- Top of page
- MATERIALS AND METHODS
- AUTHOR CONTRIBUTIONS
This study presents one of the first large-scale analyses of pediatric inpatient health care charges for children with SLE. Using the KID, we found that mean hospitalization charges were significantly greater for children with SLE compared to other pediatric discharges. Among pediatric SLE discharges, the presence of kidney involvement, especially kidney failure, significantly predicted higher charges.
A coded diagnosis of AKF was associated with the highest charges and longest hospitalizations among the SLE cohort. The KID does not provide data regarding the cause of AKF. Consequently, we cannot specify whether AKF was secondary to lupus nephritis. Regardless, these results were not surprising, since the increased charges associated with AKF were consistent with other studies of hospitalized patients who develop AKF, regardless of etiology (20, 21).
While SLE with a kidney transplant represented a relatively small proportion of SLE discharges (1.5%), the mean charge per discharge associated with a kidney transplant diagnosis was substantial and greater than SLE-associated ESKD. Approximately 30% of transplant-related discharges included the code for the kidney transplant procedure, implying that the transplant was performed during that hospitalization. It is probable that the transplant operation and postoperative care contributed significantly to the charges for the group as a whole.
Compared to AKF and transplant, ESKD accounted for the least expensive hospitalizations, but the greatest number of discharges and total charges. ESKD contributed to 15% of SLE discharges and generated 28% of all SLE-associated charges. The high charges may be a combination of the high proportion of discharges and the prolonged length of hospital stay. At 9.5 ± 0.8 days, this group had the second-longest mean ± SD length of stay. Analysis of the primary reason for admission in a prospective study may provide more insight into these hospitalizations and potentially provide a target for reduction in health care resource utilization.
Hospitalization for SLE + KI without kidney failure resulted in slightly lower charges than SLE − KI. The mean ± SD length of stay was shortest for this group at 4.7 ± 0.3 days, compared to 5.5 ± 0.3 days for SLE − KI and 6.8 ± 0.6, 9.5 ± 0.8, and 13.8 ± 1.2 days for kidney transplant, ESKD, and AKF, respectively. Some of the SLE + KI without kidney failure hospitalizations may have included brief, elective hospitalizations for diagnostic and therapeutic procedures such as kidney biopsy or medication infusion. Regardless, these data support the hypothesis that the presence of kidney failure, not solely kidney involvement, is the key determinant of increased hospitalization charges for SLE patients.
We examined the relationship between demographic and socioeconomic factors and inpatient charges, postulating that factors that lead to worse disease severity will result in greater charges. Multiple studies of patients with SLE have shown worse generalized SLE and kidney-related disease severity and outcomes in those of male sex, nonwhite race, and low socioeconomic status (8, 22–24). Data also exist demonstrating that both general SLE and kidney-related disease damage are associated with higher direct costs in the care of patients with SLE (14, 16, 25).
Based on these studies, one would conclude that sex, race, socioeconomic status, and severe kidney disease would all significantly affect health care utilization. In our study, this was only partially true. When examining all SLE discharges, sex and race were associated with higher charges in both bivariate and multivariate analyses, while socioeconomic factors such as insurance status and median income of patient zip code were not. In SLE + KI discharges, race was associated with higher charges on both bivariate and multivariate analyses, whereas socioeconomic factors were not.
Our analysis only examined inpatient charges, and it is possible that examination of outpatient expenditures would find an association between socioeconomic status and charges. It may be that once a patient is admitted to the hospital, their disease has become severe enough that socioeconomic differences affecting disease severity become less important in predicting charges. It may also be that insurance status and income by zip code are not a complete reflection of the socioeconomic factors that affect disease management and severity. Petri et al examined the relationship between race, socioeconomic status, and morbidity in SLE and found that poor scores rating patient compliance with the prescribed treatment regimen were associated with kidney disease severity, while other standard measures of socioeconomic status such as income were not directly associated with disease status (26). Data on patient compliance were not available in the KID for inclusion in our analysis. Another explanation may be that the KID employs categorical data to characterize socioeconomic status and that utilization of continuous variables would reveal a linear association or threshold effect on hospitalization charges.
It was not surprising that discharges from a teaching hospital were associated with higher charges. Hospitalization at a teaching hospital has traditionally been associated with higher costs and charges due to more complex cases, specialized services, and time directed to graduate medical education (27, 28).
Prior studies have demonstrated an association between disease severity and age at presentation. Descloux et al examined the effect of age at disease onset on disease severity, dividing by prepubertal (<9 years), peripubertal (9–14 years), and postpubertal (>14 years), and found an inverse relationship between age and extent of overall disease damage regardless of disease duration, although this relationship did not apply to renal damage specifically (29). In our study, to analyze whether age effect translated into higher charges, age was divided into 3 groups: <15 years to represent the pre- and peripubescent population, 15–18 years to represent the postpubescent population, and 19–20 years to represent young adults. Initially, the age cutoff for the youngest group was <10 years, but this resulted in a very small sample size, precluding the ability to make meaningful associations with this group. We also deviated from prior publications, since this HCUP study included prevalent and incident SLE patients with age at hospitalization documented, not age at SLE onset.
On multivariate analysis, older age was associated with higher charges for all SLE discharges, but this was not true for the SLE + KI subgroup. This contradicts the presumed hypothesis that the younger patients with more severe disease would generate higher charges. Further analysis of this question would require more detailed analysis than performed here, but one could postulate that this contradiction may be explained by non–SLE-related factors such as location of service (19–20-year-olds may be admitted to adult wards), higher rates of obstetrical care in 19–20-year-old patients, or a greater burden of SLE-associated complications in survivors of early-onset SLE. The lack of an association on multivariate analysis for age in the SLE + KI subgroup is consistent with past publications (by Baqi et al and Descloux et al) that also failed to demonstrate an association between age and development of ESKD (5, 29).
One study has reviewed pediatric SLE health care utilization in the past. Brunner et al evaluated SLE costs from 2 pediatric centers, examining variables such as inpatient versus outpatient care, medications, laboratory testing, and dialysis (17). Our analysis was different in that we examined inpatient care only, analyzing demographic and kidney disease–specific variables. Both the study by Brunner et al and our study examined the role of dialysis/ESKD on charges. Brunner et al examined the cost of dialysis in their population (n = 119), finding that dialysis patients constituted 2.5% of their study population, but contributed to 11% of the total direct costs. In our study of 7,558,812 pediatric and 7,390 pediatric SLE hospitalizations in 2006, we found that ESKD contributed to 15% of SLE discharges and 28% of all SLE-associated charges. While the 2 analyses are not directly comparable, they both demonstrate that ESKD is responsible for a disproportionate percentage of total SLE health care resources.
Several studies have examined health care utilization by adults with SLE on a national level and demonstrated results consistent with our study. Pelletier et al reported data from a 2006–2008 US claims database that adult SLE patients with kidney involvement were hospitalized more frequently than those without kidney involvement (30.3% versus 13.6%; P < 0.001) and had greater annual followup costs in both inpatient and outpatient settings ($30,652 versus $12,029; P < 0.001) (15). Clarke et al surveyed adult SLE patients at 6 centers in the US, Canada, and the UK on their reported use of medical services and calculated an estimated cost based on market costs. They reported that the 4-year (1995–1999) cumulative direct medical cost of SLE patients with kidney involvement was $27,869–99,544 versus $20,337 in those without kidney involvement (16). Li et al examined a Medicaid database (combined pediatric and adult data) and reported that the mean annual cost over a 5-year period ranged from $27,463–50,578 in patients with kidney involvement versus $13,014–16,638 in patients without kidney involvement. Patients that developed ESKD had mean costs ranging from $47,660–106,982 per annum (13).
Our data indicated that the fraction of pediatric SLE hospitalization charges compared to other pediatric hospitalization charges was rising between 2000 and 2006 (P < 0.0001). Consequently, we cannot attribute the rising charges to inflation alone. Review of the data demonstrated a small, but significant, increase in the length of stay between 2000 and 2006 (from 5.6 days to 6.4 days). Possible explanations may be that novel therapies targeted for SLE are disproportionately expensive and that survival of complex patients is increasing the length of stay.
The greatest strength of this study was the size and scope of the KID. The use of a national claims database offered data that eliminate single-center biases. The KID provided sufficient data that permitted us to move beyond demographic summaries and examine disease-specific factors such as kidney failure and its contribution to charges. Apart from the analysis by Li et al examining the effects of ESKD on charges reported in a Medicaid database (13), few other studies have been able to subdivide the kidney involvement population to examine this specific factor.
However, there were also limitations to using this database. First, assignment of billing codes is operator dependent, and therefore may be subjective. For that reason, we kept our definition of kidney involvement broad to ensure we included anyone who may have experienced any kidney complication. We reviewed ICD-9-CM codes in other articles to inform the codes used to define SLE and kidney involvement in this study (12, 13, 15). Additionally, a separate study by Chibnik et al validated the efficacy of ICD-9-CM codes in a Medicaid population to identify patients with lupus nephritis (30). In the study by Chibnik et al, the positive predictive value for a patient having lupus nephritis was 88%. They required greater than two 710.0 codes for SLE, greater than 2 nephrology visits, and greater than 2 renal ICD-9-CM codes. Our methods were different in that we were unable to require 2 separate encounters, since the KID data are shown by discharge rather than by patient. Also, the sensitivity and specificity for the pediatric population in the present study, which includes all payers, may be slightly different from the predominantly adult Medicaid population in the study by Chibnik et al.
It is notable that ICD-9-CM codes do not distinguish between neonatal lupus and SLE. This is complicated by the fact that SLE can present as early as infancy. Zulian et al summarized the clinical courses of 13 patients with infantile SLE presenting between ages 6 and 11 months (31). An examination of the age distribution of patients in our cohort demonstrated 57 discharges for patients ages <3 years who were at risk for misclassification. When these discharges were removed from the analysis, no significant change in charges was found. Given that SLE can present in infancy, the total number of discharges at age <3 years was so small and their impact on charges so insignificant that all pediatric ages were included in the analysis.
Other limitations of this data set include the fact that it is based on the number of discharges, not the number of patients. As a result, we cannot make conclusions on medical care charges per capita. Finally, this database is limited to an analysis of inpatient charges and is not inclusive of outpatient or emergency costs. Consequently, our analysis reflects only a portion of the full expense of managing patients with SLE.
This study represents a step toward a better understanding of in-hospital health care utilization by children with SLE and SLE-associated kidney disease. Our data demonstrated that hospitalization of SLE patients with kidney involvement, and more specifically kidney failure, resulted in significant health care expenditures and higher charges. A more detailed examination of resource utilization for SLE patients with kidney failure in addition to an assessment of outpatient expenditures for all aspects of SLE would complement this study and provide a more complete description of SLE-related health care utilization and charges.