Prompting to enhance evidence-based practice in the ICU Jeremy M. Kahn, MD MS Associate Professor of Critical Care Medicine Director, Program on Critical Care Health Policy University of Pittsburgh
Disclosures Grant funding from NIH and HRSA Consulting fees from the US Department of Veterans Affairs ICU telemedicine In kind research support from Cerner
Outline Gap between evidence and practice in the ICU Prompting as a tool to narrow that gap Theory Application Limitations and future directions
Outline Gap between evidence and practice in the ICU Prompting as a tool to narrow that gap Theory Application Limitations and future directions
Kahn Int J Qual Safe Health 2007
Cooke, Critical Care 2008 Day 3 Ventilator settings Open (n=277) Closed (n=482) P Value Mean tidal volume 10.8 ± 2.9 9.3 ± 2.3 <0.001 Tidal volume <6.5 ml/kg 5% 11% <0.01 Tidal volume > 12 ml/kg 31% 10% <0.001 Mean Plateau Pressure 27 ± 8 25 ± 8 <0.001 Median PEEP 5 [5 8] 5 [5 10] 0.22
Traditional multi-faceted QI is both challenging and weak Intervention N Measure of effect P-value Protocol 6 0.99 (0.92 1.08) 0.88 Protocol + education 7 1.03 (0.98 1.08) 0.26 Education + audit & feedback 3 1.58 (0.99 2.51) 0.06 Protocol + audit & feedback 7 1.04 (0.94 1.14) 0.48 Protocols/guidelines/bundles + education 9 1.08 (1.01 1.16) 0.02 Protocols/guidelines/bundles + education + reminders Protocols/guidelines/bundles + education + audit & feedback Protocols/guidelines/bundles + education + reminder + audit & feedback 4 1.28 (0.99 1.65) 0.06 3 1.5 (0.72-3.13) 0.27 4 2.23 (1.08 4.59) 0.03 Sinuff CCM 2013
Outline Gap between evidence and practice in the ICU Prompting as a tool to narrow that gap Theory Application Limitations and future directions
Prompting as a solution Signal detection theory Active errors: Heuristics and biases Passive errors: failure to bring decisions into the active realm
Examples from daily life
Clayton J Appl Behav Analysis 2006
% washing hands 100 80 60 40 20 0 Women No sign Sign Men Johnson Percept Mot Skills 2003
Intervention N % Increase Paper-based 19 14 ± 15 Computer generated 34 12 ± 13 Fully computerized 8 13 ± 18 Dexheimer JAMIA 2008
Decreased total transfusions Decreased inappropriate transfusions (threshold Hgb > 7 mg/dl) Rana CCM 2006
Weiss AJRCCM 2011
7 ICUs, 2 campuses, single hospital Remote monitoring plus remote screening and prompting Lilly JAMA 2011
Improvements in processes of care Practice/Outcomes Pre Post P value Stress ulcer prophylaxis 83% 96% <0.001 DVT prophylaxis 85% 99% <0.001 β-blockers after CV surgery 80% 99% <0.001 HOB elevation 33% 52% <0.001 Lilly JAMA 2011
Lilly JAMA 2011
ICU telemedicine and mortality Wilcox, Critical Care 2012
8 ICUs in single hospital EHR facilitated, nurse-led screening and prompting IHI ventilator bundle No remote monitoring Kahn Crit Care Med (in press)
Improvements in process and outcomes Duration of mechanical ventilation Before (N=4339) After (N=8938) Difference* (95% CIs) 5.5 days 5.1 days -0.61 days <0.01 P ICU length of stay 4.1 days 3.9 days -0.22 days 0.02 * Adjusted for patient demographics and severity of illness
Outline Gap between evidence and practice in the ICU Prompting as a tool to narrow that gap Theory Application Limitations and future directions
Limitations Human prompters not scalable Computerized prompts are annoying Solutions Smart computerized prompts Prevent alarm fatigue
Novel informatics approaches Hi, You I m look Clippy, like your want Office to Assistant. be annoyed. Would Can you I annoy like some you today? assistance today?
Smart Bayesian belief networks for prompting Would not come up in emergencies Could determine when practice is unlikely to be indicated Would learn from your past history Priors shift as you improve your ordering The better we are, the less we get prompted
Conclusions Traditional QI is not enough Prompting is an behavioral economics intervention to increase best practice Best suited for low-risk interventions little uncertainty about use
www.ccm.pitt.edu/crisma