The study design was a cross-sectional enquiry by a questionnaire mailed to addresses selected randomly to be representative of Australia’s population aged 18 years or older. The methodology was informed by the principles and recommendations [10] set out by the Global Campaign against Headache [11].

The overall purpose was to inform the design, logistic planning, and effective conduct of a future nationwide headache epidemiological study. The aims of this pilot study were to: (1) establish the participation rate, preferred response method (paper-based vs. online), and acceptability of mailed self-report study questionnaires; and (2) provide estimates of prevalence, burden, and treatment gaps to inform power calculations in a future study.

We conducted the study in New South Wales and Victoria, the two most populous states in Australia.

We used a two-stage cluster sampling approach to implement equal probability sampling for households. We first sampled 100 local government areas (LGAs) from the 207 LGAs in Victoria and New South Wales using the probability proportional to size with probability minimum replacement sampling method, where the probability of selecting an LGA was proportionate to the total number of private dwellings in the LGA. Depending on the number of private dwellings in the LGA, it was possible that large LGAs would be sampled more than once and each sample was considered an independent cluster. The maximum number of times an LGA could be sampled was restricted by the probability minimum replacement method. In the second stage, 200 individual households in each sampled LGA were selected using simple random sampling method.

HopeWiser, a company providing an Australia Post-accredited address matching approval system, provided the addresses of 20,000 randomly selected households from the study LGAs across Victoria and New South Wales. HopeWiser extracted all valid addresses (i.e., those that had been identified on more than one source for the GeoCoded National Address File dataset), then used software to match these addresses against Australia Post’s Postal Address File (a highly sampled, mature dataset, having been commercially available for > 23 years with monthly updates). The resultant matches were enhanced by markers for ‘residential, ‘non-residential’, ‘deliverable’, and ‘non-deliverable’ provided by Australia Post.

This process captured the vast majority of deliverable residential addresses in Australia with high reliability in terms of completeness and accuracy (< 2% of households use postal services such as roadside mailbags or post office boxes rather than a letter box at their residential address).

We used validated instruments where possible for each section of our study questionnaire (Appendix 1).

The core of the present questionnaire utilised modules from the Headache-Attributed Restriction, Disability, Social Handicap and Impaired Participation (HARDSHIP) questionnaire [11], which has already been used in > 20 countries to measure headache prevalence and attributed burden in non-clinical settings [12]. Although only initially validated for face-to-face administration by trained lay interviewers, HARDSHIP has also been adapted as a self-report instrument, in the EUROLIGHT questionnaire [13]. Enquiry into basic demographic data (age, gender, postal/zip code, preferred language, Aboriginal and/or Torres Strait Islander status) was followed by headache screening questions (ever, and in the preceding year) and diagnostic questions based on ICHD-3. We used the Headache-Attributed Lost Time questionnaire (HALT) for capturing headache-attributed lost productivity [14], and the generic health-related quality of life (HR-QOL) EQ-5D-5 L instrument. Further questions addressed healthcare utilisation (headache-related outpatient visits, tests, and emergency department and hospital attendances within the previous year), medications (type and frequency of symptomatic headache medication used in the preceding 31 days, and currently used preventative medications), out of pocket costs (headache-related healthcare expenses within the previous three months, not covered by health insurance), barriers to accessing care (questions on self-recognition of migraine, diagnosis ever of migraine from a healthcare provider, difficulties in accessing a healthcare provider for headache, and any previous cessations of symptomatic, and/or, preventative migraine therapy, with the reasons why, all based on previous studies [15, 16]), and informal care needs (questions from European HIROZON-funded studies by co-author ZA on unpaid care from family or friends, and, if so, how many hours per week [not yet published]).

We estimated the sample size needed to establish migraine prevalence as N = 1,750, basing the calculation on the estimated prevalence of migraine for Australia of 0.18 from the Institute of Health Metrics and Evaluation [17], with a relative 10% margin for error. We estimated the number of mail outs needed as N = 19,445, assuming that 90% of households would have at least one eligible adult (see below), but a response rate of only 10%. We inflated this to N = 20,000 in anticipation that some study letters would inadvertently be sent to non-deliverable addresses.

We outsourced the mail out (including printing invitation letters and questionnaires, addressing envelopes, inserting reply-paid envelopes, and oversight of the process) to Direct Mail Solutions, a well-established company.

Potentially eligible participants were adults aged 18 years and over. From these, only the person who had most recently had their birthday was asked to respond.

Respondents needed to opt in as participants within the study timeframe, either by entering data directly into the secure online Research Electronic Data Capture (REDCap) platform, accessed via a QR code or weblink included in the study invitation letter, or by returning their hard-copy questionnaires by reply-paid post, with researchers entering the responses into the platform.

Please see Fig. 1 for the study workflow.

Study data were collected and managed using the REDCap electronic data capture tool hosted and managed by Helix (Monash University) [18, 19]. REDCap is a secure, web-based software platform designed to support data capture for research studies, providing: (1) an intuitive interface for data capture; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for data integration and interoperability with external sources.

Participants reporting any headache within the preceding year were considered to have an active headache disorder, with all others considered to be headache-free. Diagnoses were made during analysis, and not at the time of data collection, using the HARDSHIP algorithm [11], applied to the most bothersome headache type when more than one type was reported. The algorithm first identified those reporting headache on ≥ 15 days/month, diagnosing pMOH when acute medication use on ≥ 10 days/month was also reported and otherwise “other headache on ≥ 15 days/month” (other H15+). In all others with active headache disorder (episodic headache), the algorithm diagnosed, in hierarchical order, definite migraine, definite TTH, probable migraine, probable TTH, and unclassified. Used in this way, the algorithm identifies migraine (definite or probable) with a sensitivity > 70% and a specificity > 70% [12].

Analysis included all participants who answered at least one question.

Participation rate and preferred response method were assessed with reference to age, gender and state. Categorical variables were summarised using frequency and percentage. Continuous variables with approximately normal distribution were summarised using mean and standard deviation (SD), or, otherwise, median and interquartile range (IQR). Prevalence estimates of each headache type were adjusted according to the age and gender distributions of each state. A bootstrapping method with 1,000 iterations was used to estimate variances and calculate 95% confidence intervals (CIs). In the analysis, we combined definite and probable cases of migraine, as well as definite and probable cases of TTH, respectively.

Missing data were summarised using frequency and percentage. The missingness of data was assessed using Little’s chi-squared test for missing completely at random test or covariate-dependent missingness. Since missingness of data was found to be dependent on age, gender, and socioeconomic status (represented in Index of Relative Socio-economic Advantage and Disadvantage [IRSAD] quintile), these covariates were adjusted in the analyses, where applicable.

Statistical significance was set at p < 0.05. Holm-Bonferroni’s method was used to control for 5% family-wise error rate in subgroup pairwise comparison, if applicable. All statistical analyses were performed using Stata version 16.1 (StataCorp).

The study was granted multisite ethics approval by the Alfred Hospital Ethics Committee (HREC reference: 87,013, Local number: 305/22). Governance approval was granted by the Offices for Research at the individual study sites.

Participation was voluntary, requiring respondents to opt in, with consent therefore presumed.

Only non-identifiable data were captured and therefore participants could not be identified from their responses.