Characteristics Associated With Positive Social Determinants of Health Screening in Patients Admitted to Pediatric Hospital Medicine

This research was a descriptive retrospective cohort study of patient data conducted at a suburban, quaternary care, free-standing 130-bed children’s hospital with ∼5000 admissions per year and an average LOS of 2 days. In this population, 47.1% identify as white or Caucasian, 54.6% identify as Hispanic, 17% have a preferred language other than English, 72% of patients are insured by Medicaid, and 41% of patients are considered medically complex. This project was reviewed and granted an exemption by the institutional review board. From July 2020 to October 2021, a 14-question SDOH screening tool (v2019) was used to expand the social history of all pediatric patients (aged 0–18 years) admitted or transferred to the inpatient hospital medicine service. The SDOH screening tool was developed by our institution’s Value-Based Services Organization on the basis of modified questions from previously validated tools^7,17–21  by using input from families (Fig 1). The Value-Based Services Organization is an interdisciplinary team coordinating population health and care management initiatives. October 2021 was chosen as the endpoint, and the tool was subsequently revised.

Families were screened by members of the hospital medicine team, including attending physicians, nurse practitioners, residents, and medical students. Screening occurred at any point during the hospital stay, although the aim was to complete screening at admission to provide time for social work consultation. The screening was administered at the provider’s discretion to the families in paper format in English or Spanish or verbally in the family’s preferred language using a video interpreter if needed. Screening results were entered into the electronic health record in a standardized template. For this study, our team extracted the screener using a chart review based on the search term “social determinants of health” to detect the standardized template. A positive screen result was defined as answering “yes” to 1 or more questions. Social work consultation was offered to families who screened positive. Interventions included referrals to community resources. Through a pilot with a medical–legal partnership, we developed a workflow solution to allow for the dual enrollment of families into Medicaid and the Supplemental Nutrition Assistance Program. Assistance was provided for applications for financial assistance and other nutrition programs. Regardless of whether a social worker was requested, contact information for 211 (a nonprofit helpline for community services) was added to all discharge paperwork regardless of screening results.

Retrospective data extraction was performed for demographics (ie, age, sex, race, ethnicity, primary language, and insurance), membership in problem list registries (see Table 1 for a complete list of registries), and patient complexity (ie, 3M Clinical Risk Grouper clinical complexity score ≥5²²  and All Patient Refined-Diagnosis Related Group severity of illness [SOI] score and risk of mortality [ROM] score²³  for inpatient class). The criteria for inclusion in the medically complex registry was either a 3M complexity score of >5 or a manually entered “medically complex” problem list item. Acute care metrics included LOS, 7-day readmission rate, 30-day readmission rate, and the number of emergency department visits and hospitalizations between January 1, 2020, and December 31, 2021.

Continuous variables were summarized as medians and interquartile range (Q1–Q3), whereas categorical variables were reported as frequencies (percentages). Unadjusted comparisons were made between families that were not screened versus screened and between families with negative SDOH screen results (no needs identified) versus positive SDOH screen results (at least 1 need identified). Screening questions 1 and 2, related to food insecurity, were combined for statistical analysis.

Because some patients had multiple documented encounters, unadjusted and multivariable generalized estimation equation (GEE) logistic regression models were used to compare the 2 groups. Odds ratio (OR) estimates and 95% confidence intervals (CIs) were calculated from the models and interpreted as the multiplicative increase in the odds of having a positive screen result. P values < .05 were considered statistically significant, and all statistical tests were 2-sided. The multivariable analysis was adjusted for any baseline variable with a P value of < .05 in the unadjusted analysis.

GEE models were fitted for patient LOS, readmission rates, and 2-year acute care utilization. The models compared any positive SDOH screen to multiple positive SDOH screen result needs. The LOS outcome was analyzed with a generalized linear model, assuming a log link and constant coefficient of variance and fitted using GEE with an exchangeable correlation structure. For the binary outcomes (7-day and 30-day readmission rates), logistic regression models were fit by using GEE with an exchangeable correlation structure. For 2-year acute care utilization, first-visit observations were considered for the number of emergency department visits and hospitalizations. For these 2 outcome variables, the generalized linear model assuming a log link and constant coefficient of variance was fitted by using GEE with an independent correlation structure. Statistical analyses were performed by using R statistical software (version 4.0.3; R Foundation for Statistical Computing, Vienna, Austria).

Between July 2020 and October 2021, there were 3769 patient encounters with admission or transfer to the pediatric hospital medicine service. Of those, 2454 (65%) were screened during the hospitalization with a sustained rate of >70% in the last 10 months of the study. The odds of receiving a screening were not significantly different based on primary language (English compared with Spanish), insurance type, and inpatient versus observation stays (Table 1). There was also no difference in odds of being screened on the basis of filing time of history and physical (H&P; 7:00 am to 6:59 pm: 73.0% versus 7:00 pm to 6:59 am: 72.5% screened, P = .756). Of all the screens conducted, 72.1% were performed as part of the H&P, whereas the remaining were distributed among various other types of encounters, including progress notes, consult notes, and medical student notes.

Significant odds of not being screened in a multivariable analysis include primary language other than English or Spanish (OR 0.416, 95% CI 0.211–0.818), inclusion in the complex care registry (OR 0.800, 95% CI 0.652, 0.980), extreme SOI (OR 00.642, 95% CI 0.413–0.999), and extreme ROM (OR 0.276, 95% CI 0.136–0.561), compared with minor scores. Older age and 3M clinical complexity scores ≥5 were significantly different in the unadjusted analysis but no longer remained significant in the multivariable analysis.

There were significantly greater odds of being screened in families of Asian patients compared with white families (OR 2.946, 95% CI 1.251–6.938) and in families of Hispanic patients compared with non-Hispanic children (OR 1.276, 95% CI 1.067–1.525). Patients who were included in the asthma (OR 1.319, 95% CI 1.085–1.604) and hypertension (OR 1.579, 95% CI 1.058–2.356) registries also had higher odds of being screened in multivariable analysis.

Of the 2454 encounters screened, 662 (27%) families endorsed ≥ SDOH needs, with 311 (47%) families reporting ≥2 needs. (Fig 2) The most frequently identified needs were food insecurity (14.8%), assistance with paperwork (14.6%), financial insecurity (12.2%), and health literacy (verbal 9.0%, written 7.5%). Insurance type, patient’s sex, and inclusion in other disease registries were not found to be associated with increased odds of SDOH screening. (Table 2)

The highest significant odds of screening positive for SDOH needs in the multivariable analysis was if the primary language was other than English or Spanish (OR 4.269, 95% CI 1.731–10.533). A primary language of Spanish was also associated with higher odds of screening positive (OR 1.419, 95% CI 1.050–1.918). Other factors associated with increased odds of screening positive included patients identifying as Black (OR 1.675, 1.237–2.266), compared with white patients identifying as Hispanic (OR 1.347, 95% CI 1.057, 1.717), compared with non-Hispanics, and families with patients in the complex care registry (OR 1.466, 95% CI 1.120–1.918). Children in the hypertension (OR 0.453, 0.254–0.809) and obesity (OR 0.758, 0.588–0.977) registries were significantly less likely to screen positive.

Three GEE models were used to analyze hospital utilization, and none of these metrics were found to be significantly different between families who screened negative versus positive (LOS, 7- and 30-day readmission rates, 2-year acute care utilization; Table 3). The average LOS for families with negative screen results was 43 hours (Q1–Q3: 24–70) versus 46 hours (Q1–Q3 25–82) for positive screen results (OR 1.028, 95% CI 0.886–1.193). The 7-day readmission rate was 6% versus 8% (negative versus positive; OR 1.193, 95% CI 0.787–13.819). The 30-day readmission rate was 18% versus 19% (negative versus positive) (OR 1.083, 0572–2.049). The average number of emergency department visits from January 2020 to December 2021 per family was 1 (Q1–Q3: 1–3) versus 2 (Q1–Q3: 1–3; negative versus positive; OR 0.968, 95% CI 0.832–1.127). The average number of hospitalizations in the same period was 1 (Q1–Q3: 1–2) versus 1 (Q1–Q3: 1–3; negative versus positive; OR 0.997, 95% CI 0.898–1.108). In addition, we stratified our analysis by isolating families with multiple SDOH needs and found no significant associations in any metric (Table 3).

SDOH Screening has been encouraged by the American Academy of Pediatrics and other pediatric advocates. Changing health care reimbursement structures away from fee-for-service is pushing health systems to increase SDOH screening to identify modifiable needs and address unmet SDOH needs in different settings.²⁴  In 2023, the Centers for Medicare and Medicaid Services adopted SDOH needs as a component of medical decision-making in the inpatient setting.^25,26  However, in a 2020 survey conducted by Schwartz et al at 4 children’s hospitals, only 29% of hospitalists and 41% of nurses regularly screened for SDOH needs.²⁷  Cited barriers to screening included a lack of time, training, resources, and standardized screening tools²⁷ . The availability of skilled staff for screening has also limited studies of SDOH screening to business hours from Monday through Friday using research assistants or pediatric residents.^28–31  There has been a self-reported potential bias of significantly less screening in publicly insured patients due to business hours.²⁹  When screened, publicly insured patients were more likely to screen positive (57% versus 43%, P < .01).²⁹  One study did not include non-English speakers because of inadequate staffing.³¹ 

A strength of this study was that we screened patients at all hours of admission. Including an SDOH screener as part of social history taking by a team of pediatric hospitalists, nurse practitioners, residents, and medical students achieved a sustained 70% screening rate of all admissions and transfers, with no difference in screening rate based on the timing of H&P. We did not show a difference in the odds of screening or in screening results based on insurance, race, or Spanish versus English speakers. Increasing patient complexity and language other than English or Spanish resulted in a decreased likelihood of completing screening. Significant differences in screening rates in patient characteristics may be due to availability bias and selection bias. Other protocolized studies have also been met with the same limitations in screening rates in these demographics.^28–29  Barriers to screening could include the time required for providers to spend with the families in interpreting and processing clinical data, the availability of caregivers at the bedside, general workload, or origination in the intensive care units that resulted in a different intake workflow.

In our population, 27% (n = 662) patient encounters screened positive for ≥1 SDOH need, with 47% of those reporting ≥2 needs. In comparison, the authors of 3 studies of SDOH screening in hospitalized pediatric patients identified positivity rates of 33%,²⁸  38%,²⁹  and 83%.³¹  Differences in rates may be due to underlying population characteristics, screening tools, or screening methods. Vaz et al²⁸  screened 265 families and found that the most frequently identified needs were financial stress and medical bills. Fritz et al²⁹  screened 374 hospitalized families and identified the most common needs as financial, difficulty making health care appointments, and benefits. Lopez et al³¹  screened 417 mothers, with the most prevalent needs being employment, health insurance status, and needing a primary care provider. We identified that >50% of our community’s needs were related to food insecurity, assistance with paperwork, financial insecurity, lack of social support, and health literacy. These findings suggest solutions involving public assistance and social services, along with opportunities for the innovative training and education of health care providers regarding communication styles and modes.

Variations in screening tools have made populations difficult to compare. The US Preventative Services Task Force’s position paper in 2020 advocated consolidating various tools to create a validated set of questions and improve the ability to examine research on outcomes across different studies.³²  However, to date, the American Academy of Pediatrics has recommended no such tool. This leaves institutions that desire to implement SDOH screening to develop new screeners or adopt or adapt previously existing tools, which may place emphasis on different domains and result in differences in positivity rate based on content.

Leary et al³³  surveyed hospitalized families and found that families preferred to be screened in person. Our screens were conducted in a mix of face-to-face interviews and paper format. The method of administration was not collected. Gottlieb et al³⁴  compared face-to-face screening with a private electronic format. The disclosure of more sensitive topics, such as domestic violence, substance abuse, and financial insecurity, was higher in a private electronic format. Positive screening results in the electronic group (70%) were higher than in the face-to-face group (30%).³⁴  Face-to-face screening may result in underestimating social needs because the parents or guardians are more likely to answer with socially desirable responses.³⁵ 

Leary et al³³  also found that families may prefer to wait until later in the admission process to be screened. However, we found that if patients were not methodically screened at admission, they were more likely not to be screened because we seldom found SDOH screens documented after admission. Additional workflows must be studied to match family preference to clinical practice.

With our research, we hope to help guide health systems that do not have the resources to universally screen patients for SDOH needs to help identify most at-risk populations served by a health care group. Factors such as language, medical complexity, and demographics may be used as a kind of proxy to aid in decision-making regarding where to begin funneling resources, screening, and research, keeping in mind this is not a clear substitute for specific SDOH screening.

Although “other primary language” resulted in the highest odds of screening positive for SDOH needs, speaking Spanish also resulted in higher odds of screening positive. Preferred language has been associated with increased health care costs, even after adjusting for LOS and medical complexity.³⁶  We did not investigate English proficiency. Limited English proficiency has been shown to have a dose–response relationship with poorer socioeconomic status and health care behaviors.^37–38  Language barriers may further limit families’ knowledge, referral, and access to community resources, as evidenced by our own screening rates. With domains in health literacy combined being our second highest identified need, we emphasized the use of interpreters during rounds, medical translation applications for prescription instructions, referrals to English proficiency classes, and consultation with case workers to guide the identification of community resources. In addition, offering the health care web portal in languages other than English is being addressed at an institutional level to mitigate disparities.

Nationally, CMC comprise only 6% of the pediatric population but account for ∼40% of pediatric health care spending, providing an important opportunity for cost savings.³⁹  Although patients in our complex care registry were more likely to screen positive for SDOH needs, higher 3M complexity, SOI, and ROM scores, while significant in the unadjusted analysis, were no longer significant in the multivariable analysis. This may be a result of underscreening, especially in children with extreme SOI and ROM scores. We hypothesize that this may be due to the process of intensive care transfers rather than direct hospital medicine admission (ie, transfer note type versus H&P) or lack of caregivers at the bedside. In 1 study, 6 of 10 caregivers of CMC endorsed at least 1 SDOH concern.⁴⁰  The authors of previous studies have indicated that parents of CMC have higher rates of psychological stress,^41–42  housing,⁴³  and financial insecurities.^44–46  In the inpatient setting, previous studies have also revealed an increased likelihood of positive screening in CMC.^28–29  We suspect that SDOH needs may be similar or even higher for patients or families who are unable to be at the bedside to participate in SDOH screening. Collaborative models between interdisciplinary clinicians and community health workers can significantly decrease admissions, bed days, and costs.^47–49  This is an area worthy of further intervention and study.

Our screened population also had increased odds of positive screening results based on race and ethnicity. Although we report on race and ethnicity in this study, we fully acknowledge that they are social constructs.⁵⁰  However, historical socioeconomic forces have impacted patterns of racial and ethnic disparities in the United States. Structural racism, manifested through practices such as redlining and insufficient funding for public education, is likely a contributing factor to the observed disparities in race-related outcomes in screening positive for SDOH needs. Hence, these results should be taken in context with other social risk factors. These differences are shown to help direct thought to help advance equity in health care delivery and health outcomes.

Only 1 other study has investigated the relationship between direct SDOH screening and hospital metrics in the literature review. Fritz et al showed that families with a positive screen result had a higher 30-day readmission rate (10% vs 5%, P = .05).²⁹  Contrary to our expectations, we did not find a positive SDOH screening result or multiple positive answers on a screen to be associated with significant differences in LOS, readmission rates, and 2-year health care utilization.

Despite these findings, societal and individual returns from addressing SDOH needs may be realized beyond the hospital setting or the time limits of this study. Specific domains may have greater associations, but this analysis would need further study with a larger sample size.

Families that screened positive were offered social work consultation, which included interventions such as referrals to community resources. We did not have the resources to follow up with families to evaluate the effects of provided resources. How our interventions confounded our acute care metrics is unclear. Some of these interventions may have mitigated some risks and reduced differences in our health care metrics. However, it would have been unethical to screen and not intervene. We hope to expand the follow-up of resource utilization.

This research was conducted at a single institution, which is 1 of the 3 major pediatric hospitals located in the area. Patients discharged from our hospital may have sought care at other facilities, leading to missed readmissions. In 1 study, using single-center readmission data may have missed as many as 14% of readmissions.⁵¹  It should also be noted that the screening method, whether paper-based or oral, was chosen at the provider’s discretion and could introduce variability in the results.

SDOH screening took place from July 2020 to October 2021. Data for health care utilization (LOS, readmissions, ED visits, and hospitalizations) are from January 2020 through December 2021. It was thought that this timeframe for health care utilization could be representative and comparable to the population screened for SDOH. Because the timeframes are not exactly the same, however, this could introduce confounding and measurement biases in the populations.

The generalizability of our findings may have been affected by the coronavirus disease 2019 pandemic during the study period. Locally, we had a 25% decrease in admissions in 2020 compared with 2019. Although coronavirus disease 2019 disproportionately affected those with higher needs, families were affected across the economic spectrum, which may have led to a higher percentage of patients screening positive.⁵² 

Providers can conduct large-scale SDOH screening in the inpatient hospital medicine setting. Populations at the greatest risk for screening positive include families whose primary language is not English and those with children who have medical complexities. Positive SDOH screening results in the inpatient setting may not be associated with differences in LOS, readmission rates, and 2-year acute care utilization.

We would like to thank the following for their contributions to this study: Nemours Children’s Hospital Departments of Pediatric Hospital Medicine and Graduate Medical Education, Pediatric Residency Classes of 2022 to 2024, including Akrishon Kirk, MD, University of Central Florida Medical Students, Kanekal Gautham, MD, the social work team, with special recognition to Elizabeth Wester and Brenda Marin, Dan Eckridge from Nemours Enterprise Intelligence, and Na-Tasha Williams and Kelli Thompson from Nemours Value-Based Services Organization.