Evaluating Screening to Assess Endorsement of Food Insecurity in the Inpatient Setting Free

We surveyed caregivers across 4 inpatient units at both the main and satellite campuses of a freestanding children’s hospital located in the Midwest. We screened caregivers on (1) a surgical short stay unit (SSS), (2) the transitional care center (TCC) unit, which admits children assisted mechanical ventilation outside of the intensive care unit, (3) the neuroscience unit (NNS), which includes both neurology and neurosurgery patients, and (4) the hospital medicine unit (HM) located at our satellite campus.

We conducted a prospective study using a combination of 3 surveys and 4 methods to measure FI rates across our 4 study units. We approached all eligible caregivers of children hospitalized on a single unit 4 days a week (1 unit per day) using a screening method that was randomly-selected for each day; the order of the units was random such that a single unit was not always approached on the same day of the week. We used random number generation to randomize the unit and the method within week-long blocks. A research assistant and 2 Spanish-speaking qualified bilingual staff members conducted surveys. The survey was introduced to families as a tool for us to learn more about hospitalized patients and families.

We asked each caregiver questions from 3 surveys: (1) US Household Food Security Survey Module: 6-Item Short Form¹³  (6-Item), (2) the Children’s Health Watch Hunger Vital Sign¹⁴  (a 2-item survey recommended by the American Academy of Pediatrics as a method to screen for FI),¹⁵  and (3) questions currently utilized by our social workers (Supplemental Fig 2). Furthermore, we screened families via 4 methods: face-to-face, phone, paper, and tablet. The face-to-face method consisted of a study team member approaching a caregiver at bedside and reading questions to them. For the phone call method, the study team called families, either on the in-hospital phone or the phone number listed in the electronic medical record. For the paper method, a study team member provided caregivers present in the hospital with a paper copy of the survey. The study team member then left the room and came back to collect the survey? 15 to 20 minutes later. Finally, for the tablet method, caregivers were given a tablet on which to complete the survey in REDCap. The study team member waited in the room while the caregiver completed the tablet survey.

We included all English and Spanish speaking caregivers of children hospitalized on 1 of the study units from February to April 2022. We also included patients >18 years who lived independently, except those hospitalized for treatment of an eating disorder. We excluded caregivers of patients who lived in a skilled nursing facility, patients who had been hospitalized since birth, and those in the custody of child protective services. We classified patients as “unavailable” if the family was not in the room for medical reasons (ie, imaging, surgery, etc) or if the medical or nursing team was with the family while the study team was recruiting. Patients were classified as “no caregiver” if a caregiver was not present when the study team was on the unit or if there was no answer on either phone (phone call method). Patients were classified as “declined” when a caregiver declined to complete the survey. Each day we attempted to approach all eligible caregivers and patients on the unit (including those previously missed). Caregivers who had previously completed the survey or declined to do so were not approached again. This study was deemed exempt by Institutional Review Board review.

Defining positivity for FI: if a caregiver endorsed any FI item on any of the 3 surveys, the survey was considered positive. If any of the 3 surveys were positive, we classified the caregiver as endorsing FI (overall positive). For the caregivers that screened positive, we asked if they would like to speak with a social worker and provided them with a resource sheet.

For bivariate analyses, we conducted chi square tests using STATA version 14 to examine for differences in overall positivity by survey method, clinical unit, race, ethnicity, language, and receipt of Supplemental Nutrition Assistance Program (SNAP) or Special Nutrition Assistance Program for Women, Infants, And Children (WIC) benefits. We also conducted a sensitivity analysis using multivariable logistic regression examining the association between method and overall positivity while accounting preferred language and unit. We selected these 2 covariates as it is reasonable that caregivers who prefer different languages may interpret the questions differently, making language an important covariate to include. Further, the method offered was not balanced across all hospital units but was balanced across all other potential covariates. (Supplemental Table 3). We considered P < .05 statistically significant.

We assessed eligibility in 646 patients, and of those, 562 were eligible to be approached. We completed 361 surveys (64% of all eligible), 343 in English and 18 in Spanish. Of families unable to complete surveys, 55 patients were unavailable, 133 patients had no caregiver, and 13 patients declined (2.3% of eligible) (Fig 1). Overall, of the 361 caregivers who completed surveys, 19.4% (N = 70) endorsed FI.

The 6-Item survey had 17.7% positivity rate, Hunger Vital Sign questions had a 16.0% positivity rate, and the social work questions had a 3.1% positivity rate. Regarding the method of screening, face-to-face had a positivity of 16.0%, paper 25.6%, tablet 17.2%, and phone 18.1%. The overall positivity rate did not vary significantly by method. (Table 1). In sensitivity analyses, the method of screening was not statistically associated with positivity when accounting for language and hospital unit in multivariable analyses. When method and survey question were considered together, the 6-Item survey administered via paper had the highest positivity rate of 20.9%. The Hunger Vital Sign administered on paper had a very similar positivity rate of 19.8%.

The rates of FI did not vary significantly across the different hospital units; the TCC had a 31.43% positivity rate, the SSS had a 13.25% positivity rate, the NNS unit had a 17.88% positivity rate, and the satellite HM unit had a 22.83% positivity rate (p = .10) (Table 2).

Rates of FI varied significantly by preferred language, with 61.1% Spanish-speaking caregivers endorsing FI compared with only 17.2% of English-speaking caregivers (P < .01), despite a small number of caregivers who spoke Spanish (n = 18). The positivity rate also varied significantly by self-identified race (P < .01); 16.5% of caregivers who identified as white endorsed FI compared to 26.7% of caregivers who identified as Black; 61.1% of caregivers who selected other race on the survey also endorsed food insecurity. Significantly more caregivers that identified as Hispanic, Spanish, or Latino ethnicity endorsed FI (48.5%) compared with caregivers that did not identify as these ethnicities (P < .01). A total of 33.7% of respondents that receive SNAP or WIC benefits endorsed FI compared with only 14.5% of those that do not receive these benefits (P < .01) (Table 2).

Almost 20% of the families across our 4 study units endorsed FI. Screening using paper and the 6-Item had the highest positivity rate at 20.9%, though screening using paper and Hunger Vital Sign was also nearly identical. The highest rates of FI were seen in families whose primary language was Spanish, those identifying as nonwhite, those identifying as Hispanic, and those who receive SNAP or WIC benefits.

Overall, we found higher rates of FI (∼20%) than most previously completed studies, including our own internal institutional screening.^16,17  The overall rate of FI in this work is consistent with the rate of FI in the community surrounding the hospital. Our study contrasts to prior screening efforts in that screenings were completed by research staff unaffiliated with the medical team and at a different time than medical intake questions. Often FI screening questions are asked by clinical staff (usually nursing) upon admission. In another study that also used research staff for screening in an emergency department documented a FI rate similar to ours, 17.5%.¹⁸  Importantly, our study also had a very low refusal rate (2.3% of those approached); this rate contrasts to previous studies with as many as 17% refusing screening.¹⁷  Therefore, having an individual who is not affiliated with the clinical team completing screening separated from clinical care may be critical for families to feel comfortable endorsing FI. Although there are expenditures associated with having a dedicated individual to screen for FI and other social needs, hospitals may find hiring a dedicated care navigator to screen and offer resources a worthwhile investment. At minimum, this work offers an evidence-based intervention to improve FI screening rates to mirror community rates through a dedicated individual offering these screening questions.

We demonstrate disparities in FI rates across self-identified race as well as between Hispanic and non-Hispanic families. This finding is similar to prior research in the emergency department.¹⁸  The intersection between social determinants of health, like FI, and structural racism has been well documented^19,20 ; further work is needed to reduce these disparities and understand contextual factors that may exacerbate differences. Additionally, our study demonstrated that families at our hospital who receive supplemental food services, like SNAP and WIC, are still food insecure.

Our study has several limitations that should be noted. It was performed at a single hospital across 2 campuses. Further, the majority of caregivers in this population identified as white and had a preferred language of English, limiting generalizability. Second, all screening was done by a research team member; therefore, our rates could be difficult to replicate in a clinical environment without dedicated personnel; although, this may be needed to accurately identify families with FI. Third, as is inherent in all prospective studies, we do not know the rates of FI in nonresponders. Fourth, we did not pilot the survey as a whole, though both the 6-Item Survey and the Hunger Vital Sign are validated instruments. The additional social work questions were administrated after the other 2 surveys and should not threaten validity of either instrument. Additionally, although we screened over 360 families for FI, it is possible that we were underpowered to detect small differences in positivity by modality. Finally, our statistical analyses did not account for all possible confounders, including age, insurance, diagnosis, and medical complexity. Although some of these attributes may impact a child’s risk for FI (eg, children with significant medical complexity are known to have increased medical expenses that may lead to or worsen FI²¹ ), it is unlikely that these attributes impacted the caregivers’ understanding of the questions or the preferred modality.

The 6-Item Survey, when administered on paper, had a positivity rate of at least 20%, the a priori threshold that was established by the research team as it approximated our area prevalence. The Hunger Vital Sign administered on paper had a similar positivity rate (19.8%). Both surveys could be considered for future use. Future studies should evaluate interventions available to improve FI as it affects a significant proportion of inpatient families.

We want to acknowledge the following study team members who contributed to the success of this project: Dr Anita Shah, Dr Andrew Beck, Dr Ndidi Unaka, Stacey Litman, Samantha Ingham, Ericka Fortson, Stephanie Powers, and Dr Megan Smith. The use of REDCap was provided via grant support (UL1TR001425).