surveillance of peripheral arterial disease cases …...right lower extremity critical limb...
TRANSCRIPT
©2017 MFMER | slide-1
Surveillance of Peripheral Arterial
Disease Cases Using Natural Language Processing of Clinical
Notes Naveed Afzal, Sunghwan Sohn, Christopher G. Scott, Hongfang Liu, Iftikhar J. Kullo, Adelaide M. Arruda-Olson
©2017 MFMER | slide-2
Peripheral arterial disease (PAD)
• Affects 8.5 million in US, 200 million worldwide • Associated with high risk of mortality and morbidity • Under-diagnosed and undertreated • Lack of awareness of PAD-associated risks for
adverse cardiovascular outcomes • Diagnosed by ankle-brachial index (ABI) test but this
method remains underutilized
Kullo & Rooke: NEJM, 2016; Hirsch et al: JAMA, 2001; Gerhard-Herman et al: J Am Coll Cardiol, 2016
©2017 MFMER | slide-3
Background
• Manual methods for disease surveillance are costly, time-consuming and inconsistent
• Automated surveillance of clinical notes from EHRs may effectively identify PAD cases
• Use NLP to extract PAD related information from narrative clinical notes
©2017 MFMER | slide-4
Objectives • To develop and validate a PAD surveillance
system using NLP for detection of PAD symptoms from narrative clinical notes
©2017 MFMER | slide-5
PAD diagnosis based on ABI test
• Reports were in PDF format and were not part of clinical notes
• PAD cases • ABI ≤ 0.90 rest or post-
exercise • ≥20% decrease ABI after
exercise • High ABI >1.40 = poorly
compressible arteries
• Controls • Normal ABI results
• Date of ABI testing = index date
Kullo & Rooke: NEJM, 2016
©2017 MFMER | slide-6
Concept features and rules
Documents, patient IDS, dates
NLP Algorithm
System output
EHR System
Sentence Detection
Tokenization
Concept Identification
Assertion
Patient Classification
Evaluation
ABI (Gold
Standard)
©2017 MFMER | slide-7
PAD-NLP Algorithm • Expert clinician manually reviewed clinical notes
• 20 patients with PAD • 20 patients without PAD
• Created list of PAD-related keywords for NLP algorithm prototype
• List of keywords refined by manual review of charts by a board certified cardiologist
Afzal et al: J Vasc Surg, 2017
©2017 MFMER | slide-8
Examples of keywords for confirmation of PAD Right lower extremity critical limb ischemia. She has history of peripheral vascular disease
She has been followed in the Wound Care Center this year for ulcerations to the toes of her right foot
Ultrasound shows severe right lower extremity atherosclerotic disease, beginning at the level of the right common femoral artery Noninvasive lower extremity arterial study demonstrated indeterminate ankle-brachial indices due to poorly compressible vessels bilaterally His past medical history is also significant for peripheral arterial occlusive disease with aorto-bifemoral bypass, Peripheral vascular disease Critical limb ischemia right lower extremity LLE ischemia Noncritical infrapopliteal level lower extremity arterial occlusive disease bilaterally Lower extremity atherosclerosis obliterans Severe lower extremity arterial occlusive disease
Examples of keywords for exclusion of PAD There is no evidence of critical limb ischemia No evidence of PAD Vascular lab - normal lower extremity arterial study No evidence of arterial occlusive disease
©2017 MFMER | slide-9
Keywords - NLP algorithm Confirmation - Disease Location
lower extremities/extremity; lower limbs/limb; Leg /legs; Iliac/femoral/tibial/popliteal artery/ arteries; Distal/ infrarenal /abdominal aorta/aorto (bi)iliac/ aorto (bi)iliac/aorto(bi)-iliac; aorto-(bi)femoral; foot, toe, toes, shin; plantar, heel, ankle, interdigital; below/above knee, Claudication /calf pain; Ischemic ulcer/ulcers; ASO/Arteriosclerosis obliterans/ arterial sclerosis obliterans/ atherosclerotic disease; PAD/ Peripheral arterial disease/Peripheral vascular disease /Peripheral arterial occlusive disease
Confirmation - Diagnosis Arterial occlusive disease/occlusion/occluded; Stenosis; non compressible vessels (NCV), non-compressible arteries (NCA), poorly compressible vessels (PCV), stiff vessels/ arteries ischemia; positive ABI/ankle brachial index/ vascular labs/ extremities study /arterial studies; revascularization/recanalization/bypass/angioplasty/PTA/stenting/stent/graft/endarterectomy/ endarterectomies; thrombectomy/thrombosis/thromboembolectomy/embolectomy/ embolectomies
Exclusion Family history of, Upper extremities/Upper extremity; Arm/arms, hand(s); Brachial artery, axillary artery, radial artery, ulnar artery; carotid, innominate artery, subclavian artery; mesenteric artery; celiac artery; AAA, abdominal aortic aneurysm/abd aortic aneurysm; renal arteries/artery; coronaries, coronary arteries/ artery /cerebrovascular-disease /arteries/artery; Amputation; traumatic/trauma; sarcoma/osteoma; pseudoclaudication/pseudoclaudicatory pain; diabetic foot, hammer toe/ toes; vascular calcification; varicose veins
©2017 MFMER | slide-10
PAD-NLP Algorithm Rules
PAD One disease location keyword + one diagnostic
keyword within two sentences
Non-PAD 1) No PAD criteria 2) One exclusion keyword
©2017 MFMER | slide-11
Note Types Note Sections Service Groups Consult Impression / Report / Plan Primary Care
Subsequent Visit Diagnosis Hospital Internal Medicine
Patient Progress Principal/primary Diagnosis General Medicine
Supervisory Secondary Diagnoses Family Medicine
Limited Exam Past Medical/Surgical History Critical Care
Specialty Evaluation Ongoing Care Urgent Care
Multisystem Evaluation Immunizations Cardiology
Injection Key Findings / Test Results Vascular
Educational Visit Pre-Procedure Information Pulmonary
Hospital Service Transfer Post-Procedure Information Oncology
Vital Signs Nephrology
Current Medications Neurology
Revision History Pathology
Special Instructions Gastroenterology
Advance Directives Vascular Wound Care
Discharge Activity Vascular Surgery
Final Pathology Diagnosis Cardiac Surgery
Included
©2017 MFMER | slide-12
Note Types Note Sections Service Groups Miscellaneous Chief Complaint Orthopedic
Test- Oriented Miscellaneous History of Present Illness Podiatry Dismissal Summary Family History Endocrinology Therapy System Reviews Emergency Medicine Emergency Medicine Hospital Admission Note Visit
Anticipated Problems and Interventions
Allergy
Hospital Admission Note Informed Consent Dermatology Emergency Medicine Visit Patient Education Sports Medicine
Physical Examination Spine Center Work Rehabilitation Plastic Surgery
Nursing Home Social Services Addiction
Excluded
Afzal et al: J Vasc Surg, 2017
©2017 MFMER | slide-13
REP PAD Cohort • 1569 patients
• 806 cases • 763 controls
• Average age: 71 years • 44% women • 90% white
2000 2005
ABI test date 21-day
2015 2017 2010
©2017 MFMER | slide-15
Performance of PAD-NLP algorithm vs. gold standard (ABI)
PPV 0.93
Sensitivity 0.70
NPV 0.80
Specificity 0.95
©2017 MFMER | slide-16
Temporality of PAD-NLP Algorithm • Compared temporal association between PAD-
NLP algorithm inception date (date on which NLP algorithm classified patient as PAD) with gold standard index date for each PAD patient
• For true positive cases, difference between NLP algorithm inception date and gold standard index date was measured in days
©2017 MFMER | slide-17
Temporality of PAD-NLP Algorithm
Before gold standard index date 329 cases (41%)
At gold standard index date 93 cases (12%)
After gold standard index date but within 21 day-window
141 cases (18%)
©2017 MFMER | slide-18
Temporality of PAD-NLP Algorithm
0
20
40
60
80
100
120
140
160
180
<-3000 -2999 to -2000
-1999 to -1000
-999 to -500 -499 to -100 -99 to -1 0 1 to 20
Number of patients
true positives
Number of days from ABI test
©2017 MFMER | slide-19
Discussion • PAD-NLP algorithm identified PAD cases from
clinical notes with high PPV • Prior to ABI test date in 41% of the cases from the
community • Delay in establishing PAD diagnosis despite
presence of PAD symptoms
©2017 MFMER | slide-20
Study Limitations • Data were retrieved from a single
academic medical center • Results may not be generalized to other
practice settings
©2017 MFMER | slide-21
Clinical Implications • Accurate and timely detection of actual or possible
PAD will: • Remind clinicians to order ABI testing confirming
diagnosis of PAD • Implement risk modification strategies in PAD
patients • PAD-NLP algorithm may be incorporated to clinical
CDS to identify PAD patients • Future studies will evaluate impact of CDS system on
outcomes in PAD patients
©2017 MFMER | slide-22
Conclusions • PAD-NLP surveillance algorithm enabled
early identification of PAD with high PPV • This may promote diagnosis earlier in the
course of PAD
©2017 MFMER | slide-23
Acknowledgements
• Grants • NHLBI: K01HL124045 • NHGRI: HG04599 and HG006379 • NIA: R01AG034676 • NIGMS: R01GM102283A1