The Role of Caregiver-Reported Risks in Predicting Adverse Pediatric Outcomes Free

Abstract

Objective

Certain social risk factors (e.g., housing instability, food insecurity) have been shown to directly and indirectly influence pediatric health outcomes; however, there is limited understanding of which social factors are most salient for children admitted to the hospital. This study examines how caregiver-reported social and medical characteristics of children experiencing an inpatient admission are associated with the presence of future health complications.

Methods

Caregivers of children experiencing an inpatient admission (N = 249) completed a predischarge questionnaire designed to capture medical and social risk factors across systems (e.g., patient, caregiver, family, community, healthcare environment). Electronic health record (EHR) data were reviewed for child demographic data, chronic disease status, and subsequent emergency department visits or readmissions (i.e., acute events) 90 days postindex hospitalization. Associations between risk factors and event presence were estimated using odds ratios (ORs) and confidence intervals (CI), both unadjusted and adjusted OR (aOR) for chronic disease and age.

Results

Thirty-three percent (N = 82) of children experienced at least one event. After accounting for child age and chronic disease status, caregiver perceptions of child’s health being generally “poor” or “not good” prior to discharge (aOR = 4.7, 95% CI = 2.3, 9.7), having high care coordination needs (aOR = 3.2, 95% CI = 1.6, 6.1), and experiencing difficulty accessing care coordination (aOR = 2.5, 95% CI = 1.4, 4.7) were significantly associated with return events.

Conclusions

Caregiver report of risks may provide valuable information above and beyond EHR records to both determine risk of future health problems and inform intervention development.

Returns to the hospital or emergency department (ED) following discharge represent undesired pediatric health outcomes that negatively impact children and families while disproportionately affecting certain populations (e.g., children on public insurance, children of color). Hospital and ED returns can disrupt routine and family functioning, cause distress, signal health deterioration, and hinder school and work attendance (Nakamura et al., 2014; Van Horn & Kautz, 2007). Approximately 2–6% of child hospitalizations result in readmissions within 30 days (Berry et al., 2013; Jiang & Wier, 2006), and about 22% of children admitted to children’s hospitals experience a readmission within one year (Berry et al., 2011; Nakamura et al., 2014). Children with chronic and/or complex medical conditions and those of younger age (e.g., infants) are particularly at risk for poor health outcomes and associated ED visits and/or readmissions (Kuzniewicz et al., 2013; Leyenaar & O’Brien, 2017; Ray & Lorch, 2012; Stephens et al., 2017). Interestingly, while pediatric admission rates seem to have decreased over time, those for children with chronic and/or complex medical conditions have actually increased (Bucholz et al., 2019). In addition, children are more likely to have higher rates of ED visits or readmissions if they are publicly, under-, or uninsured, of minority racial and ethnic background, experience lower socioeconomic status (SES), or have diagnosed mental health concerns (Berry et al., 2011; Jiang & Wier, 2006; Kenyon et al., 2014; Lavigne & Meyers, 2019; Parikh et al., 2017; Riese et al., 2014). These associations are especially concerning given that children are at high risk of living in poverty (17.5%) and often have inadequate health insurance coverage (19–42% across states; Office of Disease Prevention and Health Promotion, 2019). Taken together, these findings indicate that ED visits and readmissions are the result of a combination of medical and social factors.

Social determinants, “the conditions in which people are born, grow, live, work, and age,” shape the health of a population (Marmot et al., 2008) and have a significant effect on family functioning and pediatric health, with an estimated 80% of health outcomes resulting from social, economic, and behavioral factors (Fieldston et al., 2013; Hood et al., 2016; Office of Disease Prevention and Health Promotion, 2019; Raphael & Beal, 2010). Social determinants, such as SES, shape family and caregiver social needs (e.g., housing, utilities, transportation) which in turn can affect whether and how child healthcare is accessed, delivered, and received (National Academies of Sciences, Engineering [NASEM], 2019; Office of Disease Prevention and Health Promotion, 2019). Although social needs are subjectively defined, social risks are objective and include adverse social contexts known to increase the likelihood of a negative or poor health outcome (Gottlieb & Alderwick, 2019). Health disparities arise when exposure to chronic or multiple social risks (e.g., food insecurity, housing instability, low education level) disproportionately affects specific, vulnerable populations (e.g., children, those with low SES and limited resources; Alvidrez et al., 2019; American Psychological Association, 2019; Kilbourne et al., 2006).

There is an emerging recognition that integrating screening for social risk into healthcare delivery can begin to address and reduce preventable health disparities for populations experiencing social vulnerability (Marmot et al., 2008; Modi et al., 2012; NASEM, 2019). To effectively develop tools to assess social risk and intervene, researchers must first determine the most relevant variables and their relative contribution to preventable health disparities. However, research on the influence of social risk on health disparities is limited by a number of factors. For example, there is a disproportionate focus on screening and intervention in the primary care setting (American Academy of Pediatrics, 2016; Garg et al., 2012; Gottlieb et al., 2016; Pantell et al., 2019). While meaningful, this research focuses on a population that typically can already access and engage in preventive services and may exclude families at greater risk who rely on more acute care settings (e.g., ED visits) or access care only when their child is extremely ill (e.g., admitted to hospital). Also, many studies are cross-sectional or retrospective in nature, limiting their ability to determine how social risks may influence future health disparities (Karlson & Rapoff, 2009). In addition, studies often rely on electronic health records (EHRs) to understand social risk prevalence and corresponding associations with health outcomes. EHR-based social risk data often (a) are buried in problem lists, social histories, and diagnoses; (b) may be unreliable or incomplete; (c) are dependent upon systematic screening efforts; and (d) may not reflect social risks in real time (Navathe et al., 2018; Oreskovic et al., 2017). Few studies have explored the association of specific caregiver-reported social risk data on health outcomes in the pediatric inpatient setting (Desai & Starmer, 2019; Jani et al., 2019; Lammers et al., 2019), and these efforts rarely capture the potential interactions between systems (Clauss-Ehlers et al., 2019; NASEM, 2019).

Several theoretical models have been proposed to explain the role of multiple factors on pediatric outcomes (Fieldston et al., 2013; Kazak, 1989), and those considered most comprehensive are rooted in the bioecological model (Bronfenbrenner, 1979, 2006). This model postulates that individuals’ outcomes are influenced by the systems (e.g., family, neighborhood, school system, healthcare environment) in which they are embedded and the interactions between these systems. This model accounts for the interplay between social, economic, behavioral, and biological factors known to impact health behaviors, improving upon more reductionistic models, especially those which rely solely upon individual factors for predicting health outcomes. Such a framework is useful to apply when attempting to understand what is known about the relationships between social factors and future health disparities. In particular, this model captures the influence of both proximal (caregiver and family risk) and more distal social risks (health system and community) and is the guiding framework for pediatric interventions (e.g., Multisystemic Therapy [Ellis et al., 2008]; Novel Interventions in Children’s Healthcare [Harris et al., 2013]) designed for youth with high social risk experiencing preventable health problems.

To address research gaps, our team (Vaz et al., 2020) assessed the feasibility of inpatient pediatric social risk screening via caregiver report and the prevalence of caregiver-reported risk in the inpatient setting. Consistent with Kilbourne et al.’s (2006) phases of detecting prevalence of and understanding contributing factors to health disparities at multiple levels, we designed a questionnaire informed by socioecological models of behavior that was intended to capture medical and social risk across systems. This study found high acceptability of social risk screening among caregivers of children experiencing an inpatient admission and demonstrated that these caregivers reported a high prevalence of socioeconomic stress (34%) and caregiver mental health concerns (33%). Additionally, social risks were uniformly common and did not disproportionately affect those with chronic and/or complex medical diseases, such as children with seizure disorders, chronic lung disease, or congenital anomalies.

Study Aims and Hypotheses

In this study, we conducted a longitudinal follow-up of the children included in Vaz et al. (2020) to assess which individual and grouped risk factors were associated with ED visits or readmissions in the 90-day period postdischarge. We obtained numbers of these events from medical records. We then used data from questionnaires and EHR review to identify potential risk factors, both separately and aggregated across systems (i.e., patient, caregiver, family, community, healthcare system). Our first aim was to identify which risk factors were most strongly associated with risk of future health complications (i.e., as evidenced by ED visits or readmissions). Second, we explored whether using factor analysis methods to generate domain scores, in combination with counts of events as an outcome, would be informative about these risks.

We hypothesized that while certain previously studied individual factors (e.g., chronic medical condition presence, child age) will be strongly associated with future health problems, specific systemic risks most proximal to the child and family (e.g., microsystem risks, such as caregiver mental and physical health,) will predispose families to have greater difficulty managing child health in outpatient settings above and beyond the influence of disease status and child age, and result in future complications.

Materials and Methods

Participants

Participants included child–caregiver dyads who were recruited and enrolled during admission to a pediatric hospitalist service at an academic children’s hospital in the Pacific Northwest (for a thorough description of Methods, see Vaz et al., 2020). This study sample is representative of the type of population served by the hospitalist service in this region, with attempts to address potential bias in sampling including daily recruitment 7 days per week and exclusion of the medical team from consent and questionnaire administration. Caregivers had to be 16 years or older (i.e., recognized as either an adult or emancipated minor in the state of Oregon), be the parent or legal guardian accompanying a child (i.e., <18 years of age) experiencing a hospital admission, and have English or Spanish proficiency. Eighty-seven percent of caregivers (N = 265 of 304) approached during the study period consented to participate. Caregivers (N = 39) who declined participation shared similar child demographic characteristics with those who consented (age, sex, race/ethnicity, and insurance). Of those consented, 10 caregivers did not complete the predischarge questionnaire, and six were excluded due to changes in guardianship or discharge to another facility after enrollment. A total of 249 caregiver–child dyads were analyzed (94% retention rate), with caregiver age ranging from 16 to 70 years (median, 35 years). The majority of caregivers was non-Latino White (75%), biological or adoptive mothers (81%), and had a high school education or less (52%), all of which are generally representative of statewide demographic patterns (United States Census Bureau, 2019). Child age ranged from 2 days to 17.9 years (median, 3 years).

Procedures

Study procedures were approved by the Institutional Review Board at the study site. During their child’s inpatient stay, participating caregivers completed a brief self-administered social risk questionnaire and were compensated with a $5 gift card. The questionnaire was available in English or Spanish and was designed to a fifth-grade readability level. The child’s EHR was reviewed to capture chronic diseases (CDs) and basic demographic data present during their index inpatient stay and any acute utilization (ED or readmission) 90 days postdischarge.

Measures

Caregiver-Reported Social Risk

Development of the social risk questionnaire was influenced by existing outpatient social screening tools from the literature, informed by our prior clinical and research efforts with youth experiencing avoidable health outcomes (Barry et al., 2017), and adapted for the needs of this study and population of interest (see Vaz et al., 2020). In addition to demographic information, caregivers responded to 26 root questions capturing medical (e.g., caregiver perception of child health, previous healthcare utilization) and social (e.g., food insecurity, family financial status) factors. These were based on published items from the National Survey of Child Health (Data Resource Center for Child and Adolescent Health, 2017), Children’s HealthWatch Survey (Children’s HealthWatch, 2017), social risk batteries created by Gottlieb et al. (2014) and Harris et al. (2013, 2015), and the adverse childhood experiences (ACEs) questionnaire (Centers for Disease Control and Prevention, 2017). Briefly, the survey contains both multiple-choice (N = 10; e.g., “What is the most common way that you get your child to clinic or hospital visits? A: Personal vehicle B: Ask a family member or friend C: Taxi D: Bus, etc…”) and dichotomous questions (N = 14; “Would you say that type of transportation is reliable [meaning you can count on it working 100% of the time]?”). For consistency in this analysis, variables were coded as either 0 (no risk identified) or 1 (risk identified), which in a few cases meant combining multiple-choice categories as reflected in their descriptive labels (e.g. “Child health ‘not good’ or ‘poor’”). Given the potential sensitivity of the ACEs questions (exploring witnessing violence in home/community, family member death, and substance misuse), items which asked ACEs were coded with an additional “prefer not to answer” option. These responses were treated as missing. Most (94%) respondents answered all 10 ACE items with either yes or no.

System and Subsystem Categorization

Questionnaire responses were categorized in systems (e.g., microsystem, exosystem) and subsystems (e.g., family, caregiver employment), informed by previous studies categorization of factors within systems (Graves & Sheldon, 2018; Paat, 2013; Senefeld & Perrin, 2014; Wagner et al., 2019).

EHR-Derived Social and Medical Risk

Demographic data (e.g., primary language, child age, caregiver race/ethnicity) and child medical characteristics (e.g., specialty services involved) were collected from the EHR. Child medical complexity/chronic illness status was assigned using the Pediatric Medical Complexity Algorithm (PMCA version 3.1) based on ICD9/10 codes from the child’s EHR problem list and discharge diagnoses and was categorized as (a) no CD (e.g., bronchiolitis), (b) noncomplex CD (NC-CD; e.g., type 1 diabetes), or (c) complex CD (C-CD; e.g., child with cerebral palsy, seizures, and requiring tube feedings; Kaiser Permanente Washington Health Research Institute, 2019; Simon et al., 2014). The PMCA is designed to assess disparities in care according to level of medical complexity for children with special healthcare needs and is well validated (Simon et al., 2018). Children were further dichotomized as either having any CD (NC-CD or C-CD, N = 117) or no CD (no-CD, N = 132). This dichotomization was due to sample size considerations and limitations of EHR-based diagnoses.

Both based on previously referenced literature (Berry et al., 2013; Bucholz et al., 2019; Goudie et al., 2014; Kuo et al., 2011) and our own examination of these relationships (Vaz et al., 2020), analyses specifically accounted for how CD and child age influenced these relationships. Infancy and chronic condition presence were associated with event presence and inversely correlated with each other. Given that chronic conditions are likely under-ascertained in infants (e.g., diagnosis is forthcoming, may not have developed yet, are under investigation, or may evolve over time [Compas et al., 2012; Perrin et al., 2014]), a combined three-level variable (infant <1 year old at index admission, child with chronic condition, and child with nonchronic condition) was created.

Outcomes

Each child’s EHR was reviewed at 90 days postdischarge for ED visits and hospital readmissions during this time period. Greater than 90% of the hospitals in this state share an EHR that allows for access to regional hospital or ED visits (EPIC, 2019).

Data Analysis

Primary outcomes were presence of ED visits and hospital readmissions 90 days postindex hospitalization discharge. This “event” variable was coded as present if a patient had at least one ED visit or readmission in the 90 days postdischarge.

Odds ratios (ORs) were estimated for each risk factor individually using logistic regression, both alone and adjusted OR(aOR) for age/CD in the three-level categorical formulation, to identify individual risk factors for readmission or ED visits after hospitalization.

To examine correlations of specific variables within systems and subsystems, we generated a tetrachoric correlation matrix for each a priori set of variables in a system, fit a single factor using that matrix, and reviewed the eigenvalue and factor loadings. Factor scores were generated which were then rescaled to the interval between 0 and 1 so regression coefficients would reflect the difference between the lowest and highest factor scores for each system. For each factor, we fit a negative binomial regression model with number of events as the dependent variable, including the factor score and indicator variables for infants or children with chronic conditions. The exponentiated coefficient for the factor score is the rate ratio for a one-unit increase in the factor score, which has been scaled as the highest versus the lowest, whereas the intercept is the rate of events for children without chronic conditions and low to no social risk factors in the given domain. A few of the items did not fit with others in systems and were evaluated on their own. This approach allowed us to explore relationships as we might with a structural equation model, which we considered but were unable to fit to this dataset.

Statistical analyses were performed using Stata/IC version 15.1 (StataCorp LLC, College Station, TX) software for Windows. The PMCA algorithm was executed in SAS version 9.4 (SAS Institute, Cary NC) using the macro (version 3.1) published by its authors (Simon et al., 2014, 2018). Missingness was evaluated for each system and observations were omitted pairwise; for all but one system, this was <5% of the sample, and the maximum was 7%.

Results

Among 249 children, 82 (33%) experienced one or more ED or readmission events in the 90 days postdischarge. Forty-three children (17%) experienced at least two events (either ED or readmission). The total number of readmissions per child ranged from 0 to 5, and ED visits per child ranged from 0 to 10. Baseline frequencies regarding prevalence of caregiver-reported social risks were previously published (Vaz et al., 2020). Both infancy (child <12 months of age; OR 2.1, 95% CI = 1.1, 3.9) and chronic illness status (OR = 2.2, 95% CI = 1.3, 3.7) were associated with an ED or readmission event. There were no significant differences found among child/caregiver demographic characteristics and an ED or readmission event, even after adjusting for CD status and age (Table I).

EHR and social risk factors from the questionnaire were classified into systems, with the OR for any ED or readmission event adjusted for chronic illness and infancy status (Table II). Among individual characteristics, a caregiver’s description of their child’s health prior to discharge as “not very good or poor” was most associated with an event (aOR = 4.7, 95% CI = 2.3, 9.7) even after adjusting for infant status and CD.

Overall, most microsystem-level variables (e.g., caregiver health, family organization) were not associated with an event. Under the schooling subsystem, missing more than 10 days in the prior 12 months was associated with an ED or readmission (aOR = 2.6, 95% CI = 1.1, 6.5).

In contrast to our hypothesis, the highest number of associations with an ED or readmission event occurred within the mesosystem, specifically the family medical system interaction. Caregiver report of needing help with coordinating care in the previous 12 months and high care coordination needs (>5 h/week providing healthcare and managing care coordination) were associated with an event (aOR = 2.5, 95% CI = 1.4, 4.7; aOR = 3.2, 95% CI = 1.6, 6.1). More than one ED visit or readmission in the 12 months prior to index admission was also associated with future ED visits or readmission. There were no significant associations between individual social risks within the macro- and exosystems and future presence of an ED visit or readmission event (Table II).

When the individual elements were combined into subsystem factor scores (Table III), we again found that elements seemingly related to CD were associated with higher mean numbers of ED or readmission events. Children without chronic conditions averaged 0.4 events each in the 90 days after index admission; children with chronic conditions experienced 1.6 times as many events as indicated by the incidence rate ratio (IRR), and infants experienced 3 times as many events. After adjusting for infancy and chronic condition presence, children whose health was described as “poor” had 5 times as many events as those whose health was “excellent.” Children with high factor scores for outpatient and inpatient visits in the previous year experienced 1.94 times as many events; having reliable transportation, as proxy to how families are able to access to care, seemed almost unrelated to these visits (factor loading 0.01). Children requiring 5 or more hours of care coordination had IRRs of 2.13 and 2.04, respectively. Factors for social risks, such as ACEs (IRR 1.55, NS) and problems paying for food, housing, or utilities (IRR 1.42, NS) were associated with higher rates of ED/readmission events, but not at a statistically significant level. Some items reflecting challenges for caregivers were associated with lower rates of readmission events, including parent poor health (IRR 0.25), split custody arrangements (IRR 0.44), employment concerns (IRR 0.86), and parent minority race (IRR 0.64), but were not statistically significant.

Discussion

This study represents a first effort to characterize the association between caregiver-reported social risk and the long-term health outcomes of children experiencing an inpatient admission, utilizing a framework designed to look at specific risks at a system level. Our original hypothesis that proximal influences in the microsystem (related specifically to the caregiver/family/school), were not substantiated. Instead, we found several specific items among each of the systems (e.g., help needed for coordinating care [meso], missing more than 10 days of school in past year [micro], caregiver perception of child’s health as poor [individual]) were significantly associated with presence of future ED visits and readmissions, even after adjusting for chronic illness and age. Multiple mesosystem factors, however, were significantly associated with likelihood of experiencing a return to hospital.

When examining items within the mesosystem, those with seemingly the strongest impact on return visit likelihood arguably represent previous child health status. Thus, combined with other significant findings, it appears that the best potential predictor of future health events, even after accounting for CD and infancy, are items representing the child’s past and current health status. This corroborates with findings in the literature regarding predictability of readmissions/ED visits from prior healthcare utilization (Feudtner et al., 2009; Leary et al., 2019; Poole et al., 2016).

Two questions asked of caregivers, one related to perception of the child’s health (individual system) and the other, the ability to coordinate needs for the child (mesosystem), reflect real-time responses at the time of discharge and may be meaningful as a point of intervention for medical teams. The caregiver-reported child health status, is a general summary of children’s health prior to discharge, and also seems meaningful. Poorer overall child health status has been associated with other markers of child health disparity (Alaimo et al., 2001; Mehta et al., 2013). Perhaps a caregiver’s so-called “gut instinct” about their child’s health may influence care-seeking behavior postdischarge (Brittan et al., 2015; Urbane et al., 2019), or may simply represent additive child vulnerability information beyond medical coding. Caregivers may feel they have a special understanding of their child’s norms. Coupled with some clinical knowledge and expertise in their child’s clinical history and clinical condition, this may promote an intuitive sense of the child’s wellbeing (Birchley, 2015). This parental concern regarding child health, a subjective feeling that “something is wrong” or that their child is “unwell,” highlights that this particular episode of illness is different from previous episodes, and may promote additional care-seeking behavior.

Limitations

The uniqueness of this study includes the prospective nature of this cohort, the collection of social risk information from the caregiver themselves, as opposed to the EHR or other large database collection, and the length of time for follow-up (90 days, compared with 30 days, which is more commonly used as a hospital-quality metric but may be too short to identify long-term effects of a particular social risk). Although our findings do not coincide with previous research suggesting that demographic and social risk factors such as insurance status, SES, race and ethnicity, and the presence of a mental health concern predict future hospital returns (Doupnik et al., 2018; Parikh et al., 2017; Stephens et al., 2017; Victorino & Gauthier, 2009), this may be due to several limitations.

Currently, there are no validated or widely used social risk screening questionnaires for children experiencing an inpatient admission. As such, our questionnaire was not a validated measure and was a first attempt to understand the breadth of social risk in this population. Despite high prevalence of particular social risk factors in this sample (e.g., food insecurity, caregiver mental health diagnosis Vaz et al., 2020), it is possible that this sample may not be representative of the type of children at higher risk for poor outcomes, limiting our ability to draw conclusions about this group. Furthermore, it is possible that the length of our follow-up did not allow enough time for some of the impact of social factors to result in return to hospital. Finally, the relatively small sample size for this exploratory analysis limited the number of variables that could be included in a single model and meant that only relatively large effects (roughly speaking, OR > 2) for more common risk factors were statistically significant at the 0.05 level. With a much larger sample in future studies, we might estimate the effects of less prevalent risk factors with more precision and use approaches such as SEM to better understand how risk factors relate to each other.

Perhaps the heterogeneity of our sample in terms of age range and medical conditions made it difficult to detect significant contributions of these demographic and social factors. Our sample was largely homogeneous in terms of race and ethnicity, which likely hindered our ability to find significant associations between race and ethnicity and future events.

Additionally, we examined all cause readmissions, regardless of whether they could have been prevented. Aside from “ambulatory sensitive conditions,” (e.g., hospital admission is potentially preventable by timely and effective ambulatory care; Bettenhausen et al., 2017; Sanderson & Dixon, 2000), currently there is no validated rubric or measure to classify pediatric preventable events (ED or readmission). Attempts at doing so have focused on shorter times to readmission (e.g., 7 and 30 days) and have issues with inter-user reliability (Gardner et al., 2020; Hain et al., 2013). A deeper dive to examine nonmedical contributors to a returned visit may uncover more social risks. Also, caregiver-reported social risk may be biased or unreliable, although it may also be more sensitive than querying the EHR alone. Some important nonmedical risk factors (e.g., high parent–child conflict, family disorganization, provider biases) were not included. Further, reliance upon EHR data for presence and number of acute utilization events likely under captures true utilization; however, 90% of hospitals in this state share an EHR. In addition, coding CD dichotomously based on EHR limitations likely impeded our ability to capture the true strength of medical risk on future events. Although our sample reflects a typical breadth and depth of inpatient pediatric medicine service, future studies would benefit from larger samples, longer follow-up periods, inclusion of multiple regions and healthcare systems, and firmer cutoffs related to patient age and medical condition.

Clinical Implications

Two of the seemingly main takeaways from this study are to appreciate the potential strength of medical system–family interactions as well as to value the perceptions of caregivers in addition to physical findings and medical history. Given the findings, medical professionals, pediatric psychologists and other behavioral health providers may consider including some of the brief questions shown to predict readmissions or ED visits, including the caregiver’s perception of their child’s health and need for care coordination. These may represent opportunities for timely intervention prior to discharge. When endorsed, providers should further assess the perceived meaning of this for the caregiver and their perception of how to best address this, including incorporating social workers or care managers into the care of the child. In addition, future research efforts should examine the mechanism by which these statements represents increased risk, either by highlighting whether a caregiver will rely upon ED visits when concerned for child health, whether it represents a caregiver’s concern over the ability of the medical team or family to adequately address the child’s health needs, or whether this represents a more nuanced understanding of their child’s health above and beyond their chronic/complex medical status. Taken together, these items may represent candidates for future brief screening measures, further exploration, and possible intervention targets.

Conclusion

This study reflects a first attempt to understand risk factors across a heterogeneous (i.e., in age and condition) population of children experiencing an inpatient admission within a health disparity framework (Kilbourne et al., 2006). Regarding Kilbourne’s phase one (i.e., detection of disparities), systematic social risk screening of children experiencing a hospital admission promises to include families unable to easily access office-based care and related research, augmenting the capacity to include those likely at greatest risk of future health problems/disparities. There is continued emphasis from influential organizations (e.g., the American Academy of Pediatrics, Society of Pediatric Psychology, American Academy of Family Physicians) to screen patients for social risk; however, systematic screening and intervention implementation targeting risk across healthcare systems, while integrating the inpatient and outpatient sector, remains challenging and is seldom done (Billioux et al., 2017; Fraze et al., 2019). Related to Kilbourne’s phase two (i.e., understanding disparities), inpatient screening of risk across systems supersedes many of the limitations routinely encountered in clinical care and research, thus improving our ability to determine the relative weight of social and medical factors in perpetuating health disparities for this vulnerable population. Building upon these efforts, next steps could include effective intervention development and dissemination to reduce or eliminate health disparities (i.e., phase three), informed by the items found in our study. Ultimately, efficient risk identification could lead to the tailoring of recommendations and targeting of resources to improve clinical outcomes while reducing disparities (Beck & Klein, 2016). Above and beyond a return ED visit or readmission, caregiver-specific items may prove beneficial to teams anticipating discharge of patients. This information may help teams determine how best to incorporate a child’s social risks to craft improved medical care plans and ensure appropriate handoffs between providers in inpatient settings and primary care medical homes (Andermann, 2016; Marsac et al., 2016; Spencer et al., 2019). Given that children are particularly prone to the adverse effects of repeat ED visits and hospitalizations (Melnyk, 2000) as well as the increased emphasis placed at a health system level on reducing this type of utilization, effectively identifying the level and influence of a child and family’s level of social risk prior to discharge, and potentially intervening, has the potential to reduce avoidable health complications while improving child and caregiver quality of life. Finally, such information should bolster the advocacy for healthcare systems to support access to and evaluation of programs designed to meet these needs for children at risk of health disparities, as well as support the families and providers who strive to care for these children.

Acknowledgments

We are grateful for our patients and caregivers for sharing their experiences with us. Thank you to fellow Most Vulnerable Project (MVP) study team members Mauricio Gomez, Raul Vega-Juarez, and Alyssa Libak for their dedication to this project.

Funding

This work was supported by the Friends of Doernbecher (institutional) and Collins Medical Trust (foundational) grants, and the BUILD-EXITO research training program, supported by the National Institutes of Health Common Fund and Office of Scientific Workforce Diversity (UL1GM118964, RL5GM118963, and TL4GM118965). The institution’s REDCap and Biostatistics program is supported through grant (UL1TR002369).

Conflicts of interest: None declared.

References