Letter by Bachhuber and Lott Regarding Article, “Characteristics of Children Hospitalized With Infective Endocarditis”
To the Editor:
A recent study by Day and colleagues1 describes the inpatient prevalence and associated characteristics of pediatric patients hospitalized for infective endocarditis using data extracted from the 2000 and 2003 Kids’ Inpatient Database (KID). As described by the authors, KID is a stratified survey sample database of hospital discharges from 2784 institutions in 27 states in 2000 and 3438 institutions in 36 states in 2003.2 Because of the stratified design, a weight is assigned for each discharge in the dataset to produce national estimates.
The authors do not report whether the results of their analyses are weighted or unweighted. Using the International Classification of Diseases, 9th revision, clinical modification codes for infective endocarditis, the authors found 697 discharges in the 2000 KID and 891 in the 2003 KID. However, using the same codes the authors selected, we found 975 discharges in the 2000 KID and 1104 in the 2003 KID using discharge weights, indicating that the results reported by the authors were possibly unweighted and may not accurately reflect national estimates.
Unweighted analyses of stratified data must be interpreted with caution. For example, if patients with Staphylococcus as the causative organism were more frequent in the unweighted analysis than those with Streptococcus but had relatively lower discharge weights, the unweighted proportion of Staphylococcus might differ significantly from the weighted proportion. Such differences may be of considerable relevance to study conclusions and clinical practice.
Day MD, Gauvreau K, Shulman S, Newburger JW. Characteristics of children hospitalized with infective endocarditis. Circulation. 2009; 119: 865–870.
Healthcare Cost and Utilization Project (HCUP). Design of the HCUP Kids’ Inpatient Database (KID), 2003. Rockville, MD: Agency for Healthcare Research and Quality, 2006.