Improving Arterial Blood Gas Sampling Practice in the Emergency Department: A Two-Cycle Clinical Audit

Background Arterial blood gas (ABG) sampling is commonly performed in emergency care but is invasive and painful, particularly when repeated sampling is required. In selected patients with acute respiratory disease, including chronic obstructive pulmonary disease (COPD) exacerbations, venous blood gas (VBG) analysis can provide useful screening information and may avoid unnecessary arterial puncture. A venous PCO₂ below 6.0 kPa, when interpreted alongside satisfactory pulse oximetry and the clinical context, can help exclude significant arterial hypercapnia, while venous pH is generally sufficiently close to arterial pH for initial assessment. Despite this, clinicians may still default to arterial sampling, and measures intended to reduce procedural pain, such as local anaesthesia, are not always routinely used. We evaluated whether a targeted educational intervention could improve clinicians' knowledge and reported practice relating to ABG sampling. Methods We conducted a two-cycle clinical audit in a UK emergency department. Cycle 1 included 59 clinicians and identified baseline gaps in VBG knowledge, local anaesthetic use and patient-centred sampling practice. A multimodal intervention was subsequently introduced, combining departmental teaching, a short educational video, visual prompts and reinforcement during clinical work. Cycle 2 included 30 clinicians. Prespecified measures assessed knowledge of when VBG may provide adequate initial information, routine skin antisepsis and local anaesthetic use. Ultrasound use for difficult arterial access and enquiry about patient pain or experience were also assessed. As responses were anonymous and could not be paired between cycles, the two samples were analysed independently. Results Mean VBG knowledge increased from 51.7% (n/N = 61/118) in Cycle 1 to 75.0% (n/N = 45/60) in Cycle 2, an absolute improvement of 23.3 percentage points (95% CI 7.2-39.4; p = 0.005; Hedges g = 0.61). The proportion of clinicians answering both VBG knowledge items correctly increased from 32.2% (n/N = 19/59) to 60.0% (n/N = 18/30). Recognition that a normal venous pH and PCO₂ does not automatically require arterial confirmation increased from 67.8% (n/N = 40/59) to 90.0% (n/N = 27/30), while recognition of the potential role of VBG with pulse oximetry increased from 35.6% (n/N = 21/59) to 60.0% (n/N = 18/30). Local anaesthetic use increased from 6.8% (n/N = 4/59) to 23.3% (n/N = 7/30) (absolute difference 16.6 percentage points; 95% CI 1.7-34.6; p = 0.039), although it remained the largest gap from the prespecified standard. Routine skin antisepsis remained consistently high at 96.6% (n/N = 57/59) versus 96.7% (n/N = 29/30). Enquiry about patient pain or experience increased from 50.8% (n/N = 30/59) to 80.0% (n/N = 24/30) (difference 29.2 percentage points; 95% CI 7.9-45.4; p = 0.011). Ultrasound use for difficult arterial access increased from 47.5% (n/N = 28/59) to 66.7% (n/N = 20/30). Conclusions After the multimodal intervention, VBG knowledge and several aspects of reported ABG practice improved. Local anaesthetic use remained low despite some improvement, suggesting that education alone may not be enough to change this procedural behaviour. Appropriate use of VBG in selected patients, together with better analgesia when ABG sampling is required, may reduce unnecessary arterial puncture and procedural pain. Future cycles should directly measure ABG indications, procedural practice and patient-reported pain.