The typical descriptive surgical pathology report begins with measurements and other visual observations from gross dissection. Also frequently included is a block summary which serves as a key for interpreting the submitted histologic sections. Dictated portions of the report, including gross descriptions, microscopic descriptions, and diagnoses are often manually transcribed into the LIS. The actual typographical error rate of manual transcription is difficult to ascertain because there are generally multiple opportunities for proofreading by different people (transcription supervisors, pathologists’ assistants, and residents) with error correction before report verification. A metric that is somewhat more available is the typographical error rate observed in verified pathology reports. Several groups have conducted manual retrospective reviews of verified reports with the purpose of identifying all typographical errors—regardless of severity or clinical significance. Observed error rates in this context range from 6 to 12.3 % [2, 3]. Out of 4446 cases, Zardawi et al. reported a 3 % typographical error rate and a 3 % “other clerical error” rate for a total of 6 % in a large teaching hospital [3]. An examination of 1756 verified surgical pathology reports (generated on cases containing intraoperative consultations) at our institution demonstrated a typographical error rate of 11.7 % when all typographical and formatting errors are included (unpublished data). Malami and Iliyasu reported a typographical error rate of 12.3 % after examining 2877 cases from their teaching hospital [2]. Thus, unsolicited retrospective review of surgical pathology reports will identify typographical errors at a rate of up to one error per nine surgical pathology reports.

Fortunately, in the vast majority of cases such typographical and clerical errors are minor, have little to no impact on the interpretive meaning of a pathology report, and do not result in harm to the patient. Infrequently however, these errors may have a significant clinical impact. A study of 272 consecutive surgical pathology-related malpractice insurance claims over a 6 year period demonstrated one malpractice claim directly related to a surgical pathology transcription error. In the relevant case, the word “no” was omitted before the words “malignant cells” [4]. In another study, Hocking and colleagues randomly audited 250 reports from their hospital and graded identified reports as satisfactory, borderline, or unsatisfactory from a clinical care perspective. They identified an 8.8 % borderline rate in these reports. This borderline group included borderline errors of misinterpretation as well as borderline report defects (such as omissions in cancer reporting or use of incorrect classification schemes). A subset of this group represented borderline typographical errors identified in microscopic descriptions which potentially altered interpretive meaning. However, the authors are careful to note—the proper meaning could generally be gleaned from the remainder of the report . Of note, 2 % of their cases (n = 5) were graded as unsatisfactory from a clinical care perspective; none of these were cases of typographical errors.

These types of careful, detailed retrospective audits of surgical pathology reports are labor intensive and can be expensive to perform. A different metric, and one that is more commonly included as part of routine quality assurance in surgical pathology , is the rate of report amendment/correction following verification, for report defects. This metric is easily calculated with the assistance of the anatomic pathology LIS. Amendment rates in surgical pathology are considerably lower than the typographical error rate identified in manual retrospective audits of all verified pathology reports. A College of American Pathologists Q-probes study examining surgical pathology report defects asked participants to review all surgical pathology reports that underwent changes to correct defects. The median defect rate in this study of 73 institutions was 5.7 corrected defects per 1000 reports, or 0.57 %. In this study, 1.2 defects per 1000 reports were classified as a dictation or typographical error [5]. This is consistent with other published data showing about 18 % of amended reports represent typographical errors [6]. Utilizing the report defect taxonomy developed by Meier et al., other reporting defects, defects in specimens, errors of misinterpretation, and errors of misidentification are also included in the amended report rate [7]. Therefore, a considerable gap exists between the typographical error rate observed in unselected retrospective audits and the typographical error rate obtained by examining only corrected reports. Quite simply, many surgical pathology typographical errors are never identified and so are never corrected.

Raab and colleagues reported the results of an earlier College of American Pathologists Q-Probe in which 74 participating institutions, mostly from the USA and Canada, reviewed 5268 previously verified surgical pathology reports as part of their regular practice processes (multidisciplinary conferences, externally reviewed cases, regular QA processes, and physician requests for second review). In this context, the surgical pathology overall discrepancy rate was 6.8 %. In this study, the report defects were examined by organ system. None of the organ systems demonstrated a significantly different discrepancy rate. However, within the five most common subgroups (gastrointestinal and other, female genital, breast, lung, and male genital), there were differences between the types of discrepancies identified. The majority of discrepancies in male genital reports were misinterpretations (95.8 %), with 4.2 % of discrepancies being due to typographical errors. By contrast, 31.8 % of the discrepancies identified in lung reports were due to typographical errors, with 63.6 % classified as misinterpretations and 4.6 % misidentification errors. These results suggest there may be inherent differences in the difficulty of manual transcription between organ systems reflecting complexity of terminology or transcriptionist level of familiarity with the vocabulary.

Anecdotally, the typographical error rate seen in surgical pathology reports at our institution represents the work of 25 pathologists whose typographical error rates vary but individually held fairly steady over a time period of 5 months (unpublished data). Our surgical pathology practice is subspecialized; therefore, it is uncertain whether differences in typographical error rates seen among individuals in our practice represent inherent differences in pathologists’ proofreading and detail orientation, or whether they represent organ system-specific characteristics as suggested by the multi-institutional Q-Probe study above. Data are lacking as to whether individual typographical error rates correlate with other quality assurance metrics such as turnaround time, report defects including incomplete cancer reporting , or misinterpretation errors.

When specifically targeting typographical errors , it is clear that study design is important. As has been previously observed, the closer one looks for error, the more likely one will find it [6, 8, 9]. The data also seem to indicate that the vast majority of typographical errors have minimal if any clinical impact and so only very infrequently lead to adverse outcomes. In an era of limited resources and cost-containment, a strong argument can be made that programs targeting errors in medicine should focus on errors associated with adverse patient outcomes, rather than all errors [10]. However, one recent development that could change this paradigm is patient access to EHR data. At our institution, patients have access to the full text of their surgical pathology reports through an online patient portal, also called a personal health record (PHR). Typographical errors which would have previously been overlooked by surgical and clinical colleagues during routine practice are now subject to the scrutiny of individual patients and their families. Surgical pathology reports containing typographical error(s) when found within a PHR could lead to a perception of lower quality by patients, altering patient satisfaction. The potential also exists for patients to attempt to rectify identified errors through provider communication . A recent review of the literature on the subject of PHR access suggests both areas (effect on patient satisfaction and effect on physician workload) are understudied [11].

Beginning in the surgical pathology laboratory, prosectors should work, as much as possible, in an area free from extraneous background noises and stressors. In a review of patient safety and error reduction in surgical pathology, Nakhleh recommends that interrupting phone calls and other distractions be separated from tissue sectioning (and presumably the associated dictation process) because these functions require focus [12]. Rather than emphasizing speed, prosectors should carefully enunciate into recording devices, spelling long or difficult words as necessary.

In the manual transcription area, a quiet physical environment is essential. Transcriptionists should be seated at workstations with headphones, free from distraction by interrupting phone calls. Documents provided with the dictation file should be proofread and double checked, especially for demographic and specimen information including laterality. Medical dictionaries and additional support materials should be available. Transcriptionists should not be hesitant to request assistance from a prosector if there is a problem with a dictation.

Proofreading at all steps is essential. In the case of manual transcription, the transcriptionist is the first person to see the content of the report text. Every effort should be made to have transcriptionists proofread their reports, or a representative sample of their reports, to insure continuous quality improvement. A second opportunity for proofreading of the demographic, clinical, and gross description portions of the report occurs before analysis of the slides. The prosector (pathologist, pathologists’ assistant, staff member, or resident) should be required to proofread the transcribed product of their dictation. Again, this provides valuable feedback to the prosector about the dictation process. In academic medical centers, residents who are given a chance to “preview” slides and dictate or type their own surgical pathology diagnoses have another chance to review and correct transcription errors. Of course, before report verification, the attending pathologist responsible for the report should review the report in its entirety.

If auditing unselected verified reports for typographical errors is not part of the routine QA practice in your department, consider running such an audit on a subset of cases periodically. Reports at our institution found to contain typographical errors during such audits are forwarded to the verifying pathologist for review. The verifying pathologist may, at their discretion, amend the report to correct the typographical error. Attending pathologists may choose to share these reports with pathologist’s assistants and residents as a periodic additional educational and quality assurance opportunity. If typographical errors are routinely audited in your practice, consider examining the data to identify correlations between typographical error rate and organ system, pathologist, or report type, and intervene as appropriate.

If transcription errors in clinical history are problematic, consider whether your practice could utilize standardized pathology requisition forms tailored to specific clinics. In this case, check boxes next to common symptoms associated with a particular medical specialty can be utilized. This limits the amount of additional vocabulary, abbreviations, and acronyms that the transcriptionists must learn to produce error-free reports. Similarly, use gross dissection and diagnostic templates whenever possible. Transcriptionists can pull up a template and use the case dictation only to fill in the blanks; this can reduce typographical errors .

Finally, although uncommon, critical errors in resulting in misinterpretation of benign versus malignant are potentially avoided by using best practices for diagnostically critical dictations. Specifically, one should avoid the use of the word “no” as in “no tumor cells present” and “no evidence of malignancy.” Opt instead for terms that are less likely to be incorrectly transcribed, such as “negative for tumor” or “negative for malignancy.”

Voice recognition software has been in existence for more than two decades; however, systems used before 1994 were not “continuous speech” voice recognition systems. These early systems required the speaker to insert unnatural pauses between words while speaking; causing a slower and more frustrating dictation process. They were not widely used. During the late 1990s, however, hardware and software advancements enabled reasonable-cost, continuous speech systems to come to market [13].