Chronic obstructive pulmonary disease (COPD) is among the most common chronic illnesses encountered in internal medicine practice, responsible for high rates of morbidity, mortality, and healthcare utilization. Acute exacerbations of COPD (AECOPD) contribute disproportionately to this burden, representing up to 70% of hospital admissions among COPD patients and often resulting in prolonged hospital stays and avoidable readmissions [¹,²,³,⁴]. Within our institution, informal staff discussions and case reviews identified inconsistencies in the inpatient management of AECOPD, particularly in the timely administration of key therapies, adherence to oxygen titration guidelines, and discharge planning.

National and international guidelines, particularly the GOLD and NICE NG115, outline best practices for AECOPD management [⁵,⁶,⁷,⁸]. These include early arterial blood gas (ABG) testing, rapid administration of short-acting bronchodilators and corticosteroids, judicious antibiotic use based on sputum characteristics and clinical findings, controlled oxygen delivery targeting SpO\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$_2$\end{document}2 88–92%, timely imaging to exclude other diagnoses, and referral to pulmonary rehabilitation. Despite this guidance, studies suggest that real-world practice often diverges from these standards due to a variety of system- and provider-level barriers. Recent reviews and audits have highlighted frequent issues with oxygen overuse, poor documentation, and under-referral to pulmonary rehab—findings consistent with our internal observations [⁹,¹⁰,¹¹,¹²,¹³,¹⁴].

Our team hypothesized that many of the observed care gaps stemmed from workflow inefficiencies and missed clinical prompts during high-pressure admissions. Given that our hospital had already adopted electronic prescribing and documentation systems, we reasoned that integrating clinical decision support tools, such as automatic prompts or checklists, into existing digital workflows could enhance compliance with evidence-based standards. In addition, educational reinforcement through departmental meetings and targeted communication was anticipated to improve clinical staff engagement and awareness.

The primary aim of this quality improvement project was to assess our baseline compliance with key process standards for AECOPD management and implement low-cost, replicable interventions to address identified care gaps. We also aimed to measure the effectiveness of these interventions through a structured re-audit and generate actionable insights for future system-wide changes.

The project took place on a 650-bed tertiary academic hospital general medicine ward with an existing integrated EHR and in-house clinical informatics support. The clinical team included on-call respiratory specialists, ward physicians, nursing staff, physiotherapy (pulmonary rehabilitation), pharmacy, and an antimicrobial stewardship pharmacist who participated in the QI steering group.

We included consecutive adult patients (≥18 years) admitted to the general medicine ward with a primary diagnosis of acute exacerbation of COPD (AECOPD). Case identification relied on either clinician documentation of “COPD exacerbation” in admission or ED records or ICD-10 coding consistent with AECOPD (J44.1). Exclusion criteria were: primary diagnosis other than AECOPD, invasive mechanical ventilation instituted prior to the EHR trigger or admission for palliative care only. Some process measures had reduced denominators due to documentation availability (for example, antibiotic indication was not appraisable in 3 baseline patients, and smoking status was unavailable in 10 patients); we report the denominators for each measure in Tables 2–3 and describe the handling of incomplete documentation in the Results.

Based on root cause analysis and staff feedback, four main interventions were introduced between November and December 2024. First, we implemented an oxygen prescription checklist embedded into the electronic prescribing system to improve documentation of target SpO₂ ranges and FiO₂. Second, we developed an automatic referral prompt for pulmonary rehabilitation that triggered at discharge for eligible patients. Third, we delivered guideline-focused education sessions to all medical and nursing staff. Finally, a pop-up reminder was added to the discharge summary module to ensure that documentation of smoking cessation advice is provided for current smokers. These interventions were co-designed with the respiratory team, clinical informatics unit, and physiotherapy department.

The electronic orderset recommended a metered-dose inhaler (MDI) with a spacer as the preferred bronchodilator delivery method because of equivalent efficacy and logistical advantages (including lower aerosol generation). Nebulized bronchodilator therapy remained available and was permitted when MDI/spacer was not feasible or tolerated (for example, where patients were unable to coordinate inhalation, were highly agitated, or at the clinician’s discretion). We clarified this preference within the orderset and included a Table 1 footnote to reflect the local policy.

The EHR prompt did not automatically prescribe antibiotics. Instead, the prompt and orderset presented guideline-based criteria to prompt clinician consideration of antibiotics (presence of increased sputum purulence, increased sputum volume, increased dyspnoea, temperature > 38°C, or radiographic consolidation). Final prescribing decisions remained at the discretion of the treating clinician. The hospital antimicrobial stewardship pharmacist subsequently reviewed all antibiotic prescriptions for the audit cohorts during the QI period, and stewardship feedback was communicated to treating teams when prescriptions appeared discordant with the indicated criteria.

The orderset included a recommended short-course systemic corticosteroid regimen consistent with contemporary guideline practice: oral prednisolone 40 mg once daily for 5 days (or an equivalent intravenous regimen when oral administration was not possible). An informational notice within the orderset prompted clinicians to review recent corticosteroid prescriptions in the record to avoid unnecessary cumulative oral corticosteroid exposure and to discourage prolonged tapers unless clinically justified.

Targets for each process indicator were selected by a multidisciplinary QI steering group (respiratory physicians, ward physicians, nurses, pharmacy, physiotherapy, and clinical informatics) using a pragmatic balance of ambition and feasibility informed by baseline performance. We used staged compliance thresholds (85% → 90% → 95%) to secure early, achievable wins and then progressively raise them in subsequent Plan-Do-Study-Act (PDSA) cycles. The choice of these operational targets was therefore locally informed rather than derived from a single external numerical standard; this approach followed the Model for Improvement and aimed to sustain momentum through iterative improvement.

The QI project employed the Model for Improvement, utilizing rapid Plan-Do-Study-Act (PDSA) cycles. A multidisciplinary team (physicians, nurses, pharmacists, physiotherapists, and informatics specialists) met weekly during the implementation to review process measures, frontline feedback, and iterate on prompt wording, order defaults, and workflow. Figure 1 summarises the PDSA timeline.

We employed a before-and-after audit design to assess the effectiveness of the intervention. Data were collected retrospectively for the baseline cohort (July–October 2024) and prospectively for the re-audit cohort (January–March 2025). Observed outcomes were compared to established targets to assess changes in compliance. Attribution of outcomes to interventions was supported by temporality, stability of external conditions, and qualitative feedback from clinical staff. Allocation and blinding. This was an uncontrolled before-and-after QI audit, and there was no randomization or blinding. Staff were aware of the interventions and audits (education sessions and EHR prompts), which may have influenced behaviour (see Limitations). To enhance transparency and reproducibility, a visual summary of the QI cycle is provided in Figure 1. This flowchart outlines the sequential steps of the quality improvement process, beginning with the baseline audit conducted from July to October 2024, followed by the identification of care gaps, the co-design of targeted interventions, and the implementation phase. The cycle concludes with the re-audit from January to March 2025 and subsequent planning for sustainability and scale-up. The figure illustrates how each phase informed the next, in line with the iterative principles of quality improvement methodology.

Eight process indicators were selected in accordance with international guidelines. These included: ABG within one hour of admission; nebulized bronchodilator administration within one hour; systemic corticosteroid prescription within four hours; appropriate antibiotic use based on clinical indication; oxygen therapy titrated to target saturation of 88–92% with documentation of FiO₂; chest X-ray within four hours; pulmonary rehab referral at discharge; and smoking cessation counseling for current smokers as shown in Table 1. Their guideline origins supported the validity of these measures, and data were abstracted using a structured form. Completeness and accuracy were assured through dual validation by two clinicians. Implementation fidelity and monitoring. We tracked fidelity using multiple process indicators: staff education attendance (percentage of targeted staff attending training sessions; 100% coverage reported); EHR log data confirming the display of the bedside prompt for triggered admissions; and orderset opening/acceptance rates (EHR log metric). In 88% of triggered admissions, the orderset was opened and used. These fidelity metrics are summarised in Supplementary Table S1.

Descriptive statistics were used to summarize performance against each indicator. Pre- and post-intervention comparisons were made using percentage point differences. Confidence intervals were calculated for proportions. Given the sample size of 50 per audit cycle, we prioritized changes of more than 5% as potentially meaningful. Variations over time were examined through comparative tabulation. Analyses were primarily descriptive. For each process measure, we calculated proportions with 95% Wilson confidence intervals. For pre- and post-comparisons, we used two-sided Fisher’s exact tests, given the sample sizes and the presence of small cell counts in some measures. We did not apply multiplicity correction because these analyses were exploratory and hypothesis-generating; we therefore interpret p-values cautiously, mindful of the potential for type I error across multiple comparisons. No formal prospective sample-size calculation was performed; this activity was planned and conducted as a pilot/service evaluation to inform larger, more definitive future studies.

The project was approved as a service evaluation and registered with the hospital’s clinical audit department (Registration# 2025-202). Formal ethics review was waived in accordance with institutional policy, as the project involved routine care data with no patient identifiers. The authors declared no conflicts of interest.

This quality improvement project is reported in accordance with the Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0) guidelines [¹⁷]. The SQUIRE framework is specifically designed to guide the reporting of systematic efforts to improve healthcare quality, safety, and value in real-world clinical settings. A completed SQUIRE Checklist is provided in the supplementary materials to accompany this manuscript.

At baseline (as shown in Table 2), compliance was highest for early administration of bronchodilators (92%) and initiation of corticosteroids (94%), indicating good adherence to acute pharmacotherapy. ABG sampling within one hour was achieved in 82% of patients. However, oxygen titration to SpO₂ 88–92% was documented in only 80% of cases, and only 64% of eligible patients were referred to pulmonary rehabilitation at discharge. The order set recommendation for a 5-day short course was followed in the majority of patients: 94% of treated patients received the recommended prednisolone 40 mg daily for 5 days. During the audit periods, 24 patients required noninvasive ventilation. A chest X-ray was obtained within four hours in 78% of cases. Among current smokers, documentation of cessation advice was noted in 87.5% of cases.

Following the interventions, compliance improved across all measures (shown in Table 3). The pulmonary rehab referral rate increased to 86%, meeting the target of 85% or higher. Oxygen titration documentation improved to 92%, reflecting gains but remaining just below the 95% goal. ABG and corticosteroid benchmarks were met, reaching 90% and 98%, respectively. Smoking cessation advice documentation exceeded the 90% target, rising to 95%.

To illustrate the comparative impact of our interventions across key clinical standards, a clustered bar chart (Figure 2) was constructed showing baseline and re-audit compliance percentages for all eight audit measures. This visualization highlights notable improvements in pulmonary rehabilitation referrals, corticosteroid timing, and smoking cessation documentation, with more modest gains in oxygen titration and imaging. While most standards exceeded their target thresholds in the re-audit cycle, a few still fell short, underscoring the importance of continued iterative improvements.

To summarize overall performance during the re-audit period, a pie chart (Figure 3) displays the proportion of care standards that met or did not meet predefined targets. Six of the eight standards achieved full compliance with benchmark criteria, demonstrating broad improvement across multiple domains of AECOPD care. The two remaining standards, oxygen titration and chest X-ray timing, showed progress but did not reach full compliance, indicating areas for further refinement.

To better understand the impact of each intervention and assess progress toward predefined goals, we tracked key performance indicators aligned with our action plan. These included oxygen titration compliance, staff education coverage, pulmonary rehabilitation referral rates, and smoking cessation documentation. For each domain, we compared baseline performance with targets and re-audit results. Table 4 summarizes these metrics and highlights which objectives were fully achieved and which remained areas for continued improvement.

The following bar chart (Figure 4) displays the trajectory of four key intervention metrics: baseline status, predefined targets, and post-intervention compliance. Improvements in staff education, smoking cessation documentation, and pulmonary rehab referral rates all met or exceeded targets. Oxygen titration, while improved from baseline, fell slightly short of its target, highlighting an ongoing challenge in embedding oxygen therapy best practices into real-time clinical workflows.

The automated EHR prompts were consistently identified by staff as helpful in remembering care tasks during busy admissions. Feedback collected during clinical huddles confirmed that the oxygen checklist and discharge reminders resulted in improved documentation and care planning. Delays in chest X-ray performance were attributed to persistent bottlenecks in radiology transport and technician availability during peak hours. There were no unintended consequences, although two staff members noted occasional “alert fatigue” from excessive on-screen prompts.