Session #7 Ambulatory Quality: Returning to the Essence of Our Work Neil W. Wagle, MD, MBA Associate Chief Quality Officer Partners HealthCare, Center for Population Health Lara Terry, MD, MPH Medical Director, Clinical Analytics Partners HealthCare, Center for Population Health
Learning Objectives Assess current barriers to successful quality improvement. Describe the key ingredients required to achieve successful improvement. Explain how to construct analytics tools to identify areas for improvement that will have a high impact.
Poll Question #1 How far along is your organization in pursuing data-driven quality improvement? 1) Not yet started 2) Getting our feet wet 3) Growth phase 4) Robust deployment 5) Unsure or not applicable
Partners HealthCare System Partners HealthCare is an integrated system consisting of: Two large academic medical centers (Massachusetts General Hospital and Brigham and Women s Hospital). Six community hospitals. Five community health centers. Five major multispecialty ambulatory sites. Inpatient and outpatient psychiatric and rehabilitation specialty services. Homecare. More than 6,000 physicians.
The Taxonomy of Stupid Stupid Denominator Numerator Data Sources Operations Attribution. Inclusion. Accurate. Comprehensive and nuanced. Up-to-date. Claims + EHR. Problems list. Medications. Trendable: Rolling 12. Real-time feedback. Context (trend + comparison). Allows for judgment.
The Streetlight Effect
The 4th Quarter Push for Quality Metric Reporting
(Bad) Measure Proliferation Is Increasingly Well-Recognized Evidence mount[s] that even superb and motivated professionals [have] come to believe that the boatloads of measures, and the incentives to look good, [have] led them to turn away from the essence of their work. Robert M. Wachter, Interim Chairman, UCSF Dept. of Medicine New York Times Most Emailed Article (1/17/2016)
Don Berwick: Current Measurement Era Isn t Going to Work Era 1 (until late 1900s) Professional dominance. Era 2 (current) Measurement, carrots, and sticks. Era 3 (Future?) Moral Era. Steps to Move to Era 3 1. Fewer measures. 2. Simplify incentives. 3. Decrease focus on $ (incentives). 4. Avoid doctor as Lord. 5. Employ improvement science. 6. Embrace transparency. 7. Protect civility. 8. Listen. 9. Reject greed (as an industry). Institute for HealthCare Improvement Keynote, December 2015
Physician burnout has become a crisis.
The Taxonomy of Stupid Stupid Denominator Numerator Data Sources Operations Attribution. Inclusion. Accurate. Comprehensive and nuanced. Up-to-date. Claims + EHR. Problems list. Medications. Trendable: Rolling 12. Real-time feedback. Context (trend + comparison). Allows for judgment.
Better Hypertension Measure Definition Denominator: All primary care patients who have hypertension as defined by multiple clinical and billing sources. Numerator: 140/90; if age > 60 150/90. Credit if DBP 70. Use better of last blood pressure (BP) or the average of last 3 BPs over 18 months. Credit if on 3 anti-hypertensive agents. Persell, S. D., Kho, A. N., Thompson, J. A., & Baker, D. W., (2009). Improving hypertension quality measurement using electronic health records. Medical Care, 47(4), 388-394.
Exceptions Preserve Autonomy Numerator Terminally ill. Adverse reaction to medication. Anatomically not applicable. Competing comorbidity. Patient declined. Patient cannot afford. Denominator Deceased. Not a patient of this PCP. Not a patient of this clinic. Misdiagnosis.
Ingredients for Success 1 The Measures Clinicians must believe they are important. 2 The Tool Registry : a tool to close gaps in care. 3 The Team People to use the tool (hopefully not frontline docs). 4 5
Behavior Change = Feedback + Motivation
Ingredients for Success 1 The Measures Clinicians must believe they are important. 2 The Tool Registry : a tool to close gaps in care. 3 The Team People to use the tool (hopefully not frontline docs). 4 The Data Near-real time measurement, run-charts, and comparison. 5 Motivation Social pressure, Transparency, Financial, Shared purpose
Clinical Registries and the Quality Insights Analytics Change the way we measure.
1. Measurement That Reflects Reality
Clinical Registry-Based Measures More clinically relevant measures. Increased buy-in from clinicians. Increased investment in tools and effort. Improve on clinically relevant measures. Better care.
2. Real-time, Actionable Data
3. Easy to Access and Use
Delineate and Define Roles and Responsibilities Executive sponsor. Data analyst. 1 Business owner. 6 2 5 3 Project manager. 4 Subject matter experts. Data architect.
Engage Your Stakeholders Early in the Process Role for Stakeholder Group Inform development team of stakeholder needs. Who is the audience? How do they use the data? What do they need to see? What do they want to see? What is their level of analytical sophistication? How to best display this? What is the existing workflow and how do they anticipate it being integrated? Local champions when application is released.
Identify Sources for Data Elements Hard/structured data elements from electronic health record (EHR). Visits, bills, labs, vitals, health maintenance, immunizations, specialized flowsheets, PROMs. Metrics = calculations based on the data elements. Soft/unstructured data elements from EHR as available and needed. Findings from radiology, pathology, imaging findings. Claims for risk populations only (commercial at-risk, ACO, Neighborhood Health Plan, Medicaid). Manually added elements as available and needed. SmartForms, bar code scanners.
Registry Primer: Types of Registries Internal Registry For clinical care and internal quality improvement purposes measuring performance, identifying variability, focusing on improvement. Underlying data elements and inclusion rules which determines measure denominators. Data could be used for research retrospectively with Internal Review Board approval. External Registry For submission to national or research registry.
Identify Numerator and Denominator for Measure Numerator = Patients for whom measurement is expected. Exclusions: Terminally ill. Adverse reaction to medication. Anatomically not applicable. Competing comorbidity. Patient declined. Patient cannot afford. Denominator = Patients who meet inclusion. e.g., all diabetics. Exclusions: Deceased. Not a patient of this PCP. Not a patient of this clinic. Misdiagnosis.
What Is the Target For the Measure? Is there an industry target to be met? Is there a benchmark? Local? National? Other? How do the providers compare with their peers in similar settings?
Data Validation: Ensuring Accurate Data Is KEY to Provider Engagement Validate data throughout the process. Registry denominator accuracy. Errors of inclusion and inadvertent exclusion (type 1 and type 2 errors). Numerator accuracy. Were exclusion criteria correctly applied? Calculation. Is the math right? Provider Attribution. Did the patient get linked to the right provider?
Embed It in the Existing Workflow Reduce the number of clicks to get to the data. Ideally, the data should be actionable from the same site that it s viewed by the same person who views it. Find a gap. Implement improvement at same time in same place. For Clinical Care in EHR Patient-level detail. Run as user or clinic. Sort/Filter. Real-time. Close gaps in care.
Embed It in the Existing Workflow Use analytics applications based on data sources within the EDW for: Quality improvement. Aggregate data. Compare clinics/rsos. Weekly updates. Identify variability. Discover best practices.
Poll Question #2 How effective is your organization at identifying and using impactful quality measures using your own data? 1) Not effective 2) Somewhat effective 3) Moderately effective 4) Very effective 5) Unsure or not applicable
Success Story
In the last year CA - Colorectal Cancer Screening CA - Breast Cancer Screening CA - Cervical Cancer Screening 74% 72% 70% 68% 66% 64% 62% +7.5% on 140,000 people 82% 81% 80% 79% 78% 77% 76% 75% 74% 73% 72% +5% on ~78,000 women 74% 72% 70% 68% 66% 64% 62% 60% +7.5% on ~150,000 Women CVD - Lipid Control HTN - BP Control 74% 72% 70% 68% 66% 64% 62% 60% +8% on ~25,000 CVD Patients 78% 77% 76% 75% 74% 73% 72% 71% +3% on ~100,000 HTN Pts 78% 76% 74% 72% 70% 68% 66% 64% DM - Lipid Control +8% on ~28,000 Diabetics 83% 82% 81% 80% 79% 78% 77% 76% 75% 74% DM - BP Control +3% on ~28,000 Diabetics
Are Exceptions Driving Performance? Breakdown of Gains in CVD Lipid Control Breakdown of Gains in Hypertension Control 100% 100% 90% 80% 70% 60% 64.9% 0.1% 74.1% 4.2% 90% 80% 70% 60% 74.2% 0.0% 77.2% 0.3% 50% 50% 40% 30% 64.8% 69.9% 40% 30% 74.2% 77.2% 20% 20% 10% 10% 0% September 2016 September 2017 24,969 patients 24,911 patients 0% September 2016 September 2017 100,158 patients 97,381 patients Clinical passing Exception passing Clinical passing Exception 36 passing
Measure in Lives Measure NNT / NNS Patients Newly Passing Lives Saved or Stroke / MI prevented Hypertension BP Control 1:125 (death) 1:67 (stroke) 1:100 (MI) 2,816 ~93 CVE Lipid Control 1:27 (composite death, MI, stroke) 1,973 ~73 Diabetes Lipid Control 1:28 (composite death, MI, stroke 2,354 ~87 Diabetes BP Control 1:125 (death) 1:67 (stroke) 1:100 (MI) 895 ~29 Colorectal Cancer Screening 1:107 (death from colon cancer) 10,559 ~99 Cervical Cancer Screening 1:1000 (death from cervical cancer) 10,975 ~11 Breast Cancer Screening 1:368 (death from breast cancer) 4,084 ~11 Total: 403
Providers and Managers Happier Physician Survey: Overall, what impact did these activities have on the care provided to your panel of patients? Positive Impact: 85% (Large or Small = 102/120) Doctor: That [population health manager] is worth her weight in pure Spanish saffron! Staff: This is life-changing. I did in minutes what it used to take me weeks to do.
Lessons Learned and Key Takeaways Not all measures are created equal. 1 2 Engage stakeholders early in the process. High quality, accurate data is required to engage stakeholders. 3 4 Measures must be embedded in existing workflow.
Thank You