Identifying Food Insecurity in Cardiology Clinic and Connecting Families to Resources

This study was part of a larger quality improvement (QI) initiative to improve FI screening at Children’s Hospital of Pittsburgh (CHP) in conjunction with a FI task force. CHP is a large, urban academic medical center in Allegheny County. Of 1.2 million people living in Allegheny County, nearly 1 in 7 are FI, constituting 42 000 children.²⁵  The hospital-based outpatient cardiology office primarily serves patients living in Pennsylvania (>90%), with a small percentage from other referral locations. The payor mix is 42% publicly insured, 56% private insurance, and 2% self-pay.

A pediatric cardiology fellow served as team leader and attended departmental meetings for QI support and CHP FI task force meetings for resource expansion. Clinic stakeholders were identified and included clinic intake providers, frontline physicians and advanced practice providers, and our social worker (SW). Data was monitored and collected by our QI team.

Critical key drivers were identified to achieve our aim to screen CHD patients for FI and to connect families to resources. Given the human nature of our intake process, we anticipated a maximum level of reliability (LOR) of 1 to 2. A LOR of 1 allows 2 failures out of 10 opportunities. We set our SMART aim to screen 80% of patients from a baseline of 0% in 1 year and connect all FI families to resources. Six primary drivers were identified: (1) increase FI awareness, (2) identify FI families via a defined protocol, (3) stakeholder buy-in, (4) connect FI families to resources, (5) follow-up with FI families, and (6) assess risk of FI based on CHD severity (Fig 1).

The intake team received training on FI and the HVS via a verbal presentation supplemented by information sheets. The FI task force and nursing leaders trained new staff and updated staff on electronic medical record (EMR) changes and new resources. Frontline providers (FLP) were educated about FI and informed of the screening processes and resources via E-mail training. Clinic hand-outs included basic needs support information, emergency food resources, national program resources, and programs specific to Western Pennsylvania and other counties (Supplemental Fig 6).

The HVS were implemented utilizing a verbal screening protocol during clinic intake. Early insights revealed discomfort by medical intake staff and low response rate from families. A standardized verbal screening script was trialed with limited success for positive screens.²⁶  On the basis of stakeholder input, the process was transitioned to a written screen with verbal screen back-up. The initial HVS were collected on paper and recorded in a temporary location in the EMR, which later became standardized.

SW was notified of positive screens via EMR messages and followed up with FI families by phone. This transitioned to paging SW during the clinic visit for an in-person meeting or introduction. If a meeting was not possible, the family was offered a resource sheet with the SW’s introductory note. The SW attempted to call FI families; those families who could not be reached were mailed any additional resources on the basis of their address on file. Eventually, we also offered an assessment by our community partner, Just Harvest. With consent, even in the absence of a SW, Just Harvest could perform a needs assessment of FI families.

Demographic data including age, gender, and race and ethnicity in addition to anthropometric body mass index (BMI) data and disease severity were collected to assess FI risk in our cohort. This data were included in our heart institute meetings to promote provider buy-in through data sharing.

We executed tests of change by conducting rapid cycle plan-do-study-act (PDSA) ramps with change concepts focused on staff education, screening process, documentation, and resource distribution (Table 1). Each ramp was composed of 6 PDSA cycles on the basis of hypothesis-driven interventions to implement HVS. We adopted, adapted, or abandoned changes after data review and stakeholder insights at each stage of interventions. We evaluated cardiology visits at our CHP downtown prospectively from January 29, 2018 to February 8, 2019. Baseline data were established from January 29 to March 5, 2018, and testing started thereafter. We tracked EMR data and paper back-up copies for FI screening and extracted EMR data for demographic information and CHD security.

The primary process measure was proportion of patients screened for FI to total number of patients seen at each visit.

Our outcome measure was the proportion of all FI families connected to resources. Initially, all FI families were offered a general resource handout that met this definition. Our clinic process was refined to offer more individualized resources to FI families. As such, our outcome measure was redefined reflecting proportion of FI families offered individualized resources. All data presented reflects this new definition. As a measure of screening success our secondary outcome measure was the proportion of FI patients identified per patients screened.

This project was implemented in a busy cardiology clinic. For our balancing measure, we assessed acceptance of FI screening and perception of clinic workflow impact via an anonymous survey.

A time series analysis was performed for our process and outcome measures. Measures were plotted on statistical process control (SPC) charts. Because these measures were yes or no responses, the chart selected was a p-chart.

For our process measure (patients screened for FI), each point on the respective SPC chart represents a combination of 5 clinic days. SPC methods were used to identify significant changes in system performance. The process was initially assessed biweekly. Centerline (mean) was shifted, and control limits were recalculated when persistent special cause variation was seen, defined as at least 8 consecutive points above the mean. Control limits were set at ±3 σ and, when recalculated, the original centerline was continued as a dotted line to visualize system change. Control limits were recalculated when stable process change was achieved with each PDSA cycle. Our goal line was set at 80% patients screened and 14% positive screening rate to match Allegheny County’s FI rate.

For our primary outcome measure (FI families connected to resources), each plot on the corresponding annotated p-chart represents a total of 7 clinic days with positive FI screens. Our goal was set at 100% families connected to individualized resources. Our secondary outcome measure (FI families identified) was tracked on the respective SPC chart points, which represent a combination of 5 clinic days. SPC methods were used to identify significant changes in system performance.

Addressing our balancing measure, we assessed acceptance of FI screening and perception of workflow after implementation. Via an anonymous survey tool, we assessed staff acceptance by (1) if providers felt clinic workflow was negatively impacted as a result of screening, (2) if providers felt that their patients have benefited from screening, and (3) additional comments on the process.

To promote buy-in and to better understanding our patient population’s FI resource needs, we analyzed FI risk on the basis of patient demographics and CHD disease severity. We modified the stratification scheme used by Hoffman et al to include all structural and acquired CHD (Supplemental Table 3).²⁷  Those patients without CHD were labeled as “none.” Mild CHD included processes requiring cardiology follow-up without intervention. Moderate disease required intervention or more frequent cardiology follow-up. Severe disease included those anticipated to have intervention in the newborn period and frequent cardiology follow-up.

Approval was given on the basis of the hospital QI project submission guidelines, which did not require institutional review board (IRB) submission.

Of 6885 cardiology patients seen, 5064 were screened for FI. Of the families screened, children in FI households were similar in age to children in food secure families. There was no association between FI and gender (43.5% vs 45.7% female, P = .65) or BMI classification (54.2% vs 59.1% normal BMI). Most families (3933 of 5064, 77.7%) seen in clinic identified as White, non-Hispanic. FI families were more likely to be Black (29.0% vs 12.8%) or identify as multiple or other racial and ethnic groups (4.6% vs 1.4%) and were less likely to be White (58.8% vs 78.7%) (Table 2).

Baseline screening was 0%. On implementation with verbal screening, there was a rapid, initial shift of FI screening rate to 60% within 2 months. When transitioned to a written screening tool, the process improved to 87% screened (Fig 2). At baseline, our clinic did not identify any patients experiencing FI. At the conclusion of our testing period, identification of FI rose to an average of 6%. After 3 months, we identified at least 1 family struggling with FI every week (Fig 3).

Initially, all FI families were offered a general paper resource, and a note was sent to SW for follow-up allowing us to meet our goal. In the early stages, many families were lost to follow-up. Insights from early outliers with highly successful resource distribution and input from the SW team revealed that face-to-face introduction on the day of the clinic greatly reduced loss to follow-up as compared to phone call follow-up alone. After the incorporation of paging the SW during the clinic visit for an in-person assessment or, if unavailable, after an introduction to the SW follow-up phone call, our centerline shifted to 84% individualized resource distribution. After the addition of Just Harvest consent, all families were offered what were considered appropriate, individualized resources by our team (Fig 4).

Individualized assessments revealed potential areas of intervention. For example, lack of Medicaid enrollment in children with special health care needs or lack of knowledge about resources (Women, Infant, and Children [WIC], Supplemental Nutrition Assistance Program benefits, or food pantries) and immediate food support needs, including food pantry locations, were targeted areas of support. Families also endorsed difficulty obtaining healthy food recommended by the provider (Produce to People), job instability surrounding hospital stays, or shifts in income as family members stayed home to take care of their child. This helped create our finalized resource sheet (Supplemental Fig 6) and ability to offer all FI families individualized resources.

As a balancing measure, we evaluated staff acceptance of screening. Our clinic had 26 cardiologists and 9 cardiology fellows at screening initiation. Four attending physicians had moved at the time of survey, and 4 did not have clinic at our main hospital. Twenty-three individuals completed the survey (completion rate of 85%). One hundred percent of respondents felt that implementation of FI screening never negatively impacted workflow. Seventy percent strongly agreed that FI screening is of great benefit to all patients, 30% agreed that FI screening benefited some patients, and no respondents responded neutral, disagreed, or strongly disagreed that FI screening was beneficial. All comments shared by physician respondents were positive (Supplemental Table 4).

FI was more frequent in patients with moderate CHD, although the difference was not statistically significant (OR 1.24 [0.85–1.81], P = .28). FI was most frequent in patients with severe CHD compared to any other group, and 7% of families (n = 106) screened positive for FI (OR 1.67 [1.21–2.29], P = .002) (Fig 5 A and B). This association held true regardless of race and/or ethnicity; however, minority patients with severe CHD may be particularly vulnerable (Supplemental Table 5).

To our knowledge, this report is the first describing implementation of FI screening in a pediatric cardiology clinic. We demonstrated the ability to reliably assess FI in a busy outpatient cardiology office. Our screening rate and FI identification improved using a written version of HVS, which mitigated provider-reported and subjective parental discomfort. There was overwhelming support of FI screening by our clinic physicians, and they did not feel their clinic workflow was negatively impacted by screening.

We reached our screening goal during our study period, but there were limitations in FI screening and risk assessment. Our study relied on human screeners. Implementation of an automated intake form directly imported into the EMR may further increase screening rates through improved LOR. A specified EMR location for screening helped improve our screening rate, but hard stops in the EMR may improve completion rates. Whereas we identified multiple families struggling with FI, our positivity rate did not reach that of Allegheny County (14%).²⁵  Although our clinic is located in Allegheny County, our referral basis is from surrounding areas and even outside states. Additionally, although we screened every patient to create normalization, there continues to be stigmatization surrounding FI.

Our primary outcome was to connect all FI families to resources. Early distribution of a generalized FI form resulted in limited success for follow-up and revealed that FI is not a 1-size-fits-all solution. Needs assessments uncovered underused resources, which exacerbated risk for FI. Collaboration with SW and community partners allowed our families to connect with local, individualized food resources and greatly improved our success in resource connection. Partnership with community programs may be a way to address FI issues in the absence of a SW. A meeting with our SW near the time of screening was integral to connecting families with appropriate resources and prevented loss to follow-up. An established FI team and SW allowed our clinic to have refined, individualized resources for families impacted by FI. As a result, our clinic served as a standard for screening and resource distribution for expansion to other clinics.

The results of our study reveal that patients with severe CHD may be at particular risk for FI. This held true across race and ethnicity groups, except those who did not specify race, and it highlighted that minority patients with severe CHD may be particularly vulnerable. This information was used to promote provider buy-in highlighting the risk to our most vulnerable clinic population. In addition, it helped focus individualized resource connection for moderate and severe CHD patients, including SW assessment for Medicaid enrollment and assessment surrounding hospitalizations, which were cited by families as a particular source of resource strain. A limitation to this analysis was that it did not address possible confounding factors, such as household income for FI families. Further studies are needed to assess FI in our patient population to expand on opportunities for intervention.

The impact of CHD severity on FI is not unexpected. Changes in financial stability may limit families’ ability to gain access to adequate food.^2,28  Families of children with cancer lost >40% of their annual household income 6 months after their child’s diagnosis.²⁹  Severe CHD patients undergo multiple procedures requiring hospitalizations with homebound recovery for varying lengths of time. As we continue to search for ways to reduce morbidity in our patient population, addressing the impact of complex CHD on social influencers of health (including FI) by creating individualized resources is integral. This can be done through community partnerships and social work support.

Whereas the majority of our patient population identified as White, our study highlights racial disparities in FI families in our clinic population. As a medical community, targeting racial inequities that may impact medical outcomes is essential. In CHD, low neighborhood socioeconomic status is associated with worse outcomes after the Norwood procedure, suggesting that socioeconomic factors are important, potentially modifiable determinants of outcome.³⁰  Further, FI families are more likely to delay medical care because of cost.^20,21  Thus, addressing FI in CHD patients by using the HVS may improve access to timely care in these high-risk patients.

FI screening is feasible in an outpatient cardiology office. Severe CHD patients may be at higher risk for FI, and more focused studies are needed for this population. As FI is a modifiable risk factor, identifying and addressing FI may improve outcomes for CHD patients. Community partnerships and SW support improves individualized, targeted FI interventions.

We thank the Heart Institute at CHP, including our social work team and our FI Task Force.

AAP

American Academy of Pediatrics

BMI

body mass index

CHD

congenital heart disease

CHP

UPMC Children’s Hospital of Pittsburgh

EMR

electronic medical record

FI

food insecurity

FLP

frontline providers

HVS

Hunger Vital Signs

LOR

level of reliability

PDSA

plan-do-study-act

QI

quality improvement

SPC

statistical process control