We conducted a representative, cross-sectional, anonymous, household survey of persons older than 40 years living in all five governorates of the Gaza Strip. We used the 2017 Population and Housing Census as the sampling frame to select enumeration areas (the primary sampling units) from each sampling stratum (North Gaza, Gaza, Dier Al Balah, Khan Yunis, and Rafah governorates) using a systematic cluster random sampling method. We calculated the sample size as 4520 participants from 2443 households based on an estimated prevalence of coronary artery disease (11.3%) [9], a response rate of 90.0%, and design effect of 1.5.

We selected 163 clusters proportionate to the size of the population in the five governorates. In each cluster, the PCBS provided a starting point and data collectors approached every 10th household until 15 were sampled. In each household, we interviewed one eligible male and one eligible female older than 40 years old. If multiple eligible participants of the same sex lived in the same household, we randomly selected one based on a Kish selection grid. We replaced non-respondent or ineligible households using the same selection criteria.

We trained interviewers for four days prior to data collection between March and July 2020, which was delayed for eight weeks due to COVID-19 lockdown measures. A pair of interviewers visited each household and obtained verbal informed consent prior to data collection. We checked participants’ responses for completeness daily, included built-in quality control measures into our survey, and revisited subset of 220 households to verify responses. Where interviewers identified urgent clinical support required for participants, a referral pathway was put in place for to link them to accredited service providers. The study was approved by the Imperial College Research Ethics Committee (20IC5733), the American University of Beirut Institutional Review Board, and the Gaza Helsinki Committee (PHRC/HC/483/19).

We adapted our survey from previously validated surveys, prioritizing those conducted on Palestinian or Arab populations. Briefly, we collected information on participants’ demographics, social assistance, food insecurity (based on the Food Insecurity Experience Scale [10], mental wellbeing (based on the General Health Questionnaire-12 [11] and self-rated health. We asked diagnostic, management, and health service utilization questions about diabetes, hypertension, raised total cholesterol, cardiovascular diseases, chronic respiratory diseases, and cancer [12]. We assessed physical activity (based on the International Physical Activity Questionnaire short form [13], salt intake, tobacco use (based on the World Health Organization-WHO Steps Survey [14] and a detailed semi-quantitative food frequency questionnaire [14, 15] as key NCD risk factors.

We measured height to the nearest centimeter using a Seca 217 stadiometer (Seca GmbH & Co, Hamburg, Germany) and weight to the nearest 10 g using an adult, portable electronic Seca 876 scale. We measured waist circumference mid-way between the lateral lower rib margin and the iliac crest using a Seca 201 meter. We measured blood pressure using the Omron M3 automatic upper arm blood pressure device (Omron Healthcare Co., Ltd. Kyoto, Japan). We developed written guidelines on blood pressure measurement which were used to train data collectors. We ensured participants were at rest for at least 10 min, had abstained from exercising, smoking or drinking coffee for at least 30 min prior to measurement, had removed clothes that could constrict blood vessels, had an empty bladder, and sat straight in a relaxed position with uncrossed legs with their left arm at the heart level. We also ensured an appropriate size positioned one inch above the elbow, which the forearm supported on a table with an upward palm upward, and at least five minutes interval between the two blood pressure readings.

We took the average weight, height, and waist circumference from two consecutive readings. For both systolic blood pressure (SBP) and diastolic blood pressure (DBP), we also relied on two readings. Where there were inconsistencies between the two readings (more than 5 mmHg), we performed a third measurement. For each participant, average SBP and DBP was computed as the average of the recorded readings. The average blood pressure measures were those computed from all readings (two consecutive, or three if inconsistencies) performed. We excluded pregnant women (n = 12) and participants who had an amputation (n = 26) from anthropometric measurements.

Three outcome measures, reflecting the cascade of care, were total hypertension, diagnosed hypertension, and controlled hypertension. Participants were classified as hypertensive if having an average SBP ≥ 140 mmHg or average DBP ≥ 90 mmHg or a hypertension diagnosis. We defined diagnosed hypertension as anyone being informed by a doctor or other health worker that they had hypertension, or if they had taken medication for hypertension as prescribed by their doctor or another health care worker in the previous 2 weeks. We defined treated hypertension as taking, within the past 2 weeks, prescribed antihypertensive medications. We defined controlled hypertension as anyone with diagnosed hypertension with <140 mmHg systolic and <90 mmHg diastolic blood pressure readings.

Sociodemographic independent variables included age (10 year brackets), refugee status (refugee/non-refugee), sex (male/female), governorate (North Gaza/Gaza/Deir Al Balah/Khan Younis/Rafah), education (basic/intermediate/secondary or higher), wealth quintiles based on household assets, income per capita, household size, crowding index (<1/2–3/>3 people per room), locality (camp/non-camp), marital status (not married/engaged or married), employment (yes/no), food insecurity (secure [score 0–3]/mild-to-moderately insecure [4–6]/severely insecure [7,8]), health insurance (yes/no), and past-year social assistance (yes/no).

Risk factor independent variables included physical activity (low/moderate/high, based on the metabolic equivalent minutes per week), tobacco (never/current/quit), the Dietary Interventions to Stop Hypertension (DASH) index [16], and salt addition to food (always/often/sometimes/rarely/never). Independent variables included the presence of a co-morbidity (defined as the presence of at least one disease: diabetes (yes/no), elevated cholesterol (yes/no), heart attack/chest pain (yes/no), stroke (yes/no), cancer (yes/no) pulmonary disease (yes/no), body mass index (underweight [<18.5 kg/m2]/normal [18.5–24.9]/overweight [25.0–29.9]/obese [>30.0]), mental wellbeing (minimal mental illness or psychosocial distress [score <5]/mild [6,7]/moderate-severe [>7]), self-rated health (not good or not good at all/half-half/good or very good) [17].

We performed descriptive statistics to characterize our sample, stratified by sex, and to present the cascade of hypertension care. We used multiple logistic regression models to assess the crude and adjusted associations between independent variables and each outcome (total hypertension, diagnosed hypertension, controlled hypertension). Variables were selected based on potential risk factors, previously identified in the literature, related to hypertension regardless of their statistical association with the outcome [18]. We presented models clustered at the household level. We presented odds ratios, their 95% confidence interval (95% CI) and p value. We also presented the effect modification of age and sex on the prevalence, diagnosis, and control of hypertension. We calculated variance inflation factor to test for multicollinearity. Around 5% of data were excluded from the models due to missing data. We ran the models with and without missing data and found no differences in our findings so present models with missing data. We used Stata 15 (StataCorp) and Statistical Package for Social Sciences (SPSS) 25 (IBM) for these analyses.