Food insecurity is associated with greater difficulty accessing care among people living with HIV with or without comorbid non-communicable diseases in western Kenya Authors: Mariam Ardehali, Catherine Kafu, Manuel Vazquez Sanchez, Marta Wilson-Barthes, Ben Mosong, Sonak D Pastakia, Jamil Said, Dan N Tran, Juddy Wachira, Becky Genberg, Omar Galarraga, Rajesh Vedanthan Published: 2024 ABSTRACT AbstractIntroductionThe relationship between food insecurity and access to healthcare in low-resource settings remains unclear. Some studies find that food insecurity is a barrier to accessing care, while others report that food insecurity is associated with a greater need for care, leading to more care utilisation. We use data from the Harambee study in western Kenya to assess the association between food insecurity and difficulty accessing care among people living with HIV (PLWH) with or without comorbid non-communicable diseases (NCDs).MethodsThe Harambee study is a cluster randomised trial that tested the effectiveness of delivering integrated HIV and NCD care for PLWH. In this cross-sectional analysis, we examined baseline data from Harambee participants to investigate the relationship between household food insecurity and difficulty accessing care, using multivariable logistic regression models, controlling for sociodemographic factors and care satisfaction. We tested for effect measure modification by gender and household wealth and stratified analyses by NCD status.ResultsAmong 1039 participants, 11.1% reported difficulty accessing care, and 18.9% and 51.9% of participants had moderate and severe food insecurity, respectively. Among those with difficulty accessing care, 73.9% cited transportation issues as the major barrier. Difficulty accessing care was greater with higher levels of food insecurity: among participants with low, moderate and severe food insecurity, 5.9%, 9.7% and 14.4% reported difficulty accessing care, respectively. After adjusting for confounders, severe food insecurity was independently associated with difficulty accessing care (adjusted OR=2.5, 95% CI 1.4 to 4.4). There was no statistical evidence for effect measure modification by gender or wealth.ConclusionsWe found that greater food insecurity was associated with greater difficulty accessing care among PLWH with or without NCDs in rural western Kenya. These findings suggest that addressing social determinants of health may be necessary when implementing integrated HIV and NCD care programmes. WHAT IS ALREADY KNOWN ON THIS TOPIC There is limited and conflicting data on the relationship between food insecurity and access to healthcare in low resource settings, especially among people living with HIV (PLWH) with or without non-communicable diseases (NCDs). There is limited and conflicting data on the relationship between food insecurity and access to healthcare in low resource settings, especially among people living with HIV (PLWH) with or without non-communicable diseases (NCDs). WHAT THIS STUDY ADDS We investigated the nature of this relationship through multivariable analysis of baseline surveys from participants in the Harambee randomised control trial, who are living with HIV with or without comorbid NCDs in rural, western Kenya. Our results found that increasing food insecurity was associated with greater difficulty accessing care among PLWH with or without NCDs in western Kenya, regardless of wealth, gender or NCD status. We investigated the nature of this relationship through multivariable analysis of baseline surveys from participants in the Harambee randomised control trial, who are living with HIV with or without comorbid NCDs in rural, western Kenya. Our results found that increasing food insecurity was associated with greater difficulty accessing care among PLWH with or without NCDs in western Kenya, regardless of wealth, gender or NCD status. HOW THIS STUDY MAY AFFECT RESEARCH, PRACTICE OR POLICY We anticipate that our findings will inform future integrated HIV and NCD care delivery programmes to better integrate social determinants of health. We anticipate that our findings will inform future integrated HIV and NCD care delivery programmes to better integrate social determinants of health. Introduction Inconsistent access to healthcare remains a significant challenge worldwide, which contributes to excess global mortality.1 Access to healthcare is influenced by numerous multilevel factors, including distance to health facilities, transportation challenges, socioeconomic status and patients’ educational status.2 Disparities in social determinants of health (eg, reduced access to education, lack of income-generating opportunities, lack of social support and transportation barriers) may hinder access to longitudinal healthcare. Long-term, routine engagement in care is particularly important for people living with HIV (PLWH), non-communicable diseases (NCDs) and other chronic conditions,3 and inconsistent utilisation of care can lead to poor health outcomes and greater overall costs.410 Food insecurity, defined as inconsistent access to an adequate supply of food,11 is a social determinant of health that may influence routine engagement in healthcare. Food insecurity is highly prevalent internationally, with nearly one-third of the global population, approximately 2.4 billion people, living without adequate food access in 2020.12 The published literature examining the relationship between food insecurity and access to, and utilisation of, healthcare has generated conflicting results. Some studies report that food insecurity is associated with greater healthcare utilisation (indicating greater health precariousness), as those with food insecurity are more likely to experience chronic health conditions and need more frequent care.13 Other studies, in contrast, report that food insecurity is associated with decreased access to care, as those experiencing food insecurity are more likely to have a lower socioeconomic status (indicating greater economic vulnerability) and face more obstacles to receiving care.1417 The nature of this relationship in low-resource settings for PLWH with or without NCDs is not well characterised, especially when accounting for factors that may be associated with both food insecurity and accessing healthcare, such as socioeconomic status, gender, and occupation.1820 Our goal, therefore, was to investigate whether food insecurity is associated with difficulty accessing care among PLWH with or without NCDs in western Kenya. We hypothesised that food insecurity in this setting would be associated with greater difficulty accessing care due to fewer resources among those with food insecurity. We conducted a cross-sectional analysis of baseline data collected from participants in the Harambee study to assess this relationship and test this hypothesis. These findings can inform the design of integrated HIV and NCD care models that better address social determinants of health. Methods Study setting The Harambee study employed a cluster randomised trial to evaluate the effectiveness, mechanisms and cost-effectiveness of using community-based microfinance groups as a platform for integrated care delivery among PLWH with or without NCDs in rural, low-resource communities in western Kenya.21 The trial was implemented within the infrastructure established by the Academic Model Providing Access to Healthcare (AMPATH), a global health partnership between Moi University College of Health Sciences, Moi Teaching and Referral Hospital and a consortium of North American and European universities, which aims to strengthen health systems, mitigate health disparities and improve healthcare access globally through academic and community partnerships.22 Over 110 000 PLWH are currently receiving clinical care from AMPATH’s HIV programme.23 AMPATH’s approach to HIV care is designed to provide comprehensive services which include, but are not limited to, provision of clinical care, laboratory monitoring, medication distribution and social determinants of health interventions to respond to the needs of vulnerable populations in the community.23 In addition, AMPATH has established a chronic disease management programme, in partnership with the Kenyan Ministry of Health, for management of hypertension and diabetes.24 The multicomponent package includes task redistribution,25 clinical decision support using health information technology,26 consistent and secure medication supply,27 linkage and retention activities,28 community stakeholder engagement,29 and social support including MF groups. Participants and data source Full details of the Harambee study protocol and procedures are described elsewhere.21 Eligible participants were at least 18 years of age and living with HIV, actively receiving HIV care through AMPATH, on antiretroviral therapy for at least 6 months and provided informed consent.21 A total of 1115 individuals were enrolled in the Harambee study. Baseline surveys were administered from 5 March 2021 to 24 February 2022, and captured self-reported information about demographics, health insurance, depression and anxiety via the Patient Health Questionnaire-4 (PHQ-4),30 barriers to accessing care, satisfaction with care, food insecurity via an adapted version of the Household Food Insecurity Access Scale (HFIAS),31 household economic status, employment and income. Full data for variables of interest were available for 1039 participants, and the analyses presented here are restricted to these participants. The Harambee study was approved by the Moi University and Moi Teaching and Referral Hospital Institutional Research and Ethics Committee (IREC Approval #0003054) and by Brown University through an IRB Authorisation Agreement (IAA #18-90). Patient and public involvement The Harambee study involved community members in the both the design of the trial as well as the recruitment of participants. First, microfinance group mapping exercises were conducted prior to the trial, which aimed to determine whether community visits were logistically feasible, to understand if microfinance group members would be interested in participating in this type of study and to address concerns from group members and the local health ministry leadership prior to trial start.32 In addition, the trial utilised a community-engaged randomisation approach during recruitment to increase transparency and cultivate trust between participants and the study team. This involved public randomisation ceremonies with the presence of participant representatives, directors of the health facilities and Harambee research staff. Finally, results from the trial will be presented to county-level ministers of health, community members and participants with time for discussion and joint planning for next steps. Primary-dependent variable The primary-dependent variable in this analysis was difficulty in accessing care during the 6 months prior to survey completion. Participants were asked ‘In the past 6 months, has it ever been hard for you to receive care for HIV?’ Primary explanatory variable The primary explanatory variable in this analysis was food insecurity, assessed using three questions from the original nine-item HFIAS. The HFIAS questionnaire includes questions encompassing three domains of food insecurity: anxiety and uncertainty, insufficient quality of food (including food preferences, social acceptability and variety) and insufficient food intake and its physical consequences.31 We followed the analytic approach outlined in a previously published article that used a three-item version of the HFIAS questionnaire.33 The values of the responses to the three questions were summed, and the sum was divided into three categories: little/no food insecurity corresponded to scores of 0–1, moderate food insecurity corresponded to scores of 2–3 and severe food insecurity corresponded to scores of 4–6. While the three items in our questionnaire were different than those in the previously published article, we felt that the analytic approach was appropriate. Covariates Covariates included patient-reported information on demographics, socioeconomic status and satisfaction with healthcare, all of which have been shown to have associations with healthcare access in the published literature.3437 Demographic covariates included participant age at study enrolment, gender (male or female), whether or not they were the head of the household and their household’s primary healthcare decision-maker (self, partner/spouse or other). Economic covariates captured participants’ primary form of employment (farmers, daily labourers, small business owners, artisans, salaried employees, unemployed/not working, students and retirees), their household wealth status and whether or not they have ever been enrolled in the Kenya National Health Insurance Fund. Household wealth status was assessed using a five-quintile wealth index following methods established by the Demographic and Health Survey.38 Data regarding participant satisfaction with healthcare were taken from the Patient Satisfaction Survey, which was adapted from existing literature.39 Questions asked how often, when receiving HIV care, the staff at the facility treated the patient with courtesy and respect, the doctors or health providers listened to them carefully, the doctors or health providers asked for their opinion regarding their healthcare and how often the patient was satisfied with the amount of time the doctor/provider spent with them. Response options for each question assessed the frequency with which the participant felt satisfied. Following prior methods, we created a continuous summary indicator for participant satisfaction.39 Statistical analysis Descriptive statistics of the study population included frequencies of care access categories and means and SD for age. Initial exploratory analyses included univariate tabulations of all study variables. Cross-tabulations of the primary explanatory variable and covariates with the primary-dependent variable were conducted to compare whether the distribution of these variables differed by level of difficulty accessing care. Statistically significant differences in these bivariate tables were determined using χ2 tests at the 0.05 level. Unadjusted ORs for difficulty accessing care were generated for the primary explanatory variable and all covariates using bivariate logistic regression analysis. Multivariable logistic regression analysis was then used to control for all hypothesised confounders. P-value for trend was used to assess whether there was a monotonic trend in the relationship between ordinal food insecurity categories and difficulty accessing care, as well as between ordinal wealth index categories and difficulty accessing care. Complete case analysis was utilised, omitting rows with missing values for dependent variables from analysis. Based on the existing literature regarding gender, wealth, food insecurity and access to care,4044 we hypothesised that there may exist effect measure modification by gender and wealth status. We therefore conducted stratified analyses to assess whether the relationship between food insecurity and access to care varied by gender and wealth index quintiles. We also created three-way and two-way interaction terms to test for statistical interaction in the adjusted analysis. The likelihood ratio test was performed a total of three times to compare interaction models to the original, fully adjusted multivariable model. We also conducted stratified analysis by NCD status to evaluate whether the relationship between food insecurity and access to care differed by whether the participants were living with HIV only or were living with HIV and hypertension and/or diabetes. All analyses were conducted using Stata (StataCorp. 2021. Stata Statistical Software: Release V.17. College Station, TX: StataCorp LLC). Results The average age of the 1039 participants at baseline was 50.7 years (SD: 11.2), and 75.2% of participants were women (table 1). There was no difference by age or gender between our study participants and those who enrolled but did not have complete baseline data. Over 70% of individuals had at least moderate food insecurity, with 51.9% reporting severe food insecurity. Regarding healthcare access, 11.1% of participants reported difficulty accessing care over the past 6 months. Of those who reported difficulty accessing care, 73.9% attributed difficulties to transportation issues, 1.7% attributed difficulties to competing obligations and 13.0% attributed difficulties to both (figure 1). Difficulty accessing care was higher among participants with greater food insecurity: among participants with low, moderate and severe food insecurity, 5.9%, 9.7% and 14.4% reported difficulty accessing care, respectively. Additionally, among the 676 participants who reported their NCD status, 23.2% had hypertension, 0.9% had type 2 diabetes and 0.4% had both. Table 1Distribution of baseline characteristics according to extent of difficulty accessing careNo difficulty accessing care (n=924)Difficulty accessing care (n=115)Age (years) Mean (SD) (50.7 (11.2))50.8 (11.2)49.4 (11.7)Gender Female (n=782)691 88.4%91 11.6% Male (n=257)233 90.7%24 9.3%Primary employment* Salaried employee (n=56)52 92.9%4 7.1% Non-salaried worker (n=207)174 84.1%33 15.9% Farmer (n=258)223 86.4%35 13.6% Other (n=45)42 93.3%3 6.7% Small business owner (duka, kiosk, trade in farm produce) (n=198)172 86.9%26 13.1% Unemployed/not working (n=267)254 95.1%13 4.9%Wealth index quintile 1 (n=208)179 86.1%29 13.9% 2 (n=207)180 87.0%27 13.0% 3 (n=208)184 88.5%24 11.5% 4 (n=208)186 89.4%22 10.6% 5 (n=208)195 93.8%13 6.3%National Health Insurance Fund (NHIF) Enrolment* No (n=744)655 88.0%89 12.0% Yes (n=294)268 91.2%26 8.9%Participant is head of household Yes (n=743)665 89.5%78 10.5% No (n=296)259 87.5%37 12.5%Household healthcare decision-maker Self (n=933)837 89.7%96 10.3% Other (n=106)87 82.1%19 17.9%Healthcare staff satisfaction All of the time (n=948)842 88.8%106 11.2% Some of the time (n=91)82 90.1%9 9.9%Provider satisfaction* Very satisfied (n=866)769 88.8%97 11.2% Somewhat satisfied (n=171)153 89.5%18 10.5%Trust in confidentiality Agree (n=981)875 89.2%106 10.8% Not fully agree (n=58)49 84.5%9 15.5%Equal treatment by healthcare staff Agree (n=866)786 90.8%80 9.2% Not fully agree (n=173)138 79.8%35 20.2%Overall satisfaction with care* Very satisfied (n=719)645 89.7%74 10.3% Not satisfied/moderately satisfied (n=317)276 87.1%41 12.9%Food insecurity category Little/no household food insecurity (n=304)286 94.1%18 5.9% Moderate household food insecurity (n=196)177 90.3%19 9.7% Severe household food insecurity (n=539)461 85.5%78 14.5%Absolute numbers and row percentages reported, except for age where mean and SD are presented.*Values may not add to total due to missing values. Distribution of baseline characteristics according to extent of difficulty accessing care Absolute numbers and row percentages reported, except for age where mean and SD are presented. Values may not add to total due to missing values. Reasons for difficulty accessing care among participants. Crude and adjusted ORs are shown in table 2. After adjusting for potential confounders, severe food insecurity was independently associated with difficulty accessing care (adjusted OR (aOR)=2.5, 95% CI=1.4 to 4.4). Increasing severity of food insecurity was significantly associated with greater difficulty accessing healthcare (Ptrend=0.001) (figure 2). Additionally, participants living with HIV who reported not being treated equally to patients without HIV were significantly more likely to report difficulty accessing care (aOR=2.5, 95% CI=1.4 to 3.6). Table 2Unadjusted and adjusted ORs for variables associated with difficulty accessing care, with respective CIsVariableUnadjustedOR (95% CI) n=1039AdjustedOR (95% CI) n=1039Food insecurity category Little/no household food insecurityReferenceReference Moderate household food insecurity1.7 (0.9, 3.3)1.6 (0.8, 3.3) Severe household food insecurity2.7 (1.6, 4.6)2.5 (1.4, 4.4)P trend: 0.001Gender FemaleReferenceReference Male0.8 (0.5, 1.3)1.0 (0.6, 1.6)Wealth index quintile Quintile 1ReferenceReference Quintile 20.9 (0.5, 1.6)1.0 (0.6, 1.9) Quintile 30.8 (0.5, 1.4)0.9 (0.5, 1.6) Quintile 40.7 (0.4, 1.3)0.8 (0.4, 1.6) Quintile 50.4 (0.2, 0.8)0.5 (0.2, 1.1)P trend: 0.08Primary employment Salaried employeeReferenceReferenceFarmer2.0 (0.7, 6.0)2.1 (0.7, 6.6) Small business owner2.0 (0.7, 5.9)1.9 (0.6, 6.1) Unemployed/not working0.7 (0.2, 2.1)0.6 (0.2, 1.9) Non-salaried worker2.5 (0.8, 7.3)2.0 (0.6, 6.0) Other0.9 (0.2, 4.4)0.9 (0.2, 4.6)National Health Insurance Fund Enrolment YesReferenceReference No0.7 (0.5, 1.1)0.9 (0.6, 1.5)Participant is head of household YesReferenceReference No0.8 (0.5, 1.2)0.9 (0.5, 1.6)Household healthcare decision maker SelfReferenceReference Other1.9 (1.1, 3.3)1.7 (0.9, 3.2)Healthcare staff satisfaction All of the timeReferenceReference Some of the time0.9 (0.4, 1.8)0.9 (0.4, 2.2)Provider satisfaction Very satisfiedReferenceReference Somewhat satisfied0.9 (0.5, 1.6)0.6 (0.3, 1.2)Trust in confidentiality AgreeReferenceReference Not fully agree1.5 (0.7, 3.2)1.4 (0.6, 3.1)Equal treatment by healthcare staff AgreeReferenceReference Not fully agree2.5 (1.6, 3.9)2.2 (1.4, 3.6)Overall satisfaction with care Very satisfiedReferenceReference Not satisfied/moderately satisfied0.8 (0.5, 1.2)0.9 (0.6, 1.4) Unadjusted and adjusted ORs for variables associated with difficulty accessing care, with respective CIs Adjusted ORs (aORs) of food insecurity and difficulty accessing care. AORs of food insecurity and difficulty accessing care among those with moderate and severe food insecurity, respectively, with associated p values. Trend across the food insecurity categories is illustrated in the dashed blue line. The reference value is ‘little/no food insecurity’. Gender and wealth index were tested as effect measure modifiers of the relationship between food insecurity and difficulty accessing care, and stratified adjusted ORs are shown in online supplemental tables 1 and 2. While point estimates indicate the potential for quantitative effect measure modification between food insecurity and difficulty accessing care by gender in particular, the tests for interaction were not statistically significant (p=0.3 for gender and p=0.8 for wealth index quintile). The analyses stratified by NCD status showed no apparent difference in the relationship between food insecurity and access to care among those with an NCD compared those without an NCD (online supplemental table 3). Formal interaction testing for NCD status was not performed due to small sample sizes. Discussion This cross-sectional analysis of baseline data from the Harambee study in western Kenya demonstrates that more severe food insecurity is associated with greater difficulty accessing healthcare among PLWH with or without comorbid NCDs. This relationship did not vary by gender, wealth or NCD status. A majority of participants reported at least moderate or severe household food insecurity at study enrolment. Among those who expressed difficulty accessing care, the vast majority cited transportation issues as the primary barrier to care access. In light of the varied results from the literature regarding whether economic vulnerability (greater difficulty accessing care) or health precariousness (greater healthcare utilisation) of food insecurity is more important, our findings suggest that, for PLWH with or without NCDs who are eligible to receive integrated care in western Kenya, the economic vulnerability associated with food insecurity may be a key barrier to accessing care. This may be due to the myriad of obstacles that PLWH confront in low-resource settings, such as lack of access to transportation, high medication costs and the opportunity cost of procuring medications from the health facility, which can force patients into choosing between buying either food or medication (treat or eat).45 Previous literature has shown that food insecurity is linked with decreased adherence to antiretroviral therapy for PLWH,46 47 and similar healthcare access barriers have been shown for people living with NCDs.48 These barriers to treatment may be more considerable for people living in low-resource settings.4951 These findings add to the existing analyses exploring the relationship between food insecurity and access to care, particularly among PLWH with or without NCDs. Taken together with the prior literature, our findings strengthen the evidence base that addressing food insecurity and other poverty determinants have the potential to improve care access. Difficulty accessing care was reported by 11.1% of our cohort, which is within the range of~5%–20% reported by other studies from various countries worldwide, including other African nations.5254 There is no uniform definition of access to healthcare, accounting for the variability reported among the different studies.55 56 A consensus definition of this important issue is needed and worth pursuing by multilateral global agencies and research consortia. Notably, a majority of those in our study experienced some form of food insecurity, which is higher than the worldwide rate of food insecurity, but consistent with that of Kenya and Africa as a whole.57 Over 50% of Harambee participants experienced severe food insecurity, pointing to the critical need for initiatives to increase consistent food access and improve nutrition for communities in rural Kenya, particularly among PLWH with or without NCDs. Greater food insecurity has been shown to have an association with lower CD4 counts, higher morbidity and numerous other negative health outcomes.5 18 20 47 Since over 75% of rural Kenyans rely on agriculture as their primary source of food,58 improving agricultural infrastructure and collaborating with local farmers and markets could be a feasible approach to improving food security.5860 There is a growing body of literature regarding integrated HIV and NCD care.921 6164 Our findings demonstrate that it is vital to consider social determinants of health when delivering integrated HIV and NCD care in LMICs. By addressing food insecurity and other social determinants of health for similar populations, a greater number of individuals may be able to access care and experience improved health outcomes. This may increase the impact of programmes delivering integrated HIV and NCD care, potentially saving money, time and resources.616567 In addition, transportation posed a major barrier for our cohort as reported by an overwhelming majority (86.9%) of the participants. This is consistent with findings from existing studies investigating barriers to care access in western Kenya, LMICs, and globally, where transportation challenges remain a significant issue.6874 This highlights the need for future care implementation programmes to incorporate transportation challenges faced by patients when aiming to expand care access. Our results also indicated that the perception that PLWH were not treated equally to patients with non-HIV illnesses by healthcare staff was significantly associated with greater difficulty accessing care. This finding is consistent with the existing literature that shows a relationship between perceived discrimination and decreased healthcare access.75 This may be due to a variety of reasons, including a true lower quality of care received by the patient and greater vigilance and distrust of the healthcare system among certain individuals, both of which would make the patient less likely to engage with the healthcare system or follow provider recommendations.76 Future studies may explore this association to better inform integrated care models. Although no definitive evidence was found for effect measure modification by gender and wealth index, the stratified results indicate the potential of a relationship of differential magnitude by these variables. Previous studies have shown that gender is associated with varying levels of food insecurity and access to care, with women utilising outpatient health services more frequently than men and women being more frequently food insecure than men.4042 Since gender has an independent relationship with each of these variables, there is potential that it could affect the relationship between food insecurity and access to care. This is the same case for the effects of wealth on food insecurity and on healthcare access, with lower wealth being associated with decreased care access and more severe food insecurity.43 44 While, to the best of our knowledge, there is not a large body of literature that looks at these relationships directly,77 future studies can specifically investigate whether there is effect measure modification of gender or wealth on food insecurity and access to care. As reported above, our stratified analysis by NCD status demonstrated no apparent effect modification, indicating that those with comorbid NCDs had similar relationships between food insecurity and access to care as those participants with only HIV. Additional research is warranted to determine if this relationship varies by subgroups. We acknowledge the following limitations in this study. First, we used a self-reported measure of difficulty in accessing care, since there is no uniform definition of care access.55 56 There is therefore the potential for under-reporting decreased care access due to possible stigma associated with admitting inconsistent healthcare utilisation.78 Second, although we approached our analysis of food insecurity based on a scale used in previous literature,33 the original study utilised a different set of three questions from the HFIAS than we used in our questionnaire. Thus, whether our adopted coding scheme is fully appropriate and validated for the questions used in our study is not certain.31 Additionally, our study participants are composed of people who were highly engaged in HIV care at study enrolment, which may represent a group that has higher-than-average healthcare access. Fourth, we acknowledge that our analysis was restricted to participants with complete data for the variables of interest. However, our participants were representative of the overall cohort of individuals enrolled in the Harambee study, reflected by a similar gender and age distribution. Finally, 75% of participants were already members of microfinance groups at enrolment,21 offering greater access to financial liquidity than comparable populations,7983 which potentially reduces generalizability. Conclusion Food insecurity is associated with decreased access to healthcare among PLWH with or without comorbid NCDs in western Kenya. This study contributes to the existing literature regarding the relationship between food insecurity and healthcare access in LMICs, particularly among PLWH with or without NCDs. Given the independent and strong association of food insecurity with access to care, our findings underscore the importance of taking into account social determinants of health while designing and implementing integrated HIV and NCD care delivery for populations in low-resource settings. We anticipate that our findings will help to inform future interventions to improve health equity in low-resource settings worldwide. Supplementary material