Introduction

Treatment dropout rates among adults with anorexia nervosa (AN) are significant [1]. For adolescents, rates of dropout in outpatient randomized controlled trials are approximately 10–15% [2,3,4]. Identifying factors associated with treatment dropout across treatment settings is an important therapeutic goal, as modifiable factors may be found that could improve treatment compliance and outcome. Recent studies assessing factors associated with treatment dropout in outpatient settings for adults have found that dropouts have more physical and mental health problems, are more likely to live alone, are less likely to be employed [5], and have higher scores on novelty seeking [6, 7]. In addition, treatment dropouts have been found to be younger, have a higher body mass index (BMI), and in randomized controlled trials, be randomized to a nonpreferred treatment [7]. Little is known, however, about rates of dropout in other treatment settings, such as inpatient programs or day treatment programs/partial hospitalization programs, or the factors associated with treatment dropout in these settings, particularly for young patients.

A study of adults receiving either outpatient treatment, partial hospitalization, or full hospitalization found that, across diagnoses, treatment completers attended more treatment sessions, and treatment dropouts had a higher BMI at assessment. Compared to patients with AN, 23.1% of whom dropped out of treatment, 35.2% of patients with a diagnosis of bulimia nervosa (BN) dropped out of treatment, and 50.5% of patients with a diagnosis of eating disorder not otherwise specified (EDNOS) dropped out [8].

In a study of adolescents with AN in an inpatient setting, factors associated with treatment dropout included being part of a single-parent family, more severe dietary restriction, and a low score on the Eating Disorders Examination-Questionnaire restraint subscale, possibly indicating reluctance to acknowledge restricted food intake [9]. A study of adolescent and young adult inpatients with AN found that dropout was associated with older age, longer duration of amenorrhea, being part of a single-parent family, previous hospital admissions, a lower BMI at admission, and previous suicide attempts [10].

Rates of treatment dropout for adolescents in a day hospital program range from 7 to 42% [11,12,13,14,15]. To our knowledge, only one study has reported on factors associated with treatment dropout for adolescents in a day hospital program. In a study of 65 adolescents, Grewal et al. [13] found that, when compared to treatment dropouts, treatment completers were less likely to have a history of purging and were more likely to have been on a selective serotonin reuptake inhibitor when entering the program.

In outpatient treatment for adolescents with anorexia, parental expressed emotion has been found to be associated with treatment dropout [16, 17]. Expressed emotion (EE) is a measure of a relative’s attitudes and behaviors toward an ill family member [18] and consists of five components: critical comments, emotional overinvolvement, hostility, warmth, and positive remarks. Family members considered to be high on EE score above certain cutoffs on critical comments, emotional overinvolvement, or hostility. It is these families who tend to drop out of treatment prematurely [16, 17].

The purpose of the current study is to add to the limited literature on factors associated with treatment dropout for adolescents and young adults with AN in a family-based partial hospitalization program, and to assess whether parental EE plays a role in treatment dropout in this setting.

Method

Participants and procedure

Participants were 55 adolescents (mean age 14.1, SD 1.7, range 10–17) and 32 young adults (mean age 19.6, SD 1.6, range 18–24) meeting DSM-IV criteria for AN, receiving treatment between 2012 and 2015. Patients and parents participated in a partial hospitalization program (PHP) based on family-based treatment (FBT) principles [19] and completed interview-based and self-report measures at baseline and at end-of-treatment in the PHP (EOT PHP). Parents and patients over the age of 18 signed informed consent, and patients under the age of 18 signed assent to participate in the study. The study was approved by the Institutional Review Board of the University of Michigan.

Treatment setting

Patients in the PHP participated in programming 5 days per week for 6 h a day, during which they participated in group therapy, two meals, and a snack. The PHP was designed based on FBT principles [19]. In the first of three phases of FBT for AN, parents are put in charge of the weight restoration process by making all food-related decisions for their child and monitoring all meals and snacks. This is done because the ego-syntonic nature of the illness makes it difficult for the patient to willingly eat the healthy portions needed for weight gain and physical recovery. Thus, parental involvement in treatment is crucial and the PHP was designed with this FBT tenet in mind. Families were involved in the PHP in several ways: prior to joining the program, all families participated in the first two sessions of manualized FBT, where attendance by all family members was strongly encouraged. In addition, parents made all food-related decisions for their children while in the program, in keeping with manualized FBT principles, and daily parental participation at meals was required. A “debriefing” session after breakfast was intended to help parents problem-solve around strategies to help their child achieve recovery. It was emphasized to parents that their participation in the program was crucial to their child’s success and they were considered the main agents of change in the recovery process. Parents were also required to attend a weekly skills group designed to help them be effective in the weight restoration process. Please see Hoste [20] for a more thorough description of the treatment program. Patients were assigned to one of two tracks depending on their age: patients under the age of 18 participated in Track 1, and patients 18 and over participated in Track 2. Tracks were similar in terms of programming and treatment philosophy, but with some differences due to the different ages. For example, participants in Track 1 participated in school for part of the day, whereas participants in Track 2 had an extra group during that time.

Measures

Demographics

As part of the intake assessment, information was gathered on age, family status, duration of illness, and previous hospitalizations. It was also noted during treatment whether the patient was hospitalized during their stay in the PHP.

Mini International Neuropsychiatric Interview (MINI) [21] or Mini International Neuropsychiatric Interview for Children and Adolescents (MINI-Kid) [22]

The MINI and MINI-Kid are structured diagnostic interviews used to assess a range of Axis I disorders, and were conducted at baseline to assess comorbidity. The MINI was administered to patients ages 18 and above, and the MINI-Kid was administered to patients 17 and under. The interviews have been shown to have good reliability and validity [21,22,23].

Eating Disorder Examination-Questionnaire (EDE-Q) [24]

The EDE-Q is a 36-item self-report measure completed at baseline and EOT PHP. The EDE-Q measures the behavioral and cognitive aspects of eating disorder psychopathology, including frequency of behaviors such as binge eating and purging.

Family Questionnaire (FQ) [25]

The FQ is a 20-item self-report measure assessing parental expressed emotion. It is composed of a total score and two subscales: critical comments and emotional overinvolvement.

Percent of expected body weight (%EBW) and Body Mass Index (BMI)

%EBW was calculated for patients under the age of 18 based on height, age, and gender, using CDC growth charts (%EBW = BMI/50th percentile BMI for age and height × 100) [26]. BMI was calculated for patients in Track 2. Weight and height were taken by trained assessors. Weight was taken in light indoor clothing, without shoes.

Treatment dropout

Treatment dropout was defined as leaving treatment prior to the recommendation of the treatment team.

Data analysis

Chi-square tests were used to assess differences in categorical variables, and multivariate ANOVAs were used to assess differences in continuous variables.

Results

Demographics and baseline symptomatology

Participants included 79 females (90.8%) and 8 males (9.2%). The majority of participants were white (n = 79; 90.8%); two (2.3%) were Hispanic, two were Asian, two identified as biracial, one (1.1%) identified as “other” and one as unknown. Seventy patients (80.5%) came from intact families, 15 (17.2%) did not (of these, 7 were divorced, 7 had remarried, and 1 was widowed), and two did not have this information available. The length of time in the PHP ranged from 1 to 74 days, with an average of 25.1 days (SD 12.9). For treatment completers, the average length of stay was 25.7 days (SD 11.9) and for treatment dropouts, the average length of stay was 18.1 days (SD 11.1) (p = 0.006). Forty patients (46%) had been hospitalized previously for an eating disorder, and 26 (29.9%) came directly from an inpatient unit to admission to the PHP. Thirty-nine (44.8%) had received previous outpatient treatment for an eating disorder. Forty-two (48.3%) had a comorbid Axis I diagnosis. The average duration of illness was 18.7 months (SD 23.3). For Track 1, the mean %EBW at intake was 84.96 (SD 7.7). For Track 2, the mean BMI at intake was 17.54 (SD 1.7).

The total number of objective binge episodes over the previous 3 months averaged 4.4 (SD 23.5), the total number of vomiting episodes over the previous 3 months was 6.0 (SD 19.4), and the total number of laxative misuse episodes over the previous 3 months was 0.3 (SD 1.6).

Completers versus dropouts

Sixty-eight patients (78.2%) were identified as treatment completers, and 19 (21.8%) were identified as treatment dropouts. Few significant differences were found between groups (see Table 1). For patients in Track 1, %EBW was lower at EOT PHP for dropouts than for treatment completers. Across tracks, there was a trend toward patients who had purged in the last month to be less likely to drop out of treatment, and a trend toward patients who dropped out of treatment to have more critical fathers.

Table 1 Treatment completers in comparison to treatment dropouts
Full size table

Discussion

The current study sought to assess factors associated with treatment dropout from a family-based PHP for adolescents and young adults. Overall, few differences were found between treatment completers and treatment dropouts. Treatment dropouts had lower weights at EOT PHP than treatment completers, which would be expected given that they also had shorter lengths of stay in the PHP. In contrast to Grewal et al. [13], who found that treatment completers were less likely to have a history of purging, treatment completers in the current study were more likely to have purged in the past month than treatment dropouts. However, Grewal et al. included a range of eating disorder diagnoses, possibly accounting for this difference between studies. It is possible that patients with AN who are purging are considered to have more severe symptoms than those who are not purging, perhaps making it less likely for parents to discontinue treatment prematurely. Further research is needed to better determine the role of purging in treatment dropout.

In the current study, treatment dropouts tended to have more critical fathers than treatment completers. Parental EE has not been examined in other studies of day programs/PHPs, but it has been found to play a role in treatment dropout in outpatient samples. Rienecke et al. [16] found that families with critical mothers were more likely to drop out of treatment prematurely, and Szmukler et al. [17] found that compared to treatment completers, both mothers and fathers of treatment dropouts had higher levels of criticism, and fathers were higher on emotional overinvolvement. Parental EE is a modifiable factor. In the schizophrenia literature, familial interventions designed to reduce high EE have been developed, are associated with improved patient outcome, and are widely used [27, 28]. Effective psychosocial interventions often include psychoeducation about the disorder, which can lead caregivers to change their attributions about illness-related behavior. Caregivers who are high on EE tend to attribute unwanted behavior to the person rather than to the illness [29]. Changing this perception can lead to a lowering of the emotional climate in the home, which may be less stressful for the patient and improve relapse rates [28]. Other interventions for families of patients with schizophrenia include improving relatives’ problem-solving skills, having realistic expectations of the patient, and encouraging appropriate limit-setting on the part of caregivers [28].

Efforts to reduce parental EE in the eating disorders field, however, have seen mixed results, with some studies finding no changes in EE over time [30, 31], and two studies finding improvements in emotional overinvolvement but not critical comments [32, 33]. Brown et al. have concluded that critical comments are the most important index of EE [34]; thus there is a need in the eating disorders field for interventions that can effectively reduce parental criticism.

Further, there is need for a better understanding of the mechanisms by which EE impacts treatment outcome [35]. Parental EE has been shown to be associated with treatment outcome, treatment dropout, and relapse across a number of different physical and mental health conditions [36], but insight into how EE influences these outcomes has been lacking. In the schizophrenia literature, it has been suggested that living with a relative high on EE is a stressor for the patient and if this stress reaches a high enough level, relapse would occur [34]. The ways in which EE impacts the course and outcome of other disorders, including eating disorders, is not as well understood.

Given previous findings [9, 10], it was somewhat surprising that being part of an intact family was not associated with treatment completion, although with a larger sample size it may have been. Family status seems a particularly relevant issue in a family-based treatment program. Daily parental participation was required in this program, and it was not unusual for one parent to take time off work to attend the program while the other parent continued to work and spent less time involved in treatment. For a single-parent family, daily attendance may prove especially challenging. Treatment programs involving a family component may need to work closely with single-parent families to ensure that they have adequate support and can take the necessary time to be involved in the program while minimizing other stressors such as missing work.

Strengths of the study include the use of well-validated measures, a contribution to the limited literature on factors associated with treatment dropout in these programs, and a somewhat larger sample size than has been included in previous studies of adolescent day treatment programs. However, with a larger sample size some trends may have become significant. Limitations include the lack of information on characteristics such as socioeconomic status, being on medications at baseline, family history of an eating disorder, and other family characteristics that may have been informative in this regard. In addition, the current study assesses patients in a family-based partial hospitalization program; findings from this study may not be generalizable to patients in other partial programs or in outpatient treatment, although the finding that critical parents are associated with treatment dropout has also been found in studies of outpatient treatment [16, 17].

Identifying risk factors for treatment dropout is an important endeavor. Although some, such as previous hospitalizations [10], are not modifiable factors once a patient arrives for treatment, others, such as parental EE, may be. Further research with larger sample sizes is needed to identify factors associated with treatment dropout, so that treatment providers can intervene and possibly reduce the chances of patients prematurely discontinuing treatment.