Clinical Area of Focus: Radiology Group Members: Dr. Nadja Kadom, Mohammadali Mojarrad, Kristin McDougall Faculty Advisor: Dr. Nadja Kadom By Mohammadali Mojarrad
SYSTEM ERRORS Organizational Clustering Faulty medical history Teamwork/communication Inefficient processes Management/supervision Policy/procedures Technical errors Unavoidable errors COGNITIVE ERRORS Faulty information processing Faulty interpretation Premature closure Over-interpretation Faulty context generation Failure to order f/u Faulty data gathering Faulty test technique Ineffective history review/exam Faulty knowledge base Inadequate skills Perceptual error Under-interpretation 2011 G. Taylor
SYSTEM ERRORS Organizational Clustering Faulty medical history Teamwork/communication Inefficient processes Management/supervision Policy/procedures Technical errors Unavoidable errors COGNITIVE ERRORS Faulty information processing Faulty interpretation Premature closure Over-interpretation Faulty context generation Failure to oredr f/u Faulty data gathering Faulty test technique Ineffective history review/exam Faulty knowledge base Inadequate skills Perceptual error Under-interpretation 2011 G. Taylor
Hawkins 2014
Radiology reports are an important means of communication between radiologists and other health care providers. Voice recognition (VR) software has largely replaced transcriptionists and decreased report turnaround times. Voice recognition software can cause errors in radiology reports that can … Harm patients Cause a negative perception of the radiologist Delay report finalization, cause clinician calls
Error rates using voice recognition (VR) versus conventional transcriptions Pezzullo et al VR reports take 50% longer to complete despite being 24% shorter compared to transcribed reports 5.1 errors /case 90% contain errors before sign-off (vs 10% transcribed) 35% errors AFTER sign-off
Error rates: 35% of reports with errors (Pezzulo et.al. 2008) 36 % of reports with errors (Chian et al. 2010) 22% of reports with errors (Quint* et al. 2008) *Radiologist error rates from 0-100% *No difference native vs. non-native *No difference faculty alone vs faculty/trainee
Radiologist Accuracy at stake Reputation at stake Kristin Girard-McDougall (GE) RIS manager Software expert Mohammadali Mojarrad Willing to put in hours
What are we trying to accomplish? What? For whom? How good? By when? I am going to improve my reports.
We will reduce the number of voice recognition errors in Dr. Kadom’s reports by 20% by August 31, Possible applicability to other radiologists though errors, dictation styles, and proofreading habits differ Plan to share results locally/nationally to inspire other radiologists to do QI on their VR errors
Decrease VR errors by 20% Radiologist’s performance Technology performance
2013 ARRS lecture video David L. Weiss, Imaging Informatics, Carilion Clinic, Roanoke, VA Dictation style Navigation Microphone Macros & templates Vendor selection Plosives, know your software Keep eyes on study Switch off, noise masking, room design Helpful, use as many as possible Important
Decrease VR errors by 20% Radiologist’s performance Technology performance Quality of mic Quality of software Background noise Trainable Quality of mic Quality of software Background noise Trainable Proof reading Enunciation Use of macros Switch off mic Proof reading Enunciation Use of macros Switch off mic
Decrease VR errors by 20% Radiologist’s performance Technology performance Quality of mic Quality of software Background noise Trainable Quality of mic Quality of software Background noise Trainable Proof reading Enunciation Use of macros Switch off mic Proof reading Enunciation Use of macros Switch off mic
VR error rate System PeopleTechnology Background noise interference Ineffective Voice recognition Accent Limited Time Lack of adequate proof reading Work space noise Work load Fatigue No resident Lack of transcriptionist or Proof reader Cost Microphone
IdeasSelectedReasoning Proof readingXNecessary ChecklistXHelp build new proof reading habit Fix “disk” errorXEasy fix in software Fix “insert macro” errorXInsert manually rather than dictate User profile resetXGE request Conscious dictationXNecessary Background noiseCannot improve Notify GE of issuesVoice files deleted, GE cannot review New softwareNo money
Outcome measure % of reports with errors Process measure Proofreading evidenced by observer Balancing measure Time spent proofreading
Monday (n=428) Saturday (n=518) Total (n=946) Number of reports with errors 157 (37%)269 (52%)426 ( 45%) July 2013 – June Monday and 1 Sunday each month= 12 Mondays & 12 Sundays
Baseline Goal
May 1: Medicolegal ARRS May 22: GE reset July: Checklist Goal
History: spelling errors Technique: contrast dose Findings: Proofread No displaced Nondisplaced Conus colon is Colon : Comma Common No Do ‘Slight’ ‘Marked’ Insert macro
Saturday, n=33 reports (only Kadom) Error rate: 16/33 = 48% (Baseline 52%) Monday, n=53 reports (co-authored only) Error rate: 15/53 = 28% (Baseline 37%) Overall error rate: 38% (Baseline 45%)
Process measure Weekday, 100% of reports were proofread Despite proofreading, there were some errors remaining in the report before signoff Balancing measure 15% time spent proofreading Time spent proofreading each report = 2min (mean) Macro: Manual insertion takes only 2s longer RVUs do NOT cover VR (used to be covered by technical fee)
Continue to monitor % error in reports Aim at error rate < 22% Reflections: I thought I was proof reading, but not focused I did not know I had to serve the dictation software; I thought the dictation software served me.
Galen of PergamonThomas Sydenham Hippocrates
IOM Reports
ResearchQI PurposeDiscover new knowledgeBring knowledge to practice TestsLarge blinded testMany sequential tests BiasesControlStabilize from test to test Data Gather as much as possible Learn just enough for modification DurationLongShort IRB required at BMC if publication considered
WALTER A. SHEWHART March 18, 1891 – March 11, 1967 American physicist, engineer and statistician Western Electric Company Importance of reducing variation Continual process-adjustment Control chart “Shewhart cycle” W. EDWARDS DEMING October 14, 1900 – December 20, 1993 American engineer, statistician, professor, author, lecturer, and management consultant Plan-Do-Study-Act (PDSA)-cycle Highly influential on post WWII economic growth in Japan Toyota’s famous production system (“A3”)
Appreciation of a system Understanding the overall processes involving suppliers, producers, and customer/recipients Knowledge of variation Use of statistical sampling to determine range and causes of variation Theory of knowledge The limits of what can be known Knowledge of psychology Concepts of human nature
Langley GL, Moen R, Nolan KM, Nolan TW, Norman CL, Provost LP. The Improvement Guide: A Practical Approach to Enhancing Organizational Performance (2nd edition). San Francisco: Jossey-Bass Publishers; 2009.The Improvement Guide: A Practical Approach to Enhancing Organizational Performance Forming the Team Setting Aims Establishing Measures Selecting Changes Testing Changes Implementing Changes Spreading Changes } Plan DoStudy Act
Instructions Pre-Survey My IHI certificates Planning Phase Current state & Data display PDSA cycles Summary Post-survey Assess & Improve ACGME & Great teaching
Instructions Pre-Survey My IHI certificates Planning Phase Current state & Data display PDSA cycles Summary Post-survey