The contribution of sociodemographic and clinical factors to length of stay in hospitalized children
MetadataShow full item record
BACKGROUND: There is continued attention towards using patient demographic and clinical characteristics available in health administrative data when case mix adjusting the measurement of length of stay (LOS) for hospitalized children. However, little is known about what proportion of children’s LOS is explained by these characteristics. OBJECTIVES: The objectives of the study were to quantify the amount of variation in LOS within and across hospitals that is explained by demographic and clinical factors of hospitalized pediatric patients. METHODS: A retrospective cohort analysis was completed of 818,848 hospitalizations for any reason occurring from 1/1/2014 to 12/31/2014 in one of 44 freestanding children’s hospitals in the Pediatric Health Information Systems (PHIS) dataset. A generalized linear model was derived to simultaneously regress demographic factors [age, race/ethnicity, payer, rural residence, health professional shortage area (HPSA) residence, income, and distance traveled], and clinical factors (severity of illness, type and number of chronic conditions) on LOS. The percentage of LOS attributable to each characteristic within each hospital was quantified using the covariance test of the hospital random effect. RESULTS: The factors with the greatest impact on LOS were severity of illness and chronic condition type and number, with a median (interquartile range) of 16.8% (IQR 15.0%-19.4%) and 4.0% (IQR 2.9%-4.5%) of LOS, respectively, explained by these characteristics across hospitals. LOS varied significantly (p<0.05) with both severity of illness and chronic condition type and number for all 44 hospitals in the cohort. All patient demographic factors, (age, race/ethnicity, payer, rural residence, HSPA residence, income, and distance traveled) had minimal impact on LOS, with <0.1% of LOS explained by each characteristic. Across hospitals, 78.3% (IQR 75.8-80.2%)] of LOS remained unexplained by the patient characteristics under study. CONCLUSIONS: Patients’ clinical characteristics ascertained from administrative data account for approximately one-fifth of LOS whereas their demographic characteristics account for a negligible amount. Efforts to optimize the efficiency of inpatient care for hospitalized children might benefit from uncovering how much of the vast amount of unexplained LOS is due to modifiable aspects of care quality.