Address Changes Are Associated With Unmet Glycemic Targets: Opportunities to Improve Processes and Outcomes of Care Among People With Type 2 Diabetes

Housing security is a continuum. On one end is no access to reasonable quality housing (eg, homelessness). At the other end is access to reasonable quality housing in the absence of threats (ie, housing stability). ¹ Along this continuum is housing insecurity, which has been defined by the US Department of Health and Human Services as insecurity of shelter and includes threats to housing stability such as frequent moves, high housing cost to income ratio, poor quality housing, overcrowding, unstable neighborhoods, or homelessness. ² Homelessness and housing insecurity are known social determinants of health (SDOH) and are linked to poorer health outcomes and increased health care costs. ^3–6 As an important SDOH, health care systems should consider strategies to address housing insecurity and homelessness among patients to improve the health of their populations. A first step of this effort is to accurately identify patients who are, or are likely of, being housing insecure by assessing threats to housing stability. This identification could lead to more efficient targeting of screening to confirm housing status and interventions to address housing insecurity, mitigate its effects, and prevent homelessness. ⁷

One important aspect of housing insecurity is moving or residential mobility. Residential mobility at its core is defined as the movement of households (ie, address changes). ⁸ Residential mobility has been variably defined across the literature using descriptors such as frequency of moves, time since move, distance moved, reason for move, and characteristics of neighborhood moved to or from. ⁹ Residential mobility has been associated with negative mental health outcomes and less healthy behaviors (eg, higher incidences of smoking and alcohol onset). However, the important effects of residential mobility have been investigated more thoroughly in children compared to adults. In children, residential mobility has been associated with accelerated initiation of illicit drug use, higher levels of behavioral and emotional problems, reduced continuity of health care, increased teenage pregnancy rates, and adolescent depression. ⁹ However, the effects of residential mobility on health outcomes for adults is an important but under-studied area of research. ¹⁰ Therefore, it is important to understand if residential mobility, as defined by address changes, is associated with poorer health outcomes among adults with diabetes.

Several studies have attempted to comprehensively identify the housing insecure or homeless population utilizing electronic health record (EHR) data. Attempts to compile comprehensive lists of likely homeless addresses or shelters, development of programming to analyze free text in EHR data for evidence of homelessness, or attempts to quantify the number and types of addresses related to homelessness in EHR data have been studied but gaps remain. ^11–13 Most of the previous research did aim to identify homeless individuals, but did not focus on those who were housing insecure. Additionally, many of these studies were conducted with Veteran populations which may not be generalizable to the general population. To fill these gaps, our study sought to identify potentially housing insecure patients within an integrated health system that is similar both demographically and socioeconomically to the populations that we serve.

Prior research has identified numerous challenges to effective diabetes management for persons experiencing homelessness or housing insecurity. For example, persons experiencing homelessness or housing insecurity are more likely to have anxiety about chronic disease management, have increased barriers to treatment, and indicate worse health-related quality of life. ^14,15 Potential barriers to effective chronic disease management for this population include lack of safe spaces for storing medications, absence of routine for medication adherence, and difficulty prioritizing disease management. ¹⁶ Research has also found that secure housing is indicative of lower health care costs and more preventive diabetes treatment and patients with diabetes who received supportive housing are more likely to have scheduled outpatient visits with increased hemoglobin A1C (A1C) and lipid testing and prescriptions. ^17–19

Research highlights the complexity and heterogeneity underlying housing insecurity and homelessness as well as in housing insecure and homeless populations. ²⁰ Considering the importance of housing insecurity as an SDOH, particularly among patients with diabetes, it is important for physicians and health care systems to operationalize EHR-based metrics that can be used to identify patients who may be at risk. Therefore, the objectives of this study were to identify and operationalize measures of housing insecurity, with a focus on residential mobility, using existing EHR data within a large, integrated health system and to quantify the association between address changes and glycemic and other diabetes management goals.

We identified a cross sectional cohort of 271,277 adults (≥18 years) with type 2 diabetes in 2018. Patient sample characteristics are reported in Table 1. Approximately, 6 percent (N= 17,317) of the cohort had at least one address change. Women were 48% of the cohort. Approximately 80% of the cohort was over 50 years old with a mean age of 63–43 years. The cohort was 21.8% Hispanic, 41.9% White, 24.4% Asian/Native Hawaiian/Pacific Islander/Native American, and 9.8% Black. Patients with address changes were younger and more likely to be Black, have Medicaid insurance, and have a diagnosis of depression, substance abuse, or severe mental illness.

In univariate analyses, we found that one or more address changes was associated with increased prevalence of not meeting glycemic targets (A1C < 9) (24.2% vs 19.3%, p = 0.001) and decreased prevalence of meeting glycemic targets (A1C < 8) (63.4% vs 69.1%, p = 0.001). We also found that that one or more address changes were associated with increased prevalence of emergency department visits (40.1% vs 27.5%, p = 0.001), decreased prevalence of blood pressure control (78.7 vs 80.3, p = 0.001), and decreased prevalence of having a flu shot (57.0% vs 63.5%, p = 0.001) (Table 1). These relationships remained significant after adjustment via logistic regression, with address change associated with greater chance of A1C < 9 (ARR: 1.12, 95% CI: 1.09, 1.15), lower chance of A1C < 8 (ARR: 0.95, 95% CI; 0.94, 0.96), lower chance of controlled blood pressure (ARR:0.99, 95% CI: 0.98, 0.99) greater chance of emergency department visits (ARR: 1.25, 95% CI: 1.23, 1.27), and lower chance of having a flu shot (ARR: 0.94, 95% CI: 0.93, 0.95) when compared to no address change (Table 2).

We also examined whether the relationship between address change and our 5 outcomes (A1C < 8, A1C < 9, receipt of flu shot, ED visit and BP control) differed based on patient age ( ≤ 50 or < 50 years old) or patient race/ethnicity (white, Black, Asian, and Hispanic) (Table 3). We found similar relationships between address change and all outcomes expect blood pressure control in patients under age 50 and those over age 50. Patients with diabetes under age 50 with an address change were more likely to have controlled blood pressure compared to those without an address change, but patients with diabetes over age 50 with an address change were less likely to be in control of their blood pressure compared to those without an address change. Regarding race/ethnic differences, we found similar results by patient race/ethnicity for A1C < 9 and emergency department visits but we found that for Black patients with diabetes, there were no significant differences for A1C < 8 and flu shot receipt based on having an address change or not. For Hispanic and Asian patients with diabetes, there were no significant differences for blood pressure control based on address change or not.

In a large cohort of patients with type 2 diabetes within an integrated health system, we found that patients with type 2 diabetes and at least one address change were more likely to not meet glycemic targets. These patients were also more likely to have an emergency department visit, less likely to have a flu shot, and less likely to have well controlled blood pressure than those with no address change. We also found for patients with diabetes 50 years or older, an address change was associated with increased risk for uncontrolled blood pressure. These results suggest that housing insecurity is associated with not meeting glycemic and blood pressure targets, greater emergency department utilization, and less use of preventive services in patients with diabetes.

Our study examining the relationship between residential mobility, as a potential proxy for housing insecurity, and diabetes management targets aligns with previous research highlighting challenges to effective diabetes management among housing insecure and homeless populations. Prior research among homeless veterans with diabetes found that homelessness was a risk factor for increased A1C levels, with increasing indicators for homelessness associated with increased risk for not meeting glycemic targets. ²⁴ Berkowitz et al found that unstable housing was associated with greater odds of diabetes-related emergency department visits and hospitalization. ¹⁷ Vijayaraghavan et al found that housing insecurity led to decreased self-efficacy in diabetes management. ²⁵ Berkowitz et al also found that the number of unmet material needs is related to not meeting glycemic targets and housing instability is associated with increased 30-day readmission in unadjusted results. ¹⁸ Taken together, our studies suggest that both housing insecurity and residential mobility may lead to poorer diabetes-related outcomes. Our study fills an important gap in the literature by utilizing address changes in EHR data to identify potentially housing insecure patients within a large, integrated health system as well as important subgroups, such as adults 50 years and older, that should be further assessed. By assessing residential mobility via address changes, we expand the literature on housing factors that negatively affect diabetes-related outcomes and discern one easily identifiable indicator which could be used to target SDOH screening efforts. Understanding the effects of housing insecurity on effective diabetes management, particularly as related to the receipt of preventive services such as a flu shot, is especially important during the coronavirus 2019 (COVID-19) pandemic. Identifying and addressing housing insecurity is especially relevant as housing insecurity and its associated health risks have been potentially amplified due to the COVID-19 pandemic. The COVID-19 pandemic has led to an increase in unemployment, which could result in increases in housing insecurity and homelessness. ²⁶ For those already experiencing homelessness, the ability to follow COVID-19 social distancing and quarantining orders is difficult. Research has already found that homeless individuals who acquired COVID-19 are more likely to require critical care and 2 to 3 times more likely to die compared to the general population. ²⁶ Furthermore, hyperglycemia is both a risk factor for COVID-19 infection and has been associated with poorer outcomes for those who have acquired COVID-19, making the effective management of type 2 diabetes for homeless and housing insecure individuals particularly important. ^27–29

There are also well-documented race and ethnic disparities in housing insecurity, homelessness, and diabetes. Nationally representative data from the Centers for Disease Control show that non-White people have nearly twice the rates of type 2 diabetes as White people, with Native American/Alaskan Native people having the highest age-adjusted prevalence of diabetes at 15.9% compared to 13.2% of non-Hispanic Black people, 12.8% for Hispanic people, 9.0% for Asian American people and 7.6% for non-Hispanic White people. ³⁰ Regarding housing insecurity and homelessness, a higher proportion of low-income, Black, Hispanic and Native American households live in homes with inadequate conditions and are rent-burdened compared with higher income households and White people. ³¹ Research has also identified dramatic disparities in lifetime prevalence of homelessness with non-Hispanic Black people and Hispanic people of any race having higher rates of homelessness at 16.8% and 8.1% respectively compared to non-Hispanic White people at 4.8%. ³²

Additionally, there are also important age-related factors that affect housing insecurity and homelessness. Over the past 2 decades, the average age of single adults experiencing homelessness has increased and half of single homeless adults are now aged 50 and older. ³³ Furthermore, a larger proportion of older homeless adults become homeless for the first time in late middle age and these newly homeless older adults are generally low-income adults who have sustained a health or financial crisis. Older homeless adults (aged 50 or older) have unique health concerns compared to younger homeless adults, including many conditions more commonly seen in housed adults aged 70 or older such as memory loss, falls, and functional impairment. ³³ This makes understanding the effects of housing insecurity on diabetes management especially important for older adults.

This study has several limitations. First, 1 address change within 12 months may not signal housing insecurity. However, an address change may signal stress related to moving, changes in lifestyle behaviors, or geographic differences associated with adverse outcomes. For example, Baum et al utilized nationally integrated EHR data for a retrospective cohort study of patients receiving care at the Veterans Health Administration and found that exactly 1 move from a 10th to a 90th percentile zip code was associated with a greater prevalence of uncontrolled diabetes. ³⁴ This suggests that chronic disease outcomes such as diabetes control can be sensitive to even one address change. Secondly, since our study was conducted within an integrated health system, the availability of comprehensive address data outside of our health system or for other health systems may be variable as we only have address change data on patients who updated their address by engaging with our health care system. It is also important to note that our study was conducted in Northern California including the San Francisco Bay Area where the expensive housing market dramatically narrows the margin between housing insecurity and homelessness. ³⁵ This may limit the generalizability to other localities with more affordable housing. Lastly, 2 or more address changes occurred infrequently among our cohort (0.6%) indicating that more frequent moves within one year, which may be a more robust indicator of housing insecurity, were uncommon among our study population. Therefore, we were not able to assess differences among this group with frequent moves. Lastly, we cannot make causal inferences with our data.

Implications from our study extend beyond the Kaiser Permanente Northern California health system. Although Kaiser Permanente Northern California is an integrated health system, our strategy for identifying patient address changes is applicable to other health systems. Furthermore, one strength of our current study is the finding of the association between address change and suboptimal disease management in a sample of patients with diabetes with health insurance coverage highlighting the need for health system-level approaches to identifying potential housing insecurity. In their 2021 Standards of Care recommendations, the American Diabetes Association recommends assessing housing insecurity and homelessness as part of diabetes care and using that information to inform treatment decisions. ³⁶ Address changes may be an early warning sign of housing insecurity or stress that could be a used as a trigger for screening or interventions. Better knowledge of who might be at high chance of screening positive for housing insecurity could help health systems more efficiently utilize clinician and system resources. Therefore, interventions to screen patients with address changes for housing insecurity may be needed to better identify patients who may need additional support and could be used to inform treatment decisions. Identifying patients with housing insecurity and providing resources aimed at continuity of care and stable health care access could potentially mitigate the chance of uncontrolled A1C even when housing is insecure. Additional health system tools for social needs interventions, particularly related to housing insecurity, will become available as Kaiser Permanente Regions operationalize community-based resources through Thrive Local, a comprehensive social needs network that integrates community referrals to social service organizations and integrates referrals into the EHR.

Address changes obtained from the EHR represent an innovative strategy for identifying potential housing insecurity at a health care system level. By identifying the growing, vulnerable, housing insecure population and its relationship to glycemic control in patients with type 2 diabetes, we address a current gap in literature and attempt to quantify the need for increased resources and stable access to care for patients with diabetes experiencing housing insecurity.