Multiple segmented multilevel Poisson regression models showing changes in nosocomial MRSA rates for the subgroup analysis of clean surgery only*
| Variable | MRSA clinical isolates (per 100 susceptible patients) | Total MRSA infections (per 100 admissions) | MRSA surgical site infections (per 100 procedures) | ||||||
|---|---|---|---|---|---|---|---|---|---|
| aIRR | 95% CI | p Value | aIRR | 95% CI | p Value | aIRR | 95% CI | p Value | |
| Baseline phase | |||||||||
| Trend | 1.05 | 0.93 to 1.18 | 0.41 | 1.10 | 0.94 to 1.28 | 0.23 | 1.11 | 0.93 to 1.33 | 0.26 |
| Intervention phase | |||||||||
| Change in level | |||||||||
| Enhanced hand hygiene | 1.31 | 0.75 to 2.30 | 0.34 | 1.06 | 0.52 to 2.16 | 0.88 | 1.09 | 0.47 to 2.53 | 0.83 |
| Screening and decolonisation | 0.87 | 0.41 to 1.85 | 0.71 | 1.03 | 0.39 to 2.69 | 0.96 | 0.92 | 0.29 to 2.92 | 0.89 |
| Combined | 1.79 | 0.86 to 3.74 | 0.121 | 1.15 | 0.44 to 2.96 | 0.78 | 1.21 | 0.39 to 3.73 | 0.75 |
| Change in trend | |||||||||
| Enhanced hand hygiene | 0.89 | 0.78 to 1.01 | 0.063 | 0.88 | 0.75 to 1.04 | 0.127 | 0.89 | 0.73 to 1.07 | 0.21 |
| Screening and decolonisation | 0.85 | 0.74 to 0.97 | 0.019 | 0.83 | 0.69 to 0.99 | 0.041 | 0.81 | 0.66 to 1.00 | 0.054 |
| Combined | 0.82 | 0.71 to 0.95 | 0.007 | 0.84 | 0.70 to 1.00 | 0.055 | 0.84 | 0.68 to 1.03 | 0.095 |
| Washout phase | |||||||||
| Change in level | 3.01 | 1.05 to 8.63 | 0.041 | 2.21 | 0.61 to 8.04 | 0.23 | 2.59 | 0.59 to 11.46 | 0.21 |
| Change in trend | 0.96 | 0.81 to 1.15 | 0.67 | 0.91 | 0.73 to 1.12 | 0.37 | 0.86 | 0.67 to 1.09 | 0.21 |
aIRR, adjusted incidence rate ratio; MRSA, meticillin-resistant Staphylococcus aureus.
*MRSA was defined as nosocomial if it was isolated from specimens collected more than 48 h after admission or within 30 days (or 12 months for infections of prosthetic devices) after discharge from the surgical ward. Clean surgery included cardiothoracic, neuro, orthopaedic, plastic and vascular surgery subspecialties. The model used a lagged dependent variable to account for autocorrelation and adjusted for staffing (patient-to-nurse ratios), seasonal effects, type of surgical ward and baseline hand hygiene compliance rates. The model also accounted for overdispersion. Random effects for intercepts at the hospital and ward levels and random baseline trends at the hospital level were all significant and baseline trends were negatively correlated with intercepts (ie, hospitals with higher baseline MRSA rates tended to have larger decreases in baseline rates).