rating Scale for Depression (HAM-D; I. Miller, Norman, & Bishop, 1985) is a 25-item interview-administered rating scale that has also been shown to possess good reliability and validity (I. Miller et al., 1985). Scores range from 0 to 72, with greater scores representing greater depression.

The Decisional Balance Scale (short form) is a 6-item measure of the pros and cons of smoking (Velicer, DiClemente, Prochaska, & Brandenburg, 1985). Participants endorsed agreement with each item (e.g., “Smoking cigarettes relieves tension”) on a 5-point scale (range: 1 = not at all to 5 = very much). The scale was divided into Pros and Cons subscales, both of which have high internal validity (αs = .88 and .89, respectively; Velicer et al., 1985). The cons were subtracted from the pros to yield a difference score, with positive scores indicating a greater number of pros and negative numbers indicating a greater number of cons. We assessed smoking cessation self-efficacy using the 9-item short form of the Smoking Situations Confidence questionnaire (Velicer, DiClemente, Rossi, & Prochaska, 1990), which has demonstrated good validity and reliability (αs = .80–.90; Velicer et al., 1990). The questionnaire was scored to yield an average self-efficacy index ranging from 1 to 5, with higher scores reflecting greater self-efficacy.

Results

Preinclusion Attrition

Approach to analyses

We examined preinclusion attrition by comparing smokers who appeared eligible on the basis of the telephone screening but did not attend the initial assessment session (nonattendees; n = 100) with those who did attend the initial assessment session (attendees; n = 258). We used demographic variables, smoking behavior, present and past substance use and abuse, and psychiatric treatment participation—assessed during the initial phone screening—to predict attrition prior to treatment using analysis of variance (ANOVA) for continuous dependent variables and chi-square analyses for categorical dependent variables. See Table 1 for means, standard deviations, and percentages on telephone screening measures.

Table 1.

Telephone Screening Variables as a Function of Assessment Session Attendance

Measure	Attendeda	Did not attendb	Between-group differences
Gender	55.4% female	68.0% female	χ2(1, N = 358) = 4.7*
Age (years)c	45.0 (9.5)	43.3 (11.0)	ns
Minutes to first cigarettec	19.0 (38.5)	21.1 (45.6)	ns
No. cigarettes/dayc	27.9 (11.9)	27.3 (12.7)	ns
Years regular smokerc	25.6 (11.5)	24.6 (10.5)	ns
Cigarette nicotine levelc	.79 (.33)	.89 (.34)	F(1, 325) = 6.18*
Currently using alcohol (%)	67.3	67.7	ns
Currently using  nonprescription drugs (%)	6.7	5.1	ns
Past alcohol problem (%)	30.9	32.9	ns
Past nonprescription drug  problem (%)	13.3	15.2	ns
Past use of psychiatric  medication (%)	44.3	56.6	χ2(1, N = 352) = 4.3*
Current psychotherapy (%)	17.3	10.1	ns
Past psychotherapy (%)	71.5	64.2	ns

^an = 258.

^bn = 100.

^cM (SD).

^*p < .05.

Preinclusion attrition results

A chi-square analysis indicated that women were significantly less likely to attend the in-person assessment (68%) than were males (32%), χ²(1, N = 358) = 4.71, p < .03. No differences were found between attendees and nonattendees in terms of age, F(1, 356) = 2.06, ns; minutes to first cigarette after waking, F(1, 309) = 0.17, ns; daily number of cigarettes, F(1, 356) = 0.20, ns; or years of regular smoking, F(1, 343) = 0.55, ns. Nonattendees did, however, report smoking cigarettes with significantly higher nicotine levels compared to attendees, F(1, 325) = 6.18, p < .05.

Attendees and nonattendees did not differ in current use of alcohol, χ²(1, N = 356) = 0.004, ns, or current use of nonprescribed or street drugs, χ²(1, N = 352) = 0.31, ns, variables that were assessed with yes–no questions. Similarly, attendees and nonattendees did not differ on yes–no responses to questions concerning past problems with alcohol, χ²(1, N = 322) = 0.11, ns, or past problems with nonprescribed or street drugs, χ²(1, N = 340) = 0.21, ns.

Participants who answered “yes” to the question “Have you ever taken any psychiatric, nerve or mood medicine?” were significantly less likely to attend the assessment (57%) than participants who responded “no” to this question (43%), χ²(1, N = 352) = 4.31, p < .05. On the other hand, attendees did not differ from nonattendees on current interventions other than individual or past psychotherapy, χ²(1, N = 353) = 2.87, ns, or past participation in therapeutic interventions, χ²(1, N = 334) = 1.73, ns.

Postinclusion Attrition

Approach to analyses

We examined post-inclusion attrition by comparing early dropouts (N = 33, 19%), late dropouts (N = 27, 15%), and treatment completers (N = 117, 66%). These categories were based on the structure of the program using the quit date as a defining mark as well as examination of the distribution of sessions attended. Early-treatment dropouts attended at least one session and no more than four sessions and did not attend Sessions 5 (quit date), 6, 7, or 8. Late-treatment dropouts attended fewer than six sessions with at least one of those sessions being Sessions 5–8. Treatment completers attended six or more total sessions and at least one of the two final treatment sessions (Session 7 or Session 8). Two of the 179 participants who participated in treatment could not be classified according to the above definitions (1 participant attended one individual makeup session but no group sessions; the other attended one session prior to the quit date and one session after the quit date). Thus, we conducted postinclusion attrition analyses on the 177 classified participants.

We tested predictors of postinclusion attrition by comparing early-treatment dropouts, late-treatment dropouts, and treatment completers on baseline measures. Chi-square analysis, χ²(2, N = 177) = 1.26, ns, revealed that dropout status (early dropouts, late dropouts, treatment completers) was not associated with treatment condition (ST vs. CBT-D); therefore the results reported here were collapsed across treatment condition. With the exception of gender (chi-square) and age (ANOVA), all measures were evaluated for systematic differences as a function of dropout status (early dropouts, late dropouts, treatment completers) using multivariate analyses of variance (MANOVAs).

Demographics

Postinclusion attrition did not vary reliably as a function of gender, χ²(2, N = 177) = 0.94, ns, or participant age, F(2, 174) = 0.10, ns. Similarly, income and education, as measures of socioeconomic status, did not differ among early dropouts, late dropouts, and treatment completers, Pillai’s statistic = 0.02, approximate F(4, 340) = 0.68, ns.

Smoking

Baseline FTND, cotinine levels, and number of cigarettes smoked per day during the week prior to starting treatment were tested together as indicators of nicotine dependence. No significant multivariate differences were found as a function of dropout status, Pillai’s statistic = 0.06, approximate F(6, 344) = 1.61, ns. However, a significant univariate effect emerged for number of cigarettes smoked per day, F(2, 173) = 3.85, p < .05. Early dropouts smoked more cigarettes per day than treatment completers (p < .05 using the Scheffé test). Years of smoking and number of previous quit attempts were similarly evaluated together as measures of smoking history, yielding no significant differences as a function of dropout status, Pillai’s statistic = 0.01, approximate F(4, 346) = 0.37, ns. See Table 2 for descriptive statistics on variables across dropout status.

Table 2.

Means and Standard Deviations of Baseline Variables by Dropout Status

	Early dropoutsa		Late dropoutsb		Treatment completersc
Measure	M	SD	M	SD	M	SD
Demographic variables
% Female	66.7		59.3		57.3
Age	44.6	10.8	45.0	8.4	45.4	9.1
Household income scaled	3.9	1.8	4.3	2.5	4.3	1.9
Highest grade completed	14.3	3.0	14.1	1.9	14.7	2.5
Smoking variables
No. cigarettes/day	32.0a	15.1	27.5	7.8	25.8b	10.3
FTND	7.51	1.92	6.78	2.41	6.62	1.77
Cotinine level	377.1	132.2	382.1	127.3	386.8	189.7
Years smoking	27.4	11.0	26.1	8.5	27.1	9.4
No. previous quit attempts (Mdn)	3.0		4.0		4.0
Depression variables
BDI	10.9a	7.1	4.7b	4.2	6.6b	6.2
HAM-D	7.0a	6.3	3.9	3.7	4.5b	4.7
Social–cognitive variables
Self-efficacy	2.1	0.6	2.3	0.8	2.2	0.7
Pros–cons	−0.76	1.2	−0.65	1	−0.65	1

Note. Means with different subscripts are statistically different from each other in the same row at p < .05.

FTND = Fagerstrom Test of Nicotine Dependence; BDI = Beck Depression Inventory; HAM-D = Hamilton Rating Scale for Depression.

^an = 33.

^bn = 27.

^cn = 117.

^d1 = 0–$9,999/year; 2 = $10,000–$19,999/year; 3 = $20,000–$29,999/year; 4 = $30,000–$39,999/year; 5 = $40,000–$49.999/year; 6 = $50,000–$59,999/year; 7 = over $60,000/year.

Depressive symptoms

We evaluated symptoms of depression with a MANOVA, using both the BDI and modified HAM-D as dependent variables. A significant multivariate effect of dropout status emerged, Pillai’s statistic = 0.09, approximate F(4, 348) = 4.31, p < .01. Early dropouts endorsed a significantly greater level of depressive symptoms than both late dropouts and treatment completers on the BDI (p < .05 using the Scheffé test), and early dropouts reported a significantly greater level of depression than treatment completers on the modified HAM-D (p < .05 using the Scheffé test).

Social–cognitive variables

The social–cognitive variables we considered included the difference between the pros and the cons of quitting as well as perceived confidence in ability to quit. No significant differences in these social–cognitive variables were found as a function of dropout status; Pillai’s statistic = 0.02, approximate F(4, 348) = 0.64, ns.

Correlational analyses

We also examined determinants of postinclusion attrition through correlations between the number of sessions attended and pretreatment variables, because of the lack of a consistent definition of dropout in the literature (Breteler, Rombouts, & van der Staak, 1988). These correlational analyses corroborated the previously presented analyses. Greater daily pretreatment smoking (r = −.22, p < .01) and greater endorsement of depressive symptoms (BDI: r = −.25, p < .01; HAM-D: r = −.22, p < .01) correlated significantly and negatively with number of treatment sessions attended. Gender, age, household income, highest grade completed, FTND, cotinine level, years smoking, previous quit attempts, difference between pros and cons of quitting, and perceived confidence to quit smoking failed to correlate significantly with number of sessions attended.

Discussion

Participants who did not attend an in-person assessment after a telephone screening were significantly more likely than those who did attend to be female, to have taken psychiatric medication in the past, and to smoke cigarettes with a higher nicotine content. With regard to postinclusion attrition in this sample of smokers with a history of MDD, early dropouts endorsed significantly more symptoms of depression and smoked significantly more cigarettes per day at baseline than treatment completers.

Twenty-eight percent of potential participants did not attend the in-person assessment session, a rate similar to preinclusion attrition reported in other studies (Curry et al., 1989; Kviz et al., 1994). Women were significantly less likely to attend the in-person assessment session than men, a finding consistent with that noted in a marijuana dependence treatment outcome study (Curtin, Stephens, Cleaveland, & Roffman, 1993). Nonattendees also reported using tobacco products with higher nicotine content than attendees. Although differences were not found in regard to number of cigarettes smoked per day or minutes until first cigarette after waking, these participants may represent a more nicotine-dependent subsample. We did not support past findings that younger smokers (Zbikowski, Williamson, Klesges, & Eck, 1995) or smokers who had smoked fewer years (Klesges et al., 1988) were less likely to participate; however, this is not greatly unexpected given our high-risk sample of participants who, on the basis of the telephone screening, appeared to have experienced a history of MDD and who smoked more cigarettes per day than participants in previous evaluations of preinclusion attrition (Klesges et al., 1988; Zbikowski et al., 1995).

Although women were less likely than men to attend the in-person assessment, women were no more likely than men to drop out after beginning treatment (Garfield, 1986; Stark, 1992). A lack of environmental support, such as availability of transportation or day care, may represent more of a barrier to women than to men in initiating treatment. It is also possible that higher rates of depression in females (Kessler et al., 1994) contribute to their higher rate of pre-inclusion as well as postinclusion attrition. Because degree of current depressed symptoms was not assessed during the phone screening, we were not able to test this hypothesis. We did find, however, that women were more likely than men to have taken psychotropic medication in the past, with antidepressant medication being the most frequent type. Overall, nonattendees may be contemplating change and beginning to explore options, as evidenced in their inquiring about treatment, but they may not be ready for action-oriented treatment (Prochaska, DiClemente, & Norcross, 1992).

The current postinclusion attrition rate of 34% fell within the range of previously reported rates of 10% (e.g., Zelman et al, 1992) to 50% (e.g., Curry et al., 1988). Inconsistencies in definitions of attrition and treatment completion (Breteler et al., 1988) limit direct comparisons and likely contribute to this wide range of attrition rates. Rigorous follow-up procedures may reveal limited attrition rates in the present sample of smokers who would be expected to be at high risk for dropping out of treatment. Consistent with determinants of attrition from smoking and other drug intervention studies, early dropouts reported smoking significantly more cigarettes per day than treatment completers (Hansen, Collins, Malotte, Johnson, & Fielding, 1985), and they endorsed significantly more depressive symptoms than late dropouts or treatment completers (Stark, 1992). Identification of depressive symptoms and heavier smoking as risk factors for attrition with this high-risk sample of smokers is a replication of similar findings and suggests that these are key determinants of attrition and fairly robust findings. Given that depressive symptoms predicted early dropout, one might have expected a between-group difference on dropout status, because the CBT-D included the addition of depression coping skills. However, in light of the fact that the CBT-D was intended to prevent increased depressive symptoms on quitting (Week 4 and beyond), CBT-D participants most likely did not receive an adequate “dose” of the depression coping skills, which may account for this lack of a group difference.

The present results suggest that difficulties in retaining participants who smoke more on a daily basis and who report greater symptoms of depression may challenge the effectiveness of current smoking cessation programs. Equivalent attrition rates across the two treatment conditions suggest that experiencing symptoms of depression may interfere with active participation in any goal-oriented treatment. Individuals with higher levels of depressive symptoms may require treatment for these symptoms prior to, rather than concurrent with, participation in smoking cessation interventions. Attrition-prevention interventions, such as brief motivational enhancement techniques (W. R. Miller & Rollnick, 1992), pregroup preparation (Yalom, 1985), between-session telephone or e-mail contact, or aggressive follow-up procedures for missed appointments may prove helpful.

The present findings and comparisons of them to the results of similar investigations are limited by a lack of an accepted definition of dropout (Breteler et al., 1988). Furthermore, several of the variables were assessed with single-item, face-valid questions with unknown reliability and validity. Nevertheless, the present results identify characteristics that appear to heighten the risk of pre- and postinclusion attrition. Future investigations should attempt to replicate and extend our current understanding of risk factors for attrition, investigate attrition-prevention interventions, and explore the relationship between these risk factors and long-term outcomes. For example, further investigations of the match between client variables, such as negative affect, and specific intervention components, such as supportive counseling (Zelman et al., 1992), may guide the development of interventions for groups that are difficult to retain and treat Similarly, extending outcome research to process research (e.g., examining the relationship between molecular measures, such as weekly measures of mood or social support, and treatment participation) may further elucidate factors that place smokers who already are at high risk at an even greater risk to continue smoking.