Predicting response to interpersonal counselling (IPC) from case formulation: a systematic comparison between recovered and unchanged depressive cases

ABSTRACT We sought to explore how the process between the counsellor and patient for arriving at a case formulation may predict the outcome of manualized interpersonal counselling (IPC) for depression in primary care. Qualitative content analysis and applied conversation analysis (CA) were used to achieve depth in the understanding of case formulation process among five patients who recovered and five who were unchanged according to quantitative post-treatment change rates derived from Clinical Outcomes in Routine Evaluation– Outcome Measure (CORE-OM). Interaction in the case formulations for the recovered group was generally characterized by a joint construction effort between the counsellor and the patient centred on one problem area. The ability to delimit problems to one area was associated with the patient’s role disputes in social relationships. For the unchanged patients, the case formulation typically reflected unilateral construction of the problem area, and more than one problem area was selected as the focus. The problem areas in the unchanged group were associated with complicated grief or loneliness. The process between counsellor and patient of arriving at and agreeing on a case formulation might potentially contribute to recovery, and it deserves greater attention in training counsellors and conducting research.

2007). We examine here how the process of case formulation during the initial two sessions contributes to the outcome of IPC in the treatment of depression.
"A psychotherapy case formulation is a hypothesis about the causes, precipitants and maintaining influences of a person's psychological, interpersonal, and behavioural problems" (Eells, 2007a, p. 4). The process of formulation provides an opportunity for a shared understanding of the patient's difficulties and can offer a way of tailoring treatment to the individual and his or her singular situation that diagnosis alone does not (Eells & Lombart, 2011;Macneil, Hasty, Conus, & Berk, 2012;Sturmey, 2009). Although case formulation models in psychotherapy and counselling share many common features, each is also distinct from the others (Eells, 2007a). IPC case formulation is based on empirical research demonstrating an association between patients' interpersonal circumstances that appear to be temporally related to the onset of their depression and how complicated bereavement, role disputes, role transitions or interpersonal deficits may predispose patients to depression in these situations (Markowitz & Swartz, 2007).

Steps in IPC case formulation
In its original form, IPC was developed to serve as a simplified version (lasting from three to seven sessions (Weissman & Klerman, 1993)) of interpersonal psychotherapy (IPT) to be administered within primary care. IPT is one of the most empirically validated shortterm treatments for diagnosed depression (Cuijpers et al., 2011). It has been tested on different age and target groups, in different treatment settings and against different cultural backgrounds (Markowitz & Weissman, 2012). IPT usually consists of 12-16 sessions (Mufson, Moreau, Dorta, & Weissman, 2004;Weissman, Markowitz, & Klerman, 2000). IPT is designed for use by health professionals who have already achieved proficiency in some form of psychotherapy, whereas IPC is designed for those who lack psychotherapeutic training (Weissman et al., 2000). At the outset, IPC was used with patients who have low levels of depressive symptoms or distress (subsyndromal symptomatic depression), but in recent years it has also been used with patients who have met the criteria for major depressive disorder (Kontunen et al., 2016;Menchetti et al., 2014). As the IPC procedures, although simplified, are derived directly from interpersonal psychotherapy (IPT) (Weissman & Klerman, 1993;Weissman et al., 2000), the structure of IPT and studies concerning it also deserve to be considered here. The structure of IPC is based on the IPT manual (Klerman, Weissman, Rounsaville, & Chevron, 1984;Weissman et al., 2000), i.e. it is divided into assessment, middle and termination phases, although the treatment can be shorter if the patient had made adequate progress by the sixth session. The structure and time limit of IPC are aspects that have scripts to follow, and it require that the case formulation should emerge no later than the second session. The conduct of IPC case formulation occurs through the following steps (Weissman et al., 2014;Weissman & Klerman, 1993).
Step 1: clarification of symptoms and diagnosis Symptom identification is accomplished by having the patient complete a self-report measure such as the Beck Depression Inventory (BDI) (Beck, Ward, Mendelson, Mock, & Erbaugh, 1961). Then, after reviewing the symptoms, it is important to discuss what depression is. Giving the patient a sick role is also a crucial part of this diagnostic clarification, as this role allows the patient set aside temporarily some activities which seem to be overwhelming or pass certain responsibilities to someone else while recovering. The sick role allows the patient to receive help from others and to cooperate with the counsellor in the process of recovering.
Step 2: evaluating interpersonal relationships It is important in IPC to find out what was going on in the patient's social and family life at the time the symptoms began, what may have been the triggers of the symptoms and who are the important people in the patient's life. The counsellor should conduct a brief "Interpersonal Inventory", i.e. a review of the people who are involved in the patient's life. In addition to knowing more about what problematic relationships there may be, this review will also allow a discussion to take place on the emotional support, social companionship, or practical help available to the patient while coping with the depressive episode.
Step 3: establishing an interpersonal problem area Case formulation should concentrate on current relationships and link the onset of the patient's mood disorder or distress to one of four foci, i.e. interpersonal problem areas (Markowitz & Swartz, 2007). Interpersonal role disputes as a problem area implies conflicts or disagreements within a significant relationship. Here role expectations or values are non-reciprocal and communication problems are serious. Role transitions may be normative and developmental (e.g. graduation, becoming a parent, retirement) or else unwished for or unexpected (e.g. divorce, being diagnosed with a severe illness, becoming unemployed). In a role transition the nature of relationships changes. For example, receiving a diagnosis of a serious illness can involve changes in familial responsibility or treatments that may isolate the patient from sources of social support. Loss of work often involves ending close relationships at the workplace and may also complicate other relationships associated with diminished self-esteem. Complicated grief refers to depressive symptoms that result from serious difficulties in going through various phases of the normal mourning process following the death of a significant person in the patient's life. Interpersonal deficits, loneliness, isolation or sensitivity is chosen as the focus of treatment when a patient presents with a long-standing pattern of impoverished social relationships. One (or at most two) of these problem areas may be labelled and explicitly included in the case formulation which the therapist and patient together agree to work on before the IPC proceeds to its middle phase.
Step 4: making the interpersonal formulation An IPC case formulation is a collaboratively constructed summary of earlier discussions about the symptoms and their relation to interpersonal events and social relationships. Deciding which focus will be the most appropriate for the patient involves identifying options. The choice should be guided by its relevance to the patient's depression, its overlap with the time for which the patient has been feeling depressed and the potential support available to the patient in making a change in the problem area (Law, 2013). The counsellor must check that the patient agrees on the case formulation if it is expressed in the counsellor's own words, as it affirms the therapeutic alliance and underscores the patient's active role in the treatment.

The present study
Case formulation has been highlighted as central to the IPT approach (Markowitz & Swartz, 2007), but although case formulations are achieved in and through interaction, there has been no previous systematic research into the process of IPC or IPT case formulation. Previous research into other forms of therapy has shown that case formulation is a complex interactional activity (Antaki, Barnes, & Leudar, 2004;Davis, 1986) and this reality may explicate the limited amount of direct evidence linking case formulation with outcome. A new insight into how the process of case formulation is associated with the outcome of the counselling would clarify our picture of IPC and the factors that influence a positive response to counselling.

Study design
A systematic case comparison method was chosen for this study (Iwakabe & Gazzola, 2009. First, quantitative outcome measures were used to select recovered and unchanged cases. Then, qualitative analyses by multiple judges in each case were used to capture factors associated with different outcomes.

Setting
The present sample of patients was drawn from a broader group who were participating in a major outcome study (Kontunen et al., 2016) designed to compare interpersonal counselling (IPC) with interpersonal psychotherapy (IPT). The participants for this broader project (N = 40) were recruited from among those seeking treatment at primary care units in the hospital district of East-Savo (population approximately 50 000), Finland. The participants were required to have a diagnosis of major depressive disorder (mild or moderate) assigned by the screening psychiatrist (T.L.). Demographic factors (current age, marital status, educational level and job status) were assessed at the screening visit. The use of antidepressant medication and data concerning continuing or auxiliary psychotherapeutic treatment for depression were assessed one year after the end of the 12-month follow-up by conducting a retrospective review of the medical records of each patient. The protocol was approved by the medical district's ethical committee, and informed consent was obtained from all the patients. The patients' names quoted in this study are pseudonyms and all the details of the individual cases are obscured so that the subjects could not be identified.

Measures and categorization of outcomes
The quantitative outcome data on the treatments were assessed using the Clinical Outcome in Routine Evaluation -Outcome Measure (CORE-OM) (Barkham et al., 2001) and Beck Depression Inventory (BDI) scales (Beck et al., 1961). The BDI is a 21-item self-report instrument, in which the items are scored on a four-point scale from 0 to 3 and summed to obtain a total score ranging from 0 to 63. The BDI is one of the most common self-reported measures of depression, and has been viewed as the gold standard in depression outcome research. The strength of the CORE-OM lies in the breadth of its coverage of welfare and psychological health. Its 34 items are designed to assess the level of psychological distress and the outcome of psychological therapy. The items are scored on a five-point scale from 0 to 4, ranging from "Not at all" to "Most or all of the time" during the previous week. Thus higher scores indicate greater distress. The items comprise four domains: 1) Specific problems (depression, anxiety, physical problems, trauma), 2) Functioning (general day-today functioning, close relationships, social relationships), 3) Subjective well-being (feelings about oneself and optimism about the future), 4) Risk (risk to oneself, risk to others). The sum of the total scale is divided by the number of items. We followed the advice of Leach et al. (2006) and multiplied the CORE-OM points by 10, yielding a more convenient range of 0-40, because it is easier to perceive and assign meanings to scores expressed in whole numbers.
The CORE-OM and BDI scoring systems have shown excellent psychometric properties. Validation of the CORE-OM instrument for the Finnish population has demonstrated similar results to those found for the UK population: Internal consistency (Cronbach's alpha) for the total CORE-OM score was excellent in both clinical (α = 0.94) and nonclinical (α = 0.91) samples (Honkalampi et al., 2017), and correspondingly, the internal consistency of the BDI-21 was 0.89 in a large nationwide population-based sample in Finland (Elovainio et al., 2009). The CORE-OM and BDI scores also exhibited good internal consistency in the original sample of the current study (N = 36) as Cronbach's alpha varied from 0.90 at baseline to 0.96 at 12 months of follow-up in CORE-OM and from 0.82 to 0.93 in BDI. The Pearson correlation coefficient between CORE-OM and BDI was .70 at baseline and .82 at the 12-month follow-up in this sample.
The method of Jacobson and Truax (1991) was used to examine clinically significant changes, where a significant change means one that is unlikely to be due to simple measurement unreliability. This method includes two steps for evaluating individual recovery. The first step calculates the reliable change index (RCI) which specifies the amount of change a patient must show between the pre-and post-test situations for that change to be larger than that reasonably expected due to measurement error alone. The second step calculates the cut-off value to find a weighted midpoint between the means for a patient and a non-patient population. In the current study, the RCI for CORE-OM was set at 6.1 points and the cut-off value at 10.6. These two steps were used to classify individuals into one of four categories: recovered (the patient has passed the cut-off and the RCI in a positive direction), improved (has passed the RCI in a positive direction but not the cut-off), unchanged (has passed neither criterion), or deteriorated (has passed the RCI in a negative direction).

Inclusion of patients
The attrition from admission to the selected subsample is shown in Figure 1. For the present purposes, 10 patients were selected to be integrated into the sample for the current qualitative study: all five cases meeting the criteria for unchanged cases and five recovered cases in terms of their CORE-OM scores using the criteria set out by Jacobson and Truax (1991). The selection of five out of the ten recovered cases was made in an attempt to render the two groups comparable in terms of psychological health at baseline but with a different outcome at the end of treatment. The CORE-OM score at baseline should be moderate (not mild), because none of the unchanged patients had a mild baseline score and such a patient would have needed no auxiliary treatment for depression, thus preventing any confounding treatment effect on the outcome. Using these criteria before any qualitative analysis, we found 5 patients who had recovered to serve as counterparts to the unchanged patients.

Counsellors
Six psychiatric mental health nurses from primary health care units constituted the sample of counsellors for the study. One nurse treated 3 patients, 2 of whom recovered and 1 remained unchanged, one treated 1 recovered and 1 unchanged patient and one treated 2 recovered patients. The remaining three nurses treated only unchanged patients in this sample. The nurses had received 3 days of theoretical training in IPC and had undergone a supervision period of 40 hours with at least one pilot case before the research began. All the nurses had had at least 10 years of outpatient or in-patient experience with depressed patients.

Researchers
The research team was composed of a clinical psychologist and psychotherapist (J.K.), a sociologist and occupational therapist (E.W.), a physician specialized in psychiatry and family therapy (T.L.), a professor specialized in general practice and psychiatry (M.T.) and a professor, psychoanalyst (IPA) and family therapist (J.A.). In terms of biases, all five researchers liked training community therapists in psychosocial treatment skills, although they varied in how comfortable they felt using brief psychotherapies or counselling.

Treatments
The treatments consisted of 7 weekly sessions following the structure of IPC as laid down by Judd, Weissman, Davis, Hodgins, and Piterman (2004) and the protocol of Menchetti et al. (2010), Menchetti et al. (2014). The sessions lasted 45 minutes and the purposes of the visit were outlined in a 30-page session-by-session checklist for the structure of IPC in Finnish modelled on Weissman and Klerman (1993) and Judd et al. (2004). The first two pages introduced the basics of IPC, the assessment process and how to orient the patient with respect to the subsequent IPC sessions. The provision of a written description of the conduct of the IPC for the patient (a patient information sheet) was included here. The third page guided the counsellor in identifying depression symptoms, educating the patient with regard to depression and giving the patient permission to adopt a "sick role", i.e. taking a break, asking for help and accepting responsibility for working towards recovery. It was also possible for the counsellors to give the patients homework sheets (pages 4-7) at the end of the first IPC session and review the answers during the second session. These patients' self-report forms comprised questions concerning life events in the IPC problem areas (Weissman, 1995). The following three pages contained a closeness circle and the charts for summary notes based on an interpersonal inventory in order to obtain an overview as to who were the important people in the patient's life, which relationships would give them support and which might be most closely tied to the depressive symptoms. Based on all the information obtained in the interpersonal inventory, and reviewing what was going on in the patient's life at the time the symptoms began, the counsellors were guided to suggest a relationship between the patient's symptoms of distress and current life stress, focusing on one (or at most two) IPC problem areas (grief, interpersonal disputes, role transitions, or interpersonal deficit/isolation) (on page 11). The manual emphasized that these connections should be made explicit, but also open-ended when talking with the patient. The patient should agree with the case formulation in a manner which signals that it is jointly shared by the patient and counsellor. After the pages concerning the initial phase, the next pages targeted the middle phase containing guides to encourage the patient's capacity for coping with the problem area that had been identified, using such techniques as clarification, communication analysis or role play (pages 12-28). These pages included optional worksheets for the four problem areas to be completed along with the sessions. On the last pages the counsellor was guided to discuss the major problem area in depth and review the development achieved in the course of the treatment and the patient's current state. Adherence to the treatment protocol was ensured by using session-by-session checklists, audiotaping all treatment sessions and discussing the treatment protocols in regular supervision groups. Any complaints or severe side-effects were also discussed with the therapists in the regular supervision groups.

Qualitative data and data analysis
The qualitative data analysis was carried out through four stages.
First stage: the rich case record As a useful starting point, the first author (J.K.) gathered all the data obtained from the 10 patients in rich case record form (cf. Elliott, 2002), comprising basic facts demographic and characteristics forms filled in by the patient at the screening visit, the researcher's counselling process notes about the patient's history, present illness, quantitative outcome measures and an overview of the initial sessions from the transcribed audiotapes.
The second stage: content analysis of the patients' descriptions of their close relationships and the problem areas Since it is stated in the IPC manual that the case formulation needs to be established and presented during the initial two IPC-sessions, the audio recordings of these sessions (N = 20) were selected for each of the ten patients for more precise qualitative analysis. As each session lasted 45 minutes, the data examined for this study involved approximately 15 hours of interaction (10 patients x 2 sessions x 45 minutes).
In the data analysis, the original recordings were listened to a number of times by the first (J.K.) and the second (E.W.) authors separately and all sequences in which interpersonal relationships, the problem areas or the goals of the treatment were explicitly mentioned by the counsellor were identified, transcribed, and entered into Atlas.ti 7.5.16 (1993-2017) (http://atlasti.com/) software for coding and analysis. Atlas.ti is a workbench for the qualitative analysis of textual, graphical, audio, or video data. Although the coding techniques are based on the ideas and terminology used in grounded theory (Glaser & Strauss, 1967), it is possible to use this methodology with any systematic approach to unstructured data.
Next, the first author (J.K.) used directed content analysis (Hsieh & Shannon, 2005) to evaluate the patients' descriptions of their close relationships and the problem areas from the transcribed audiotapes. This directed content analysis, also referred to as deductive category assignment (Mayring, 2014), means that categories are deduced from the theory, other studies or previous research. In this study, the categories of close relationships were deduced from the theory and practices of IPC (Weissman et al., 2000): Who are important people in the patient's life and is it possible for the person to perceive or evoke support from a close person? The categories of the problem areas were derived from the four problem areas of IPC (Klerman et al., 1984;Weissman & Klerman, 1993;Weissman et al., 2000): Which problem areas were negotiated and decided upon as foci for the intermediate sessions?
Using the qualitative software program Atlas.ti, the first author (J.K.) coded the transcribed material into topic segments according to content shifts in the case formulation phases (discussing symptoms, evaluating interpersonal relationships and establishing an interpersonal problem area). To explore the segments more specifically, each topic segment was coded in terms of a key issue. Code names for the key issues were first taken as far as possible from the patient's own words, e.g. the code for the closest interpersonal relationship: "husband similar to dad and unable to tolerate worries". Then this and other similar codes were collated into a more general code: "Inadequate perceived support from spouse" (See Appendix 1).

The third stage: conversation analysis of the case formulations
In the third stage of analysis, the second author (E.W.) conducted an applied conversation analysis (CA) concerning what the therapists actually did in the session when they were making the interpersonal formulation. At this point the analyst was unaware of the outcomes of the cases.
Conversation analysis is a qualitative micro-analytical approach to the study of the organization of human interaction at its finest level of detail. CA highlights the fact that different kinds of social actions are organized into sequences (Sidnell & Stivers, 2013). This means that each utterance gains its meaning in relation to the prior utterance and poses implications for subsequent utterances (Heritage, 2011). Following this idea, the analyst explored the process of case formulation as an interchangeable conversation in which the patient and the counsellor relate their conversational moves to the preceding context. In our analytical procedure the recordings were listened to several times and passages in which the counsellor explicitly refers to the problem areas or treatment goals ("case formulation segments") were identified and collected from the data. The transcribed collection from a dataset for CA consisted of 25 case formulation segments (2-3 per case). These segments involved 1) the counsellors turn explicitly referring to a problem area or treatment goal, 2) the patient's response, and 3) the counsellor's next remark dealing with the patient's response. The detailed conversation analysis transcripts display the words as they were said and indicate pauses within and between utterances and overlapping speech (see Hepburn & Bolden, 2013, and the simplified transcription symbols provided in the footnote 1 ).
Next, all these segments were qualitatively analysed case by case, to specify the nature and variation of the phenomenon in question. Attention was paid to their content and reoccurring lexical design (the vocabulary, words or morphemes of a language used). At this point in the analysis, the patients' orientation with respect to the clinicians' problem formulation was also explored. This was done by focusing on the patients' turns after the counsellors' topicalizations of a problem area.
The fourth stage: validation of the qualitative data analysis The last step of the data analysis was aimed at increasing the trustworthiness of the results. The first author (J.K.) compiled the findings from the rich case record, the content analysis and the conversation analysis (See Appendix 1). To demonstrate the extent to which these represented the sample of participants (See Table  1), they were divided into categories labelled as general (including all or all but one of the cases, 4-5), typical (more than half of the cases, 3), and variant (at least two cases) (Hill et al., 2005;Knox, Schlosser, Pruitt, & Hill, 2006). Analytical observations regarding the case formulation segments that had been analysed by means of conversation analysis were also discussed at two group meetings (data sessions) attended by trained CA researchers (unaware of the outcomes). In these discussions, the analytical findings from the conversations were tested against analytical observations made by other researchers. Data sessions are a standard means of quality assurance for CA data analysis. Then the first (J.K.) and the second (E.W.) authors listened to the audiotapes while reading the transcript again and checked the categories of the content and process of case formulation once again. Finally, the selected examples and extracts and the outcome data were discussed several times with all the research team members, who shared opinions, disagreements and feelings regarding the findings, thus reducing the biases inherent when just two persons were analysing the data. These discussions were inspired by the notion of consensual qualitative research (Hill et al., 2005).

Results
Following the steps in IPC case formulation, the results are presented in four categories (symptoms, interpersonal relationships, interpersonal problem areas and making the interpersonal formulation). The data are presented below using the frequency labels general, typical and variant (see Table 1). In order to exemplify the variation in the categories six data examples from the recovered and unchanged groups will be presented in the following sections. A detailed summary of the findings for the recovered patients (Paula, Daniel, Carolyn, Joanna and Mary) and unchanged patients (Ann, Dorothy, Helen, Lisa and Alex) is presented in Appendix 1. Table 1. Core categories, subcategories and frequencies regarding the process and content of case formulation for the patients in the recovered group (n = 5) and unchanged group (n = 5). Note. General = 4-5 cases, typical = 3 cases, variant = 2 cases. Findings representing 0-1 = -Results in bold indicate those categories that most clearly separate the groups (labelled following Knox et al., 2006).

Demographic factors, symptoms and outcome data
The recovered and unchanged groups were comparable in terms of sex (4 females, 1 male) and educational level (4 vocational qualifications, 1 master's degree). Four members of the recovered group were employed or studying, and one was on sickness benefit, while 3 of those in the unchanged group were employed or studying and 2 were on sickness benefit. Three of the recovered patients were diagnosed as having recurrent depression, as compared with 2 of the unchanged patients. None of the patients had problems with alcohol consumption. The changes were significant in that there were no statistically significant differences in the pre-treatment scores between the recovered and unchanged groups (CORE-OM: p = 0.841; BDI: p = 0.905), but the differences between the groups at the 12-month follow-up examination were significant (CORE-OM: p = 0.008; BDI: p = 0.016). No subsequent psychiatric or psychotherapeutic treatment had been needed by the recovered patients one year after the last IPC followup session, whereas three of the unchanged patients were continuing treatment with the same counsellors, one was on a "coping with depression" course and the fifth patient had become motivated to continue in intensive psychotherapy.

Evaluating interpersonal relationships
All the patients who recovered were married, and they generally mentioned that they were able to discuss things with their spouses and felt they had social support available outside the treatment. For example, Mary had experienced many difficulties at work, but she felt that her husband supported her. "When I told him about my work, he said that the organization I work for is truly incomprehensible. When I showed him the emails that I'd received from the management, he said they really have some sick people working there." In contrast, the unchanged patients had generally had less social support available. One example of the generally poor social support in the unchanged group, was Helen. She had had a very close relationship with her late mother but described her husband and father as being distant people for her. Helen described her relationship with her husband as follows: "The children say that when their father's eyes roll around, then one has to be quiet. He does not hit anybody, but he gets really mad if I ask too many questions. . .That's what I've been saying to mom, did I have to take a man who was similar to dad? Is it always so that a daughter looks for someone similar to her father, even if it means that she goes for an alcoholic."

Establishing an interpersonal problem area
In the recovered cases a single problem area was generally formulated as the focus for the treatment. Four of the foci were role disputes, including three disputes at the patient's workplace or family business (Mary, Daniel, Paula). Mary also emphasized role transition due to her husband's unemployment, Carolyn's problem area was arguments with her husband, and Joanna's depression was associated with role transition after childbirth.
"It (childbirth) turned everything upside down, like, so you didn't know what to expect at all. That's the reason why I apparently started getting stressed about everything too much. Because it was that everything came as a total surprise." In contrast to the patients who recovered, those whose condition remained unchanged generally had problem formulations that involved more than one focus. Typically, the problem area was role transition (major life change). Ann had suffered from cancer and was very anxious about her daughter, who had moved away from home, Dorothy's situation was unrelieved from the beginning of the treatment because of her fatal disease and she had been isolated from her friends during her sickness leave, while Alex had returned to his family home after his studies and had socially alienated when unemployed, so that he could not name a problem area of his own. As a variant, Lisa and Helen in the unchanged group represented the complicated grief category. They had felt depressed after their mothers had died. Soon after the IPC treatment started Lisa's father also died and she had conflicts with her sisters about their parents' estate.
" There's been this kind of misunderstanding, well, from relatives about the funeral, when there were a couple of aunts there and then my brother there and when I kind of started crying more, they said just let it out. But they kind of misunderstood that I was sort of mourning over how difficult it was to be with my sister, but they kind of understood that I was crying for my father."

Making the interpersonal formulation
The CA analysis yielded two main types of interaction in connection with case formulation: joint and unilateral construction of the problem area. The interaction aspect of problem formulation in the recovered group was generally characterized by joint construction of the problem area. The problem area was collaboratively discussed and the counsellors were generally responsive to the patients´emotional expressions. They adapted to the rhythm of the patient, gave sufficient space and allowed for breaks and verbalized feelings. They also explored and deliberated over ideas together with the patient and presented their suggestions as tentative ones that were open to joint exploration. The first case example, representing commonalities among the recovered patients, is a woman (Paula) whose problem area arose naturally and the conversation analysis reflects strong affiliation and agreement between the patient and her counsellor. This extract was selected on the basis of presenting the sequences in which the problem area was labelled and the decision made concerning the focus of the therapy.
Paula was working full-time in customer service when she was diagnosed with moderate depression. Her interpersonal inventory demonstrated that she had good relationships with her husband and children, but she had had conflicts and disagreements with a co-worker that significantly affected her job satisfaction and well-being, resulting first in burnout and then in depression and sick leave. Several rounds of negotiations had been held at her workplace to resolve the problems, but no progress had been made.
In the extract the counsellor is reviewing a questionnaire on interpersonal problem areas that the patient has filled in as a home assignment. They are discussing first the question related to grief, the death of one of Paula's relatives. After this discussion, the counsellor returns to the role of the disputes at her workplace, the interpersonal issue they had been discussing previously. [indeed 11 Couns: .hhhmm it says here ((on the questionnaire form)) that you feel 12 some distrust for him ((omitted 1.5 min of talk on meetings arranged at the workplace to resolve the problems)) 13 Couns: mm (0.2) krhm-hm and indeed these (0.2) because of these things 14 ((problems at the workplace)) (0.5) you ran out of strength and then 15 this sick leave came along 16 Paula: mm 17 Couns: .hh yes and then we can presumably decide quickly that this is it 18 Paula: yes In line 4 the counsellor concludes that, based on their discussions, the role disputes (at the patient's workplace) are a suitable starting point. She uses the personal pronoun we, which highlights the shared process of case formulation. The counsellor also invites the patient to adopt a reflective stance towards her problems. She describes the problem area as something that can be looked for and unravelled from several perspectives (l.9). The patient responds with strong agreement (indeed, l.10). In lines 11-15 the counsellor explicitly links the symptoms of depression and onset of the patient's sick leave to the interpersonal problems at the workplace. The patient confirms this, and the counsellor moves on to make a decision on the problem area, stating explicitly that it is a joint decision, "we decide", and using a firm declaration, "this is it" (l.17), which the patient confirms (l.18).
When the discussion was continued later, they proceeded to investigate the patient's thoughts concerning the goal of the treatment, whereupon the patient described her workplace problems in detail. In the first lines (1-4) the counsellor formulated the patient's description. 01 Couns: so that somehow you're sad about having a quarrel about some 02 things and then your working has become troublesome and you 03 have needed to avoid each other and maybe needle a bit and act 04 unprofessionally in these personal relationships 05 Paula: yes 06 Couns: so we wish that that would change and the goal would be for 07 instance that that (0.5) could you say it in your own words so that 08 Paula: how would I say it the goal would be perhaps that I could have 09 some collaboration with my closest co-worker and we could even 10 like advan[ce our working practices so 11 Couns: [yes indeed 12 Paula: so that we wouldn't need to think that as an obstacle 13 that which one of us is better in a way or 14 Couns: simply equal 15 Paula: which one of us shines more brightly 16 Couns: yes The formulation (l.1-4) highlights the core feeling experienced by the patient. In this way the counsellor is emphasizing the importance of the patient's subjective experience when investigating the problem area. The counsellor also highlights the shared nature of the process. In lines 6-7 she first moves towards suggesting that she shares the patient's desire for a change and then invites the patient to define the focus of the therapy in her own words. The patient finds it difficult to respond at first, but then provides an elaborate description of the desired state of affairs (l.8-10). The counsellor provides minimal responses (l..11) to support the drift of the patient's argument, and then collaboratively complements the patient on her description (l.14).
Interaction within problem formulation varied more in the unchanged group than in the recovered group. As a variant, the interaction was similar to that in the recovered group, in that the counsellors collaboratively constructed the problem areas and emotional responsiveness prevailed in the interaction. However, the interaction typically reflected difficulties between the patient and the counsellor: unilateral construction of the problem area, counsellors' difficulties in responsiveness and fact that the counselling manual was adhered to strictly. The counsellors asked the patients to name the problem area or goal of the therapy but did not investigate their replies collaboratively or develop them any further. They also tended to overlap with the patient's speech, effectively preventing the patient from saying anything more on the subject, and then returned to reviewing the questionnaire form, bypassing the patient's expressions of emotional experiences. The case of Alex captures the commonalities among the unchanged patients for whom the conversation analysis reflects unilateral construction of an ambiguous problem area and weak agreement between the patient and counsellor.
Alex was a young man who simply felt that life had nothing to give him. He had a vocational qualification in woodworking and carpentry, and after his studies he had returned to his family home, where he remained unemployed for months. While still studying he had met a girl and they had soon decided to live together, although he felt that they had no real relationship. He drank a lot of alcohol with his companions, and when the relationship with his girlfriend ended, he felt that this was a relief: "one area of life less that has to be coped with." He passed the days with other men in bars "chewing the fat", but he no longer perceived substance use as a serious problem; he simply felt that life had nothing to give him. The following extract is an example of an interaction in which the problem area is discussed. As in the previous example, the counsellor is reviewing the questionnaire on interpersonal problem areas that had been a home assignment. were talking about the different areas that we went through and 07 then there was this question of taking responsibility 08 (0.5) 09 Couns: .hhhh well erm (0.5) based on that I was thinking that (0.2) that 10 (0.5) could it be connected to what we speak of (0.2) in this context 11 with the term role transition 12 (3.0) 13 Couns: so that at the end of one's studies one somehow gets that anxiety 14 and that kind of erm that kind of .hh yes anxiety and it feels difficult 15 like (0.2) to think about and make choices concerning the future 16 so erm (0.2) so it's of course natural that 17 at the end of one's studies the future is already kind of looming 18 there and where to find a job and (0.2) 19 where to settle down and those so[rts of things 20 Alex: [mm 21 Couns: so would it be that (0.5) that kind of erm 22 (3.0) 23 Couns: so in a sense it's all very natural that we have these changes in our 24 lives it is all the time we have these changes and we kind of give up 25 some old things and gain something new in their place 26 (2.0) 27 Couns: we find new things to replace the ones we have lost or 28 given up so that .hh 29 (3.0) 30 Couns: what do you think could it simply be that 31 from the role of a student (0.2) from the freedom of being a student 32 and then that kind of erm (0.2) it just came to mind from that 33 responsibility it just came to mind that 34 Alex: I gu[ess so 35 Couns: [that having a more responsible life and erm 36 quite often it is connected to the stability and a bit erm 37 like a bit more permanent 38 Alex: mm The counsellor concludes that the patient seems to get along with his friends and it is not interpersonal sensitivity that needs to be worked on (l.1-5). He then goes on to suggest that taking responsibility could be the focus of the therapy (l.6-7). The patient does not respond, and the counsellor goes on to connect the theme of responsibility with the concept of role transition (l.9-11). The patient remains silent and the counsellor further connects the concept with the patient's current situation (l.13-19). At this point, the patient weakly agrees (mm, l.20). The patient's minimal response comes right after the counsellor's list of students' concerns. This may indicate that the patient recognizes these concerns as his own. The counsellor does not elaborate on this, however. Instead, she makes a move to normalize the patient's situation, treating it as an example of more general, normal changes that people tend to experience in life (l.23-25, 27-28). The counsellor then suggests again that role transition should be selected as the focus of the therapy (l.30-33), and this time she explicitly elicits the patient's agreement (what do you think, could it be). The patient hesitantly aligns himself with the counsellor's suggestion (I guess so, l.34) but the counsellor continues to talk, overlapping with the patient's response, and thus misses the opportunity to pick up on the patient's theme. Later in the session the counsellor engaged the patient in discussing his frustrations and suicidal thoughts, and this discussion revealed his alienation, which went much deeper than the focus of role transition. He talked about how pointless people's lives are and how he had constantly had suicidal thoughts. At this point, it seemed to be extremely difficult to get the patient motivated and engaged in limiting himself to one interpersonal problem area.

Summary of the data
The frequencies of categories are presented in Table 1, and more detailed findings are given in Appendix 1. Joint construction of the problem area between the counsellor and the patient was general in the recovered group. When making the interpersonal formulation, the counsellors were generally highly responsive to their patients and their patients manifestly agreed on the problem area. Also, it proved possible in the recovered group to jointly limit the patients' problem to one area, generally role disputes in social relationships. By contrast, unilateral construction of the problem area by either the counsellor or the patient was typical in the unchanged group, whereupon the counsellors bypassed the patient's responses and ended up with more than one problem area (complicated grief and role transition or role disputes), or else failed to define a manifest problem area during the initial two sessions. Simultaneously with the findings of differences in the process of case formulation, the recovered and unchanged groups also proved to have different backgrounds, the recovered patients generally feeling that they had adequate social support from their family and while the unchanged patients generally felt that they lacked external social support.

Discussion
To our knowledge this is the first study of IPC from a process standpoint, in order to explore predictive factors in the context of case formulation, which has been highlighted as central to the IPT approach (Markowitz & Swartz, 2007). Our novel findings were that the interaction that took place in the recovered group at the case formulation stage was generally characterized by joint construction of one problem area associated with role disputes in social relationships. The process of case formulation varied more in the unchanged patients, and it was impossible to define a focus on any one problem area., The interaction typically reflected difficulties between the patient and the counsellor: unilateral construction of the problem areas associated with complicated grief and major life changes.
An interpersonal formulation procedure requires that the resulting formulation should be convincing for both the counsellor and the patient (Markowitz & Swartz, 2007), but we have known little about how an understanding of the patient's problems is negotiated in actual counsellor-patient interaction. Our conversation analysis of IPC case formulation provides a working alliance perspective on this data. It has been shown that if an emotional bond and a shared idea of the goals of the therapy cannot be formed during the initial three sessions, it is very likely that the patient will not benefit from the intervention (Horvath & Bedi, 2002). Our findings indicate that the counsellors in the recovered group rhythmically attuned their expressions to the patients' speech, and in this way they validated the patients' emotional expressions and emphasized the importance of their subjective meanings with regard to their experiences in the investigation of the problem area (cf. Weiste & Peräkylä, 2014).
It should be acknowledged that our conversation analysis findings were reliant on several other factors that may have contributed the various processes of case formulation. The heterogeneity of depression may be one explanation. There were no statistically significant differences in the pre-treatment CORE-OM or BDI severity scores between the recovered and unchanged groups, but the present unchanged patients seemed to be experiencing a quite different form of depression from the recovered patients. Conflicts at their workplace or disputes with their spouse preceded depression in the recovered group, whereas depression in the unchanged group was associated with irreversible life changes (life-threatening illnesses, leaving work due to illness or the death of a close relative) or a lack of prospects for the future (withdrawal from life and suicidal ideation). Rantala, Luoto, Krams, and Karlsson (2018) have argued that depression is not a single disease. It is a group of separate syndromes, with patients differing remarkably in symptom profile, pathophysiology and treatment responsiveness. They classified depression into 12 subtypes based on evolutionary explanations, focussing on the underlying reasons (triggers) for depression. Of these proximate mechanisms, hierarchy conflict refers to events such as unemployment or professional hierarchy conflicts. Our data support the link between focussing on and resolving hierarchy conflict at work and improvement in IPC, since three of the patients who recovered established this as the problem area. This is in line with earlier findings in IPT studies that have suggested that IPT may be most suitable when a conflict with a significant person is in focus (Gunlicks-Stoessel, Mufson, Jekal, & Turner, 2010). On the other hand, loneliness, traumatic role transition after a serious loss of health and grief as underlying reasons for depression predicted a poorer outcome in our data. These observations support earlier findings that suffering from a concurrent physical illness may limit the benefit of IPC (Holmes et al., 2007;Menchetti et al., 2014). Complicated grief (combined with role disputes or role transition) was also associated with a poorer outcome, which is in line with the findings of Markowitz, Bleiberg, Christos, and Levitan (2006) in dysthymic patients.
Another factor related to the patient may also be an explanation for the varied processes of case formulation. All the patients who recovered had social support from their spouses. This is consistent with some previous studies of IPC concerning who might benefit most from this counselling, in that being in stable and supportive relationships was associated with a better outcome (Badger et al., 2013;Badger, Segrin, Meek, Lopez, & Bonham, 2006). Two of the unchanged patients were single, having isolated themselves from close relationships and three were married but without social support from their spouses. The overall evidence shows that spousal support is important and the most salient social support for adults and older adults against depression (Gariépy, Honkaniemi, & Quesnel-Vallée, 2016). In particular, the spouse's or partner's role as a provider of social support is central during the recovery phase of depression after a somatic disease (Salakari et al., 2017). This was lacking for two of the unchanged patients in this study.

Strengths and limitations
The strength of this systematic multiple case comparison study lies in its qualitative enrichment of the quantitative findings of our randomized controlled trial (RCT) in a naturalistic clinical setting (Kontunen et al., 2016), and thus bridges the research-practice gap (Iwakabe & Gazzola, 2009). It illustrates how a manualized treatment protocol can be applied to specific individuals (Iwakabe & Gazzola, 2014). This allows one to demonstrate the complexity of the change that takes place in individuals (cf. Hill, Chui, & Baumann, 2013). Our findings here highlight the diversity of the interactional factors lying behind the statistical outcomes and explain therapeutic interaction at its finest level of detail. Thus, the findings add to our understanding of the therapeutic relationship, explaining why some patients recovered and others were unchanged, at least as far as the particular treatment method was concerned (Norcross & Wampold, 2011).
Although qualitative enrichment is the strength of this study, it encompasses a number of limitations that have to be taken into account. Our study design yielded a complex picture of the interplay between the contributions of the patient and the counsellor to the therapeutic relationship and to patient change. The study design did not allow us to isolate the effect of the counsellor on the outcome from that of the patient, but our analysis did reveal, for instance, that the recovered and unchanged groups differed with regard to the level of external social support at the pre-treatment stage. This may explain why the patients recovered or were unchanged, as may the interaction between the counsellor and the patient.
Evaluating patients' descriptions of their close relationships and problem areas according to IPC focuses by means of directed content analysis (Hsieh & Shannon, 2005) has its limitations. Although relationships and focuses are labelled explicitly in the IPC case formulation, the fact that only one person (J.K.) was identifying the categories might have directed us to find evidence that is biased to support preconceptions (Hsieh & Shannon, 2005). The same problem concerns CA analysis, even though the second author (E.W.) was unaware of the outcomes of the cases when conducting the analysis. We strived to increase the trustworthiness of the analysis by listening to the audiotapes while reading the transcript through a number of times and discussing selected segments of the recordings amongst the whole research team and at two group meetings (data sessions) attended by trained CA researchers who were unaware of the outcomes. The selection of case examples and excerpts from the cases is also a critical matter. It ultimately remains for the readers, however, to decide whether the choice of data and excerpts illustrates the formulation as it took place with the recovered and unchanged cases in a plausible manner.

Implications for practice
One essential issue for clinicians is to what degree a case formulation model is applicable and useful in ordinary practice (Godoy & Haynes, 2011). Our qualitative findings address the underlying concepts and rationale for case formulation at the primary health care level, for the circumstances in which patients actually seek treatment for the first time and are treated by nurses who are sharing the responsibility for the treatment of depression with general practitioners. At its best, the content and process of IPC case formulation can be time-saving, can rely on general knowledge of depression, and can help the counsellor to identify what is workable for the patient and avoid areas that may be interesting but do not further recovery (cf. Eells, 2007b). In this regard, our findings support, at least in the context of less complex problems (cf. Groenier, Pieters, Witteman, & Lehmann, 2014;Kuyken, Fothergill, Musa, & Chadwick, 2005) the suggestion that good outcomes may be achieved by novice therapists' by staying close to the accepted protocols and manuals.
But there are also several limitations to the following of a manual-guided case formulation procedure in IPC. Since the counsellor has to label problem areas explicitly and limit the choice to one or at most two of these (Markowitz & Swartz, 2007), the case formulation may not fit the individual situation and there is a risk of not recognizing, or of misunderstanding, the patient's problems (cf. Eells, 2007b). Three IPC case formulations in this subsample were characterized by unilateral construction of the problem area and difficulties in responsiveness. The problems in these cases were complicated ones, and more than one problem area was labelled, or the problem area was indefinable. The content and process of IPC case formulation may be an insufficient approach in complex cases of this kind, and we can postulate that some other high-quality case formulation protocol implemented by advanced practitioners would have improved the outcomes (cf. Kuyken et al., 2005).
One option for increasing the utility of IPC case formulation would be to develop the manual further. In the light of the review by Carroll and Rounsaville (2008) the challenges inherent in moving a manual from research to clinical practice and making it more "clinician-friendly", our data strongly suggest that the IPC manual should place greater emphasis on collaboration and joint construction of the case formulation, as this is a crucial element in the treatment outcome. Also, the manual should provide more guidance in troubleshooting, especially on how the counsellor can help the patient to choose the most salient problem area even when there may be several problem areas altogether. We conclude that classification of the problems in accordance with the IPC procedure should suffice if the counsellor and patient are jointly able to limit the problems to one area. Although IPC can provide structure and content to enable primary care practitioners to evaluate, support and prioritize patients with depressive symptoms (Weissman & Verdeli, 2012), counsellors would need guidance on how to manage when the patient cannot select a problem area and cannot set achievable goals and on how to make practical changes to treatment strategies. Counselling psychologists may be particularly well suited as providers of process-oriented research for this task and training for counsellors in how to use the IPC manual in a more "clinician-friendly" manner (cf. Allan, Campos, & Wimberley, 2016).
In summary, this multiple case study provides new detailed information about the content and process of case formulation which may be helpful for counsellors working with IPC or training others in its use for the treatment of depression at the primary care level. The present findings emphasize the importance of joint construction of the problem area between the counsellor and the patient, and how the heterogeneity of depression and external social support may account for the efficacy of counselling. Further qualitative process research regarding case formulation and the mechanisms of change in IPC can be expected to provide a more comprehensive understanding of the factors that influence a positive response to counselling. Note 1. Transcription symbols (simplified from Hepburn & Bolden, 2013): [] Overlapping talk (0.0) Pause: silence measured in seconds and tenths of a second hh An in breath / hh An out breath (h) Laughter particles ((word)) Transcriber's comments.

Disclosure statement
No potential conflict of interest was reported by the authors. Jukka Aaltonen, M.D., Ph.D., is psychoanalyst (IPA) and professor of family therapy (emeritus) at the University of Jyväskylä, Finland. His main scientific interest has been in the development of the Need-adapted approaches of the treatment of schizophrenia within municipal psychiatry. Table A1. Main findings regarding case formulation for the patients in the recovered group (n = 5) and unchanged group (n = 5).