Module 6 Discussion: Initial Post
A Free Algorithm for Five Years of Records: Privacy, Security, and Fairness in a Data-Sharing Offer
Our health system was recently offered a free readmission-risk prediction tool by a technology company, on the condition that we share five years of de-identified inpatient records to help train and improve its models. The offer is attractive: we cannot afford to build such a tool ourselves, and readmissions carry penalties. But it raises three questions a DNP-prepared leader should insist on answering first.
Privacy. De-identified data can be shared under federal rules without patient authorization, but de-identification reduces risk rather than eliminating it, especially when detailed records are combined with other data sources a technology company may hold. Our patients did not expect their records to train a commercial product. Rosenbaum (2010) argues that data stewardship involves obligations beyond legal compliance, including transparency and accountability to the people the data describe, and she describes stewardship entities that balance the value of access against those obligations. A stewardship review should ask whether patients would reasonably expect this use, whether the contract forbids re-identification and resale, and whether the health system retains rights over derived models.
Security. Five years of records transferred to an outside company create a large target. The agreement should specify encryption in transit and at rest, access limited to named personnel, audit rights, breach notification timelines, and destruction of data at the end of the agreement.
Fairness. The deepest risk may be in the model itself. Obermeyer et al. (2019) showed that a widely used algorithm for identifying patients who need extra care exhibited significant racial bias: at a given risk score, Black patients were considerably sicker than White patients, because the algorithm predicted health care costs rather than illness, and unequal access means less is spent on Black patients. Correcting the bias would have raised the share of Black patients identified for additional help from 17.7 to 46.5 percent. An algorithm trained on our records will learn our inequities unless someone checks what it is actually predicting. Before adoption, the model's target, the variables it uses, and its performance across racial, ethnic, and insurance groups in our own population should be tested. Accuracy cannot be taken on trust either: when one widely implemented proprietary sepsis model was validated externally, it performed far below its developer's reported figures and missed most sepsis cases (Wong et al., 2021).
My recommendation would be to decline the offer as written and propose a counteroffer: a data use agreement with stewardship protections, security terms, and a requirement for local fairness testing before clinical use. Has anyone's organization accepted a similar arrangement, and what protections did it negotiate?
References
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
Rosenbaum, S. (2010). Data governance and stewardship: Designing data stewardship entities and advancing data access. Health Services Research, 45(5, Pt. 2), 1442-1455. https://doi.org/10.1111/j.1475-6773.2010.01140.x
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
How this DNP 865 Module 6 example is structured
DNP865 Module 6 discussions often weigh privacy, security and ethical use of patient data. Aspen does not publish module deliverables, so check your classroom for the exact prompt. This example frames one decision, addresses privacy, security and fairness in turn with evidence, makes a recommendation and ends with a question for peers.
DNP865 Module 6 questions, answered
What does DNP865 Module 6 usually ask for?
The module's discussion often asks you to weigh privacy, security and ethical questions in the use of patient data, such as data sharing, secondary use or algorithms. Aspen does not publish module deliverables, so your classroom's instructions govern.
Can de-identified data be shared without patient permission?
Federal rules allow sharing of properly de-identified data without authorization, but re-identification risk remains, and stewardship principles call for transparency, contractual protections and consideration of patients' reasonable expectations.
How can an algorithm be racially biased without using race?
By predicting a proxy that reflects unequal access. A widely used algorithm predicted health care costs rather than illness, and because less was spent on Black patients, it underestimated their needs at the same risk score.
Write yours, or have the desk draft it
This paper is an original model document written by our desk, not a submitted student paper and not an official Aspen University document. Read it for the moves, then write your own to the instructions in your classroom. If you want one built to your exact prompt and rubric, the first custom sample is free and arrives in 24 to 48 hours.