The target population for Malaysian GATS survey were non-institutional population of men and women aged 15 years or older, who considered Malaysia to be their usual place of residence⁴. The sampling did not include those who were visitors (e.g. tourists), institutionalized in hospitals, or residing in an assisted living facility/nursing home, on a military base, or in group quarters or a prison.

In Malaysia, the sampling for GATS was a multistage stratified cluster sampling that was representative of Malaysia’s population (15 years or above) all over the country except very remote areas in Sabah and Sarawak, where the only access to the areas is by boat or air transport. The sampling strategy of the survey was to generate precise cross-sectional estimates at the national level, by gender and by geographical (urban/rural) localities at the national level, and to allow for comparison of the estimates between different countries conducting the survey.

According to the GATS sampling protocol, a sample size of at least 4,000 respondents is required (2,000 males and 2,000 females, and 2,000 from both urban and rural areas)⁴. The GATS sample size was adjusted for potential ineligibility and non-response in determining the number of households to be selected in order to get the required number of respondents. After the adjustment, the final sample size was 5,112 households. Of this number, 2,676 were from urban areas; while 2,436 were from rural areas (the response rate in urban areas was usually lower than rural areas in Malaysia). There was no gender assignment made for the selected households.

The sampling process for GATS was done with the following sample design and technique:

GATS Malaysia included a household questionnaire and an individual questionnaire, both based on the GATS Core Questionnaire with Optional Questions, which was designed for use in countries implementing GATS⁵.

The household questionnaire, which solicited information on household members was used to randomly select an eligible household member (aged 15 years or above) to complete the individual questionnaire. The individual questionnaire, which was administered to the randomly selected adult through handheld machines, had 10 sections: Background Characteristics; Tobacco Smoking; Shisha/Hookah; Smokeless Tobacco; Electronic Cigarettes; Cessation; Second-hand Smoke; Economics; Media; Knowledge, Attitudes and Perceptions.

The adapted questionnaire was translated and back-translated to ensure the quality of the translation. The questionnaire was then pre-tested in the field in July 2011. Based on the pre-test, the questionnaire was modified and finalized for full study implementation. The final version of the questionnaire was approved by the GATS Questionnaire Review Committee in August 2011.

GATS was the first national community survey conducted by the Institute for Public Health via electronic data collection. General Survey System (GSS) software, developed by Research Triangle Institute (RTI) International for the Global Tobacco Surveillance System was used⁶. The software system is designed to support the collection of data in the field where interviewers collect data using handheld computers. The programming of the questionnaire using GSS was carried out in collaboration with information technology personnel associated with GATS Malaysia. The systems were developed and tested using Hewlett-Packard iPAQ 210 handheld devices.

Repeated quality control mechanisms were employed to test the quality of questionnaire programming, in accordance with the GATS Programmer’s Guide to General Survey System manual⁷. The main steps involved in checking quality control were version control/verification for the household and individual questionnaires, date and time verification, verification of skip patterns, and validation checks. The entire process, including administration of the questionnaires, data collection using handheld machines, and data management and aggregation (preparing raw data for analysis), was pretested before the actual survey process began. Handheld programming was finalized in July 2011. The electronic case file (used to identify the selected household addresses) was finalized in October 2011 and uploaded to the handheld devices during the training program for field staff.

GATS Malaysia carried out a pre-test in both urban and rural settings on 19 to 20 July 2011 using a sample of 120 respondents who were equally distributed by gender and smoking status and with individuals from all relevant age groups. The pretest was conducted with close cooperation from CDC and WHO experts, especially in terms of wording and comprehensibility, inconsistencies in skip patterns, the sequencing of questions, completeness of response categories, workload, interview time, availability and call-backs, and other issues. Other important objectives of the pretest were to test procedures for handheld data collection, assessed problems in the process of data transfer and aggregation, and developed a data management system for implementation of the GATS Malaysia.

A total of 8 field managers, 23 field supervisors and 41 field interviewers were selected to participate in a centralized training workshop held in Kuala Lumpur from 17 to 21 October 2011. Participants were trained on the contents of the questionnaire, how to complete the questionnaires on paper as well as by using handheld devices, paired mock interviews between participants, and role-plays. After the training workshops, all field interviewers and supervisors were provided with iPAQs loaded with lists of assigned household addresses.

To standardize the survey procedures and minimize non-sampling errors, three manuals and a picture book were prepared. All the manuals were developed first in English and then translated into the Malay language.

Fieldwork took place from 24^th October to 18^th December 2011. Field interviewers and field supervisors who had participated in the training workshop were posted throughout the country to carry out data collection. The field interviewers collected survey information using the handheld devices and submitting the data saved on a secured digital card (SD card) to their respective field supervisor on a daily basis. The data collection teams will visit the selected Living Quarters for at least four visits before classifying it as a non-respond LQs. The same procedure also being followed for individual respondent before classifying it as a non-respondent. Field supervisors were responsible for the overall operation of the field team and for maintaining the time schedule of data collection in the field; as well as transferring data to the national data coordinating centre via internet at least twice a week. The IT/data management team at the Institute for Public Health provided technical support with respect to concerns raised during fieldwork and for troubleshooting any issues with the handheld devices.

Field-level data were aggregated on a daily basis and analysed twice a week to identify data collection errors, problems with skip patterns, and conduct consistency checks. Field-level feedback forms were analysed, and the information was provided to interviewers and supervisors to improve their performance. Quality control techniques were implemented to ensure good and high quality data via direct observation and assessment of interview process in the field by Institute for Public Health supervisors and short telephone interviews for verification with 10% of the completed households.

All estimates that derived from the GATS were in the form of percentage distributions, proportions, ratios, or means. Complex survey analysis was used to obtain prevalence and population estimates with 95% confidence intervals. To improve the representativeness of the sample in terms of the size, distribution, and characteristics of the study population, sample weights were calculated for each respondent prior to the analysis. The analysis was carried out using SPSS version 19 and SUDAAN version 10.1 software; standard errors were calculated using Taylor series linearization. Sample weighting and analysis were done with the help of the Centers for Disease Control and Prevention, Atlanta, USA. For the GATS sample estimates to be representative of the population, it was necessary to multiply the data by a sampling weight, or expansion factor. The basic weight for each sampled household would be equal to the inverse of its probability of selection (calculated by multiplying the probabilities at each sampling stage). As per the GATS manual, a three-step process for GATS sample weighting was conducted for GATS Malaysia¹¹.

Table 2 presents the number of households and persons interviewed and the response rate by residence. Of the 5,112 sampled households, 4,389 completed the household questionnaire, and the calculated household response rate was 88.1%. The household response rate was higher in rural areas (93.4%) compared to urban areas (83.2%). From the 4,389 completed household interviews, there were 4,250 selected individuals who completed the individual interview (calculated person-level response rate was 96.9%). The person-level response rate was slightly higher in rural areas compared to urban areas (98.1% vs. 95.6%). The overall response rate was computed as the product of the household response rate and the person-level response rate. This rate was 85.3%, where the response rate was 79.6% and 91.7% for urban and rural areas respectively.

Overall, 4,250 individuals representing a population of 20.53 million men and women aged 15 years or older were surveyed. By gender, 2,104 men and 2,146 women completed the survey. Respondent from rural area was slightly higher compared to urban (2,185 vs. 2,065). Almost half of the respondents were from the age group of 25–44 years old. By ethnic groups, almost one third was Malays. Table 3 presents the selected demographic characteristics of the respondents.