The burden of headache disorders in India: methodology and questionnaire validation for a community-based survey in Karnataka State

The Institutional Ethics Committee of National Institute of Mental Health and Neurosciences (NIMHANS), Bangalore, India approved the study protocol. Informed consent was obtained from participants before interviews commenced. In addition, to obtain wide acceptance of the study, a brief overview of the study and intended interviews, and their purpose and objectives, were presented to local community leaders, and their queries answered and doubts clarified.

The adult population between 18 and 65 years in both urban and rural Bangalore was the population identified for the study. Institutional households (paying-guest accommodation, hostels, etc.) and immigrants (defined as staying for <6 months in the household and locality) were excluded, as were bachelor households (those containing two or more bachelors belonging to different families staying together, whether or not sharing the preparation or partaking of food) on the basis that they were not permanent residents. Those not conversant with the local language (Kannada) were also excluded.

The study was a cross-sectional survey, using cluster sampling among urban and rural populations representative of the population of Karnataka State.

Applying an expected prevalence of headache of 35 %, based on review of previous studies [6], and with 95 % confidence and 10 % relative error margins, the minimum sample requirement was 1,000 biologically unrelated individuals from each (urban and rural) stratum.

The study was undertaken in both urban and rural areas of Bangalore. For the urban component, after excluding predominantly commercial areas within Bangalore city, specific geographical areas with relatively stable population consisting of a mix of upper, middle and lower classes were listed and one area (Kemepegowdanagara [http://​g.​co/​maps/​ungmu]) was randomly selected. For the rural component, one revenue subdivision (taluka) in Bangalore rural district was randomly chosen and, from a list of villages therein obtained from Census 2001 data, two villages (Uyamballi and Doddaaladahalli [http://​g.​co/​maps/​9gbnt]), located almost 110 km from Bangalore city and with 500 and 800 households respectively, were randomly selected.

An individual household (defined as a group of people living together and sharing a kitchen) was considered as the Primary Sampling Unit. Local residential maps were developed to identify area boundaries and household distribution, and all households then given unique numbers. Uninhabited or abandoned houses, institutional households and commercial establishments were excluded for survey purposes (see Fig. 1).

In each surveyed locality, the first household was identified using the random direction method. The survey team, on reaching the centre of the locality, numerically marked the major roads along all four directions. One road was selected corresponding to the last digit of a currency note. Households in this road were visited consecutively (door-to-door) by knocking at their doors unannounced (cold-calling). The first household on the right side of the road became the first in the survey. The second survey household was the one on the left side of the first, and the third was the household on the left side of the second. This process continued until the requisite sample was achieved.

All households approached were accounted for in the survey (Fig. 1). When the doors were locked, or no responsible person (either the head of the household, or any other adult member of the household who could provide reliable information for all members) was available, the household was visited again on the same evening or within the next 3 days (‘mop-up visits’). When necessary, the co-operation of neighbour households was sought to make contact with the required household members. Extra efforts were made to include such households during the last week of the survey in each area (‘final mop-up round’).

In each surveyed household, a responsible person was asked to list all members (defined as residing there for >6 months) in a specific order (oldest male first, followed by oldest female, and so on to the youngest male and youngest female). From this list, one individual aged between 18 and 65 years, male or female, was randomly selected as the participant in the study. Those who were not conversant in the local language (Kannada) were excluded and replaced by further random selection from the household list. The selected participant was interviewed immediately if present and consenting; when he or she was not present, an appointment was made for a return visit at a convenient time. In case of refusal, only age and socio-economic status were documented along with the key reason(s) for refusal. No replacements were made at the household level for selected respondents who withheld consent or failed to keep three appointments, two being on Sundays or holidays; these people were listed as non-participants (Fig. 1).

The survey instrument was a structured questionnaire developed for the study after a review of the literature; it has meanwhile been used, in appropriate translations, in community-based Global Campaign surveys of headache disorders in China [9] and Russia [10].

The first part contained socio-demographic questions (on age, gender, education, income). The Kuppuswamy scale for classification of socio-economic status of individuals [11] was utilized after updating income levels to the year 2008. The second part began with the screening questions for life-time (“Have you ever had a headache?”) and 1-year prevalence of headache (“Have you had headache during the last year?”), and these were asked of all participants; it concluded, for those answering “yes” to the second question, with an enquiry into frequency of headache. Participants reporting headache on ≥15 days/month were asked a subset of questions designed to diagnose MOH. The third part consisted of diagnostic questions for migraine and TTH based on the International Classification of Headache Disorders, 2nd edition (ICHD-2) [7]. Before these were applied, an introductory question asked whether all headaches experienced by the participant were of one or >1 type: those with >1 type were asked to keep in mind only the subjectively most bothersome type. The fourth part comprised questions on headache with a 1-day timeframe (“Did you have headache yesterday?”), and duration, intensity and burden of that headache. The fifth part enquired into health-care utilization for headache. The sixth was a single question on willingness to pay (WTP) for effective headache treatment (“I want you to imagine that there is a treatment you can buy. If you take it, your headaches will not bother you. I am going to ask you how much you would be willing to pay for this treatment”). The seventh part assessed disability burden using the Headache-Attributed Lost Time (HALT) index [12]. The eighth was the World Health Organization’s 8-item Quality-of-Life scale of (WHOQoL-8) [13], and was applied to all participants to permit comparison between those with and without headache.

All parts were translated into Kannada following Lifting The Burden’s translation protocol [14].

Diagnoses were not made by the interviewers; instead, responses to the diagnostic question set were fed into an algorithm developed for the purposes of the study by LJS and TJS. This algorithm was converted to a computer program in Epi-info version 3.5 [15].

In participants reporting more than one headache type during the last year, only the subjectively most bothersome was diagnosed. As for the headache yesterday, it was asked only whether this had been of the same type as the most bothersome headache. Cases were first identified of headache occurring on ≥15 days/month and, amongst them, those with medication overuse (MOH). The algorithm was applied to the remainder of the participants, applying ICHD-2 criteria [7] in hierarchical sequence first for migraine, then for TTH, then for probable migraine and finally for probable TTH. Cases remaining were considered unclassifiable. In the later analysis, migraine and probable migraine were grouped as all-migraine, and TTH and probable TTH as all-TTH.

Six field interviewers with a background in social sciences, and experienced in the methods and techniques of community-based surveys, were recruited for data collection. They were trained through multiple methods: lectures, demonstrations, supervised and independent skill-development sessions. The Manual for descriptive studies developed by NIMHANS [16] for epidemiological studies in developing countries was a key resource document. Training was conducted in the local language (Kannada). Interviewers were taken through the final version of the questionnaire so that they understood the nature and sequence of questions, and which were of particular importance (e.g., the screening and diagnostic questions). As hands-on instruction, they undertook mock and supervised interviews, and, finally, independent interviews on patients with and without headache in the outpatient department of NIMHANS. These steps aimed to standardize the method of data collection, reducing intra- and inter-observer variability.

The pre-pilot study was undertaken in the clinical setting. A neurologist and headache expert (GBK) administered the study questionnaire to 40 purposively selected individuals to demonstrate acceptability and inoffensiveness of the questionnaire.

The pilot study was population based, applying the full study methodology for selection of participants and data collection. It was conducted in both rural and urban areas, and interviewed 224 individuals aged 18–65 years, representative of the different socioeconomic strata. Its objectives were to test procedures for selection of households and of participants within households, field test the study instrument, estimate the participation rate and identify and solve logistic problems that might occur in the main study.

During the pilot study, the KISH method [17] of selecting participants from individual households gave a very low interview-completion rate (34 %) compared with simple random selection from the household member list (89 %). Hence, it was decided to use the latter in the main study. Experience from the pilot study also led to editing of the questions on headache characteristics and willingness to pay to suit the cultural context. HALT index and WHOQoL questions needed more explanations to the interviewee to clarify the concepts behind them and their foci of enquiry.

The trained interviewers, calling door-to-door in the manner described above, visited households in the identified urban and rural communities. They conducted interviews during morning hours and at mutually convenient times by appointment. In the rural area, the field staff took up temporary residence nearer to the study locality to ensure that early morning, late evening, Sunday and holiday appointments were not missed.

Weekly monitoring meetings were held at NIMHANS, Department of Epidemiology to review the progress of work. The investigators conducted monthly refresher training for the field staff. Surprise field visits were undertaken, and a 10 % sub-sample of participants were re-interviewed within 3 weeks of first interview. The discrepancies were very few and, as they pertained to the nature of costs incurred, they were not considered to be significant.

Diagnostic validation required that a sub-sample of participants were re-interviewed, soon after, by a headache expert (GBK) who was blind to the questionnaire diagnoses and applied his clinical skills as well as ICHD-2 criteria to make his own “gold-standard” diagnoses. The 15 % sub-sample was randomly selected. Reviews were undertaken within 3–6 weeks of the primary interviews.

In the beginning, letters were sent inviting selected participants from urban areas to NIMHANS and those from rural areas to its extension services. Because of very poor response, general health camps were held in both urban and rural areas, in the surveyed localities (with a maximum walking distance of 1 km). They were conducted on holidays or Sundays by arrangement with local leaders. The response among selected participants was only marginally better (fewer than one quarter attended, the majority of whom were badly affected by migraine). However, turnout was huge as people found it a good opportunity to seek care for general medical problems and/or other neurological problems. This posed logistic difficulties in planning the camps and, especially, to arrange supplies of free medication.

Finally, the door-knock approach was adopted, with the headache expert calling a second time at the houses of selected participants, and this provided acceptable coverage rates.

Completed survey forms, as received, were scrutinized for accuracy, completeness, inconsistencies, wrong or illegible markings and missed entries first by the team coordinator (Mr. Lokesh) and then by study team member (KR). When discrepancies were observed, study team members (primarily KR but also GNR and GG) discussed them with the field research officers during the weekly meetings. Minor discrepancies (wrong markings, illegible entries) were corrected after discussion but, if there were more major discrepancies, or important data were not collected, the team coordinator was asked to re-visit the household to ascertain what was correct. In fact there were no major discrepancies from the second week of the survey until completion.

Data were entered into a secure database using Epiinfo [15]. A random 10 % of the entered data was cross checked against the original forms. There were very few wrong entries, and therefore no concerns about the overall accuracy of data entry. The high level of accuracy was possible due to the ‘check-conditions’ and ‘jump-conditions’ included in developing the data entry form in Epiinfo.

Analyses were undertaken using Epiinfo [15] and SPSS [18]. In the validation study, sensitivities, specificities and positive (PPV) and negative predictive values (NPV) [19] and Cohen’s Kappa scores [20] were calculated using freely available on-line statistical calculators.