Biosurveillance Evaluation of SNOMED CT’s Terminology (BEST Trial): Coverage of Chief Complaints Peter L. Elkin, MD, FACP, FACMI Professor of Medicine Mayo Clinic College of Medicine Vice-President, Translational and Biomedical Informatics Director, Center for Biomedical Informatics Vice-Chairman Department of Internal Medicine Mount Sinai School of Medicine Co- Authors: Steven H. Brown, MD, Andrew Balas, MD, Ph. D, Zelalem Temesgen, MD, Dietlind Wahner-Roedler, MD, David Froehling, MD, Mark Liebow, MD, Brett Trusko, Ph. D, Katie Skeen-Morris, MPH, S. Trent Rosenbloom, MD, Greg Poland, MD © Mount Sinai School of Medicine 2008
Overall Goal Data Information Knowledge INTELLIGENCE Public Health Response © Mount Sinai School of Medicine 2008
Outbreak One Outbreak Two Multi-Center Data Sharing and Interchange © Mount Sinai School of Medicine 2008
© Mount Sinai School of Medicine 2008
© Mount Sinai School of Medicine 2008
Intelligent Agents © Mount Sinai School of Medicine 2008
Level One Ontology HEALTH LEVEL 7 REFERENCE INFORMATION MODEL RIM_0100 released January 2001 reflects RIM changes through Harmonization on 11/17/2000 Enitites Acts (Services) Roles Message_control Role-role relationships Healthcare_finances Billboard produced by: Rochester Outdoor Advertising Participation Role Notary_public notary_county_cd : CE notary_state_cd : CE Place gps_txt : ST position_txt addr : AD directions_txt Health_chart 1 Individual_healthcare_practitioner fellowship_field_cd : CE graduate_school_nm : ON graduation_dttm : TS board_certified_ind : BL is_assessed_against 0. . * Health_chart_deficiency assessment_dttm : TS desc : ED level_cd : CV type_cd : CV Healthcare_facility licensed_bed_nbr : REAL mobile_ind : BL is_site_for Person disability_cd : CE ethnic_group_cd : CE race_cd : CE ambulatory_status_cd : CV birth_order_nbr : INT education_level_cd : CV living_arrangement_cd : CV marital_status_cd : CV religious_affiliation_cd : CV student_cd : CV credit_rating_cd : CV addr : SET<AD> special_accommodation_cd : SET<CV> 1 Role_relationship 0. . * type_cd : CC effective_tmr : IVL<TS> has_as_target id : SET<II> status_cd : CS responsibility_cd : SET<CE> position_nbr : LIST<INT> has_as_source qty : PQ 0. . * certificate_txt : ED Entity Manufactured_material expiration_dttm : TS lot_nbr : ST is_communicated_by 0. . * Person_Language 1 Access gauge_qty : PQ entry_site_cd : CD body_site_cd : CD Clinical_document_header availability_status_cd : CV change_reason_cd : CV completion_status_cd : CV confidentiality_status_cd : CV content_presentation_cd : CV document_creation_dttm : TS file_nm : ST last_edit_dttm : TS reporting_priority_cd : CE results_report_dttm : TS storage_status_cd : CV transcription_dttm : TS document_change_cd : CV version_nbr : INT version_dttm : TS Role Relationship Container capacty_qty : PQ height_qty : PQ diameter_qty : PQ barrier_delta_qty : PQ bottom_delta_qty : PQ separator_type_cd : CD cap_type_cd : CD Clinical_document Working_list ownership_level_cd Referral authorized_visits_qty : REAL desc : ED reason_txt : ED 1 is_utilized_during 0. . * utilizes 1 Supply qty : PQ Transportation Consent Procedure entry_site_cd : SET<CD> method_cd : SET<CV> body_site_cd : SET<CD> is_target_for 1 Act_relationship type_cd : CS inversion_ind : BL has_source sequence_nbr : INT 0. . * priority_nbr : INT pause_qty : PQ checkpoint_cd : CS split_cd : CS has_target join_cd : CS 0. . *negation_ind : BL conjunction_cd : CS is_authorized_by Diet energy_qty : PQ carbohydrate_qty : PQ Observation value : ANY derivation_expr : ST method_cd : SET<CV> body_site_cd : SET<CD> interpretation_cd : SET<CS> Act Relationship originates_in_context_of 1. . * Medication form_cd : CD route_cd : CD dose_qty : PQ strength_qty : PQ rate_qty : PQ dose_check_qty : PQ method_cd : SET<CV> body_site_cd : SET<CD> substitution_cd : CV Document_service completion_cd : CV set_id : II storage_cd : CV version_nbr : INT copy_dttm : TS origination_dttm : TS is_part_of Financial_act effective_tmr : IVL<TS> reason_cd : CE status_dttm Act_context level_cd Message_interaction is_communicated_as 0. . 1 Public_health_case detection_method_cd transmission_mode_cd disease_imported_cd 1 Patient_Provider 0. . * Device manufacturer_model_nm : ST last_calibration_dttm : TS software_nm : ST local_remote_control_state_cd : CE alert_level_cd : CE 1 Healthcare_benefit_product_policy Diagnostic_related_group_definition Patient_billing_account assignment_of_benefits_ind : BL base_rate_qty : MO benefit_product_desc : ED adjustment_cd : CV capital_reimbursement_qty : MO benefit_product_nm : ST certification_required_ind : BL cost_weight_qty : MO benefit_product_type_cd : CE current_unpaid_balance_qty : MO major_diagnostic_category_cd : CE benefits_coordination_ind : BL expected_insurance_plan_qty : REAL operating_reimbursement_qty : MO cob_priority_nbr : REAL expected_payment_source_cd : CV reimbursement_qty : MO combine_baby_bill_ind : BL notice_of_admission_dttm : TS standard_day_qty : PQ notice_of_admission_ind : BL group_benefit_ind : BL standard_total_charge_qty : MO patient_financial_class_cd : CV mail_claim_party_cd : CE trim_high_day_qty : PQ price_schedule_id : II release_information_cd : CE trim_low_day_qty : PQ report_of_eligibility_dttm : TS status_cd : CS coverage_type_cd : CE retention_ind : BL defines 1 agreement_type_cd : CE signature_on_file_dttm : TS policy_category_cd : CE special_program_cd : CV is_defined_by 0. . * access_protocol_desc : ED stoploss_limit_ind : BL suspend_charges_ind : BL Encounter_drg total_adjustment_qty : MO approval_ind : BL total_charge_qty : MO confidential_ind : BL total_payment_qty : MO cost_outlier_qty : MO separate_bill_ind : BL desc : ED bad_debt_recovery_qty : MO grouper_review_cd : CE Champus_coverage grouper_version_id : II bad_debt_transfer_qty : MO handicapped_program_cd : CE outlier_days_nbr : REAL outlier_reimbursement_qty : MO non_avail_cert_on_file_ind : BL outlier_type_cd : CV retirement_dttm : TS station_id : II manages 0. . * Resource_slot status_cd : CS time_slot : GTS 0. . * is_source_for 0. . * Schedule status_cd : CS slot_size_increment_qty is_managed_by specifies_ability_in Language_ability mode_cd : CV proficiency_level_cd : CV has_parts 0. . 1 Act id : SET<II> mood_cd : CS type_cd : CC txt : ED status_cd : CS activity_time : GTS critical_time : GTS confidentiality_cd : SET<CV> max_repeat_nmr : IVL<INT> interruptible_ind : BL priority_cd : SET<CV> orderable_ind : BL availability_dttm : TS Act 1 provides_context_for Patient_encounter discharge_disposition_cd : CV acuity_level_cd : CV birth_encounter_ind : BL status_reason_cd : CV classification_cd : CV encounter_classification_cd : CV practice_setting_cd : CV valuables_desc : ED pre_admit_test_ind : BL 1 source_cd : CV special_courtesies_cd : CV valuables_location_desc : ED effective_tmr uses Inpatient_encounter length_of_stay_qty : PQ Specimen body_site_cd : CE communicates_in is_specified_by 1 Non_Person_living_subject taxonomic_classification_cd : CE breed_cd : CE strain_txt : ED euthanasia_ind : BL production_class_cd : CE gender_status_cd : CE has Military_person Healthcare_provider military_branch_of_service_cd : CV specialty_cd : CV military_rank_nm : ST military_status_cd : CV has_an_assessment_of Living_subject birth_dttm : TS deceased_dttm : TS Material deceased_ind : BL administrative_gender_cd : CE Organization form_cd : CV organ_donor_ind : BL danger_cd : CE org_nm : SET<ON> multiple_birth_ind : BL standard_industry_class_cd : CE effective_tmr : IVL<TS> handling_cd : CE addr : SET<AD> Participation has_as_participant type_cd : CS tmr : IVL<TS> note_text : ED 0. . * signature_cd : CV for function_cd : CD awareness_cd : CV 0. . * signature_txt : ED encounter_accommodation_cd : CV status_cd : CS 0. . * Role participates_in is_played_by type_cd : cc is_source_of effective_tmr : IVL<TS> 0. . 1 is_target_for 1 addr : SET<AD> 1 1 telecom : SET<TEL> Entity 1 id : SET<II> plays_a_role type_cd : CC determiner_cd : CS importance_status_txt : ED has qty telecom : SET<TEL> 0. . * 1 desc status_cd : CS sends 1. . 1 shall_receive 1. . * Entity_name effective_tmr : IVL<TS> nm : EN purpose_cd : CV is_for Appointments & scheduling is_sited_at Practitioner_provider position_cd : CV primary_care_ind : BL Employee_Employer addr : SET<AD> hazard_exposure_txt : ED job_class_cd : CV job_title_nm : ST telecom : SET<TEL> protective_equipment_txt : ED salary_qty : MO salary_type_cd : CV status_cd : CS job_cd : CE Practitioner_Certifier board_certification_type_cd : CV certification_dttm : TS residency_field_cd : CE Unmapped_financial_classes (from RIM_Healthcare_finances) Outbreak tmr Encounter_facility_association effective_tmr : IVL<TS> is_used_by status_cd : CS transfer_reason_cd : CV 0. . * © Mount Sinai School of Medicine 2008 Preauthorization authorized_encounters_qty : REAL authorized_period_begin_tmr : IVL<TS> id : II issued_dttm : TS 0. . 1 requested_dttm : TS restriction_desc : ED authorizes status_cd : CS status_change_dttm : TS Insurance_certification has_coverage_affirmed_by 1 0. . * Guarantor_contract certification_duration_qty : PQ billing_hold_ind : BL effective_tmr : IVL<TS> affirms_insurance_coverage_for billing_media_cd : CE id : II charge_adjustment_cd : CE insurance_verification_dttm : TS contract_duration_cd : CE modification_dttm : TS contract_type_cd : CE non_concur_cd : CE effective_tmr : IVL<TS> non_concur_effective_dttm : TS interest_rate_nbr : REAL penalty_qty : MO periodic_payment_qty : MO report_of_eligibility_dttm : TS priority_ranking_cd : CV report_of_eligibility_ind : BL Billing_information_item condition_cd : CE occurrence_dttm : TS occurrence_span_cd : CE occurrence_span_from_dttm : TS occurrence_span_thru_dttm : TS quantity_nbr : REAL quantity_type_cd : CV value_amt value_cd : CE Healthcare_benefit_coverage_item service_category_cd : CV service_cd : CE service_modifier_cd : CE authorization_ind : BL network_ind : BL assertion_cd : CE covered_parties_cd : CE qty : REAL quantity_qualifier_cd : CE time_period_qualifier_cd : CE range_low_qty : PQ range_high_qty : PQ range_units_cd : CV eligibility_cd : CE policy_source_cd : CE eligibility_source_cd : CE copay_limit_ind : BL Financial_transaction extended_qty : MO fee_schedule_cd : CE insurance_qty : MO posting_dttm : TS qty : MO transaction_batch_id : II unit_qty : MO unit_cost_qty : MO
Level Two Ontology © Mount Sinai School of Medicine 2008
Level Three Ontology • Fully Encoded Health Record • Consistent with the Level One and Two Ontologies for Health • Compositional Expressions are assigned Automagically • Information is gathered through the usual documentation of patient care. • Example…………. . © Mount Sinai School of Medicine 2008
Placing a Stake in the Ground by Don Berwick, MD Clinicians and Researchers need a comparable mechanism for accessing Medical Record Data. Let’s place that stake in the ground where we think it Really Ought To Be!!! © Mount Sinai School of Medicine 2008
© Mount Sinai School of Medicine 2008
Record Entry Process Web Server Enterprise Java Bean Container Transcribed Record Processed Record to Storage EMR Repository Map Text To Terminology and Terminology Store Intelligent Query to Database Handle Query and Explode Matches Query Record Retrieval Process © Mount Sinai School of Medicine 2008 Record to Processor and Return Terminology Server Query to Processor and Return
Comparable Data • SNOMED-CT – Description Logic-Based Terminology – Compositional System – ~370, 000 Concepts – ~1, 000 Terms – LBI version adds 790, 000 Terms – Over 30, 000 Indices to the SNOMED-CT Terminology © Mount Sinai School of Medicine 2008
Compositional Systems © Mount Sinai School of Medicine 2008
“Acute myocardial “Heart attack, infarction of the Anterolateral cardiac anterolateral wall” wall, Acute” “AMI, Anterolateral Wall” “Acute MI of the Anterolateral Wall” MCVS Output SNOMED-CT © Mount Sinai School of Medicine 2008 - Myocardial infarction (disorder) [22298006] - [has Finding Site]. Entire myocardium of anterolateral region (body structure) [190762001] - [is Modified By]. Acute (qualifier value) [53737009]
Bird Flu © Mount Sinai School of Medicine 2008
"DOC_ID SEC_ID PROP_ID CONCEPT_CODE INEX_TYPE PTS" "2 52 1 23685000 7 1" "2 52 1 255302009 7 1" "2 52 1 79619009 7 1" "2 52 2 53059001 7 1" "2 52 2 237679004 7 1" "2 52 2 11092001 7 1" "2 52 3 49436004 7 1" "2 51 13 43364001 7 1" "2 51 14 258707000 7 1" "2 51 14 255260001 7 1" "2 51 14 102522009 7 1" "2 51 15 226630009 7 1" "2 51 15 51440002 7 1" "2 51 15 182334003 7 1" "2 51 15 43364001 7 1" "2 56 43 229799001 7 1" "2 56 43 258684004 7 1" © Mount Sinai School of Medicine 2008
Data Representation • Pseudo-Anonymized Records – Translation tables held locally in a secure fashion – Trackable across visits and episodes of care • No Text => Not Human Readable • The data exists without the story • Facts are recoverable – Patient • Document – Section • Sub-Section / Problem / Sentence • Compositional Expression / Concept © Mount Sinai School of Medicine 2008
BEST Trial: Chief Complaint Surveillance • Sample – 50, 000 Mayo Outpatient Clinical Records from Pediatrics, Adult General Medicine • 36, 097 Records had non-Null CCs – Only H&Ps and Limited Evaluations – All Visits from December through February (2003 to 2005) – 1035 Randomly selected chief complaints © Mount Sinai School of Medicine 2008
BEST Trial: Chief Complaint Surveillance • Method – Two Board Certified Internists (one Infectious Diseases Specialist and one General Internist) reviewed the mappings of the 1035 chief complaints – Disagreements were adjudicated by a Third Board Certified Internist © Mount Sinai School of Medicine 2008
BEST Trial Results: Chief Complaint Surveillance • Complete Match with one concept 35 TP FP TN FN C buttock I abscess D Abscess of buttock (disorder) [64576003] [K] © Mount Sinai School of Medicine 2008
BEST Trial: Results: Chief Complaint Surveillance • Match using a Compositional Expression 64 TP FP TN FN C Leg ulcers in both the lower E extremities. Ulcer of lower extremity (disorder) [95344007] [K] • - - [has Finding Site] . Both lower extremities (body structure) [4180000] [M] © Mount Sinai School of Medicine 2008
BEST Trial: Results: Chief Complaint Surveillance • More Complicated CE 128 TP FP TN FN PAR TIAL Rash right hand, left ankle (see HPI). -Eruption (disorder) [271807003] [K] - - [has Finding Site] Entire hand (body structure) [302539009] [M] - - [has Laterality] Right (qualifier value) [24028007] [L] . - [has Finding Site] - Entire ankle region (body structure) [361292008] [M] - [has Laterality] Left (qualifier value) [7771000] [L] © Mount Sinai School of Medicine 2008 .
False Positives (interesting) 3 1 2 T P F P T N F N C Mrs. Volkmann is a very pleasant 49 -year. E old female who presents today for followup abdominal pain consistent with presumed diverticulitis. - [AND] . Mrs -Volkmann's contracture (disorder) [111247001] [K] - [is Modified By] - . Very (qualifier value) [260358002] [M] . pleasant . Age 19 to 59 years (finding) [102522009] [K]. Female (finding) [248152002] [K] - [WHO] - [AND] - TODAY (substance) [101953001] [K] - [is Qualified By] . Present (qualifier value) [52101004] [Q] - [is Modified By] . Following (attribute) [255260001] [M] . Abdominal pain (finding) [21522001] [K] - [CONSISTENT WITH] - [PRESUMED] © Mount Sinai School of Medicine 2008 . Diverticulitis (disorder) [307496006] [K]
Results © Mount Sinai School of Medicine 2008
Results: MCVS Biosense NLP Chief Complaint SNOMED CT Engine • • • Sensitivity (Recall) - 98. 7% PPV (Precision) - 97. 4% NPV 89. 5% Specificity 81. 0% Pos LR 5. 181 NLR 0. 016 © Mount Sinai School of Medicine 2008
Results: MCVS Biosense NLP Clinical Problems SNOMED CT Engine • Sensitivity (Recall) - 99. 7% • PPV (Precision) - 99. 8% Peter L. Elkin, MD 1, Steven H. Brown, MD, Casey Husser, MD, Brent A. Bauer, MD, Dietlind Wahner-Roedler, MD, S. Trent Rosenbloom, MD, Ted Speroff, Ph. D; “An Evaluation of the Content Coverage of SNOMED-CT for Clinical Problem Lists“, Mayo Clin Proc. 2006 Jun; 81(6): 741 -8. © Mount Sinai School of Medicine 2008
Value of Chief Complaints for Biosurveillance (VCB) Trial • Case Controlled Study – 1455 Cases of PCR positive Influenza (A or B) – 1467 Controls who were tested for Influenza but who tested negative • Chief Complaints were reviewed by two independent clinicians and disagreements were adjudicated by a third clinician – Was there any manifestation of the flu? – Was there a good case definition for the flu? © Mount Sinai School of Medicine 2008
Case Definition © Mount Sinai School of Medicine 2008
CC Methods 1455 Influenza Cases 1467 Control Patients 1371 had a note from The day of or the Day before the sample 705 Cases had a record On the day of the Influenza swab or the Day prior to Obtaining the sample 1063 had non-null Chief Complaints 308 had blank or null Chief Complaints 683 had non-null Chief Complaints © Mount Sinai School of Medicine 2008 22 had blank or null Chief Complaints
CC Results – Automated Analysis • Chief Complaints containing any concept consistent with the flu in Influenza cases SNOMED CT Concept Fever -- 248425001 Cough -- 49727002 Diarrhea -- 62315008 Dyspnea – 267036007 Hemoptysis -- 66857006 Sore Throat -- 162397003 Number of Exploded Concepts 9 63 5 50 Number of Exploded Terms 28 169 30 194 Number of Records Containing the Concept 413 333 6 72 5 16 1 4 35 113 © Mount Sinai School of Medicine 2008
CC Results – Automated Analysis • Notes containing a reasonable case definition for the flu in Influenza cases cough / fever / sorethroat 34 Cough / sorethroat 22 dyspnea / sorethroat fever / sorethroat Cough / fever 1 34 162 Cough / hemoptysis 1 cough / diarrhea / fever 2 Cough / diarrhea / fever / sorethroat 1 dyspnea / cough / fever 4 dyspnea / cough 13 dyspnea / fever 9 © Mount Sinai School of Medicine 2008
CC Results – Automated Analysis • Rate of Identification of Flu Manifestations Patients with one flu concept 612 % One Flu Concept Reviewed 57. 60% % One Flu Concept Total 44. 60% Patients with more than one Flu Concept 283 % > 1 Flu Concept Reviewed 26. 60% % > 1 Flu Concept Total 20. 60% Patients with >2 Flu concepts 41. 00% % > 2 Flu Concepts Reviewed 3. 86% % > 2 Flu Concepts Total 2. 99% © Mount Sinai School of Medicine 2008
CC Results - Clinical Review • Rate of identification of Influenza manifestations by the Clinicians in true Influenza cases Good Case definition % of Cases Reviewed % of Cases Total At least one Flu Concept % At least one Flu Concept Reviewed % At least one Flu Concept Total © Mount Sinai School of Medicine 2008 280 26. 30 % 20. 40 % 724 68. 10 % 52. 80 %
CC Results – Case / Control Study Flu Status Text Diagnosis Not Flu Any Sign Flu Good Case Definition Flu Total Control 447 202 56 705 Case 647 444 280 1371 Total 1094 646 336 2076 © Mount Sinai School of Medicine 2008
CC Results – Case Control Study © Mount Sinai School of Medicine 2008
CC Results – Case / Control Study Pearson 1 Flu Chi-Square Concept Test At Least One Flu Concept Sensitivity 32. 3% Cases (p<0. 001) 52. 8 (49. 9, 20. 4 (18. 2, 55. 5) 22. 6) (p<0. 001) Sensitivity 28. 7% Controls 36. 5 © Mount Sinai School of Medicine 2008 Good Case Definition of the Flu 7. 9%
Case Predictive Model • Compare Sensitivity / Specificity of CC, HPI, PE and Impression for each case definition – Sensitivity – Specificity • Overall Case Model Identification © Mount Sinai School of Medicine 2008
Next Steps • Finding Case Definitions in the Clinical Record – Compare Whole Record Surveillance for the Influenza case / control study population – Avian Flu AH 5 N 1 – Anthrax – Radiation Exposure – Smallpox – Ricin Exposure • Building a regression model to identify the best approach to surveilling these case definitions from clinical records – Look at the impact of common miss-diagnoses asserted when records are near misses for one of the case definitions (Logistic Modeling) © Mount Sinai School of Medicine 2008
Clinical Note Surveillance • Influenza: Physicians around hospital and hospital ED get rapidly increasing number of patients with respiratory symptoms suggestive of a viral infection, but no increased prevalence of similar symptoms in surrounding hospitals. AH 5 N 1 is suspected due to association of patients with each other and “dead chickens”. All specimens are sent to state DOH ASAP for ID. State lab identifies AH 5 N 1. The followup once notification is disseminated from health department(s) to local providers, is similar to the presumptive diagnosis information transmission to public health BIS. A more robust method for collection of presumptive diagnoses in either scenario is to use standardized “problem” terms (using SNOMED) for selection of presumptive problems as part of routine operations of a CIS for physician order entry and for physician and nursing documentation. © Mount Sinai School of Medicine 2008
Avian Flu Case Definition Figure 1: Example case definition for Avian Flu. Please note that blue concepts are positive assertions, red are negative assertions and green concepts are uncertain assertions. Radiography of chest (procedure) [55587003] [K] is Modified By Revealed (qualifier value) [263857004] [M] has Procedure Site Interstitial space (body structure) [53742001] [M] is Modified By Infiltration (morphologic abnormality) [47351003] [M] Figure 2: Automated compositional expression generated by the second to last sentence by the Mayo Clinic’s Vocabulary Server. These structures are stored in the database and facilitate very specific surveillance of clinical record data. © Mount Sinai School of Medicine 2008
Anthrax Exposure: Figure: This is the history from a patient coded in SNOMED-CT. The bold terms are valid concepts in SNOMED-CT, the blue concepts are positive assertions, the red concepts are negative assertions and green concepts when present represent uncertain assertions. . For example, Nausea is a (finding) and has code [73879007] and Severe is of semantic type (severity modifier) and has code [24484000] and they are linked by the Has. Severity semantic operator. © Mount Sinai School of Medicine 2008
Acute Radiation Exposure © Mount Sinai School of Medicine 2008
Smallpox • Figure 7: This is the history from a patient coded in SNOMED-CT. The bold terms are valid concepts in SNOMED-CT, the blue concepts are positive assertions, the red concepts are negative assertions. This example includes common diagnoses which can be missed diagnosed when a case of small pox presents to a healthcare facility. © Mount Sinai School of Medicine 2008
Acute Ricin Exposure © Mount Sinai School of Medicine 2008
The Need for a Rapid Response © Mount Sinai School of Medicine 2008
BEST Trial - Biosurveillance: SNOMED CT based NLP • • • More Complete Data Fully Automated Approach Standards Based Parse Once => Use Many We believe that surveillance using case definitions provides more relevant and actionable data Public Health needs an accurate fully automated mechanism for Standards-based Biosurveillance and Situational Awareness. © Mount Sinai School of Medicine 2008
Скачать презентацию Biosurveillance Evaluation of SNOMED CT s Terminology BEST Trial
34f0086069a1692bc9cda89a4ec15967.ppt