The dataset contained records for 877,201 hospitalizations of children under 18 at time of admission with 50,879 (5.8%) being repeat admissions. The study population was 50.9% male and 49.1% female, 48.5% white, 16.1% African-American, 12.2% Hispanic, 3.6% Asian/Pacific Islander, 18.7% other and less than 1% American Indian. The average length of stay was 3.7 days (median 2 days, interquartile range 2–3 days). Private insurance paid for 46.2% of the hospital stays, government paid for 45.9% and self-pay, no charge or other sources accounted for 7.9% of the hospital stays. Most of the hospitalizations took place in urban areas (89.2%) compared to rural areas (10.8%).
We validated the Premier sample of hospitalizations by comparing characteristics of the sample to the HCUP KID sample of pediatric hospitalizations for 2006 and details of the validation are available in a previous publication (Lasky et al., 2011). The Premier sample included a greater proportion of infants born in the hospital, from Southern hospitals, from non-teaching hospitals, and from large size hospitals compared to the HCUP KID sample. The two samples were similar with regards to proportions male, routine discharge status, APR-DRGs severity and proportions urban. We did not compare the proportions of different racial and ethnic groups in Premier to the KID because of well documented limitations in racial and ethnic data within the KID, high rates of missing data resulting from state differences in collection and reporting of race and ethnicity. Vancomycin was administered in 19,775 hospitalizations, or 2.3% of 877,201 pediatric hospitalizations in the database in 2008. In 98% of cases, vancomycin was administered parenterally or “other” (this includes ophthalmic solutions, intraocular/intravitreal injections, catheter flushes, inhalation formulations, rectal formulations and topical gels compounded by pharmacy) and in less than 2% of cases, vancomycin was administered orally. Half (10,033 or 50.7%) of the courses of vancomycin were less than 3 days duration. Males had higher prevalence of use than did females (2.5%, 95% CI 2.5–2.6 compared to 2.0%, 95% CI 2.0–2.0), African-Americans had higher use than did whites or other groups (3.1%, 95% CI 3.0–3.2, 2.3%, 95% CI 2.3–2.4 and 1.8%, 95% CI 1.8–1.9 respectively), and children in the age groups 2–4 and 5–11 had higher prevalence compared to children under 2, or children age 12–17 (6.8%, 95% CI 6.5–7.1, 6.6%, 95% CI 6.4–6.8, 1.5%, 95% CI 1.5–1.5, and 4.3%, 95% 4.2–4.4). The greatest number of hospitalizations with vancomycin use occurred to children under 2 (10,282 or 52% of hospitalizations), however the highest prevalence of use occurred in children age 2–4, and 5–11 (). Children age 2–4 were 4.5 times more likely, children 5–11 were 4.4 times more likely, and children 12–17 were 2.9 times more likely to receive vancomycin compared to children under 2. In hospitalizations of children under one year with vancomycin use, the four most frequent ICD-9 group diagnoses were: “Liveborn infants according to type of birth” (ICD-9-CM V30–39) (51.83%), “Other conditions originating in the perinatal period” (ICD-9-CM 760–779) (14.29%), “Infections of skin and subcutaneous tissue” (ICD-9-CM 680–709) (6.07%), and “Congenital anomalies” (5.91%) (ICD-9-CM 740–759). In children age 1 or over, the four most frequent diagnoses were: “Infections of skin and subcutaneous tissue” (23.32%), “Pneumonia and influenza”, (9.41%) (ICD-9-CM 480–488), “Complications of surgical and medical care, not elsewhere classifiable” (7.58%) (ICD-9-CM 996–999), and “Other bacterial diseases” (4.41%) (030–041).
Vancomycin was administered to children at 374 hospitals in the Premier hospital database; another 47 hospitals with 17,271 pediatric hospitalizations (13,233 under age 2) reported no vancomycin use during 2008. The number of hospitalizations with vancomycin use ranged from 0 to 1225 at individual hospitals, and percentage of hospitalizations with vancomycin use ranged from 0.0% to 33.3%. Twenty one hospitals (5.6%) had more than 200 hospitalizations with vancomycin use, accounting for 9,979 (50%) of the pediatric hospitalizations with vancomycin use. Because of the skewness in distribution of vancomycin use, we stratified hospitals by number of hospitalizations with vancomycin use. Low volume was defined as 0 to 10, medium volume as 11–100, and high volume as over 100. Most hospitals were categorized as low volume (221 hospitals), 155 hospitals were categorized as medium volume, and 45 hospitals categorized as high volume. Within high volume hospitals, percentage of hospitalizations with vancomycin use ranged from 1.3–12.9 (mean percentage was 4.6, 95% CI
3.9–5.4). Within medium volume hospitals, percentage of hospitalizations with vancomycin use ranged from 0.3 to 9.5 (mean percentage was 1.7, 95% CI
1.5–1.9). The hospitals with high volume of vancomycin use were predominantly large (73.3%), teaching (68.9%), and urban (97.8%) compared to the hospitals with low volume of vancomycin use, which were 45.7% large, 19.91% teaching, and 74.21% urban, and 95% Confidence Intervals of estimates generally did not overlap. In 2008, 47 hospitals, or 11.16% of the hospitals in the database, reported no vancomycin use in the entire year.
The logistic mixed effects modeling showed hospital variation in vancomycin use. The estimated variance of the random effects hospital intercepts for models run for children under 1, and for the four most frequent ICD-9 Code Groups are summarized in . The intercept estimates and the 95% Confidence Bounds for each model do not include zero, indicating hospital variation that is statistically significant after controlling for the other variables in the model. The lower limits of the 95 percent confidence intervals are above zero, indicating statistically significant variability in the use of vancomycin depending on the hospital in which a patient was treated.
The estimates and 95% Confidence Bounds of the variance of the random hospital intercepts on the logit scale in four models for individual diagnostic groups in children under 1.
The Odds Ratios for hospital and patient fixed effects for two of the models are plotted in and . Variables associated with vancomycin use were different in each of the age and ICD groups. For example, in children under 1 year with ICD-9 group “Liveborn infants according to type of birth” increased APR-DRG severity of illness was associated with over 6 times the use of vancomycin, but in children under 1 year with ICD-9 group “Infections of skin and subcutaneous tissue” use of vancomycin was almost 2 times as frequent in children with increased APR-DRG severity (although the 95%CI slightly overlapped 1). Both models showed an association between increased APR-DRG severity of illness and vancomycin use, but the magnitude of the effect differed greatly in the two patient groups. Another example is the effect of rural vs. urban status of the hospital. Rural or urban status of the hospital was statistically significant in predicting vancomycin use in children under 1 year with ICD-9 group “Liveborn infants according to type of birth” and with ICD-9 group “Infections of skin and subcutaneous tissue”, but the direction of the effect was different in each group of patients. For children under 1 year with ICD-9 group “Liveborn infants according to type of birth” rural hospitals had lower vancomycin use than did urban hospitals, but for children under 1 year with ICD-9 group “Infections of skin and subcutaneous tissue” rural hospitals had higher vancomycin use than did urban hospitals. Another example can be seen with patient’s insurance coverage. In children under 1 year with ICD-9 group “Liveborn infants according to type of birth” those with government insurance had slightly higher vancomycin use than did those with private insurance, and those with no insurance had less vancomycin use than those with private insurance. In children under 1 year with ICD-9 group “Infections of skin and subcutaneous tissue” insurance coverage was not associated with vancomycin use.
Odds Ratios for models in patients under 1 year of age ICD-9-CM group “Liveborn infants according to type of birth”.
Odds Ratios for models in patients under 1 year of age ICD-9-CM group “Infections of skin and subcutaneous tissue”.