Integration of the AHRQ Clinical Classification System and the Pareto Principle to Study Children with Multiple Chronic Conditions Jay G. Berry, MD MPH Division of General Pediatrics I Complex Care Service Boston Children’s Hospital I Harvard Medical School
Disclosures • Funding Source: Agency for Healthcare Research and Quality (R21 HS023092-01) • No conflicts of interest • My wife says that I ramble
Background Children with Multiple Chronic Conditions • Definition – 2 or more co-occurring chronic conditions • Prevalence and impact – 1% of all children – 30% of healthcare spending on children • Quality of care – Fragmented, uncoordinated, crisis-driven
Multiple Chronic Conditions Example: Children with Cerebral Palsy Cerebral Palsy : movement disorder caused by brain dysregulation Blindness Epilepsy Deafness Cardiac Apnea, autonomic Dysphagia, instability Chronic Aspiration Gastroesophageal reflux, constipation Asthma, Respiratory Urinary retention, Insufficiency recurrent infection Kyphoscoliosis Hip dysplasia, Osteopenia, Malnutrition, Fractures, Skin breakdown Spasticity
Population Health Initiatives Children with Multiple Chronic Conditions • Assess which conditions affect the health outcomes and resource use the most • Target conditions for clinical interventions – Prevent occurrence – Control severity – Optimize care management – Contain cost
Population Health Initiatives Children with Multiple Chronic Conditions • Assess which conditions affect the health outcomes and resource use the most – Vilfredo Pareto’s “80 - 20” Rule (ca. 1896) 80% of land in Italy owned by 20% of the population – Perhaps Pareto’s rule might apply to healthcare spending in co-occurring conditions Most of the spending is driven by a few conditions
Population Health Methods Children with Multiple Chronic Conditions • Administrative healthcare claims are often the most forthcoming source of data – State Medicaid program claims – Billing codes recorded for each encounter • Diagnosis classification systems are often applied to the claims data – Use ICD9 diagnosis codes – Identify and count an individual’s conditions
Population Health Methods Children with Multiple Chronic Conditions • AHRQ Clinical Classification System – Transparent, open source diagnosis system developed for adult patients – Classifies ~14,000 ICD9 codes as chronic or not chronic – Maps each code to a mutually-exclusive diagnosis category
Population Health Methods Challenges Children with Multiple Chronic Conditions • Epidemiology of chronic conditions in children – Heterogeneous array of conditions – Low prevalence of many conditions • Existing diagnosis classification schemes for use with administrative data – Incomplete listing of childhood conditions – Non-specific categorization of the conditions
Objectives 1. To adapt for children the adult-based AHRQ Clinical Classification System to identify and count multiple chronic conditions in children 2. To test the adapted system on children with cerebral palsy, identifying which co- occurring conditions influence healthcare spending the most using the Pareto Principle
Study Design, Setting, and Population • Retrospective cohort analysis • Truven Medicaid Marketscan Database – 10 state Medicaid programs in 2013 – Health administrative claims and spending • Across the care continuum • ICD9 diagnosis codes • Children with cerebral palsy – ICD9 codes 343.XX
Main Outcome Measure • Total healthcare spending for health services across the care continuum – Payment made by Medicaid – Health services: inpatient, outpatient, community, home, pharmacy, durable medical equipment, etc. – Out-of-pocket spending not included
Main Independent Variable • Comorbid Health Conditions in Children with Cerebral Palsy (CP) – AHRQ Clinical Classification System – Chronic condition definition “Lasts 12 months or longer and (a) places limitations on self-care, independent living, and social interactions; or (b) results in the need for ongoing intervention with medical products, services, and special equipment” – Identified with ICD9 codes across all healthcare encounters
Pediatric Adaptation AHRQ Clinical Classification System (CCS) • Two complex care pediatricians reviewed CCS for adaptation – Diagnosis categories – Individual diagnoses within each category • Findings presented to a larger study team – Revision and affirmation of adaptations – Adjudication of few instances of differences in judgment
Statistical Analysis Three staged approach to assess the impact of comorbid conditions on healthcare spending for children with cerebral palsy 1. Pareto principle ranking 2. Multivariable linear regression 3. Classification and regression tree analysis
Pareto Principle Ranking Impact of Comorbidities on Healthcare Spending • Goal – To create a ranked list of the comorbid conditions in the population of children with cerebral palsy that account for the most healthcare spending • Procedure – Distinguish the top-ranked comorbidity – Remove all children with it from the cohort – Repeat with the 2 nd -ranked comorbidity
Multivariable Linear Regression Impact of Comorbidities on Healthcare Spending • Goal – To quantify the contribution of each comorbid condition on healthcare spending for individuals with cerebral palsy • Model derivation – Outcome = total healthcare spending – Primary cofactors = comorbid conditions – Confounders = demographic characteristics – Backward elimination approach
Classification & Regression Tree Impact of Comorbidities on Healthcare Spending • Goal – To distinguish which combinations of comorbid conditions account for the greatest amount of healthcare spending for children with cerebral palsy • Model derivation – Assesses every possible combination of conditions – Creates a tree with a branch for the presence of each condition string with the greatest spending increase
Results Pediatric Adaptation of AHRQ CCS 159 new chronic condition categories • Identified from non-specific categories in the original system – Dysphagia, gastroesophageal reflux, scoliosis • Created de novo – Enterostomy, tracheostomy
Results Children with Cerebral Palsy (CP) in Medicaid Attribute Finding Number of individuals 16,695 Median (IQR) No. of Chronic Conditions 6 (4-10) Total Annual Healthcare Spending $610 million Median (IQR) Annual Spending $10,700 Per Individual ($4,070-35,570)
Population Spending for CP Comorbid Conditions with Greatest Impact Epilepsy Intellectual disability Enterostomy Asthma Prevalence Neurogenic bladder Spending 0 20 40 60 Percent of Total
Population Spending for CP Comorbid Conditions with Greatest Impact Epilepsy Intellectual disability Enterostomy 78% of Total Spending Asthma Prevalence Neurogenic bladder Spending 0 20 40 60 Percent of Total
Population Spending for CP Comorbid Conditions with Greatest Impact 38% Prevalence Epilepsy 60% of Total Spending Intellectual disability Enterostomy Asthma Prevalence Neurogenic bladder Spending 0 20 40 60 Percent of Total
Population Annual Spending for CP With vs. Without Epilepsy Epilepsy Healthcare Setting Present Absent Specialty Care $101 $70 Inpatient Care $92 $46 Pharmacy $33 $13 U.S. $ millions
Individual Annual Spending for CP Comorbid Conditions with Greatest Impact Tracheostomy Enterostomy Intellectual disability Neurogenic bladder Esophageal reflux Epilepsy Scoliosis Dysphagia Asthma 0 20,000 40,000 60,000 Mean (SE) Additional Spending
Individual Annual Spending for CP Comorbid Conditions with Greatest Impact Tracheostomy Enterostomy Intellectual disability Neurogenic bladder Esophageal reflux Epilepsy Scoliosis Dysphagia Base Spending = $1950 Asthma 0 20,000 40,000 60,000 Mean (SE) Additional Spending
Individual Annual Spending for CP Comorbid Conditions with Greatest Impact + $56,670 Tracheostomy Tracheostomy Enterostomy Enterostomy + $25,710 Intellectual disability Neurogenic bladder Esophageal reflux Epilepsy Scoliosis Dysphagia Base Spending = $1950 Asthma 0 20,000 40,000 60,000 Mean (SE) Additional Spending
Individual Annual Spending for CP Comorbidity Combination with Greatest Impact Combination Spending Enterostomy, $225,590* Tracheostomy, & Chronic respiratory failure *Derived from Classification and Regression Tree Model
Main Findings • Integration of the adapted AHRQ CCS and the Pareto Principle – Distinguished with clinical fidelity the co-occurring conditions responsible for the most healthcare spending in children with cerebral palsy – Performed well on the population and individual patient levels
Limitations • Variation in coding practices may affect identification of chronic conditions • Restrictiveness of study setting to children in Medicaid from 10 states may limit generalizability • Subjectivity of the AHRQ CCS adaptions cannot be understated
Recommend
More recommend
