BIG ISSUES IN THE NEXT TEN YEARS OF IMPROVEMENT

BIG ISSUES IN THE NEXT TEN YEARS OF IMPROVEMENT Academy for Health Services Research and Health Policy Annual Meeting Washington, DC: June 24, 2002 Donald M. Berwick, MD, MPP

Patient and Community The Chain of Effect in Improving Health Care Quality Experience Aims (safe, effective, patient-centered, timely, efficient, equitable) Micro-system Process Simple rules/design Concepts (knowledge-based, customized, cooperative) Organizational Context Facilitator of Processes Design Concepts (HR, IT, finance, leadership) Environmental Context Facilitator of Facilitators Design Concepts (financing, regulation, accreditation, education)

Three Levels of Aggregation Interactions between patient and care system Design of organizations Design of the environment of care

Frameworks for Patient Segmentation Stratification models robust, but dynamic Not stereotyping, but a prelude to individualization Beyond age, gender, race, and diagnosis

Plausible Candidates for Segmentation Dimensions Predilection for self-care Attitude toward time Attitude toward continuity Need for information Social support structures

Incorporating Knowledge of Variation into Care Statistical process control in physiological monitoring Special and common cause Graphical methods Treatment as prediction Explicit theories Tracking across boundaries

Establishing Continuous Flow Engineer-clinician as a specialty Prototype systems Wait-free, defect-free, inventoryfree Involve the varsity from engineering sciences

Emergency Medical Services Diversion Hours -- Massachusetts 1200 1000 800 600 400 2000 2001 200 0 JAN M AR M AY JUL SEP NOV

Amidst Variable Inflow, How Crowded Can Hospital Units Be? 100% 90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 M onth Predicted Blocking Rate Observed Blocking Rate Utilization As random demand meets a fixed capacity, rejections are accurately predicted by queueing theory (r = 0.9).

Lessons from Queueing Theory: Predictions Calculated performance measures: Utilization (% time servers are busy) Waiting time Length of waiting lines Probability of balking or rejection Response of a queueing system to increasing utilization is nonlinear As utilization rises, waits and rejections increase exponentially Lq re q u e s ts re jected 100 0 80% 70% 60% 50% 40% 30% 20% 10% 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% utilization 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% utilization

Lessons from Queueing Theory: Observations 50% 45% 40% 35% Rejection rate 30% 25% 20% 15% 10% y = 0.0003e 7.8221x 5% R 2 = 0.5294 0% 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% MICU utilization

Elective Surgical Requests vs Total Refusals elective surgical patients seeking ICU admission patients diverted or rejected from the ICU 10 9 8 7 6 5 4 3 2 1 0 4/1/00 4/6/00 4/11/00 4/16/00 4/21/00 4/26/00 5/1/00 5/6/00 5/11/00 5/16/00 5/21/00 5/26/00 5/31/00

Possibilities and Limits of Self-Care Everyone is a doctor as a concept Obstacle-free innovation settings Start with lead customers

Achieving a Truthful System Open the discussion: How truthful are we? Leaders must start it

Stages of Facing Reality: To live divided no more The data are wrong The data are right, but it s not a problem The data are right; it is a problem; but it is not my problem. I accept the burden of improvement

Designing the 21 st Century Medical Record National moon shot Public-domain default option Scalable from the point of care Define quality characteristics Reduce medical record costs 50% Rediscover Larry Weed: Medical Records that Guide and Teach

Designs of Healing Places Innovation laboratories Avoid political correctness Substitute learning Assume nothing from the status quo

Placing Knowledge at the Point of Care Honor thy user Not primarily compliance, but rather human factors design Pull as well as push Create public-domain options

Where Is the Muda in Health Care? Models of muda in health care The business case for waste reduction Exploit international variation for learning Question habits, rules, and beliefs Involve the front line New accounting processes to unmask muda

Percent of Medicare Decedents Admitted to ICU During the Terminal Hospitalization (1995-96) 30.0 % Admitted to ICU (1995-96) 96) 25.0 20.0 15.0 10.0 5.0

Distinguish Need More from Need Different Learn from the variation Population care models should help Transparency to enlighten us W h a t i s e n o u g h?

How Can Joy in Work Be Reliable? Seek joy. High quality cannot come from a demoralized workforce Ethnography, anthropology, stories Need a system-level strategy, not just a discipline-specific strategy

Barriers to Quality that Lie in Health Care Financing Document the toxicities Understand the entire chain of financial flow Explore capitation. Again Define quality characteristics of payment Social experiments on payment

Fair, Equitable, and Open Approach to Medical Injuries Define the quality characteristics of a liability system Social experiments Consider a policy of extreme honesty Consider no-fault, enterprise liability Consider reliable compensation of victims of medical injury

Preparing Young Professionals for Citizenship in Improvement Separateness and fragmentation as learned skills Cooperation as a topic for study Cover the continuum (AAMC, ACGME, etc.) Team training Reconsider professional ethics