Reporting of Critical Findings in Neuroradiology

Viertel, Valentina G.; Trotter, Stacey A.; Babiarz, Lukasz S.; Alluwaimi, Fatma; Nagy, Paul; Lewin, Jonathan S.; Yousem, David M.

doi:10.2214/ajr.12.9041

Cited by 12 publications

(10 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Whenever we discuss "labels" in this article, we are referring to these human-generated report annotations. In total, there were 55 labels included in the analysis, 20 of which were classified as critical labels on the basis of the criteria used at our hospital, which are derived from a previously published report (36). At our institution, reporting of acute ischemic stroke is required within 15 minutes of image acquisition, and reporting of intracranial hemorrhage, airway compromise, acute spinal cord compression, ruptured aneurysm, and markedly misplaced lines and tubes is mandated within 1 hour of image acquisition.…”

Section: Complexity Analysismentioning

confidence: 99%

See 1 more Smart Citation

Natural Language–based Machine Learning Models for the Annotation of Clinical Radiology Reports

et al. 2018

View full text Add to dashboard Cite

Purpose To compare different methods for generating features from radiology reports and to develop a method to automatically identify findings in these reports. Materials and Methods In this study, 96 303 head computed tomography (CT) reports were obtained. The linguistic complexity of these reports was compared with that of alternative corpora. Head CT reports were preprocessed, and machine-analyzable features were constructed by using bag-of-words (BOW), word embedding, and Latent Dirichlet allocation-based approaches. Ultimately, 1004 head CT reports were manually labeled for findings of interest by physicians, and a subset of these were deemed critical findings. Lasso logistic regression was used to train models for physician-assigned labels on 602 of 1004 head CT reports (60%) using the constructed features, and the performance of these models was validated on a held-out 402 of 1004 reports (40%). Models were scored by area under the receiver operating characteristic curve (AUC), and aggregate AUC statistics were reported for (a) all labels, (b) critical labels, and (c) the presence of any critical finding in a report. Sensitivity, specificity, accuracy, and F1 score were reported for the best performing model's (a) predictions of all labels and (b) identification of reports containing critical findings. Results The best-performing model (BOW with unigrams, bigrams, and trigrams plus average word embeddings vector) had a held-out AUC of 0.966 for identifying the presence of any critical head CT finding and an average 0.957 AUC across all head CT findings. Sensitivity and specificity for identifying the presence of any critical finding were 92.59% (175 of 189) and 89.67% (191 of 213), respectively. Average sensitivity and specificity across all findings were 90.25% (1898 of 2103) and 91.72% (18 351 of 20 007), respectively. Simpler BOW methods achieved results competitive with those of more sophisticated approaches, with an average AUC for presence of any critical finding of 0.951 for unigram BOW versus 0.966 for the best-performing model. The Yule I of the head CT corpus was 34, markedly lower than that of the Reuters corpus (at 103) or I2B2 discharge summaries (at 271), indicating lower linguistic complexity. Conclusion Automated methods can be used to identify findings in radiology reports. The success of this approach benefits from the standardized language of these reports. With this method, a large labeled corpus can be generated for applications such as deep learning. RSNA, 2018 Online supplemental material is available for this article.

show abstract

Section: Complexity Analysismentioning

confidence: 99%

“…two-dimensional visualizations in which the distance between points reflects their similarity in 400-dimensional space (37). clusters among learned word embeddings in the head CT corpus (36). We used t-distributed stochastic neighbor embedding (t-SNE) to generate (Fig 4).…”

Section: Modelingmentioning

confidence: 99%

Natural Language–based Machine Learning Models for the Annotation of Clinical Radiology Reports

et al. 2018

View full text Add to dashboard Cite

show abstract

“…A critical test result is defined as a finding that requires direct notification of a member of the patient's care team, including findings that could result in mortality or significant morbidity if appropriate steps are not undertaken [7]. In the United States, TJC trusts each hospital's leadership to approve and validate their own definition of critical test results [7,8]. In the Philippine General Hospital, no vetted critical radiology findings list or protocol for neuroradiology or cranial CT findings exists.…”

Section: Introductionmentioning

confidence: 99%

Critical Findings Reporting in Cranial CT Scans of Patients in a Tertiary Government Hospital Setting:A Pilot Study

Licaros¹,

Cañal²

2020

Acta Medica

View full text Add to dashboard Cite

BACKGROUND AND PURPOSE: Effective communication among health care providers within an institution is vital for ensuring patient safety. A radiologist’s competence in identifying critical findings is just as important as being able to relay this information to the primary provider in a timely manner so that appropriate interventions can be initiated in life-threatening emergencies. This study aims to obtain baseline data regarding effective communication of critical findings between radiology and the primary health care providers. MATERIALS AND METHODS: Based on the Massachusetts’s Coalition for Prevention of Medical Errors safe practice recommendations, a communication protocol was developed. Certain selected cranial critical findings (new hemorrhage, new herniation, acute brain edema, acute stroke and findings suggestive of meningitis or abscess) were used in the critical findings protocol. Total communication time (TCT) in the relay of critical findings included time from CT scan order to scan (OS time), time from scan to interpretation (SI) and time from interpretation to relay (IR). All these times were recorded and compared to the standard 60 minutes using percent compliance, mean time, median time and standard deviation. RESULTS: Seventy-nine (79) critical findings were recorded and relayed in a 30-day period. There was 100% (79/79) compliance with the relay of critical findings to the primary care team and 97.5% (77/79) success rate in direct communication of critical findings within 60 minutes of identification. The mean OS time was 155.09 (±134.43) minutes, mean SI time was 46.54 (±44.01) minutes, and mean IR time was 9.13 (±12.04) minutes. The average time elapsed from CT study acquisition to direct notification was 57.0 (± 45.80) minutes. CONCLUSION: Effective communication of critical findings using a protocol adopted from set standards of safe practice recommendations is achievable in our institution. The proposed protocol exhibited compliance to and performed well against established benchmarks. The timely identification and delivery of critical findings to the primary care team is central to patient management and should be practiced in our setting. Keywords: critical finding, effective communication, cranial CT, standards

show abstract

“…1 There have subsequently been protocols created to reach this goal, including those for the following: 1) defining a critical finding (CF), 2) communicating the critical finding to the physician, and 3) indicating what makes a report "timely." [2][3][4][5][6] TJC has never defined a CF, stating only "each organization can define for itself the circumstances under which a test result is considered 'critical.'" 1 The American College of Radiology identifies findings that require nonroutine communication as those "that suggest a need for immediate or urgent intervention."…”

mentioning

confidence: 99%

“…3,7 Another publication studied the existing list of critical findings in neuroradiology and revised that list to include frequent findings that were not previously identified in an effort to be more inclusive of lesions that warrant immediate communication. 4 Others have suggested that careful review of performance on an ongoing basis with regard to communication of radiologic critical findings can result in significant improvement in adherence to institutional guidelines.…”

mentioning

confidence: 99%

Critical Findings: Timing of Notification in Neuroradiology

Honig

Babiarz

et al. 2014

American Journal of Neuroradiology

View full text Add to dashboard Cite

show abstract

Reporting of Critical Findings in Neuroradiology

Abstract: Compliance with reporting and communicating critical findings must be monitored. Calling clinicians to report noncritical findings may result in unnecessary interruptions in work flow for radiologists and referring health care providers.

Cited by 12 publications

References 7 publications

Natural Language–based Machine Learning Models for the Annotation of Clinical Radiology Reports

Natural Language–based Machine Learning Models for the Annotation of Clinical Radiology Reports

Critical Findings Reporting in Cranial CT Scans of Patients in a Tertiary Government Hospital Setting:A Pilot Study

Critical Findings: Timing of Notification in Neuroradiology

Contact Info

Product

Resources

About