Distributed Subgradient Methods for Convex Optimization over

33
HAI (VAP) and the patient Alison S. Clay, MD, FCCP Duke University Healthcare Team Training

Transcript of Distributed Subgradient Methods for Convex Optimization over

Page 1: Distributed Subgradient Methods for Convex Optimization over

HAI (VAP) and the patient

Alison S. Clay, MD, FCCP Duke University

Healthcare Team Training

Page 2: Distributed Subgradient Methods for Convex Optimization over

Disclosures

• This is my personal experience, told with the support of my administration

• Although I am a physician, I am a fierce patient advocate

• I believe teamwork is the only way to innovate in healthcare – I do consulting work with healthcareteamtraining

(HTT)

Page 3: Distributed Subgradient Methods for Convex Optimization over

Goals

• Discuss the impact of HAI on patients – with particular attention to VAP – Considering Post-ICU Care Syndrome (PICS)

• Share the impact of Duke initiatives to prevent HAI, from both the individual’s and system’s perspectives

Page 4: Distributed Subgradient Methods for Convex Optimization over

The start

Page 5: Distributed Subgradient Methods for Convex Optimization over

On Call

Page 6: Distributed Subgradient Methods for Convex Optimization over

The work up

Page 7: Distributed Subgradient Methods for Convex Optimization over

The diagnosis

Page 8: Distributed Subgradient Methods for Convex Optimization over

Iwashyna, T. Am J Resp and CCM 2012; 186: 302-304

Page 9: Distributed Subgradient Methods for Convex Optimization over

The cost of a port infection

1 1 1 1 2 6

+

X2 ports 12

X2 per day 24

X6 weeks >1000

Page 10: Distributed Subgradient Methods for Convex Optimization over

Trajectory of Illness

Page 11: Distributed Subgradient Methods for Convex Optimization over
Page 12: Distributed Subgradient Methods for Convex Optimization over

Trajectory of Illness

Page 13: Distributed Subgradient Methods for Convex Optimization over
Page 14: Distributed Subgradient Methods for Convex Optimization over
Page 15: Distributed Subgradient Methods for Convex Optimization over

Trajectories of Care for patients with PMV

Unroe, M et al. Annals of Internal Medicine 2010; 153:167-175

Page 16: Distributed Subgradient Methods for Convex Optimization over

Trajectory of Illness

Page 17: Distributed Subgradient Methods for Convex Optimization over

A few other discussion points

Page 18: Distributed Subgradient Methods for Convex Optimization over

Delirium Subtypes

From Vanderbilt ABCDE educational module

Page 19: Distributed Subgradient Methods for Convex Optimization over

ICU Delirium • Increased ICU length of stay (8 vs 5 days) • Increased hospital length of stay (21 vs 11

days) • Increased time on ventilator (9 vs 4 days) • Higher ICU costs ($22,000 vs $13,000) • Higher ICU mortality (19.7% vs 10.3%) • Higher hospital mortality (26.7% vs 21.4%) • 3-fold increased risk of death at 6 months

Ely, et al. ICM2001; 27, 1892-1900 Ely, et al, JAMA 2004; 291: 1753-1762 Lin, SM CCM 2004; 32: 2254-2259 Milbrandt E, et al, Crit Care Med 2004; 32:955-962. Ouimet, et al, ICM 2007: 33: 66-73. From Vanderbilt ABCDE educational module

Page 20: Distributed Subgradient Methods for Convex Optimization over

Mobility

Page 21: Distributed Subgradient Methods for Convex Optimization over

Duke Initiatives

• Wake Up and Breathe (and redefine VAP) • Early Mobility • Patient Safety Education (in the professional

schools) • Patient Centered Care

Page 22: Distributed Subgradient Methods for Convex Optimization over
Page 23: Distributed Subgradient Methods for Convex Optimization over

“SAT + SBT” Was Superior to Conventional Sedation + SBT Intervention (SAT) group = Less benzodiazepine

Girard et al. Lancet 2008; 371:126-34

Extubated faster Discharged from ICU sooner

From Vanderbilt ABCDE educational module

Page 24: Distributed Subgradient Methods for Convex Optimization over

“SAT + SBT” Was Superior to Conventional Sedation + SBT

Intervention (SAT) group = More unplanned extubation, but not more reintubation

Discharged from hospital sooner Better survival at 1 yr

Aliv

e

Girard et al. Lancet 2008; 371:126-34 From Vanderbilt ABCDE educational module

Page 25: Distributed Subgradient Methods for Convex Optimization over

Early Exercise in the ICU

• Early exercise = progressive mobility • Study design: paired SAT/SBT protocol

with PT/OT from earliest days of mechanical ventilation

Schweickert WD, et al. Lancet. 2009;373:1874-1882.

Wake Up, Breathe, and Move

From Vanderbilt ABCDE educational module

Page 26: Distributed Subgradient Methods for Convex Optimization over

Early Exercise Study Results

Schweickert WD, et al. Lancet. 2009;373:1874-1882.

Page 27: Distributed Subgradient Methods for Convex Optimization over

Mobility- No excuses

Cardiopulmonary Phys The J 2012 March; 23(1):30-35

Page 28: Distributed Subgradient Methods for Convex Optimization over

Mobility- No excuses

Cardiopulmonary Phys The J 2012 March; 23(1):30-35

Page 29: Distributed Subgradient Methods for Convex Optimization over

Duke Data • Protocol started 2.5 years ago

– Competencies for returning PT (ortho->MICU) • Suctioning • Familiarity with Duke protocol

• Mini-QI prior to start: – 40-50% being seen by therapist (long delay in ordering; nurses

not activating order) – PT/OT rounding with team did NOT work

• Post Protocol data underway – % patients getting PT, how many times a day seen, highest level

of activity, etc. – reasons not referred, % referred but not treated – Delirium Screen

Page 30: Distributed Subgradient Methods for Convex Optimization over

IPE: Patient Safety (PT and SON)

Page 31: Distributed Subgradient Methods for Convex Optimization over

Team Training with the patient

Page 32: Distributed Subgradient Methods for Convex Optimization over

True Patient Centered Care

Page 33: Distributed Subgradient Methods for Convex Optimization over

Great Team (family and HS!) Patient Centered Care Health System invested in Continual Improvement

Bad Disease Many Health System Encounters A little PTSD