Low-Hanging Fruit: Analytic Best Practices for Physician-Led ACOs
MY BACKGROUND Practicing General Internal Medicine Physician Hospitalist at Newton-Wellesley Hospital Researcher at Brigham and Women s Hospital and Harvard Medical School Leader of the Physician-Led ACO Workgroup for the Brookings Institution s ACO Learning Network
ACO LEARNING NETWORK A collaboration between Dartmouth and Brookings Institution Partnering with ACOs to help them successfully implement accountable care Components include: Webinars Meetings, workshops, and conference calls Issue-focused workgroups Online resources Policy suggestions ACO best practice recommendations www.acolearningnetwork.org
PHYSICIAN-LED ACOs Outpatient physician groups Large percentage of primary care physicians Typically do not include a hospital Lack the financial resources of larger, hospital-led ACOs Nimble organizations that know their patients well Can leverage the primary care relationship which is at the heart of medicine
NUMBER OF PHYSICIAN-LED ACOS GROWING AT A FASTER RATE THAN HOSPITAL-LED ACOS Health Affairs Blog, 2014
PRELIMINARY RESULTS OF MEDICARE SHARED SAVINGS PROGRAM ARE PROMISING FOR PHYSICIAN-LED ACOS Governance of MSSP ACOs Receiving Shared Savings in Year One Hospital-Led 47% Physician-Led 53% Source: CMS.gov
ACO LEARNING NETWORK: PHYSICIAN-LED ACO WORKGROUP MEMBERS Over 90% are participating in Medicare Shared Savings Program Most also participating in commercial insurance ACO contracts PCP Participants: 60 200 Attributed Lives: 5,000 13,000 Care Manager FTEs: 0 15 In-house Data Analyst FTEs: 0-8 (Average of 2) Claims Data Analysis: 50% outsource to a vendor
CHALLENGES OF BIG DATA PREDICTIVE ANALYTICS FOR PHYSICIAN-LED ACOs ACO may include multiple EHRs Outpatient providers may lack the resources to invest in complex data-analytics platform ACO may not have access to full spectrum of clinical data Inpatient records Post-acute care data Specialty and procedural data IT vendors often over-promise and under-deliver
HOW ARE MANY PHYSICIAN-LED ACOs RESPONDING TO THE CHALLENGES OF BIG DATA HEALTH ANALYTICS? A. Invest more money and resources in Big Data solutions B. Bide their time until predictive algorithms improve C. Get started today with LITTLE DATA analytics
HOW ARE MANY PHYSICIAN-LED ACOs RESPONDING TO THE CHALLENGES OF BIG DATA HEALTH ANALYTICS? A. Invest more money and resources in Big Data solutions B. Bide their time until predictive algorithms improve C. Get started today with LITTLE DATA analytics
LITTLE DATA: TAKING A SIMPLE APPROACH TO POPULATION HEALTH ANALYTICS Analytics does not need to be complex or costly There are many simple opportunities Take advantage of the resources and staff you already have Understand your analytic goals Don t wait for all the data. Get started NOW!
LESSON #1 TAKE ADVANTAGE OF DATA ALREADY AVAILABLE WITHIN YOUR ORGANIZATION
SIMPLE ANALYTIC METHODS TO IDENTIFY HIGH-RISK PATIENTS Utilization Patterns Patients who have had 2 or more hospitalizations in the past 6 months Diagnoses Patients with 3 or more chronic conditions (CHF, COPD, Diabetes) Medications Patients on multiple high risk medications (anticoagulants, immunosuppressants, insulin)
LESSON #2 START BY DEFINING YOUR INTERVENTIONS
DEFINING YOUR INTERVENTIONS What do you plan to DO with the results of your data analysis? Complex care management Home-based care Telephonic patient outreach Disease-specific interventions: CHF clinic Behavioral health program Diabetes education
EXAMPLE: COMPLEX CARE MANAGEMENT $250,000 available for care manager salaries Each care manager will follow 100 high-risk patients Care manager salary + benefits $80,000 Need to identify 300 high-risk patients for enrollment in complex care management
LESSON #3 SEGMENT YOUR HIGH-RISK PATIENTS INTO ACTIONABLE SUBGROUPS BASED ON COMMON CLINICAL FEATURES
IS THIS THE MOST USEFUL WAY TO UNDERSTAND THE HEALTH NEEDS OF A POPULATION? Costly, high utilizers Patients with multiple chronic conditions Patients at Risk Everyone Else
ACTIONABLE POPULATION HEALTH BUCKETS CHF with exacerbation in the past six months Diabetics with HbA1c > 9 Stage IV cancer More than 2 ambulatory sensitive ER visits in the past six months
LESSON #4 COMBINE RAW ANALYTICS WITH PROVIDER INTUITION
PROVIDER INTUITION: ASK YOUR CLINICIANS TO IDENTIFY HIGH-RISK PATIENTS Primary care physicians know details about patients that may not be captured in algorithms Living situation Social support network Ability of patient to achieve health goals Recent study found that providers identify many patients as complex who were not identified by commonly used risk-stratification algorithms * * Grant et al. Annals of Internal Medicine, 2011
LESSON #5 USE DATA COLLECTED FROM PATIENTS THEMSELVES
PATIENT ENGAGEMENT AND ACTIVATION ARE CORRELATED WITH HEALTH OUTCOMES AND COST (Health Affairs, 2013)
PATIENT SELF-REPORTED HEALTH STATUS AND PHYSICAL FUNCTIONING CAN PREDICT UTILIZATION J Am Heart Assoc 2014
EXAMPLE: INSTITUTE FOR FAMILY HEALTH FQHC that participates in Medicare Shared Savings Program Collects data from patients on social determinants of health: Homelessness Access to healthy food Substance use Family and social support Goal is to gather additional data outside of the traditional medical record that can allow the ACO to best identify patients at risk
CONCLUSIONS & RECOMMENDATIONS
CONCLUSIONS & RECOMMENDATIONS 1) Define your care management capabilities and disease specific goals
CONCLUSIONS & RECOMMENDATIONS 1) Define your care management capabilities and disease specific goals 2) Create a list of high-risk patients based on simple metrics such as utilization patterns and diagnoses
CONCLUSIONS & RECOMMENDATIONS 1) Define your care management capabilities and disease specific goals 2) Create a list of high-risk patients based on simple metrics such as utilization patterns and diagnoses 3) Let your providers modify the list using their own clinical judgment
CONCLUSIONS & RECOMMENDATIONS 1) Define your care management capabilities and disease specific goals 2) Create a list of high-risk patients based on simple metrics such as utilization patterns and diagnoses 3) Let your providers modify the list using their own clinical judgment 4) Use patient engagement and patient-reported metrics to enhance your population health efforts
CONCLUSIONS & RECOMMENDATIONS 1) Define your care management capabilities and disease specific goals 2) Create a list of high-risk patients based on simple metrics such as utilization patterns and diagnoses 3) Let your providers modify the list using their own clinical judgment 4) Use patient engagement and patient-reported metrics to enhance your population health efforts 5) Segment high-risk patients into actionable buckets
CONCLUSIONS & RECOMMENDATIONS 1) Define your care management capabilities and disease specific goals 2) Create a list of high-risk patients based on simple metrics such as utilization patterns and diagnoses 3) Let your providers modify the list using their own clinical judgment 4) Use patient engagement and patient-reported metrics to enhance your population health efforts 5) Segment high-risk patients into actionable buckets 6) Use care management and other tools to improve health outcomes for your high-risk patients
CONCLUSIONS & RECOMMENDATIONS 1) Define your care management capabilities and disease specific goals 2) Create a list of high-risk patients based on simple metrics such as utilization patterns and diagnoses 3) Let your providers modify the list using their own clinical judgment 4) Use patient engagement and patient-reported metrics to enhance your population health efforts 5) Segment high-risk patients into actionable buckets 6) Use care management and other tools to improve health outcomes for your high-risk patients
James Colbert, M.D. Harvard Medical School Brookings Institution ACO Learning Network @jcolbertmd