HCS 483 Week 5 Information Systems Management Example

Managing the Machine That Listens: Governance, Risk and Measurement for Ambient AI Documentation Tools in a Multispecialty Physician Group

[Student Name]

University of Phoenix

HCS/483: Health Care Information Systems

Week 5 Assignment

[Instructor Name]

[Date]

The physician group, its pilot and its figures are composites written for a model paper; research findings come from the sources listed.

Physicians in a composite multispecialty group of 220 physicians report spending one to two hours after clinic each evening finishing notes in the electronic health record. Burnout surveys rank documentation as the top frustration, and two primary care physicians left last year citing it. A vendor has offered an ambient artificial intelligence scribe: with patient consent, a smartphone application listens to the visit and drafts a note that the physician reviews and signs. The group is piloting it with 40 physicians. This paper considers the tool as an information systems management problem and sets out how the group will govern and evaluate it.

Early Evidence

Tierney et al. (2024) described one large integrated health system's rapid rollout of ambient AI scribes, making the technology available to thousands of physicians and reporting wide early use, physician enthusiasm and reductions in time spent on documentation, along with the need for physicians to review and correct drafts. Their report is encouraging but early: it comes from a single, well-resourced system, relies partly on self-reported benefits and has limited data on note accuracy and patient outcomes.

Promise and Limits of AI in Medicine

Topol (2019) reviewed the convergence of human and artificial intelligence in medicine and argued that the most immediate benefit may be restoring time for the human side of care, while cautioning that many algorithms had not been validated prospectively in real clinical settings and that issues of bias, privacy and transparency remained unresolved. Ambient scribes fit his optimistic case: they aim to give physicians back time and eye contact with patients.

A Cautionary Example

Vendor claims about artificial intelligence do not always hold up. Wong et al. (2021) externally validated a proprietary sepsis prediction model used by hundreds of hospitals and found much weaker performance than advertised: it identified sepsis poorly, missed about two thirds of sepsis cases and generated alerts for 18% of all hospitalized patients. The lesson for the physician group is that an AI tool's quality must be measured locally, on the group's own patients and specialties, before and after it is trusted, whatever the vendor's brochure says.

Governance

The group creates a clinical AI oversight committee including the chief medical information officer, two physicians from the pilot, a nurse, the privacy officer, a compliance officer and a patient representative. The committee approves the pilot's scope, reviews monitoring data monthly and decides whether to expand, modify or stop. It also sets a policy for any future AI tool: no clinical AI goes live without local validation and a named owner.

Vendor Contracting and Privacy

Recordings of visits are protected health information. The business associate agreement requires that audio is encrypted, stored only as long as needed to draft the note, not used to train the vendor's models without the group's written permission and deleted on a set schedule. The contract requires the vendor to report accuracy problems and security incidents promptly and allows the group to audit its data practices.

Patient Consent and Communication

Patients will be asked at check-in and again by the physician whether they agree to the tool listening. A one-page explanation, in English and Spanish, describes what is recorded, how it is used and how to decline. Declining will not affect care, and physicians will turn the tool off for sensitive discussions on request.

Monitoring Note Quality and Bias

Every AI-drafted note must be reviewed and signed by the physician, who remains responsible for its content. Each month, a physician reviewer and a coder audit a random sample of 50 notes against the audio for omissions, errors and invented details, and for problems that appear more often with patients who speak with accents or use interpreters. Coding accuracy is checked because notes drive billing.

Cost and Contract Terms

The vendor charges a monthly fee per physician, which for the full group would cost roughly $500,000 a year. The pilot contract runs six months with no automatic renewal, so the group can walk away if results disappoint. Pricing for expansion is fixed in advance, and the contract states that the group owns its notes and can export them if it changes vendors. Tying the decision to measured results rather than to enthusiasm protects the budget, since a tool that saves physicians time but reduces note accuracy could raise costs elsewhere through coding errors or liability.

Equity and Access

Ambient tools may work less well for patients whose first language is not English, for visits conducted through interpreters and for physicians with strong accents. The pilot deliberately includes physicians in clinics that serve many Spanish-speaking patients, so problems appear early rather than after expansion. Audit samples will be drawn in proportion from those clinics, and any gap in accuracy will be reported to the committee before a decision on expansion.

Training and Support

Pilot physicians receive an hour of training on consent, review and editing, and a superuser in each department handles questions. Information technology staff monitor application performance and phone compatibility.

Measures and Decision Point

Measures include after-hours documentation time from electronic health record audit logs, physician-reported burnout on a short survey, note error rates from the monthly audit, patient consent and decline rates, patient experience scores for the pilot physicians and cost per physician per month compared with any change in visit capacity. At six months, the oversight committee will decide whether to expand to all physicians, continue with changes or stop.

What the Pilot Teaches About Managing Future Tools

New technologies will keep arriving. Managing them well follows a pattern this pilot tests: define the problem first, read independent evidence, protect data in the contract, involve patients, measure locally and set a decision date.

Conclusion

Ambient AI scribes may give physicians back hours and attention, but the evidence is early and AI tools have disappointed before. Governance, careful contracting, patient consent, local monitoring of accuracy and bias, and a clear decision point let the group gain the benefits while managing the risks.

References

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 Innovations in Care Delivery, 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

Wong, A., Otles, E., Donnelly, J. P., Krumm, A., McCullough, J., DeTroyer-Cooley, O., Pestrue, J., Phillips, M., Konye, J., Penoza, C., Ghous, M., & Singh, K. (2021). External validation of a widely implemented proprietary sepsis prediction model in hospitalized patients. JAMA Internal Medicine, 181(8), 1065-1070. https://doi.org/10.1001/jamainternmed.2021.2626