Abstract Session A3: Health Disparities/Vulnerable Populations - - PDF document

Abstract Session A3: Health Disparities/Vulnerable Populations

Moderators: Gail Daumit, MD, MHS and Monica E. Peek, MD, MPH

**This session is one of two piloting a short abstract presentation style** ILLICIT BUPRENORPHINE USE, AND ACCESS TO AND INTEREST IN BUPRENORPHINE TREATMENT Aaron Fox1,2; Adam Chamberlain2; Taeko M. Frost3; Chinazo Cunningham1,2. 1Montefiore Medical Center, Bronx, NY; 2Albert Einstein College of Medicine, Bronx, NY; 3Washington Heights CORNER Project, New York, NY. (Tracking ID #1927780) BACKGROUND: In the United States, the opioid addiction epidemic is escalating; however, there is a large gap (nearly 1.5 million persons) between those in need of treatment and those who receive treatment. Primary care physicians have the opportunity to address this treatment gap by offering buprenorphine maintenance therapy (BMT), but access to treatment may not be adequate to meet the current demand. Recently, diversion of buprenorphine has received major media attention, where concerns were raised about illicit buprenorphine use to get high; however, qualitative studies have suggested that opioid users may use illicit buprenorphine to "self- treat" their opioid addiction, especially if they experience barriers to BMT. This study investigated illicit buprenorphine use among syringe exchange participants, a group with high needs for opioid addiction treatment, and explored whether illicit use was associated with access to BMT and interest in initiating BMT. METHODS: Syringe exchange participants were recruited from the offices of a harm reduction agency in New York City. Computer-based interviews were conducted to determine: 1) prior use of buprenorphine (illicit and prescribed); 2) access to BMT (perceived barriers); and 3) interest in BMT (overall interest in BMT and likelihood of initiating treatment). Overall interest was measured using a 5-point Likert scale; those rating their level of interest as 4 or 5 were considered to be interested in BMT. Access to and interest in BMT were compared between illicit buprenorphine users and non-users using chi square or t-tests. RESULTS: Of 102 opioid users, 57 had used illicit buprenorphine (34 with illicit buprenorphine use only; 23 with illicit and prescribed buprenorphine use). Nine participants had used prescribed buprenorphine only. Overall, 45% of participants were interested in BMT. Regarding access, the most common barrier to BMT was, "did not know where to get treatment," which was reported by 51% of participants. Other common barriers were costs (33%) and transportation (28%). Compared to those who had never used illicit buprenorphine, not knowing where to get treatment was more common among illicit buprenorphine users (64% vs. 36%, p < 0.01),

• verall interest in BMT was greater among illicit buprenorphine users (mean ± SD; 3.37 ± 1.29 vs. 2.80 ± 1.34,

p = 0.03), and more illicit buprenorphine users reported they would be likely to initiate BMT if it were easily accessible (82% vs. 50%, p < 0.01). CONCLUSIONS: Illicit buprenorphine use was common. A majority of illicit buprenorphine users were interested in BMT and reported that they would be likely to initiate treatment, but nearly two-thirds of illicit users did not know where to access BMT. Therefore, relatively simple interventions that address barriers to BMT (e.g. linking illicit buprenorphine users to practices that offer BMT or initiating BMT onsite at harm reduction agencies) could reduce illicit buprenorphine use, narrow the treatment gap, and diminish the tragic consequences of opioid addiction.

IMPACT OF HEALTH COACHING ON PATIENT TRUST IN THEIR PRIMARY CARE PROVIDER: A RANDOMIZED CONTROLLED TRIAL David Thom; Danielle Hessler; Rachel Willard-Grace; Thomas Bodenheimer; Adriana Najmabadi; Christina Araujo; Ellen Chen. UCSF, San Francisco, CA. (Tracking ID #1928972) BACKGROUND: In primary care, there has been a move to share tasks and responsibilities traditionally reserved for the primary care provider (PCP) with other members of the patient care team, including medical assistants, nurses, pharmacists, patent educators and coaches. Concern has been raised regarding the impact of the ‘team approach' on the quality of the patient-provider relationship. We analyzed data from a randomized controlled trial comparing health coaching to usual care to assess the impact of health coaching

• n patients' relationship with their primary care provider (PCP).

METHODS: Randomized controlled trial comparing health coaching with usual care. Participants were low-income English or Spanish speaking patients age 18 to 75 with poorly controlled type 2 diabetes, hypertension and/or hyperlipidemia. Health coaches were certified medical assistants who attended 40 hours of health coach training over six weeks using a curriculum developed by the study team that included instruction in using active listening and non-judgmental communication; helping with self-management skills including creation of action plans, and providing social and emotional support. Patient trust in their primary care provider measured by the 11-item Trust in Physician Scale, converted to a 0 to 100 scale. Patient satisfaction with their PCP was assessed by a single item, "How likely would you be to recommend your doctor to your friend or relative?" with a response scale from 1='definitely not recommend' to 5= ‘definitely recommend'. Data were analyzed using linear mixed modeling. P-values were two-tailed. RESULTS: A total of 441 patients were randomized to receive 12 months of health coaching (n=224) vs. usual care (n=217). At baseline, there were no significant differences in participant characteristics between the two study arms, including trust in their PCP (Table 1). At 12 months, trust and satisfaction were reported by 203 patients (91%) in the health coaching group and 175 of patients (81%) in the usual care group. Both the mean level of patients' trust in their PCP and the percent of patients who would definitely recommend their primary care provider to family or friends increased significantly more in patients receiving health coaching (Table 2). These differences remained significant after adjustment for number of PCP visits during the study. CONCLUSIONS: Health coaching does not appear to lower, and in fact may increase, patients' trust in their primary care providers. Clinicians should be reassured that working with health coaches does not appear to compromise, and may in fact enhance, their relationships with their patients. Table 1. Participant characteristics at enrollment by study arm (% or mean (sd))*

Characteristic Health coaching arm (n=224) Usual care arm (n=217) Age (years) 52.6 (10.7) 52.9 (11.5) Gender (female) 52% 59% Born in the US 26% 25% Years living in US** 18.5 (10.4) 17.9 (11.9) Spanish is primary language 68% 70% Race/Ethnicity: African American 20% 18% Latino or Hispanic 69% 71% White non-Hispanic/Asian/Other 11% 11% Education less than high school 44% 44% Annual household income < $10,000 60% 56% Trust score 72.4 (12.4) 72.7 (12.7) Would definitely recommend PCP 57% 59%

* There were no statistically significant differences by study arm. ** For participants born outside the United States Table 2. Change in patient trust in and satisfactions with the primary care provider (PCP) and number of visits to the PCP from baseline to 12 months Outcome Change in health coach group Change in usual care group Difference in change 95% CI p- value Adjusted p- value Patient Trust score (mean) + 3.8 +1.4 2.4 0.03 to 4.8 .047 .033* Definitely recommend PCP (%) +16.3% +4.0% 12.3% 5% to 24% .002 .015* *Adjusted for number of visits to PCP during 12 month intervention.

HEALTH LITERACY AND USABILITY OF PATIENT EDUCATION DOCUMENTS Raman Singhal1; Katya Ostrow2; Calie Santana1. 1Montefiore Medical Center, Bronx, NY; 2Sackler School of Medicine, Tel Aviv, Israel. (Tracking ID #1929654)

BACKGROUND: In 2003, the National Assessment of Adult Literacy found that 36% of the population has basic or below basic health literacy and 41% of Bronx County adults lack basic literacy skills. With the advent of electronic medical records (EMRs), the dissemination of patient education materials has become increasingly feasible and required as part of the meaningful use federal incentive program. It remains largely unknown, however, if patients can understand and act on the information presented. We sought to determine the literacy level and usability of patient handouts, which document characteristics should be improved upon to increase usability, and how documents from different vendors perform on a usability scale. METHODS: We analyzed 35 patient education documents in English from 4 vendors (A-D) and pertaining to 6 medical conditions (3 acute, 3 chronic) using the Suitability Assessment of Materials (SAM) tool. Vendors A and D provide easy-to-read ("basic") as well as regular ("standard") versions of documents - we analyzed 11 basic and 9 standard documents. The SAM tool assesses a document's usability in six domains: content, literacy demand, graphics, layout/typography, learning stimulation, and cultural appropriateness. The literacy demand domain incorporates the reading grade level of the document's text, which was calculated manually. Documents were printed in black-and-white, and were randomly numbered. Two researchers (RS and KO), blinded to each other's scores and vendor, reviewed each document. For each document, we calculated a usability score, defined as the average SAM score - usability is categorized as not suitable (SAM < 40%), adequate (40-70%), or superior (> 70%). We also calculated the average score for each domain; because of the disparate weighting of each domain within the SAM score, domain scores were converted to percentages. RESULTS: Usability ranged from 36.3% to 76.2% - 25 documents (71%) were adequate, 6 were superior, and 4 were not suitable. Figure 1 shows this distribution and the relationship between usability and reading grade level. All not-suitable documents were written at or above the 9th grade reading level and all superior documents were written between the 4th and 7th grade level. The 11 basic documents had an average reading grade level of 5.7 (range 4-7), 1 (9%) had superior usability, and zero were not suitable. The 9 standard documents averaged a 9.3 reading grade level (range 6-14), zero had superior usability, and 3 (33%) were not suitable. As expected, average scores in all domains of the SAM tool trended up with increasing usability; the highest correlation was with literacy demand and the lowest with layout/typography. By domains, literacy demand achieved the highest average score across all documents at 75.3%, while graphics achieved the lowest at 30.7%. Documents from vendor A had an average usability score of 56.4%, vendor B 70.3%, vendor C 61.9%, and vendor D 52.9%. Vendors A (25%) and D (75%) produced all not-suitable documents, and vendor B produced 66% of the superior documents. CONCLUSIONS: The majority of patient education handouts available to us are acceptable by health literacy standards. Literacy demand had the highest correlation with usability, however, our study highlights that it is just one factor in creating a superior

• document. As an example, the literacy differences between basic and standard documents are muted when these documents are scored

for usability. While several documents have mastered the literacy domain, the next area to improve is graphics - this domain had the second highest correlation with usability, but the lowest average score. As we aim to always choose usable documents, our results show that vendor B produced a majority of the superior documents. B's documents achieved the highest average score in 4 domains, excelling by the greatest absolute magnitude in the graphics domain. As we increase the use of printed materials in our daily practice, we must continue to critically appraise their value and quality to ensure that they are useful for our patients.

FOOD INSECURITY, COPING STRATEGIES, AND GLYCEMIC CONTROL IN LOW-INCOME PATIENTS WITH DIABETES Victoria L. Mayer1; Kevin H. McDonough1; Hilary K. Seligman2; Judith A. Long1,3. 1Perelman School of Medicine at the University of Pennsylvania, Philadelphia, PA; 2University of California San Francisco, San Francisco, CA; 3Philadelphia Veterans Affairs Medical Center, Philadelphia, PA. (Tracking ID #1933037) BACKGROUND: In the U.S., people who are food insecure (lack sufficient food for an active and healthy life) have a higher prevalence of diabetes and worse diabetes control. However, the mechanisms underlying these associations are not well understood. In this study we evaluate the relationship between food insecurity and coping strategies hypothesized to worsen glucose control in a population of low-income diabetics. METHODS: We performed a cross-sectional telephone survey among low-income adults with diabetes seen at a large, urban medical center. We interviewed people within 8 weeks of having a hemoglobin A1c (HbA1c) evaluation and characterized individuals as having poor glycemic control (HbA1c ≥ 8.0) or adequate glycemic control (HbA1c < 8.0). The main independent variable was a dichotomous measure of food insecurity (18-item U.S. Department of Agriculture Household Food Security Survey Module). We assessed the use of coping strategies including a diet low in fruits and vegetables and high in added sugars (using the California Health Interview Survey Dietary Screener), foregone medications, foregone medical care, use of food assistance programs (including the Supplemental Nutrition Assistance Program - SNAP) and emergency food programs, and overeating in times of adequacy. Additional independent variables included socio-demographic and clinical

• covariates. In bivariate analyses, we compared glucose control and use of coping strategies in the food insecure

and food secure groups. We developed logistic regression models to evaluate the association between food insecurity and glucose control, with adjustment for covariates and coping strategies. Finally, we assessed the role of interactions between food insecurity and coping strategies. RESULTS: Of 413 respondents (response rate 51%), 40.4% were food insecure. The food insecure group had a higher proportion of patients with poor glycemic control (68.3% vs. 53.3%, p=0.002). There were no significant differences between the food insecure and food secure groups in daily intake of fruits, vegetables, or added

• sugars. A significantly higher percentage of patients in the food insecure group reported foregone medications,

foregone medical care, participation in food assistance programs and use of emergency food programs. Food insecure patients were also more likely to report overeating at times when food is adequate. However, these coping strategies were not significantly associated with glucose control. In the adjusted model without interaction terms, when compared to food secure individuals, food insecure individuals were more likely to have poor glucose control (OR 2.83, 95% CI 1.47-5.44). However, there was a significant interaction between food insecurity and receipt of SNAP: food insecure individuals receiving SNAP benefits were less likely to have poor glucose control (OR 0.28, 95% CI 0.09-0.90), while food insecure individuals not receiving SNAP were more likely to have poor glucose control (OR 7.41, 95% CI 2.41-22.77). CONCLUSIONS: Food insecurity was associated with higher risk of poor glucose control. While more prevalent among patients who were food insecure, many coping strategies previously thought to underlie this relationship were not significantly associated with poor glucose control. However, receipt of food assistance among food insecure individuals was associated with better diabetes control, suggesting that in addition to helping individuals afford food, such programs may also play a role in improving health.

PATIENT ACTIVATION INTERVENTION IMPROVED EVIDENCE-BASED MEDICATION USE IN DIABETIC AND HEART DISEASE PATIENTS IN FEDERALLY QUALIFIED HEALTH CENTERS: THE OFFICE GUIDELINES APPLIED TO PRACTICE PROGRAM Ade B. Olomu1; Bikki Gautam1; Bethany Buda1; Wei-Wen Hsu3; Gurpreet Chahal1; Janaki Samaraweera1; Haritha Machavarapu1; Wendy Uwaje1; David Todem2; Margaret Holmes-Rovner4. 1Michigan State University, East Lansing, MI; 2Michigan State University, East Lansing, MI; 3Michigan State University, East Lansing, MI;

4Michigan State University, East Lansing, MI. (Tracking ID #1933143)

BACKGROUND: Many studies have documented suboptimal treatment for cardiac outpatients, especially for minority and low-income population. Despite the well documented efficacy of aspirin, Beta Blockers, statins and angiotensin - converting enzyme inhibitors (ACEIs), appropriate use and adherence to these therapies remains a concern in patients attending Federally Qualified Health Centers (FQHCs). Patient activation/engagement is an increasingly important component of strategies to improve outcomes for patients and reform health care. Objectives: 1) determine the impact of a patient activation intervention on guidelines based medication use for diabetic and heart disease patients in FQHCs 2) determine the predicators of medication use. METHODS: The Office Guidelines Applied to Practice (Office-GAP) study is a two-center study designed to improve cardiovascular care for minority and low-income populations in outpatient clinical settings. Clinics were assigned to intervention or control arm by tossing the coin. Office-GAP intervention included: 1) Patient activation /engagement intervention during a group visit 2) Physician training for patient activation/engagement and 3) Decision support/Checklist intervention (DSI) used in real time in the office. After a group visit, patients followed up with 2 physician visits using GAP tools. We performed chart abstraction of all enrolled patients with cardiovascular disease (CVD) and Diabetes Mellitus (DM) from September 2010 to Dec 2012 in 2

• FQHCs. Logistic Regression analysis was used to examine change over time in the proportion of patients using

Aspirin/Plavix, ACEIs /ARBs, beta-blockers and statins. RESULTS: Of 242 patients studied, 100 patients were in the intervention (Office-GAP) arm and 142 in the control arm. The control group showed no difference between baseline and 6 months medication use, while the intervention group showed the use of ACEIs/ARBs as (58.62% vs. 63.22%); Aspirin/Plavix (67.74% vs. 96.77%), Beta-blocker (60.00% vs 71.43%) and statin (62.78% vs. 80.52%) respectively. Longitudinal logistic regression revealed Office-GAP intervention significantly increased the use of ACEIs/ARBs for all eligible patients at 3 months (OR 3.93, p=0.001), Aspirin/Plavix (OR 2.43, p=0.046) at 6 months compared to control. Predictors of medication use included age, {ACEIs/ARBs(OR =1.03, p= <0.001), Aspirin/Plavix (OR= 1.04, p=<0.001), Beta Blocker( OR=1.02, p=0.042) and statins (OR=1.03, p=<0.001)} and Charlson Index {ACEIs/ARBs (OR=1.28, p= <0.001), ASA/Plavix (OR=1.4, p=<0.001), Beta Blocker (OR=1.84, p=<0.001) and statins (OR=1.38, p=<0.001) }. Whites were more likely to be on ASA/Plavix at baseline, 3 months and 6 months compared to Blacks (OR=1.78 p=0.011). Furthermore, patients that completed all 3 Office- GAP visits were more likely to be on ASA/Plavix (OR=1.73, p=0.022) compared to patients that completed only one or two Office-GAP visits. CONCLUSIONS: This Patient Activation Program led to increased use of guidelines based medications for patients with CVD and DM in FQHCs. We found that white patients were more likely to be on aspirin. Age and higher comorbidity predicted increased medication use. The Office-GAP program could serve as a model for implementation of guideline-based care for chronic diseases in outpatient clinical settings. Further study is needed to establish reach, effectiveness, and cost-effectiveness.

INDEPENDENT BARRIERS TO DIABETES CONTROL: RESULTS FROM THE MEASURING ECONOMIC INSECURITY IN DIABETES (MEND) STUDY Seth A. Berkowitz1; James B. Meigs1; Steven J. Atlas1; Darren A. DeWalt4; Hilary K. Seligman3; Deborah J. Wexler2. 1MGH, Boston, MA; 2MGH, Boston, MA; 3UCSF, San Francisco, CA; 4UNC, Chapel Hill, NC. (Tracking ID #1934690) BACKGROUND: Food insecurity, cost-related medication underuse (CRMU), and housing instability may be significant, yet modifiable, barriers to diabetes management. However, whether these barriers are independently associated with poor diabetes control, accounting for other aspects of low socioeconomic status and comorbidity, is unclear. To inform future population management interventions, we tested the hypothesis that they are independently associated with poor diabetes control. METHODS: We contacted a stratified random sample of adult (age >20 years) diabetes patients in 4 clinic sites (2 community health centers, 1 academic internal medicine practice, and a diabetes center) from June 15-Sept 15th 2013. Participants completed validated instruments assessing food insecurity, CRMU, and housing instability, in English or Spanish. We also collected information on other social circumstances, including educational attainment, limited English proficiency (LEP), health literacy, and nativity, along with age, gender, race/ethnicity, insurance, diabetes duration, Charlson comorbidity score, diabetes medications, Hemoglobin A1c (HbA1c) tests/year and number of outpatient visits. The primary outcome was poor diabetes control (most recent HbA1c>9.0% or LDL cholesterol >100 mg/dL). We estimated prevalence using inverse probability weighting. We performed multivariable logistic regression analysis with generalized estimating equations to account for clustering by clinic, and used pseudo-R2 statistics to evaluate explained variation in diabetes control. RESULTS: Overall, 412 patients were included (response rate: 62%). Of these, 19% reported food insecurity, 28% CRMU, and 11% housing instability. Patients reporting food insecurity were more likely to be younger (mean age 56 vs. 63 years, p=.01), non-white (35% vs 22%, p=.01), and have Medicaid (22% vs. 11%, p=.03). In unadjusted analyses, the prevalence of poor diabetes control was higher in those with, vs. without, food insecurity (53% vs. 28%, p=.01), CRMU (53% vs. 22%, p=.01) and housing instability (56%

• vs. 30%, p=.01). Those with, vs. without, food insecurity, CRMU, or housing instability had similar comorbidity and HbA1c tests/year

(an indicator of engagement with care), but more outpatient visits/year (Table 1). In a multivariable regression analysis, adjusted for the covariates listed above, food insecurity and CRMU were associated with worse diabetes control (Table 2). Using Pseudo-R2 statistics, our model explained 28% of the variation in diabetes control; food insecurity, CRMU, and housing accounted for 4% of total variation. By comparison, education, health literacy, LEP, and nativity (other indicators of social disadvantage) together explained only 2%. CONCLUSIONS: Food insecurity and CRMU are common, and independently identify patients at increased risk of poor diabetes control despite similar comorbidity and engagement with care. Because these patients already have higher healthcare utilization, interventions targeting food insecurity and cost-related medication underuse may have greater utility than increasing usual care.

Table 1

Food Insecure Food Secure p- value Cost-related Medication Underuse No Cost-related Medication Underuse p- value Unstable Housing Stable Housing p- value Charlson Score Median (IQR) 4 (3-7) 4 (2-7) .44 4 (3-6.5) 4 (2-7) .84 3.5 (2-6) 4 (3-7) .11 HbA1c tests/year Median (IQR) 2.5 (1.8- 3.0) 2.5 (1.9- 3.0) .90 2.4 (1.9-3.1) 2.5 (2.0-3.0) .94 2.9 (2.1-3.4) 2.5 (1.9- 3.0) .08 Outpatient Visits/year Median (IQR) 8 (5-11) 6 (4-10) .01 8 (5-11) 7 (4-10) .048 9 (6-13) 7 (4-10) .01

Table 2 Adjusted Odds Ratio (95% CI) Food Insecurity 1.62 (1.18-2.23) Cost-related Medication Underuse 2.59 (1.41-4.75) Housing Instability 0.82 (0.28-2.40)

ASSOCIATIONS BETWEEN MEDICAL HOME CHARACTERISTICS AND SUPPORT FOR PATIENT ACTIVATION IN THE SAFETY NET: UNDERSTANDING DIFFERENCES BY RACE, ETHNICITY, AND HEALTH STATUS Robert S. Nocon1; Yue Gao1; Kathryn E. Gunter1; Janel Jin2; Lawrence P. Casalino3; Michael T. Quinn1; Sarah Derrett4; Wm Thomas Summerfelt5; Elbert S. Huang1; Sang Mee Lee6; Marshall Chin1. 1University of Chicago, Chicago, IL; 2Johns Hopkins Bloomberg School of Public Health, Baltimore, MD; 3Weill Cornell Medical College, New York, NY; 4Massey University, Palmerston North, New Zealand; 5Advocate Healthcare, Chicago, IL; 6University of Chicago, Chicago, IL. (Tracking ID #1936110) BACKGROUND: Few studies have evaluated whether the patient-centered medical home (PCMH) supports patient activation, and existing evidence is mixed. No studies have evaluated whether support for patient activation differs among racial and ethnic groups in a PCMH. This is critical because activation is lower on average among minority patients. We sought to assess the association between clinic PCMH characteristics and patient perception of clinic support for patient activation, and whether that association varies by patients' self- reported race, ethnicity, and health status. METHODS: We conducted a cross-sectional analysis of PCMH characteristics and patient-assessed clinic support of patient activation in 24 safety net clinics across five states. PCMH characteristics were measured via surveys of 271 providers and staff. The provider and staff survey produced a 0 (worst) to 100 (best) PCMH score based on a scale created by the authors and described previously in the literature. Clinic scores were created by averaging provider/staff responses. Clinic support of patient activation was measured via surveys of 1,656 patients. The patient survey used the patient activation scale of the Patient Assessment of Care for Chronic Conditions to produce a 0 (worst) to 100 (best) score for clinic support for patient activation. Patient race, ethnicity, and health status were based on self-report in the patient survey. To investigate the relationship between PCMH characteristics and patient activation, while allowing for a clustering effect of patients within clinics, we fitted multivariate models using generalized estimating equation models with an exchangeable correlation structure. We analyzed interactions terms to assess how the association of PCMH characteristics and patient activation varied by race/ethnicity and health status subgroups. We interpret the association in terms of a 10-point change in PCMH score, a difference we found to be operationally meaningful in previous work. RESULTS: We received 214 (79.0%) provider and staff survey responses and 735 (44.4%) patient survey

• responses. Mean PCMH score among the 24 clinics was 58.7 (SD=6.4). The mean score for patient perception
• f clinic support of patient activation was 68.8 (SD=30.0). Across all patients, a 10-point higher PCMH score

was associated with a 5.6-point higher score for patient perception of clinic support for patient activation (95% CI, 1.3-9.9). The association between PCMH score and patients' perception of clinic support for patient activation was particularly strong among minority patients in fair or poor health: a 10-point higher PCMH score was associated with a 16.2-point (CI 1.7-30.6) higher score for clinic support of patient activation among Hispanic patients in fair or poor health and a 34.1-point higher score among black patients in fair or poor health (CI 9.1-59.1). The effect of PCMH score on patient activation score among black patients with poor/fair health status was statistically significantly different from the effect seen among non-Hispanic white patients in good or better health. CONCLUSIONS: In a population of safety net patients, PCMH characteristics showed a moderate, positive association with patients' perception of clinic support for activation; the magnitude of association was notably larger for minority patients in poor/fair health status. The PCMH may be promising for reducing disparities in patient activation for ill racial and ethnic minority patients.

THE ASSOCIATION OF FOOD INSECURITY AND DIABETES CONTROL AMONG LOW- INCOME INDIVIDUALS Julie B. Silverman1,2; Jim Krieger3,4; Meghan M. Kiefer2,1; Paul Hebert1,2; Nathan Drain3; June Robinson3; Leslie Taylor3; Janet Kapp3; Karin M. Nelson1,2. 1VA Puget Sound Healthcare System, Seattle, WA; 2University

• f Washington, Seattle, WA; 3Seattle King County Public Health, Seattle, WA; 4University of Washington,

Seattle, WA. (Tracking ID #1937147) BACKGROUND: Food insecurity is defined as the lack of dependable access to adequate, safe and nutritious foods necessary for a healthy and active life. Although there is an abundance of literature examining the relationship between diet and health, research on the health consequences of food insecurity is surprisingly

• limited. Only a few studies have demonstrated an association between food insecurity and glycemic control.

The objective of this study was to evaluate the relationship between food insecurity and glycemic control among individuals with diabetes and to explore potential mediators of this association. METHODS: We conducted a secondary analysis of baseline data from 287 low-income patients with poorly controlled type 2 diabetes (HbA1c >=8%) who were enrolled in the Peer Support for Achieving Independence in Diabetes (Peer-AID) trial, a randomized controlled study evaluating a diabetes self-management intervention delivered by community health workers. We evaluated the differences in socio-demographic and clinical characteristics, including glycemic control, between food-secure individuals and those struggling with food insecurity (based on the USDA's 6-item Food Security Survey Module). We used multivariable linear regression to model the relationship between glycemic control and food insecurity, adjusting for age, gender, race, language, education, marital status and BMI. Diabetes-related distress (moderate distress defined by a score of 2 or more on Polonsky's Diabetes Distress Scale), depression (depression defined by a score of 5 or more on the PHQ-8) and medication adherence (low medication adherence defined by a score of less than 6 on the Morisky Scale) were evaluated as potential mediators. RESULTS: The prevalence of food insecurity was 47.4%. Compared to food-secure participants, participants with food insecurity were more likely to be female (59.6% vs. 39.1%, p=0.001), non-white (66.9% vs. 44.4%, p <0.001), speak only English at home (59.6% vs. 46.4%, p=0.025), be unmarried or single (68.4% vs. 45.7%, p<0.001) and be obese (68.4% vs. 50%, p=0.001). Food-insecure participants utilized the health care system more frequently in the past year, with increased physician visits (10.7 vs. 7.3, p<0.001) and a higher percentage using the emergency department (42.7% vs. 31.1%, p=0.04). Although all subjects had poor glycemic control, food-insecure individuals had significantly higher hemoglobin A1c levels (9.4 vs. 8.8, p=0.003 unadjusted; 9.4

• vs. 8.8, p=0.002 adjusted). They also had a higher incidence of hypoglycemic episodes in the past year (57.7%
• vs. 45.2%, p=0.04). Forty percent of the food-insecure individuals who experienced hypoglycemia attributed it

to not being able to afford food, compared to none of the food-secure individuals. Food-insecure participants had increased distress regarding their diabetes care (55.1% vs. 33.7%, p<0.001), a greater prevalence of depression (74.1% vs. 35.6%, p<0.001) and lower medication adherence (52.9% vs. 37.2%, p=0.02). After adjusting for these potential mediators, the A1c levels of food-insecure individuals remained significantly higher than those of food-secure individuals (9.4 vs. 8.7, p=0.002). CONCLUSIONS: Food insecurity is associated with poorer glycemic control among low-income individuals with diabetes. Given that 1 in 6 households was food insecure during 2012, food insecurity is not an inconsequential concern. Assessing food security status and addressing patients' food needs may be an effective, non-pharmacological intervention to improve glycemic control in low-income individuals with poor diabetes control.