420 likes | 573 Views
Engineering in Public Health for “Efficient, Effective, and Equitable” Outcomes. Nicoleta Serban, PhD and Julie Swann, PhD Industrial and Systems Engineering Georgia Institute of Technology November 2013. Definitions.
E N D
Engineering in Public Health for “Efficient, Effective, and Equitable” Outcomes Nicoleta Serban, PhD and Julie Swann, PhD Industrial and Systems Engineering Georgia Institute of Technology November 2013
Definitions • Industrial Engineering, Operations Research, and Management Sciences draw upon methods from the mathematical, physical, and social to design, optimize, predict, and/or evaluatesystems • Public Health can be defined as the “science and art of preventing disease, prolonging life, and promoting health” within and across populations Material adapted from CDC, Wikipedia, Dictionary.com, Institute of Industrial Engineer websites, or others
What Problems in Health? • Decisions where there are • Limited resources, or • Trade-offs between two elements (e.g., cost and quality-adjusted life years) or • Uncertainties in the system or • Projections over space or time • There may be one or more goals or objectives, and multiple constraints in the system • Systems could take different forms • Individual (e.g., optimizing radiation treatment within body) • Provider (e.g., serving patients in hospitals) • Network (e.g., distributing emergency supplies within a state) • Population (e.g., policies around infectious diseases)
Performance Measures • Efficiency • Accomplishment of an outcome with the minimum time, effort, or resources needed • “Are we doing things the right way?” • Example: are we delivering medicines without unnecessary resources? • Effectiveness • Producing the best or desired outcome • “Are we doing the right things?” • Example: are we delivering the right medicines for the right population? • Equity • Achieving outcomes that are fair or equitable across a system • “Are we impacting all people or places?” • Example: is our distribution of medicines equitable? • Achieving all measures may be desirable, but may not always be possible Cost Outcomes Fairness
Data We Are Using Data Types Examples • Disease Registry • Disease Progression • Electronic Health Records • Facility Info • Medical Claims Data • National Survey or Examination Data • State Databases • General • Cystic Fibrosis • “Natural History” models • Queries (CHOA, VA, etc.) on specific projects • VA satellite clinics • Medicaid (children and pregnant women, GA + 13 other states, 2005-2009) • NHANES, HCUP KIDS • GA’s Oasis, HCUP SEDD and SID • Census, National Provider Index
Serban & Swann Group: IE-Health Research Estimating • Disease prevalence in small areas • Childhood obesity • Intervention Cost-effectiveness Analysis • Telemedicine • Healthcare Access • Measurement • Policy Interventions • Linking to Health Outcomes • Network Interventions • Decision-making support • Asthma Baseline • Care Pathway • Work covers multiple collaborations between GT ISyE, GT IPaT, Children’s, CDC, VHA, and other health entities Projecting Value Measuring & Improving Informing
Disease Prevalence Given limited data availability about a disease, how can we accurately estimate prevalence of a disease within small geographic areas, so interventions can be targeted effectively?
Pediatric Obesity • Pediatric obesity has tripled in the last three decades • The cost of obesity in the United States totaled about $147 billion in 2008 (Finkelstein, et al 2009) • Interventions could impact children immediately, and health systems for the long-term • Obesity differs geographically and/or by population characteristics • Targeting the interventions (e.g., by area) may be more cost-effective • But, survey and surveillance approaches are costly • Statistical modeling and simulation can be used to estimate prevalence in small geographic areas
Data on Pediatric Obesity • National Health and Nutrition Examination Survey (NHANES) contains national survey and examination data for all ages1 • National Survey of Children’s Health (NSCH) provides state-level estimates for ages 0-17 from self-reported surveys2 • Arkansas school systems started collected measurements, reported for schools or counties Davila Payan, C., M. DeGuzman, K. Johnson, N. Serban, and J. Swann (2013), “Estimating Prevalence of Overweight and Obese Children in Small Geographical Areas using Publicly Available Data”, under review, Obesity Zhang X, Onufrak S, Holt JB, Croft JB. A Multilevel Approach to Estimating Small Area Childhood Obesity Prevalence at the Census Block-Group Level. Prev Chronic Dis 2013;10:120252.
Approach: Modeling and Simulation • Step 1: Model Individual Obesity Probability given a set of characteristics (e.g. demographics) • Data: NHANES • Model: Logistic Regression • Step 2: Simulate virtual population within a small geographic area • Data: 2010 Bureau Census • Model: Multinomial Distribution • Step 3: Update individual probabilities using the model in Step 1 for the population generated at Step 2 • Numerical Summaries: Mean and Variance • Estimates at the census tract or other geographic levels
Estimates of Overweight or Obese Children Prevalence (census tracts) Number of Children Priority Areas: Areas that cover ~ 80% of the overweight children in GA Results used to target interventions by Children’s Healthcare of Atlanta
Intervention Cost-Effectiveness Analysis What are the costs and the benefits for the implementation of an intervention? What is the implementation threshold for ineffectiveness of an intervention?
Telemedicine • Originated in the Netherlands in the early 1900s • More than 100 definitions of telemedicine (WHO, 2010) • Delivery of health-related services (preventative, promotive, and curative) and information via telecommunications technologies (Wikipedia) • Time Magazine has called telemedicine “healing by wire” • Countries of implementation: • Almost everywhere on the globe! • Provider-driven implementations in the US (e.g., VHA) • Example of an implementation: • Tele-ophthalmology at VHA • Cost-effectiveness: Diabetic Retinopathy Screening Bashshur, R. L., & Shannon, G. W. (2009). History of Telemedicine: Evolution, Context, and Transformation. New Rochelle, NY: Mary Ann Liebert, Inc. World Health Organization (2010). Telemedicine: Opportunities and developments in Member States.
Approach: Modeling and Simulation • Step 1: Model Individual Disease Progression • Data: Sample of veterans with diabetes (Atlanta) • Model: Markov Decision Model • Step 2: Simulate individual disease profiles • Data: VHA & General parameter input • Model: Estimated Disease Progression Model • Step 3: Compare traditional to telemedicine • Simulate screening with and without telemedicine • Utility measures: Quality Adjusted Life Years (QALY) vs. costs of the program E. Kirkizlar, J. Swann, N. Serban, J.A. Sisson, C. Barnes, M. Williams (2013). Evaluation of Telemedicine for Screening of Diabetic Retinopathy in the Veterans Health Administration. Ophthalmology, in press.
Estimate of Cost vs. QALY Ratio • Cost-effectiveness for pool sizes of 3500 or larger (~9000 VHA patients in Atlanta area) • Cost-effectiveness for patients between the ages of 50-80
Healthcare Access What approaches should be used for quantifying access over a network? What interventions could result in the most improvements? How does access impact outcomes?
Access to Pediatric Healthcare • There is a large and increasing discrepancy between the rural and urban supply of pediatricians • In 1996, 95.5% of counties with populations below 10,000 had no pediatrician and 84.4% of counties with populations below 25,000 had no pediatrician (Randolph, Pathman 1996) • Uninsured children who received care from a pediatrician in a given calendar year were 93% less likely to visit the emergency room. (Johnson & Rimsza, 2004) • The U.S. Department of Health and Human Services identifies increasing accessibility of healthcare services as a key step in mitigating the widening disparities of health outcomes of children (Healthy People 2010) • Inequities in access to healthcare are also associated with higher costs and inconsistency in health treatments (Williams, R.A. 2007)
Pediatricians Per Child in Georgia • Specialists are even sparser • Many counties have no pediatricians What role does spatial access play in health outcomes?
Health IE and Access • Measurement Concerned with evaluating geographical access of different populations by taking into account supply and demand trade-offs and system constraints. • Inference: Equity A study of systematic disparities in access to services between population groups defined by geographic location and demographic characteristics. • Designing Interventions Optimization allows one to target interventions or estimate the impact on a complex system • Outcomes Ultimately we want to design interventions likely to positively impact health outcomes or inequities
Measurement: Our Approach We use optimization to match supply and demand and quantify access for multiple populations across the network System constraints Mobility – transportation, cost, infrastructure Capacity – access to supply, system dynamics User choice within network and willingness to travel System can be measured on distance traveled, congestion or scarcity, coverage, or others Other methods: Simple ratios (providers vs. population) Minimum distance Gravity models (modified ratios of providers vs. population)
Optimization Model • Decision Variable: • Number of children in each census tract assigned to each physician (separately for Medicaid and other populations) • Objective function: • Minimize travel distance for covered patients • Constraints: • 1) Coverage: • Achieve a minimum coverage level at the state level • 2-3) Physician Capacity: • For each physician, limit assignees by the maximum patient caseload & the estimated number of Medicaid patients they accept • 4-5) Patient Mobility: • For each census tract, limit the assignees to physicians more than 10 miles away by the percentage of the population with cars • 6) Dispersion of Congestion: • For each county, ensure that average congestion for physicians is below a threshold level • 7-9) Logic of Assignments: • Assignments less than number of children in population • No negative assignments
Estimates of Access* Distance Congestion Coverage • Coverage is either high close to 100% or very low close to 10% • Urban areas: shorter distances and higher coverage than rural • Wide variety in distances traveled • Congestion high across state except for some cities • * Nobles, M., N. Serban, and J. Swann (2012), “Measurement and Inference on Pediatric Healthcare Accessibility. Under second review at Annals of Applied Statistics ,Submitted January 2013.
PolicyInterventions: Medicaid Change in Distance Change in Congestion Locations where Medicaid have lower congestion • Most locations that show an improvement in congestion for Medicaid population are in Atlanta • Increasing participation of MDs taking Medicaid would improve travel distance and congestion for Medicaid patients without compromising access for overall population
PolicyInterventions: Caseload Change in Distance Change in Congestion • Increasing the caseload of physicians shows no significant decrease with a significant increase in congestion. • Decreasing the caseload of physicians does shows a significant increase in the travel cost but decrease in the congestion for all population
Linking Access to Outcomes • Asthma: Linking Access to Outcomes1 • Distances to asthma care vary greatly over a network • Garcia, E, N. Serban, and J. Swann (2013), “Linking Access for Asthma Care to Emergency Department Visits and Hospitalizations”, Working paper at Georgia Institute of Technology.
Distance to Asthma Care is related to Outcomes Hospitalizations per Child With Asthma Distance to an Asthma Specialist ED Visits per Child With Asthma Collected from OASIS system in GA for children aged 5- 17. Estimated by Serban & Swann research team
Linking to Outcomes: Approach Severe outcomes for asthma: • ED visits & hospitalizations • Data sources: OASIS (GA), HCUP (other states) • Estimate geographic access to asthma care • Centralized assignment optimization model • Both specialists and primary care • Modeling and Prediction • Logistic regression with replications • Controlling for other factors • Predict reduction in severe outcomes when access is improved
Predicted Reduction in ED Visits • Impact of intervention varies by county because of the interactions between access and other factors • Maximum Specialist Distance: 5 vs. 15 miles • More counties improve with the 5 mile intervention (as expected) • Counties with improvement when max specialist distance is 15 mi do not have significant improvement over distance reduced to 5 mi Primary Dist : 15 mi Specialist Dist : 5 mi Primary Dist : 15 mi Specialist Dist : 15 mi Specialist Dist : 15 mi Specialist Dist : 5 mi
Policy vs. Network Interventions Policy Intervention: Setting: current network of care Change: policies on the level of acceptability (e.g. Medicaid) and/or accommodation (e.g. after hour care) Network Intervention: Network structure alternations Approach: Intervention scenarios: mobile vs. stationary locations, physician vs. mid-level provider Linking to Outcome: emphasis on scenarios that result in reduction of severe health outcomes Linking to Disparities: emphasis on scenarios that result in reduction of healthcare disparities Optimization Model: select cost-effective scenarios
Decision-Making Support Given individual-level medical information (clinical or/and administrative), how can we translate it into meaningful knowledge not only about the patient but also about the system of care?
From Information to Knowledge to Decisions • Knowledge • Cost-Benefit Analysis • Disparities in Healthcare • Cost distribution • Data • Measurements: Outcome & cost • Summaries: Patient & Process • Interventions: Policy & Network • Information • Clinical • Administrative • Technology • R&D • Decisions • Which intervention to implement? • How to reduce cost? • What are expected outcomes? • What population to target?
Asthma Baseline: From Information to Knowledge Information: Medicaid MAX files 14 states, 5+years Patients with Asthma related ICD-9 codes Data: Measurements Cost & Outcomes Prevalence of asthma among Medicaid children Severity of condition Knowledge: Statistical Modeling Significant associations between cost and outcomes across the care network
Care Pathway: From Information to Knowledge • Information: Medicaid MAX files • 14 states, 5+years • Patients with Asthma related ICD-9 codes • Data: Measurements • Patient-level care trajectory • Severity of condition • Patient summary data • Knowledge: Statistical Modeling • Longitudinal care pathway profiles across children with Medicaid • Relationship of each profile with cost
CMS Data and Background Medicaid claims data will be used as a test bed for the decision-making support tools targeting knowledge representing the care of children with Medicaid. Medicaid Data Questions: Dr. NicoletaSerban (nserban@isye.gatech.edu) or Dr. Julie Swann (jswann@isye.gatech.edu)
CMS Medicaid Data • MAX Claims Data • Personal Summary: patients, demographics, birthdate, etc. • Inpatient: claims, diagnoses, procedures, LOS, payment • Other Therapy: claims for physician, lab, clinic, outpatient • Long Term Care:facility type, date of service, etc. • Prescription Drug:paid drug claims • (National Provider ID & Characteristics File for 2009 forward) • Person-level, Research-Identifiable-Files with locations and a provider-ID • Years 2005 – 2009 for 14 states (+2010-2011 for select states) • SE: Georgia, Alabama, Arkansas, Louisiana, Mississippi, N. Carolina, S. Carolina, Tennessee, Texas • Other: California, Minnesota, New York, Pennsylvania • Study population: children and pregnant women • Georgia Data for 2010 and 2011 should be available for purchase soon (Oct 2013 and Spring 2014) GT Project Champion: Beth Mynatt (IPaT, GT)GT Lead on Information Technology: Matt Sanders (GTRI) GT Research Leads: NicoletaSerban and Julie Swann
CMS Medicaid: Approved Topics 1) MEASURING AND EXPLAINING INEQUITIES:To assess the impact of healthcare system characteristics vs. inequities in healthcare, including geographical, use, quality, expenditure and outcomes among Medicaid children enrollees, especially in states with historic inequities like in the southeast. • To identify geographic areas with widespread and increasing Medicaid healthcare use (status quo and over time) and determine the underlying associative factors (e.g. access to healthcare facilities, race and ethnicity); • To investigate geographic variations in healthcare quality indicators (adherence to medications, emergency room visits, and other utilization measures) for high-impact diseases in children such as respiratory deficiencies, obesity, diabetes and other disabilities; • To identify geographic subdivisions which have achieved good health outcomes and low disparities despite adverse social determinants, or which have achieved poor health outcomes and high disparities worse than the social inequalities.
CMS Medicaid: Approved Topics 2) OPTIMIZING INTERVENTIONS AND DELIVERY SYSTEMSTo analyze flows and policies across the system, e.g., the match between supply and demand, and financially, both geographically and across time, along with the corresponding costs or outcomes, to analyze improved methods of delivery including medical homes. • To examine areas in the children Medicaid expenditure with the greatest costs or utilization, and assess potential interventions for reducing the healthcare costs, especially where interventions may be targeted by patient characteristics such as risk or where chronic issues like pediatric obesity can be addressed; • To evaluate the potential costs and benefits to creating a medical home or using telemedicine in the Medicaid system, where the creation may be focused on a subgroup of the Medicaid population or within specific geographical areas with great need; • To forecast the available “supply” of general or specialist providers or network services across geographical regions (e.g., counties or census tracts) as a function of socio-economic and other elements, link this factor with the costs or outcomes in the system as measured by the claims data, and examine potential interventions.; • To evaluate the impact of various public policies, such as changes in cost-sharing, on the demand for Medicaid coverage.
CMS Medicaid Data: Access to Data • Data stored in secure location at Georgia Tech, with access to the detailed data by a limited set of GT employees approved by CMS and IRB • Massive data files, with technology infrastructure for efficient access • Sharing of aggregated data is allowed with collaborators, if consistent with research purposes and research protocol • Cells should have at least 11 entries • Data undergoes review process at GT before release from data workstation • Significant liability involved if breach occurs • Data management plan revised for easier maintenance
Comments about MAX Data • Limitations • Research must fit within scope proposed to CMS • Analysis of raw data must be conducted at GT • Process for analyzing data is onerous, time-consuming, and “expensive” • Data only covers Medicaid claims • Does not include school absenteeism data • Data never includes the most recent (~2) years of data • Positives • We can benchmark GA against 13 other states • Patients and/or providers can be followed longitudinally • Includes provider visits and prescription claims • No one has to give permission for us to ask specific questions or publish related answers
Examples of Potential Research Allowed • Asthma Specific • Access • Where is access to asthma care most critical for Medicaid patients? • What factors are related to lower access, and what policies might improve access? • Does access to specialty care impact outcomes (e.g., reduce ER visits)? • Current Utilization and Costs • What are the costs, utilizations, and “outcomes” (such as hospitalizations) for Medicaid patients across the network? How does GA compare to other states? • How do estimates of prevalence compare to utilization of services for asthma care? • Where is variation in care and patient usage the greatest for medication adherence? • Projections of Changes • Which interventions are likely to have the largest impact and where? • Telemedicine? Reimbursement changes? Medical Education? Etc.
Questions • Nicoleta Serban nserban@isye.gatech.eduor 404-385-7255 • Julie Swannjswann@isye.gatech.edu or 404-385-3054 Undergraduate: Richard Zheng, Alex Terry, PravaraHarati MS students: Kevin Johnson PhD students: Ben Johnson, Erin Garcia, Zihao Li, Mallory Nobles Visiting Assistant Professor: Monica Gentilli