HIM 500 Module 8 Emerging Technology Short Paper Example

Who Wrote This Note? Artificial Intelligence and the Legal Health Record at Bramble Bay

[Student Name]

Southern New Hampshire University

HIM 500: Healthcare Informatics

Module Eight Short Paper

[Instructor Name]

[Date]

The organization, setting and figures below are a composite written as a model document. No real employer, client, colleague or patient is described.

Who Wrote This Note? Artificial Intelligence and the Legal Health Record at Bramble Bay

Artificial intelligence is arriving at Bramble Bay Medical Center through the record. An ambient documentation pilot will soon draft clinical notes from recorded conversations, the coding vendor is offering an assistant that suggests codes from notes and several physicians already paste drafts from general chatbots into patient portal replies. Most discussions of AI focus on accuracy and efficiency. This paper asks a question health information leaders are well placed to answer: what happens to the legal health record when software writes part of it?

Kinds of AI in the Record

Four kinds of AI now touch the record. Ambient documentation tools convert a recorded conversation into a draft note for the clinician to review. Natural language tools read notes to suggest diagnosis and procedure codes or to identify quality measure data. Generative tools draft replies to patient messages or summaries of long records. Predictive models estimate risks such as deterioration or readmission and display scores in the chart. Topol (2019) described the convergence of human and machine intelligence across medicine and argued that AI's most promising near-term contribution may be restoring time for clinicians to spend with patients, while cautioning that validation in real clinical settings lagged far behind the enthusiasm.

What the Evidence Shows

Evidence is growing but uneven. For ambient documentation, Tierney et al. (2024) reported a rapid rollout at a large integrated health system, where thousands of physicians used ambient scribes across hundreds of thousands of encounters and generally reported that the tool reduced documentation burden, with notes requiring review and editing. For generative replies, Ayers et al. (2023) compared chatbot and physician answers to patient questions posted on a public forum and found that evaluators preferred the chatbot responses most of the time and rated them higher for quality and empathy, although the study did not involve real patient records or clinical follow-up. For predictive models, the strongest lesson is cautionary: Obermeyer et al. (2019) traced unequal treatment in a popular population health algorithm to its target: because it forecast future spending, and unequal access had kept spending on Black patients lower, it rated them as healthier than white patients with the same burden of illness.

Risks

Three risks matter most for the record. First, generative tools can produce fluent text that is wrong, such as an examination finding that was never performed or a medication dose the patient did not mention, and a busy clinician may sign it. Second, bias can enter through training data or through the choice of what a model predicts, as Obermeyer and colleagues showed. Third, automation bias, the tendency to trust machine output over one's own judgment, can lead clinicians and coders to accept suggestions without checking them. Char et al. (2018) warned that the values and incentives of those who build and deploy machine learning determine how it behaves, meaning that software built to maximize coded revenue or visit throughput may behave differently from one designed for accuracy and care.

What AI Means for the Legal Health Record

AI raises specific questions for health information management. Table 1 sets them out with Bramble Bay's proposed answers.

Table 1. AI and the Legal Health Record

Note. Rules proposed by the author for the steering committee.

Applying the Rules to the Pending Tools

The rules translate into concrete conditions for the tools Bramble Bay is weighing. For the ambient documentation pilot, every draft note will carry an attestation line, clinicians will be trained to read the whole draft rather than skim it and health information staff will audit 40 pilot notes a month against the recording's transcript, counting additions, omissions and errors, with a stopping rule if serious errors exceed an agreed rate. For the AI coding assistant, coders will review every suggestion, and the system will log which suggestions were accepted so that compliance can compare accepted codes with documentation during routine audits; any pattern of suggested codes that consistently raise payment will be investigated. For patient messages, the hospital will block pasting from outside chatbots on hospital devices and evaluate a generative tool built into the record, with drafts clearly marked for review. Each condition protects the same principle: a human who understands the patient remains accountable for what the record says.

A Governance Framework for AI

Bramble Bay should manage AI tools through its existing evaluation process with four additions. First, an inventory of every AI tool in use, including unofficial ones such as general chatbots, so that nothing operates without review. Second, local validation before use, such as a silent trial for predictive models and a note accuracy audit for ambient documentation. Third, monitoring after launch, because models and vendors change; accuracy, bias measures and user overrides will be reviewed at least twice a year. Fourth, transparency to patients about when AI helps draft their notes or messages. The health information department should own the record rules in Table 1 and the monthly audit of AI-drafted notes, because protecting the integrity of the record is its core responsibility.

Conclusion

AI can relieve documentation burden and improve communication, but it changes how records are created and who stands behind them. Bramble Bay should adopt AI tools deliberately, keeping clinicians as accountable authors, marking machine-drafted content, validating and monitoring tools locally and treating accuracy of the record as a condition of adoption rather than an afterthought.

References

Ayers, J. W., Poliak, A., Dredze, M., Leas, E. C., Zhu, Z., Kelley, J. B., Faix, D. J., Goodman, A. M., Longhurst, C. A., Hogarth, M., & Smith, D. M. (2023). Comparing physician and artificial intelligence chatbot responses to patient questions posted to a public social media forum. JAMA Internal Medicine, 183(6), 589-596. https://doi.org/10.1001/jamainternmed.2023.1838

Char, D. S., Shah, N. H., & Magnus, D. (2018). Implementing machine learning in health care: Addressing ethical challenges. New England Journal of Medicine, 378(11), 981-983. https://doi.org/10.1056/NEJMp1714229

Obermeyer, Z., Powers, B., Vogeli, C., & Mullainathan, S. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science, 366(6464), 447-453. https://doi.org/10.1126/science.aax2342

Tierney, A. A., Gayre, G., Hoberman, B., Mattern, B., Ballesca, M., Kipnis, P., Liu, V., & Lee, K. (2024). Ambient artificial intelligence scribes to alleviate the burden of clinical documentation. NEJM Catalyst, 5(3). https://doi.org/10.1056/CAT.23.0404

Topol, E. J. (2019). High-performance medicine: The convergence of human and artificial intelligence. Nature Medicine, 25(1), 44-56. https://doi.org/10.1038/s41591-018-0300-7