What really brings you here today? Applying McWhinney’s Taxonomy of Patient Behaviour between the first waves of COVID-19

Keywords: attitude to health, behaviour, classification, COVID‐19, diagnosis, family practice, patient-centred care, physician–patient relations, primary health care, sick role.

Doctor–patient encounters are surrounded by expectations from both sides, and the driving force behind each encounter is not always self-evident, as patients’ reasons for care-seeking do not always depend on their symptoms.^1,2 Sometimes patients exhibit incongruous consultation behaviour: quickly seeking care for low-impact issues or postponing seeking care for high-impact issues.¹

The family doctor and philosopher, Ian McWhinney, has argued that the behaviour of patients in doctor–patient encounters should be classified in parallel with our usual classifications of clinical conditions.³ He proposed that behavioural and social phenomena are related to clinical issues and a separate classification may permit greater understanding of their relationship. He then developed a taxonomy of patient behavior, taking the doctor–patient contact as the reference point, and a second taxonomy for the classification of interactions between individual patients and their environment.³

Understanding patients’ motivations for seeking consultations with their primary care doctor allows doctors to be better prepared for the encounter. During this cross-sectional study, we focus on patient behaviour.

Our primary objective was to apply McWhinney’s Taxonomy of Patient Behaviour (subsequently referred to as ‘the Taxonomy’) to our clinical records and to assess its relevance and adequacy by testing whether the taxonomic categories are mutually exclusive and if they allow patients’ behaviour to be classified in every appointment, as McWhinney proposed. We also aimed to explore associations of the Taxonomy with patients’ age and gender, type, modality and the weekday of the appointment. We reflect on the influence that the idiosyncrasies of the coronavirus disease 2019 (COVID-19) pandemic may have had on our results.

We designed a cross-sectional study with data retrospectively collected from electronic medical records. Our team is composed of three physicians (one family doctor and two family medicine residents) and we work in a Community Health Centre in Florianópolis, Brazil. The Ethical Research Committee from the State Secretariat of Health of Santa Catarina, Brazil, approved the research.

We reviewed the Taxonomy and coded every patient-initiated appointment as one of the following: (a) ‘limit of tolerance’, when the reason for the patient’s visit was because their symptoms were intolerable, such as pain or discomfort; (b) ‘limit of anxiety’, when the reason for the encounter was due to the patient’s concern with their possible unfavourable outcomes; (c) ‘heterothetic’, or problems of living presenting as symptoms, when the impetus for the appointment was related to a patient’s poor adaptation to their environment (e.g. symptoms with aetiology such as lumbar pain, dyspareunia, decompensation of chronic diseases); (d) ‘administrative’, when the doctor–patient encounter happened for administrative purposes, even if there was a symptom involved; and (e) ‘no illness’, for preventive care appointments or medical assessments without any presenting symptom.³

For coding, we filled an unused field in the electronic medical record, the ‘secondary ICD’, using the International Classification of Diseases (ICD) codes: ICD-10 R68.8 for ‘limit of tolerance’, Z71.1 for ‘limit of anxiety’, F43.2 for ‘heterothetic’, Z02.9 for ‘administrative’ and Z13.9 for ‘no illness’. We constantly checked agreement between the authors to reduce measurement bias. We were unable to perform double coding as we had limited time, financial and personal resources.

In December 2020, we obtained a spreadsheet containing every medical appointment made by the three doctors during the previous 4 weeks, with the variables: patient identity number (ID), day and time of the appointment, day and time at the time of scheduling (only present for previously scheduled appointments), modality of the appointment (either in person or teleconsultation), secondary ICD (filled with the Taxonomy coding) and the name of the attending doctor. The data were imported to RStudio 1.2.1335 with R 3.6.3.^4,5

The weekday of the patient visit was drawn from the ‘day and hour’ field. No variables were transformed. The variables used in the analysis were: patient ID (categorical); patient age (continuous); gender (categorical, ‘male’ and ‘female’); type of appointment (categorical, ‘scheduled’ and ‘same-day’); modality of appointment (categorical, ‘in-person’ or ‘teleconsultation’); Taxonomy classification (categorical, ‘limit of tolerance’, ‘limit of anxiety’, ‘heterothetic’, ‘administrative’ and ‘no illness’); and day of the week (categorical, with the levels: ‘Sunday’, ‘Monday’, ‘Tuesday’, ‘Wednesday’, ‘Thursday’ and ‘Friday’; as no appointments were made on Saturday).

We used Pearson’s Chi-squared to test relationships between the Taxonomy classification and patients’ gender, modality of appointment (in-person or remote), type of appointment (previously booked or same day), and day of the week. One-way ANOVA was applied to the variables age and Taxonomy classification. The results of the analysis are reported using the ggtstatplot R package.⁶ All tests were two-tailed with an alpha level of 0.05. There were no subgroup analyses.

We did one post hoc analysis, as there was only one appointment on Sunday that could compromise the Chi-squared test. We proceeded with Pearson’s Chi-squared test with simulated P-value (using Fisher’s exact would not be possible). As this delivered similar results, we report here the original analysis.

We had 818 patient file openings, of which 171 were excluded from the study due being either patient file annotations or doctor-initiated appointments, giving 647 patient-initiated medical appointments by 461 unique patients (65.2% were female). The average patient age was 32.7 years (range 0–89 years). Most (541; 83.6%) were same-day appointments. There were more in-person appointments (364; 56.3%) than remote consultations. The type of Taxonomic classifications was, in descending order: 180 (27.8%) ‘administrative’; 172 (26.6%) ‘limit of tolerance’; 141 (21.8%) ‘limit of anxiety’; 117 (18.1%) ‘no illness’; and 37 (5.7%) ‘heterothetic’.

Fig. 1 shows Taxonomy types by gender, appointment modality, and for same-day appointments. Female patients had more consultations classified as ‘heterothetic’ (29; 78.4%) and ‘limit of anxiety’ (43; 63.2%. P < 0.001). Patient behaviour also differed according to the modality of the appointments, with ‘limit of tolerance’ and ‘heterothetic’ observed more frequently in face-to-face appointments (P < 0.001). Most of the ‘limit of anxiety’ consultations were same-day appointments (P < 0.001).

Fig. 1.  Percent of consultations by Taxonomy category and appointment type, consultation modality, and gender (N = 647).

The average patient age was higher on appointments classified as ‘heterothetic’ (37.3 years) and lower on ‘limit of anxiety’ appointments (30.2 years) (one-way ANOVA, P = 0.02).

Fig. 2 shows Taxonomy classes by day of the week. Appointment modality and day of the week were unrelated (P = 0.538).

Fig. 2.  Percent of consultations by Taxonomy category and day of the week (N = 647).

This study was conducted between the first and second COVID-19 waves in Florianópolis, so the results must be considered with this in mind. We suppose that the greater frequency of ‘limit of tolerance’ and ‘administrative’ appointments reflect, at least partially, the characteristics of our population, whose socioeconomic status ranges from marginally above the poverty line to the lower-middle class. This group mainly utilises the public healthcare system for consultations, diagnostic tests, and to receive medications, which have no out-of-pocket cost at the point-of-care in the Brazilian health system (Sistema Único de Saúde (SUS)). Individuals of higher socioeconomic status often meet their healthcare needs in other ways such as by buying their medicines, and consulting with other healthcare providers, paid for by out-of-pocket payments.

The higher rates of ‘administrative’ appointments may also be a consequence of the COVID-19 pandemic, as workers with cold symptoms need a note to stay at home in order to prevent transmission of the virus. Additionally, they may also need a ‘release note’ to return to work when they finish their isolation period after being positive for SARS-CoV-2.

Even though teleconsultations may be common in primary care in some other countries, in Brazil, this began as an extraordinary measure during the current COVID-19 pandemic. This allowed us to analyse the difference in patients’ behaviours between in-person and teleconsultation appointments. We hypothesised that the opportunity to contact a doctor without leaving their house might have motivated more low-impact appointments, lessening the threshold of tolerance: of both symptoms and anxiety. We observed that ‘heterothetic’ appointments tended to occur much less on teleconsultations. The limitations of non-verbal communication might have heightened the already difficult identification of this motive for seeking care. However, the preferences of our team may also play a role, as we prefer to conduct mental health consultations face-to-face.

The rarity of ‘heterothetic’ appointments draws attention to the challenges of making this classification by considering individual appointments because the psychosocial issues that intertwine the symptoms require longitudinal care before they are revealed. However, we observed that as we practiced using this Taxonomy, we became more conscious of patients sometimes seeking medical consultations for deeper reasons than immediately apparent, giving us the impression of a more complete patient evaluation.

This also supports the need to have different ways to classify illness and behaviour if we are to better understand both illness and patient behaviour.³ Combining the clinical diagnosis and the patient’s behaviour in the same classification can confuse the truth by stimulating a vigorous search for an accurate diagnosis of a patient’s imprecise complaints. In primary care, this can be harmful to both patient and healthcare system.^7,8 Having a different classification that includes the problems of life presenting as symptoms as one of the possible motivations for the attendance is another tool to help family doctors to deal with the ‘uncertainty sea’⁸ in which they sail.^8,9

Although there is great interest in the reasons leading patients to seek a medical consultation,^1,10–13 few prior studies have assessed patients’ behaviour when choosing to seek a consultation at a given time.^3,13–20

In a study of patient behaviour published in 1975, but conducted in 1971, Stewart, McWhinney, and Buck describe the classification of 389 visits by five trained physicians using a ‘Taxonomy of Patient Behaviour’.¹⁵ They assessed the reliability of a taxonomy very similar to the one McWhinney published a year later,³ by observing consultations from cooperating physicians and their patients and comparing the assessment of the investigator and the physician. Analysing these two publications, we suppose that the taxonomy published in 1972 consolidates what was empirically tested in 1971 by Stewart et al.¹⁵ This taxonomy had seven categories, was applied in a sample of women patients, both in patient-initiated and doctor-initiated visits, and was considered reliable. The overall frequencies were grouped by the authors as: 41% for all symptoms (‘limit of tolerance’ or ‘limit of anxiety’), 21.9% for psychosocial problems that were openly presented (‘limit of tolerance’ or ‘limit of anxiety’), 14.3% for signal behaviour (defined by the authors as ‘presenting illness or symptom is used as a ‘ticket of admission’ to the doctor so that some underlying problem can be presented’), and 22.8% for no illness.¹⁵ Our findings are similar.

Between 1986 and 1988, Susan McNair, a family medicine resident at that time, also classified with this Taxonomy her patients’ behaviours when consulting, and observed quite different results from ours. The majority of McNair’s appointments were classified as ‘limit of tolerance’ (67.4%), followed by ‘no illness’ (16.9%), ‘heterothetic’ (8.2%), ‘limit of anxiety’ (7.2%), and ‘administrative’ (0.3%).²¹

The Finnish general practitioner, Simo Kokko, inductively developed a typology for long-term consultation patterns, aiming to integrate morbidity and disease factors in the typology and to analyse consultation rates.¹³ In analysis of the 2251 attendances during 9 years of a sample of 100 patients, Kokko obtained six categories that, in decreasing frequency, were ‘healthy and competent’ (patients met the physician mostly for minor ailments); ‘contented returners’ (patients with conditions requiring repeated follow up); ‘information seekers’ (individuals who ‘often seemed to be looking for medical explanation’); ‘support seekers’; ‘drifters’ (people who had not ‘found their place in life’); and ‘hard to convince’.

The classifications suggested by McWhinney and Kokko seem complementary rather than divergent. Although McWhinney’s Taxonomy allows us to investigate what motivates consultation both in a singular and episodic form of care, Kokko’s permits classification of long-term patterns of seeking care for each patient. We value McWhinney’s Taxonomy for being used concurrently with the clinical encounter and not just retrospectively, as is the case for Kokko’s typology.

McKinley and Middleton qualitatively analysed written agendas completed by 756 patients using the taxonomy of Stewart et al.¹⁵ as a framework, and found that 42% of those patients consulted for having reached their anxiety limit, and 24% for their limit of tolerance. These proportions are much higher than previously found by Stewart et al.,¹⁵ McNair²¹ and us. Although the authors report some study limitations, the difference may be mostly due to the patients’ self-report of their agenda.

A more robust method to identify the demands of patients when seeking a primary care doctor was carried out by Good et al.²⁰ After 100 in-depth patient interviews, in addition to observation of clinical interviews, they developed the Primary Care Patient Request Scale. The various requirement domains were grouped into five factors: treatment of psychosocial problems; medical explanation; supportive communication; test results; and ventilation and legitimation.²⁰ These five factors, as well as the domains within them, fall short of identifying ‘why’ this patient came for a consultation at this time.

The lack of double coding to reduce bias, as well as the sampling limited to consecutive meetings in just 1 month of care, are the main limitations to the current research. Although we analysed the data of just 4 weeks of a single primary care team and its patient population, this appears to be the third study to apply McWhinney’s Taxonomy by doctors in clinical encounters, and the first outside Ontario, Canada. We also did not find other research reporting patients’ motives for seeking care amidst the COVID-19 pandemic.

We initially decided to apply the Taxonomy only in the patient-initiated consultations, but after becoming familiar with the Taxonomy during the course of the study, we concluded that this was an unnecessary self-imposed limitation.

We made classifications of patient behaviour from individual encounters and not taking the episode of care²² as a reference. We consider that classifying the episode of care would also be of great value, but in addition to limitations of the electronic medical record, the methodological complexity is much greater. It would be interesting for future research not only to compare classifications of individual encounters with episodes of care, but also to evaluate the integration of the Taxonomy with the Reasons for the Encounter of the International Classification of Primary Care and use the Taxonomy classification to assess the pre-test probability of a given diagnosis.

This work expands the perspectives of using this Taxonomy to teach clinical interview skills. The Calgary–Cambridge,²³ a clinical interview guide used internationally, teaches how to clarify the nature of patient encounters and address each patient’s perspective of illness by exploring their ideas and beliefs, concerns, expectations, effects on life and feelings – but does not go any further. This Taxonomy can be complementary to the Calgary–Cambridge interview guide.

Several studies, with different methods, synthesise and classify the reasons that lead a person to seek medical help at a given time; however, in our experience, none is as practical and straightforward as McWhinney’s Taxonomy of Patient Behaviour. It is relevant and sufficient, even almost 50 years after its first publication. With our data, we hope to encourage other family doctors to adopt this framework in their consultations and teaching functions, and, perhaps, inspire more complex investigations.

The data that support this study will be shared upon reasonable request to the corresponding author.

The authors have no conflicts of interest to declare.

This research did not receive any specific funding.