Diagnostically, 123 patients met criteria for anorexia nervosa (AN) and 1 met criteria for anorexia nervosa like eating disorder not otherwise specified (EDNOS). Among those with AN, 56 (45.5%) met criteria for the restricting subtype of the illness (AN-R) and 67 (54.5%) met criteria for the binge purge subtype of the illness (AN-BP).

Patients ranged in age from age 18 to 63 years with an average age of 30.81 (SD = 11.43) years. One hundred and nineteen (96%) identified as female, three (2.4%) identified as male, one (0.8%) identified as transgender and one (0.8%) preferred not to disclose their gender. Ninety-eight (79.0%) identified as Caucasian or with a specific European nationality (i.e. Dutch), 1 (0.8%) identified as Black, eight (6.5%) identified as Asian, 2 (1.6%) identified as Hispanic, 3 (2.4%) identified as Indian, 2 (1.6%) identified as Middle Eastern, 7 (5.6%) identified in non-specific ways (i.e. Canadian) and 8 (6.5%) patients did not report their ethnicity, At the time of admission, patients’ average Body Mass Index (BMI) was 15.21 (SD = 1.71) and they had been unwell for an average of 10.53 (SD = 9.59) years. Preceding the study period, 27 (21.8%) individuals had had prior inpatient admissions to our inpatient program.

The average length of stay for all first admissions in the study period was 13.88 (SD = 7.30) weeks. Thirteen patients left treatment in the first four weeks of treatment (10.5%), thirty-one left after four weeks but prior to weight restoration (25.0%) and eighty (64.5%) completed treatment.

Patients in the early treatment termination group had an average inpatient length of stay of 2.26 (SD = 0.88) weeks; patients in the later treatment termination group had an average length of stay of 10.28 (SD = 6.19) weeks and patients in the treatment completion group had an average length of treatment of 17.17 (SD = 5.48) weeks.

Comparison of timing of treatment termination groups showed no statistical differences in baseline demographic variables including age at admission, duration of illness, BMI at admission or average frequency of excessive exercise prior to admission. Sixty-five (54.6%) of patients reported engaging in excessive exercise in the three months prior to admission. Significant differences in of binging, purging, diuretic or laxative use in the three months prior to admission were also not found for those with the binge purge subtype of anorexia (n = 67) (Table 1).

There was a significant relationship between eating disorder subtype and timing of treatment termination (x2(4) = 14.43, p = .006) with patients leaving treatment early having the highest incidence of AN-BP diagnoses (92%) (Fig. 1).

Timing of treatment termination and eating disorder diagnosis

There was no statistical difference in the frequency of identified genders, ethnicity or prior inpatient eating disorder admissions between the three treatment outcome groups.

Comparison of treatment outcome groups on our primary baseline symptom domains found significant differences. Timing of treatment termination was associated with severity of depressive symptoms at admission as measured by the BDI (F(2,120) = 5.83, p = .004) and severity of eating disorder psychopathology at admission as measured by the EDEQ total score (F(2,111) = 4.56, p = .013).

Difficulties with emotional regulation at admission as measured by the DERS total score did not achieve statistical significance (Table 2).

Post hoc comparisons showed that patients who left treatment early had higher BDI scores (M = 47.61, SD = 9.75) that those who completed treatment (M = 34.44, SD = 12.32, p = .003). The BDI scores of those who left treatment earlier and later (M = 37.00, SD = 12.99, p = .038) did not differ statistically. The BDI scores of patients who left treatment later also did not differ significantly from those who completed treatment. BDI scores were available for all patients save one in the treatment completion group.

Patients who left treatment early had significantly higher EDEQ total scores (M = 5.26, SD = 0.60) than those completed their admissions (M = 4.09, SD = 1.42, p = .009) but not those who left later in treatment (M = 4.38, SD = 1.16). The EDEQ scores of patients who left treatment later also did not differ from those who completed treatment. EDEQ total scores were available for all 13 patients who terminated treatment early, 29 who terminated later but before completion and 72 who completed treatment.

Analyses of the overall DERS scores showed potential differences between patients who left treatment early (M = 143.83, SD = 25.15), those who left late (M = 117.233, SD = 35.14, p = .021) and those who completed treatment (M = 119.12, SD = 26.35, p = .019) that did not achieve statistical significance. Patients who left later did not differ from those who completed treatment. DERS total scores were available for all 13 patients who terminated treatment early, 30 who terminated later but before completion and 73 who completed treatment.

EDEQ subscale scores by timing of treatment termination are presented in Fig. 2. Results show that there was a statistically significant relationship between timing of termination of treatment and only one subscale: eating concerns. Patients who left treatment early had significantly higher scores on the eating concerns subscale (M = 4.86, SD = 0.92) than those who completed treatment M = 3.56, SD = 1.52, p = .014). Additional subscale comparisons are available in Additional file 1: Table S1.

EDEQ subscale scores by timing of treatment termination

DERS subscale scores are presented in Fig. 3. Patients who left treatment early had significantly higher scores on the goals subscale (M = 22.08, SD = 3.48) than those who left treatment later (M = 17.84, SD = 5.33, p = .014). Additional subscale comparisons were not statistically significant and are available in Additional file 1: Table S1.

DERS Subscale scores by timing of treatment completion