Title: Predictors and barriers for the management of non-communicable diseases among older Syrian refugees amidst the COVID-19 pandemic in Lebanon: A cross-sectional analysis of a multi-wave survey

bootstrapping optimism-adjusted discrimination and calibration of the model using C-statistic and calibration (C-slope), respectively. one NCD 24% of and 11% CRD. There were (20%) participants who at least one of their NCDs. Predictors for manage NCDs and multiple The model’s adjusted C-statistic was and C-slope Conclusions: This study identified that the predictors of inability to manage NCDs among older Syrian refugees in Lebanon are mainly related to financial barriers, which aids the targeting of assistance and interventions. Context-appropriate assistance is required to overcome financial barriers and enable equitable access to medication and healthcare.


Introduction
Forcibly displaced populations have been recognized as vulnerable populations to the direct and indirect impacts of the COVID-19 pandemic. [1][2][3] Furthermore, older refugees are particularly vulnerable as they have a higher prevalence of pre-existing chronic disease 4 and have an increased risk of hospitalization and mortality due to COVID-19. 5 Even prior to the COVID-19 pandemic, poor management of noncommunicable diseases (NCDs) was a global public health concern. 6 NCD self-management, including lifestyle changes and adherence to medication, has been shown to reduce premature mortality and morbidity. [7][8][9] In refugee populations, however, NCD management is problematic due to the competing demands of basic needs such as access to adequate food, shelter, protection and water. Furthermore, NCD self-management in refugee populations is exacerbated by poor access to health services, language barriers between the patient and the healthcare provider, being food insecure, and having limited health literacy. 10 In Lebanon, large populations of Syrian refugees are situated in the Bekaa valley and North Lebanon and reside in informal tented settlements or residential areas. 11 These are areas where the Lebanese health system was already stretched to meet the needs of the host population, and since 2011 have faced unprecedented challenges in meeting the needs of Syrian refugees. 12 Refugees have access to subsidized healthcare services and medication mainly through the United Nations High Commissioner for Refugees (UNHCR); however, despite these subsidies they are often unable to cover the full cost of needed medication. 13 The COVID-19 pandemic and the recent major economic crisis in Lebanon have impacted access to healthcare and supply of essential medicines 14 , which is likely to further exacerbate refugees' abililty to manage chronic conditions. Studies on the management of NCDs among refugees in Lebanon and the Middle East, a region that hosts one of the largest refugee populations, are scarce. 15 Understanding the barriers and predictors of selfreported management of NCDs among older refugees, a population at high-risk, is important to allow resource allocation and contextualized humanitarian assistance to prevent premature mortality and morbidity. The present study aimed to elucidate the predictors of inability to manage NCDs in older Syrian refugees and describe barriers to accessing healthcare and managing these chronic conditions . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

Study design, sample and study population
This was a cross-sectional study nested within a multi-wave longitudinal study, which aimed to examine the vulnerabilities of older Syrian refugees residing in Lebanon during the COVID-19 pandemic. 16 The study included Syrian refugees aged 50 years and older, who were identified from a full listing of beneficiary households of a non-governmental humanitarian organization [Norwegian Refugee Council].
Within the beneficiary sampling frame, all households who had used services offered by the humanitarian organization between 2017 and 2020 and included an adult aged 50 years or older were contacted and were included in the study. If there were multiple adults aged 50 years or older within a household, one person was randomly selected and was assessed for capacity to consent, invited to participate, and provided oral consent to enter the study. The same respondent was approached to complete a telephone survey across different time points. The data for the present analysis were extracted from Waves 1 (September 2020-December 2020) and Waves 2 (October 2020-January 2021). This study was reported according to the TRIPOD reporting guideline for prediction modelling. 17 Out of 17,384 households initially contacted at wave 1, 4,010 eligible beneficiary Syrian refugees aged 50 years and over were invited to participate; of those, 3,322 beneficiaries consented and participated in both waves 1 and 2 ( Figure 1), 1893 respondents reported having at least one NCD and and 387 reported being unable to manage at least one NCD.

Data sources
The questionnaire for each wave was developed using a combination of sources including validated questionnaire modules, contextually specific questions, and community-identified priorities. The survey was co-created by academics, humanitarian actors, local government officials, and focal points from the refugee communities. Modules varied between the waves; the demographic module was in wave 1, and the NCD and access to healthcare module was included in wave 2. The questionnaire was piloted internally with data collectors and local community focal points to ensure face validity. Trained data collectors administered the surveys in Arabic and entered data into structured electronic data collection forms hosted on Kobo toolbox. Data entry checks and monitoring were performed for quality assurance.

Outcome Measures
The outcome variable was self-reported inability to manage any NCD. This included the following conditions: hypertension, diabetes, cardiovascular disease or chronic respiratory disease. For each . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 17, 2022. ; https://doi.org/10.1101/2022.04.12.22273786 doi: medRxiv preprint condition, participants were asked the following question "Are you able to manage your [Insert condition]?".

Candidate predictors
Using the literature, 16 potential predictors were identified for inability to manage NCDs and were included for model development. These included: age (continuous); sex (male/female); residence (outside/inside tented settlements); education (preparatory and elementary/never attended); smoking status (smoker/ex-smoker/never-smoker); number of chronic conditions (one/two/three or more); hypertension (yes/no); diabetes (yes/no); chronic respiratory disease (yes/no); cardiovascular disease (yes/no); living arrangement (alone/with someone); food insecurity measured using the Food Insecurity Experience Scale 18 (severe food insecurity (raw score 7-8)/mild-moderate food insecurity (raw score 4-6)/ food secure raw score 0-3)); employment status (yes/no); water insecurity measured using the short-form Household Water Insecurity Scale 19 (yes/no); receipt of cash or voucher assistance (yes/no) and family debts (yes/no).

Missing data
The largest amount of missing data in any variable was 1.8%, and these were assumed to be missing at random, so we used complete case analysis. 20

Statistical Analysis
Absolute frequencies and proportions were presented alongside odds ratios and their 95% confidence intervals using unadjusted logistic regression models, which examined the odds of inability to manage any NCD for each candidate predictor.
All variables were categorical other than age, which had a linear relationship with inability to manage NCD. All candidate predictors of inability to manage NCD were entered into multivariable logistic regression models and removed using a stepwise backwards method using a P<0.157 21 . This has been used as a proxy for the Akaike Information Criterion (AIC), where predictors are removed to obtain the lowest AIC. Multicollinearity was assessed using correlation matrices and variance inflation factor; a variance inflation factor > 5 indicated collinearity. Number of NCDs and each separate chronic condition were modelled in two separate models. Smoking status was removed from the final model to improve model fit and prevent overfitting. 22 The final model's performance in terms of discrimination was assessed using the C-statistic, which ranges from 0.5 -1.0 (this is also known as the area under the receiver operating characteristic curve (AUC-ROC)), where a value of 1.0 represents perfect discriminative ability between those with and without the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 17, 2022. ; https://doi.org/10.1101/2022.04.12.22273786 doi: medRxiv preprint outcome, and 0.5 denotes a discriminative ability equal to chance. We also assessed the calibration of the final model, which describes the agreement between observed outcomes and predictive probabilities. 23 This was assessed using calibration plots, which categorizes patients into ten groups according to predictive probabilities, where the mean predicted risk within each of these groups is plotted against the mean observed proportion of events. 23,24 If there is a perfect calibration, the graph will show a diagonal line where there is a slope of 1 and intercept of 0. A slope <1 suggests overfitting in the model, this is where respondents with high risk of the outcome have overestimated risk predictions while those with low risk of the outcome have underestimated risk predictions. In addition, we assessed whether there was an overall difference between the observed number of events and the average predictive risk using the calibration-in-the-large.
The final model's selection of predictors, discrimination and calibration estimates were internally validated using bootstrap methods; where 500 bootstrap samples with replacement were used to validate the model selection process and generate an estimate of optimism, optimism adjusted estimates of Cstatistic and optimism-adjusted calibration plot were generated. 20,23,25,26 Bootstrap shrinkage was applied to the final apparent model and the modified beta coefficients and odds ratios were presented. All analyses were conducted using Stata/SE 17.

Ethical approval
This study was approved by the American University of Beirut Social and Behavioral Sciences Institutional Review Board [Reference: SBS-2020-0329]. Refugees are considered vulnerable populations, and extensive efforts were made to ensure autonomy was respected throughout the data collection.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 17, 2022. ; https://doi.org/10.1101/2022.04.12.22273786 doi: medRxiv preprint

Characteristics of the population
Out of a total of 1,893 participants, 387 noted that they were unable to manage their condition through any means (lifestyle or medication) ( Table 1). Among these participants, 174 were unable to manage their NCD through medication. The study sample median age was 59 (IQR: 54-65), and 1089 (58%) were women. Hypertension topped the listed of reported NCDs (74%), this was followed by cardiovascular disease (42%), diabetes (41%), and chronic respiratory disease (19%).
Among 174 participants who were unable to manage their NCD through medication, the primary reasons included unaffordability of the medication (40.8%), feeling better and not needing the medication (22.4%) and medication not always available (14.4%). Unaffordability of medication remained the primary reason for hypertension, diabetes and cardiovascular disease when examining disease-specific reasons for nonadherence to medication ( Figure 2a). In addition, there were 295 participants who were unable to manage their NCDs and unable to access primary health care. The primary reason for being unable to access primary care in this group was the cost of the doctor's visits, medication or tests (80.7%) (Figure 2b).

Predictors and model performance
The final model retained eight predictors of inability to manage NCD, which included age, self-reported hypertension, diabetes, cardiovascular disease and chronic respiratory disease (the more NCDs an individual possesses the higher the predicted risk of inability to manage them), severe household food insecurity, no receipt of cash assistance, and household water insecurity ( Table 2). The final model had a moderate discriminative ability with an optimized adjusted C-statistic of 0.65 (95% CI: 0.62 -0.68) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022. ; https://doi.org/10.1101/2022.04.12.22273786 doi: medRxiv preprint (Table 2 and Figure 3a). The calibration plot without correction for optimism is shown in Figure 3b and the optimism adjusted calibration plot is shown in Figure 3a which depicts the expected performance of the final model in future samples. 23,27 The calibration slope after adjustment for optimism was 0.88 (95% CI: 0.73 -1.03) ( Table 2 and Figure 3a). The odds ratios and coefficients of the final model have been adjusted for overfitting and are presented in Table 2.
All the included predictors had the expected direction coefficient with the outcome. In particular, receiving no cash assistance, having household water insecurity or severe food insecurity had positive coefficients, and hence increased the likelihood of older Syrian refugees being unable to manage their NCDs. Age had a small negative coefficient with inability to manage NCDs.
To illustrate, the predicted risk of inability to manage NCDs for a Syrian refugee aged 60 years and over, with all four NCDs, severe household food insecurity, no receipt of cash assistance and household water insecurity was 51% (using the optimized-adjusted coefficients). Whilst the predicted risk of inability to manage NCDs for a Syrian refugee aged 60 years old with all four NCDs, who has received cash assistance and had no food insecurity and no household water insecurity was 19% (using the optimizedadjusted coefficients). For a Syrian refugee aged 60 years or old who had hypertension and diabetes without other NCDs the predicted risk was 9%.
In a sensitivity analysis, total number of chronic diseases was modelled as a categorical variable rather than each individual NCD. The predictors of this alternative model are: the number of chronic diseases, food insecurity and non-receipt of cash assistance. However, the discrimination (C-Statistics: 0.63 (95% CI: 0.60-0.66)) and calibration (C-Slope 0.86 (95%CI: 0.71-1.06)) of the optimized-adjusted model were poorer than the final model (Supplementary table 1).
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022.

Discussion
This study has identified predictors of inability to manage NCDs among older Syrian refugees in Lebanon during the COVID-19 pandemic. Younger age, not receiving cash assistance, household water insecurity, severe household food insecurity and co-morbidity were predictors of inability to manage NCDs. One of the key barriers for older Syrian refugees to adhere to medication was cost of the medication and beliefs that they did not require the medication anymore after feeling better. In addition, for older Syrians who were unable to access primary healthcare, the cost of the visit, tests or medications was the main reason.

Results in context
Our study showed that Syrian refugees who did not receive cash assistance had a higher likelihood of being unable to manage their NCDs. This finding concurs with a study conducted in Lebanon among Syrian refugees showing that receiving multipurpose cash assistance (MPC) lead to an increased access to primary health care for various illnesses, including chronic diseases. 28 The majority of Syrian refugees rely on cash assistance to cover their basic needs such as food, rent, water and healthcare. 29 Hence, one possible explanation is that without cash assistance resources are limited to the acute and immediate livelihood needs rather than managing a chronic disease. According to UNHCR, medications for chronic diseases are provided with a small handling fee to Syrian refugees in Lebanon. However, those with conditions requiring long term, high cost treatments, mainly related to chronic diseases, are not covered, which is a main gap to be addressed in the humanitarian response. 12 Furthermore, our results showed that affordability was the primary reason for not taking chronic disease medication or attending primary healthcare. Similarly, studies performed in Lebanon and Jordan among Syrian refugees showed that the inability to afford cost of the treatment and medications primarily prevent seeking health care for chronic disease. 13,30 It remains crucial for humanitarian agencies to remove the financial barriers for older Syrian populations in accessing required healthcare and essential chronic disease medication.
Water insecurity and severe food insecurity among Syrian refugees are markers of severe economic vulnerability and low socioeconomic status. 31-33 As a result, households that are unable to meet their basic needs for food and water are unlikely to be able to use a proportion of their household income to pay for subsidized NCD medication. Furthermore, food insecurity has been shown to reduce overall dietary quality and diversity. 34-36 Thus, food insecurity prevents refugees from modifying and improving their diet quality to manage their chronic disease.
Multimorbidity is common in older adults and it has implications on the self-management of these conditions. 37 Our results are consistent with previous research, which showed that older adults with multimorbidity had reduced adherence to the management of their NCDs. 38 Polypharmacy and complex . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022. ; https://doi.org/10.1101/2022.04.12.22273786 doi: medRxiv preprint dosing regimens are common in older age, 39 these reduce compliance 38 as each medication may have its own special instructions to follow, 40 which can be increasingly difficult in older age.
Although there was a statistical association between younger age and self-management of NCDs, the magnitude of the estimand was small. Previous studies have shown that younger age was associated with lower medication adherence for depression 41 and heart failure 42 . While other studies have suggested that there is no association between age and chronic medication adherence.
There is growing recognition that management of NCDs among refugee populations represents a challenge for humanitarian agencies, as these are costly to manage with limited resources available for healthcare 13

Strengths and Limitations
The present study has enhanced our understanding of the predictors of inability to manage NCDs among older Syrian refugees in Lebanon and fills a major gap in the international literature . Furthermore, this study is one of the largest on older Syrian refugees in the published literature with a high response rate of >85% among those who were eligible. 11 The study was limited as the predictive model had a moderate discriminative ability, which may be explained by missing predictors, such as perception of medication regime as being complicated, not knowing the purpose of the medication, accessibility issues, side effects due to medication, lack of healthcare support, lack of trust in doctors, stressful living conditions and time since diagnosis 15,30,44,45 . Furthermore, the calibration of the model showed overfitting and may not perform well in future samples, hence, future studies should aim to be larger if they wish to develop a predictive model. Another limitation was that data were self-reported to data collectors so misclassification in the data is possible; however, we tried to limit this through data quality and consistency checks.
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022. ;

Conclusions
This study highlights that inability to manage NCDs was mainly related to financial barriers. The predictors from this study will allow healthcare professionals and humanitarian organizations to identify older refugees who are at a greater risk of being unable to manage their NCDs. These vulnerable groups should have the necessary assistance and support to allow an improvement in medication adherence and equitable access to healthcare. Furthermore, investment in NCD healthcare services in primary care will be beneficial to the prevention of premature mortality and morbidity from NCDs in Lebanon and elsewhere.

Acknowledgements
We thank Lara Abou Ammar, Nadine Rashidi, Zeina El Khoury, Zeinab Ramadan and Stephanie Bassil for their assistance in the study. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022. ; . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022. ;

37.
Barnett . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022. ; https://doi.org/10.1101/2022.04.12.22273786 doi: medRxiv preprint Table 1  is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted April 17, 2022.