Vaccine Administration in Children’s Hospitals

We conducted a retrospective cohort study of vaccine administration in children’s hospitals using the Pediatric Health Information System (PHIS) database. PHIS is an administrative billing database administered by the Children’s Hospital Association (Lenexa, KS) that includes pharmacy data for inpatient encounters for 49 US children’s hospitals. Hospitalizations for children <18 years old admitted to a PHIS hospital between July 1, 2017, and June 30, 2019, were included. We included all children <18 years as routine and catch-up vaccinations are relevant to all ages.³⁰  We excluded children who had a condition that would alter the recommended vaccination schedule according to ACIP guidelines by using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) codes for diagnoses and Pharmaceutical Clinical Transaction Classification (CTC) codes for medications, such as chemotherapeutic drugs (Supplemental Table 5). Children who died during hospitalization or were discharged to hospice care were excluded. This study was categorized as not human subjects research by the Institutional Review Board at Seattle Children’s Hospital.

Our primary outcome was administration of ≥1 dose of any vaccine type to children who were age eligible during an inpatient hospitalization in the study period. Our secondary outcome was administration of ≥1 dose of any recommended childhood vaccine other than influenza and the hepatitis B birth dose, described as routine childhood vaccines for the remainder of this article.

Vaccine administration data were obtained from the PHIS database by using CTC codes for the following vaccines: diphtheria, tetanus, acellular pertussis (DTaP/Tdap); hepatitis B; rotavirus; Haemophilus influenzae type B; pneumococcal conjugate; inactivated poliovirus (IPV); measles, mumps, rubella (MMR); varicella; hepatitis A; meningococcal; human papillomavirus (HPV); and influenza. Hepatitis B vaccine doses were dichotomized into birth dose (if administered at <31 days of age) and nonbirth dose (if administered at ≥31 days of age).

Children were considered age eligible for a vaccine dose if their admit or discharge age was within the recommended age window for receiving a childhood vaccine dose according to the routine and catch-up vaccination schedules from the Centers for Disease Control and Prevention (CDC) (Table 1).³⁰  Children were considered eligible for influenza vaccine doses if their admission date or discharge date was during influenza season (September through April).

Demographic variables including age, race, ethnicity, and insurance status were obtained from PHIS. Medical complexity was assessed by using the Pediatric Medical Complexity Algorithm (PMCA).³¹  Age was examined as a continuous variable and categorized on the basis of age categories used by the CDC to estimate population-level vaccine coverage (0–18 months, 19–35 months, 3–6 years, 7–12 years, and ≥13 years).³² 

We obtained length of stay (LOS), ICU status, and season of admission for each hospitalization from PHIS. We examined mean LOS in days and also assessed LOS categorically (<3 days, 3–7 days, 7–30 days, and >30 days). Season of admission was categorized as influenza season (September through April) or noninfluenza season (May through August). We included these variables because we hypothesized that they may be related to vaccine administration during hospitalization; for example, children with longer LOS may be administered more vaccine doses.

We used descriptive statistics to assess (1) the number and proportion of each type of vaccine administered during hospitalization and (2) the number and proportion of hospitalizations where children were administered ≥1 vaccine dose of any type for which they were age eligible. To describe variation by vaccine type and hospital, we first calculated unadjusted proportions with 95% confidence intervals (CIs) of age-eligible hospitalizations in which ≥1 dose was administered for each vaccine for each hospital. We risk adjusted hospital-level rates of vaccine doses administered per 1000 hospitalizations by including hospital-level covariates in a generalized linear mixed-effects regression model. Covariates in the model were age, sex, race and ethnicity, insurance, PMCA, and LOS with hospital as a fixed effect. Because population-based rates of vaccination vary nationally, we adjusted the model to account for population-level rates by age based on the CDC’s population-level estimates (Supplemental Table 6).³²  To model the relationship between administration of doses of different vaccines at the hospital level, risk-adjusted vaccine administration rates were divided into quintiles on the basis of the percent difference from the risk-adjusted median rate of vaccine administration per 1000 hospitalizations by vaccine type. Given the number of hospitals and vaccines included in this study, a heat map was developed to display the risk-adjusted median rates of vaccine administration in quintiles.^33,34  Quintiles were created to classify hospitals with risk-adjusted median rates at (1) >25% below the median, (2) 10% to 25% below the median, (3) <10% above or below the median, (4) 10% to 25% above the median, and (5) >25% above the median.

Lastly, we used descriptive statistics to examine demographic, clinical and hospital visit characteristics of children who received ≥1 vaccine dose of any routine childhood vaccine (excluding influenza and hepatitis B birth dose). We examined the relationship between receipt of any routine childhood vaccine excluding influenza and hepatitis B birth dose and demographic, clinical, and hospital visit characteristics using linear regression models for continuous variables and logistic regression models for categorical variables.

We conducted a sensitivity analysis excluding hospitalizations of children who were admitted at <2 months of age. We hypothesized that there may be differences in vaccine administration for these children because they were only age eligible for hepatitis B vaccines at the time of admission. Furthermore, this population includes many children who were admitted to the NICU. Children in the NICU are at risk for undervaccination, and there may be hospital-level policies related to vaccinations for this population.^26,35–39  We examined the demographic, clinical, and hospital visit characteristics associated with administration of ≥1 noninfluenza, nonbirth dose hepatitis B vaccine dose in children ≥2 months of age at the time of hospitalization.

There were 1 536 340 hospitalizations involving 1 185 667 children who met inclusion criteria during the study period. The mean age was 5.5 years (SD 5.8 years); 46% were female, 43% were White, 18% were non-Hispanic Black, and 55% had public insurance. The mean hospital LOS was 4.8 days (SD 13.9). Twelve percent of children were admitted to the ICU; 70% were admitted during influenza season (September through April) (Table 2).

There was ≥1 vaccine dose administered in 198 076 of the 1 536 340 hospitalizations included in this study (12.9%; 95% CI: 12.8–12.9). A total of 272 361 vaccine doses were administered. The proportion of hospitalizations where ≥1 vaccine dose was administered varied by hospital from 1% to 45% (Fig 1, Supplemental Tables 7 and 8). The most common types of vaccines administered were hepatitis B (n =142 861; 9.3% [95% CI: 9.25%–9.34%] of hospitalizations with age-eligible children) and influenza (n = 48 716; 6.1% [95% CI: 6.07%–6.18%] of age-eligible children admitted during influenza season) (Table 2). Eighty-eight percent of hepatitis B vaccine doses were given to children age <31 days. In 1.9% (95% CI: 1.86%–1.90%, n = 28 912) of hospitalizations, a child was administered ≥1 vaccine doses other than influenza vaccine or birth dose of hepatitis B vaccine, most commonly pneumococcal conjugate, H influenzae type B, or DTaP (Table 1).

Overall, the adjusted hospital-level median rate of vaccine administration was 88.5 per 1000 hospitalizations (25% to 75% interquartile range [IQR]: 52.8–205.7). The adjusted hospital-level median rate of vaccine doses administered per 1000 hospitalizations varied by vaccine and by hospital (Fig 1). The most common vaccine doses administered were the birth dose hepatitis B vaccine (adjusted median rate: 193.9 [IQR: 133.3–496.9]) and influenza vaccine (adjusted median rate: 58.1 [IQR: 35.6–97.2]) (Fig 1, Supplemental Tables 7 and 8). There were a number of hospitals that delivered no vaccine doses for certain vaccine types: rotavirus (n = 7 hospitals), MMR (n = 2), varicella (n = 1), hepatitis A (n = 2), and HPV (n = 11) (Fig 1, Supplemental Tables 7 and 8).

Compared with children age 3 to 6 years, children who were admitted at age <2 months had the highest odds of being administered ≥1 dose of a routine childhood vaccine, excluding influenza and the hepatitis B vaccine birth dose (n = 14 061, odds ratio: 8.9; 95% CI: 8.4–9.4). All age groups had higher odds of being administered any routine childhood vaccine during hospitalization compared with children age 3 to 6 years (Table 3). Children who were identified as non-Hispanic Black and other race had higher odds of receiving routine childhood vaccines during hospitalization compared with children who were non-Hispanic White race and Hispanic ethnicity. Children with public insurance had higher odds of receiving ≥1 routine childhood vaccine during hospitalization compared with children with private insurance (odds ratio: 1.4; 95% CI 1.4–1.4) (Table 3). Children with higher levels of medical complexity (noncomplex chronic and complex chronic) had higher odds of receiving ≥1 routine childhood vaccine compared with children with no chronic disease. Children who had a longer LOS had higher odds of receiving any routine childhood vaccines compared with children who had a LOS of <3 days.

When we excluded children who were admitted at <2 months of age (n = 14 061), we identified similar associations between vaccine administration and age, race, ethnicity, insurance status, ICU admission, and LOS as the primary analysis of children age ≥2 months (n = 14 851) (Table 4).

In this retrospective cohort study, we examined vaccine administration during hospitalization using a national sample of US children’s hospitals. We identified that vaccine type, patient demographic, and hospital visit characteristics were all associated with receiving ≥1 vaccine dose during hospitalization. The most common vaccines administered were the birth dose of hepatitis B vaccine and influenza vaccines, with <2% of hospitalizations involving the administration of any other recommended childhood vaccines. Furthermore, there was considerable variability in the proportion of hospitalized children who received ≥1 vaccine dose by hospital.

This study highlights several important aspects of inpatient vaccine delivery. First, certain doses of childhood vaccines were administered more than others, with the birth dose of hepatitis B vaccine and influenza vaccine doses constituting the majority of vaccine doses administered to hospitalized children. This suggests that there are existing systems in place to deliver these vaccines to hospitalized children, consistent with previous literature.^{14,19,21,23,25,26,37,40} 

There are several reasons why influenza vaccination during hospitalization may be prioritized compared with other childhood vaccines. First, children may not have seen their primary care provider during the current influenza season, and therefore, hospitalization affords an opportunity to vaccinate.^19,41–43  Additionally, parent report of their child's influenza vaccine status may be more trusted^44–46  compared with other childhood vaccines^10,47–50  because parents only need to recall if their child has received a single vaccine in the past year, compared with the entire CDC vaccine schedule with >20 vaccine doses by 18 months of age.³⁰  Furthermore, the Joint Commission has identified inpatient influenza vaccination as a clinical quality measure for hospitals, which may lead to prioritization of influenza vaccine delivery.⁵¹  Expansion of quality measures related to noninfluenza vaccines may help catalyze efforts to increase inpatient vaccinations.

Of note, the PHIS database does not contain individual vaccine records or information about vaccine eligibility aside from age. This limitation highlights the well-documented challenge of identifying who is vaccine eligible during hospitalization.^{10,11,13,14,16}  With the growth of population-based IISs nationally, better integration of IISs with hospital electronic medical records and administrative databases such as PHIS could help in identifying opportunities to vaccinate hospitalized children.⁵²  Use of IIS to identify opportunities to vaccinate children during hospitalization aligns with a key objective for the Healthy People 2030 goals to increase the proportion of people with vaccination records in an IIS.⁴  Although PHIS does not contain individual vaccine eligibility data, ≥1 vaccine doses were delivered in only 12.9% of hospitalizations, well below current estimates of the proportion of hospitalized children who are undervaccinated (27% to 84%).^10–14  This discrepancy, and the variability between hospitals in vaccine administration identified in this study, suggests there are likely missed opportunities to vaccinate hospitalized children on a national scale.

Using administrative data from 49 children’s hospital, we identified significant variation in vaccine administration by hospital. Even after adjusting for state-level vaccination rates, hospital-level demographic characteristics, and hospital-level case mix, some hospitals were high performers with regards to administering vaccines. It is noteworthy that some hospitals (hospitals 46–49) were high performers across different types of vaccines. This finding suggests that these hospitals may have developed effective systems to identify undervaccinated hospitalized children and deliver needed vaccines to them. Conversely, there were several hospitals that delivered no vaccines of certain types, particularly live vaccines (rotavirus, MMR, and varicella) and HPV. A better understanding of existing mechanisms to vaccinate hospitalized children across hospitals, such as hospital-level vaccination policies and strategies (eg, vaccine prompts, workflow processes), may help elucidate why children at some hospitals were more likely to receive vaccines and identify key targets for quality improvement efforts to improve inpatient vaccine delivery.²⁹ 

There may be factors that are external to hospitals that contribute to inpatient vaccine rates. For example, although 49 states have an IIS, there is variation by state in the proportion of children who participate.⁵³  A higher proportion of children participating in IIS may make it easier for hospitals to identify needed vaccinations and to communicate inpatient vaccinations to the medical home. Another important factor that has yet to be fully explored is the cost to hospitals and patients who receive vaccines during hospitalization. Future work is needed to identify effective strategies used by high-performing hospitals and key contextual factors associated with high performance to improve vaccine delivery to hospitalized children nationally.

Another key finding was that certain children had higher odds of being administered ≥1 vaccine during hospitalization: children <3 years or >6 years of age, children who were identified as Black or “other” race, children with public insurance, those with higher levels of medical complexity, and those who had a LOS >3 days. These characteristics, including non-Hispanic Black race and having public insurance, are consistent with population-level estimates associated with lower vaccine coverage.⁵⁴  We hypothesize that differences in vaccine receipt during hospitalization may also be due to younger children and children with longer LOS becoming vaccine eligible during hospitalization or the medical team having additional time to identify the need for vaccination. Furthermore, barriers to accessing vaccinations in traditional settings, such as primary care, may be higher for children with public insurance^55–57  and medical complexity,^58,59  which may increase the need for catch-up vaccines in these populations and make inpatient vaccination a higher priority.

Since the onset of the coronavirus disease 2019 pandemic in March 2020, the need for this work is particularly relevant. Childhood vaccination rates have declined nationally because of a reduction in outpatient vaccinations during the coronavirus disease 2019 pandemic,^60,61  increasing the risk of vaccine-preventable disease in children.⁶²  As a result, there is renewed interest in strategies to capture all available opportunities to provide routine childhood vaccinations, including hospitalization. With 3 million pediatric hospitalizations annually,^63,64  developing robust systems to identify and provide vaccines during hospitalization has the potential to improve childhood vaccination rates.

This study has several limitations. The primary limitation is the lack of individual vaccine eligibility within the data set. To help mitigate this, we only included children who were age eligible for vaccines and adjusted for state-level vaccination rates. By including all children who were age eligible per the CDC routine and catch-up vaccination schedules, rather than only those who were vaccine eligible, the calculated vaccination rates are likely lower than the true rates in vaccine-eligible children. Secondly, exclusion criteria for the study were broad to encompass all children who had conditions that would alter the routine childhood vaccination schedule per the ACIP recommendations.¹⁸  Realistically, many of these children would still have been eligible for several routine vaccines. Without vaccine eligibility data, we are unable to quantify missed opportunities by hospital or by individual.

Another limitation is that the PHIS database contains administrative billing data and is subject to the data quality issues inherent in database management.⁶⁵  The Children’s Hospital Association, which administers PHIS, conducts data quality checks to minimize errors to the accuracy and completeness of the data.⁶⁶  Lastly, PHIS is specific for children’s hospitals and findings may not be generalizable to other settings such as community hospitals or children’s units within general hospitals.

In this national study of US children’s hospitals, we examined vaccine administration of all recommended childhood vaccinations during hospitalization. We identified several significant factors associated with vaccine administration in hospitalized children: the type of vaccine, patient demographic characteristics, and hospital visit characteristics. Additionally, there was significant variability in vaccine delivery by hospital. Future work should identify strategies to improve rates of vaccines delivered during hospitalization nationally. This study highlights the need to improve integration of vaccine eligibility data into inpatient clinical and administrative databases to accurately identify vaccine-eligible children.

ACIP

Advisory Committee on Immunization Practices

CDC

Centers for Disease Control and Prevention

CI

confidence interval

CTC

Clinical Transaction Classification

DTaP

diphtheria, tetanus, acellular pertussis

HPV

human papillomavirus

ICD-10

International Statistical Classification of Diseases and Related Health Problems, 10th Revision

IIS

Immunization Information System

IPV

inactivated poliovirus

IQR

interquartile range

LOS

length of stay

MMR

measles, mumps, rubella

PHIS

Pediatric Health Information System

PMCA

Pediatric Medical Complexity Algorithm

UTD

up to date