This was a pre–post observational audit study conducted at a private level 1 trauma hospital in Gauteng, South Africa, using the PAIN OUT registry. The national PAIN OUT South Africa study included 11 sites. The perioperative pain management bundle was introduced at this site in April 2023.

Eligible patients were ≥ 18 years of age, admitted for orthopaedic surgery and able to provide informed consent. Exclusion criteria included deafness, cognitive impairment, inability to communicate in English, Afrikaans, isiXhosa or isiZulu, or refusal to participate:

Baseline recruitment was closed in March 2023 to allow bundle implementation, despite not reaching the target sample size of 100. Numbers of excluded patients was not consistently recorded; this is acknowledged as a limitation (Figure 2).

Clinicians were trained in bundle principles via infographic handouts and structured sessions led by a physiotherapist and data collector. Training covered pain assessment, documentation and treatment.

Data were collected on postoperative day 1, after patients had returned to the ward for ≥ 6 h, by a trained data collector per the PAIN OUT protocols.⁸ Data were collected using a Process and Outcomes questionnaire:

Data were entered into the PAIN OUT registry, which assigns unique identifiers to ensure confidentiality and anonymity.

A pragmatic audit approach was adopted. The target sample size of 100 patients per phase was not achieved at baseline (n = 86) because of recruitment challenges. Recruitment was not extended further because of resource constraints and hospital workflow limitations.

Registry data were reviewed and classified according to perioperative phases and classes of analgesia. The primary outcome was postoperative pain severity. Secondary outcomes included bundle adherence, pain interference, affective impairment, adverse effects and patient perceptions of care.

Descriptive statistics (median [interquartile range {IQR}]) summarised demographics, comorbidities, chronic pain profile, duration and type of surgery and bundle elements based on the non-normal distribution of the data. Categorical data were presented as percentages. Mann–Whitney U tests assessed whether postoperative pain severity decreased when all four bundle elements were implemented. Cohen’s d was calculated (Z scores) when p < 0.05, to establish effect size. Chi-square (χ²) tests compared categorical variables between audit phases.

Regression analysis was initially planned, but not feasible because of missing baseline data on non–opioid use. Missing data were handled by case–wise exclusion; no imputation was performed. Potential confounders (comorbidities, surgical complexity and chronic pain status) were described but not formally adjusted for because of limited sample size and non–randomised design.

Ethical approval for the national PAIN OUT registry, including this hospital site, was granted by Stellenbosch University (Ref: N19/10/140) with approval from the local hospital ethics board. Approval for the extraction and analysis of the collected data for this study was granted by the University of Cape Town (Ref: 455/2024). The study complied with the ethical principles of the Declaration of Helsinki.⁹

There were no statistically significant difference between participants in the baseline and post-implementation groups for age, weight or height (Table 1). The majority of participants completed the survey in English, with no statistically significant difference between groups in the language used to complete the survey.

Of the 89 (46.60%) participants presenting with comorbidities, the majority had more than one comorbidity. The most common presenting comorbidity was hypertension (n = 66; 34.55%) (Table 2). There was no statistically significant difference between groups in the number of comorbidities or in pre-existing chronic pain.

The participants had undergone a range of orthopaedic surgical procedures with no difference between groups. They spent a median of 1.67 h (1.15–2.25) in theatre with participants in the follow-up phase of the study, spending significantly less time in theatre (1.38 h [0.97–2.08] vs 2.02 h [1.52–2.50]; U = 2965.50; p < 0.01, Cohen’s d = 0.27). The participants had been back on the ward for a median of 24.33 h (21.52–27.32).

Following the introduction of the perioperative pain management bundle, adherence to its components improved substantially (Table 3). Almost all patients in the post-implementation group received education on pain management (98.1% vs 65.1%; p < 0.01), and regular pain assessment increased markedly (95.2% vs 58.1%; p < 0.01). Use of regional anaesthesia or wound infiltration rose from 30.2% at baseline to 100% post-implementation (p < 0.01), and 72.4% of patients received the maximum dose of non-opioid analgesics (baseline data unavailable).

Overall, 65.7% of patients received all four bundle elements compared to none at baseline (p < 0.01). When considering only the three elements with complete data, 94.3% of post-implementation patients received all three versus 11.6% at baseline (p < 0.01), indicating a significant improvement in bundle delivery.

Implementation of the bundle was associated with a significant improvement in overall pain composite scores (median 1.75 vs 2.25; p < 0.01) with a small effect size (Cohen’s d = 0.29) (Table 4). Although pain intensity scores did not differ significantly between groups, pain interference decreased markedly (median 0.25 vs 1.38; p < 0.01) with a moderate effect size (Cohen’s d = 0.48), indicating better functional recovery. Sleep interference showed the largest improvement (Cohen’s d = 0.52).

Side effects of pain management also improved, with reductions in dizziness and drowsiness (overall side effect score median 0.75 vs 1.25; p < 001; Cohen’s d = 0.30). Additionally, fewer patients requested additional analgesia post-implementation (21.9% vs 48.8%; p < 0.01), suggesting greater satisfaction with pain control.

Strengths include participation in an international registry, enabling benchmarking and standardised data collection.^5,8 Limitations include missing baseline data for non-opioid dosing, small sample size and the non-randomised design, which introduces potential confounding. Baseline recruitment was extended for 16 months but closed to enable the timely implementation of the intervention. Differences in case mix and comorbidities cannot be fully excluded as alternative explanations for observed improvements.^16,17 Future research should include larger samples, complete data capture and adjustment for confounders.