Phase 1: Research question and data collection | |||
Step | Example from Stortenbeker et al. (2022) | ||
Research question | “To what extent do linguistic markers in utterances differ between general practice patients presenting MUS and MES?” | ||
Data collection | Verbatim transcripts of general practice consultations were derived from an existing research project [36]. | ||
Phase 2: Codebook development | |||
Step | Issue | Action | Example from Stortenbeker et al. (2022) |
Selection criteria | Inclusion and exclusion | Define research scope | Language use of patients presenting medically explained or unexplained symptoms to GPs. |
Read through training consultations | Patients talk about their past (‘but it was always low’) or current health problems (‘I am unstable’) as well as about potential future health issues (‘I think it could go wrong’). | ||
Redefine selection criteria | Scope was limited to include only utterances relating to current or past condition of patients, not prospective conditions. | ||
Unit of analysis | Turn constructional unit | Define unit of analysis | Grammatical finite clauses served as unit of analysis in earlier stages. |
Read through training consultations | A more flexible unit of analysis was needed for subjectivity markers in cases such as ‘[I notice though] [that I’m getting sensitive to it]’. | ||
Redefine unit of analysis | Turn constructional unit was selected as the new unit of analysis. | ||
Deductive categorization | Retain predefined category | Scan literature for relevant linguistic elements | Patients with MUS use more negations when describing (non-) occurrences of symptoms than patients with MES [37, 38]. |
Formulate code | Negation – a) absent; b) syntactic; c) morphological | ||
Read through training consultations | Plenty of examples were found, such as ‘I am unstable’ and ‘I cannot move comfortably’, so negation was retained in the revised codebook. | ||
Deductive categorization | Exclude predefined category | Scan literature for relevant linguistic elements | Doctors use more ‘illness terms’ (e.g. urination problems) towards MUS patients, whereas MES patients are often described with ‘disease terms’ (e.g. bladder infection) [39]. |
Formulate code | Terminology – a) illness; b) disease | ||
Read through training consultations | Differentiating between the two was not easy (e.g. ‘I got dizzy’, ‘well then you’re all worn out’) and remained subjective. As an objective definition of the boundaries was not possible, the category was removed from the codebook. | ||
Inductive categorization | Include category based on observations | Read through training consultations | Salient utterances such as ‘that ear keeps on whizzing’ were marked, suggesting ‘that ear’ operating as a separate agent as opposed to ‘I can hear pretty badly’. |
Scan literature for relevant studies | Patients can be disconnected from emotional and/or somatic experiences in various degrees [40]. | ||
Formulate new code | Grammatical subject – a) first person (the patient, ‘I’); b) third person (patient’s biomedical or psychosocial state, ‘that ear’). | ||
Iterative refinement | Add subcategory after test coding | Define code | Grammatical subject – a) first person; b) third person. |
Read through training consultations | Some utterances could not be indicated as having a first- or third-person subject, such as ‘[positive though] [that I do not have any new lesions]’ in which no subject is present in the first TCU. | ||
Redefine code | “empty subject” was included as a subcategory in the revised version of the codebook. | ||
Phase 3: (double) coding | |||
Step | Issue | Action | Example from Stortenbeker et al. (2022) |
Double-coding | Refine coding categories | Double code session | Intensity displayed a Kappa of .66. |
Explore systematic differences | One coder did not interpret certain time words as intensifiers, whereas the other coder did, e.g. ‘sometimes’, ‘all of a sudden’. | ||
Fine-tune codebook and coders | Remarks were added to the codebook. Words denoting an in- or decrease in time/frequency words are only marked when intensified such that ‘after that it was wrong again’ is not intensified, ‘all the time I think oh I’m getting tired’ is intensified. | ||
Coding | N/A | N/A | Final coding was performed by the main researcher in various separate coding sessions. Cases of doubt were marked and evaluated at a later point in time. |
Phase 4: Analysis and reporting | |||
Steps | Example from Stortenbeker et al. (2022) | ||
Analysis | Logistic binary random intercepts models with various linguistic markers as outcome variables, and consultation type (unexplained or explained symptoms) and codes related to message content as predictor variables, controlled for various relevant confounders. | ||
Reporting | Distinguished between hypothesis-based and explorative analyses. For more information, see Stortenbeker et al. (2022). |