TABLE 3

Main Findings of Included Observational Studies

First Author (Year)DesignSample Size/PopulationIntervention/Study FocusMedical ComplexityPreventable HospitalizationAssociation DirectionaMajor Findings
Armour (2009)26Retrospective cohort4395 individuals aged 0–64 y with spina bifida and private employer-sponsored insurance (or their dependents)Hospitalization rates and hospital expenditures due to urinary tract infections in this population versus population without spina bifidaSpina bifidaUTIPatients with spina bifida had 22.8 UTI discharges/1000 persons over 3 y compared with 0.44/1000 persons without spina bifida. Of 81 hospitalized for UTI, 73 had ambulatory claim within 7 d before hospitalization (34 diagnosed with UTI). Reduction in UTI hospitalization by 50% may reduce expenditures by $4.4 million/1000 patients.
Berry (2011)23Retrospective cohort317 643 children hospitalized at 1 of 37 freestanding children’s hospitals across the United StatesCharacterize children with recurrent 365-d readmissions to children’s hospitals1. CCCACSC readmissionSmaller % of hospitalizations associated with ACSCs in those with ≥4 readmissions (14% of hospitalizations versus 23.1% among those with 0 readmissions, P < .001). Among those with ≥4 readmissions (<3% of sample): accounted for 18.8% admissions, 23.4% bed days, and 23.2% charges, 52.6% had technology assistance (vs 5.3% with 0 readmissions, P < .001), technology complications noted in 8.7%, CCC in 89% (vs 22.3% with 0 readmissions, P < .001). Most prevalent CCCs were neuromuscular (39.6%) and cardiovascular (22.4%). Asthma, pneumonia, gastroenteritis, and seizure = most common ACSC hospitalization among patients with ≥4 readmits and account for 80% of their ACSCs.
2. Technology assistanceReadmission
Berry (2011)32Retrospective cohort1083 hospitalized patients who participated in a structured inpatient and/or outpatient complex care program in 4 different children’s hospitalsHospitalization characteristics among patients receiving care coordination in complex care programs1. CCC30-d Readmissions30 d readmit rate = 25.4% (mean 3.1 ± 2.8 admits per patient over 2 y). Highest rates among patients with technology assistance, neurologic impairment or other CCCs, compared with patients in complex care program without these characteristics and compared with nonprogram patients without these characteristics. Most common principal diagnoses for all hospitalizations were respiratory (29%; pneumonia/bronchiolitis 9.4%, asthma only 1.8%), followed by GI nutrition (15.8%; vomit and feeding difficulties at 3.4%), and followed by technology problem (9%).
2. Neurologic impairment
3. Technology assistance
4. Program enrollment criteria = different combinations of high levels of subspecialty use, organ system involvement, technology assistance, past utilization
Cooley (2009)37Retrospective cohort43 primary care practices (of 60 identified) in networks from 7 health plans in 5 states that volunteered to participate (6 plans exclusively Medicaid, and seventh was commercial)Hospital utilization associations with MHI scoresInvestigator selected conditions: cerebral palsy, epilepsy, autism, ADHD, asthma, diabetesVariation in hospitalization ratesReduced hospitalizations with higher MHI score (β − 0.19, P < .01), and subcomponents: organizational capacity, chronic condition management, care coordination and data management (β −0.14 to −0.2) after controlling for chronic condition. Majority of sample for utilization analysis had asthma or ADHD (90.9%), <6% had other diagnoses.
Dosa (2001)33Prospective cohort248 children under 18 y with chronic conditions and unscheduled ICU admissions related to the chronic condition admitted to the only ICU in a largely rural 17-county region in upstate New YorkRelative risk of unscheduled pediatric ICU admissions in children with chronic health conditions compared with those without1. CSHCN, definition from Maternal and Child Health BureauChart review by investigators: events leading to hospitalization categorized into 1 of 6 categories, 5 of which considered potentially preventableThirty-two percent of admissions were potentially preventable in chronic illness group. Chronic versus healthy preventable hospitalizations not reported. Among chronically ill: fewer admissions in the technology assisted group were potentially preventable: 19% vs 38% with chronic illness but no tech, P < .05). Potentially preventable family and environmental factors in 18% (noncompliance, inappropriate supervision, tobacco smoke); health system deficiencies involved in 21% (inadequate care coordination, failure to provide mental health or hospice). More ICU admissions in chronic illness compared with healthy, RR, 3.3 (95% CI: 2.5–4.2), and RR, 373 (95% CI: 330–422) for technology assisted subgroup.
2. Technology assistance
Feudtner (2009)35Retrospective cohort186 856 patients ages 2–18 y discharged from 1 of 38 children’s hospitals across US during 2004 (Pediatric Health Information System)Predictors of 365-d readmissionCCC365-d readmission16.7% of all discharges had readmit within 1 y. CCC readmit AORs ranged from 1.2, 95% CI: 1.1–1.3 (cardiovascular) to 1.6, 95% CI: 1.5–1.7 (neurologic) excluding malignancy and compared with no CCC. Discharge with home health AOR 1.2, 95% CI: 1.1–1.3). Also higher odds with older age, African American race/ethnicity, public payer, longer LOS, high number and more recent past admissions. Model has good positive predictive value, limited sensitivity.
Frei-Jones (2009)34Retrospective cohort100 patients with sickle cell disease younger than 21 y admitted to St. Louis Children’s Hospital over 12-mo periodPredictors of 30-d readmission in patients with sickle cell disease.Sickle cell30-d readmissionBivariate readmission associated with no heme follow-up, OR, 7.7, 95% CI: 2.4–24.4. In multivariate model, ≥3 admits in past year and discharge follow-up were significantly associated, P < .01 R2 = 0.41. Fifty percent were given heme follow-up appointments (53% nonreadmit, 43% readmit, P = .8), 42% went (54% nonreadmit, 13% readmit, P < .001).
Gay (2011)31Retrospective cohort2546 cases, 1435 unique patients, with 15-d readmission to Children’s Hospital at Vanderbilt (of 30 188 total admissions)Characterize patients readmitted within 15-d of dischargeCCCs plus asthma, diabetesUnplanned 15-d readmissionMost readmits were in patients with chronic illness (78% of total). Compared with those with acute illness, chronic illness patients were median 5 y older, with median 2 d higher index and readmit LOS (P < .001). Higher proportion of readmits in chronic illness were planned (25.5% vs 3.2%), a complication of a device (8% vs 1.2%), or related to index admission (95.3% vs 81.9%). Lower proportion of 0–7 d readmits were planned or for chronic illnesses than at 8–15 d (planned 18% vs 24.5%, P < .001, chronic illnesses 75.4% vs 82%, P < .001). The 14.4% of patients with ≥3 readmits accounted for 43.7% of all readmits. Oncology patients had highest unplanned readmits (20.7%), then neurologic conditions (9.5%).
Hain (2013)28Retrospective cohort200 randomly selected pairs of index admission/15-d readmission to Children’s Hospital at VanderbiltPreventability of 15-d readmissionsCRGs (3M)Investigator-defined 5-point Likert scale independently scored by 4 pediatriciansPotentially preventable readmit rate = 20%, 95% CI: 14.8–26.4. Central venous catheter infection or ventricular shunt malfunctions in 8.5%. Surgical admits more often “more likely preventable” readmission (38.9% vs 15.9%, P = .002). Preventability was similar for each of the CRG groups except malignancy (5% more likely preventable in malignancy versus 24.7% to 26.7% more likely preventable in other CRGs, P = .003). Most agreement when rating planned readmits. More than 2 of the 4 reviewers never agreed on rating “preventable in most cases.” Significant chronic illness in 74% of sample.
Kun (2012)30Retrospective cohort109 children ages 0–21 y started on home mechanical ventilation for chronic respiratory failure at Children’s Hospital Los AngelesIdentify risk factors for nonelective 12-mo readmission in patients with newly initiated home ventilationChronic respiratory failure with newly initiated home mechanical ventilationNonelective 12-mo readmissionNonelective 12-mo readmission rate = 40%, and 28% occurred in first month. Pneumonia (28%), tracheitis (17%), and tracheostomy decannulation/obstruction (11.5%) were most common readmit causes. Change in care management 7-d before discharge had bivariate association with readmit (18% vs 3%, P = .014). In multivariate analysis, no statistically significant associations observed between readmission and demographic or clinical characteristics.
Lu (2012)24Cross-sectional1 326 650 weighted admissions ages 3 mo to 17 y from Kids Inpatient Database (KID), containing discharge data from 38 states based on sample from 3739 US community hospitals and 45 children’s hospitals; national sampleHospital days, hospital charges, and demographic characteristics of nationally representative sample of potentially preventable hospitalizationsCCCsACSC as primary diagnosis; list adapted to remove adult conditions except Pelvic Inflammatory Disease and add failure to thrive, vaccine-preventable diseasesFewer ACSC admissions had CCC than non-ACSC (10.8% vs 16.1%, P < .001). Children with ACSC admission less likely to have comorbid CCC (AOR, 0.64; 95% CI: 0.60–0.68). Asthma and pneumonia = 48.4% and 46.7% of ACSC hospital charges and days, respectively. Children hospitalized with ACSC were more likely boys, nonwhite, publicly insured, or admitted through the ED. Lowest income quartile not associated.
Raphael (2011)39Prospective cohort1591 CSHCN under 18 y in Medical Expenditure Panel Survey (MEPS); national sampleParent-report quality of primary care (family-centered, timely, accessible) and subsequent hospitalizationCSHCN Screener (MEPS does not include “at risk” component)Different hospitalization rates across different CSHCN groups suggests some hospitalizations may be preventableHospitalization rate = 4.3%, High-quality family centeredness observed 68.3%, timeliness 51.5%, access in 80.4%. Private insurance: Low family-centered care associated with more hospitalizations (incident rate ratio (IRR), 3.87; 95% CI: 1.23–12.13); low accessibility associated with more hospitalizations (IRR, 3.45; 95% CI: 1.30–9.19). Timeliness not associated with hospitalizations. Public insurance: no relationships with quality.
Todd (2006)36Cross-sectionalDischarges from Colorado hospitals, ages 28 d to <18 y (2003, n = 26 931); Discharges in Kids Inpatient Database (KID), ages 28 d to <18 y, (2000, n = 1 932 883), national sampleACSC hospitalization relative rates stratified by insurance and patient characteristicsICD-9 codes for neuromuscular, brain and spinal cord malformations, mental retardation, central nervous system disease, cerebral palsy, muscular dystrophiesACSC modified list: asthma, diabetes, vaccine preventable disease, psychiatric disease, and appendectomy with perforated appendix or peritonitisColorado: public/no versus private insurance had higher hospitalization rates for chronic disease (RR, 1.76; P < .001). Higher RR also observed across subgroups including ACSC hospitalizations and All Patient Refined Diagnosis Related Group severity >2. Independent associations with hospitalizations of public/no insurance and nonwhite race/ethnicity, age, chronic disease and ACSC (all P < .002). Similarly, US sample, public/no versus private insurance had higher hospitalization RR for chronic disease (2.20, 95% CI: 2.18–2.21), asthma (2.37, 95% CI: 2.34–2.40), and diabetes (1.39, 95% CI: 1.36–1.42).
  • ADHD, attention-deficit/hyperactivity disorder; GI, gastrointestinal; LOS, length of stay; MEPS, Medical Expenditure Panel Survey; RR, relative risk; UTI, Urinary tract infection.

  • a ↑, Main predictor or intervention associated with increased preventable hospitalizations; ↓, Main predictor or intervention associated with reduced preventable hospitalizations; ↔, No significant (or inconsistent) association between main predictor or intervention and preventable hospitalizations.