1 INTRODUCTION

Mental disorders are a leading health concern globally (Rehm & Shield, 2019; Scott et al., 2018; Vigo et al., 2016; Whiteford et al., 2016) and of increasing interest within the Arab world (El Rassi et al., 2018; Karam & Itani, 2015; Maalouf et al., 2019; Zeinoun et al., 2020). Disability-adjusted life years attributable to mental disorders in Arab countries have been reported to be higher than the global average (Charara et al., 2017; Maalouf et al., 2019; Mokdad et al., 2016). The Global Burden of Disease study estimates that mental and substance misuse disorders account for a higher proportion of global years lived with disability in Qatar (36.7%) than any other country (21% median, 15.4-36.7% range) (Vos et al., 2015). However, this estimate is a projection based on a simulation across many countries and have been criticized for yielding inaccurate estimates for individual countries (Vigo et al., 2016). Actual needs assessment survey data collected in the country are necessary to determine the true burden of mental disorders. To date, no national epidemiological study of this sort has been carried out in Qatar.

Qatar is a small country in the Arabian Peninsula that has witnessed a rapid pace of urbanization, modernization, and economic growth in the last three decades accompanied by major shifts in its socio-demographic profile towards a younger and more highly educated population. These demographic shifts include changes in birth rate and migration. From the beginning of 2000, newborns of foreign nationals increased in parallel with the growing population of foreign residents. Around 90% of Qatar's current population consists of expatriates. Two-thirds of newborns in the country are non-Qatari (De Bel-Air, 2018). The need for high quality epidemiological baseline data has been recognized as necessary for purposes of benchmarking, planning, implementation, and evaluation of a national policy for providing the necessary community-based mental health programs detailed in Qatar's National Mental Health Strategy (Supreme Council of Health, 2013).

Methodology for doing this has been established by the World Health Organization (WHO) World Mental Health (WMH) Survey initiative. The methodology relies on the Composite International Diagnostic Interview (CIDI), a highly structured diagnostic interview, typically administered by trained lay interviewers (Kessler & Ustün, 2004; World Health Organization, 1990). The CIDI has been shown to have good reliability and validity for estimating prevalence of common mental disorders across cultures (Andrews & Peters, 1998; Haro et al., 2006; Kessler et al., 2020; Peters & Andrews, 1995; Robins, 1988; Wittchen, 1994). To date, WMH surveys have been implemented in over 30 countries worldwide, building the necessary infrastructure to collect comprehensive comparative data in countries around the world, several of which are in the Arab world, including Lebanon, Iraq, and Saudi Arabia (Al-Habeeb et al., 2020; Altwaijri et al., 2020; Aradati et al., 2019; Harris et al., 2020; Hcp.med.harvard.edu, n.d.; Karam et al., 2019; Shahab et al., 2017).

Qatar's national mental health study, the World Mental Health Qatar (WMHQ) is carried out by the Social and Economic Survey Research Institute (SESRI) at Qatar University in collaboration with strategic local stakeholders that include the Ministry of Public Health and Hamad Medical Corporation. Technical support is being provided by the WMH Data Collection Coordination Center at the Institute for Social Research, University of Michigan, and the WMH Data Analysis Coordination Center at Harvard Medical School. The diagnostic interview in the WMHQ is based on the most recent international version of the CIDI, CIDI 5.0 Version 3.3, which comprises assessment of mental disorders using the definitions and criteria of the Diagnostic and statistical manual of mental disorders, fifth edition or DSM-5 (American Psychiatric Association, 2013).

The WMHQ study will be based on a representative sample of n = 5000 Arabs, including both Qataris and non-Qataris, living in Qatar. The inclusion of non-Qataris is important because other Arab WMH surveys were either carried out in war-afflicted countries (Iraq and Lebanon) or focused solely on nationals (Saudi Arabia). Inclusion of non-nationals, which make up a large proportion of the population in many Arabic countries (e.g., about one-third of the populations of Lebanon and Saudi Arabia and 90% of the Qatari population) is essential to ensure complete coverage of the mental health care delivery needs within the country. Furthermore, the WMHQ fills in an important need for reliable, valid, and culturally appropriate assessment tools for DSM-5 disorders (Mokdad et al., 2016). Lastly, the study aims to support research training opportunities for investigators from the Arab region, which remain substantially underrepresented in the mental health research arena to date (Zeinoun et al., 2020).

This study describes the original methodology of the WMHQ face-to-face pilot study conducted just prior to the COVID-19 pandemic. We also describe the survey processes used in adapting, translating, testing, administering, and monitoring data quality in the context of the rapidly developing urban yet traditional setting of Qatar.

2 METHODS 2.1 Sample design

The state of Qatar is divided into eight administrative municipalities. The municipalities are further subdivided into 98 zones, which are in turn subdivided into districts and blocks. A sampling frame of all housing units in Qatar was used to draw a representative sample of Qatari nationals (Qataris) and Arab expatriates (non-Qataris), who were 18 years or older and lived in residential housing units in Qatar during the survey reference period. The sampling frame was developed by SESRI with the assistance from Qatar General Electricity and Water Corporation (Diop et al., 2017). In this frame, all housing units in Qatar were listed with information about the housing address and information to identify if residents in the housing units are Qataris (non-migrants) or non-Qataris (migrants).

To assure representation of all municipalities, stratified sampling was used whereby each municipality was treated as one stratum. Inside each municipality, housing units in the zones, subdivisions of municipalities, were ordered by geographic location to permit well distributed sampling of housing units from different areas or blocks. A systematic probability sample with probabilities of selection proportionate to size was constructed for Arab households (Qataris and non-Qataris). Inside each selected household, one eligible adult was randomly selected using a computerized within-household selection method appropriate to Middle East culture (Le-Trung et al., 2013).

The collected data were weighted to account for variation in within-household probabilities of selection based on household size, non-response, and post-stratification calibration to known population targets to help reduce residual effects of non-response and under-coverage of the sampling frame. Our weighting variable ranged from 441 to 13,279 with a variance of 1360 and mean of 1744.

Non-response propensity score weighting at the household level was applied to adjust the within-household inverse probability of selection to account for non-response by municipality and pre-specified gender sampling information (Trung Le et al., 2014), using the following formula: where , the adjustment factor for non-response, was based on the propensity that a sampled unit was likely to respond to the survey (Ridgeway et al., 2015).

Post-stratification was then carried out with these weighted data using a raking method to align survey distributions with known population characteristics for age, gender, nationality (Qatari/non-Qatari), and marital status obtained from Census bureau (Ministry of Development Planning and Statistics, 2015).

2.1.1 Response rates

The final status of sampled households and two corresponding response rates (RR1 and RR2) were calculated using standards set by the American Association for Public Opinion Research (AAPOR, 2015). First, RR1 which is the ratio of the number of completed interviews to the total sample size after excluding those who were ineligible: where C is the number of completes, E is the number of eligible responses, and UE is the number of unknown eligibility. Second, which adjusts the denominator after estimating the proportion of eligible participants from those of unknown eligibility where e is the estimated proportion of eligibilities given by where IE is the number of ineligibles.

2.2 Instrument translation and adaptation

Arabic is a polyglossic language with dialects that differ across regions of the Arab world. As our target population consisted of Qatari and non-Qatari Arabs, the language of the instrument had to be understood by all Arabic speaking respondents alike. The conceptual equivalence of the Arabic CIDI has already been established in prior WMH surveys conducted in Arabic-speaking countries like Lebanon (Karam et al., 2006), Iraq (Alhasnawi et al. 2009), and the latest in Saudi Arabia (Altwaijri et al., 2020). But to further ensure conceptual equivalence in our Arabic version of the CIDI 5.0, which was translated from English to a standard modern conversational form of Arabic, we used the process of translation and adaptation of instruments as outlined by the WHO guidelines (WHO, n.d.) and the Translation, Review, Adjudication, Pretesting and Documentation method (Kessler et al., 2008).

2.2.1 Forward-back translation

The first team in charge of the forward-translation from English to Arabic consisted of five bilingual team members who are from different Arabic-speaking countries (Iraq/Kuwait, Syria, Palestine, Sudan, and Morocco) and different disciplines (Public Health, Anthropology, Linguistics, and Social Sciences). Different members of the translation team independently translated different assigned modules. The team met every week for 4–5 h over 3 months to review and discuss the translation of each new module and reach consensus where there was any disagreement on terminology or phrasing of certain questions.

We relied on extensive back-translation process to assess conceptual equivalence to the original English version of the CIDI 5.0. The back-translation stage from Arabic to English lasted approximately 2 months and was conducted in parallel (as newly translated Arabic modules became available) by a second team of two other bilingual researchers who had not seen the original English version of the CIDI 5.0. The back-translation team lead by a senior researcher resolved any inconsistency in the back-translated English versions and produced a single unified back-translated English version.

In the last consolidation stage, the two leads from the two teams met and reviewed the entire Arabic and English versions of the instrument. For the most part, minor discrepancies in translation arose and were resolved by consensus among the leads of the bilingual teams. There were only two instances where the leads resorted to the larger translation team for advice in reaching agreement on unresolved issues.

2.2.2 Cultural adaptation and pre-testing

We also used findings from the cognitive interviews conducted by the Saudi National Mental Health survey to make cultural adaptations to sensitive questions in the survey (Shahab et al., 2019). Qatar and Saudi Arabia—both Arab states that are part of the Gulf Cooperation Council—share the same traditional conservative Islamic culture and their people would have similar sensitivity concerns and misunderstandings of certain concepts. Furthermore, the pre-testing stage entailed this relatively large pilot study during which detailed feedback about our version of the instrument was received from the interviewers who administered the instrument in face-to-face interviews leading to subsequent modification to certain problematic (comprehension) or potentially insensitive or offensive wordings.

2.3 Instrument programming and administration

A team of programmers from SESRI and the Harvard-Michigan WMH coordinating centers programmed the questionnaire in Blaise 5.2 (Blaise, 2017). Trained interviewers used Computer-assisted personal interviewing (CAPI) to administer the instrument during face-to-face interviews conducted in households of Arabic speaking residents and nationals of Qatar. These respondents were recruited from January 2020 through February 2020.

In order to customize our version of the questionnaire to Qatar's context, we adapted culturally sensitive questions, entire optional sections, and added non-CIDI sections using a CAPI modularization program. The final survey instrument consisted of 25 sections including 20 sections from the CIDI 5.0. For more details about the content of the WMHQ survey, please refer to Table 1.

TABLE 1. CIDI and non-CIDI modules used in Qatar's WMH study pilot Module name Respondent contact Your background Your health MoCA—Montreal cognitive assessmenta Depression Persistent depression High mood Worry and anxiety Anger attacks Panic attacks Social anxiety Stressful experiences Self-harm Tobacco, alcohol, and drugs Unusual experiences Treatment of emotional problems Employment Finances Personal relationships Social networks Childhood experiences Schizotypal Personality Questionnairea Family history of psychological, emotional or behavioral problemsa Resiliencea Abbreviations: CIDI, Composite International Diagnostic Interview; WMH, World Mental Health. a Non-CIDI modules.

In addition to the CAPI mode, SESRI IT team used the audio computer-assisted self-interviewing (ACASI) Blaise feature to program an ACASI version for two modules of CIDI 5.0 that were particularly sensitive in the cultural context of Qatar: the Self Harm and Tobacco, Alcohol and Drugs modules.

ACASI enabled respondents to listen to the questions through headphones and to enter responses using a touch-screen. The questions of the sensitive CIDI modules were presented as recorded audio voice-overs. As mispronunciation and unfamiliar accents could influence the participants' responses, two recorders (one female and one male) fluent in the Arabic language clearly enunciated the questions without any heavy accent to ensure the audios were understandable to the study population.

2.4 Study-related training 2.4.1 Train-the-trainer sessions

The study team, including the principal investigator of the study, three research assistants, field operation team, and two experienced interviewers, attended 2 days of webinar-based training on the administration of the CIDI 5.0 instrument conducted by the CIDI training center at University of Michigan Institute for Social Research. Participants were provided with password-protected access to online training modules for the CIDI and were required to participate in live interaction webinars. Upon completion of the CIDI training, participants received their certificates, which enabled them to train the local field interviewers.

2.4.2 Field interviewers selection and training process

Initially, 31 field interviewers were nominated for training based on their extensive experience and performance in previous surveys conducted by SESRI. Upon further screening, 29 interviewers (20 females and 9 males) were invited to attend the WMH survey-training workshop.

Culturally, it is more acceptable for female interviewers to visit households and interview male respondents. As such, more females were trained than males.

Prior to the workshop, potential interviewers received the study material, which included a hard copy of the questionnaire, the training slides, and the respondent package, as presented in details in Appendix A.

The training workshop spanned over 5 days, starting at 8 a.m. and finishing at 4 p.m. The training was conducted at SESRI's headquarters at Qatar University. Specific topics covered during the training and other training-related details are summarized in Figure 1.

World Mental Health (WMH) Study—Interviewer's selection and training process

Role-playing using scripted interviews prepared in advance were used in the last 3 days of the training. By the end of the training course, each interviewer completed at least three scripted interviews.

Out of the 29 interviewers, 27 passed the evaluation and successfully completed the study training. The evaluation was based on performance during the training, which included assessment for IT skills and degree of understanding of the CIDI administration (question reading and probing techniques) and rounding rules.

2.5 Field operation 2.5.1 Team structure

The final survey fielding team consisted of 27 interviewers (20 females plus 7 males), 8 head of groups (HoGs), and 4 supervisors. Each interviewing group consisted of two female interviewers and one male HoG except for male interviewers who worked alone, that is, one male interviewer only. The main role of HoGs is to transport and accompany each pair of female interviewers to their respective sampled households for conducting the interviews. Two out of the four supervisors directly oversaw the male interviewers, while the two remaining supervisors oversaw female interviewers and their HoGs.

2.5.2 Sample allocation

As the sampled households were spread geographically all over Qatar and across zones; the supervisors were responsible for allocating to each male interviewer or HoG a detailed list of sampled households, their location, and a timeline for completing these potential interviews within each zone. Each HoG then allocated a specific number of sampled households within the same zone to each group of interviewers based on sample location, which was often close to the interviewers' home locations, starting with sampled households that are closest to each other.

2.5.3 Fielding operations team roles

As described earlier, the HoG role fulfilled the cultural expectations of a male chaperone of female interviewers and ensured their safety in case of altercations or other issues upon reaching or contacting the sampled households for the first time that is without arranging appointments a head of household visit.

Upon reaching the first destination, the HoG and the first female interviewer would contact the sampled household. If the contact was successfully established, the HoG would proceed along with the second female interviewer to the next sampled household, while the first female interviewer would conduct the interview with the selected respondent in the first household. At the end of each working day in the field, the HoGs and interviewers would complete a detailed report about the status of the sampled households that were visited. This reporting to the supervisor was done on daily basis.

In turn, supervisors monitored fielding progress in terms of achieving target completed interviews by teams and ensured smooth survey fielding experience for participants and survey fielding team alike. In particular, supervisors monitored and evaluated interviewers as they collected data to ensure data integrity by conducting face-to-face verification of visited households and by conducting random checks to verify adherence to fielding protocol and practices including wearing appropriate uniform and carrying valid ID, study permit, and consent forms.

Interviewers worked from 4 to 8 p.m. Sunday to Friday, but also worked to accommodate scheduled appointments with potential respondents. Interviewers were paid per working day rather than per competed interview, which has been shown previously to yield better data quality (Kessler & Üstün, 2004).

2.5.4 Fielding procedures

Field procedures and strategies utilized by interviewers in handling most scenarios commonly encountered in the field are shown in Figure 2.

2.6 Safety protocol

As part of the study's safety procedures, interviewers provided all participants with contact sheets that included information about: (1) the Mental Health Service hotline; (2) the emergency department contact number; and (3) contact details for local primary healthcare centers.

2.7 Quality control indicators

We aimed to ensure that the interviewers' performance adhered to preset quality criteria. To this end, SESRI IT team, in collaboration with University of Michigan, developed and implemented a Quality Control (QC) system to monitor interviewer activity in the field.

2.7.1 CIDI-specific quality metrics

The QC indicator database, with visualizations of flagged activities based on preset criteria in a workbook called Power BI (Powerbi.microsoft.com, n.d.), served as a primary source in the investigation of data quality in relation to the CIDI. Values of flags were updated once per day, at 23:00 (UTC +3). SESRI QC monitoring team investigated daily output (tables and charts) from the QC system and reported to the principal investigator and the Field Operations manager the issues found. The Field Operations team compared the results with their field reports and reported back to the principal investigator and research team. These indicators included the following flags: low response rate; too many completed interviews per day; short question field time; short interview length; low prevalence rate; and long interview pause. For more information on these indicators, please see Appendix B.

2.7.2 Verification indicators & random visits

In addition to the QC indicators with flags, performance indicators were also calculated and monitored regularly. Furthermore, a random sample of interviews were selected for face-to-face or phone verification.

Phone verification was conducted by interviewers from the computer-assisted telephone interviewing lab at SESRI. Two demographic questions and three general health variables (height, weight, and presence of one or more life-threatening or seriously impairing chronic physical health problem) were re-asked during the phone verification, as the mental health variables were considered too sensitive to verify over the phone.

We also developed and monitored indicator for discrepancy between the frame (expected) location and the actual location of interviews or geolocation. Two methods allowed the interviewers to capture their geolocations: mobile application developed for the purpose of capturing interviewers' visit points and the CAPI system installed on their laptops. Field Operations Manager reviewed discrepancies daily and discrepancy representing a distance of greater than 30 m was flagged for verification.

Supervisors also conducted in-person random visits of interviewers in the field and evaluated the status of each visit as good, warning, or suspended from field based on preset scored criteria (such as adherence to consenting protocol, carrying an identifying badge, and physical appearance).

3 RESULTS 3.1 Sample distribution & response rates

The total pilot sample consisted of n = 1076 households selected proportional to size of municipalities in Qatar. Response rates ranged from 41.2% to 54.9% across municipalities based on 395 completed survey interviews (see Table 2).

TABLE 2. Distribution of total pilot sample and response rate by municipality Municipality Number of sampled units Percentage of total sample (%) Response rate (%) Ad Dawhah 437 40.6 43.8 Al Rayyan 399 37.0 51.1 Al Wakrah 77 7.2 54.9 Umm Salal 62 5.8 43.5 Other 101 9.4 41.2 Note: Small municipalities were grouped into “Other,” including Al Shamal, Al Khor, Al-Shahaniya, and Al Daayen.

The overall response rates and final status of sampled households in the survey are shown in Table 3. RR1 and RR2 were 46.1% and 56.9%, respectively. Eligible households that did not complete the interview included housing units of Arab residents who refused to participate in the study. Approximately, 3.5% of the total sample refused to participate in the study. Ineligible households included mostly non-Arab households, as well as housing units under maintenance, vacant housing units, and nonresidential buildings. Households with no one at residence (closed) or falling under the sample error subcategory were also not eligible for the survey. Households with unknown eligibility included housing units that were initially contacted but without success in reaching the participant even after three attempts or where a contact was faced with obstructive or uncooperative household keepers (servants, drivers, and security personnel) so eligibility status of potential participants could not be determined.

TABLE 3. Final interview status of sampled households and response rates Interview status Households Frequency Total sample 1076 Completed 349 Not completed 727 Eligible householda 73 Ineligible householdb 319 Unknown eligibilityc 335 Response rate (RR1)d 46.1 Response rate (RR2)e 56.9+ Note: Response rates (RR1 and RR2) were calculated using standardized coding and interpretation procedure for different dialing outcomes as set by the AAPOR (2015). a Eligible households that did not complete the interview included housing units of Arab residents who refused to participate in the study and those who agreed to an appointment, but the appointment was not fulfilled upon follow-up. b Ineligible households included mostly non-Arab households, as well as housing units under maintenance, vacant housing units, and nonresidential buildings. Households with no one at residence (closed) or falling under sample error subcategory were also not eligible for the survey. c Households with unknown eligibility included housing units that were initially contacted but without success in reaching the participant even after three attempts or where a contact was faced with obstructive or uncooperative household keepers (servants, drivers, and security personnel) so eligibility status of potential participants could not be determined. d RR1 is the ratio of the number of completed interviews to the total sample size after excluding those who were ineligible: RR1 = C/(C + E + UE), where C is the number of completes, E is the number of eligible responses, and UE is the number of unknown eligibility. e RR2 = C/(C + E + eUE) adjusts the denominator after estimating the proportion of eligible participants from those of unknown eligibility, where e is the estimated proportion of eligibilities given by e = (C + E)/(C + E + IE), where IE is the number of ineligibles. 3.2 Sample characteristics

Weighted descriptive statistics of the overall sample of respondents who completed the interview are presented in Table 4. The mean age was 37.6 years (SD = 12.1). Overall, three-quarters of respondents were married and 39.7% reported receiving an undergraduate degree. Approximately, 51.4% of respondents were male and 48.6% were female, which is representative of Qatari population (Planning and Statistics Authority, 2018). Also as observed in our sample, but not officially made available by Qatar's Planning and Statistics Authority, Qataris are minority in the population of Qatar (Winckler, 2015). Most of the respondents in our sample (82.0%) were non-Qataris (Snoj, 2019), while 18.0% of respondents were Qataris, with almost equal gender distribution in both groups. As the census bureau in Qatar (Qatar Census, n.d.) does not publish a lot of official demographic data by nationality (Qataris vs. non-Qataris) or ethnicity (Arab vs. non-Arab), it was not possible to directly compare the distribution of our sample broken down by these variables with official population distributions.

TABLE 4. Characteristics of final sample of respondents who completed CIDI Frequency (n) (%)a % Gender Male 162 51.4 46.0 Female 187 48.6 54.0 Nationality Qatari 63 29.0 18.0 Non-Qatari 286 71.0 82.0 Marital status Married 286 74.9 82.0 Divorced/Separated 12 2.5 3.0 Widowed 6 1.7 2.0 Never married 44 20.8 13.0 Education level None 7 2.9 2.0 Primary school (1st–6th grade) 31 12.3 9.0 Secondary school (6th–12th grade) 71 20.6 20.0 Post-secondary school/Diploma 45 15.2 13.0 Undergraduate degree 158 39.7 46.0 Graduate (MA/PhD) 34 9.2 10.0 Abbreviation: CIDI, Composite International Diagnostic Interview. a Percentages based on weighted proportions. 3.3 Performance metrics

On average 66 contacts attempts were made per day. These contact attempts lead to an average of 66 completed interviews per day, with a mean of one completed interview per interviewer. The average length of time per interview was 97 min.