©2013 MFMER | slide-1 Learning from Every Death: Moving From Counting to Improvement Jeanne M Huddleston, MD, MS Associate Professor of Medicine Chair, Morbidity & Mortality Council Mayo Clinic, MN, USA
©2013 MFMER | slide-1
Learning from Every Death: Moving From Counting to Improvement
Jeanne M Huddleston, MD, MSAssociate Professor of Medicine
Chair, Morbidity & Mortality CouncilMayo Clinic, MN, USA
©2013 MFMER | slide-2
Agenda
1. Mayo Clinic and US Healthcare Performance
2. Mortality Review System (MRS) History
3. MRS as a Safety Learning System
4. Case Study and Results
5. Possibilities for Denmark
St Marys Campus, Mayo Clinic Hospital
Total clinic patients:
1,260,000
Hospital admissions:
131,000
Hospital days:
608,000
Employees:
> 60,000
Annual Patient Encounters
©2013 MFMER | slide-5
©2013 MFMER | slide-6
©2013 MFMER | slide-7
21st century health care- Inefficient processes- Poor system integration- High levels of variation
- Care delivery - Outcomes
- Suboptimal value
©2013 MFMER | slide-8
©2013 MFMER | slide-9
©2013 MFMER | slide-10
©2013 MFMER | slide-11
11
12
What are you doing to learn from groups of
patients at high risk of harm?
What could you do to improve learning?
©2013 MFMER | slide-13
12 years in the making….
©2013 MFMER | slide-14
Iterative Learning: 12 year journey
M&M Tradition
100 death review –
2004
4th quarter review –
2005
Continuous review – 4th
quarter 2006
MCA (100% cases reviewed)
early 2009
MCF (targeted reviews)
later 2009
MCHS review roll-out
2012
Enterprise-wide
integration
M&M Tradition
100 death review –2003
4th quarter review –2005
Continuous review – 4th
quarter 2006
©2013 MFMER | slide-15
Original Charge from Hospital Leadership
1. To create a meaningful mechanism to review deaths at MCR hospitals:
• Thoroughly understandable
• Measurable
• Improvable
2. To identify and quantify unanticipated deaths
3. To identify rate of adverse events in patients who die in MCR hospitals
4. To classify and quantify system level changes which will improve mortality rate.
©2013 MFMER | slide-16
“No one should ever suffer or die as a result of failures in our systems or processes of healthcare delivery.”
MC Mortality Review Subcommittee, May 2007
16
©2013 MFMER | slide-17
Guiding Principles:Not Negotiable
1. System review (not peer review)
2. Every case is reviewed by a practicing nurse and physician
3. All findings are recorded in the central registry
4. Multidisciplinary, multispecialty sessions used to build consensus re: findings
5. Implementation is local
©2013 MFMER | slide-18
Aggregate learning
Aggregate learning
Is there anythingthat could mitigate
future events?
Is there anythingthat could mitigate
future events?
Prioritization of information
Prioritization of information
Identification of issues
Identification of issues
Raw InformationRaw Information case reviews
case reviews
ProblemProblem
Further reviewFurther review
YesYes
ReportReport
NoNo
No further review
No further review
No problem
No problem
Re
vie
we
r W
ork
Co
mm
itte
e W
ork
Clinical
PracticeQuality
©2013 MFMER | slide-19
Primary Reviewers (MD and RN)
Encompass the entire hospitalization
Include pre-hospitalization information as available
Identify possible system and process failures
Identify possible instances when standard of care was not followed
• Documentation, supervision, clinical guidelines, etc
Identify possible delays in treatment or diagnosis
Identify possible communication or documentation issues
©2013 MFMER | slide-20
Information Recorded into the Registry
Qualitative and quantitative
Qualitative: Patient’s story of healthcare experience
• Level of detail
• Describe timelines of issue identified
• Enough to remember for future presentation
• Enough for leadership to understand concern
Quantitative: Indicators
Case classification
©2013 MFMER | slide-21
Multidisciplinary Team Discussions
Many brains are better than one
Reviews are inherently subjective• Retrospective
• Rely on documentation
Clinical expertise
Cultural understanding
Consensus
Classification of death21
©2013 MFMER | slide-22
Overall Case Classification
Iterative learningIterative learningIterative learningIterative learning
NOT NOT NOT NOT related to to to to preventabilitypreventabilitypreventabilitypreventability
Consensus driven optionsConsensus driven optionsConsensus driven optionsConsensus driven options• Opportunity Opportunity Opportunity Opportunity for improvementfor improvementfor improvementfor improvement
• NO NO NO NO opportunity for improvementopportunity for improvementopportunity for improvementopportunity for improvement
22
©2013 MFMER | slide-23
Morbidity and Mortality Council: The Think Tank (of data geeks)
• Integration of safety and quality leadership with practicing providers and analysts
• Aggregate and reconcile all safety data
• Strategic data analysis
• Thoughtful interpretation
• Integrate recommendations to practice leadership
©2013 MFMER | slide-24
Omission vs. CommissionOmission vs. CommissionOmission vs. CommissionOmission vs. Commission
Mayo Clinic, Mortality Review System
©2013 MFMER | slide-25
Example: delayed diagnosis of sepsis & delayed recognition of a postoperative complication
59 year old female underwent TAH
POD #3 – AKI, urinary retention with new abdominal distension and pain
POD #4 – AKI worse, significant abdominal pain – narcotics stopped. Episode of PAF (130’s)
POD #5 – hypotensive (70/45) with diaphoresis and nausea
RRT called but no blood pressure on their arrival
Code called with > 1hr of resuscitation efforts
On autopsy, abdomen filled with pus and a knick in the small bowel.
©2013 MFMER | slide-26
82 year old male with severe COPD and pancreatic cancer was hospitalized for bowel obstruction.
Postoperative delirium
Postoperative respiratory failure
Pain meds held
Per nursing notes
- patient routinely called out in pain
- family members consistently asked that he be kept comfortable.
Average pain score was 8/10 in the 24 hours preceding death.
Patient was made comfort care only and died within hours.
Joshua Bright: A Good Death
Can patients
have a “good”
death?
©2013 MFMER | slide-27
©2013 MFMER | slide-28
P-chart of % of deaths with issues identified
Making a Difference through 2009
Nursing sepsis education
Wipe C-diff initiative
Sepsis order set
Admission office
Airway management protocol
Other localized efforts: official reads on outside films, consultant notification of patient condition change, NGT suction and chest tube in-services, DOM encouragement of RRT utilization
Avg=0.2262
LCL
UCL
Data source: MRS Mortality Registry
©2013 MFMER | slide-29
MRS System Findings (2010)
Narcotic-induced respiratory depression• postoperative
• OSA
Failure-to-Rescue
Failure-to-Recognize• Septic shock
• Hypovolemic shock
• Mesenteric ischemia
Triage of unstable patients
Supervision
©2013 MFMER | slide-30
SPC p-chart of MCR mortality rate
Data source: DSS and UHC
This information is confidential and protected from disclosure by Minnesota Statute 145.61 et seq.
©2013 MFMER | slide-31
P - chart of % of deaths with issues identified
Data source: MRS Mortality Registry
This information is confidential and protected from disclosure by Minnesota Statute 145.61 et seq.
©2013 MFMER | slide-32
Days Between Deaths (during 2007-2014 there were 669 deaths)
©2013 MFMER | slide-33
Slightly Deeper Dive
Data all from mortality review registry
2008 – 2014 (7 full years)
Inclusion criteria• Patients with either of two indicators:
1. Delayed diagnosis
2. Resuscitation
3. Only one event per patient per day per service
©2013 MFMER | slide-34
FTRR events by nursing unit
©2013 MFMER | slide-35
FTRR events for non-ICU nursing units by year
©2013 MFMER | slide-36
RRT called
Clear change in physiologic
state
Met RRT
Criteria
44 hours
27 hours
Labs:
Hb: 7.1, Leukocytes: 2.3
Creat (9/29): 3 (2-month baseline): 1.6
55-year-old male with CLL with MUD peripheral blood stem cell
transplant 7/10 (2 months before). Dismissed 9/14
Case Example
©2013 MFMER | slide-37
©2013 MFMER | slide-38
©2013 MFMER | slide-39
HSE Approach: Bedside Patient Rescue
• FMEA
• ID & understand stakeholder needs
• Classify/quantify user requirements
• Data mining: Identification of covariates predictive of deterioration
• Simulation modeling
• Integrate stakeholder needs, results of analytics and operational use specifications
• FEA
• Iterative sensitivity & specificity testing of predictive algorithms
•Simulation modeling
• Develop pilot
• Evaluate pilot
• Determine if & when implementation appropriate
• Develop implementation plan
• Develop monitoring plan
• Implement
Planning PhaseImplementation
Phase
Execution
Phase
BPR Project
Management Team
BPR Project
Management Team
Requirements
Team
Requirements
Team
Analytics
Team
Analytics
TeamSystems
Architecture
Team
Systems
Architecture
Team
Clinical
Verification &
Validation
Team
Clinical
Verification &
Validation
Team
Clinical
Pilot Team
Clinical
Pilot Team
Clinical
Integration
Team
Clinical
Integration
Team
©2013 MFMER | slide-40
Input from > 2,000 providers: Top 5 Failure Modes
1. The patient's clinical condition is not re-assessed at the bedside following new interventions (medication, fluid bolus, tests results).
2. Care providers of all types can feel that there are too many complex things to do in a short period of time.
3. A physician does NOT review nursing notes documented in the electronic medical record.
4. Care team attributes a patient's acute physiological deterioration to the wrong cause.
5. Some care providers (nurses or physicians) believe that a standard, or clear definition of acute patient deterioration does NOT exist.
©2013 MFMER | slide-41
Mortality Review taught us:
Our MAT was not being called.
Research supported by CSHCD
Our providers told us:
Criteria could not be trusted and it did not apply to all patients.
©2013 MFMER | slide-42
Research supported by CSHCD
Best performing score had a positive predictive value of < 1.8%....
98 of every 100 alarms would be wrong!!
How about a different Early Warning Score?
©2013 MFMER | slide-43
Traditional statistics failed to improve positive predictive value. Machine
learning methods to were much more successful (PPV from 2% to 20%).
AUC = 0.93
©2013 MFMER | slide-44
Adding nursing “worry” to the mathematical model drastically
improved the performance (300%).
©2013 MFMER | slide-45
Driving Principle, Locally OperationalizedTiered, time-limited escalation of
expertise at the bedside
RN Validation
Tier 1• First responder to
bedside & notifies senior member of team
Tier 2• Second responder to bedside
for eval of clinical condition, differential, goals of care, and plan for reassessment
Tier 3
• MAT
Evaluate/treat/resolve
or silence/snooze
Max 2-6 hours
Evaluate/treat/resolve
or silence/snooze
Max 2-3 hours
Concept derived from 7 RN focus groups & discussions with MERS, Sepsis MTR, CC-IMP &
HIM quality committee including their practice leadership. MANY details to sort through as
portion of Phase 2 BPR Charter. The practice must design the process and policies.
RN clinical
verification of
vital signs
©2013 MFMER | slide-46
How could you improve learning from mortality analysis (or other forms of
case review) in your hospital?
Name 3 suggestions
©2013 MFMER | slide-47
Why does MRS work?
• Move away from insular peer review
• Embrace concepts of process of care and system failure
• Culture change
• It’s not about adverse events • It’s about learning to identify process of
care failures• It’s about inspiring action
• Right size quality improvement initiatives
©2013 MFMER | slide-48
Multidisciplinary reviews identify opportunities for improvement.
• Ongoing culture shift with issue identification and open discussion
• Long gone are the days of “nurses’ jobs” or “physicians’ responsibilities”
• Increased physician involvement
• Integral part of how business is done• Leadership involvement
• Leadership accountability
©2013 MFMER | slide-49
“Mortality Review takes too much time and uses too many resources.”
©2013 MFMER | slide-50
©2013 MFMER | slide-51
Possibilities for “The Five”
• Work together• Common approach and registry
• Common definitions
• Standardized training
• Start small (100 consecutive patients)
• Reconcile findings with existing reporting mechanisms
• Meet monthly (at hospital level) to share learning
• Benchmark learnings amongst each other• Positive deviants?
“I have always thought a good deal of Lincoln’s Gettysburg address. There’s a line in it which
explains why we want to do this thing. It is ‘that
these dead shall not have died in vain.’ We
know how hard it is for those who have had the misfortune of deaths in their families, of deaths
that might have been avoided. What better could we do than take young men and help them become proficient in the profession so as to
prevent needless deaths?”
Foundation life goal, Dr. William J. Mayo, tells senators.
Minneapolis Morning Tribune, March 23, 1917.
©2013 MFMER | slide-53
Questions & Discussion