Hospital Characteristics and Quality of Pediatric Inpatient Care: A Multi-Level Analysis

Pediatric patients with severe illnesses or with multiple, complex chronic conditions are most likely to experience hospital-acquired conditions (HACs) (Dynan et al, 2013), an indication of low-quality care.  Identification of patient characteristics associated with HACs provides opportunities for increased care vigilance.  However, pediatric HAC rates vary substantially across US hospitals. Understanding the nature of hospital variation associated with pediatric HAC can assist quality improvement efforts across hospitals in identifying evidence-based best practices and procedures for patient care and training.

In this study, we identified hospital characteristics associated with the presence of pediatric HAC after adjusting for discharge-level patient demographics, insurance status, and increased risk of HAC due to medical complexity. Our outcome was a dichotomous composite variable that identified discharges with at least one of eight associated Agency for Health Care Research and Quality's pediatric quality indicator (PDI) events. (Smith et al, 2012) We adjusted for discharge-level risk using a high dimensional propensity score compiled from 278 clinical classification diagnoses. Multi-level statistical analysis was performed.

We analyzed data from the 2009-2011 Healthcare Cost and Utilization Project annual Nationwide Inpatient Sample databases.  We link these data with 2010 American Hospital Association annual data.  Preliminary analyses suggested that HACs occurred disproportionately and predominantly in teaching hospitals.   As a result, we restricted our study to US teaching (major, minor and children’s) hospitals treating at least 10% or 100 pediatric patients.  At the discharge-level, we included all pediatric discharges that were at risk for any of 8 PDIs comprising our dichotomous composite measure of HAC. We excluded children born in hospital, and discharges where a present on admission condition or transfer (in or out) was recorded. 

Hospitals with less than 0.25 residents per non-long term care bed (OR=0.70; p=0.009) have lower rates of adjusted HACs than hospitals with 0.25 residents or more per bed. Hospitals with 5 measures of hospital technology have higher adjusted HAC rates than hospitals with 1 (OR=0.18, p<0.001), 2(OR=0.49, p<0.001), 3 (OR=0.50 p<0.001), or 4 (OR=0.75, p=0.318) measures. Compared to discharges from children’s hospitals, children with inpatient stays in non-major teaching hospitals experience lower adjusted HACs (OR=0.38, p<0.001).   Hospitals in the Northeast and South have lower adjusted HAC rates than hospitals in the West region (OR=0.51; p=<0.001 and OR=0.70; p=0.006, respectively). 

We find no difference in adjusted HAC rates over the period studied. Contrary to previous findings, hospital size, discharge quarter (“new residents” July 1 effect), and safety net status based on share of total discharges paid by Medicaid are not associated with adjusted HAC rates.

            Although teaching hospitals are essential for training the next generation of physicians, our study indicates that after adjusting for patient-level factors there remains inherent increased patient safety risks for children seen in teaching hospitals associated with factors such as congestion (higher resident rate per bed) and the increased use of technology. We discuss the potential for process improvement by focusing on identifying positive deviance—that is, by determining if there are hospitals that have achieved high value pediatric inpatient care under such conditions.