|Home | About | Journals | Submit | Contact Us | Français|
There has been much outcome research on interpersonal psychotherapy (IPT) but little investigation of its components. This study assessed interrater reliability of IPT therapists in identifying interpersonal problem areas and treatment foci from audiotapes of initial treatment sessions. Three IPT research psychotherapists assessed up to 18 audiotapes of dysthymic patients, using the Interpersonal Problem Area Rating Scale. Cohen's kappa was used to examine concordance between raters. Kappas for presence or absence of each of the four IPT problem areas were 0.87 (grief), 0.58 (role dispute), 1.0 (role transition), and 0.48 (interpersonal deficits). Kappa for agreement on a clinical focus was 0.82. IPT therapists agreed closely in rating problem areas and potential treatment foci, providing empirical support for potential therapist consistency in this treatment approach.
Psychotherapy research historically focused predominantly on treatment process rather than outcome. Interpersonal psychotherapy (IPT), a time-limited treatment for major depression and other disorders,1,2 has been an exception to that rule. IPT is based on empirical research findings, and its investigators have always stressed outcome research. As a result we know that IPT works, but relatively little about how it works. Research has shown that IPT reduces depressive symptoms, but has barely explored the specific choices and interventions that therapists employ to achieve this clinical improvement. Since IPT works, it makes sense to study its workings.
One basic process question involves formulation of the IPT treatment focus. In the early sessions of the 12- to 16-session acute therapy, the IPT therapist must quickly elicit information about recent interpersonal problems from the patient, condense it into a formulation, and obtain the patient's agreement to this formulation, which then becomes the focus of the ensuing treatment.3 Although every patient's story differs, IPT formulations link the targeted mood disorder to one of four interpersonal problem areas: grief (complicated bereavement), role disputes, role transitions, and interpersonal deficits. The IPT manual1,2 provides definitions and examples of problem areas and focuses on observation rather than inference, minimizing subjective factors in diagnosis. Relative to other psychotherapies—and particularly to psychoanalysis, whose richness has made clinician agreement on formulation notoriously difficult4,5—the limited options for IPT should facilitate choosing a focus. The question remains: can clinicians agree in diagnosing interpersonal problem areas?
The four problem areas have been paradigmatic of IPT since its inception, and IPT supervision has anecdotally suggested convergent clinician assessment. Yet therapist concordance in diagnosing these problem areas has never been studied. Absence of proven problem area reliability has not evidently hindered IPT efficacy to date, but evaluating reliability might enhance understanding of treatment process and outcome for IPT3 and for case formulation generally.5,6 In attempting not a process study but an interrater reliability study to help define this important process issue, we hypothesized that agreement by trained IPT therapists would be high.
IPT problem areas were assessed by three IPT research psychotherapists (a psychiatrist, a psychologist, and a social worker) from a comparative treatment study of dysthymic disorder. All were women; they had a mean (±SD) of 7.7±5.5 years of clinical experience (range 2–13) and 5.3±3.1 years of IPT experience (range 2–8). All three separately rated the same 16 audiotaped initial IPT sessions, and two raters each reviewed two additional tapes, yielding 52 ratings that could be paired for comparison. Data were missing on interpersonal deficits on one rating, and on role disputes on two ratings.
Initial session tapes elicit an anamnesis of the patient's history but predate therapist and patient agreement on a problem area. Three tapes came from another site, from patients with major depression. The remaining 15 were tapes of dysthymic subjects in a randomized 16-week clinical trial who gave informed written consent for treatment and audiotaping. Thus a rater had often herself done the interview on the latter 15 tapes.
Demographics were available on subjects from 15 of the 18 tapes. They were typical of subjects in dysthymic treatment trials: 73% were women, of mean age 44.6±10.1 years, with a range from 33 to 60 years. Eighty-seven percent were white, with 1 (7%) Hispanic and 1 (7%) Asian subject; 20% were married, 60% separated or divorced, and 20% never married. Eighty percent had completed at least a two-year college degree; 87% were employed outside the home, and 1 (7%) was a housewife.
Raters used the Interpersonal Problem Area Rating Scale7 (IPARS; Appendix A), a checklist of previously untested psychometric properties, to rate the presence or absence of each of the four problem areas. They also chose which problem area appeared primary; that is, which they would select as a treatment focus for the patient (IPARS section B.2). When rating patients they had themselves treated, raters were asked to determine problem areas based on the evidence of the single tape rather than on additional knowledge of the patient. Audiotapes spanned the range of IPT problem areas.
The reliability among the three raters was evaluated separately for the classification of each problem area and for classification of the primary problem area, using the analysis of variance approaches to kappa for multiple raters presented by Fleiss.8 Because problem areas are not mutually exclusive, and because we were interested in rater agreement on each of them, separate kappas were computed for each problem area. Landis and Koch9 suggest that a kappa above 0.8 be considered “almost perfect,” between 0.6 and 0.8 “substantial,” 0.4 to 0.6 “moderate,” 0.2 to 0.4 “fair,” and below 0.2 “poor.”
The three raters detected grief on 10 tapes, role disputes on 20, role transitions on 47, and interpersonal deficits on 9. Multiple problem areas were recorded by at least one rater on 12 tapes (67%). The primary clinical focus was classified as grief on 4 ratings, role dispute on 9, role transition on 38, and interpersonal deficits on 1. The interrater reliability kappas were as follows: grief, 0.87; role dispute, 0.58; role transition, 1.0; and interpersonal deficits, 0.48. Kappa was 0.82 for agreement on the primary problem area that would constitute a clinical focus for treatment. On four tapes, one rater felt that the actual therapist's comments influenced her choice of problem area; on one tape, more than one rater felt this. Deleting the latter tape from analyses produced only trivial changes (kappa: dispute, 0.53, deficits, 0.47).
This study, the first of its kind for IPT, found substantial overall therapist agreement on the presence of interpersonal problem areas. This makes clinical sense: eliciting the history of a death should raise the issue of grief; interpersonal struggles should raise consideration of role disputes; and so forth. Interpersonal deficits, which had the lowest reliability, is the least well defined of the interpersonal problem areas.
Results support the utility of the IPARS instrument, which IPT therapists may find helpful to case formulation. Results also indicate the likelihood of therapist agreement on a key component of the diagnostic phase of the IPT process: the identification of interpersonal problem areas. Moreover, therapists had “near perfect” agreement on the more important clinical issue of choosing a treatment focus for IPT. This second issue depends on the first: had therapists disagreed on the problem areas themselves, they would have had greater difficulty in agreeing on a focus. We infer that therapists clearly agreed on the interpersonal focus of the putative treatment course, and that disagreements were more likely to arise over ancillary interpersonal problem areas: for example, a role dispute peripheral to the patient's central issue of grief.
This study does not answer the larger question of whether agreement on treatment focus matters—whether the choice of treatment focus actually influences treatment outcome. Presumably it is important that the therapist choose a problem area that the patient finds convincing. That therapists agreed on problem areas in these cases supports the credibility of the focal IPT problem areas. Patients, too, may easily accept these formulations, which would build the treatment alliance and start therapy off strongly.
Limitations of the study include its small sample sizes of audiotaped sessions and raters. The study also assessed tapes mostly of dysthymic patients. Another diagnosis (e.g., major depression or bulimia) or a treatment group for which additional IPT problem areas have been constructed (e.g., depressed adolescents10) might affect agreement. The therapist on the audiotape at least occasionally might have slanted the interview, biasing raters and increasing convergence; this is always an issue in reliability studies based on taped interviews. We asked raters to score this very question. That raters occasionally noted the influence of the interviewer on their judgment does not mean that they acted on that influence. Raters try to rate objectively; an awareness of influence—like awareness of transference—may in fact enhance an objective perspective.
That some raters had actually treated the subjects, and hence had knowledge extraneous to the taped session, should only have decreased rater agreement. The study by design tested reliability in a “naturalistic” setting: in all but three cases, tapes were drawn from the study and one of the raters had just been doing her job: getting a history from the patient. The other raters were testing their reliability against that of the actual therapist. Arguably the results support the expected: that given only four general life-event-based choices to select among, trained psychotherapists can agree on clinical choices well beyond what would be expected by chance. Part of the beauty of IPT is its simplicity, for patients and therapists alike.
Research has demonstrated the reliability of core conflictual relationship theme (CCRT) ratings, with kappas similar to those reported here.11 Studies like these complement outcome research by empirically supporting the mechanics of the therapeutic process. More such research is needed for IPT.
The authors thank Gregory Hinrichsen, Ph.D., for his generous provision of taped IPT sessions. This work was supported by Grant MH49635 (Dr. Markowitz) from the National Institute of Mental Health, and by a fund established in The New York Community Trust by DeWitt-Wallace. The findings were presented in part at the Society for Psychotherapy Research meeting, Braga, Portugal, June 1999.