Comparison of Parent Report with Administrative Data to Identify Pediatric Reutilization Following Hospital Discharge
Healthcare providers rely on historical data reported by parents to make medical decisions. The Hospital to Home Outcomes (H2O) trial assessed the effects of a one-time home nurse visit following pediatric hospitalization for common conditions. The H2O primary outcome, reutilization (hospital readmission, emergency department visit, or urgent care visit), relied on administrative data to identify reutilization events after discharge. We sought to compare parent recall of reutilization events two weeks after discharge with administrative records. Agreement was relatively high for any reutilization (kappa 0.74); however, this high agreement was driven by agreement between sources when no reutilization occurred (sources agreed 98%-99%). Agreement between sources was lower when reutilization occurred (48%-76%). Some discrepancies were related to parents misclassifying the site of care. The possibility of inaccurate parent report of reutilization has clinical implications that may be mitigated by confirmation of parent-reported data through verification with additional sources, such as electronic health record review.
© 2019 Society of Hospital Medicine
Prior healthcare utilization predicts future utilization;1 thus, providers should know when a child has had a recent healthcare visit. Healthcare providers typically obtain this information from parents and caregivers, who may not always provide accurate information.2-4
The Hospital to Home Outcomes study (H2O) was a randomized controlled trial conducted to assess the effects of a one-time home nurse visit following discharge on unplanned healthcare reutilization.5 We assessed reutilization through two sources: parent report via a postdischarge telephone call and administrative data. In this analysis, we sought to understand differences in reutilization rates by source by comparing parent report with administrative data.
METHODS
The H2O trial included children (<18 years) hospitalized on the hospital medicine (HM) or neuroscience (Neurology/Neurosurgery) services at Cincinnati Children’s Hospital Medical Center (CCHMC) from February 2015 to April 2016; they had an English-speaking parent and were discharged to home without skilled nursing care.6 For this analysis, we restricted the sample to children randomized to the control arm (discharge without a home visit), which reflects typical clinical care.
We used administrative data to capture 14-day reutilization (unplanned hospital readmissions, emergency department [ED] visits, or urgent care visits). CCHMC is the only pediatric admitting facility in the region and includes two pediatric EDs and five urgent care centers. We supplemented hospital data with a dataset (The Health Collaborative7) that included utilization at other regional facilities. Parent report was assessed via a research coordinator phone call 14-23 days after discharge. Parents were asked: “I’m going to [ask] about your child’s health since [discharge date]. Has s/he been hospitalized overnight? Has s/he been taken to the Emergency Room/Emergency Department (didn’t stay overnight)? Has s/he been taken to an urgent care?” We report 14-day reutilization rates by source (parent and/or administrative) and visit type.
We considered administrative data the gold standard for documentation of reutilization events for two reasons. First, all healthcare encounters generate billing and are therefore documented with verifiable coding. Second, we had access to data from our center and other regional healthcare facilities. Any parent-reported utilization to a facility not documented in either dataset was considered an unverifiable event (eg, outside our catchment region). Agreement between administrative and parent report of 14-day reutilization was summarized as positive agreement (reutilization documented in both administrative and parent report), negative agreement (no reutilization reported in either administrative or parent report), and overall agreement (combination of positive and negative agreement). We classified discrepancies as reutilization events in administrative data without parent report of reutilization or vice versa. We performed medical record review of discrepancies in our institutional data.
We summarized agreement by using the Cohen’s kappa statistic by reuse type (hospital readmission, ED, and urgent care visit) and overall (any reutilization event). Strength of agreement based on the kappa statistics was classified as poor (<0.20), fair (0.21-0.40), moderate (0.41-0.60), good (0.61-0.80), and very good (0.81-1.00).8 We used McNemar’s test to evaluate marginal homogeneity.
