Capstone Project Sample: A Quality-Improvement Project to Reduce Medication Errors

Model Answer

Abstract

Medication errors remain among the most common preventable causes of patient harm in acute hospital settings, with dosing and administration errors particularly prevalent on high-turnover medical wards. This capstone project reports a quality-improvement (QI) initiative undertaken on a 28-bed acute medical ward to reduce the rate of medication errors reported through the local incident-reporting system. Using the Institute for Healthcare Improvement’s Model for Improvement, a baseline error rate was established over four weeks before three sequential Plan-Do-Study-Act (PDSA) cycles were implemented: a structured medication safety huddle, a redesigned double-checking process for high-risk insulin administration, and refinement of electronic prescribing decision-support alerts. Weekly error rates, expressed per 1,000 doses administered, were tracked throughout a twelve-week improvement phase. The error rate fell from a baseline mean of 8.25 per 1,000 doses to 2.9 per 1,000 doses by the final week, a reduction of approximately 65%. Staff feedback indicated improved safety culture and confidence in escalation. The project demonstrates that structured, low-cost QI methodology can produce clinically meaningful reductions in medication error rates, and it concludes with recommendations for embedding and spreading the changes across the wider directorate.

Introduction and Problem Statement

Medication-related patient safety incidents account for a substantial proportion of avoidable harm reported across NHS acute trusts, with hospital prescribing errors alone estimated to affect around one in twenty inpatient prescriptions in UK hospitals (Ashcroft et al., 2015), and national commentary continues to identify medicines as one of the most persistent sources of preventable harm in the health service (King’s Fund, 2019). On the acute medical ward at the centre of this project, a review of six months of Datix incident reports identified medication errors as the single largest category of reported patient-safety incidents, with omitted doses, wrong-dose administration and insulin-related errors the most frequent sub-types. Ward-level discussion with nursing and pharmacy staff suggested that workload pressures, interruptions during medication rounds, and inconsistent double-checking practice for high-risk drugs were contributing factors.

The ward itself is a 28-bed acute medical unit admitting predominantly older adults with complex comorbidities, many of whom are prescribed five or more regular medicines on admission, a population in which polypharmacy is itself a recognised risk factor for administration error. The ward operates two medication rounds per shift, typically staffed by two registered nurses and, on early shifts, a rotating junior doctor undertaking the drug chart review. Informal conversations conducted as part of the project’s scoping work suggested that staff were broadly aware that medication safety was an area for improvement but lacked a structured, shared understanding of which specific failure points were driving the ward’s incident rate, which the baseline data collection described below was designed to clarify before any change was introduced.

This quality-improvement capstone project was therefore designed to answer the question: can a structured, ward-based improvement initiative reduce the rate of medication errors on an acute medical ward over a twelve-week period? The aim was expressed as a SMART objective: to reduce the ward’s reported medication error rate by at least 25% within twelve weeks of implementation, measured as errors per 1,000 doses administered, without increasing staff-reported workload burden. Secondary objectives were to strengthen the ward’s safety culture around escalation and to produce a set of changes that could plausibly be sustained and spread to comparable wards.

The scope of the project was deliberately bounded to a single ward and a twelve-week improvement window, following four weeks of baseline data collection, to make the initiative deliverable within the constraints of a taught capstone module while still generating data of practical value to the host organisation. Prescribing errors originating outside the ward were excluded from scope, as were errors relating to controlled drugs, which are governed by a separate local protocol. The project was also deliberately limited to changes that the ward team could implement and sustain with existing staffing and budget, rather than proposals requiring additional establishment or capital investment, since a central aim of the capstone was to demonstrate that meaningful improvement was achievable within the resource envelope realistically available to a single ward.

Background and Literature

Medication safety has been a persistent policy priority since the World Health Organization’s Global Patient Safety Challenge identified medication error as a leading avoidable cause of harm worldwide (World Health Organization, 2017), and successive UK reports have reinforced that a meaningful proportion of medication errors are preventable through system-level, rather than purely individual, intervention (NHS Improvement, 2018; Care Quality Commission, 2021). NICE guidance on medicines optimisation emphasises structured processes for high-risk drugs, including insulin, anticoagulants and opioids, and recommends that organisations use local incident data to target improvement work where harm potential is greatest (NICE, 2021).

The evidence base on double-checking is more equivocal than ward culture often assumes. Systematic reviews have found that independent double-checking can reduce error rates for high-risk medicines such as insulin, but only when the second check is genuinely independent rather than a rushed co-signature performed under time pressure (Alsulami et al., 2019). This distinction reflects a broader human-factors principle that safety systems relying on vigilance alone are inherently fragile (Reason, 2000; Vincent, 2010), and was directly relevant to the ward’s existing practice, where double-checking was nominally required but frequently performed as a formality alongside the primary nurse.

Electronic prescribing and decision-support systems have been shown to reduce prescribing error rates substantially compared with paper-based systems, though poorly tuned alert systems can generate “alert fatigue”, where clinicians begin to override or ignore warnings, eroding the system’s protective effect (Franklin et al., 2020; Nanji et al., 2018). This tension between alert sensitivity and staff tolerance informed the project’s approach to refining, rather than simply adding, e-prescribing alerts.

Improvement science literature consistently favours small, tested changes over large-scale redesign. The Institute for Healthcare Improvement’s Model for Improvement, built around iterative Plan-Do-Study-Act cycles, was selected as the project’s methodological framework because it is well suited to ward-level QI work with limited resources (Boaden et al., 2008), allows changes to be tested and adapted quickly, and has an established track record in UK medication-safety projects (Langley et al., 2009; Institute for Healthcare Improvement, 2023).

Ward safety huddles, the intervention selected for the project’s first PDSA cycle, have a growing evidence base in acute nursing settings as a low-cost mechanism for surfacing risks that might otherwise only be communicated informally or missed entirely at handover, particularly where they are kept brief, standardised and led consistently by the same role each shift rather than treated as an optional extra dependent on individual staff initiative. Taken together, this body of literature pointed towards a project design combining three complementary, low-cost interventions, communication (the huddle), process redesign (the double-check), and system tuning (the alerts), rather than relying on any single change to carry the full burden of improvement, consistent with the multi-layered defence model widely used in patient-safety improvement work.

Approach and Methodology

The project used the Institute for Healthcare Improvement’s Model for Improvement, structured around three fundamental questions: what are we trying to accomplish, how will we know a change is an improvement, and what changes can we make that will result in improvement. The measurable aim was defined in the objectives above; the improvement measure was the weekly medication error rate per 1,000 doses administered, drawn from Datix incident reports supplemented by a weekly structured medication-round observation audit conducted by the ward’s practice development nurse, to capture near-misses and minor errors that staff might not formally report.

A four-week baseline period preceded the improvement phase, during which no changes were introduced, to establish a stable measure of current performance. Three sequential PDSA cycles were then tested over twelve weeks. Cycle 1 introduced a brief, structured medication safety huddle at the start of each shift, led by the nurse in charge, in which any medication concerns from the previous shift were flagged and any high-risk administrations for the coming shift were highlighted. Cycle 2 redesigned the insulin double-checking process, replacing the informal co-signature with a mandatory independent second calculation performed away from the bedside before the two nurses compared results, supported by a laminated prompt card at the medication trolley. Cycle 3 worked with the trust’s pharmacy informatics team to review and retune three e-prescribing alerts that were firing at rates staff described as excessive, reducing nuisance alerts while retaining hard-stop warnings for genuinely high-risk combinations.

Stakeholders included the ward manager, a link pharmacist, the practice development nurse, junior medical staff prescribing on the ward, and the trust’s patient safety team, who reviewed the project design and agreed data-collection methods before implementation began. The project was registered locally as a service evaluation rather than research, in line with the Health Research Authority’s decision-support tool (Health Research Authority, 2023), as it involved evaluating an existing service rather than testing a new treatment against a control; no patient-identifiable data were collected, and all incident data were reviewed in anonymised, aggregate form.

Data were analysed descriptively using a simple weekly rate and presented as a run chart to allow visual identification of trend and shift, consistent with standard QI practice, which favours run-chart interpretation over formal statistical significance testing for small, time-limited datasets (Provost and Murray, 2011). Staff feedback was gathered informally through short structured conversations with ward staff at the end of the project.

Each PDSA cycle followed the same internal structure: a brief planning discussion with the relevant staff group to agree the change and how it would be tested; a two-to-three-week test period during which the change was applied and the weekly error rate and observation-audit data continued to be collected; a short study step in which the practice development nurse and project lead reviewed the emerging data together with informal staff feedback; and an act decision on whether to adopt the change as tested, adapt it, or abandon it before moving to the next cycle. This structure was chosen deliberately so that each cycle generated a small, interpretable dataset before the next change was layered on top, reducing the risk of introducing several simultaneous changes and being unable to attribute any resulting improvement to a specific cause.

Implementation / Findings

Baseline data collection (weeks 1–4) showed a mean medication error rate of 8.25 per 1,000 doses administered, consistent with the trust’s directorate-wide average for comparable wards and confirming that the ward’s error rate, while not exceptional, offered meaningful scope for improvement. Cycle 1 (weeks 5–6) introduced the medication safety huddle, an intervention with an established evidence base for improving situational awareness and communication in ward teams (Ilott et al., 2013); the error rate fell to a mean of 6.5 per 1,000 doses, and staff reported that the huddle surfaced concerns, such as a patient’s new allergy status, that had previously been missed at handover.

Cycle 2 (weeks 7–9) introduced the redesigned insulin double-checking process. The error rate fell further, to a mean of 4.6 per 1,000 doses, with insulin-specific errors, previously the ward’s largest single sub-category, falling from an average of 1.8 to 0.4 per week. Two nurses noted in feedback that performing the calculation independently, rather than glancing at a colleague’s figure, made them more confident that the check was meaningful rather than procedural.

Cycle 3 (weeks 10–12), the retuned e-prescribing alerts, produced a further reduction to a mean of 3.1 per 1,000 doses. Junior doctors reported a noticeably lower rate of alert overrides and described the remaining alerts as easier to take seriously, supporting the alert-fatigue mechanism identified in the literature review. Table 1 summarises the weekly error rate across the full data-collection period.

By the final week of the improvement phase, the error rate stood at 2.9 per 1,000 doses, a reduction of approximately 65% on the baseline mean and comfortably in excess of the project’s 25% target.

Evaluation of Outcomes

Against the project’s SMART objective, the initiative substantially exceeded its target, reducing the medication error rate by around 65% rather than the planned 25%, though this figure should be interpreted cautiously given the small scale and short timeframe of the project. A run-chart shift of this magnitude, sustained across two consecutive PDSA cycles, is generally considered a meaningful signal in QI methodology, but a twelve-week single-ward dataset cannot rule out seasonal variation in patient acuity, staffing levels, or a Hawthorne effect arising from staff’s awareness that their practice was being observed and measured, a caution well established in the QI literature (Dixon-Woods and Martin, 2016).

Secondary objectives were also broadly met. Structured feedback conversations with eleven ward staff at project end indicated that most felt the huddle and double-checking changes had not meaningfully increased their workload, and several described feeling more confident escalating concerns about prescriptions they judged unsafe. This is consistent with the project’s aim of strengthening safety culture alongside the numerical error-rate target.

Not every element performed as anticipated. The e-prescribing alert retuning, while ultimately effective, took longer to implement than planned because pharmacy informatics capacity was limited, meaning Cycle 3 ran for only three weeks rather than the intended four; a longer test period would have strengthened confidence in this element’s independent contribution to the overall reduction. Additionally, the observation-based audit component of data collection was resource-intensive for the practice development nurse and would not be sustainable as a permanent measurement method without redesign, for example by sampling a smaller number of rounds per week.

It is also worth noting that the three interventions were not tested in isolation from one another once introduced; because each cycle built on the previous one, later error-rate figures reflect the cumulative effect of the huddle, the double-check redesign and the alert retuning acting together, rather than the independent contribution of Cycle 3 alone. This is a common and generally accepted trade-off in sequential PDSA work, where the priority is demonstrating that a bundle of related changes improves an outcome, but it does mean that a claim that the alert retuning specifically accounted for the final 1.5 point reduction in the error rate would overstate what the data can support with confidence.

Recommendations and Reflection

Three recommendations follow directly from the project’s findings. First, the medication safety huddle and the redesigned insulin double-checking process should be formally embedded into ward standard operating procedure, given their demonstrated association with error reduction and their neutral-to-positive reception by staff. Second, the retuned e-prescribing alerts should be reviewed again after a longer bedding-in period to confirm that the reduction in overrides is sustained rather than a novelty effect, and the informatics team should be asked to extend a similar review to comparable wards. Third, given the resource intensity of the observation audit, the trust’s patient safety team should consider whether Datix incident data alone, supplemented by a lighter-touch monthly spot-check rather than a weekly full audit, could provide adequate ongoing measurement once the initial improvement has been confirmed as sustained.

Reflecting on the project as a piece of applied improvement work, the principal methodological strength was the use of a recognised, iterative framework that allowed each change to be tested on a small scale before being reinforced or adjusted, reducing the risk of implementing an ineffective change trust-wide. The principal limitation was scale: a single ward, twelve-week project cannot establish with confidence that the observed reduction would replicate elsewhere, or that it will persist once the novelty of the project period has worn off, and a longer follow-up audit at three and six months would strengthen the evidence base considerably.

On a personal level, the project developed practical skills in stakeholder negotiation, particularly in securing pharmacy informatics time for Cycle 3, and in translating literature-based evidence, such as the equivocal evidence on double-checking, into a locally credible case for change. It also reinforced the importance of involving frontline staff in designing changes to their own workflow, since the prompt card and independent-calculation redesign in Cycle 2 originated from a suggestion made by a staff nurse during an early huddle rather than from the project plan itself.

Finally, the project offered a useful reminder that sustaining an improvement is often harder than achieving it. Several similar ward-level initiatives described informally by colleagues elsewhere in the trust had produced an initial improvement that gradually eroded over subsequent months once the original project team moved on to other priorities. Guarding against this pattern here will depend on genuine ownership by the ward manager and link pharmacist, visible inclusion of the huddle and double-check steps in ward induction for new starters, and a simple, low-burden method of periodically re-checking the error rate so that any drift is noticed early rather than only becoming apparent through a future adverse incident.

References

Alsulami, Z., Choonara, I. and Conroy, S. (2019) ‘Double checking the administration of medicines: a systematic review’, Archives of Disease in Childhood, 104(2), pp. 178–183.

Ashcroft, D.M., Lewis, P.J., Tully, M.P., Farragher, T.M., Taylor, D., Wass, V., Williams, S.D. and Dornan, T. (2015) ‘Prevalence, nature, severity and risk factors for prescribing errors in hospital inpatients’, Drug Safety, 38(9), pp. 833–843.

Boaden, R., Harvey, G., Moxham, C. and Proudlove, N. (2008) Quality Improvement: Theory and Practice in Healthcare. Coventry: NHS Institute for Innovation and Improvement.

Care Quality Commission (2021) Medicines in Health and Adult Social Care: Learning from Risks and Sharing Good Practice. London: CQC.

Dixon-Woods, M. and Martin, G.P. (2016) ‘Does quality improvement improve quality?’, Future Hospital Journal, 3(3), pp. 191–194.

Franklin, B.D., Reynolds, M., Sadler, S. and Hibberd, R. (2020) ‘The impact of closed-loop electronic prescribing and administration on prescribing and administration error rates’, Journal of the American Medical Informatics Association, 27(4), pp. 618–626.

Health Research Authority (2023) Is My Study Research? London: HRA.

Ilott, I., Gerrish, K., Booth, A. and Field, B. (2013) ‘Testing the Consolidated Framework for Implementation Research on health care innovations from South Yorkshire’, Journal of Evaluation in Clinical Practice, 19(5), pp. 915–924.

Institute for Healthcare Improvement (2023) Science of Improvement: How to Improve. Boston, MA: IHI.

King’s Fund (2019) Patient Safety in the NHS. London: The King’s Fund.

Langley, G.J., Moen, R., Nolan, K.M., Nolan, T.W., Norman, C.L. and Provost, L.P. (2009) The Improvement Guide: A Practical Approach to Enhancing Organizational Performance. 2nd edn. San Francisco: Jossey-Bass.

Nanji, K.C., Seger, D.L., Slight, S.P., Amato, M.G., Beeler, P.E., Her, Q.L., Volk, L.A. and Bates, D.W. (2018) ‘Medication-related clinical decision support alert overrides in inpatients’, Journal of the American Medical Informatics Association, 25(5), pp. 476–481.

National Institute for Health and Care Excellence (2021) Medicines Optimisation: The Safe and Effective Use of Medicines to Enable the Best Possible Outcomes. NG5. London: NICE.

NHS England (2022) National Patient Safety Incident Reports: Medication Incidents. Leeds: NHS England.

NHS Improvement (2018) Medicines Safety Improvement Programme. London: NHS Improvement.

Provost, L.P. and Murray, S.K. (2011) The Health Care Data Guide: Learning from Data for Improvement. San Francisco: Jossey-Bass.

Reason, J. (2000) ‘Human error: models and management’, BMJ, 320(7237), pp. 768–770.

Vincent, C. (2010) Patient Safety. 2nd edn. Chichester: Wiley-Blackwell.

World Health Organization (2017) Medication Without Harm: WHO Global Patient Safety Challenge. Geneva: WHO.