Home Food Delivery for Diabetes Management in Patients of Rural Clinics

Type 2 diabetes mellitus (T2DM) is a growing public health concern affecting 37.3 million people (11.3%) in the United States (US). At the same time, 13.5 million (10.2%) US households experience food insecurity. People living in rural areas are particularly challenged by the intersection of food insecurity and T2DM - both food insecurity and T2DM prevalence in rural areas are higher than metropolitan areas. Arkansas has the sixth highest proportion of rural population (44.5%) of all US states, with rates of food insecurity and T2DM of 14.7% and 12.5%, respectively. Rurality and food insecurity are associated with social determinants of health that impede diabetes management (e.g., limited transportation, longer distance to healthcare clinics, lack of access to healthy food options). People experiencing food insecurity have poor diet quality (e.g., high intake of sugar-sweetened beverages and red/processed meat, low intake of fruits and vegetables), which is a key risk factor for unmanaged T2DM. Unmanaged T2DM can lead to increased risk of serious complications. Improved diet quality is a crucial factor to enhance T2DM management. Millions of people with T2DM experiencing food insecurity turn to food pantries for food; however, food pantries are not a long-term solution to improve health for rural people with T2DM experiencing food insecurity. Certified Diabetes Care and Education Specialists (CDCES; i.e., diabetes educators) have expertise in working with patients to instill knowledge and skills necessary to live an active and healthy lifestyle with T2DM.

This study used a two-arm randomized controlled comparative effectiveness design to compare the effectiveness of a Healthy Food Delivery Intervention (HFDI) plus a consultation with a CDCES Consultation Only in improving diabetes-related outcomes among rural people with T2DM experiencing food insecurity. We hypothesized HFDI to be comparatively more effective in improving HbA1c and other outcomes for rural people with T2DM experiencing food insecurity.

The primary outcome was change in hemoglobin A1c (HbA1c) between baseline and post-intervention. Secondary outcomes included diet quality (Healthy Eating Index-2015 and Veggie meter scores), body mass index, blood pressure, lipids, glucose, diabetes management self-efficacy and diabetes-related distress. We also examined the duration of treatment effects by comparing outcomes at 6 months post-intervention and 12 months post-intervention.

Participants were randomly assigned to either the HFDI arm or the CDCES Consultation Only arm. Participants randomized to the CDCES Consultation Only study arm received the standard of care at the clinic from which they were recruited, plus a 60 minute healthy eating consultation with a CDCES and a food pantry resource guide specific to their community. In addition to standard of care and the CDCES consultation, participants randomized to the HFDI arm received weekly home delivery of type 2 diabetes-appropriate food boxes over 3 months (12 food boxes in total). Each food box included approximately 9000 calories of food. Food boxes were designed by a Registered Dietician. Food boxes also included recipes and cooking tips specific to foods included in that week's box.

Participants were initially recruited only from family medicine clinics within the University of Arkansas for Medical Sciences (UAMS) network that serve the highest percentage of rural patients with T2DM. Additional clinics and Federally Qualified Health Centers were added as recruitment sites in order to meet study recruitment goals.

Inclusion criteria included: 1) 18 years of age or older; 2) food insecure (Hunger Vital Sign™ 2-item screener); 3) HbA1c ≥ 6.5%; 4) speak English or Spanish; and 5) live at a rural address. Note: the food insecurity and rural residence criteria were removed during the course of the study in order to meet study recruitment goals.

Exclusion criteria included: 1) having conditions making it unlikely the participant will be able to follow the protocol, such as terminal illness, severe mental illness, severely impaired vision or hearing, eating disorder, or plans to move out of the geographic region; 2) pregnant; 3) only one adult with T2DM per household was allowed to enroll.

Investigators worked with clinical stakeholders at each clinic site to ensure the study was integrated into existing workflows. Study flyers were placed throughout the clinics, in the exam rooms, and in the clinic waiting areas. All clinics had electronic health record systems that were queried monthly for recruitment. Research coordinators at each clinic site identified potential participants by reviewing monthly patient queries and daily clinic schedules. The research coordinators provided study information to potential participants in a pre-visit call or during the clinical visit. If the patient was interested, the research coordinator scheduled a 30-minute meeting to discuss the study and conduct eligibility screening in a private room in the clinic. Research staff were trained to flag potential participants that may not meet eligibility, which were then sent to the study physician for medical determination. Those deemed eligible to participate reviewed study information with the research coordinator, were given the chance to ask questions, and signed an informed consent. All recruitment and consent information was available in English and Spanish.

Randomization was conducted by the study biostatistician, who had no interactions with potential participants and no supervisory role with study staff responsible for recruiting, consent, and intervention processes. Stratified random assignment using permuted blocks of size 2, 4, and 6 was used to randomize participants 1:1 to the treatment arms using lists created with the blockrand R package. Strata of participant sex, level of T2DM control (6.5-9.0%; >9.0%), and age group (18-64; 65+ years) were used to ensure that similar numbers within each characteristic are assigned 1:1 to the treatment arms.

Data collection occurred in-person at the clinic from which participants were recruited, co-occurring with participants' primary care provider visits when possible. Biometric and survey measure outcomes were collected at baseline (pre-intervention), immediate post-intervention (approximately 12 weeks after baseline), 6 months post-intervention, and 12 months post-intervention.