HI150 Unit 8 Seminar Reflection Example

Hypertension Dictated, Hypotension Signed: A Unit 8 Seminar Reflection Following One Unedited Speech Recognition Error Through a Coding Query, a Correction Entry and an Afternoon of Health Information Work

[Student Name]

Purdue University Global

HI150: Automation of Health Information 1

Unit 8 Seminar Reflection

[Instructor Name]

[Date]

The seminar case, the hospital and the clinician are composite; the error rates and the sources cited are real.

The Case as Posed

A hospitalist dictated a discharge summary using speech recognition and signed it the same afternoon. The software heard "hypotension" for "hypertension" in the past medical history, and nobody read the line again before the signature went on. Elsewhere the summary listed two blood pressure medications. The coder assigned diagnoses from the summary, noticed the contradiction and sent a query.

Where the Time Went

I tallied it from what the seminar case described. The coder spent about ten minutes finding the contradiction and drafting a query. The account waited four days for an answer, holding the claim. The physician spent perhaps five minutes reviewing the summary and writing a correction. The correction had to be an amendment entry, dated and timed, preserving the original text, because a signed document cannot simply be overwritten, and federal hospital rules expect every entry to stay dated, timed and authenticated (42 C.F.R. § 482.24, 2024). A health information analyst then checked that the correction linked to the original and that the corrected version would be the one released in future. Rough total: more than an hour of staff time across four people, plus four days of delayed billing, to undo an error the physician could have caught in ten seconds of reading before signing. The software saved the physician typing time and moved the proofreading to the coding desk, where it cost far more.

What the Evidence Says

A classmate brought a study that followed dictated documents from three stages: raw speech recognition output, the transcriptionist-edited version and the signed note. The error rate was 7.4 per 100 words in raw output, falling to 0.4 after transcriptionist review and 0.3 in signed notes. Even so, an error survived somewhere in 42.4 percent of the signed notes, and about 6 percent of errors in signed notes were clinically significant (Zhou et al., 2018). That changed how I thought about the case. A hypotension-for-hypertension swap is not a rare slip. It is the expected result of signing without reading, multiplied across thousands of notes.

What I Believed and Where I Ended

I came in thinking the fix belonged to coding: coders should catch contradictions, and queries are how the system corrects itself. A classmate who codes for a living disagreed sharply. Coders see only what affects codes; a wrong medication dose or a wrong allergy in a discharge summary would never reach their desk. Relying on coding to proofread means relying on chance for everything that does not change a code. I now think the fix belongs at signing. Two things would help. First, a short required review step before signature on speech-recognized documents, which some systems support by highlighting low-confidence words. Second, feedback to physicians showing how often their signed notes generate queries for transcription-type errors, since many have no idea. Health information departments can supply that data; they see every correction, including those requested by patients themselves, who have a right to ask for amendment when they read an error in their own record (45 C.F.R. § 164.526, 2024).

What the Department Could Do Next Month

Our department already logs every correction entry. With a filter for corrections that replace a single word or phrase in a signed dictated document, we could count transcription-type errors by author each month at almost no cost. The first report would likely show a handful of physicians producing most of the corrections, which is the pattern any quality measure tends to reveal. That list is not a disciplinary tool. It is a way to start a conversation with the physicians who would benefit most from a review step, and to show clinical leaders how much health information time those errors consume. I have proposed it to my supervisor as a three-month trial with the hospitalist group, since the seminar case came from their documents.

What Remains Open

The seminar did not resolve who pays for the time. The hour spent fixing the error was charged to the coding and health information budgets, not to the clinical department that created it. As long as the cost lands elsewhere, the incentive to proofread stays weak. I do not know whether departments should report correction time back to clinical leaders, or whether that would only produce resentment. I would like to try it for one month with one group and see whether the conversation changes, or only the mood.

References

42 C.F.R. § 482.24 (2024). https://www.ecfr.gov/current/title-42/section-482.24

45 C.F.R. § 164.526 (2024). https://www.ecfr.gov/current/title-45/section-164.526

Zhou, L., Blackley, S. V., Kowalski, L., Doan, R., Acker, W. W., Landman, A. B., Kontrient, E., Mack, D., Meteer, M., Bates, D. W., & Goss, F. R. (2018). Analysis of errors in dictated clinical documents assisted by speech recognition software and professional transcriptionists. JAMA Network Open, 1(3), Article e180530. https://doi.org/10.1001/jamanetworkopen.2018.0530