Quality Improvement to Identify and Address Food Insecurity During Pediatric Hospitalizations Free

Monroe Carell Jr. Children’s Hospital at Vanderbilt is a 343-bed academic tertiary care pediatric hospital located in Nashville, Tennessee.²³  The hospital’s patient population includes children from urban, suburban, and rural areas. In 2022 to 2023, 54.3% of admitted patients were insured by Medicaid. English (87.6%), Spanish (9.1%), and Arabic (1.4%) were the 3 most common preferred languages. The prevalence of FI in Tennessee at the time of this project was 11.5%.²⁴  This QI project took place on the 42-bed pediatric medicine acute care (PMAC) unit, a general pediatric inpatient unit with patients on medical or surgical services. The unit is staffed by attending physicians, advanced practice providers, residents, medical students, and pediatric nurses. Among pediatric nurses, the admission/discharge/transfer (ADT) nursing team on the PMAC unit is primarily responsible for patient intake between 11 am and 9 pm on weekdays. Bedside nurses perform intake outside of these hours and during times of high admission volume. Previously, there was no routine screening for FI or other social needs occurring at any time during hospitalization. In the absence of routine screening, FI was only addressed if needs identified by a patient’s care team prompted a social work (SW) consult. Our hospital system utilizes EPIC (Verona, Wisconsin) as its electronic health record (EHR).

We used the Model for Improvement as a framework to guide this project.²⁵  The interprofessional QI team included unit physician and nursing leadership, attending pediatric hospital medicine (PHM) physicians, a PHM advanced practice provider, ADT nurses, social workers, pediatric residents, and medical students. Bedside nursing champions and SW leadership were also involved in key meetings and decisions. The team developed a key driver diagram to guide development of a FI screening process and referral to resources (Fig 1). Primary drivers included:

1. routine screening that accurately identifies FI;

2. clear communication between care team members and families;

3. buy-in from providers, nursing, SW, and families on the importance of addressing FI in the hospital;

4. availability of user-friendly food resources; and

5. documentation of screening results and response in the EHR.

Plan-Do-Study-Act (PDSA) cycles were used to identify, implement, and assess interventions targeting these key drivers.

Our hospital has a well-established family advisory council. This project’s team leader worked with the council and developed a subcommittee focused on providing insight into FI screening and resource provision. The team leader partnered with them throughout the project to codesign interventions and receive feedback about the process from the caregiver perspective.

In February 2021, PDSA cycles began for families cared for by the PHM service on PMAC. In the initial tests of change, ADT nurses provided families a paper screening form that included the hunger vital sign questions²⁶  at the time of nursing admission intake. See Fig 2 for full process maps of initial and current screening processes. Admission was chosen as the time to conduct screening for the following reasons:

1. ability to integrate into an existing nursing workflow;

2. hypothesized increased likelihood of caregiver presence during admission; and

3. to allow time to address positive screens.

A paper screening form completed by the family was intentionally prioritized over verbal screening on the basis of evidence of improved social risk disclosure when families do not have to discuss them aloud.^27,28  The paper form was available in the 3 most preferred languages at the hospital: Arabic, English, and Spanish. Families were considered to have a positive FI screen if they answered “sometimes” or “often” to at least 1 question. Screens were considered high risk if they answered often to at least 1 question, implying the experience of chronic FI. For positive screens, the physician team manually inserted an Epic SmartPhrase listing state-based food resources in the patient’s preferred language into the after-visit summary (AVS) (Supplemental Information). Nurses reviewed this resource handout with families along with other discharge paperwork. In response to a high-risk screen, the physician team placed an SW consult to review available food resources with the family and assess for and address cooccurring social risks. Bedside nurses could also place an order for 1 daily, free guest tray at the family’s request.

Subsequent interventions included standardizing our screening process, sending audit and feedback e-mails to PMAC nursing, integrating screening questions into the nursing admission history in the EHR and automating AVS resource provision for positive screens (Supplemental Information, AVS Food Resources and SmartText), expanding screening to patients on all services, making screening questions required documentation, and providing education to clinical team members throughout the project. Each PDSA cycle and its associated key driver(s) and lessons learned are described in Table 1.

To gauge nursing buy-in and parent perceptions of the screening process, we also conducted stakeholder interviews with a subset of 10 randomly selected PMAC nurses and 10 screened families. Each stakeholder was approached by a team member and completed a short questionnaire (Supplemental Information, Interviews) in a private location.

Eligible patients initially included Arabic-, English-, and Spanish-speaking families admitted to a PHM team on PMAC, then families preferring those same languages admitted to any team on PMAC, and finally all patients on PMAC. Once all patients on the unit were eligible, families that preferred languages other than English, Spanish, or Arabic had their screening form completed and resource handout information reviewed with the assistance of an interpreter. Screening was conducted once during each admission for any patient with multiple admissions during the study time frame.

Data, including screening results and social work consult (yes/no), were collected for eligible patients prospectively throughout the project via a report generated within the EHR by a member of the improvement team.

The primary measure was the percentage of eligible families with FI screening documented during their admission. Data from each unique admission were included in analyses. To monitor for inequity in the screening process, the percentage of eligible families with documented FI screening was also stratified by preferred language (ie, number of screened families who prefer Spanish/number of eligible families who prefer Spanish).

We measured the percentage of screened patients who reported FI. To ensure resource information was being provided to families, our secondary measure was the percentage of families with a positive FI screen documented who received food resource information. Inclusion of food resource information in the AVS was determined through chart review for each patient with a positive screen.

The balancing measure was the percentage of families screened for FI with an SW consult ordered during their admission. This measure was included to monitor for a substantial increase in SW consults because of the screening process.

Descriptive statistics were calculated for insurance status and preferred language overall, and by FI screening result.

Data for all measures were grouped weekly except for the Arabic language stratification, which was grouped into consecutive groups of 5 given the small sample. Appropriate statistical process control p-charts were used to analyze the impact of interventions on process, outcome, and balancing measures and to identify special cause variation.²⁹  For the secondary measure of food resource provision in the AVS, the team followed the data prospectively in a p-chart but transitioned to a g-chart for the purposes of the manuscript given the abundance of weekly data points at 100%. Statistical process control charts were generated using QI charts (Process Improvement Products, San Antonio, TX).

The Vanderbilt University institutional review board approved this QI project as a nonresearch study. This manuscript was written according to the Standards for Quality Improvement Reporting Excellence 2.0 guidelines.³⁰ 

A total of 8850 admissions between February 15, 2021, and August 6, 2023, were eligible for FI screening. Among the 2801 patients screened, 1% preferred Arabic, 91% preferred English, 7.5% preferred Spanish, and 0.5% preferred another language. Over half (56.1%) were insured by Medicaid. A total of 5.1% of screened patients had >1 admission during the study period.

Five instances of special cause variation resulted in centerline shifts with cumulative improvement from 0% to 77% during the study period. Key interventions included testing of the FI screening process, standardization of daily screening form collection, audit and feedback e-mails to the nursing team, expansion of screening to patients admitted to all services, and making FI screening questions required, but not “hard-stop,” nursing admission documentation (Fig 3). There was 1 point outside the control limits associated with the initial testing of EHR-based screening questions conducted among the ADT nursing team in preparation for implementing the new screening process with all PMAC nurses.

Improvements were observed in screening percentages for English-, Spanish-, and Arabic-speaking families. However, the means for families who preferred Spanish (59.4%) and Arabic (30.8%) were lower than the English-speaking (78.0%) families at the end of the data period (Supplemental Figs 5a–5c).

Among families screened, 11% (310) screened positive for FI, with 1.5% (42) reporting chronic FI. Among those screening positive, 1.3% preferred Arabic, 73.4% preferred English, 24.9% preferred Spanish, and 0.3% preferred another language. The majority (82.9%) were publicly insured. Special cause variation on the g-chart was associated with the automated EHR resource provision for positive FI screens, with the study concluding with 234 consecutive successful information provisions for families with a positive screen, well above the upper control limit (Fig 4).

The rate of SW consults among screened patients remained stable at a mean of 13% throughout the study period (Supplemental Fig 6), indicating that the FI screening process did not meaningfully impact SW consult volume. Two observed instances of special cause variation were separated by over a year, but investigation into these weeks did not reveal any insight into the potential causes of the increased SW consults.

Interviews with bedside nurses revealed support for screening and appreciation for the paper screening form, as well as the opportunity to play a role in helping families get needed resources. Families reported being receptive to, and even appreciative of, being asked the screening questions (Supplemental Information, Interviews).

Our multidisciplinary team successfully implemented FI screening at the time of hospital admission as part of routine clinical care in a tertiary, freestanding children’s hospital. Using QI methodology and PDSA ramps, we scaled screening conducted by a specialized nursing team on a single unit among Arabic-, English-, and Spanish-speaking patients on PHM teams to all patients admitted to the hospital unit. We achieved our Specific, Measurable, Achievable, Relevant, Time-bound (SMART) aim by screening 77% of eligible patients. Using EHR automation, we provided food resource information to 100% of families (234 consecutive families) with a positive screen without a significant increase in SW consults among screened patients. This work demonstrates that FI screening can be completed routinely as part of hospital care with reliable resource provision.

Despite growing interest, limited reports exist describing efforts to integrate FI screening into routine clinical care for all patients on a general inpatient unit that can be used to guide hospitalwide FI screening. QI efforts focused on comprehensive social risk screening that included FI as 1 of the domains that have been implemented on specialized units, but these only screened English-speaking families via research assistants or residents.^31,32  Another QI initiative aimed to identify and address caregiver hunger during admission, with screening and interventions provided by a research assistant.³³  One freestanding academic children’s hospital in the Southeast initiated hospitalwide screening for FI by incorporating the hunger vital sign questions into nursing admission intake. A positive screen triggered an SW consult to connect families with resources. During their study period, 61.6% of eligible admissions had an FI screen completed, with 4.9% of screened patients reporting FI.³⁴ 

We began our screening pilot on a small scale using an approach similar to Hanna et al,³⁴  while also expanding on their work by including patients who prefer languages other than English and employing QI methodology to develop and test our approach. Because it was built using existing infrastructure, our initial screening approach involved a convoluted process with many steps reliant on multidisciplinary, rotating care team members integrating new tasks into their workflow (Fig 2a). This complexity is reflected in the low screening rates throughout the duration of this process. However, these initial data were instrumental in demonstrating feasibility, nursing buy-in, and the need for a more streamlined process to leadership. This ultimately resulted in our ability to integrate screening questions into the EHR nursing admission intake.

EHR integration represented a notable improvement in the reliability of our process but did not result in a sustained increase in screening rates like we had anticipated. Our most impactful interventions were expanding screening to patients on all services on PMAC and making screening questions required admission documentation. This highlights the benefit of decreasing the cognitive load required to identify eligible patients by making screening part of the standard admission process for every patient. Additionally, the use of a visual reminder (green checkmark) that accompanies the required documentation workflow, even without the questions being a true hard stop to the admission process, is an effective tool to remind and motivate staff to complete tasks. Early buy-in from nursing leadership and longitudinal support from nursing champions were also key components of successfully implementing and expanding this process. Other teams aiming to change behavior may consider incorporating these intervention approaches.

We saw a lower screening rate among families who prefer Spanish when examining data across the full study period (Supplemental Fig 7). There were sustained improvements with the Spanish population associated with adding screening questions to the EHR and making them required documentation and the Arabic population when screening expanded to all nurses, but not in response to other interventions that resulted in special cause among the full cohort and families who prefer English (Fig 3, Supplemental Fig 5 A–C). Possible explanations for this disparity were provided by patient and bedside nurse feedback throughout the project and are a focus for future interventions. These include limited introduction to why the forms are being provided when using an interpreter to complete an admission, unclear translation, inaccurate documentation of preferred language (ie, actually prefer Chuj instead of Spanish), or team difficulty in transcribing responses into the EHR. Importantly, this discrepancy in screening rates highlights the need for constant monitoring to ensure disparities do not develop as a result of interventions that lead to success within the overall cohort.

To improve screening rates among patients who prefer languages other than English, and to get closer to universal screening completion in general, additional measures such as better understanding of how to introduce screening forms in a culturally acceptable manner, additional translation services, and EHR reminders to circle back to complete unanswered questions outside of the admission period for patients with parents who are not at bedside during admission are likely needed.

Eleven percent of screened families reported FI, within the range of 5% to 38% in the existing literature^15,34,35  and similar to the local prevalence of 11.5%.²⁴  Our system of automated resource insertion into the AVS for patients screening positive resulted in consistent provision of food resource information for families experiencing FI. Because of limited ability for our SW team to support an increased consult volume, we developed a triage system to address positive screens that involved flagging only high-risk patients for an SW consult. Anecdotally, team members were entering consults for these high-risk patients, though our system does not allow us to capture whether an SW consult was placed specifically for FI. Although the most effective and desirable method of connecting families to resources from the hospital remains unknown, there is evidence from the acute care setting that provision of targeted resource handouts is noninferior to navigator assistance in decreasing social risks and improving child health.³⁶  Despite serving a population across a wide geographic area, we tried to ensure all families had relevant resource information by including quick-response codes that allow families to search for resources by zip code.

Limitations of this work include possible missing data during the time frame of our initial screening process that required transfer of screening forms between team members before entry into the EHR. We would expect this to result in an underestimation of our screening rate during that time, and there are clear trends toward improvement associated with specific interventions after transition to a more reliable, EHR-based screening process. Because this study was conducted on 1 unit at a single academic, freestanding children’s hospital with an ADT nursing team, our specific process may not be generalizable to other institutions with different nursing infrastructure. Additionally, our most reliable interventions were designed within Epic, which may not translate directly to other EHRs. Nevertheless, this approach can be adapted to the specific nursing workflow and EHR infrastructure within any institution. Finally, following up with patients after discharge to determine who used the provided information to access resources was outside the scope of this QI project, but is an important focus for future research to guide effective inpatient social resource interventions.

We used QI methodology to test and implement FI screening as part of routine clinical care during nursing admission intake. Building this screening process has supported ongoing work to create a more robust response to positive screens including hiring an inpatient resource specialist. Health care systems with limited SW support may consider automated inclusion of food resource information in a family's preferred language in their discharge papers as an initial step toward addressing needs within their population. This is a feasible approach for hospitals to consider as they develop systems to screen for and address social risks in response to Centers for Medicare and Medicaid Services and Joint Commission mandates.

We thank bedside champions Dr Courtney Svenstrup, Shelby Gunther, Beth Loats, and the PMAC ADT and bedside nursing team; Julie Garcia, Brittany Davis, and Nyah Cade for assistance developing patient resources and SW support throughout the project; and Karen Wilson and the health information technology team that built our EHR-related interventions.