Access to Care and Prevalence of Hypertension and Diabetes Among Syrian Refugees in Northern Jordan

Key Points Question What is the prevalence of hypertension and diabetes among long-displaced Syrian refugees in northern Jordan and what is their level of access to care? Findings In this cross-sectional study of 1022 randomly sampled households of Syrian refugees, the biologically based prevalence of hypertension and diabetes was moderately higher than self-reported prevalence. Among the participants, 57.4% had 1 or more complication, 82.8% were obese or overweight, 49.1% sought care in the past month, and 26.8% missed their medications in the past week. Meaning These findings suggest that long-term disease management is inadequate, in that Syrian refugees were generally aware of their diagnoses and had access to medication, but complications and factors associated with severe disease were highly prevalent.


Introduction
In the Eastern Mediterranean region, the transition from a burden of primarily infectious diseases to noncommunicable diseases (NCDs) has been associated with increased population growth and longevity. 1,2 Proportional mortality from NCDs has been projected to increase from 62% in 2015 to 70% in 2030. 3,4 Conflicts in Iraq, Syria, and Yemen have made the inadequate management of NCDs among conflict-affected and displaced populations a major public health issue. 5 NCD management in humanitarian settings is poorly studied, and health responses have been slow to move away from the paradigm of episodic clinical care. [5][6][7] Health systems and humanitarian organizations are challenged to provide integrated and cost-effective approaches to stabilize acute presentations, ensure continuous treatment, provide access to medications and insulin, provide patient education, and manage acute complications. 5,7 Specifically, the crisis in Syria has greatly impacted regional health trends and national health systems. 1 As of January 2020, 5.6 million refugees were displaced to Turkey, Lebanon, Jordan, and Iraq. 8 In Jordan, 1 in 14 people is a registered refugee, and 79% of refugees live outside camps in urban and periurban areas. 8,9 Household surveys have documented that one-half of refugee households have 1 or more adult with an NCD. 10 A 2016 household survey among Syrians in northern Jordan found the most prevalent diagnoses to be hypertension (14.0%) and diabetes (9.2%). 11 A 2015 clinic-based survey among Syrian individuals with diabetes in Bekaa Valley, Lebanon, found that 30% of patients received a diagnosis during displacement, decreasing the likelihood that they had received comprehensive education on disease management. 12 Because Syrian refugees may have developed NCDs after an extended displacement and may lack a diagnosis and awareness of their condition, it follows that neither the disease burden nor health care utilization is well-understood.
Syrian refugees in Jordan access primary care from clinics run by the Jordanian Ministry of Health, nongovernmental organizations (NGOs), and the private sector. In January 2018, facing budget shortfalls, the Ministry of Health reduced subsidies for refugees at public clinics (reinstated in March 2019). 13,14 Household surveys have cited costs, lack of knowledge of services, and availability of services as primary barriers to NCD care. 10,11,13 Interruptions likely affect disease control; in 2016, 25% of surveyed patients with NCDs in northern Jordan reported medication interruptions longer than 2 weeks during the past 6 months, primarily because of costs. 11 There is emerging evidence that community health worker (CHW) models that focus on NCDs can facilitate linkage and continuity of care. [15][16][17][18] The International Rescue Committee, a humanitarian organization that has provided primary care for Syrian refugees since 2012, has integrated community health into the primary care model. As part of a study to design and evaluate a CHW model for the management of NCDs among refugees, we conducted a household survey among Syrian refugees living outside camps in northern Jordan. The primary objectives were to quantify prevalence using biological measures of hypertension and diabetes, determine the proportion of known and unknown diagnoses, and evaluate access to care for diabetes and hypertension in the catchment area.

Study Design, Setting, and Participants
Ethical approval for this cross-sectional study was granted from the institutional review boards of the

Cluster Development, Sample Size, and Sampling
Because no database of refugee households was available, a cluster sampling design was used. To construct clusters, a grid was superimposed over maps developed in Quantum GIS software version 3.6.0 (Open Source Geospatial Foundation Project) using shape files created with the LandScan database and a Google Satellite layer. 11 The areas within the grid borders defined a cluster (see the Figure). 20,21 To detect the prevalence of self-reported hypertension and diabetes among adults (aged Ն18 years) (23% and 16% respectively, based on averaging estimates from surveys in northern Jordan 11 and precrisis Syria 22 ) such that the 95% CI had a precision of 5%, a sample size of 1050 households was calculated. A design effect of 1.5, 10% nonresponse rate, and 2 adults aged 18 years or older per household were assumed. 11 A large number of clusters relative to households per cluster (70 clusters of 15 households) was used to increase the likelihood of finding sufficient refugee households embedded within host communities.
Sampling occurred in 2 stages: sampling of clusters and households within clusters. To select clusters, 70 geospatial coordinates were randomly allocated across the administrative areas, proportional to the population size of refugees. A cluster was sampled if a coordinate fell within its boundary. To sample households within a cluster, community health volunteers first located the household, shop, or mosque closest to the sampled geospatial coordinate. A chain referral process, wherein each household was asked about the next nearest Syrian household, was used to identify 20 to 30 households within each cluster. 10,11,23 For multidwelling buildings, 1 household was randomly selected from the enumerated number of Syrian households. The next day, survey teams used the map to locate the first 15 households. To reduce selection bias, if households were absent, a follow-up appointment was made to revisit that same day. If households were unavailable after repeated attempts, they were replaced by the next mapped household. If 15 households could not be located, the cluster was considered complete.

Data Collection and Variables
A questionnaire was designed in KoBoCollect software version 2.019.07 (KoboToolbox), which Jordanian nurses administered in Arabic using tablet computers. Each team consisted of 2 nurses.
Nurses took a household census and enquired about prior diagnoses among adults aged 18 years or older to estimate prevalence in the adult population. To estimate biologically based prevalence among the higher-risk individuals aged 30 years or older and knowledge of relevant diagnoses, an adult aged 30 years or older was randomly selected. Nurses used auscultation and an electric sphygmomanometer to measure blood pressure (BP) 3 times, 5 minutes apart. Patients were seated with an unclothed arm supported at the level of his or her heart. A glucometer and testing strips were used to measure random blood glucose (RBG).Those with above-threshold BP and/or RBG were referred for care. An electronic weighing scale and measuring tape were used to measure body mass index (calculated as weight in kilograms divided by height in meters squared). Nurses asked about current medication and complications of disease, including heart problems, stroke, extremity numbness, poorly healing wounds, renal problems, and amputations (see eTable 1 in the Supplement for details).
An available adult aged 18 years or older with self-reported hypertension and/or diabetes was randomly selected to answer questions concerning access to care, current use and adherence to medication, and socioeconomic factors. Supervisors checked questionnaires before leaving the household. A 5-day training covered sampling, interviewing, role-playing, standardized biological measurement, and pilot testing of questionnaires in areas outside the sampling frame.

Statistical Analysis
Analysis was conducted in Stata statistical software version 14.2 (StataCorp), using the svyset command to produce design effects and point estimates with appropriate 95% CIs. Two sets of prevalence estimates for hypertension, diabetes, and both conditions were calculated. Among adults aged 18 years or older, self-reported prevalence was calculated to estimate the known disease burden. This used the total number of self-reported diagnoses (numerator) and the size of the population aged 18 years or older (denominator). 11,24,25 Among adults aged 30 years or older, abovethreshold BP (mean of last 2 of 3 systolic BP and diastolic BP measures Ն140/90 mm Hg) and above-threshold RBG (Ն200 mg/dL, regardless of fasting status; to convert blood glucose to mmol/L, multiply by 0.0555) were estimated. 24 To estimate the total burden, above-threshold estimates were added to the number of respondents currently taking medication for each condition (numerator), along with the screened population size of adults aged 30 years or older (denominator) (World Health Organization STEPS method). 24 To analyze age and sex as determinants of prevalence among adults aged 30 years or older, prevalence ratios (PRs) and 95% CIs were calculated using a generalized linear Poisson regression with robust variance. 26,27 Multivariable logistic regression was used to investigate determinants (age and sex) of having an undiagnosed disease (ie, undiagnosed and positive result vs diagnosed regardless of screening result) achieving statistical significance (2-sided P < .05). For adults aged 18 years or older reporting hypertension and/or diabetes, sex-specific point estimates for access to care were calculated. Missing data are noted in the tables. Data analysis was performed from May to September 2019. household size was 6 (2.5) persons.

Self-reported Diagnoses Among Adults Aged 18 Years or Older
Among adults aged 18 years or older, the self-reported prevalence of hypertension was 17.2% (95% CI, 15.9%-18.6%), that of diabetes (both insulin dependent and non-insulin dependent) was 9.8% (95% CI, 8.6%-11.1%), and that of both conditions was 7.3% (95% CI, 6.3%-8.5%). This equated to 1 in 5 adults reporting any diagnosis.          concomitant diabetes to be identified and controlled at an early stage. Among patients with known diagnoses, there is a critical need to treat complications to prevent severe disease and avoid excessive health care costs. This is important given the prolonged nature of displacement globally and limited resources for refugee health. 32 Population-based primary prevention policies targeting diet, smoking, and physical activity in Jordan should consider the socioeconomic environments of refugees. 33 However, NGOs and district health systems could also sharpen their focus on high-risk groups. First, this includes broadening the group considered to be at risk of developing NCDs to age 30 years and older, especially women, during clinic-based screening. Second, given the poor access to secondary care, primary care must improve awareness among practitioners of early detection and management of simple complications (eg, diabetic foot), similar to primary care approaches in Iran. 17,29,30 Third, high-risk patients (eg, with poorly controlled disease, serious complications, or type 1 diabetes) could be counseled on both medical and preventative interventions to reduce their risk of deterioration. 7 This includes selfmanagement protocols for monitoring and recognition of danger signs, exercise classes for women in private spaces, and counseling on reducing salt intake. 12,34,35 Fourth, psychosocial care related to conflict and displacement can reduce hopelessness and increase motivation for seeking care. 29,36 Given the presence of CHW networks in protracted crises, task-shifting of nonclinical activities, such as routine monitoring, patient education, and psychosocial and peer support to CHWs, would relieve staff. 7,16,34,37 Limitations

JAMA Network Open | Global Health
This study has limitations that should be addressed. The chain referral method introduces selection bias because it is dependent on the respondents' knowledge of neighboring Syrian households and possible predilection to refer to family or friends. Unhoused persons and refugees who are intentionally hidden are excluded. Nonetheless, nearly all clusters had sufficient households, and data collectors were indeed referred to persons living in tents, shacks, and empty buildings.
Two-thirds of respondents were female, and the health of working men may be different from that of those available during the daytime. The biological assessments cannot be used to confirm diagnoses. Estimation of diabetes prevalence is a known problem for surveys. 38 The use of RBG instead of fasting blood glucose is nonstandard; therefore, prevalence is approximated. 24

Conclusions
This study documents the impact of long-term displacement on Syrians who sought refuge in Jordan for a duration long enough where achieving disease control should become feasible. 7 By focusing programs on early identification through clinical screening and improving adherence to continuous care and secondary prevention among patients, severe morbidity among refugees could be minimized here and in other protracted crises.