Analyzing medical dialogues: strength and weakness of Roter’s interaction analysis system (RIAS)

https://doi.org/10.1016/S0738-3991(02)00014-9Get rights and content

Abstract

Roter’s interaction analysis system (RIAS) is analyzed in this article. Ground rules of linguistic interaction analysis, emphasizing meaning as a product of interaction and turn taking as a basic principle for the understanding of interaction are briefly introduced. Specific aspects of the application of RIAS are discussed and a number of adjustments and/or specifications suggested: (1) utterances should be defined in terms of content and turn taking criteria; (2) the recording system should allow for registering interruptions; (3) pauses or silences should be scored on the basis of functional criteria and not as demarcation in the communication; (4) clear distinctions should be made between the categories of “backchannel” and “agree”; (5) questions should be coded according to function rather than linguistic form; (6) some of the socioemotional categories may appear too narrow, others too wide; (7) crying should be included in the coding scheme as a separate category.

Introduction

Several coding systems have been applied in research on medical communication, and different aspects of their applicability to doctor–patient interaction have been discussed in the literature [1], [2]. The “Roter’s interaction analysis system” (RIAS) is probably one of the most widely used methods [3]. RIAS has proved to be highly reliable when applied by trained coders, and the system has proved to be clinically meaningful in a number of empirical studies [4], [5].

In the RIAS the verbal dialogue is analyzed utterance by utterance, in terms of mutually exclusive categories: 29 task-focused categories (mainly asking questions and giving information) and 14 socioemotional ones (Table 1) [6]. For each consultation, or larger segment of a consultation, the main data consist of the number of utterances in each category. In its original form, RIAS does not facilitate sequential analysis. For instance, information is provided on the number of different types of questions from the physician and answers from the patient, but the system is not designed to assess what type of answer is typically given to a specific type of question.

Different approaches have been taken to the analysis of social interaction. In this paper we shall draw on linguistic interaction analysis, as exemplified in particular by conversation analysis [7]. One of the most important insights from conversation analysis—not limited to medical consultations—is that meaning is not given in advance, but created in co-operation throughout a conversation [7]. A conversation—and language use in general may be characterized as a joint action, i.e. an undertaking by a group of two or more persons who negotiate and co-ordinate their individual acts [7], [8]. Thus, in social interaction meaning is never produced in isolation. The speaker is never alone in authoring his contribution; rather he is always in dialogue with his interlocutor and the context [7]. The elementary contributions are “inter-acts”—they are of a social nature. While the speaker’s action is listener-oriented, the listener is speaker-oriented, in that he aims at understanding what the speaker intends to communicate. Grauman has described this principle as a reciprocal setting and taking of perspectives [9].

A basic pattern in interpersonal interaction is turn taking, a continuous shift of the “floor” between the participants in the dialogue [8]. An analysis of a conversation thus implies a double perspective, including an analysis of the individual acts as well as an analysis of the social exchange. A main challenge of an interaction analysis system is to capture the dialogic nature of interaction. It is obviously difficult to achieve this with a system such as RIAS, in which the units of observation are utterances spoken by the individuals who are engaged in the consultation.

Roter is well aware of this problem, and in the RIAS manual (1995) she advises the coder in a number of ways to consider interactional qualities of the dialogue in coding the verbal text of doctor–patient interchange. When in doubt about the coding, the coder is advised, as a rule of thumb, to take the next utterance, the response, into consideration when deciding the coding category. Roter thus implicitly takes the view that in interaction analysis each utterance must be understood in its sequential, textual context, the so-called co-text. This analytical guideline rests upon the theoretical consideration that the utterance is produced in co-operation with the other participant(s) in the dialogue. We fully agree with Roter on the general principle that the co-text provides important information when coding utterances in conversations.

However, coders may often find that strict adherence to the manual conflicts with a more functional approach based on the principles of conversation analysis. The purpose of the present paper is to elucidate and discuss some features of the RIAS, with an emphasis on how RIAS can be applied in describing and analyzing the dialogic nature of interaction.

This paper represents the results of a cross-disciplinary research group (two linguists, two psychologists, one psychiatrist, and one dentist) who have collaborated in applying RIAS to five Norwegian studies of provider–patient interaction. Transcripts and videos of consultations from one of the projects, an experimental study on physician communication styles in relationship to patient anxiety, [10], [11] were scored by all the members of the group and then thoroughly discussed. The aim of this collaboration was to study RIAS in order to illuminate its strength and weakness from different professional angles as a preparation to the actual application of RIAS in the five studies. Subsequent to the group collaboration, the consultations were coded by trained coders.

In recording and coding on the material, the group has used “state-of-the-art” software, allowing analysis of the consultation in terms of both the number of utterances in the dialogue and their duration. We have applied the “Observer” software for behaviour observation and video analysis [12]. In the Observer system all tapes were digitized and copied to CDs. This made it easy to review the recordings, in that specific sections or categories could easily be found and replayed during the scoring. The data were stored in sequence in two separate channels, one for the physician and one for the patient, thus providing a method for registering parallel speech, as opposed to the original Roter recording procedure, in which the frequencies (but not the duration) of utterances were scored in one channel only. As in most studies, RIAS was used to code directly from audio or videotape, rather than from a written transcript.

Section snippets

The analytical unit in spoken interaction

In this section we will focus on the segmentation and operationalisation of analytical units in spoken interaction.

In Roter’s system the communication units are defined as “utterances”: “the smallest discriminable speech segment to which a classification may be assigned. An example of a definition of an utterance given by Harris (1951, quoted from Schiffrin [13] 1987): “an utterance is any stretch of talk by one person, before and after which there is silence on the part of that person.” In

Pauses and filled pauses

In conversation analysis a pause is more than a demarcation point between two utterances [18]. The pause is in itself a dynamic element of the conversation, also when shorter than one second. In conversational interaction pauses of only 0.2 s are frequent, and can be enough to signal a possible change of floor, or give room for an agreement response.

Sometimes a single word or a fragment of a sentence may “fill” a pause in an otherwise uninterrupted utterance. In the RIAS, the TRANS-category

Interruptions

Interruption is a frequent conversational phenomenon, but it is not commented on in the RIAS manual. A seminal paper on interruptions showed that male physicians interrupt patients more often than they are interrupted, while encounters between female physicians and patients displayed a different, more equal pattern. [19] One should be very careful, however, not to conclude that interruption is merely an indicator of dominance, power and control. Closer analyses have shown that overlapping

Conflicting criteria

For several coding categories, the RIAS manual lists more than one criterion for coding. In some cases, these criteria may contradict one another. In this section we will discuss two examples of such conflicting criteria. We will argue that when such conflicts occur, one should choose the more functional rather than the more formal criterion for coding.

The empathic process

Empathy is usually seen as a complex and multi-dimensional phenomenon. A recent definition of the essence of curative empathy in psychotherapy is the following: “therapists steep themselves in the world of the other attempting to understand how others see and experience themselves and their world, putting this into words and checking their understanding”[21]. The empathic process thus often involves several responses from the listener leading towards an understanding of the experiences of the

Conclusion

In this article we have outlined some problems caused by the lack of strict definitions of the coding categories in RIAS. We have pointed to some of the problems we encounter in coding, and discussed them in the light of principles of linguistic interaction analysis, most specifically the need to observe the principle of turn taking and the fact that meaning is constructed during the interaction. We believe that with a greater emphasis on functional criteria in the coding process, the Roter

References (25)

  • L.M. Ong et al.

    Doctor–patient communication: a review of the literature. [Review] [112 refs]

    Soc. Sc. Med.

    (1995)
  • T.S. Inui et al.

    Outcome-based doctor–patient interaction analysis. Part 1. Comparison of techniques

    Med. Care

    (1982)
  • Stiles WB, Putnam SM. Coding categories for investigating medical interviews: a metaclassification. In Lipkin Jr M,...
  • D.L. Roter

    Patient participation in the patient-provider interaction: the effects of patient question asking on the quality of interaction, satisfaction and compliance

    Health Educ. Monogr.

    (1977)
  • D. Roter et al.

    Communication patterns of primary care physicians

    J. Amer. Med. Assoc.

    (1997)
  • Roter D. The roter method of interaction process analysis. 1995, unpublished...
  • Linell P. Approaching dialogue on monological and dialogical models of talk and interaction. Amsterdam, Philadelphia:...
  • Clark H., Using language. Cambridge: Cambridge University Press,...
  • Grauman CF, Commonality, mutuality, reciprocity: A conseptual introduction. In: Marková I, Foppa K, editors....
  • Graugaard PK, Eide H, Finset A. Interaction analysis of physician-patient communication: The influence of trait anxiety...
  • P.K. Graugaard et al.

    Trait anxiety and reactions to patient-centered and doctor-centered styles of communication: An experimental study

    Psychosom. Med.

    (2000)
  • L.P. Noldus et al.

    The observer video-pro: new software for the collection, management, and presentation of time-structured data from videotapes and digital media files

    Behav. Res. Meth. Instrum. Comp.

    (2000)
  • Cited by (69)

    • Correlates of observer-rated active involvement in psychiatric treatment visits

      2017, Psychiatry Research
      Citation Excerpt :

      To our knowledge, behavioral measures of patient active involvement in a treatment context have rarely been examined, and extant studies have used different methodologies. An early example of this type of method is Roter's Interaction Analysis System (RIAS; Roter, 1977), which focuses on questions asked by the patient; this method has been subject to criticism (e.g., see Sandvik et al., 2002). Other behavioral coding schemes have been developed more recently.

    View all citing articles on Scopus
    View full text