Systematic Review of In-Session Affect Experience in Cognitive Behavioral Therapy for Depression

Abstract

One way of attempting to improve the efficacy and effectiveness of Cognitive Behavioral Therapy (CBT) for depression is to identify the processes of change that contribute towards its positive outcome. In addition to well-researched cognitive processes, another possible change process is affect experiencing (AE); i.e., a patient’s affective experience in-session. Theorists, clinicians and researchers have emphasized the role of affective traits, tendencies and symptoms in the development, maintenance, and treatment of depression. We make the case that it may be important to also consider patients’ full range of affect experiencing (AE), as a changeable in-session process that may relate to CBT treatment outcome. This systematic review aimed to clarify what is already empirically known regarding in-session AE in CBT for depression and which gaps in empirical research need to be filled by future studies. The reviewed studies on AE in CBT for depression suggest that it is possible to identify and measure AE. In-session experiencing of positive and negative affect (when it includes cognitive processes) relates to and may predict a reduction of symptoms. We encourage researchers to develop and refine multifaceted process measures and analyses to explore when, how and how much AE can be effectively experienced by patients, and how optimal levels of AE may be facilitated by the therapist.

Appendices

Appendix A

Detailed Description of the Database Searches

The literature review was conducted using the following databases: ScienceDirect, ERIC, SocINDEX, MEDLINE, PsychInfo, Academic Search Complete, ProQuest Central, PsycArticles and Scopus. Search terms included variations on the terms for: (a) in-session affective experiencing (emotion*, affect*, feeling, experien*, process*, in-session, within-session), (b) cognitive behavioral therapy (cognitive therapy, cognitive behavio* therapy, CBT), and (c) depress*.

The search was conducted on abstracts of peer-reviewed journals with ‘AND’ entered into the database search to link the different categories (a, b and c) of search terms. This means that 21 separate searches were conducted for all the variations of the terms for in-session affective experiencing (a), as well as for all the terms for cognitive behavioral therapy (b):

• 1. (a) Emotion* AND (b) cognitive therapy AND (c) depress*

• 2. (a) Affect* AND (b) cognitive therapy AND (c) depress*

• 3. (a) Feeling AND (b) cognitive therapy AND (c) depress*

• 4. (a) Experien* AND (b) cognitive therapy AND (c) depress*

• 5. (a) process* AND (b) cognitive therapy AND (c) depress*

• 6. (a) In-session AND (b) cognitive therapy AND (c) depress*

• 7. (a) Within-session AND (b) cognitive therapy AND (c) depress*

• 8. (a) Emotion* AND (b) cognitive behavio* therapy AND (c) depress*

• 9. (a) Affect* AND (b) cognitive behavio* therapy AND (c) depress*

• 10. (a) Feeling AND (b) cognitive behavio* therapy AND (c) depress*

• 11. (a) Experien* AND (b) cognitive behavio* therapy AND (c) depress*

• 12. (a) Process* AND (b) cognitive behavio* therapy AND (c) depress*

• 13. (a) In-session AND (b) cognitive behavio* therapy AND (c) depress*

• 14. (a) Within-session AND (b) cognitive behavio* therapy AND (c) depress*

• 15. (a) Emotion* AND (b) CBT AND (c) depress*

• 16. (a) Affect* AND (b) CBT AND (c) depress*

• 17. (a) Feeling AND (b) CBT AND (c) depress*

• 18. (a) Experien* AND (b) CBT AND (c) depress*

• 19. (a) Process* AND (b) CBT AND (c) depress*

• 20. (a) In-session AND (b) CBT AND (c) depress*

• 21. (a) Within-session AND (b) CBT AND (c) depress*

Appendix B

Measurements of AE

Operationalization of AE

The reviewed studies used different measures to operationalize different aspects of AE, focusing on its intensity, mixed cognitive-affective process (e.g. affective exploration) or a sole affective process (expressed bodily arousal or inhibition), with either positive and/or negative valence.

Intensity of Cognitive-Affective Process

The intensity of the patient’s AE in-session was measured by three process measures, whereby AE was conceptualized as an integrated cognitive-affective process. For example, the Experiencing scale (EXP; Klein et al. 1986), which appeared the AE measure with the most robust psychometrics, was used by five studies to measure the patient’s depth of moment-to-moment processing during the therapy hour. Patient statements are rated on a 7-point scale in terms of the extent to which they talk about or use their affective experience as a referent during therapy, and explore and reflect on their inner experience to achieve self-understanding and problem resolution (Klein et al. 1986). Studies differed in the timeframe they used for coding the EXP; rating patient’s AE every minute (e.g. Wiser and Goldfried 1993), per 10-minute segment (e.g. Castonguay et al. 1996), per 20-minute segment (e.g. Watson et al. 2006), session excerpts (e.g. Rudkin et al. 2007). The authors report on the construct validity of the EXP as well as its predictive validity in emotion-focused treatments, with published interrater reliability coefficients ranging from 0.76 to 0.91, with rating-rerating correlation coefficients around 0.80 (Klein et al. 1986). In their respective studies, the authors report interrater reliabilities in the good (e.g. CCI of 0.73 and higher in Watson and Bedard 2006) or excellent range (e.g., CCI of 0.83 in Rudkin et al. 2007; 0.88 in; Castonguay et al. 1996).

In addition to the EXP, Watson et al. (2011) also measured the intensity of patient’s in-session emotional processing (which includes cognitive and affective aspects) on the Observer-Rated Measure of Affect Regulation (O-MAR; Watson and Prosser 2004), whereby observers rated the patient’s (1) Level of Awareness, (2) Modulation of Arousal, (3) Modulation of Expression, (4) Acceptance of Affective Experience, and (5) Reflection on Experience on a 7-point Likert scale. They report on preliminary evidence of internal consistency, and construct and predictive validity (Prosser and Watson 2007) as well as significant interrater reliabilities in their early sessions (ICC = 0.78) and late sessions (ICC = 0.87).

The CHANGE rating scale (Hayes et al. 2006) is an observational coding system designed to measure the frequency and extent of change processes in psychotherapy on a 4-point Likert scale. Previous research indicates good interrater agreement and predictive validity for common disorders, including depression (Adler et al. 2013). Its ‘Emotional Processing’ subscale was used by Abel et al. (2016) to score the overall session on patient insight into and exploration of emotional experiences and affective arousal. They reported good interrater agreement on the raw scores for this subscale for all sessions that were coded after training (ICC = 0.80).

Intensity and Valence of Cognitive-Affective Process

Two AE measures that also conceptualized AE as an integrated cognitive-affective process, extended the measurement of AE intensity by including a rating of valence of the affect. For example, in the Assimilation of Problematic Experiences Scale (APES; Stiles et al. 1991), used by Rudkin et al. (2007) and Basto et al. (2016), raters rated the overall (non) verbal patient expressions per session, based on the assimilation model (Stiles 2002; Stiles et al. 1990). The APES describes the relation of an identified problematic experience to the rest of the patient’s presentation on a spectrum from negative to positive valence and intensity on an 8-point Likert scale ranging from warded off/dissociated, and understanding/ insight, to integration/mastery. The other AE measure that also rated the valence of the affect was the rating scale of Therapy Change Processes (TCP; Hayes and Goldfried 1996), used by Hayes and Strauss (1998) to code each session on different subscales of therapist intervention and patient reaction on a 3-point Likert scale. The patient-reaction subscales on Affective intensity and Destabilization both cover negative AE. Affective intensity was measured as the degree of discomfort (non) verbally expressed by the patient in the session when addressing therapeutic issues. Destabilization was defined as the extent of variability or turbulence in somatic, behavioral, cognitive, and affective functioning, and included signs of emotional distress, anxiety, panic, or the emergence of new emotions. Hayes and Goldfried do not report psychometric properties but do report on the estimates of agreement on the coding categories between two raters in their study (ranging from 0.78 to 0.90), as well as the modal agreement for categories (0.82).

Intensity and Valence of Affective Process

The four other measures not only considered the valence of the affect but also measured AE as a purely affect-related process. For example, the Client Emotional Arousal Scale—III (CEAS-III; Warwar and Greenberg 1999) was used by Basto et al. (2016) to assess the valence and intensity of patient emotions, in multiple passages of text within the session. In each identified emotional passage of the transcript, the patient’ s primary emotion was identified and coded for its intensity on a 7-point Likert scale (modal and peak emotional arousal) and its negative (pain/hurt; sadness; hopelessness/helplessness; loneliness; anger/resentment; contempt/disgust; fear/anxiety; shame/guilt, anger and sadness); or positive valence (love; joy/excitement; contentment/calm/ relief; pride/self-confidence). They report previously established interrater reliability coefficients of 0.70 for modal and of 0.73 for peak arousal ratings on the CEAS-III but do not report on its validity.

Another measure of intensity and valence of affective process is the Psychotherapy Process Q-sort (PQS, Jones 2000). The PQS is rated by clinical judges and yields a score from 1 to 9, from extremely characteristic to extremely un characteristic for a given transcribed treatment session, for each of 100 items describing the patient’ s attitudes, behaviors, or experience; the therapist’ s actions and attitudes; and the nature of their interaction. Although the reviewed studies report good construct and discriminant validity of the PQS, and high interrater reliability among 2 raters (0.82 in Coombs et al. 2012; 0.83 in Ablon and Jones 1999; 0.84 in; Jones and Pulos 1993), this referred to the PQS scores of all 100 items. The reported interrater reliability at the individual Q-item level was diverse, ranging from 0.50 to 0.90, however, it was unclear which items had the low reliability ratings. This is concerning, because the three studies that used the Process Q-set to measure AE, did not use this total score and reported on different combinations of items. For example, Coombs et al. (2012) identified two factors of negative affect in their factor analyses, labeled ‘Patient Painful Affect’ (composed of the PQS items: Patient experiences discomforting or troublesome (painful) affect, & Patient is anxious and tense, & Patient struggles to control feelings or impulses), and ‘Patient Inhibition’ (items: Patient feels shy or embarrassed, & Patient does not express anger, & Patient is unanimated or unexcited). Coombs and colleagues also identified a factor of positive emotions labeled ‘Collaborative Emotional Exploration’ (which amongst other cognitive items includes the affective items: Level of catharsis of the patient, &, Patient does not feel wary or suspicious), items also reported on by Ablon and Jones. (1999). Jones and Pulos (1993) also used items from the Process Q set but looked at negative emotions only. They conceptualized a factor (without reporting on the factor analysis) of ‘Negative Patient Affect’ reflecting the extent to which patients felt depressed and anxious or experienced other troublesome affect during therapy sessions (based on the items: Patient feels sad or depressed & Patient feels inadequate and inferior & Patient experiences discomforting or painful affect, & Patient is self-accusatory; expresses shame or guilt & Patient is anxious or tense). Ablon and Jones (1999) did not report on one factor but also reported on these same individual items of negative expressed emotions.

Also, Mackay et al. (2002) measured the patients’ emotional tone (its valence and intensity of the affective process), for one session of each patient using their protocol (Mackay et al. 1998). Observers in their study rated audiotapes and transcripts for each sentence of patient speech on the dimension of pleasure-displeasure (How pos/neg is the patient’s emotion in this sentence?) and the dimension of arousal (What is the level of activation of the patient’s emotion in this sentence?) on a 9-point Likert scale. Although no psychometric properties of these items were reported, and these dimensions were not used in other studies previously, raters established sufficient inter-rater reliability on the two dimensions (ICC of 0.76 for pleasure and 0.75 for arousal).

Negative Valence of Affective Process

Furthermore, the only study that used a self-report measure (Stringer et al. 2010), to measure the intensity and negative valence of affective process, used the Negative Emotions Subscale of the Emotional Arousal Session Report Measure (EASRM; Warwar and Greenberg 2002). At the end of the third session, patients were asked to indicate on a 7-point Likert scale the extent to which they felt the negative emotions: (a) emotional pain, (b) sadness, (c) helplessness, and (d) hopelessness) during session. These four items had an inter-item reliability of 0.81 at the third or fourth session and 0.89 at the final session in their dataset. According to the authors the subscale demonstrated high construct validity and appears to be a reliable measure of negative emotion.

Observer-Ratings of AE

Despite using multiple raters and establishing sufficient reliability across raters, with intraclass correlations of 0.70 and higher (defined as ‘good’ by Shrout and Fleiss 1979), the exact training procedures for the raters remained unclear in most of the twelve observer-rated studies. Ablon and Jones (1999), for example, reported that all raters were trained in the use of the Q technique and Castonguay et al. (1996) mentioned that the coders received 60 h of training but did not report on formal training procedures or criterion scores of training tapes of established expert benchmarked ratings. Also, Basto et al. (2016), simply reported that when raters reached consensus and mastered the coding of the CEAS-III they were given the sessions to code. This means that although raters might agree with each other (high ICC), they might both have misinterpreted the item. Wiser and Goldfried (1993) addressed this concern by reporting mean reliability with the manual criterion (ICC of 0.80). The training procedures reported by Abel et al. (2016), appeared to be most robust, as it included practice of pre-rated training and criterion sessions that continued until interrater reliability of ICC 0.80 was achieved on all items in the criterion coding, as well as regular supervision meetings between the coders and scale authors to review discrepancies and prevent rater drift.