Trends and Resource Utilization in Kawasaki Disease Hospitalizations in the United States, 2008–2017 Free

Patients under 18 years of age hospitalized with KD were identified from a retrospective, serial cross-sectional analysis of inpatient discharges between 2008 and 2017 using the Healthcare Cost and Utilization Project’s (HCUP) National Inpatient Sample (NIS) database. The NIS database is one of the largest publicly available all-payer inpatient care databases in the US and is part of a family of databases and software tools developed for the HCUP and sponsored by the Agency for Healthcare Research and Quality.¹¹  Inpatient stay records in the NIS includes clinical and resource use information typically available from discharge abstracts created by hospitals for billing. The NIS is published yearly and contains data of approximately 7 million unweighted hospitalizations per year, which estimates approximately 35 million weighted national hospitalizations. The 2017 NIS sampling frame includes data from 48 statewide data organizations, covering more than 97% of the US inpatient population.¹²  We used the sampling weights provided by the NIS to generate national estimates.¹¹  The NIS database has been used to study several newborn and pediatric conditions.^13–17 

We queried the NIS databases between 2008 and 2017 and extracted all pediatric hospitalizations assigned the International Classification of Diseases, 9th/10th Revision, Clinical Modification (ICD-9/10-CM) codes for KD in the primary or secondary diagnoses fields. ICD-9-CM code 446.1 was used from January 1, 2008 through September 30, 2015, and ICD-10-CM code M30.3 was used from October 1, 2015 through December 31, 2017. The ICD-9-CM and ICD-10-CM codes used for identifying CAA included 414.11 and I25.41, respectively.

We studied baseline patient characteristics of the study population such as age, gender, race, comorbidities (history of obesity, history of hypertension, deficiency anemias, history of chronic pulmonary disease, valvular heart disease, history of fluid and electrolyte disorders, history of neurologic disorder, coagulopathy, weight loss), median household income (presented as first, second, third, and fourth quartile), and primary payer (Medicare or Medicaid, private including HMO, uninsured or self-pay) as supplied by the NIS.¹⁸  Hospital-level characteristics included hospital bed size (small, medium, or large), hospital location (rural or urban), teaching status, day of admission (weekday or weekend), admission type (emergent or urgent and elective), and hospital region (Northeast, Midwest, South, or West). Comorbidities were estimated using Elixhauser comorbidity software supplied by HCUP tools and software.¹⁹  Specific concurrent medical conditions and procedures of interest were identified by ICD-9/10-CM diagnosis and procedure codes.

KD hospitalization rate was calculated by dividing the weighted number of KD hospitalizations (numerator) by the corresponding subgroup population derived from the Center for Disease Control and Prevention’s wide-ranging online data for epidemiologic research database as denominator and expressed as per 100 000. Population estimates in these online databases are bridged-race population estimates produced by the US Census Bureau in collaboration with the National Center for Health Statistics.¹⁸  To establish trends over time, the Cochrane-Armitage test for dichotomous dependent variables and Jonckheere-Terpstra test for continuous dependent variables were used. One-way ANOVA and post-hoc Tukey’s Honestly Significant Difference test were used to evaluate any significant differences between subcategories, such as race and US census region.

Descriptive statistics were used to summarize baseline patient and hospital level variables. Categorical variables are represented as proportions and continuous variables are represented as median and interquartile range. The exposure variable was a calendar year, and the outcomes of interests were KD hospitalization rate, frequency of CAAs accompanying KD hospitalizations, length of stay (LOS), and inflation-adjusted cost of hospitalization. Designated weight values provided by the NIS were used to produce nationally representative estimates. Due to changes in sampling and weighting strategies after 2012, the HCUP has provided trend weights for years 1993 through 2011 to make estimates comparable to the new design (2012 and after).¹⁹  In sensitivity analysis, we determined the effect of the transition from ICD-9 to ICD-10 codes in October 2015, and we examined the proportion of KD hospitalizations assigned ICD-10 codes for CAA for each month of 2014 through 2016.

To calculate the estimated cost of hospitalization, the NIS data were merged with cost-to-charge ratios available from the HCUP. We estimated the cost of each inpatient stay by multiplying the total hospital charge with the cost-to-charge ratios provided by the HCUP.²⁰  Adjusted cost per year was calculated in terms of the 2017 cost after adjusting for inflation according to the latest consumer price index data released by the US government. This permitted us to standardize the costs over the study. Statistical Analysis System 9.3 (SAS Institute, Cary, NC, USA) was used for all analyses. We considered a two-tailed P value <.05 as statistically significant.

A total of 43 028 pediatric hospitalizations were assigned an ICD-9 or -10 diagnostic code for KD between years 2008 and 2017. The overall hospitalization rate for KD in this population was 5.5 per 100 000 children over the entire study period. The baseline characteristics of these hospitalizations are presented in Table 1. Briefly, 75% of KD hospitalizations were under the age of 5, 58.7% were males (male to female ratio of 1.5:1), 35.7% were White, and the majority (61.4%) were hospitalized in large teaching hospitals (85%).

The trends in KD hospitalization rates (per 100 000) according to gender, age, race, and the US census region are shown in Table 2. Hospitalization rate was higher in males, children aged ≤4 years, and Asian or Pacific Islanders. The KD hospitalization rate was higher in Blacks than in Whites (P = .01). The hospitalization rate also differed by census region with the highest in the West (6.1), followed by the Northeast (5.9).

There was no significant change in the KD hospitalization rate across the years of the study (4.6 in 2008 to 5.7 in 2017, P_trend = 0.18). Except for Native Americans, there were no significant changes in the KD hospitalization rate in all other subgroups when stratified by age, gender, race, and US census region. The KD hospitalization rate for Native Americans increased from 1 in 2008 to 4.5 in 2017 (P_trend = 0.048). The occurrence of CAAs were assessed in this population cohort as an indicator of CALs and presented as a percentage of the respective KD hospitalizations in Table 3. CAAs were most common in children <1 year of age (7.6%) and in males (4%). Among racial subgroups, the occurrence of CAAs was highest in Native Americans (9.2%), followed by Hispanics (4.6%), and Asian and Pacific Islanders (4%). The trend of overall occurrence of CAAs, shown in Fig 1, was noted to be mostly steady between years 2008 and 2014 (P_trend = 0.36), followed by a sudden steep rise to 6.8% in 2017 (P_trend = 0.04). In sensitivity analysis to determine the effect of the transition from ICD-9 to ICD-10 codes in October 2015, we examined the proportion of KD hospitalizations assigned ICD-10 codes for CAA for each month for the years 2014 through 2016. The proportion of KD hospitalizations with CAA was 5.2% in September 2015 (the last month for ICD-9 codes) and it increased to 7.1% in October 2015 (first month of ICD-10 codes), 5.3% in November, and 3.3% in December 2015 (Supplemental Table 1). Thus, the transition to ICD-10 codes did not significantly lead to a sudden increase in CAA coding in the NIS. There was no mortality noted among KD hospitalizations in children during the study period. Figure 2 illustrates the trends in the median inflation-adjusted hospital costs and LOS over the study period. While there was no significant change in the median LOS (2–3 days, P_trend = 0.3), the median cost of hospitalization increased significantly from $6819 in 2008 to $10 061 in 2017 (P_trend <0.001).

This nationwide study is an updated analysis of the trends in KD hospitalizations, prevalence of CAAs, and resource use from 2008 to 2017. We found that although the overall KD hospitalization rate remained stable, there was a significant increase among Native American children. The frequency of occurrence of CAA complicating KD hospitalizations remained stable until 2014, after which a significant increase was observed. LOS was stable, but a significant increase occurred in inflation-adjusted hospital costs. This study extends our knowledge on KD and highlights the novel finding that CAAs among KD hospitalizations may be increasing in the US.

The current study demonstrated a stable trend in KD hospitalizations from 2008 to 2017. This is in agreement with another population-based study from Olmsted County in Minnesota, which demonstrated a plateau in the incidence of KD from 1994 to 2016.²¹  Similarly, another population-based study from Ontario, Canada demonstrated a stable incidence rate of KD between 2006 and 2014 among children under 10 years old.²²  Our findings contrast with a previous report by Okubo et al, which was based on another HCUP administrative database, the Kids Inpatient Database. In that report, the KD hospitalization rate decreased marginally from 6.54 to 6.11 per 100 000 children between 2003 and 2012.⁷  The differences in these findings may be related to the different time periods covered by each study and differences in the two administrative databases. Further, our results confirm earlier reports of higher incidence of KD in children <5 years, males, and those of Asian and Pacific Islander ancestry.^7,9,23–25  However, when we analyzed trends for hospitalizations by race and ethnicity, there were no significant changes except for Native Americans. The reason for this trend is unclear, but it could be related to racial disparities and inequities in healthcare relative to the environmental and other risk factors for KD.

This study is also one of the largest to track the frequency of occurrence of CAAs associated with KD over an extended period. The occurrence of CAAs is reported to be <5%.¹  In the current study, the occurrence of CAAs remained stable between 1.8% to 2.9% between 2008 and 2014. This was surprisingly followed by a significant increase to 4% in 2015 and subsequently to 6.8% in 2017. To our knowledge, this is one of the few studies from the US demonstrating an increase in the occurrence of CAAs in children with KD. However, a recent study between 2017 and 2019, with 2080 children from 44 children’s hospitals in the US, reported that the occurrence of CAAs ranged between 7.7% and 9.4% across various models of care.²⁶  The reason for the increase in the administrative diagnosis of CAAs among KD hospitalizations is not clear from the published literature. Refractory KD (unresponsive to IVIG) is a risk factor for CAA.²⁷  A report by Moffet et al using data from 43 children’s hospitals within the Pediatric Health Information System in the US from 2004 to 2012 demonstrated no significant change in the proportion of patients with KD who had refractory KD.²⁸  Another study based on the Pediatric Health Information System spanning 2006 through 2015 reported an increase in the administration of infliximab, a second line agent for patients refractory to IVIG.²⁹  This probably suggests an increase in the proportion of patients with refractory KD, but it is not clear how this translates into increased frequency of occurrence of coding for CAAs. Son et al reported an increase in diagnostic coding for KD and CAA after the release of the American Heart Association guidelines on the diagnosis and management of KD.¹  The study period under reference was 2008 through 2017, and the most recent guidelines were published in 2017, thus making it unlikely to positively affect coding for CAA in the NIS.³⁰  However, our findings could suggest a delayed adoption of guidelines in 2004 recommending echocardiogram as soon as KD is suspected with more frequent follow-ups.³¹  Furthermore, Coon et al found that the rate of coronary artery abnormalities in KD patients doubled between 2000 and 2014, driven entirely by nonsevere forms, without measurable changes in the rate of adverse cardiac outcomes, suggesting an overdiagnosis of coronary artery abnormalities. It is plausible that the increased coding for CAAs in this study is the result of overdiagnosis.¹⁰ 

Finally, the HCUP transitioned from ICD-9 to ICD-10 diagnostic codes in the last quarter of calendar year 2015.³²  Our sensitivity analysis shows that this transition had little effect on the coding for CAA associated with KD. This rise in the administrative diagnosis of CAA associated with KD has several implications. First, the ICD-10 diagnostic code must be validated in future studies for future studies to be reliable. Second, an increase in CAAs implies that there is a growing population of children at risk for acquired heart disease such as angina and myocardial ischemia, for which these patients need long term surveillance to ensure better outcomes.

Native Americans and Native Alaskans comprise only <2% of the US population³³  and are often not analyzed or combined with other racial groups in medical research. It is of no surprise Native Americans were not studied in previous national surveys on KD hospitalizations.^7,8,9  A novel finding from the current study suggests an increase in the KD hospitalization rate among Native Americans during the study period, a finding not previously reported. Additionally, we found that the Native Americans had the highest frequency of coding for CAAs. This finding comports with those of 2 previous reports, though the number of Native Americans in these 2 studies were too small to make the findings statistically meaningful.^23,34  There is evidence to suggest that more Native Americans are living in metropolitan areas and accessing nonfederal hospitals (data captured in NIS) for their healthcare.³⁵  This might account for the increased trend in KD hospitalizations among Native Americans. An additional explanation for this trend is an increased commitment and an improved ability to capture race and ethnicity data in medical records. The disproportionately high coding for CAAs among Native Americans could represent yet another example of the racial and ethnic disparities in healthcare and healthcare quality that exists in the US.^36,37  This may stem from several issues related to barriers to access spelled out by the American Academy of Pediatrics Committee on Native American Child Health.³⁸  Further studies are needed to establish the true frequency of occurrence of CAAs in Native American children with KD.

Of the many publications on KD in the US, only a few have examined the economic burden associated with KD hospitalizations.^9,39–41  Hospital charges or costs are directly related to LOS.^42–44  The median LOS from this study was 2 to 3 days and is consistent with previous reports.^{23–25,40,45}  Although the median LOS remained stable over time, the inflation-adjusted hospital cost increased significantly during the study period. First, this could be due to increased use of advanced cardiac imaging, such as computed tomography angiography and magnetic resonance angiography of the coronary vessels in the latter part of the study period. Second, substantial variability exists in the treatment of refractory KD.³⁰  Studies have demonstrated that that the use of infliximab or steroids as first line agents for refractory KD is more effective that repeat IVIG, but most of the patients with refractory KD receive IVIG first.^10,46  Further, Carmen Johnson et al compared the cost-effectiveness of infliximab versus IVIG for refractory KD and found that infliximab treatment was associated with significantly shorter LOS and lower direct hospital costs.⁴⁶  Further comparative cost-effectiveness studies and the standardization of the treatment of refractory KD can yield substantial healthcare savings.

The major strength of the study stems from the use of a population-based database covering the entire US population, making the findings from the current study nationally representative. Additionally, we have provided insights into the resource use by KD hospitalizations and can serve as a reference point for future studies, as well as helping administrators and policymakers with resource allocation. The limitations of this study, like all others conducted using administrative databases are well described.⁴⁷  Large databases, such as the NIS, are susceptible to coding errors and duplications. However, the HCUP has instituted mechanisms to ensure the validity of the data in the NIS.⁴⁸  Each discharge entry in the NIS is a hospitalization and not a patient since a patient can be readmitted for recurrent KD. Thus, the hospitalization rate might overestimate the true incidence of KD in the US. Furthermore, there are no studies to date that have validated the ICD-9 and ICD-10 diagnostic codes for KD and CAA. Three studies from Georgia and California found the positive predictive values for the KD diagnostic codes to be 74% to 86%. Though this may indicate the hospitalization rates presented in this study are underestimated, further studies exploring the negative predictive value and sensitivity of ICD codes is warranted.⁴⁹  However, most of the previous studies with large sample sizes used these diagnostic codes, making the current study comparable in methodology.^{7–9,28,34,50}  The dataset also limits exploration of patients with KD admitted primarily for a cardiovascular procedure since it would be difficult to differentiate an initial presentation from a readmission. Future studies that compare patient cohorts with short versus longer hospitalizations for KD may help delineate clinical, geographic, and social patterns that contribute to the duration of hospital stay.

This population-based study demonstrated that between 2008 and 2017, the overall KD hospitalization rate in the US remained stable, but the frequency of coding for CAAs increased significantly after 2014. There was no mortality recorded. While LOS remained stable, inflation-adjusted hospital cost increased significantly. The increased coding for CAAs require further investigation and validation.