Table 4.
Negative Binominal and System GMM Models Explaining MRSA Incident Counts
| Model 1 | Model 2 | |||
|---|---|---|---|---|
| Fixed Effects Negative Binominal Estimates | System GMM Estimates | |||
| IRR | 95% CI | Coeff. | 95% CI | |
| Control variables | ||||
| Lagged MRSA incident count | 1.0004 | [0.9983, 1.0024] | 0.6765*** | [0.5409, 0.8121] |
| Trust size | 0.8401* | [0.7356, 0.9595] | 10.3952* | [2.0964, 18.6941] |
| Foundation trust | 0.9694 | [0.8844, 1.0626] | 3.0843 | [−1.0350, 7.2036] |
| Specialist trust | – | 16.9482 | [−10.4058, 44.3023] | |
| University affiliation | – | 9.0002 | [−11.1853, 29.1858] | |
| Trust location dummies | – | Yes | ||
| Times dummies | Yes*** | Yes*** | ||
| Patient mix characteristics | ||||
| Severity of illness | 0.9875 | [0.8758, 1.1134] | −2.7412 | [−6.4184, 0.9361] |
| Emergency admission ratio | 0.8994 † | [0.8072, 1.0021] | −0.9520 | [−5.6264, 3.7223] |
| Length of stay | 0.8908 | [0.7684, 1.0327] | 1.7419 | [−3.3311, 6.8149] |
| Age at admission | 1.1368 | [0.8996, 1.4367] | 0.7886 | [−5.8281, 7.4052] |
| Resource endowments | ||||
| Medical workload | 0.9194 † | [0.8444, 1.0011] | 0.9541 | [−1.7401, 3.6484] |
| Nursing workload | 1.0217 | [0.9377, 1.1132] | 0.5980 | [−2.6846, 3.8807] |
| Single room ratio | 0.9468* | [0.8997, 0.9963] | −0.0608 | [−2.2308, 2.1092] |
| Error reporting climate | 0.9455* | [0.8956, 0.9981] | −2.0607* | [−4.0864, −0.0350] |
| Infection control practices | ||||
| General cleaning | 0.9535* | [0.9116, 0.9973] | −3.5265* | [−6.6120, −0.4409] |
| Hand hygiene | 0.9135** | [0.8617, 0.9684] | −2.7206** | [−4.7616, −0.6796] |
| Infection control training | 1.0296 | [0.9868, 1.0742] | −3.2910*** | [−5.2175, −1.3645] |
| Trusts | 171 | 173 | ||
| Years | 4 | 4 | ||
| Total observations | 684 | 692 | ||
| Log likelihood | −1,522.0523*** | – | ||
| Wald chi‐squared | – | 827.46*** | ||
***p < .001; **p < .01; *p < .05; † p < .10.
Model 1: IRR based on standardized negative binominal estimates for panel data reported. Yearly number of bed days used as exposure variable. 95% confidence interval in parentheses.
Model 2: Coefficient estimates for standardized independent variables reported. All variables for resource endowments and infection control practices as well as emergency admission ratio and length of stay are treated as endogenous. All other variables are treated as exogenous. The Hansen J‐test of overidentifying restrictions and the Arellano–Bond test for zero second‐order autocorrelation remain insignificant. The dependent variable in both models is the unstandardized yearly MRSA incident count in each organization.
GMM, generalized method of moments; IRR, incidence rate ratios; MRSA, methicillin‐resistant Staphylococcus aureus.