Lost Earnings and Nonmedical Expenses of Pediatric Hospitalizations

We performed a cross-sectional analysis as part of the Hospital-to-Home Outcomes Study (H2O). H2O was a prospective, randomized controlled trial that was designed to determine the effects of a single-nurse home visit after a standard pediatric discharge.¹⁶ H2O took place at Cincinnati Children’s Hospital Medical Center (CCHMC), a large, academic, free-standing pediatric facility that cares for ∼95% of all children who are hospitalized within Hamilton County, Ohio. Enrollment occurred from February 2015 to April 2016. In 2015, the median household income for Hamilton County was $49 013 (range across census tracts: $6688–$167 500),¹⁷ and the unemployment rate was 4.6%.¹⁸ 

Children were eligible for inclusion in H2O if they were admitted to the Hospital Medicine or Neurosciences (Neurosurgery and Neurology) teams. Patients were excluded if they were discharged to a residential facility, lived outside the home health care nurse service area, were eligible for skilled home health care services, or if the participating parent or caregiver (hereafter referred to as parent) was non-English speaking.¹⁶ Recognizing that nonmedical costs that are incurred during extended hospitalizations may not be representative of typical hospitalizations for common pediatric conditions, we excluded children with prolonged lengths of stay for these analyses. We defined prolonged stays a priori as ≥2 SDs above the mean. To ensure that we captured nonmedical costs throughout a hospitalization, we also excluded children who were enrolled in the study and were surveyed before completing at least 20% of their total length of stay (eg, enrolled on the first day of a 6-day hospitalization).

A total of 1500 children were randomly assigned during H2O; 2 children withdrew. The mean length of stay was 3.1 days (SD: 4.9 days). We excluded 31 children with a length of stay >13 days and 95 children because they completed <20% of their length of stay at the time of survey. Thus, the analyses that are detailed below include 1372 children. The CCHMC Institutional Review Board approved H2O.

We conducted a face-to-face in-hospital survey as part of H2O and collected information on the child’s demographic characteristics, including age (<3, 3–10, or >10 years), race (white, African American, or other), and insurer (private or public or self-pay).^2,15 Parents self-reported their own sex, marital status (single, widowed, separated, divorced, or married or living with partner), educational attainment (less than a college degree or college or more), employment status, and work flexibility. We categorized the employment variable according to the number of parents who were employed (0, 1, or 2) and whether that employment was full- or part-time. Respondents characterized their work as “flexible” or “not flexible” on the basis of their ability to rearrange their work schedule and receive paid sick leave.

We collected the number of adults and children living in the home (1, 2, or ≥3 for both), annual household income, and presence of financial and social hardships. Household income was reported in 7 categories, ranging from <$15 000 to ≥$120 000. Financial and social hardships were characterized by using previously validated questions (Table 1).^19,–21 We categorized these measures into groups of 0, 1 to 2, or ≥3 financial hardships and 0 or 1 to 2 social hardships.^9,10,22 

We used survey questions to assess nonmedical costs that were related to the hospitalization from the time of entry into the acute care setting (ie, emergency department or inpatient unit) to the survey completion. Most of our surveys were completed approximately two-thirds of the way through the hospitalization, but there was variation across the sample. For surveys that were completed <8 hours before discharge (610 observations), we assumed that the respondents gave us a complete accounting of their costs and that any remaining hours until discharge had no additional costs. For the remainder of the sample (762 observations), we calculated an hourly hospitalization cost rate on the basis of the reported nonmedical costs at time of the survey administration. We then used this hourly rate to estimate additional nonmedical costs accrued over the remaining hospitalization. For consistency with the part of the sample that needed no extrapolation, we excluded the last 8 hours of the hospitalization (Fig 1). We expected that nonmedical costs would vary during the course of the day (ie, overnight relative to day costs) and opted to exclude 8 hours because many discharges occurred early in the morning. Given that this was an assumption that we could not test with our data, we opted to conduct sensitivity analyses to evaluate other cut points (ie, 4 and 12 hours).

We anticipated that nonmedical costs would affect both sides of a family’s economic ledger (ie, lost earnings and additional expenses). Thus, we captured costs related to both lost wages and/or tips due to missed work and accrued costs due to parking or travel, meals, child care, or other expenses (Table 1). We calculated the total cost burden (TCB) for the entire hospitalization using the sum of lost earnings, accrued costs, and the extrapolation method described above. We then calculated the daily cost burden (DCB) by dividing the TCB by the length of stay in days measured in 24-hour increments. Finally, we calculated the daily cost burden as a percentage of daily income (DCBi) by dividing the DCB by the category maximum of the family’s reported income also standardized for a 1-day period (Fig 1). We dropped 6 observations with missing income data. For the top-coded income category, we assigned an income of $150 000 to be consistent with income differences across categories. For example, the DCBi for those with a reported income between $15 000 and $29 999 was calculated by dividing their DCB by $82.19 ($29 999 per 365 days). With the possible exception of the top-coded category, the use of the categorical maximum allows the DCBi to be interpreted conservatively as the minimum percentage of household income spent on nonmedical costs.

We used descriptive statistics to assess the distribution of demographic characteristics across our sample. Comparisons of characteristics between the sample with and without extrapolated nonmedical costs were made by using the χ² test. Then we calculated cost burdens (TCB, DCB, and DCBi) across child, parent, and household characteristics using the Kruskal-Wallis test to assess for differences across groups. We report simple P values, indicating those that are <.05 after adjusting for multiple comparisons using the false discovery rate (FDR) method.²³ We also quantified components of the cost burden, reporting medians and interquartile ranges (IQR). For sensitivity analyses, we followed similar methods.

Of the 1372 children included in this study, 56% were <3 years old (Table 2). A slight majority were white and publicly insured. The average length of stay was 2.4 days (SD: 1.7 days); the median was 1.8 days (IQR: 1.2–2.9 days). Responding parents were mostly women, were married or living with a partner, and without a college degree. A variety of work arrangements were noted; 47% reported 2 employed parents. Households generally included >1 adult and >1 child. Parents reported household incomes across all categories; 21% reported an income <$15 000. Nearly 70% of parents reported ≥1 financial hardship, and ∼38% of parents reported ≥1 social hardship. Characteristics of the 762 participants from whom we extrapolated data were similar to those of the 610 participants from whom we did not extrapolate data except that they were more likely to have longer lengths of stay (as expected) and private insurance (Supplemental Table 5).

The median TCB was $112.8 (IQR: $40–$321.5; range: $0.0–$11 742.3). The median DCB was $51.4 (IQR: $23.6–$159.2) (Table 3). The median daily earnings loss was $0 (IQR: $0–$139), and additional expenditure was $29.8 (IQR: $16.7–$47.4). Parents of African-American children were more likely to report earnings loss (median $53.6) than parents of white children (median: $0; P < .01). However, parents of African-American children also reported lower additional expenditures than parents of white children and children of other races ($25.7, $32.3, $33.9, respectively; P < .01). Households with reported incomes in the $15 000 to 29 999, $30 000 to 44 999 and $45 000 to $59 999 categories were more likely to report earnings loss compared with households with reported incomes in other categories (median loss of $26–$66 compared with households with reported incomes in all other categories, which reported a median of $0 in lost wages; P < .01). The DCB did not differ by primary insurer. However, it did differ by marital status, parental education, employment status, work flexibility, and household income (all: P < .01). Households with more reported financial hardships had a significantly higher DCB (median: $60.5 for those with ≥3 financial hardships, $64.4 for those with 1–2 financial hardships, and $40.8 for those with 0 financial hardships; P = .02). There were no differences in the DCB by social hardship.

The median DCBi was 45% (IQR: 12.8%–149.5%); in other words, the median household saw their daily income depleted (ie, lost, spent, or both) by ∼45% because of nonmedical costs related to the hospitalization (Table 4). Within racial categories, African Americans were most affected, with the DCBi depleted by >70% (IQR: 22.1%–190.9%). Similarly, those with public insurance or no insurance had a significantly higher DCBi than those with private insurance (81.5% vs 17.6%; P < .01). Likewise, single parents and those with less education, less employment, and less work flexibility all had a higher DCBi (each: P < .01). Although there were no differences in the TCB and DCB by a parent’s sex, the DCBi was significantly lower for male parents compared with female parents. This was because of differences in the distribution of income by parent sex: men were significantly less likely to report income <$15 000 (9.4% vs 22.6%) and significantly more likely to report income ≥$120 000 (23.7% vs 12.1%) compared with women. Men were also more likely to be white, married, and in households with 2 working adults (all: P ≤ .01). Finally, those with ≥3 financial hardships had a DCBi of 86.4% compared with 64.5% for those with 1 to 2 financial hardships and 14.8% for those with 0 financial hardships (P < .01). Similarly, those with ≥1 social hardships had a DCBi of 70.3% compared with 32.8% for those with 0 social hardships (P < .01).

We conducted a series of sensitivity analyses in which other extrapolation cut points were used, including 4 and 12 hours (instead of 8 hours). None of these changes altered the findings in substantive ways (Supplemental Table 6).

In this study of 1372 children who were hospitalized for common pediatric conditions, we found that the nonmedical cost burden varies considerably across child, parent, and household characteristics. The nonmedical cost burden for the median household is ∼$50 for each day their child is hospitalized; for a quarter of households, this is >$165 each day. Nonmedical costs disproportionately affect those with competing challenges. For example, those who report ≥3 financial hardships lose or spend 6 times more of their daily income on nonmedical costs than those without financial hardships. Those with ≥1 social hardship lose or spend double their daily income compared with those without social hardships. We propose that the degree of hardship along with other factors related to work flexibility and monetary reserves could either mitigate or exacerbate a child’s hospital-to-home transition.

Hardships related to food, housing, and energy insecurity have each been linked to short- and long-term pediatric outcomes.¹² Such competing priorities decrease adherence to treatment plans¹¹ of substantive importance to those transitioning from hospital-to-home. Our previous focus groups with caregivers of children who were recently hospitalized highlighted the relevance of these hardships during this vulnerable period.^2,24 Families that are already at greater risk for undesirable posthospitalization outcomes^9,10 on account of pre-existing hardships also must confront the reality of losing more of their daily income than their more affluent peers. Having less income and/or more costs could lead families to have less to spend on posthospitalization needs; recommended treatments and follow-up visits could be delayed or missed. On the other hand, we expect that those with fewer hardships, more flexibility, and more reserves would be less affected by hospitalization-related nonmedical costs and thus be more likely to experience desired outcomes.

Families that reported annual incomes between $30 000 and $44 999 reported the highest DCB of all: ∼$93 per day. This income range is just above the federal poverty level, which is $24 300 for a family of 4.²⁵ Those in this income category were more likely to report earnings loss, suggesting that the near poor may only be able to earn a wage if they are physically present at work. This could represent lower work flexibility, thereby adding further constraints during this vulnerable hospital-to-home transition period. These families could have additional medical cost burdens resulting from less generous insurance plans that require greater cost sharing.²⁶ We also found that female parents reported significantly lower household incomes and higher DCBi than male parents. This finding is consistent with overall demographic and economic trends. Most single parents are women, and men are usually higher earners.^27,28 

The underlying contribution of potentially exacerbating factors, including financial and social hardships, is important given their established associations with adverse, undesirable outcomes (eg, treatment nonadherence and readmissions).^9,–11 We expect that nonmedical costs may compound the effect of pre-existing hardships. In turn, this could place families at an even greater risk for subsequent morbidity. Indeed, every additional dollar that is either lost or not earned could place considerable challenges on a family’s ability to navigate postdischarge care needs. Moreover, the need to make such challenging decisions may magnify the stress already felt by families, potentially deepening the fog they experience during and after hospitalization and impeding their return to normalcy.²⁴ 

The cost burdens we uncovered and their associations with underlying child, parent, and household factors may be relevant to hospital-wide metrics. Pediatric readmissions are increasingly being used as a quality marker for hospitals, but they vary on the basis of the population the hospital serves.^29,–31 Many hospitals have implemented readmission reduction efforts to support patients and families during the stressful periods of hospitalization, transition, and postdischarge.^24,32 Specifically, interventions that enhance education and support postdischarge needs (eg, follow-up support in person or via phone, nurse home visitation, and connections to community-based resources) are being trialed to assist families and, ideally, reduce readmission rates.^16,33,34 Our findings reveal the potential need for additional supports to be integrated into routine hospital-based care (eg, parking vouchers, meals for family members, and transportation reimbursement).

As hospitals and health systems aim to deliver patient- and family-centered care, the role of the nonmedical cost burden bears consideration. In light of the growing push to understand outcomes that are meaningful to patients and families,³⁵ Cheng et al³⁶ highlighted the need to provide connections to relevant resources that can address the social determinants of health. Chassin and Galvin³⁷ defined quality of care as “the degree to which health services for individuals and populations increase the likelihood of desired health outcomes.” We suggest that quality care should extend to the consideration of the nonmedical cost burden alongside mitigating and exacerbating factors that influence desired health outcomes.^38,–40 

This study has several limitations. First, our calculations are based on a face-to-face survey. It is possible that certain individuals remember their costs more readily than others and are subject to social desirability bias; some may not want to share information about income lost or expenses incurred. Second, we performed these analyses within a cohort of children who were hospitalized in general inpatient units at 1 pediatric institution in a midsize metropolitan area in Ohio. Within CCHMC, there are existing supports, such as free parking for all parents and meal bags that are offered sparingly to help families. Thus, our findings may undercount the burden experienced by families at institutions that charge for parking or do not provide in-hospital assistance. In addition, we only focused on the English-speaking population although future research needs to include or perhaps focus specifically on the immigrant population. Finally, our approach to enrollment led to variability in survey timing relative to hospital discharge. We recognize that the end of a hospitalization can often include some of the most intense and perhaps costly hours, requiring last-minute arrangements related to the transition to home. Still, we opted to extrapolate data consistently across our sample while also pursuing sensitivity analyses to vary the point at which extrapolation was performed. Clearly, it is possible that costs are not linear over the course of a hospitalization, so our extrapolation method may not yield precise cost burden estimates.

Nonmedical costs related to the hospitalization of a child disproportionately affect those experiencing financial and social hardships. Equity gaps in lost earnings and incurred expenses during pediatric hospitalizations may influence parallel and persistent equity gaps in pediatric morbidity surrounding hospitalizations. The development of interventions that are focused on supporting families during and after hospitalization should account for both medical and nonmedical cost burdens.

CCHMC

Cincinnati Children’s Hospital Medical Center

DCB

daily cost burden

DCBi

daily cost burden as a percentage of daily household income

FDR

false discovery rate

H2O

Hospital-to-Home Outcomes Study

IQR

interquartile range

TCB

total cost burden

This article is submitted on behalf of the H2O study group. All authors and study group members had full access to the data and have approved of this work as submitted.

The following H2O Study Group members are nonauthor contributors: JoAnne Bachus, BSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Kathleen Bell, MS, CCHMC, Division of Hospital Medicine, (Cincinnati, OH); Monica L. Borell, BSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Patricia Crawford, BSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Jennifer M. Gold, BSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Judy A. Heilman, RN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Jane C. Khoury, PhD, CCHMC, Division of Biostatistics and Epidemiology, (Cincinnati, OH); Pierce Kuhnell MS, CCHMC, Division of Biostatistics and Epidemiology, (Cincinnati, OH); Karen Lawley, BSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Allison Loechtenfeldt, BS, CCHMC, Division of Hospital Medicine, (Cincinnati, OH); Colleen Mangeot, MS, CCHMC, Division of Hospital Medicine, (Cincinnati, OH)* Margo J. Moore, BSN, CCRP, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Lynne O’Donnell, BSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Cory Pfefferman, BS, CCHMC, Division of Hospital Medicine, (Cincinnati, OH); Rita H. Pickler, RN, PNP, PhD, College of Nursing, The Ohio State University (Columbus, OH); Hadley S. Sauers-Ford, MPH, Pediatric Telemedicine Analyst, Department of Pediatrics, University of California Davis Health System (Sacramento, CA); Susan N. Sherman, DPA, SNS Research (Cincinnati, OH); Lauren G. Solan, MD, MEd, University of Rochester Golisano Children’s Hospital, Division of Hospital Medicine (Rochester, NY); Angela M. Statile, MD, MEd, CCHMC, Division of Hospital Medicine (Cincinnati, OH); Heidi J. Sucharew, PhD, CCHMC, Division of Biostatistics and Epidemiology (Cincinnati, OH); Karen P. Sullivan, BSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); Heather L. Tubbs-Cooley, RN, PhD, CCHMC, James M. Anderson Center for Health Systems Excellence, Home Care Services and Division of Patient Services (Cincinnati, OH); Susan Wade-Murphy, MSN, CCHMC, Home Care Services and Division of Patient Services (Cincinnati, OH); and Christine M. White, MD, MAT, CCHMC, Division of Hospital Medicine, (Cincinnati, OH).