Study design

The FABIANA study is divided into the three consecutive substudies I–III corresponding to the outlined study aims. An overview of the FABIANA substudies including the planned recruitment phases is provided in Fig. 1. All substudies use an observational, cross-sectional, retrospective, multicentre design. The qualitative substudy I and the quantitative substudy III are multi-informant studies and aim to include triads of patients with anorexia nervosa, carers and physicians. A detailed description of the substudies follows below.

DUI, duration of untreated illness.

Fig. 1 Overview of the FABIANA substudies.

Study centres

The main study centre of the FABIANA study is located at the Institute for Psychosomatic Medicine and Psychotherapy of the University Medical Center Hamburg-Eppendorf, Germany. Additionally, 20 cooperating partners are involved in the recruitment of eligible patients. The exact locations of the cooperating study centres are shown in Fig. 2. The cooperating study centres consist of 12 in-patient clinics, 2 medical care centres, 2 psychotherapy training centres, several psychotherapists providing out-patient care, and 1 counselling centre. Three of the in-patient clinics, one of the medical care centres and all remaining out-patient care providers are located in Hamburg. The remaining cooperating centres are located throughout Germany. During the course of the study additional out-patient care providers might be recruited. As specified in Fig. 2 only selected study centres will participate in substudy I, whereas all study centres will be involved in substudy II and substudy III.

Fig. 2 Location of FABIANA study centres across Germany.

All cooperating study centres provide and/or refer into specialised eating disorder treatment in the context of routine care provision in Germany. This means for all settings that psychotherapy is the integral part of the treatment. All in-patient and day-patient care providers offer multimodal treatments consisting of individual psychotherapy, group psychotherapy, nutritional therapy including weight contracts, psychoeducation and different non-verbal therapies. Outpatient treatment is either cognitive–behavioural or psychodynamic individual psychotherapy. For adolescents, family therapeutic sessions are usually added in both settings. The intensity of out-patient treatment usually ranges from 24 to 100 sessions. All cooperating study centres have experience of several years in the treatment of patients with eating disorders. The costs of out-patient and in-patient psychotherapeutic treatments are fully covered by all statutory health insurances in Germany. Individually specific reimbursement regulations apply for individuals who have private health insurance (approximately 9% of all insured people).^{Reference Priebe and Wright23}

Eligibility and recruitment

The inclusion and exclusion criteria, that apply to all substudies equally, are presented in Table 1. The rationale for the criteria in detail is as follows: the minimum age limit of 14 years was chosen in light of the sufficiently consolidated development of higher-order cognitive functions (such as abstract, logical thinking and reasoning, the ability to reflect on oneself or complex problem-solving) at the end of early adolescence.^{Reference Eccles24} This criterion seems especially important for the participation in the qualitative interviews in substudy I. Additionally, the chosen age limit allows for greater consistency within the applied instruments (in particular Structured Clinical Interview for DSM-IV Axis I/II diagnoses (SCID) assessments^{Reference Salbach-Andrae, Burger, Klinkowski, Lenz, Pfeiffer and Fydrich25,Reference Wittchen, Zaudig and Fydrich26} ).

Table 1 Inclusion and exclusion criteria

In addition, we specified criteria regarding the timing of the first psychotherapeutic treatment. Accordingly, the treatment commencement should not be longer than 3 months previously, in order to minimise the risk of recall biases and response shifts.^{Reference Kjellsson, Clarke and Gerdtham27} However, patients should also not be included in the study at the very beginning of their treatment (i.e. before the completion of 7 days as an in-patient or five sessions of out-patient treatment) in order to exclude patients on a waiting list or individuals with early treatment drop-out. In line with this reasoning, earlier psychotherapeutic treatments only constitute as such, if they lasted for or exceeded this duration (7 days inpatient treatment or five sessions outpatient psychotherapy).

Eligible study participants need to fulfil all inclusion criteria and no exclusion criteria. In the first step, the cooperating partners will screen for eligible patients in the context of their regular admission interviews or initial consultations. They will inform eligible patients about the study, invite study participation and hand out study materials (i.e. written informed consent, questionnaires). Afterwards, the study team examines the final eligibility of each patient in the context of an interview for example validation of anorexia nervosa diagnosis using SCID-I, fulfilment of the specified criteria regarding first psychotherapeutic treatment).

The recruitment of carers and involved physicians is based on the included patients. Therefore, study patients will be asked to name one carer, who supported or followed their path from illness onset to treatment initiation, and one physician, who was involved in the process of treatment initiation (for example diagnosis, referral). After obtaining the patient's written consent to contact the carer or the release of patient–doctor confidentiality the study team will invite the carer or physician to participate in the study.

Assessment of DUI

DUI will be assessed in all participating patients throughout all substudies and will be specified in months. The DUI is operationalised as the time interval between the anorexia nervosa illness onset and the beginning of a psychotherapeutic anorexia nervosa treatment.

The retrospective assessment of the illness onset is based on a combination of questionnaire data and SCID-I lifetime interview.^{Reference Wittchen, Zaudig and Fydrich26} The two-staged procedure follows existing recommendations to minimise recall biases^{Reference Anderluh, Tchanturia, Rabe-Hesketh, Collier and Treasure28} and was tested within a small pilot study (n = 14: n = 3 adolescent patients, n = 11 adult patients). First, the patient answers questionnaire items on the lowest and highest weight, episodes of rapid weight loss and the lifetime occurrence of different anorexia nervosa symptoms (for example amenorrhea, binge eating). Afterwards, the patient is asked to map each item on a life chart. The life chart is constructed to shortly describe important life events or stages of life as anchor points in the free space below the timeline and the first occurrence of single anorexia nervosa symptoms in the space above the timeline. This questionnaire data is used as preparation for the SCID-I anorexia nervosa lifetime interview, which is used to finally determine the month of illness onset. As recommended, the assessment will focus especially on the onset of somatic and behavioural symptoms (such as first occurrence of underweight, amenorrhea, beginning of purging behaviours) rather than on psychological symptoms (such as body image disturbance, fear of weight gain).^{Reference Anderluh, Tchanturia, Rabe-Hesketh, Collier and Treasure28}

The beginning of the first psychotherapeutic anorexia nervosa treatment is either the date of hospital admission or the date of the first out-patient psychotherapy session. It will be assessed using a short questionnaire, which is filled out by the treating psychotherapist. In case of a missing therapist report, the patient-reported date will be used as a surrogate.

Substudy I

The main aim of the qualitative substudy I is to identify facilitators and barriers in anorexia nervosa treatment initiation from the perspectives of patients, carers and involved physicians using personal interviews. For sample description purposes, various other variables will be assessed. An overview of the complete data collection of substudy I is provided in the Appendix. Based on the therapist-reported weight and height, we will calculate the body mass index (BMI) as kg/m². In adolescent patients we will additionally calculate the age- and gender-adjusted BMI percentiles and the percentage of expected body weight (observed BMI/50th percentile BMI × 100) using the latest representative growth charts for children and adolescents in Germany.^{Reference Neuhauser, Schienkiewitz, Schaffrath-Rosario, Dortschy and Kurth29,Reference Le Grange, Doyle, Swanson, Ludwig, Glunz and Kreipe30}

The interviews on facilitators and barriers will be semi-structured using an interview guideline created for each perspective. The guidelines begin with open questions to reflect on personal experiences during the process from illness onset to treatment initiation, and get increasingly more problem-centred and focused on potentially modifiable facilitators and barriers. Each interview is expected to last 30 to 90 min. The interviews with the patients will be face to face, and the interviews with carers and physicians will be conducted by phone. All interviews will be audio-taped and transcribed.

According to the applied grounded theory approach, a theoretical sampling method with the goal to include patients whose cases are heterogeneous (particularly with respect to DUI and age group) will be used. Therefore, the sampling procedure can be considered as an iterative process that will be finished with theoretical saturation.^{Reference Strauss and Corbin31,Reference Strübing32} A total number of 25 to 40 interviews is expected (i.e. 10 to 15 triads of patients, carers and physicians including non-response). Each interview will be analysed independently by two members of the study team using grounded theory as the theoretical approach and MAXQDA as analysis software.^{Reference Software33} Grounded theory distinguishes three modes of coding the interview material (open, axial and selective coding). The identification of key facilitators and barriers will be a result of the selective coding, for which we apply established criteria as representativeness (i.e. quantity of material implying the relevance of a factor) and interconnectedness (i.e. relevance for other factors).^{Reference Breuer34}

Substudy II

The key factors identified in substudy I form the basis to subsequently develop an instrument for the combined assessment of multiple facilitators and barriers in anorexia nervosa treatment initiation, which is the main aim of substudy II.

As recommended, anchor examples (i.e. prototypical statements) from the qualitative interviews will be inspected regarding their use for item formulation.^{Reference Pentzek, Wollny, Herber, Porst, Icks and Abholz35} The feasibility of this newly constructed instrument will be tested using cognitive interviews in a sample of ten patients with anorexia nervosa. The cognitive protests aim to detect items that may not be sufficiently easy to understand, clear or recallable. These items will be re-formulated.

Afterwards, the first comprehensive psychometric test of the instrument will be conducted in 40 to 100 patients using principal component analysis. The sample size is based on the recommendation to include at least four participants for each item in order to carry out the planned principal component analysis.^{Reference Tabachnik and Fidell36} Patients will be recruited consecutively in all cooperating study clinics during a time period of 4–5 months. Study patients will be asked to fill out the newly developed instrument as well as the basic information questionnaire from substudy I (Appendix, excluding SCID-II). Moreover, established measures (such as on treatment motivation, perceived stigma) will be applied to validate the instrument depending on the identified key factors. Finally, the instrument will be adapted based on the results of the psychometric item analysis (for example the parameters mean, standard deviation, range, skewness, kurtosis, item-total correlation, Cronbach's alpha) and the principal component analysis.^{Reference Moosbrugger and Kelava37}

Substudy III

In order to finally derive recommendations and prioritise early-intervention approaches, it is essential to quantify the effect of individual facilitators and barriers on the DUI. This is the major aim of substudy III.

Apart from the qualitative interviews, data collection is analogous to substudy I but supplemented by the newly developed instrument to assess facilitators and barriers in treatment initiation, and the German eight-item short version of the Eating Disorder Examination – Questionnaire (EDE-Q) or Child EDE-Q (ChEDE-Q) depending on the age group.^{Reference Kliem, Mossle, Zenger, Strauss, Brahler and Hilbert38,Reference Kliem, Schmidt, Vogel, Hiemisch, Kiess and Hilbert39} The application of further questionnaires will be influenced by the results of substudy I and II.

In the prediction of DUI, we will control for fixed, non-modifiable variables with a potential influence (Table 2). The non-modifiable control variables were a priori selected based on the existing evidence. Specifically, we included all non-modifiable variables that demonstrated a significant influence on the DUI or the treatment initiation in at least two eating disorder studies or one anorexia nervosa study. The variables were theoretically grouped into predisposing, enabling and need factors according to Andersen's behavioural model of health service use.^{Reference Babitsch, Gohl and von Lengerke40} Strong evidence exists for some of the prespecified covariates (i.e. migration background, purging behaviour as operationalised via the anorexia nervosa subtype), whereas the influence of other covariates is only supported by single studies (i.e. insurance status, BMI, eating disorder pathology). Conflicting evidence exists regarding the role of age and comorbidity with some indications of higher age and comorbidity in treatment seekers and other findings that imply the opposite direction.^{Reference Regan, Cachelin and Minnick20} Consequently, no directed hypothesis was formulated. In general, it is expected that a significant amount of variance will already be explained by the prespecified covariates.

Table 2 Prespecified fixed/non-modifiable covariates in the prediction of the duration of untreated illness (DUI)

EDE-Q, Eating Disorder Examination Questionnaire; ChEDE-Q, Child EDE-Q; SCID-I/II, Structured Clinical Interview for Axis I and II DSM-IV diagnoses.

Prior to the prediction, a non-responder analysis including age as independent variable will be conducted. Furthermore, the missing data mechanism (missing completely at random, missing at random or not missing at random) will be analysed and an imputation model depending on the existing mechanism will be selected. Intercorrelations of predictor variables will be examined, and highly intercorrelated variables may be excluded. Finally, a Cox regression model will be computed including the prespecified control variables, the cluster variable ‘setting’ (in-patient versus out-patient), and the key facilitators and barriers as potential predictors of the DUI in anorexia nervosa. The Cox regression model was chosen, because the DUI is a usually heavily skewed, exponentially distributed time-to-event variable.^{Reference Dell'Osso, Glick, Baldwin and Altamura14}

For sample size calculation, we performed a simulation based on the assumptions of an exponentially distributed DUI with an intercept of log^{Reference Strauss and Corbin31} (i.e. mean DUI of 31 months)^{Reference Weigel, Rossi, Wendt, Neubauer, von Rad and Daubmann17}and six uncorrelated predictors with a moderate effect (β= log(1.3)). In order to reach a power of 80%, n = 130 study patients need to be included. Assuming a non-response rate of 25%, n = 173 patients need to be initially approached in substudy III.

Ethics and dissemination

Written informed consent will be obtained from all study participants prior to study inclusion. In minor participants, the consent of all legal guardians, or the consent of at least one legal guardian and written explanation regarding any missing consents are additionally required. Carers of study patients will be invited for study participation only after obtaining the patient's written consent to contact the respective carer. Along the same lines, physicians will only be contacted if the patient has signed a release from medical confidentiality. The physicians will receive a copy of this confidentiality release. The data will be handled confidentially. Reciprocal information between participants of individual triads is not intended. After study participation, all study participants will receive a financial compensation.

The authors assert that all procedures contributing to the FABIANA study will comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2013. Ethical approval was obtained from the Ethic Committee of the Medical Association Hamburg (PV5108) on 11 November 2015. An amendment including an update of the study materials to comply with the European Union general data protection regulation was approved on 18 July 2018. Secondary votes were obtained from the ethic committees of the medical associations in all relevant German federal states (i.e. Lower Saxony, Hesse, Baden-Wurttemberg and Schleswig-Holstein; the primary vote is generally accepted in Bavaria). The substudies will be conducted and reported in accordance with relevant reporting guidelines.^{Reference Collins, Reitsma, Altman and Moons42–Reference von Elm, Altman, Egger, Pocock, Gotzsche and Vandenbroucke44}