Patient advisory group
Prior to commencement of this study, the research team began working with a 20-member Patient Advisory Group. The aim of this was to ensure the voices of patients were included and their insights reflected in all stages of this study. The feedback provided was that many people living with ME/CFS do not attend general practitioners (GPs; i.e. primary care physicians) for ME/CFS-related care due to a lack of treatment(s) and stigma/disbelief associated with the condition. To address the potential under-estimation of prevalence due to such factors, we conducted focus groups and long interviews with ME/CFS patients to further understand patterns of GP attendance. As such, this study will present crude prevalence estimates generated from a nationally representative primary care dataset along with qualitative data to provide greater context for these estimates.
Quantitative analysis
Setting and study population
The datasets for this study were extracted from the MedicineInsight database, which is managed by NPS MedicineWise, an independent, not-for-profit organization [15, 16]. The database contains de-identified patient data from primary care/general practices across Australia which has been described elsewhere [16]. The patients included in this database are similar in terms of age, sex and socio-economic status to all Australian patients who have received at least one general practitioner (GP) Medicare Benefits Schedule (MBS) subsidized consultation (i.e., the universal health insurance program) [17].
A unique patient identifying number allows patients within practices to be tracked over time to produce longitudinal data. As of 31/12/2019, 671 general practices were participating, with > 2.2 million active patients.
The study period was defined as 1st January 2014 to 31st December 2019, and we defined patients using observable person-time [18]. The study population was defined as those meeting the following criteria:
-
Visited a practice site and met specific MedicineInsight data quality requirements.
-
At least one clinical encounter with the practice, including face-to-face or phone call during the study period. A lookback period was adopted for clinical encounters, see section below for details.
-
Have valid information for age (aged ≥13 years in each year of interest).
-
Observable person-time during the study period commenced from the patient’s first recorded clinical encounter at the practice and ended at their last recorded clinical encounter with a one-year ‘lookback’ period to improve case ascertainment [18].
Case definition
In addition to the criteria above, patients were included as cases if they had:
-
1. A diagnosis of “myalgic encephalomyelitis”, “chronic fatigue syndrome”, “ME”, “CFS”, or “ME/CFS”. Algorithms were used to identify coded and free-text information with the key words in the fields of either ‘encounter_reason’, or ‘diagnosis_reason’
-
2. Encounter/diagnosis of ME/CFS from the first lookback year (01/01/2014) up to 31/12/2019.
Records of ‘chronic fatigue’ were excluded as this symptom alone does not provide evidence of an ME/CFS diagnosis.
Lookback period and case definition assumptions
Based on discussions with the Patient Advisory Group, a 12-month lookback period was adopted to improve case ascertainment. This was based on the understanding that some ME/CFS patients do not attend a GP clinic for ME/CFS-related problems on an annual basis. Therefore, the prevalence estimates for 2015 includes all ME/CFS encounters that occurred from 01/01/2014 to 31/12/2015. Based on the focus groups discussions and feedback from the Patient Advisory Group, assumptions were also made that in the situation that a patient had ME/CFS encounters several years apart, they would be considered a prevalent case for all years in between the encounters. For example (see Fig. 1), if a patient had an ME/CFS encounter in 2014 and then again in 2018, they will be considered a case in:
-
2015 (due to the lookback period, i.e. 2014);
-
2016 and 2017.
-
2018 (as they had an encounter in 2018)
-
2019 (due to the lookback period, i.e., 2018).
In this example, for both 2016 and 2017, we assumed that the patient remains an ME/CFS case, reflecting the chronic nature of the condition and the very unclear and limited possibility of patient recovery from ME/CFS within the given period under review.
Statistical analysis
Data were analysed in the Secure Unified Research Environment (SURE), a highly secure virtual environment, using STATA software (version 16, Stata Corp. College Station, TX, USA). Descriptive statistics were calculated for cases for each period of interest from 2015 through to 2019. For each year of interest, descriptive characteristics of the cohort are presented as means (standard deviations (SD)) for normally distributed continuous variables, along with medians (interquartile ranges (IQR)), and frequency and percentages as appropriate.
Annual crude prevalence rates per 100,000 persons (95% CI) were calculated for each year of interest, i.e. 2015–2019, along with rates by sex, 10-year age groups, Index of Relative Socio-economic Advantage and Disadvantage (IRSAD) quintiles, and remoteness areas as measured by the Australian Bureau of Statistics’ Australian Statistical Geography Standards [19]. As MedicineInsight data is representative of the Australian primary care patient cohort, age standardization was deemed unnecessary [15].
Qualitative analysis
Validated guidelines, information power and ethics
We adopted qualitative research methods to supplement and augment the quantitative analysis by providing deeper contextualization and nuance to the quantitative findings [20]. ‘Mixed methods’ research capitalizes on the strengths of quantitative and qualitative data within a single study by integrating the two data types [21, 22]. We used the Standards for Reporting Qualitative Research that is a list of 21 items considered essential for complete transparent reporting of qualitative research and the Consolidated Criteria for Reporting Qualitative Research for focus groups and interviews [23].
We adopted the theory of ‘information power’ when constructing the size and representation of the focus groups and the number of interviews [24]. Information power is underpinned by a negative association between the depth of information provided by the participant(s) and the number of participants required. That is, a high level of information by a small number of participants provides strong qualitative evidence, particularly where data saturation is achieved, as in our study. More specifically, information power of an interview sample is determined by items such as: study aim; sample specificity; use of established theory; quality of dialogue; and analysis strategy [24]. Quality of dialogue was important for our sample as informed by our discussions with the Patient Advisory Group. The group indicated strong dialogue was expected during the investigation of the relatively narrow study aim for this current study of establishing prevalence in Australian primary care patients. To generate discussion amongst age demographics with similar lived experience, focus groups were stratified into the age groups of older, middle-aged, and younger people with ME/CFS. Carers were allocated to a separate focus group to also facilitate strong dialogue with shared experience.
Recruitment
Participants were recruited from an advertisement on Emerge Australia’s website, radio, and newspaper advertising, and ME/CFS Facebook platforms to ensure a wider reach. Inclusion criteria were a self-reported diagnosis of ME/CFS and aged 18 years or older. People who indicated their interest in the study were provided with a detailed information sheet and consent form. A pre-screening interview was then conducted by telephone (5–10 min) to answer questions, assess eligibility for the study, collect informed written consent and socio-demographic information for the recruitment database that would be used for sampling of focus groups and individual interviews (email address; age; sex; postcode; time of onset of first symptoms; time of diagnosis).
Data gathering and analysis
As part of the broader qualitative study regarding ME/CFS in Australia, we used semi-structured focus groups triangulated with both long interviews and the quantitative analysis to also investigate timing, barriers, and enablers to diagnosis of ME/CFS in the Australian primary healthcare setting. Focus groups were conducted virtually using computer assisted meeting technology to simulate a focus group environment and were audio recorded and transcribed verbatim with identifying material (names) removed. Semi -structured questionnaires were used to guide the discussion with each focus group and the same interviewers and observers participated in the focus groups and interviews. Focus group data were triangulated with individual interviews to explore emergent themes [25]. Given that debilitating fatigue is a key symptom of the chronic condition, two focus group interviews (of the same focus group) of no longer than 45 min each were conducted with the opportunity to also provide free text responses between the focus groups (2 days) to the questions that guided the interviews.
Thematic analysis was conducted inductively [26, 27] with the assistance of NVivo software. Emergent themes were discussed between co-authors and the Patient Advisory Group. Reflective notes (by the interviewers and observers) and meeting notes (for example with the Patient Advisory Group) were kept.
Ethics approval
Ethics approval for the quantitative study was granted by the University of Tasmania’s Health and Medical Research Ethics Committee (H0018473) and the qualitative component of the through the University of Tasmania (H0018683). Approval to conduct this study was granted by the MedicineInsight independent Data Governance Committee (reference number: 2019–026).