Developing and validating natural language processing algorithms for radiology reports compared to ICD-10 codes for identifying venous thromboembolism in hospitalized medical patients

Verma, Amol A.; Masoom, Hassan; Pou-Prom, Chloé; Shin, Saeha; Guerzhoy, Michael; Fralick, Michael; Mamdani, Muhammad; Razak, Fahad

doi:10.1016/j.thromres.2021.11.020

Cited by 20 publications

(21 citation statements)

References 32 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Again, they report highly specific results but limited sensitivity, in line with our findings (74.83% (95% CI, 67.01 to 81.62) and 75.24% (95% CI, 65.86 to 83.14) for PE and DVT, respectively) 21. Verma et al report using natural language processing algorithms for digital interrogation of radiology reports in a large cohort of hospitalised medical patients to identify VTE outcomes 19. The authors conclude a high level of accuracy, reporting sensitivities of 94%/91% and PPVs of 90%/89% for DVT and PE, respectively.…”

Section: Discussionsupporting

confidence: 87%

“… 21 Verma et al report using natural language processing algorithms for digital interrogation of radiology reports in a large cohort of hospitalised medical patients to identify VTE outcomes. 19 The authors conclude a high level of accuracy, reporting sensitivities of 94%/91% and PPVs of 90%/89% for DVT and PE, respectively. Finally, Klil-Drori et al have recently validated an algorithm for confirmation of suspected PE, combining emergency department diagnosis coding, imaging coding and dispensed prescription or hospital treatment.…”

Section: Discussionmentioning

confidence: 85%

“…These two factors are the most important contributors to poor internal validity of efficient data methods, reflected in several studies across different countries. 19 23 24 In particular, Fang et al highlight the poor performance of outpatient coding to predict VTE in a separate cohort of 4642 adult patients. 25 Finally, coding teams may fail to document subsequent HAT (false negative) following index admission with alternative pathology (such as pneumonia) and prior diagnosis of VTE can often be coded during repeat hospital attendance, mistaken for HAT (false positive).…”

Section: Discussionmentioning

confidence: 99%

“…PPV values for DVT codes ranged from 31% to 97%. More recently, a cross-sectional North American study compared ICD-10 codes for VTE in hospitalised medical patients to a ‘gold standard’ manual review of clinical data in 4000 patients 19. The authors report a sensitivity of 63% for any DVT and a sensitivity of 83% for PE, implying further discrepancy between types of VTE.…”

Section: Discussionmentioning

confidence: 99%

See 3 more Smart Citations

Accuracy of efficient data methods to determine the incidence of hospital-acquired thrombosis and major bleeding in medical and surgical inpatients: a multicentre observational cohort study in four UK hospitals

et al. 2023

View full text Add to dashboard Cite

ObjectivesWe evaluated the accuracy of using routine health service data to identify hospital-acquired thrombosis (HAT) and major bleeding events (MBE) compared with a reference standard of case note review.DesignA multicentre observational cohort study.SettingFour acute hospitals in the UK.ParticipantsA consecutive unselective cohort of general medical and surgical patients requiring hospitalisation for a period of >24 hours during the calendar year 2021. We excluded paediatric, obstetric and critical care patients due to differential risk profiles.InterventionsWe compared preidentified sources of routinely collected information (using hospital coding data and local contractually mandated thrombosis datasets) to data extracted from case notes using a predesigned workflow methodology.Primary and secondary outcome measuresWe defined HAT as objectively confirmed venous thromboembolism occurring during hospital stay or within 90 days of discharge and MBE as per international consensus.ResultsWe were able to source all necessary routinely collected outcome data for 87% of 2008 case episodes reviewed. The sensitivity of hospital coding data (International Classification of Diseases 10th Revision, ICD-10) for the diagnosis of HAT and MBE was 62% (95% CI, 54 to 69) and 38% (95% CI, 27 to 50), respectively. Sensitivity improved to 81% (95% CI, 75 to 87) when using local thrombosis data sets.ConclusionsUsing routinely collected data appeared to miss a substantial proportion of outcome events, when compared with case note review. Our study suggests that currently available routine data collection methods in the UK are inadequate to support efficient study designs in venous thromboembolism research.Trial registration numberNIHR127454.

show abstract

Section: Discussionsupporting

confidence: 87%

Section: Discussionmentioning

confidence: 85%

Section: Discussionmentioning

confidence: 99%

Section: Discussionmentioning

confidence: 99%

See 2 more Smart Citations

Accuracy of efficient data methods to determine the incidence of hospital-acquired thrombosis and major bleeding in medical and surgical inpatients: a multicentre observational cohort study in four UK hospitals

et al. 2023

View full text Add to dashboard Cite

show abstract

“…The positive predictive value of ICD coding for VTE has previously been validated. [14][15][16][17][18][19][20][21][22] thrombosis), and intra-thoracic thrombosis (superior vena cava, brachiocephalic, other unspecified thoracic veins). In instances of multiple entries, the initial occurrence of VTE was utilized.…”

Section: Study Design and Patient Populationmentioning

confidence: 99%

The prevalence of thrombocytopenia in patients with acute cancer-associated thrombosis

2023

View full text Add to dashboard Cite

Introduction: Venous thromboembolism (VTE) and thrombocytopenia are frequently encountered complications in cancer patients. While there are several studies evaluating the safety and efficacy of anticoagulation regimens in patients with cancer associated thrombosis (CAT) with thrombocytopenia, there is a paucity of data assessing the scope of the concurrent diagnoses. This study evaluates the prevalence of thrombocytopenia among patients with acute CAT. Methods: A retrospective cohort analysis of adult cancer patients was conducted at Beth Israel Deaconess Medical Center between 2010 and 2021 with CAT (acute VTE within 6 months after new diagnosis of malignancy). VTE included acute deep vein thrombosis (DVT), pulmonary embolism (PE), abdominal or thoracic venous thrombosis, and cerebral sinus thrombosis (CST). The lowest platelet count within 2 weeks of (before or after) the index VTE event was identified to assess the frequency and grade of concurrent thrombocytopenia. Results: We identified 3,635 patients with CAT (79.8% solid tumors, 17.8% hematologic malignancies, and 2.4% multiple concurrent cancer diagnoses). Thrombocytopenia (defined as platelet count <100,000/µL) occurred in 22% (95% CI 21%-24%) of patients with CAT with solid tumors diagnoses and 47% (95% CI 43%-51%) of patients with CAT and hematologic malignancies. Severe thrombocytopenia (platelet count <50,000/µL) occurred in 7% (95% CI 6%-8%) of patients with solid tumors and 30% (95% CI 27%-34%) of patients with hematologic malignancies. Conclusions: Concurrent diagnoses of CAT and thrombocytopenia are very common, especially among patients with hematologic malignancies.

show abstract

Development and Validation of a Natural Language Processing Model to Identify Low-Risk Pulmonary Embolism in Real Time to Facilitate Safe Outpatient Management

Amin,

Weissler,

Ratliff

et al. 2024

Annals of Emergency Medicine

View full text Add to dashboard Cite

Developing and validating natural language processing algorithms for radiology reports compared to ICD-10 codes for identifying venous thromboembolism in hospitalized medical patients

Cited by 20 publications

References 32 publications

Accuracy of efficient data methods to determine the incidence of hospital-acquired thrombosis and major bleeding in medical and surgical inpatients: a multicentre observational cohort study in four UK hospitals

Accuracy of efficient data methods to determine the incidence of hospital-acquired thrombosis and major bleeding in medical and surgical inpatients: a multicentre observational cohort study in four UK hospitals

The prevalence of thrombocytopenia in patients with acute cancer-associated thrombosis

Development and Validation of a Natural Language Processing Model to Identify Low-Risk Pulmonary Embolism in Real Time to Facilitate Safe Outpatient Management

Contact Info

Product

Resources

About