Patient-Centered Analytics Presented to ASHNHA Alaska Partnership for Patients Advisory Group February 4, 2015 Gloria Kupferman

DataGen Medicare advocacy analytics for 46 State Hospital Associations, 6 multi-state systems Data partner for 30+ BPCI awardees including AAMC convened group Readmissions diagnostic reporting for 7 states AHRQ reporting NYS Partnership for Patients

Today’s Agenda Reasons to use patient-centered analytics Data types and sources Metrics Tools Case studies Questions

Reasons for Patient-Centered Analytics

The Current Health Care Delivery System Practitioner Office Visit Labs, Scans, Screens Readmission Labs, Scans, Screens ED Visit Practitioner Office Visit Labs, Scans, Screens Labs, Scans, Screen Practitioner Office Visit Non-traditional care setting Practitioner Office Visit Initial Inpatient Stay Other Services (Hospital Outpatient, Medical Equipment, etc.) Practitioner Office Visit Rx Rx Rx Rx Inpatient Post-Acute Stay (Rehab, Psych, LTC, SNF, HH)

The Population

Population Health The health status and outcomes of individuals within a group Patients you see People who are not yet your patients The distribution of the status and outcomes within the group These groups can be defined by geographic boundaries, employer, ethnicity, health factors, or any other defined group.

Population Health Management Managing, addressing, and improving the health status and outcomes for individuals within a group Emphasis on the “triple aim” Access to care and the patient experience Quality of care Efficiency of care

Why Do We Need to Look at Patient-Centered Data? To assess the current “state of play” Identify, measure and address opportunities for change Track progress Examples: Hot-spotting Gap spotting Identify best practices, top performers Identify opportunities

Delivery / Payment Systems Support for successful population health management There needs to be a sustainable financial model Accountable Care Organizations Medical Homes Episodes of Care / Bundled Payments Capitation

Data and Sources

Data and Sources Administrative Data Payer Data Internal Data Mandatory State-collected data Agency for Healthcare Quality Research H-CUP Payer Data Medicaid/Medicare Direct from Payer APCD Claims Aggregators Consulting Firms Internal Data EMR System Billing System RHIO/HIE Self-insured Program Community Health Data Census / Inter-census Survey Public Health Data CDC, etc.

Data Coverage Initial Inpatient Stay ED Visit Readmission Practitioner Office Visit Readmission Labs, Scans, Screens ED Visit Initial Inpatient Stay Practitioner Office Visit Labs, Scans, Screen Non-traditional care setting Labs, Scans, Screens Other Services (Hospital Outpatient, Medical Equipment, etc.) Practitioner Office Visit Practitioner Office Visit Labs, Scans, Screens Inpatient Post-Acute Stay (Rehab, Psych, LTC, SNF, HH) Rx Rx Practitioner Office Visit Rx Rx

Administrative Data Initial Inpatient Stay ED Visit Readmission Practitioner Office Visit Labs, Scans, Screens Labs, Scans, Screen Practitioner Office Visit Non-traditional care setting Other Services (Hospital Outpatient, Medical Equipment, etc.) Labs, Scans, Screens Practitioner Office Visit Practitioner Office Visit Labs, Scans, Screens Includes all care provided in only those settings Inpatient Post-Acute Stay (Rehab, Psych, LTC, SNF, HH) Rx Rx Practitioner Office Visit Rx Rx

Payer Data Initial Inpatient Stay ED Visit Readmission Practitioner Office Visit Labs, Scans, Screens ED Visit Initial Inpatient Stay Readmission Practitioner Office Visit Labs, Scans, Screen Non-traditional care setting Labs, Scans, Screens Other Services (Hospital Outpatient, Medical Equipment, etc.) Practitioner Office Visit Practitioner Office Visit Labs, Scans, Screens Dependent upon coverage Won’t reflect non-covered services Inpatient Post-Acute Stay (Rehab, Psych, LTC, SNF, HH) Rx Rx Practitioner Office Visit Rx Rx

Internal Data Initial Inpatient Stay ED Visit Readmission Practitioner Office Visit Labs, Scans, Screens ED Visit Initial Inpatient Stay Readmission Practitioner Office Visit Labs, Scans, Screen Non-traditional care setting Labs, Scans, Screens Other Services (Hospital Outpatient, Medical Equipment, etc.) Practitioner Office Visit Practitioner Office Visit Labs, Scans, Screens Depends on who it is internal to Inpatient Post-Acute Stay (Rehab, Psych, LTC, SNF, HH) Rx Rx Practitioner Office Visit Rx Rx

Community Health Data Initial Inpatient Stay ED Visit Readmission Practitioner Office Visit Labs, Scans, Screens Labs, Scans, Screen Practitioner Office Visit Non-traditional care setting Other Services (Hospital Outpatient, Medical Equipment, etc.) Labs, Scans, Screens Practitioner Office Visit Labs, Scans, Screens Practitioner Office Visit Prevalence data Won’t give insights into care management Inpatient Post-Acute Stay (Rehab, Psych, LTC, SNF, HH) Rx Rx Practitioner Office Visit Rx Rx

Data Pros and Cons There is increasing interest in transparency and data sharing, but availability is still spotty There is no one-stop shopping Most under-represented in the data sets are uninsured and people who have not needed or sought out care Bureaucratic and HIPAA constraints

Patient-Centered Data Metrics

Some Patient-Centered Data Metrics Chronic conditions Stratification of population into disease cohorts Risk scores Stratification of population into risk cohorts Episodes of care Stratification of population into care cohorts PMPM Total healthcare spend per member (i.e. person) per month Quality metrics Avoidable events per person

Primary Care Shortages Chronic Conditions Hot Spotting Market Planning Primary Care Shortages

Chronic Conditions - Sources Survey Example - Behavioral Risk Factor Surveillance Survey (BRFSS) CDC annual phone survey “Have you ever been told you have diabetes?” Claims data Must have physician encounters Stratified sample, a

Chronic Conditions - Sources Survey vs. Claims Tell the story of

Chronic Conditions – Best Practices Routine Diabetes Care in NYS - Physician Office & Outpatient Setting, 2011 A1C Test 1,478 (4.1%) 1,437 (4.0%) 10,200 (28.4%) 15,552 (43.3%) 1,384 (3.8%) 1,316 (3.7%) Eye Exam Lipid Profile 1,527 (4.2%)

Risk Scores Quantify the increase in future health care costs based on demographic factors, chronic conditions, and interactions of chronic conditions Example: Hierarchical Conditions Categories CMS method for adjusting payments to Managed Care plans based on score for each beneficiary Discuss risk scoring methods Add smart art – circles for the demographic factors, most severe manifestation of each disease, interactions, predictive power Risk scores to put population into cohorts to compare care patterns. Identify who needs more or less care. Informs care-givers of need for interventions, care management. Risk scores to predict total spend for negotiations between providers and payers. Used to target individuals for preventative care.

Risk Scores - Factors Factors must be predictive of variation in cost of health care Hierarchies of dx codes that flag the most severe manifestation of a disease Interaction of chronic conditions such as CHF, ESRD, Diabetes Demographic – Age, Gender, etc. Discuss risk scoring methods Add smart art – circles for the demographic factors, most severe manifestation of each disease, interactions, predictive power Risk scores to put population into cohorts to compare care patterns. Identify who needs more or less care. Informs care-givers of need for interventions, care management. Risk scores to predict total spend for negotiations between providers and payers. Used to target individuals for preventative care.

Risk Scores Factors can gain or lose predictive power and so must regularly adjust Ex. Chronic kidney disease - lower level manifestations were removed from 2014 HCCs because they no longer contribute to prediction of costs Risk scores to put population into cohorts to compare care patterns. Identify who needs more or less care. Informs care-givers of need for interventions, care management. Risk scores to predict total spend for negotiations between providers and payers. Used to target individuals for preventative care.

Payment is an all-in price for the bundle DRGs on steroids Episodes of Care “Bundle” all services related to a particular condition, diagnosis or procedure Payment is an all-in price for the bundle DRGs on steroids Create financial incentives for providers to work together CMMI Bundled Payments for Care Improvement Arkansas Medicaid Program Bundled payments Funds improvement in care management Payment can be retro or prospective Precedence and “attributed” provider – Arkansas makes the physician the responsible provider, CMMI experimenting with physician, hospital, post acute provider

Episodes of Care Common episodes Surgical Medical Total hip/knee replacements Spinal fusion Cardiac valve replacements PCIs Medical Stroke Heart failure Bundled payments Funds improvement in care management

Episodes of Care What physician specialties are involved in caring for stroke patients? Bundled payments Funds improvement in care management

Episodes of Care Bundled payments Funds improvement in care management

A measure of insurance spend PMPM Per member per month A measure of insurance spend Looks at all healthcare encounters by insured person Not limited by diagnosis or procedure Based on at least one year of data Can be combined with stratifications of the population to compare ACOs and capitation PMPM stratifies the member spend into care settings Compare PMPM stratifications and totals across/between risk categories, age cohorts, product line, primary care physician, etc. Comparison can lead to decisions about high performing providers, incentive payments, etc.

Payer Incentives/Penalties Continuous Improvement programs Quality Measures Public Report Cards Payer Incentives/Penalties Continuous Improvement programs CMS Partnership for Patients Leapfrog

Quality Measures – Risk Adjustment Example - Direct Standardization population rate = expected rate at varying levels Adjusts for differential case mix No adjustment for Socioeconomic Factor CMS HF mortality measure uses the US avg as the population rate

Quality Measures – How many ways can you define Readmission? Potentially Preventable vs. All-cause Condition Specific vs. Hospital-wide Chain vs. not-chained 7 Day vs. 30 Day How many different ways can you measure readmissions

How Do they all fit together? Chronic Conditions Risk Score Frequency of events Higher Cost of Events Measurement Quality Metrics PMPMs Episodes of Care Demographic and socioeconomic factors are predictive of chronic conditions, chronic conditions are predictive of frequency of events and higher cost of care both in total and per event. We quantify this in total with Risk Scores. We can measure frequency of events and prevention of events with quality metrics; we can measure higher cost in total (for a population) with PMPMs, measure cost per event with episodes of care.

Tools for Evaluating Patient-Centered Analytics

Analytic Tools Web-Based BI Platforms “BI-lite” Internal Resources Speak to this a little and move on.

The Analytics Team Medical/clinical person as advisor and user – Operational Financial Analytics Medical/clinical person as advisor and user – Operational person – user, also knows about people and processes Financial – someone who understands the financial model and how to measure results Analytics person – we’ll discuss in a minute

The Analytics Team Connector Medical Operational Financial Analytics Connector needs to know enough to fill in the gaps among the other team members Is probably already a team member Inland – was operations person Data doc – DataGen -

The Analytics Guy* NOT the guy who maintains your PCs NOT a spreadsheet guy NOT a guy who only uses desktop-type tools (Excel, Access) Knows how to use serious database tools (SQL, Oracle, SAS) Understands healthcare data terminology (with help) Can think like your users PC guy doesn’t know how to clean up ICD-9 codes, what proc code modifiers are, difference between BNDD and NPI provider identifiers PowerPivot * In the generic, non-gender specific sense

Case Study 1

Alaska city hospital vs. large metropolis medical center A Tale of Two Cities Patient-centered analytics to evaluate opportunities for care redesign and shared savings Alaska city hospital vs. large metropolis medical center Use graphics from past presos and current SAF work.

Major Joint Replacements – Alaska Hospital

Major Joint Replacements – Small City Hospital Opportunities and Strategies Opportunity Maximize internal cost savings to improve margin under the bundled payment Strategies Minimize risk by seeking protection for high cost outliers Use safe harbors to incentivize physicians Work with vendors

Major Joint Replacements – Metropolis Medical Center

Major Joint Replacements – Metropolis Medical Center Opportunities and Strategies Retain Savings Under New Care Delivery Model Reduce readmissions and improper utilization of SNF care Enhance Revenue Increase market share for Medicare managed care, Medicaid, commercial payers – “Center of Excellence” Free up capacity for more intensive rehab services Engage Physicians Through Gainsharing Expand “pay for performance” Reduce device costs Improve Patient Care Quality and Outcomes Coordinate delivery of services across entire continuum Direct patients to most appropriate care settings

Case Study 2

A Look at Heart Attack Care One market, two hospitals Community hospital Tertiary care facility Heart attack patients arrive at both hospitals Cost to the system varies

Heart Attack (AMI) – Community Hospital

Heart Attack (AMI) – Community Hospital US

Heart Attack (AMI) – Community Hospital

Heart Attack (AMI) – Tertiary Care Hospital

PCI – Tertiary Care Hospital

PCI – Tertiary Care Hospital US

PCI – Tertiary Care Hospital

Vice President, DataGen Questions? Gloria Kupferman Vice President, DataGen gkupferm@hanys.org 518-431-7968