Developing VA GDx: An Informatics Platform to Capture and Integrate Genetic Diagnostic Testing Data into the VA Electronic Medical Record Scott L. DuVall Jun 27, 2014 1
Julie Lynch Vickie Venne Dawn Provenzale Dan Berlowitz Michael Kelley Laurence Meyer Michael Icardi Bedford VA Team VINCI Team VA Genetic Counselors 2
Outline Department of Veterans Affairs (VA) Genetic Diagnostic Testing (GDx) The importance of GDx GDx used in practice GDx workflow Motivation to incorporate GDx in EMR The building of VA GDx Genotype and Phenotype from EMR text 3
Genetic Diagnostic Tests (GDx) Identification of Germ line mutations Somatic mutations Using PCR FISH DNA sequencing 5
Increasing Number of GDx November 2012: 2,878 genetic tests August 2013: 10,000 genetic tests June 2014: 19,057 genetic tests 6
Importance of GDx Specific diagnosis Accurate prognosis Proper treatment strategy 7
Importance of GDx Reducing Disparity Prostate cancer surveillance guidelines Lung cancer EGFR testing Reducing Variations in Care 8
Value Driven Care Importance of GDx 9
Importance of GDx Patients are unnecessarily treated with toxic or ineffective medications. Patients are not treated with targeted medications that improve survival. Patients are treated with molecularly targeted medications that are ineffective and expensive. VA spends too much money on drugs or genetic tests that provide no benefit or clinically useful information and may even harm patients. 10
GDx Used in Practice (now) Tier I / Level I / Green FDA label requires test to inform drug or dose Clinical practice guidelines based on systematic review supporting testing (CMS covers testing) Breast Ovarian Colorectal Gastric Leukemia Lymphoma Lung Melanoma Cardiovascular Infectious Other 11
GDx Used in Practice (soon) Image from Institute of Medicine 12
GDx Used in Practice (soon) 13
GDx Workflow 14
Motivation for VA GDx VA directive for cancer care VA policies on specific GDx The greatest obstacle to the adoption of personalized approaches such as genomic testing is the lack of adequately designed studies assessing their clinical utility. 15
Motivation for VA GDx informatics [is] the bottleneck in translating genomic advances into clinical care The expertise required includes: clinical, genomics, business logic, informatics 16
VA GDx $ 17
VA GDx More than 110 different GDx test types GDx testing for more than 70,000 patients ~50% Cancers ~12% HCV ~12% cardiac disease ~12% coagulopathy disorders ~12% heritable conditions ~3% pharmacogenomic tests 18
Genotype and Phenotype DISCHARGE SUMMARY ALLERGIES: Tylenol, Advil, Nafcillin. MEDICATIONS ON ADMISSION: diazepam 25 mg daily, aspirin 325 mg daily, Lasix 40 mg daily, Clariton extended release 1 pill po prn. HOSPITAL COURSE: The patient was admitted for a brain hemorrhage and was placed on Lasix. The patient became stable and was transferred to ICU for continued monitoring. While in the hospital, she acquired MRSA and was placed on antibiotics including vancomycin and penicillin. Patient should continue aspirin daily and an oral dose of diazepam 25 mg for a total of 10 days. Structured and coded data Fasting Glucose 135 mg/dl Pulse 60 bpm Diagnoses 250.0 Diabetes 274.0 Gout 172 Melanoma Unstructured narrative data
CC F/U Depression The patient indicates that his symptoms have improved significantly, but not as much as he expected. He is still sleeping a lot (about 12 hours per day) and finds it hard to concentrate on looking for work. He denies suicidal ideation. His PHQ-9 score is 16 today. The patient has a history of binge eating episodes. He is an emotional eater and often feels out of control, but continues to eat after job search disappointments. He often binges at night and has done this 3-4 times per week for the past several years. Professional Diagnosis: Axis I: Major Depression, partial response to meds Binge eating disorder Axis II: Deferred Axis III: Hyperlipidemia, obesity synthetic medical record data 20
NLP Pipeline pre-processing Sentences Words and Tokens Parts of Speech Phrases Concepts re-usable modules post-processing Context Inference project-specific modules 21
Grammatical Parsing Neoplasm in the body of the pancreas. There is slight enlargement of the liver. The kidneys are unremarkable and the adrenal glands are within normal limits. This example shows how the system processes clinical notes and maps text to RadLex terms.
Grammatical Parsing 2 Neoplasm in the body of the pancreas. There is slight enlargement of the liver. The kidneys are unremarkable and the adrenal glands are within normal limits. 3 First, the text is split in sentences. This document would contain three sentences. 1
Grammatical Parsing Neoplasm in the body of the pancreas. Each sentence in the document can be dealt with separately. This provides context for syntactical parsing.
Grammatical Parsing Neoplasm in the body of the pancreas. The sentences are then tokenized. This is the process of splitting the text into words and punctuation.
Grammatical Parsing noun Neoplasm in the body of the pancreas. preposition determiner noun preposition determiner noun period Part-of-speech tags are then assigned to tokens. Most concepts that can be mapped to RadLex are nouns and adjectives.
Grammatical Parsing noun noun noun Neoplasm in the body of the pancreas. noun phrases Neoplasm in the body of the pancreas. Neoplasm in the body of the pancreas. Chunking provides the phrases that will be searched for in RadLex. Individual nouns and longer phrases are both searched.
Grammatical Parsing Neoplasm pancreas neoplasm: RID3957 pancreas: RID170 body of the pancreas body body: RID5950 body of pancreas: RID174 Neoplasm in the body of the pancreas no match Each of these chunks is then mapped to the closest matching RadLex term or is left unassigned when no RadLex term matches.
Grammatical Parsing neoplasm: RID3957 Neoplasm in the body of the pancreas. There is slight enlargement of the liver. The kidneys are unremarkable and the adrenal glands are within normal limits. kidney: RID205 body of pancreas: RID174 enlarged: RID5791 liver: RID58 adrenal gland: RID88 After mapping, the document text is not changed, but RadLex terms are associated as metadata.
Patterns and Rules Her vital signs were temperature of 100.8, heart rate 96, blood pressure 140/80, respirations 20. The patient had a sodium of 133, potassium 5.1, chloride 102, bicarbonate 11.4, BUN and creatinine 23 and 2.0.
Patterns and Rules Her vital signs were <vital sign> of <#>, <vital sign> <#>, <vital sign> <#>/<#>, <vital sign> <#>. The patient had a <lab test> of <#>, <lab test> <#>, <lab test> <#>, <lab test> <#>, <lab test> and <lab test> <#> and <#>.
Patient Phenotype synthetic medical record data 32
Patient Phenotype synthetic medical record data 33
Patient Phenotype synthetic medical record data 34
Patient Phenotype synthetic medical record data 35
Patient Phenotype synthetic medical record data 36
Patient Genotype 37
Disease Characteristics Treated Hep C Patients 22,249 Genotype Lab 17,540 (79%) Missing Genotype Lab 4,709 (21%) DuVall SL, LaFleur J et al. ICPE 2012
Disease Characteristics Genotype Lab Available Genotype Treated Hep C Patients 17,540 (79%) 21,632 (97%) 22,249 Missing Genotype Lab Available Clinic Notes Genotype in Notes 4,709 (21%) More than 1 mention 94% More than 10 mentions 50% More than 20 mentions 22% Missing Genotype 617 (3%) DuVall SL, LaFleur J et al. ICPE 2012
Questions? Scott DuVall scott.duvall@utah.edu 40