Introduction

Considerable research has investigated basing therapeutic interventions on a patient’s readiness to change a behavior (“stages of change”) across health behaviors such as smoking, exercising, and dietary fat intake [1–3]. Stages of change are distinct categories and range from Precontemplation to Maintenance; individuals can move from one stage to another. The Precontemplation stage includes individuals who are unaware of having a problematic behavior or who for other reasons are not thinking seriously about changing. Those in the Contemplation stage are aware that a problem exists, are thinking about changing, but have not yet made a commitment to change. Individuals in the Preparation/Action stage have a plan of action and perceive the benefits of changing as outweighing the costs. Prochaska and colleagues [4–6] investigated whether stages of change predicted treatment outcome in a smoking cessation study and found that they could categorize treatment failures and successes based upon the participants’ readiness (Precontemplation, Contemplation, Preparation/Action).

These ideas have been applied to research on the treatment of eating disorders (ED). Generally, results have indicated that patients in higher readiness to change categories show greater improvement in symptoms and cognitions than those in lower stages [7–11]. For example, in one study of residential treatment for patients with anorexia nervosa (AN), those with a higher readiness for change were more likely to achieve favorable short-term outcomes within a shorter length of stay than those with a lower readiness for change, who subsequently required a longer stay in order to achieve favorable outcome [12]. However, one concern is that the vast majority of studies investigating whether stages of change predict treatment outcome have only studied patients with AN [7, 12, 13] and have found mixed results. For example, one study of inpatients being treated for AN compared treatment-as-usual (TAU) to TAU with four sessions of motivational interviewing and found that both conditions showed improvement across the follow-up periods and there were no differences on ED symptom outcomes between treatment conditions, although those receiving motivational interviewing had a lower dropout rate and seemed to improve in their readiness for change compared to those receiving TAU without motivational interviewing [7]. In contrast, a separate study of patients with AN found that readiness to change significantly predicted treatment outcome, even when controlling for severity of the ED, in part because ED severity was also associated with readiness for change [13]. The authors concluded that treatment may be more effective by first resolving the patient’s ambivalence about recovery [13].

Only a few studies have reported results with other ED diagnostic presentations [9, 14, 15]. Specifically, among the studies of patients with diverse ED diagnostic presentations, measurement has been limited to readiness to change the eating disorder “overall” or to readiness to change efforts at dietary restraint [14, 15]; these decisions were founded, in part, on the basis that restriction precontemplation scores on the Readiness and Motivation Interview have been shown to be the most reliable predictor of treatment outcome [9]. Past research has found that interview methods provide the symptom-specific discrimination to evaluate SOC across individual symptoms [16, 17], yet interview methods require a significant amount of staff time in training and administration and therefore may not be a practical choice for assessing readiness to change among a wide variety of research and clinical settings. Use of a self-report, symptom-specific measure of SOC in eating disorders would allow for efficient evaluation of SOC, however, it is not yet known whether a symptom-specific self-report measure has adequate predictive utility.

For the many patients who engage in multiple ED behaviors, it is important to assess SOC separately for each behavior (e.g., binge-eating, purging, fasting), instead of grouping behaviors (e.g., binge-eating and purging) because patients may be at different stages for different behaviors. Furthermore, many patients experience diagnostic crossover during the years they have an eating disorder and therefore use of self-report measures of stages of change for a single or small subset of symptoms, particularly across longer periods of treatment or in longitudinal studies, could miss critical information if new symptoms were adopted while others were ceased. Wolk and Devlin [11] found that assessing different behaviors with a simple questionnaire facilitated efficient assessment of SOC. Ackard and colleagues [18] expanded on Wolk and Devlin’s assessment strategy and developed a reliable and valid self-report measure of SOC (eating disorder stage of change; EDSOC) across eight specific and separate ED behaviors (binge-eating, fasting, restricting, self-induced vomiting, laxative use, diet pill use, diuretic use, metabolic enhancement use). Using a self-report assessment allows for greater ease of evaluating changes in SOC during the course of treatment, and using an assessment that allows for measurement of specific ED symptoms captures information on diverse diagnostic presentations. Because SOC can be different across different symptoms and can change over the course of treatment, the use of an efficient self-report assessment facilitates easier translation of results into clinical practice and thus can help clinicians focus their assessments and tailor treatment interventions if indicated.

The objectives of this study are: (1) to determine the prevalence of different SOC (Precontemplation, Contemplation, and Preparation/Action) at intake assessment among individuals seeking ED treatment both across individual symptoms and across an aggregate “overall” stage of change; and (2) to evaluate whether symptom-specific and overall SOC at intake assessment predict treatment outcome in a treatment-as-usual clinical setting. We hypothesize that patients will present for ED treatment across the spectrum of SOC and that patients in Preparation/Action at the beginning of treatment will experience greater behavior and emotional change at 3, 6, and 12 months following their intake assessment compared to those in Precontemplation or Contemplation.

Methods

Procedure

Eligible patients were all male and female patients 13 years of age or older who presented for eating disorders treatment at the Park Nicollet Melrose Center between 2005 and 2007. They were invited at their intake assessment to the eating disorders treatment facility (Park Nicollet Melrose Center) to participate in the study in person or by a mailed invitation letter and study-specific consent form. The study-specific consent form provided permission for research staff to contact the patient and send a battery of questionnaires to him or her at three intervals following intake assessment. Consenting patients (N = 182) were mailed a battery of questionnaires at 3, 6, and 12 months following intake assessment and asked to complete and return these questionnaires in a postage-paid, self-addressed, stamped envelope. All follow-up data were collected between 2005 and 2008; 182 participants completed the study at intake assessment, 130 at 3-month follow-up, 103 at 6-month follow-up and 72 at 12-month follow-up. Participants were paid $20 for completing each of the three follow-up batteries of questionnaires (for a total of $60).

The Park Nicollet Melrose Center uses a multi-disciplinary treatment approach (e.g., medical, nutritional, psychological, psychiatric, and behavioral care) to treat eating disorders in a multi-modal setting that includes residential and inpatient beds and the capacity to provide individualized services across various levels of outpatient treatment (from full day treatment to intensive outpatient to standard outpatient programming). Treatment is personalized to the patient’s specific needs, but not tailored to the patient’s SOC; treatment may include family-based, dialectic-behavioral, cognitive-behavioral, medical, nutritional, spiritual, medication management, and other interventions. Patients often take part in selecting the type of programming they believe will benefit their recovery process (e.g., meeting with a Chaplin, attending occupational therapy groups, physical therapy groups and outings). Melrose Center is part of a larger not-for-profit health care organization that serves a diverse population of patients. The majority of the patients receiving treatment at Melrose Center utilize their insurance coverage for care received; however, a small percentage of patients pay out-of-pocket for services.

This study was reviewed and approved by the Park Nicollet Health Services Institutional Review Board; active consents and assents (for participants under 18 years of age) for all participants were used. It should be noted that during an audit of the study, the Park Nicollet Institutional Review Board identified that while all participants provided informed consent via signature and date on the consent form, the incorrect version of the consent form was used during the consent process. Specifically, participants were consented with an older version of the consent form and not the most current, IRB-approved version. This earlier version of the consent form included permission to extract two additional pieces of information that were determined at a later date not to be necessary, and thus language pertaining to extracting this additional information was excluded in the latter version of the consent form. In discussions about this discrepancy, the Institutional Review Board voted for the study to proceed and the study data to be used, as long as this error in the consent process was clearly disclosed in any dissemination of the study data.

Participants

Most participants were female (n = 175; 96.2 %) and Caucasian (n = 168; 92.3 %). The average age at intake assessment was 25.1 years (SD = 9.0; range 13–58 years) and age at ED onset was 16.5 years (SD = 5.2; range 5–46 years); average duration of ED was calculated to be 8.3 years (SD = 8.5; range 0–43.4 years). Body mass index (BMI) at intake assessment averaged 21.7 (SD = 5.9; range 10.6–58.1) as measured anthropometrically and calculated using the standard formula of weight in kilograms divided by the squared product of height in meters. Diagnoses were determined by psychiatric clinical interview following DSM-IV criteria; the distribution across diagnoses was 50.0 % eating disorder not otherwise specified (EDNOS; n = 91), 30.8 % bulimia nervosa (BN; n = 56) and 19.2 % anorexia nervosa—restricting subtype (AN-R; n = 26) and anorexia nervosa—binge-eating/purging subtype (AN-B/P; n = 5). An additional four participants were diagnosed with anorexia nervosa, but without specification of subtype. Of the participants diagnosed with EDNOS, 71.4 % reported fasting and/or restricting behavior, 47.3 % binge-eating, 29.7 % self-induced vomiting, 18.7 % laxative use and 7.7 % diuretic use at intake assessment.

Measures

The following information was collected at the patient’s intake assessment and extracted from the medical chart: gender, race, date of birth, date of intake assessment at the treatment facility, age at intake assessment, self-reported age at onset of ED symptoms, treatment milieu, previous eating disorder treatment, eating disorder diagnosis at intake assessment, and measured height and weight used to calculate body mass index (BMI). In addition, the following measures were collected as part of the mailed battery of questionnaires: Eating Disorders Stage of Change; Eating Disorder Examination–Questionnaire; Beck Depression Inventory; and State-Trait Anxiety Inventory.

The Eating Disorders Stage of Change measure (EDSOC, [18]) was adapted from a previously reported Stage of Change Scale [11] to assess three stages of change (Precontemplation, Contemplation, and Preparation/Action) across eight separate ED behaviors (binge-eating, fasting, restricting, self-induced vomiting, laxative use, diet pill use, diuretic use, metabolic enhancement use). It was designed to reduce limitations with other self-report instruments, notably non-specificity of ED behavior and length of questionnaire. Three questions are asked for each behavior, with the specific behavior noted in each individual question: “Do you intend to stop (behavior) in the next 6 months?” and “Do you intend to stop (behavior) in the next 30 days?” and “In the past year, before coming to this clinic, have you had definite plans to stop (behavior) (either on your own or with outside help) and have you actually attempted to carry out these plans?” Participants were asked to answer “yes” or “no” to each of the questions above. The participant could also check a box indicating that she or he did not engage in that specific behavior. A stage of change category (Precontemplation, Contemplation, or Preparation/Action) for each individual ED behavior is generated. Answering “no” to the first question yields a score of Precontemplation. Answering “yes” to the first question, but at least one “no” to the second or third questions yields a score of Contemplation. For a score of Preparation/Action, the participant must answer “yes” to all three questions. In addition to individual stages of change categories for each symptom, an overall stage of change score was identified based on the lowest stage of change across all symptoms reported by the participant. The addition of the overall stage of change score was made to allow for comparisons between this measure and the extant literature on SOC, which is predominantly non-symptom-specific. The psychometric properties of the EDSOC have been published elsewhere [18]: test–retest reliability over 1 week ranged from 0.57 (fasting) to 0.90 (purging) among children and from 0.35 (binge-eating) to 1.00 (diuretic use) among adults; symptom-specific convergent validity assessments ranged among children from 25.0 to 55.5 % and among adults from 36.3 to 56.3 % agreement with clinicians’ ratings of symptom-specific readiness to change; and divergent validity correlations ranged from 0.08 to 0.54 among children and from 0.01 to 0.42 among adults.

The Eating Disorder Examination-Questionnaire (EDE-Q, [19]) is a 41-item self-report questionnaire that asks specific questions pertaining to the presence and frequency of ED behaviors, thoughts, and feelings about one’s body over the past 28 days. The EDE-Q is derived from the interview version of the Eating Disorders Examination and has four subscales: Dietary Restraint; Eating Concern; Weight Concern; Shape Concern. Mean scale scores (and standard deviations) among young Australian adult women are: 1.30 (1.40) for Dietary Restraint; 0.76 (1.06) for Eating Concern; 1.79 (1.51) for Weight Concern; 2.23 (1.65) for Shape Concern [20]; mean scores (and standard deviations) among undergraduate women in the United States are similar: 1.29 (1.41) for Dietary Restraint; 0.87 (1.13) for Eating Concern; 1.89 (1.60) for Weight Concern; 2.29 (1.68) for Shape Concern [21]. Higher scores indicate greater pathology across the four subscales. The EDE-Q has been found to have good concurrent validity and criterion validity among community samples [22] and good internal consistency [22, 23]. Cronbach’s alpha reliability values for the current sample were 0.86 for Dietary Restraint, 0.85 for Eating Concerns, 0.94 for Shape Concerns, 0.89 for Weight Concerns and 0.97 for the global score.

The Beck Depression Inventory-2 (BDI-II, [24]) is a 21-item self-report instrument designed to measure the level of depression in adolescents and adults over the age of 13. Each of the 21 questions requires the participant to make a choice across a 4-point Likert scale. Total scores range from 0 to 63, and a higher score indicates a higher level of depression. Clinical severity categories are: 0–13 minimal; 14–19 mild; 20–28 moderate; 29–63 severe [24]. Psychometric properties that assess the reliability and validity of the BDI-II for measuring depression have been found to be strong among adolescent [25] and adult populations [26]. Cronbach’s alpha reliability for the current sample was 0.95.

The State-Trait Anxiety Inventory (STAI, [27]) is a 40-item self-report assessment that includes separate measures of state and trait anxiety. The total score ranges from 20 to 80 for each scale (trait or state anxiety), and higher scores indicate greater anxiety. Clinical severity categories, based on score, are: 20–30 low; 31–40 low average; 41–48 average; 49–60 high average; 61–70 high; 71–80 very high [27]. Psychometric properties of the STAI are sound, with test–retest among male and female high school and college students ranging from 0.65 to 0.86 for trait anxiety and 0.16 to 0.62 for state anxiety. This low level of stability for the state-anxiety scale is expected since responses to the items on this scale are thought to reflect the influence of transient situational factors that exist at the time of testing [27]. The validity correlations are 0.80 with the Taylor Manifest Anxiety Scale, 0.75 with the IPAT Anxiety Scale, and 0.52 with the Multiple Affect Adjective Check List [27]. Cronbach’s alpha reliability values for the current sample were 0.95 for State and 0.94 for Trait anxiety.

Data analysis

Descriptive summaries were generated for demographic and clinical variables overall and by overall SOC at intake assessment. Continuous variables were compared for differences between overall SOC groups using one-way ANOVA or Kruskal–Wallis tests based on normal or non-normal variable distributions, respectively. When differences were statistically significant, for normally distributed variables post hoc two-sample tests with Bonferroni correction were conducted and for non-normally distributed variables the Dwass–Steel–Critchlow–Fligner (DSCF) method for multiple comparisons was used [28]. Categorical variables were compared for differences between overall SOC groups using generalized χ 2 tests. Frequency and percent of participants reporting symptom-specific and overall SOC were calculated for each time point.

Treatment outcome was defined by frequency of specific ED symptoms (binge-eating, fasting, restricting, purging, laxative and diuretic use) as reported on the EDE-Q and by eating disorder and psychological status as measured by EDE-Q global and subscale scores, STAI subscales and BDI-II. To evaluate whether SOC predict treatment outcome, each treatment outcome was analyzed as a dependent variable using a repeated measures linear regression model with autoregressive correlation structure. Independent variables included in every model were age at intake assessment, eating disorder diagnosis, time point, SOC at intake assessment and the interaction of time point with SOC at intake assessment. Symptom-specific SOC at intake assessment was used as the independent variable in regression models with ED symptom frequency as the dependent variable; overall SOC at intake assessment was used as the independent variable in regression models with scores on ED or psychological measures as the dependent variable. For example, SOC for binge-eating at intake assessment was used as an independent variable in the model with binge-eating frequency as the dependent variable; whereas overall SOC at intake assessment was used as an independent variable in the model with depression as the dependent variable. Missing data within unique observations were excluded from analysis for each regression model; meaning, within each participant, time points with missing data were removed from each regression model. Of note, neither of the two participants in the purging Precontemplation group at intake assessment reported purging frequency, so the regression model with purging frequency as a dependent variable only uses purging Contemplation and Preparation/Action SOC groups at intake assessment as the dependent variable. Additionally, too few participants reported any laxative use (22 participants with 41 total observations) or diuretic use (6 participants with 11 total observations) to adequately power the regression model. Thus, regression models with laxative frequency and diuretic frequency as dependent variables were not generated.

All analyses were conducted in SAS version 9.3 (SAS Institute Inc, Cary, NC, USA) and a significance level of 0.05 was used throughout.

Results

Sample characteristics at intake assessment

Overall SOC was calculated for the 149 participants who had complete data on the EDSOC, and we investigated whether demographic or clinical factors (current age, age at onset of ED symptoms, duration of ED, BMI, gender, race, ED diagnosis and prior ED treatments) were associated with overall SOC at intake assessment (see Table 1). At intake assessment, BMI and eating disorder diagnosis were associated with overall SOC; specifically, those in Precontemplation had a higher BMI than those in Contemplation or Preparation/Action stages (average BMI 23.1 vs. 21.3, 21.2, respectively; Kruskal–Wallis test 6.2431, p = 0.0441); however, post hoc tests using the DSCF method indicated that this difference was not statistically significant. Results across eating disorder diagnosis were statistically significant across groups at intake assessment, revealing that those in Precontemplation were more likely to be diagnosed with EDNOS than with AN, and those in Preparation/Action were more likely to be diagnosed with AN than EDNOS (see Table 1).

Table 1 Mean scores (standard deviations) and percentages on sociodemographic and clinical characteristics across overall SOC at intake assessment
Full size table

Specific eating disorder symptom use over time and across stages of change

The percentage of participants using various ED symptoms at intake assessment ranged from 5.5 % for diuretics or metabolic enhancers to 70.3 % for restriction of dietary intake (see Table 2). The distribution of participant’s readiness to change individual ED behaviors varied considerably. At intake assessment, participants who reported engaging in binge-eating and/or vomiting also reported the highest SOC (Preparation/Action stage) for those behaviors (see Table 2). Specifically, 57.8 % of those who reported binge-eating and 56.5 % of those reporting self-induced vomiting scored in the Preparation/Action SOC for that symptom. In contrast, only 10.0 % of those who reported using metabolic enhancers scored in the Preparation/Action stage. Overall SOC, which was calculated by taking the lowest SOC for any symptom reported by the participant, was split uniformly across stages, with 30.2 % in Precontemplation, 39.6 % in Contemplation and 30.2 % in Preparation/Action stage at intake assessment (see Table 2).

Table 2 Stage of change (symptom-specific and overall) at each time point
Full size table

The percentage of individuals in Preparation/Action stage was generally stable across time points for the symptoms of binge-eating, fasting, restricting intake and self-induced vomiting (see Table 2). However, the percentage of participants in Preparation/Action stage who used laxatives, diet pills, diuretics or metabolic enhancers varied considerably across time points. For example, only 33.3 % of participants reporting laxative use were in the Preparation/Action stage at intake assessment, compared to 71.4 % at 3-months, but only 37.5 % at 6-months and 9.1 % at 12-months (see Table 2). Similar patterns were seen for participants reporting the use of diet pills, diuretics or metabolic enhancers as part of their eating disorder behaviors with highest motivation for change often occurring around 3- or 6-month follow-up.

Stage of change at intake assessment as a predictor of treatment outcome

Table 3 provides details on 28-day ED symptom/behavior use among the full sample (N = 182), as reported on the EDE-Q, across symptom-specific SOC at intake assessment and illustrates the variability in average ED behavior/symptom use by symptom-specific SOC across time points. Table 4 illustrates the average scores and standard deviations on eating disorder and psychological measures by overall SOC at intake assessment (N = 149) and shows general trends for scores on measures to be the highest at intake assessment versus other time points and also highest for those in Precontemplation compared to those in Contemplation or Preparation/Action.

Table 3 Mean (SD; n reporting that specific symptom/behavior at specific time interval) in 28-day symptom use from intake assessment to follow-up by symptom-specific SOC at intake assessment
Full size table
Table 4 Mean (SD; n at specific time interval) eating disorder and emotional state scores by overall SOC at intake assessment
Full size table

Regression models (see Table 5) were conducted to evaluate if symptom-specific SOC at intake assessment predicted change in eating disorder behavior over time. As shown in Table 5, symptom-specific SOC at intake assessment did not predict changes in ED behavior frequency over time. Frequency of fasting (p = 0.005) and restricting behaviors (p < 0.001) improved over time regardless of symptom-specific SOC at intake assessment. Furthermore, findings did indicate that the Precontemplation group had a statistically higher frequency of restricting than the Preparation/Action group overall (95 % CI for the difference −1.417, −0.005).

Table 5 Regression Results shown as (Type III Test of Fixed Effects p-value) Regression Coefficient Estimate
Full size table

Regression models were also conducted to evaluate if overall SOC at intake assessment predict changes in eating disorder and psychological measures over time, and indicated that overall SOC did not predict BDI-II, any of the EDE-Q subscales or trait anxiety as measured by the STAI (see Table 5). Interestingly, EDE-Q subscale scores, STAI trait anxiety scores and BDI-II scores showed improvement across follow-up intervals regardless of SOC at intake assessment (p < 0.001). Furthermore, those in Precontemplation at intake assessment had higher EDE-Q Weight Concern scores than those in the Contemplation group (95 % CI for the difference −1.236, −0.109). Finally, overall SOC at intake assessment significantly predicted state anxiety across follow-up intervals; the higher the stage of change at intake assessment (i.e., Preparation/Action vs. Precontemplation, Contemplation vs. Precontemplation), the lower the state anxiety over time (p = 0.0135; see Fig. 1).

Fig. 1
figure 1

State Anxiety Score over time by stage of change at intake assessment

Full size image

Discussion

The current study utilized a psychometrically sound self-report measure of stages of change to determine the prevalence of Precontemplation, Contemplation, and Preparation/Action stages across individual eating disorder symptoms and across a calculated overall stage of change score, and to evaluate whether symptom-specific and overall stages of change predict treatment outcome in a treatment-as-usual clinical setting. Findings indicate that patients seeking ED treatment present with significant variability in their readiness to change specific ED behaviors. Binge-eating and vomiting were found to be the symptoms for which the greatest percentage of participants were motivated to change. Findings from the current study indicate that SOC remained generally stable over follow-up intervals for most symptoms, but fluctuated considerably over time for laxative, diuretic, diet pill and metabolic enhancement users. As an example, the percentage of individuals found to be in Preparation/Action for laxative use increased from intake assessment to 3 months, then declined from 3 to 6 months, and declined further from 6 to 12 months. In fact, the percent of participants at 12-month follow-up in the Preparation/Action stage for laxative use was lower than that at intake assessment. Because this study included patients receiving treatment-as-usual (TAU) care, we cannot comment as to whether patients remained engaged in treatment at the time of the follow-up. Findings may reflect the difficulty that patients have in maintaining motivation to reduce symptom use in the face of multiple factors, including changes in weight and level of body dissatisfaction, fluctuations in the frequency of the use of other ED symptoms, engagement in treatment and type of treatment received.

Generally, our hypothesis that a self-report measure of symptom-specific SOC at intake assessment would predict symptom-specific treatment outcome was not confirmed. Our null findings are consistent with Allen and colleagues [14] who found that readiness to change was not associated with the single symptom of binge-eating that they evaluated, but more recent reviews have held more promise on the predictive utility of the transtheoretical model in eating disorders [29]. We assert that future research is needed in this area to help the many individuals who struggle with eating disorders improve their outcome and to better identify those therapeutic interventions that promote greater change.

Our hypothesis that overall SOC at intake assessment would predict treatment outcome was confirmed only for state anxiety. Study findings indicate that overall SOC at intake assessment did not predict changes in dietary restraint, ED-related concerns (eating, weight or shape concerns), depression or trait anxiety over follow-up intervals; patients improved in these areas regardless of SOC at intake assessment. However, analyses reached statistical significance for the association between SOC at intake assessment and state anxiety over time intervals. Given that we did not find SOC to be associated with changes in ED symptoms, but did find significant and positive results for state anxiety, we wonder if readiness to change reflects a more global motivation than just for ED behaviors, and in light of current findings with state anxiety, may reflect level of anxiety at the time of assessment and the distress associated with being evaluated for ED treatment. As such, we advocate for the continued inclusion of readiness to change measures in treatment-as-usual settings so as to be able to evaluate their utility in identifying changes in ED symptoms, broader psychopathology, and quality of life over time. We believe that additional research is needed to explore the associations between stages of change and treatment outcome across multiple factors relevant to ED treatment (i.e., anxiety, body dissatisfaction and distortion of perception, depression, quality of life, self-efficacy) and across long time intervals.

Given that half of the participants in the current study were diagnosed with EDNOS at intake assessment, we conducted post hoc tests to describe this diagnostic subsample and the changes in SOC over time. Only a small proportion (18.1 %) were in Preparation/Action stage at intake assessment, and this may be related to symptom infrequency compared to their peers meeting full diagnostic criteria for AN or BN. Specifically, there was a significant difference for binge-eating frequency at intake assessment across ED diagnosis (median binge-frequency was: AN = 6, BN = 11, EDNOS = 5; Kruskal–Wallis p = 0.0187; post hoc tests show that the frequency of binge-eating for participants with bulimia nervosa was significantly greater than that for those with EDNOS) but not for other eating disorder symptoms. Surprisingly, we found no differences across diagnostic groups by duration of eating disorder. However, it should be noted that overall SOC among those diagnosed with EDNOS remained generally stable across follow-up intervals, with 32.3, 33.3 and 30.0 % in Preparation/Action at 3, 6 and 12-month follow-up periods, respectively. Motivational enhancement therapy interventions may be particularly important for those individuals diagnosed with EDNOS who present for specialized treatment.

We believe it is important to emphasize that patients in the current study showed improvement over the course of the study in this TAU program, regardless of SOC at intake assessment. It is possible that a self-report measure of SOC is not specific enough to predict clinical outcome for individual ED symptoms compared to interview methods. Similarly, one other study using a mixed-diagnostic sample found that overall readiness to change was associated with overall EDE-Q score, but did not predict treatment completion nor cessation of binge-eating [14]. However, it is also possible that differences between the current study of treatment outcomes and those of others are related to the shorter duration of study used in other studies (e.g., outcome assessed at 2 and 6 weeks [7] or 6 months [9]) compared to 12 months in the current study. Future research should continue to evaluate the value of overall and symptom-specific readiness to change as a predictor of treatment outcome over long follow-up periods.

Nevertheless, scores on the EDE-Q subscales started at clinical levels and approached published norms (but were still slightly elevated) across the follow-up periods. Furthermore, inspection of the sample scores compared to published clinical cutoffs indicates that our study sample scored in the moderate to severe depression category at intake assessment and improved to the minimal to mild depression category by the 12-month follow-up. A similar finding was found for anxiety, with most scores in the high average or higher range for state anxiety at intake assessment, then reducing to the average range for state anxiety by the 12-month follow-up. For trait anxiety, scores for the sample were in the high average or higher range at intake assessment, and subsequently lowered to the average to high average range by the 12-month follow-up. We believe that the global patient improvement in this TAU setting which did not specifically focus on enhancing motivation is important to emphasize because efforts to design motivation-enhancing treatment programs and evaluate their effectiveness have found mixed results in ED treatment outcome across treatment conditions [14, 15, 29]; the value of motivational enhancement therapy for eating disorders warrants further exploration [29].

Stages of change did not improve substantially over the course of the study; the inclusion of motivational enhancement therapy may help increase readiness for change and improve already positive outcomes. Several clinical settings have developed strategies to enhance motivation to change under the belief that improving overall motivation may contribute significantly to treatment outcome and relapse prevention. One of the most impressive examples comes from Australia, where Touyz and colleagues [30] reported on their 5-day hospital-based treatment program for patients with anorexia nervosa who were in the Contemplation stage versus their 3-day program for patients with AN who were in the late Contemplative or Action stages. Patients completing their programs showed significant improvement from admission to discharge on many domains, including but not limited to eating disorder cognitions, core beliefs, readiness to change, quality of life and many other perpetuating factors (self-esteem, depression, anxiety, stress) as well as increases in BMI for both underweight and normal weight patients and decreases in the frequency of binge-eating, self-induced vomiting and exercise; additional improvements from discharge to follow-up were seen quality of life, depression, anxiety and stress [31]. However, others have found only modest results [14, 30, 32]. For example, Allen and colleagues compared enhanced cognitive-behavioral therapy (CBT-E) in its usual form to CBT-E preceded by four sessions of motivation-focused therapy (MFT + CBT-E) in a sample of patients with AN and found that while MFT + CBT-E was associated with increases in readiness to change (readiness to change both “overall” eating disorder symptoms as well as dietary restraint), there were no significant differences between groups on treatment completion or eating disorder symptoms over time [14]. Another study of patients with diverse eating disorder diagnostic presentations compared a five session readiness and motivation therapy (RMT) to a wait-list control group [15]. Due to the nature of the clinical setting, patients were permitted to access other psychological or medical treatments during the course of the study; the authors found equivalent improvements between treatment conditions and thus no significant differences in depression, drive for thinness and bulimia symptoms across 6-week and 3-month follow-up [15]. Thus, it is not clear that tailoring treatment to SOC is significantly helpful for symptom reduction, particularly among clinical settings with diverse diagnostic presentations or with patients whose symptom constellation changes over time. Future research should continue to evaluate whether motivation-enhancing treatments are effective among patients with any eating disorder diagnostic presentation and serve to enhance treatment outcome during the course of treatment and across long-term follow-up.

Strengths and limitations

Noteworthy strengths of the current study include the use of data from a robust sample of patients with diverse ED presentations by individual symptoms and diagnosis. These patients were treated in a clinical setting, not as part of a controlled treatment trial, thus our final sample includes those who may or may not have completed all treatment recommendations, which more accurately reflects the course of treatment and variability in motivation for many people with EDs. In addition, the measures used in the current study demonstrate sound psychometric properties, allowing for good generalizability to other similar samples. The evaluation of symptom-specific SOC allows for greater understanding of the variability in motivation for certain behaviors versus others. Finally, the long period of follow-up allows us to identify changes in motivation, symptom use and broader psychological functioning over a longer period than many studies in the area of readiness.

However, there are several limitations to this study that are worth consideration. First, there is a high probability of a response bias in this sample; those who chose to participate in the study could be more compliant or healthier than nonresponders. We attempted to navigate this problem by offering a financial incentive for study participation; however, future research may utilize other recruitment strategies to increase participation. Second, some cell sizes of interest, such as symptom-specific SOC as predictors of diet pill or metabolic enhancer use at follow-up, were too small for robust analyses; larger sample sizes may help to increase power and subsequently increase detection of meaningful differences, particularly over the later follow-up periods where we saw more attrition. Future research with larger samples would also allow for confirmation of a quadratic time trend that was found in post hoc analyses to be significant, and of SOC by ED diagnosis among mixed-diagnostic groups, such as a separate investigation of SOC for BN patients across treatment from those diagnosed with AN or EDNOS. Third, we relied on self-report instruments for assessment of stages of change and our outcome measures to ease patient burden of time; however, we recognize that interview methods are preferred. Finally, our sample was predominantly of Caucasian females; future research should seek to have broader sociodemographic diversity within the study sample.

Conclusions

Stages of change for symptom reduction vary considerably among patients seeking eating disorder treatment in a treatment-as-usual setting. Symptom-specific SOC did not predict change in symptom frequency over time. Overall SOC was a significant predictor for only one treatment outcome, state anxiety. Although stages of change do not appear to be strong predictors of treatment outcome, findings from the current study demonstrate that most patients improve over time as evident by the reduction of eating disorder behaviors and amelioration of mental health concerns within a treatment-as-usual care setting.