- Email: [email protected]
Discrepancies in Reporting the Vertebral Level of Abnormality in MR of the Spine
Discrepancies in Reporting the Vertebral Level of Abnormality in MR of the Spine Minal Jagtiani Sangwaiya, MBBS, MRCS, MD, Shyla Saini, Rajiv Gupta, MD, Albert J. Yoo, MD, Markus Stout, MBA, Keith Dreyer, MD, PhD, DO, Mannudeep Kalra, MD
Purpose: The aim of this study was to assess discrepancies in the spinal levels of abnormalities stated in the findings or impression (or both) sections of radiology reports of magnetic resonance (MR) imaging. Materials and Methods: Radiology reports from January 2006 through December 2007 (n ⫽ 2,097,966) were analyzed using an online radiology report search engine. Reports were searched for presence of the key words MR spine and addendum. The addended reports were then manually assessed for any discrepancies in the reported spinal levels between the body and impression sections; the addenda corrected these errors (identified errors). In addition, all reports with the search term MR spine from January 2006 (n ⫽ 1,183) and January 2007 (n ⫽ 1,354) were assessed manually to recognize unidentified errors in spinal locations of reported pathology. Two neuroradiologists independently graded the clinical significance of errors on a 5-point scale (1 ⫽ definitely not significant, 5 ⫽ definitely significant). Results: Of the 11,427 spinal MR reports analyzed in 2006, 7 had identified errors in the sites (levels of the spine) of the lesions. In 2007 (n ⫽ 11,785 spinal MR reports), 4 reports were detected with identified errors in spinal levels. In January 2006 and January 2007, 8 and 12 reports, respectively, had unidentified erroneous vertebral levels. Errors were related to discrepant vertebral regions (eg, cervical vs thoracic) in 16% of cases (5 of 31), the wrong number of vertebrae (eg, L2 instead of L3) in 68% of cases (21 of 31), and both in 16% of cases (5 of 31). The average time taken to issue an addendum was 5 ⫾ 7 days in 2006 and 11 ⫾ 13 days in 2007. Fifteen reports (48%) scored ⬍3 on the scale of clinical significance, 1 report scored 3, and 15 scored ⬎3. Conclusions: Errors in lesion level on spinal MR do occur in radiology reports. The number of unidentified errors is substantially higher than that of identified errors. Care should be taken before signing off on radiology reports to identify erroneous mentions of the vertebral levels of abnormalities. Key Words: Radiology reporting, quality, errors, spinal MR J Am Coll Radiol 2009;6:715-720. Copyright © 2009 American College of Radiology
INTRODUCTION The National Quality Forum, a US public and private entity, issued results in 1999 indicating that between 44,000 and 98,000 Americans die each year from preventable medical errors in hospitals alone [1]. This issue was revisited again in 2008, when a list of 28 “never events” (errors that should not occur in medical practice) was proposed. These never events included surgery at the wrong Department of Radiology, Massachusetts General Hospital, Boston, Massachusetts. Corresponding author and reprints: Mannudeep K. Kalra, Massachusetts General Hospital, Department of Radiology, 25 New Chardon Street, Suite 400, Boston, MA 02114; e-mail: [email protected]. © 2009 American College of Radiology 0091-2182/09/$36.00 ● DOI 10.1016/j.jacr.2009.06.002
site, surgery performed on the wrong patient, the wrong surgical procedure on the right patient, and patient death or serious disability associated with a medication error [2]. Various factors contribute to wrong-site surgeries, including the failure of surgeons to adopt guidelines and suggestions by authorities. In 2007, Jhawar et al [3] described the frequently reported factors for wrong-sided cranial surgery. These included emergency operating conditions, fatigue, and time pressures to begin or complete procedures. To avoid wrong-site procedures, the Joint Commission [4] recommended that every patient undergo a preprocedural verification process to ensure that all relevant documents, related information, and equipment are available before the start of a procedure. The site of the 715
716 Journal of the American College of Radiology/ Vol. 6 No. 10 October 2009
procedure should be marked by a licensed independent practitioner. Also, to conduct a final assessment, surgeons and interventionalists should conduct a “time out” before beginning a procedure. This is to ensure that the correct patient, site, positioning, and procedure are identified. To the best of our knowledge, no prior studies have assessed the frequency of errors in vertebral level in radiology reports. The purpose of our study was to examine the frequency and severity of discrepancies in the spinal levels of abnormalities stated in the findings and impression (or both) sections of spinal magnetic resonance (MR) imaging reports at a tertiary care institution. MATERIALS AND METHODS This cross-sectional retrospective study was approved by the human research committee of our institution. The requirement for informed consent from study subjects was waived. This study was compliant with the Health Insurance Portability and Accountability Act. To protect the privacy of the reporting radiologists, our human research committee did not permit us to document the names of the radiologists who dictated the reports assessed in our study. Radiology Reporting According to ACR guidelines, the communication of diagnostic imaging findings in radiology reports should include two components or sections [5]; the “body” section should describe the procedures and materials, findings, potential limitations, clinical issues, and prior comparison studies, and the “impression” section should present a succinct summary of the main findings in the study performed and a conclusion or diagnosis. All radiology reports at our institution are reported using the ACR guidelines on a voice recognition dictation system (Nuance Communications Inc, Burlington, Massachusetts). After dictation, reports are electronically verified and signed off by the radiologists and then archived in the radiology information system (RIS) and other hospital electronic medical records. If an error is recognized in the original dictation after this stage, a radiologist must dictate an “addendum” to the original report, which is automatically integrated with the original radiology report in the RIS as well as all other hospital records of the patient. This workflow holds true for all radiology reports in our department. Each addended radiology report has an additional section called “addendum,” under which further explanations, revisions, and corrections are described. From the RIS, the radiology reports (and any subsequently issued addenda) are transferred to an electronic search engine, Folio VIEWS 4 (NextPage Inc, Draper, Utah), which has the capacity to perform Boolean searches with keywords, phrases, or expressions on all reports for each year and month. The
archived data date back to 1994 and are updated once every 3 months. Radiology Reports With Identified Errors We queried all radiology reports from January 2006 through December 2007 (n ⫽ 2,097,966) using the radiology search engine with the key word MR spine. A total of 23,212 spinal MR reports were found (n ⫽ 11,427 in 2006, 6,099 in female and 5,328 in male patients; n ⫽ 11,785 in 2007, 6,366 in female and 5,419 in male patients). On these spinal MR reports, we applied the key word addendum, to identify spinal MR reports with addenda (n ⫽ 445 in 2006, 252 in female and 193 in male patients; n ⫽ 340 in 2007, 202 in female and 138 in male patients). The addended spinal MR reports were manually searched for discrepancies in the intervertebral levels of abnormalities described in the body and impression sections. Corresponding MR images for these studies were reviewed to confirm the recognized errors (Figure 1). Radiology Reports With Unidentified Errors In this step, all spinal MR reports from Janaury 2006 (n ⫽ 828, 501 in female and 327 in male patients) and from January 2007 (n ⫽ 952, 429 in female and 523 in male patients) were identified using the key word MR spine. These reports included reports with or without identified errors of vertebral level of abnormality. All of these reports were manually assessed to determine the unidentified discrepancies in vertebral levels of abnormalities between the body and impression sections (Figure 1). Analysis of Site (Vertebral Level) Errors For each report with an identified discrepancy in the vertebral level of an abnormality between the body and impression sections, we recorded the stated vertebral level in reports with errors, correct vertebral levels in the addenda or MR images, and the average time period in days between initial signoff and the dictation of an ad-
Fig 1. Flowchart showing the search for erroneous radiology reports. MR ⫽ magnetic resonance; MRI ⫽ magnetic resonance imaging.
Sangwaiya et al/Discrepancies in Reporting Spinal MR Abnormalities 717
Fig 2. Flowchart depicting the incidence of errors (identified and unidentified) in the studied cohort. MR ⫽ magnetic resonance.
ments between the two readers. A score of 1 corresponded to definitely no clinical significance (eg, mild degenerative disease of the spine reported at different levels of the spine with a discrepancy between the body and impression sections of the report). On the other extreme, a score of 5 meant definite clinical significance (eg, a discrepancy between the body and impression sections in reporting the level of a spinal tumor). If the radiologist was uncertain about the clinical significance of a particular error, it was given a score of 3. Consensus agreement between the two radiologists was taken as final. The incidence of errors at each level of the spine was studied to determine the most frequent level of error, if any. RESULTS
dendum. The electronic medical record of each patient with an erroneous vertebral level in the radiology report was assessed to determine if the patient underwent a subsequent spinal procedure or surgery. We also evaluated the RIS and electronic medical records to determine if the unidentified errors had been corrected in subsequent radiology reports or in the medical records of the patients. Finally, two board-certified and experienced neuroradiologists independently graded each report with identified and unidentified errors of vertebral levels for the clinical significance of the errors. This was graded on a 5-point scale (1 ⫽ definitely no clinical significance, 2 ⫽ probably no clinical significance, 3 ⫽ uncertain, 4 ⫽ probable clinical significance, 5 ⫽ definite clinical significance). Consensus decisions were used for any disagree-
Radiology Reports With Identified Discrepancies Of the 23,212 spinal MR studies performed from January 1, 2006, through December 31, 2007, 11 reports (0.05%) had identified discrepancies in the vertebral levels of abnormalities between the body and impression sections (Figure 2). In 2006, there were 7 of 11,427 (4 in women, 3 in men; mean age, 61 years; age range, 34-77 years) reports with identified discrepancies in vertebral levels of the lesion. In 2007, there were 4 of 11,785 (3 in women, 1 in a man; mean age, 50 years; age range, 26-67 years) reports with identified discrepancies in the vertebral levels of abnormalities in the body and impression sections (Table 1). The average time taken to issue an
Table 1. Distribution of identified and corrected errors pertaining to the levels of vertebral abnormalities in spinal magnetic resonance reports (2006-2007) Level of Abnormality Level of Abnormality Mentioned in the Mentioned in the Impression of True Level of Genderⴱ Diagnosis of Error Body of the Report the Report the Abnormality M Spinal cord volume loss T5/T6 Cervical (C5/C6) T5/T6 F Disc herniation C4/C5 Not mentioned C5/C6 F Bony destruction T3 L3 T3 F Endplate fracture C6 C6 T6 F Progressed degenerative T7/T8 T8/T9 T7/T8 disease M Postoperative laminectomies L3, L4, L5 L3, L4, L5 T3, T4, T5 M Conus medullaris infarct and Cervical Cervical Thoracic multilevel degenerative changes M Epidural abscess C6/C7 C6/C7 C5/C6 F Disc protrusion C6/C7 C7/T1 C6/C7 F Metastatic soft tissue mass L5 T1 L5 F Fractures and intervertebral C6/C7 C6/C7 C7/T1 disc injury ⴱSeven female patients, 4 male patients.
718 Journal of the American College of Radiology/ Vol. 6 No. 10 October 2009
addendum for correcting these discrepancies was 5 ⫾ 1 days in 2006 and 14 ⫾ 13 days in 2007.
Types and Significance of Errors in Vertebral Levels of Abnormalities
Radiology Reports With Unidentified Errors
Errors were related to discrepant vertebral regions (eg, cervical vs thoracic) in 16% of cases (5 of 31), the wrong number of vertebrae (eg, L2 instead of L3) in 68% of cases (21 of 31), and both in 16% of cases (5 of 31). According to the first radiologist (R.G.), 17 of 31 discrepant reports were probably or entirely clinically insignificant, and the remaining 14 discrepancies were probably or definitely clinically significant. The second radiologist (A.J.Y.) graded 12 of 31 discrepant reports as probably or entirely insignificant, 1 as uncertain, and 18 as probably or definitely clinical significant to the patients’ management. There were differences in opinions of the two radiologists on the clinical significance in 9 of 31 reports. After consensus readouts, 3 discrepant reports were deemed clinically significant, and 5 were deemed clinically insignificant. The remaining discrepant report was graded as of uncertain clinical significance.
In January 2006 and January 2007, unidentified discrepancies in the levels of vertebral abnormalities in spinal MR reports were noted in 8 of 828 reports (6 in women, 2 in men; mean age, 52 ⫾ 12 years; age range, 38-77 years) and 12 of 952 reports (8 in women, 4 in men; mean age, 52 ⫾ 17 years; age range, 31-76 years), respectively (Table 2). Thus, in January 2006 and January 2007, there were of 1.1% (20 of 1,780) unidentified errors in spinal MR reports (Figure 2). Fortunately, none of the patients with unidentified errors had subsequent procedures or surgeries on the wrong vertebral levels. However, one patient did undergo an invasive procedure at the correct level, subsequent to imaging interpretation with an unidentified error. Also, in 4 of 20 reports with discrepant vertebral levels of abnormalities, the discrepancies were identified and corrected in subsequent reports or medical notes. The remaining patients with unidentified discrepancies did not have any addenda to original radiology reports or identification of errors in subsequent clinical notes (mean follow-up duration, 12-36 months).
DISCUSSION Errors in medical practice have significant adverse effects on patient care. According to a report from the Institute
Table 2. Distribution of unidentified errors pertaining to levels of vertebral abnormalities in spinal magnetic resonance reports (January 2006 and January 2007) Level of Abnormality Level of Abnormality Mentioned in Mentioned in the the Body of Impression of True Level of Genderⴱ Diagnosis of Error the Report the Report the Abnormality F Disc bulge C3/C4 T3/T4 C3/C4 F Post laminectomy L3/L4 L4/L5 L3/L4 F Post hemilaminectomy L5/S1 L4/L5 L4/L5 M Post traumatic injury C4/C5 T4/T5 C4/C5 F Endplate edema L2/L3 L1/L2 L2/L3 F Nerve compression S1 L5 L5 M Vertebral body expanded T9 T8 T9 F Compression fracture T3 T2 T3 F Inflammatory changes S1 L1 S1 M Disc osteophyte C3/C4 C5/C6 C5/C6 M Endplate edema C3/C4 C2/C3 C3/C4 F Vertebral body height loss T12 T9 T12 M Post diskectomy C6/C7 C5/C6 C6/C7 F Canal stenosis L2/L3 L2/T3 L2/L3 F Degenerative changes T6/T7 C6/C7 T6/T7 F Tarlov cyst S1/S2 S2/S3 S1/S2 F Vertebral hemangioma L3 T3 L3 F Disc herniation L3/L4 L2/L3 L3/L4 M Nerve compression L1 S1 S1 F Soft tissue density L2/S1 L5/S1 L5/S1 ⴱFourteen women, 6 men.
Sangwaiya et al/Discrepancies in Reporting Spinal MR Abnormalities 719
of Medicine, a federal agency, medical errors result in more deaths than motor vehicle accidents, breast cancer, or acquired immune deficiency syndrome [6]. Wrongsided or wrong-site surgery is one of the most common and significant medical errors. In a recent study on the basis of a self-reported survey of Canadian neurosurgeons, the frequency of surgery at the wrong vertebral level is common. In fact, the incidence of cervical discectomy at the wrong level is about 6.8 per 10,000 surgeries [3]. The investigators noted that for spinal surgery, unusual patient anatomy and mistakes in radiographic reporting are common contributors to spinal surgeries at erroneous levels. However, to the best of our knowledge, there are no prior data on the incidence of incorrect assignment of vertebral levels of abnormalities in radiology reports. Our study found discrepancies between the vertebral levels of abnormalities stated in the body and impression sections of radiology reports. Most discrepancies were not identified and corrected by the radiologists (unidentified errors, 20 of 1,780 [1.1%]), although some were subsequently corrected (11 of 23,212 [0.05%]) to state the correct vertebral levels of abnormalities. Unfortunately, for most radiology reports with discrepant vertebral levels of abnormalities in our study, the wrong levels were stated in the impression sections (21 of 31 [68%]), whereas the body sections correctly stated the vertebral levels with abnormalities. Almost half (15 of 31 [48%]) of these discrepant reports on vertebral levels of abnormalities were clinically significant, and the time interval between the correction of a discrepant report and its initial dictation could be long (between 5 and 11 days). We were unable to compare these results with other published studies on radiology errors because, to the best of our knowledge, there are no similar studies reported in the indexed MEDLINE literature. We also observed that unidentified or uncorrected errors in vertebral levels of abnormalities were twice as frequent in female patients as in male patients (14 vs 6). This finding may have simply been related to the fact that more female patients underwent MR of the spine in both 2006 and 2007. It is interesting that a recent study on the prevalence of laterality errors (right vs left) in radiology reports also documented a greater number of errors in radiology reports for female patients compared with male patients [7]. Discrepancies noted in our study could have resulted from several factors, including busy MR services and reporting practices in our department, the presence of complex multilevel vertebral abnormalities, anatomic variability or artifacts affecting vertebral level count, human error in consistent counting of the vertebral level of abnormalities, or radiologists’ reliance on memory when dictating the impression at the end of the reports. It is
also possible that these errors were due to the use of voice recognition software; at other radiology centers, they may occur because of errors in transcription. Such errors constitute another reason for radiologists to validate report content, consistency, and concordance before final signature. Although these specific reasons were not assessed in our study, we believe that more than one of these factors may have contributed to the discrepancies observed in our study. Another limitation of our study was the lack of an analysis of the distribution of the discrepant vertebral levels of abnormalities among individual radiologists, an aspect that our human research committee explicitly prohibited us from investigating. For similar reasons, we did not assess the reasons some of the discrepant reports were corrected and others remained unidentified. A major limitation of our study was that we did not assess errors in radiology reports compared with the actual MR images for which both the body and impression sections of reports stated the incorrect levels of abnormalities in the spine. Our study also did not estimate basic interpretative mistakes of the radiologists, which could have further increased the number of unidentified errors considerably. Thus, the prevalence of errors in stating vertebral levels of abnormalities in our study likely underestimated the actual magnitude of these errors. Despite these limitations, our study has implications for both radiologists and referring physicians and surgeons. For referring physicians, we stress the importance of reading the entire radiology report instead of just the impression section. This is particularly important for surgeons or interventionalists who plan to undertake invasive procedures on the basis of MR reports. Radiologists must realize that discrepancies and errors in stating the correct vertebral level of abnormalities are frequent. They should make conscious efforts to match the vertebral levels reported in the impression and body sections of radiology reports before signing off the reports to referring physicians. Another option would be to “copy” and “paste” the body vertebral levels from the body section to the impression section. We also propose that the ACR look into the issue of discrepancies in radiology reports and perhaps make exceptions to its two-part radiology report structures for spinal MR reports, so that these reports have only one section that identifies the vertebral levels of abnormalities. Annotation of MR images for vertebral levels of abnormalities may also help reduce these errors. Efforts of the Joint Commission to reduce medical errors are noteworthy to avoid erroneous site and side of surgery. Recently, the Joint Commission [4] introduced a universal protocol for image guided and invasive surgical procedures. This document describes guidelines to be adopted before and during procedures to prevent or minimize the incidence of
720 Journal of the American College of Radiology/ Vol. 6 No. 10 October 2009
wrong-side and wrong-site procedures. The introduction of similar practices for radiologists may also help avoid or reduce the prevalence of these discrepancies in radiology reports. In conclusion, errors in the levels of lesions on spinal MR do occur in radiology reports. The number of unidentified errors is substantially higher than that of identified errors. Care should be taken before signing off on radiology reports to identify erroneous mentions of vertebral levels of abnormalities.
2. Michaels RK, Makary MA, Dahab Y, et al. Achieving the National Quality Forum’s “never events”: prevention of wrong site, wrong procedure, and wrong patient operations. Ann Surg 2007;245:526-32. 3. Jhawar BS, Mitsis D, Duggal N. Wrong-sided and wrong-level neurosurgery: a national survey. J Neurosurg 2007;7:467-72. 4. Joint Commission. Universal protocol. Available at: http://www.jointcommission. org/PatientSafety/UniversalProtocol. Accessed April 18, 2009. 5. American College of Radiology. Practice guideline for the reporting and archiving of interventional radiology procedures. Available at: http:// www.acr.org/SecondaryMainMenuCategories/quality_safety/guidelines/iv/ reporting_archiving.aspx. Accessed April 17, 2009.
REFERENCES
6. Kohn LT, Corrigan JM, Donaldson MS, eds. To err is human: building a safer health system. Washington, DC: National Academy Press; 2000.
1. National Quality Forum. Serious reportable events. Available at: http://www. qualityforum.org/pdf/news/prSeriousReportableEvents10-15-06.pdf. Accessed April 18, 2009.
7. Sangwaiya MJ, Saini S, Blake MA, Dreyer KJ, Kalra MK. Errare humanum est: frequency of laterality errors in radiology reports. AJR Am J Roentgenol 2009;192:W239-44.
Purpose: The aim of this study was to assess discrepancies in the spinal levels of abnormalities stated in the findings or impression (or both) sections of radiology reports of magnetic resonance (MR) imaging. Materials and Methods: Radiology reports from January 2006 through December 2007 (n ⫽ 2,097,966) were analyzed using an online radiology report search engine. Reports were searched for presence of the key words MR spine and addendum. The addended reports were then manually assessed for any discrepancies in the reported spinal levels between the body and impression sections; the addenda corrected these errors (identified errors). In addition, all reports with the search term MR spine from January 2006 (n ⫽ 1,183) and January 2007 (n ⫽ 1,354) were assessed manually to recognize unidentified errors in spinal locations of reported pathology. Two neuroradiologists independently graded the clinical significance of errors on a 5-point scale (1 ⫽ definitely not significant, 5 ⫽ definitely significant). Results: Of the 11,427 spinal MR reports analyzed in 2006, 7 had identified errors in the sites (levels of the spine) of the lesions. In 2007 (n ⫽ 11,785 spinal MR reports), 4 reports were detected with identified errors in spinal levels. In January 2006 and January 2007, 8 and 12 reports, respectively, had unidentified erroneous vertebral levels. Errors were related to discrepant vertebral regions (eg, cervical vs thoracic) in 16% of cases (5 of 31), the wrong number of vertebrae (eg, L2 instead of L3) in 68% of cases (21 of 31), and both in 16% of cases (5 of 31). The average time taken to issue an addendum was 5 ⫾ 7 days in 2006 and 11 ⫾ 13 days in 2007. Fifteen reports (48%) scored ⬍3 on the scale of clinical significance, 1 report scored 3, and 15 scored ⬎3. Conclusions: Errors in lesion level on spinal MR do occur in radiology reports. The number of unidentified errors is substantially higher than that of identified errors. Care should be taken before signing off on radiology reports to identify erroneous mentions of the vertebral levels of abnormalities. Key Words: Radiology reporting, quality, errors, spinal MR J Am Coll Radiol 2009;6:715-720. Copyright © 2009 American College of Radiology
INTRODUCTION The National Quality Forum, a US public and private entity, issued results in 1999 indicating that between 44,000 and 98,000 Americans die each year from preventable medical errors in hospitals alone [1]. This issue was revisited again in 2008, when a list of 28 “never events” (errors that should not occur in medical practice) was proposed. These never events included surgery at the wrong Department of Radiology, Massachusetts General Hospital, Boston, Massachusetts. Corresponding author and reprints: Mannudeep K. Kalra, Massachusetts General Hospital, Department of Radiology, 25 New Chardon Street, Suite 400, Boston, MA 02114; e-mail: [email protected]. © 2009 American College of Radiology 0091-2182/09/$36.00 ● DOI 10.1016/j.jacr.2009.06.002
site, surgery performed on the wrong patient, the wrong surgical procedure on the right patient, and patient death or serious disability associated with a medication error [2]. Various factors contribute to wrong-site surgeries, including the failure of surgeons to adopt guidelines and suggestions by authorities. In 2007, Jhawar et al [3] described the frequently reported factors for wrong-sided cranial surgery. These included emergency operating conditions, fatigue, and time pressures to begin or complete procedures. To avoid wrong-site procedures, the Joint Commission [4] recommended that every patient undergo a preprocedural verification process to ensure that all relevant documents, related information, and equipment are available before the start of a procedure. The site of the 715
716 Journal of the American College of Radiology/ Vol. 6 No. 10 October 2009
procedure should be marked by a licensed independent practitioner. Also, to conduct a final assessment, surgeons and interventionalists should conduct a “time out” before beginning a procedure. This is to ensure that the correct patient, site, positioning, and procedure are identified. To the best of our knowledge, no prior studies have assessed the frequency of errors in vertebral level in radiology reports. The purpose of our study was to examine the frequency and severity of discrepancies in the spinal levels of abnormalities stated in the findings and impression (or both) sections of spinal magnetic resonance (MR) imaging reports at a tertiary care institution. MATERIALS AND METHODS This cross-sectional retrospective study was approved by the human research committee of our institution. The requirement for informed consent from study subjects was waived. This study was compliant with the Health Insurance Portability and Accountability Act. To protect the privacy of the reporting radiologists, our human research committee did not permit us to document the names of the radiologists who dictated the reports assessed in our study. Radiology Reporting According to ACR guidelines, the communication of diagnostic imaging findings in radiology reports should include two components or sections [5]; the “body” section should describe the procedures and materials, findings, potential limitations, clinical issues, and prior comparison studies, and the “impression” section should present a succinct summary of the main findings in the study performed and a conclusion or diagnosis. All radiology reports at our institution are reported using the ACR guidelines on a voice recognition dictation system (Nuance Communications Inc, Burlington, Massachusetts). After dictation, reports are electronically verified and signed off by the radiologists and then archived in the radiology information system (RIS) and other hospital electronic medical records. If an error is recognized in the original dictation after this stage, a radiologist must dictate an “addendum” to the original report, which is automatically integrated with the original radiology report in the RIS as well as all other hospital records of the patient. This workflow holds true for all radiology reports in our department. Each addended radiology report has an additional section called “addendum,” under which further explanations, revisions, and corrections are described. From the RIS, the radiology reports (and any subsequently issued addenda) are transferred to an electronic search engine, Folio VIEWS 4 (NextPage Inc, Draper, Utah), which has the capacity to perform Boolean searches with keywords, phrases, or expressions on all reports for each year and month. The
archived data date back to 1994 and are updated once every 3 months. Radiology Reports With Identified Errors We queried all radiology reports from January 2006 through December 2007 (n ⫽ 2,097,966) using the radiology search engine with the key word MR spine. A total of 23,212 spinal MR reports were found (n ⫽ 11,427 in 2006, 6,099 in female and 5,328 in male patients; n ⫽ 11,785 in 2007, 6,366 in female and 5,419 in male patients). On these spinal MR reports, we applied the key word addendum, to identify spinal MR reports with addenda (n ⫽ 445 in 2006, 252 in female and 193 in male patients; n ⫽ 340 in 2007, 202 in female and 138 in male patients). The addended spinal MR reports were manually searched for discrepancies in the intervertebral levels of abnormalities described in the body and impression sections. Corresponding MR images for these studies were reviewed to confirm the recognized errors (Figure 1). Radiology Reports With Unidentified Errors In this step, all spinal MR reports from Janaury 2006 (n ⫽ 828, 501 in female and 327 in male patients) and from January 2007 (n ⫽ 952, 429 in female and 523 in male patients) were identified using the key word MR spine. These reports included reports with or without identified errors of vertebral level of abnormality. All of these reports were manually assessed to determine the unidentified discrepancies in vertebral levels of abnormalities between the body and impression sections (Figure 1). Analysis of Site (Vertebral Level) Errors For each report with an identified discrepancy in the vertebral level of an abnormality between the body and impression sections, we recorded the stated vertebral level in reports with errors, correct vertebral levels in the addenda or MR images, and the average time period in days between initial signoff and the dictation of an ad-
Fig 1. Flowchart showing the search for erroneous radiology reports. MR ⫽ magnetic resonance; MRI ⫽ magnetic resonance imaging.
Sangwaiya et al/Discrepancies in Reporting Spinal MR Abnormalities 717
Fig 2. Flowchart depicting the incidence of errors (identified and unidentified) in the studied cohort. MR ⫽ magnetic resonance.
ments between the two readers. A score of 1 corresponded to definitely no clinical significance (eg, mild degenerative disease of the spine reported at different levels of the spine with a discrepancy between the body and impression sections of the report). On the other extreme, a score of 5 meant definite clinical significance (eg, a discrepancy between the body and impression sections in reporting the level of a spinal tumor). If the radiologist was uncertain about the clinical significance of a particular error, it was given a score of 3. Consensus agreement between the two radiologists was taken as final. The incidence of errors at each level of the spine was studied to determine the most frequent level of error, if any. RESULTS
dendum. The electronic medical record of each patient with an erroneous vertebral level in the radiology report was assessed to determine if the patient underwent a subsequent spinal procedure or surgery. We also evaluated the RIS and electronic medical records to determine if the unidentified errors had been corrected in subsequent radiology reports or in the medical records of the patients. Finally, two board-certified and experienced neuroradiologists independently graded each report with identified and unidentified errors of vertebral levels for the clinical significance of the errors. This was graded on a 5-point scale (1 ⫽ definitely no clinical significance, 2 ⫽ probably no clinical significance, 3 ⫽ uncertain, 4 ⫽ probable clinical significance, 5 ⫽ definite clinical significance). Consensus decisions were used for any disagree-
Radiology Reports With Identified Discrepancies Of the 23,212 spinal MR studies performed from January 1, 2006, through December 31, 2007, 11 reports (0.05%) had identified discrepancies in the vertebral levels of abnormalities between the body and impression sections (Figure 2). In 2006, there were 7 of 11,427 (4 in women, 3 in men; mean age, 61 years; age range, 34-77 years) reports with identified discrepancies in vertebral levels of the lesion. In 2007, there were 4 of 11,785 (3 in women, 1 in a man; mean age, 50 years; age range, 26-67 years) reports with identified discrepancies in the vertebral levels of abnormalities in the body and impression sections (Table 1). The average time taken to issue an
Table 1. Distribution of identified and corrected errors pertaining to the levels of vertebral abnormalities in spinal magnetic resonance reports (2006-2007) Level of Abnormality Level of Abnormality Mentioned in the Mentioned in the Impression of True Level of Genderⴱ Diagnosis of Error Body of the Report the Report the Abnormality M Spinal cord volume loss T5/T6 Cervical (C5/C6) T5/T6 F Disc herniation C4/C5 Not mentioned C5/C6 F Bony destruction T3 L3 T3 F Endplate fracture C6 C6 T6 F Progressed degenerative T7/T8 T8/T9 T7/T8 disease M Postoperative laminectomies L3, L4, L5 L3, L4, L5 T3, T4, T5 M Conus medullaris infarct and Cervical Cervical Thoracic multilevel degenerative changes M Epidural abscess C6/C7 C6/C7 C5/C6 F Disc protrusion C6/C7 C7/T1 C6/C7 F Metastatic soft tissue mass L5 T1 L5 F Fractures and intervertebral C6/C7 C6/C7 C7/T1 disc injury ⴱSeven female patients, 4 male patients.
718 Journal of the American College of Radiology/ Vol. 6 No. 10 October 2009
addendum for correcting these discrepancies was 5 ⫾ 1 days in 2006 and 14 ⫾ 13 days in 2007.
Types and Significance of Errors in Vertebral Levels of Abnormalities
Radiology Reports With Unidentified Errors
Errors were related to discrepant vertebral regions (eg, cervical vs thoracic) in 16% of cases (5 of 31), the wrong number of vertebrae (eg, L2 instead of L3) in 68% of cases (21 of 31), and both in 16% of cases (5 of 31). According to the first radiologist (R.G.), 17 of 31 discrepant reports were probably or entirely clinically insignificant, and the remaining 14 discrepancies were probably or definitely clinically significant. The second radiologist (A.J.Y.) graded 12 of 31 discrepant reports as probably or entirely insignificant, 1 as uncertain, and 18 as probably or definitely clinical significant to the patients’ management. There were differences in opinions of the two radiologists on the clinical significance in 9 of 31 reports. After consensus readouts, 3 discrepant reports were deemed clinically significant, and 5 were deemed clinically insignificant. The remaining discrepant report was graded as of uncertain clinical significance.
In January 2006 and January 2007, unidentified discrepancies in the levels of vertebral abnormalities in spinal MR reports were noted in 8 of 828 reports (6 in women, 2 in men; mean age, 52 ⫾ 12 years; age range, 38-77 years) and 12 of 952 reports (8 in women, 4 in men; mean age, 52 ⫾ 17 years; age range, 31-76 years), respectively (Table 2). Thus, in January 2006 and January 2007, there were of 1.1% (20 of 1,780) unidentified errors in spinal MR reports (Figure 2). Fortunately, none of the patients with unidentified errors had subsequent procedures or surgeries on the wrong vertebral levels. However, one patient did undergo an invasive procedure at the correct level, subsequent to imaging interpretation with an unidentified error. Also, in 4 of 20 reports with discrepant vertebral levels of abnormalities, the discrepancies were identified and corrected in subsequent reports or medical notes. The remaining patients with unidentified discrepancies did not have any addenda to original radiology reports or identification of errors in subsequent clinical notes (mean follow-up duration, 12-36 months).
DISCUSSION Errors in medical practice have significant adverse effects on patient care. According to a report from the Institute
Table 2. Distribution of unidentified errors pertaining to levels of vertebral abnormalities in spinal magnetic resonance reports (January 2006 and January 2007) Level of Abnormality Level of Abnormality Mentioned in Mentioned in the the Body of Impression of True Level of Genderⴱ Diagnosis of Error the Report the Report the Abnormality F Disc bulge C3/C4 T3/T4 C3/C4 F Post laminectomy L3/L4 L4/L5 L3/L4 F Post hemilaminectomy L5/S1 L4/L5 L4/L5 M Post traumatic injury C4/C5 T4/T5 C4/C5 F Endplate edema L2/L3 L1/L2 L2/L3 F Nerve compression S1 L5 L5 M Vertebral body expanded T9 T8 T9 F Compression fracture T3 T2 T3 F Inflammatory changes S1 L1 S1 M Disc osteophyte C3/C4 C5/C6 C5/C6 M Endplate edema C3/C4 C2/C3 C3/C4 F Vertebral body height loss T12 T9 T12 M Post diskectomy C6/C7 C5/C6 C6/C7 F Canal stenosis L2/L3 L2/T3 L2/L3 F Degenerative changes T6/T7 C6/C7 T6/T7 F Tarlov cyst S1/S2 S2/S3 S1/S2 F Vertebral hemangioma L3 T3 L3 F Disc herniation L3/L4 L2/L3 L3/L4 M Nerve compression L1 S1 S1 F Soft tissue density L2/S1 L5/S1 L5/S1 ⴱFourteen women, 6 men.
Sangwaiya et al/Discrepancies in Reporting Spinal MR Abnormalities 719
of Medicine, a federal agency, medical errors result in more deaths than motor vehicle accidents, breast cancer, or acquired immune deficiency syndrome [6]. Wrongsided or wrong-site surgery is one of the most common and significant medical errors. In a recent study on the basis of a self-reported survey of Canadian neurosurgeons, the frequency of surgery at the wrong vertebral level is common. In fact, the incidence of cervical discectomy at the wrong level is about 6.8 per 10,000 surgeries [3]. The investigators noted that for spinal surgery, unusual patient anatomy and mistakes in radiographic reporting are common contributors to spinal surgeries at erroneous levels. However, to the best of our knowledge, there are no prior data on the incidence of incorrect assignment of vertebral levels of abnormalities in radiology reports. Our study found discrepancies between the vertebral levels of abnormalities stated in the body and impression sections of radiology reports. Most discrepancies were not identified and corrected by the radiologists (unidentified errors, 20 of 1,780 [1.1%]), although some were subsequently corrected (11 of 23,212 [0.05%]) to state the correct vertebral levels of abnormalities. Unfortunately, for most radiology reports with discrepant vertebral levels of abnormalities in our study, the wrong levels were stated in the impression sections (21 of 31 [68%]), whereas the body sections correctly stated the vertebral levels with abnormalities. Almost half (15 of 31 [48%]) of these discrepant reports on vertebral levels of abnormalities were clinically significant, and the time interval between the correction of a discrepant report and its initial dictation could be long (between 5 and 11 days). We were unable to compare these results with other published studies on radiology errors because, to the best of our knowledge, there are no similar studies reported in the indexed MEDLINE literature. We also observed that unidentified or uncorrected errors in vertebral levels of abnormalities were twice as frequent in female patients as in male patients (14 vs 6). This finding may have simply been related to the fact that more female patients underwent MR of the spine in both 2006 and 2007. It is interesting that a recent study on the prevalence of laterality errors (right vs left) in radiology reports also documented a greater number of errors in radiology reports for female patients compared with male patients [7]. Discrepancies noted in our study could have resulted from several factors, including busy MR services and reporting practices in our department, the presence of complex multilevel vertebral abnormalities, anatomic variability or artifacts affecting vertebral level count, human error in consistent counting of the vertebral level of abnormalities, or radiologists’ reliance on memory when dictating the impression at the end of the reports. It is
also possible that these errors were due to the use of voice recognition software; at other radiology centers, they may occur because of errors in transcription. Such errors constitute another reason for radiologists to validate report content, consistency, and concordance before final signature. Although these specific reasons were not assessed in our study, we believe that more than one of these factors may have contributed to the discrepancies observed in our study. Another limitation of our study was the lack of an analysis of the distribution of the discrepant vertebral levels of abnormalities among individual radiologists, an aspect that our human research committee explicitly prohibited us from investigating. For similar reasons, we did not assess the reasons some of the discrepant reports were corrected and others remained unidentified. A major limitation of our study was that we did not assess errors in radiology reports compared with the actual MR images for which both the body and impression sections of reports stated the incorrect levels of abnormalities in the spine. Our study also did not estimate basic interpretative mistakes of the radiologists, which could have further increased the number of unidentified errors considerably. Thus, the prevalence of errors in stating vertebral levels of abnormalities in our study likely underestimated the actual magnitude of these errors. Despite these limitations, our study has implications for both radiologists and referring physicians and surgeons. For referring physicians, we stress the importance of reading the entire radiology report instead of just the impression section. This is particularly important for surgeons or interventionalists who plan to undertake invasive procedures on the basis of MR reports. Radiologists must realize that discrepancies and errors in stating the correct vertebral level of abnormalities are frequent. They should make conscious efforts to match the vertebral levels reported in the impression and body sections of radiology reports before signing off the reports to referring physicians. Another option would be to “copy” and “paste” the body vertebral levels from the body section to the impression section. We also propose that the ACR look into the issue of discrepancies in radiology reports and perhaps make exceptions to its two-part radiology report structures for spinal MR reports, so that these reports have only one section that identifies the vertebral levels of abnormalities. Annotation of MR images for vertebral levels of abnormalities may also help reduce these errors. Efforts of the Joint Commission to reduce medical errors are noteworthy to avoid erroneous site and side of surgery. Recently, the Joint Commission [4] introduced a universal protocol for image guided and invasive surgical procedures. This document describes guidelines to be adopted before and during procedures to prevent or minimize the incidence of
720 Journal of the American College of Radiology/ Vol. 6 No. 10 October 2009
wrong-side and wrong-site procedures. The introduction of similar practices for radiologists may also help avoid or reduce the prevalence of these discrepancies in radiology reports. In conclusion, errors in the levels of lesions on spinal MR do occur in radiology reports. The number of unidentified errors is substantially higher than that of identified errors. Care should be taken before signing off on radiology reports to identify erroneous mentions of vertebral levels of abnormalities.
2. Michaels RK, Makary MA, Dahab Y, et al. Achieving the National Quality Forum’s “never events”: prevention of wrong site, wrong procedure, and wrong patient operations. Ann Surg 2007;245:526-32. 3. Jhawar BS, Mitsis D, Duggal N. Wrong-sided and wrong-level neurosurgery: a national survey. J Neurosurg 2007;7:467-72. 4. Joint Commission. Universal protocol. Available at: http://www.jointcommission. org/PatientSafety/UniversalProtocol. Accessed April 18, 2009. 5. American College of Radiology. Practice guideline for the reporting and archiving of interventional radiology procedures. Available at: http:// www.acr.org/SecondaryMainMenuCategories/quality_safety/guidelines/iv/ reporting_archiving.aspx. Accessed April 17, 2009.
REFERENCES
6. Kohn LT, Corrigan JM, Donaldson MS, eds. To err is human: building a safer health system. Washington, DC: National Academy Press; 2000.
1. National Quality Forum. Serious reportable events. Available at: http://www. qualityforum.org/pdf/news/prSeriousReportableEvents10-15-06.pdf. Accessed April 18, 2009.
7. Sangwaiya MJ, Saini S, Blake MA, Dreyer KJ, Kalra MK. Errare humanum est: frequency of laterality errors in radiology reports. AJR Am J Roentgenol 2009;192:W239-44.