Children’s Hospitals: We Get What We Pay For

In 2016, the US spent $3.34 trillion on health care, with almost one-third of it allocated to hospitalizations¹ ; of that total, $47.7 billion was for pediatric hospitalizations.²  As hospitalization expenditures continue to increase, much attention is understandably focused on how to identify the drivers of these high-cost events, in particular, on what explains the considerable variation in hospitalization costs among different hospitals caring for similar conditions. On one hand, we know that certain patient characteristics are associated with more expensive care. We have learned, for example, that children with documented high-risk social conditions have longer lengths of stay³  and those children hospitalized from low-income zip codes also incur higher costs for common pediatric conditions compared with children from high-income zip codes.⁴  But these variations in patient features are not enough themselves to account for the cost differences observed among different hospitals.

In this issue of Pediatrics, Berry et al⁵  offer another insight into the phenomenon of hospital cost variation. Beginning with the premise that hospitals that care for a greater variety of conditions likely incur a higher set of fixed costs, the authors invoke a novel instrument, the hospital diversity of diagnosis index (HDDI), to capture the effect of this diagnostic diversity. Using the 2016 Kids’ Inpatient Database, the authors conducted a retrospective analysis of 1 654 869 hospitalizations for children 0 to 21 years of age distributed among freestanding children’s hospitals (FCHs), nonfreestanding children’s hospitals (NFCHs), and nonchildren’s hospitals (NCHs). They establish first that FCHs are the most expensive venues in which to care for hospitalized children, followed by NFCHs, and then NCHs. They then go on to show that in multivariable logistic regressions controlling for patient-level characteristics, the addition of the HDDI to the regression model decreases the adjusted cost per hospitalization by 34.8% among FCHs, 32.7% among NFCHs, and 3.1% among NCHs, thus eliminating the statistical difference in costs among hospital types.

The authors’ analysis, however, may have transported us to the right destination in a flawed vehicle, because the HDDI has some curious features. It gives great weight to the number, rather than the complexity, of diagnoses treated by a given institution, and it prioritizes evenness of distribution among conditions. To appreciate the consequences of these features, consider 2 hypothetical hospitals, one of which cares for 4 common conditions and another that cares for 8 common conditions. The second hospital will register an HDDI twice that of the former so long as the conditions cared for in each hospital are evenly distributed, each diagnosis accounting for 25% of admissions in the first hospital and 12.5% of admissions in the second. But now consider the relative HDDIs of 2 hospitals, each of which care for 5 diagnoses. In the first hospital, 5 common diagnoses are again evenly distributed across categories. But the second hospital in this scenario takes care of 2 common conditions and 3 rare, highly complex, and potentially expensive conditions. In this example, the HDDI for the first hospital will be higher than that of the second because evenness of distribution weighs heavily when calculating the HDDI.

What does this mean in terms of how we interpret the findings from Berry et al?⁵  Happily, as it turns out, the number of diagnoses is highly correlated in this sample with the presence of rare and expensive conditions, such that the HDDI for all its limitations succeeds in identifying the key issue here: that hospitalization in children’s hospitals is more expensive than it is in NCHs, even for common conditions.

This excess cost, however, must be placed in the proper context. Were children’s hospitals to produce the same output as NCHs, there would be little way to justify the excess costs associated with the care they deliver. But that is not the case. Although children’s hospitals do care for common conditions, they also sustain the necessary infrastructure to treat extremely complicated diseases requiring advanced imaging, surgical, laboratory, and therapeutic applications. The fixed costs associated with the ability to deliver such care are considerable. In addition, children’s hospitals train a disproportionate number of the next generation of pediatricians, particularly subspecialists, and they also support research that advances the entire field of pediatrics.

Each of these activities is associated with what economists refer to as positive externalities: they generate value that cannot be captured by the price commanded to produce them. Once a clinical discovery is developed at a children’s hospital, the finding is free for everyone to use. The newly minted pediatric intensivist coming out of a children’s hospital does not produce a future stream of income for the institution that trained her. The trauma unit or NICU located in a specific children’s hospital provides a value for those who may never have to use it but who benefit just by having it in reserve. When positive externalities occur in other types of markets, often the government will intervene to subsidize those activities that would otherwise be produced in socially suboptimal quantities. In the market for children’s hospitalization services, those subsidies are built into the price paid by all admissions, including those for common diseases.

One might lament the existence of these cross-subsidizations just as one might lament that these higher prices often translate into higher insurance premiums for commercially insured patients. But a world without these cross-subsidies is a world without NICUs, pediatric heart transplants, fetal surgery to correct congenital defects, or the ability to treat rare metabolic disorders. And that world, namely, a world without children’s hospitals and their elevated cost structures, would be a lamentable world indeed.

FCH

freestanding children’s hospital

HDDI

hospital diversity of diagnosis index

NCH

nonchildren’s hospital

NFCH

nonfreestanding children’s hospital