|Home | About | Journals | Submit | Contact Us | Français|
The cost of pediatric stroke care has received little attention, but the available data suggest it is expensive. To determine the cost of acute stroke, we analyzed a US national database.
We used the Kids’ Inpatient Database (KID2003) to determine the hospital-based costs of acute stroke in children ages 3 months to 20 years. Discharges were selected if the first diagnostic position contained an International Classification of Diseases, 9th Revision code pertaining to ischemic or hemorrhagic stroke. We examined the relationship between cost and stroke type by adjusting for variables that predict the cost of adult stroke.
There were 2224 pediatric cases, after statistical weighting, discharged with a diagnosis of hemorrhagic or ischemic stroke in KID2003. The estimated cost of acute pediatric stroke in the United States was $42 million in 2003. For the entire cohort, the mean cost of acute hospital care was $20 927 per discharge. The mean cost for ischemic stroke was $15 003, for intracerebral hemorrhage $24 117, and for subarachnoid hemorrhage $31 653. Stroke diagnosis, length of stay, hospital ownership, rural/urban teaching status, US geographical region, and discharge disposition were significantly associated with cost. Cost remained significantly associated with stroke diagnosis after adjusting for other predictors in the final multivariable regression model.
Pediatric stroke is expensive, and the lifetime cost of care is likely greater for a child than an adult. The cost to the family and the larger society underscore the importance of pediatric stroke treatment and prevention.
Acute adult stroke care is expensive.1–3 The components of cost have been analyzed extensively,4–7 and predictors of cost have been identified.8–10 In contrast, the cost of acute pediatric stroke has received little attention. The initial management of stroke in children is similar to that for adults,11,12 so one might assume that the costs are similar. However, the causes for stroke in children differ from those for adults,13,14 and the mortality rate in children is lower.15 These clinical differences could lead to differences in the cost of care for children compared with adults.
There are few cost-of-illness studies for pediatric stroke. We reported the cost of pediatric stroke at one institution.16 Other reports were limited because they provided few details how costs were measured and did not analyze factors that might be related to cost.17,18 Cost-of-illness studies for chronic pediatric illnesses such as asthma,19 diabetes,20 and cancer21 demonstrate the substantial cost of these illnesses. Cost-of-illness studies are important for understanding the economic and societal impact of a disease, because they can be used for the development of healthcare policy22 and determining the efficiency of treatment.23
We used the Kids’ Inpatient Database (KID) to determine the hospital-based costs of acute stroke care for children in the United States. We selected key demographic and clinical variables and analyzed them to determine whether they are associated with cost.
The patient sample was taken from the KID, a component of the Healthcare Cost and Utilization Project.24 The KID is an administrative database that measures the use of inpatient hospital services by children ages 0 to 20 years old in the United States. We used the 2003 edition, KID2003, because it was the first that enabled cost estimation by providing cost-to-charge ratio files. The design of KID2003 is described in detail at the Healthcare Cost and Utilization Project–US web site (http://hcup-us.ahrq.gov/db/nation/kid/kidrelatedreports.jsp). Briefly, sampling was limited to all discharges in which a patient was aged ≥20 years at admission. Nonfederal, short-term, nonrehabilitation, community general, and specialty hospitals from 36 states participated in Healthcare Cost and Utilization Project. The KID sampled pediatric discharges from all hospitals that agreed to participate. Federal hospitals, long-term hospitals, psychiatric hospitals, drug and alcohol rehabilitation facilities, and hospital units within institutions were excluded. To assure the accuracy of the representation of each hospital’s pediatric cases, discharges were sorted by state, hospital, diagnosis-related group, and a random number within each diagnosis-related group. Systematic random sampling was used to select 10% of pediatric cases from each frame hospital. National estimates were obtained by the use of discharge weights that were developed using the American Hospital Association universe as the standard. Hospitals were poststratified on 6 characteristics (ownership/control, bed size, teaching status, rural/urban location, and US region) contained in the American Hospital Association hospital files with an additional stratum for freestanding children’s hospitals. Discharge weights were created by stratum in proportion to the total number of nonnewborn American Hospital Association discharges for nonnewborn discharges. KID2003 contains information for up to 15 diagnoses per discharge, demographics, hospital characteristics, and discharge status from 3438 hospitals.
We limited our search to discharges that were coded for stroke in the first (ie, principal) diagnostic position. We previously found that the maximum identification of discharges occurred when we searched all 15 diagnostic positions of KID2003.25 We found that when compared with all discharges with a stroke code in any diagnostic position, 53% had an ischemic stroke code and 72% had a hemorrhagic stroke code in the primary diagnostic position. We believed this strategy increased the likelihood that a child was treated specifically for stroke rather than for a diagnosis that was not directly related to stroke.
In the current study, we identified discharges for hemorrhagic stroke (HS) by searching for the International Classification of Disease, 9th Revision (ICD-9) codes 430 for (subarachnoid hemorrhage [SAH]) and 431 for (intracerebral hemorrhage [ICH]). We identified discharges for ischemic stroke (IS) by searching for ICD-9 codes 433.xx, 434.xx, 436.xx, and 437.4–6. We included sinovenous thrombosis in this analysis because other reports in the childhood stroke literature have included sinovenous thrombosis with IS.25 Discharges with ICD-9 injury codes were excluded to reduce the inadvertent inclusion of traumatic hemorrhage and stroke. We excluded patients aged <3 months to minimize including children with perinatal stroke or neonatal intraventricular hemorrhage. This last exclusion only eliminated 14 discharges (6 IS and 8 HS) before statistical weighting.
We analyzed variables that were identified as significant predictors of cost for adult stroke.8–10,27,27 Variables of interest included stroke diagnosis, sex, age, hospital ownership, urban/rural teaching status, hospital region, length of stay, and discharge disposition. In addition, we included median household income and insurance payer. Race/ethnicity was excluded from the analyses due to the large amount of missing data (33.4%). If there were ≤11 discharges before statistical weighting for a given variable, that variable was excluded from further analysis to avoid biased statistical weighting of small sample sizes and to comply with the requirements for use of the KID2003 data set.
The KID2003 provided inpatient cost-to-charge ratios for almost every hospital in the data set.24 KID2003 did not report professional fees so we did not estimate this component. If a cost-to-charge ratio was unavailable for a specific hospital, KID2003 provided a group average cost-to-charge ratio that was a weighted average for hospitals in a given group. In those cases in which a cost-to-charge ratio was unavailable for a specific hospital, then we used the group-weighted average to calculate the cost. We calculated costs by multiplying the cost-to-charge ratios by the charges for hospital services.
Data analyses were conducted using SAS (Version 9.1; SAS Institute Inc, Cary, NC). Weighting structures provided by the KID were used to account for the survey design and sampling procedures specific to the cost data. All reported estimates are national estimates calculated using the statistical weights unless otherwise noted. National estimates and 95% CIs of cost of stroke were calculated by demographic variables. Values were rounded up to the nearest individual.
The distributions of the outcome variables “cost” and “length of stay” (LOS) were skewed so all analyses were performed on the ln transformation of these 2 variables. In the case of LOS, a small constant was added to avoid the problem of a zero value when calculating the ln of LOS. Survey regression methods were used to compare the cost of the 3 types of stroke.
Univariate regression analyses were run for each variable to determine inclusion in the multivariable model with the significant test criteria of P<0.10. Stepwise multivariable modeling was performed, starting with all significant univariate variables and including any clinically relevant interactions, resulting in a final model to obtain adjusted costs for each type of stroke.
Based on the case definition of our study, we identified 1070 actual discharges with a primary diagnosis of hemorrhagic or ischemic stroke in KID2003 (Table 1), which equates to 2224 discharges after statistical weighting. (Subsequent mention of discharges refers to the number after statistical weighting.) There were 1231 discharges with a primary diagnosis of IS (55%), 480 discharges of SAH (22%), and 514 discharges of ICH (23%; Table 2). Boys were slightly more frequent than girls in the cohort, primarily because hemorrhages occurred more frequently in boys. More girls than boys had ISs. Stroke discharges occurred most frequently in the age group 15 to 20 years compared with other age groups.
The proportion of stroke discharges was slightly lower in the highest household income range (Table 3). Private insurance covered 55% of the children, government insurance (Medicaid/Medicare) covered 35%, and self-pay or other coverage accounted for 10%. The hospitals were predominantly government or publicly owned, and most patients were discharged from urban teaching hospitals (Table 4). The majority of actual discharges were reported from the mid-western, southern, or western regions of the United States. Data from 2 large northeastern states were not available, but statistical weighting enabled the estimation of charges and costs for the northeast region. Most patients (72%) were discharged home, 11% were transferred to another facility, 8% were transferred to another short-term hospital, and 9% died in the hospital. ICH was associated with the highest mortality (15%) followed by 11% for SAH and 5% for IS.
Univariate analysis identified stroke diagnosis and LOS as variables that were highly associated with cost (P<0.001 for each). Overall, the mean cost of acute hospital care was $20 927 (in 2003 dollars) with a mean LOS of 8.3 days (Figure A–B) For IS, the LOS was 6.6 days and cost was $15 003. LOS for ICH was 9.6 days and cost was $24 117, whereas LOS with SAH was 11.2 days and cost was $31 653. Other variables significantly associated with cost were hospital ownership, rural/urban teaching status, US geographical region, and discharge disposition (Tables 5 and and6)6) Costs were higher at government/public-owned hospitals and urban teaching hospitals. Average cost was highest in the western United States and lowest in the south. Transfer to another facility such as a rehabilitation or intermediate care facility was associated with the highest cost, whereas discharge home was associated with the lowest cost. Family income level was associated with cost such that costs for patients from the higher income quartile were significantly higher than costs for patients from the lowest income quartile. Insurance status and sex were not significantly associated with cost.
Multivariate analysis showed that the relationship between cost and stroke type remained highly significant after adjusting for LOS, urban/rural teaching status, hospital geographical region, age group, and discharge disposition.
This is the first analysis of pediatric stroke in the United States that examined the relationship between cost and factors associated with the cost of adult stroke. Through the use of KID2003, we were able to estimate the costs for the acute care of stroke for the entire United States in 2003. In this study, HS was listed as frequently as IS as the primary diagnosis, a disproportionate number of strokes occurred in older adolescents, and discharge dispositions were distinctly different from adults.
HS in this cohort was as frequent as IS. Although some discharges of IS may have been missed by our search strategy, this study illustrates that SAH and ICH are important causes of pediatric stroke. There was an association between age and stroke frequency. The group of adolescents aged 15 to 20 years accounted for the largest proportion of IS and HS. The reason for this predominance is not evident but is consistent with our earlier work with the KID2003 data set.16
Most children (83%) were discharged from an urban teaching hospital. This contrasts with the adult experience, in which most patients were discharged from urban nonteaching hospitals or rural hospitals.26 Discharge dispositions after pediatric stroke were different from those of adults. Most children were discharged home regardless of stroke type, larger proportions of children were transferred to short-term hospitals, and smaller proportions were transferred to nursing homes compared with adults, whereas larger proportions of adults died or were discharged to nursing homes.8,10,26
Four US adult stroke studies were similar to the present study. Exact comparisons were difficult because of different study designs and analyses. Two studies used the Nationwide Inpatient Survey, the parent source of the KID2003 data set, whereas 2 used other sources. The Nationwide Inpatient Survey and KID2003 do not report professional fees, only hospital charges. Russell et al10 analyzed the 2002 Nationwide Inpatient Survey for the costs of ICH. The mean cost for all patients was $15 256 in 2002 dollars, and the mean LOS was 7.7 days. For patients aged 18 to 44 years, the mean LOS was 9.3 days and the mean costs were $21 077, similar to the results in children. Russell et al identified multiple predictors of cost, including LOS, age >65 years, male sex, Hispanic ethnicity, teaching status of the hospital, and geographic location of the hospital. Many of these factors were significantly related to the cost of pediatric stroke. Qureshi et al26 used the Nationwide Inpatient Survey to compare charges and outcomes of stroke between the years 1990 to 1991 and 2000 to 2001 in adults. They reported that in 2000 to 2001, average charges for IS were $16 200, ICH average charges were $28 800, and SAH average charges were $65 900. If one multiplies their charges by the weighted cost-to-charge ratios of the present study, their estimated costs would have been $6966 in 2001 dollars for IS, $12 384 for ICH, and $28 678 for SAH, which were somewhat lower than the costs in children. The reasons for the lower costs are not apparent but may have reflected the greater proportions of adults who died compared with children.26 Two other US adult stroke studies were less directly comparable to the present study because of different methods; however, these 2 studies identified predictors for stroke hospitalization costs. Holloway et al9 analyzed administrative data from 5 US academic centers in 1992 to determine the cost of adult stroke, including professional fees. They used cost-to-charge ratios that were developed in a Medicare claims analysis and found average costs of $9882 for IS in 1992 dollars, $21 535 for ICH, and $39 994 for SAH. Holloway et al found that stroke type, LOS, and discharge disposition were significant predictors of cost, similar to what we found with children. Reed et al8 determined the cost of acute adult stroke from an administrative data set from 150 US community hospitals in 1998. Costs were calculated with proprietary relative value unit-based methodology and included institutional overhead costs such as utilities, housekeeping, and maintenance. The mean cost for IS was $5837 in 1998 dollars, $10 241 for ICH, and $23 777 for SAH. Reed et al found that significant positive predictors of cost were stroke type, LOS, hospital teaching status, admission to an intensive care unit, and discharge disposition, many of which were significantly associated with pediatric cost. Reed et al found an inverse relation between cost and age.
Together, these studies show that there are many similarities between the cost of acute pediatric and acute adult stroke care. The magnitudes of cost are similar, and many predictors of adult stroke cost were significantly associated with pediatric stroke cost. The trend of higher costs for ICH and SAH is similar between adults and children.
There are 3 other cost studies of pediatric stoke. We reported that the cost of initial diagnosis and care for the year after diagnosis was $42 338 at a single institution.16 Janjua et al17 analyzed the Nationwide Inpatient Survey for childhood ischemic stroke for 2000 through 2003. They reported a median charge of $38 700 but did not report costs or analyze predictors. If one applies the weighted cost-to-charge ratio of the present study to their median charge, their estimated cost was $16 641, similar to the present study. Chen et al recently reported the cost of pediatric and adult stroke care in Taiwan for 1997 through 2003.18 Although direct comparison is difficult, the cost of acute pediatric stroke care was more expensive because of more frequent HSs and associated severe illnesses. LOS, hospital type, discharge location, sex, age, and stroke diagnosis were significantly associated with cost, similar to the present study. Gardner et al28 recently presented an abstract that analyzed the direct cost for pediatric stroke in a large health maintenance organization. For IS, mean acute cost (including professional fees) was $36 073 and median cost was $21 144. For HS, mean cost was $90 693 and median cost was $34 256. Together, the previous studies suggest that our results reasonably estimate the cost of pediatric stroke care. Furthermore, the factors associated with cost in Taiwan and in the United States were similar.
There are several limitations to our study. The KID2003 is an anonymized database that is discharge-specific. Only hospitalized discharges were recorded, so children who were evaluated as outpatients were not listed. We could not review the actual medical record to verify the accuracy of the diagnoses and we could not exclude the possibility that a given individual was admitted multiple times. There are a number of limitations in ICD-9 codes identifying stroke cases.30–31 Limiting searches to the primary diagnosis position may miss cases of ischemic stroke32 and the use of ICD-9 codes to identify pediatric IS or sinovenous thrombosis can miss cases.34,34 Because we could not verify the accuracy of the diagnoses, we cannot determine whether our findings overestimate or underestimate the true hospitalization cost. The KID2003 did not include the cost of professional fees, rehabilitation, or outpatient care, so we could not calculate all the costs encountered by a consumer or a third-party payer. Lastly, multiple studies4,36,36 have shown that indirect costs contribute substantially to the cost of adult stroke. Indirect costs for chronic pediatric illnesses were not available in KID2003 and could not be estimated. Therefore, our results represent a minimum estimate of the true cost of care.
A minimum estimate of the annual cost of acute pediatric stroke hospitalization in the United States is $42 million. Although this amount pales in comparison with adult stroke, pediatric stroke is expensive. The lifetime cost of stroke care may have greater financial impact given the longer life expectancy for a child. The cost to the family and the larger society can be used as one measure of the effectiveness of current and future treatments.
Source of Funding
Supported by the Research Institute of the Nationwide Children’s Hospital.