| Course | NUR-8150 Information Systems and Technology Impacting Healthcare Delivery |
|---|---|
| Module | Module 3 |
| Paper type | Systems paper: clinical decision support and alert fatigue |
| Length | About 1,066 words, 6 pages |
| Format | APA 7 student paper |
| School | Thomas Edison State University |
| Program | Doctor of Nursing Practice |
| Updated | October 2026 |
Free sample paper for NUR 8150 Module 3
Half the Alerts Ignored: Auditing Decision Support, Testing a Sepsis Model and Redesigning Alerts in a Camden Health System
Student Name
Doctor of Nursing Practice, Thomas Edison State University
NUR-8150: Information Systems and Technology Impacting Healthcare Delivery
Mentor Name
Month Day, Year
Half the Alerts Ignored: Auditing Decision Support, Testing a Sepsis Model and Redesigning Alerts in a Camden Health System
The Problem
The health system in this paper is invented: two hospitals and twenty clinics serving Camden and nearby towns. Its record fires thousands of alerts each day. Nurses and prescribers describe them as noise. A pharmacist on the medication safety committee reported that a serious drug interaction alert was overridden before a patient developed toxicity, and the review found the prescriber had overridden the same alert many times before without harm. The vice president for nursing asked the DNP-prepared director of clinical decision support to evaluate the system's alerts and its new sepsis prediction score.
The review of that event showed how fatigue works in practice. The prescriber had seen the same interaction alert on nearly every order for a common pair of drugs, almost always without consequence, and had learned to click past it. The alert looked the same whether the combination was harmless at the doses ordered or dangerous for a patient with poor kidney function. Nothing on the screen told the prescriber which situation this was. The safety committee concluded that the alert had trained its users to ignore it, and that adding more alerts would make the problem worse rather than better.
How Often Alerts Are Overridden
Nanji et al. (2014) studied medication-related decision support alerts in outpatient practices. Of 2,004,069 medication orders, 157,483 triggered alerts, a rate of 7.9 percent, and 52.6 percent of the alerts were overridden. The most common alerts were duplicate drugs, allergies and drug interactions. Override rates were highest for formulary substitution, age-based, kidney dosing and allergy alerts. In a review of 600 overrides, about 53 percent were judged appropriate, but this varied from 12 percent for kidney dosing alerts to 92 percent for allergy alerts, meaning many allergy alerts fired when they were not needed while many kidney alerts were dismissed when they mattered.
Our system's audit of three months of inpatient and outpatient orders found a similar picture: alerts on 9.4 percent of orders and an override rate of 61 percent. Duplicate therapy alerts made up a third of all alerts and were overridden 88 percent of the time, most often for appropriate reasons such as an as-needed dose alongside a scheduled one. Kidney dosing alerts were overridden 74 percent of the time, and a pharmacist review of fifty found that most of those overrides were not appropriate.
| Alert type | Our override rate | Appropriate on review | Action |
|---|---|---|---|
| Duplicate therapy | 88 percent | Mostly appropriate | Suppress known safe pairs |
| Allergy | 79 percent | Mostly appropriate | Clean allergy lists; tier severity |
| Kidney dosing | 74 percent | Mostly inappropriate | Make interruptive with dose suggestion |
| Serious interactions | 58 percent | Mixed | Keep interruptive; require reason |
Testing a Prediction Model
Last year the system switched on its vendor's built-in sepsis score. Wong et al. (2021) tested the vendor's sepsis score on 27,697 adults with 38,455 stays at Michigan Medicine, where sepsis occurred in 2,552 stays. Its discrimination, a c-statistic of 0.63, fell far short of what the vendor had reported. At Michigan's chosen threshold the score fired on 6,971 of 38,455 stays, yet it never flagged 1,709 of the 2,552 sepsis cases, and it caught only 183 patients whose antibiotics had not already been started.
The recommendation is to check the score against our own patients before anyone acts on it. The informatics team runs it silently for three months, comparing its scores with chart-confirmed sepsis. If its performance is similar to the Michigan results, it will not be used as an interruptive alert, and the existing nurse-driven sepsis screen will continue.
Silent validation means the model runs in the background and its scores are recorded, but no alerts reach clinicians. After three months, the informatics team compares the scores with sepsis cases confirmed by chart review using the same definition Michigan used. The team reports the model's sensitivity, how many patients it flags for each true case and how many cases it catches before antibiotics are already given. Only then will the committee decide whether the score adds anything to the nurses' screen.
What Good Support Looks Like
Carroll et al. (2013) tested a decision support system in pediatric clinics that asked mothers about depression on a waiting room form and then alerted physicians to positive screens, with or without printed guidance. Families of patients up to 15 months old were randomized to these approaches or to a reminder alone. Physicians suspected maternal depression and referred mothers twice as often with the decision support, 2.4 percent compared with 1.2 percent, and documented depressed mood and loss of pleasure far more often.
The lesson for our system is design: the support gathered information from the patient before the visit, reached the clinician at the right moment, and offered a clear next step. That pattern, information gathered, delivered at the moment of decision with an action attached, guides our redesign.
The Redesign
The redesign follows five principles. Alerts that are almost always overridden for good reasons are suppressed or made non-interruptive. Alerts that matter and are often wrongly overridden, such as kidney dosing, become interruptive with a suggested dose the prescriber can accept with one click. Allergy lists are cleaned so that intolerances, such as nausea, are not recorded as allergies. Each interruptive alert requires a coded reason for override, and the reasons are reviewed monthly. Nurses' alerts are reviewed separately, since they face different alerts during medication administration.
Measures
The decision support committee will track total alerts per 100 orders, override rates by alert type, the appropriateness of a monthly sample of overrides judged by pharmacists, adverse drug events linked to alert types and clinician satisfaction. For the sepsis model, the measures are sensitivity, positive predictive value and alerts per patient-day on our own data. A balancing measure, the rate of kidney-related adverse drug events, will confirm that fewer alerts have not removed protection.
Conclusion
Alerts in this health system are overridden more often than in a large published study, with many safe overrides of low-value alerts and many unsafe overrides of important ones. A widely used sepsis model missed most sepsis cases when tested independently, so ours will be validated locally before use. A pediatric trial shows that support designed around the clinician's moment of decision can work. The redesign suppresses low-value alerts, strengthens high-value ones with clear actions and measures the results.
References
Carroll, A. E., Biondich, P., Anand, V., Dugan, T. M., & Downs, S. M. (2013). A randomized controlled trial of screening for maternal depression with a clinical decision support system. Journal of the American Medical Informatics Association, 20(2), 311-316. https://doi.org/10.1136/amiajnl-2011-000682
Nanji, K. C., Slight, S. P., Seger, D. L., Cho, I., Fiskio, J. M., Redden, L. M., Volk, L. A., & Bates, D. W. (2014). Overrides of medication-related clinical decision support alerts in outpatients. Journal of the American Medical Informatics Association, 21(3), 487-491. https://doi.org/10.1136/amiajnl-2013-001813
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
NUR 8150 Module 3 instructions, in plain terms
Evaluating technology as a tool to improve patient and system outcomes is central to TESU's NUR-8150 course description, and with the syllabus kept for enrolled students, module three audits alerts and a sepsis score. Expect a DNP systems analysis that measures alert volume and override rates with system data, judges whether overrides are appropriate, compares local findings with published studies, evaluates a prediction model's performance before trusting it, applies evidence on what good decision support looks like, proposes a redesign tied to each alert type and sets measures including a balancing measure. Some sections assign a nurse-facing alert program, an early warning score or an artificial intelligence tool instead. See if the brief names a decision support framework, and whether they want a governance structure described.
How the NUR 8150 Module 3 example is put together
An invented Camden health system and a drug interaction alert overridden before harm open the paper. The outpatient override study is reported with its figures, and the system's own audit is set beside it in a four-row table that separates alerts to weaken from alerts to strengthen. The external validation of a sepsis model is reported, and a plan for silent local validation follows. The maternal depression trial shows what good support looks like and why. Five redesign principles follow, then measures that include a kidney-harm check. The notes explain why each alert type is handled differently and why every interruptive alert carries an action.
Where the marks sit in the NUR 8150 Module 3 rubric
Decision support papers are graded on measuring alert volume and overrides with system data, judging appropriateness rather than counting overrides alone, comparing findings with published evidence, validating prediction models locally before use, applying evidence on effective design, proposing changes tailored to each alert type and setting measures with a balancing measure. Mentors credit papers that recognize both alert fatigue and the danger of removing important alerts. Recommending silent validation before a model goes live is often noted as sound governance. Including nurse-facing alerts shows attention to the whole care team. Override rates, the sepsis c-statistic and the referral figures need quoting exactly, and the argument should travel from the overridden interaction alert to the redesign and its tracking.
NUR 8150 Module 3 help: mistakes that cost marks
A common mistake is treating every override as a problem; review whether overrides were appropriate before deciding what to change. Another is trusting a vendor's prediction model without local validation. Students often propose turning off alerts without a balancing measure to catch harm. Give high-value alerts a clear action, not only a warning. Include nurse-facing alerts. Describe who governs alert changes and how often they are reviewed. An alert or decision support paper for your course can begin with the brief; draft one is free. Validate any vendor prediction score on your own patients first, and report sensitivity and alerts per patient-day in plain terms.
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 Thomas Edison State 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.
More NUR 8150 and Doctor of Nursing Practice sample papers
- NUR 8150 Module 1: EHR Burden Paper: Event Log Data, Clerical Load and a Plan to Return Clinician Time
- NUR 8150 Module 2: Implementation Risk Paper: CPOE Error Pathways, a Mortality Warning and a Safer Go-Live Plan
- NUR 8150 Module 4: Predictive Analytics Paper: High-Cost Patients, Algorithmic Bias and Fair Use of Data
- NUR 8150 Module 5: Telehealth Paper: Pandemic Use Patterns, the Digital Divide and an Equitable Virtual Care Plan
- NUR 8150 Module 6: Cybersecurity Paper: Ransomware Trends, Breach Effects on Care and Downtime Readiness
- NUR 8000 Module 5: Collaboration Paper: Team Training for Rapid Response Calls
- NUR 8250 Module 1: Value-Based Payment Paper: Porter's Value Equation, Readmission Penalties and Unintended Harm
- NUR 7630 Module 4: Diagnostic Accuracy Paper: Pretest Probability, D-Dimer and Predictive Values
- NUR 8200 Module 4: Social Needs Paper: Community Health Workers, Navigation Trials and Return on Investment
NUR 8150 Module 3 questions, answered
Where can I find a free NUR 8150 Module 3 sample paper?
The Camden alert audit, sepsis model review and redesign are all printed above.
How often are medication alerts overridden?
In a study of more than 2 million outpatient orders, 52.6 percent of alerts were overridden, and about half of overrides were judged appropriate.
How accurate is the Epic sepsis model?
Tested independently at Michigan Medicine, it discriminated poorly, with a c-statistic of 0.63, and two thirds of sepsis cases went unflagged.
What makes clinical decision support effective?
Gathering the right information, delivering it at the moment of decision and offering a clear action, as in a trial that doubled maternal depression referrals.
How can alert fatigue be reduced?
Suppress low-value alerts, clean allergy lists, make important alerts actionable, require override reasons and review results regularly.