The best magazine
Discussion
We found reductions in hospital-associated SAB infection rates that were associated with the intervention in 4 of the 6 states, with no change in rates in 2 states (Figure 4). The reductions in rates in the 4 states were all statistically significant. In future work, we will examine whether these reductions are large enough to translate into a conclusion that the NHHI was cost-effective.
In Tasmania, rates were already decreasing before the intervention was introduced (Figure 4), and the intervention failed to immediately decrease rates. Tasmania was already on a successful trajectory, and it may have been wiser to wait to intervene until rates became flatter, because the existing hand hygiene programs were already achieving the desired reductions in infections. We recommend plotting average infection rates over time before introducing any intervention aimed at reducing infection rates to avoid introducing potentially unnecessary interventions. However, it is possible that, without the introduction of the NHHI, the change in rates may have worsened. Therefore, the program may have been successful in maintaining a decreasing trajectory.
We used a statistical approach that examined a range of plausible patterns in infection rates, whereas previous studies have only examined a narrow range of possible patterns. One previous study used separate linear regressions before and after an intervention and then compared the 2 regression slopes. However, this ignores any change in the intercept (eg, model E in Figure 2) and will overestimate the effect of an effective intervention if the postintervention period has a positive intercept and underestimate it if it has a negative intercept. Another study (also examining the impact of the NHHI) assumed that the rates of infection were flat before the intervention (model G in Figure 2), but this will not give an accurate estimate of the intervention if rates were changing before the intervention (model F in Figure 2). If rates were already coming down before the intervention, then assuming that rates were flat will overestimate the effect of an effective intervention, whereas if rates were increasing before the intervention, then an effective intervention would be underestimated. We recommend trying a range of potential models when examining changes in infection rates over time, because using a narrow range of models can lead to false-positive or false-negative findings or biased estimates that under- or overestimate the intervention effect.
An interesting observation is that, in the 4 states where there was a reduction in rates, this reduction occurred immediately after the introduction of the initiative. Therefore, models B and C were always preferred to similar models with a delayed change, such as models G and J (Figure 2). The fast reaction to change may reflect the large change in practice with the initiative, including education and regular audits.
Our study design did not have any control hospitals that did not receive the intervention. This means that our study is vulnerable to time-dependent confounders, such as other changes to national, state, or hospital-level infection policy that may have reduced infection rates. This potential confounding was somewhat reduced, because the intervention was introduced at multiple times, which decreases the overall correlation between the intervention and other changes. However, the timing of the intervention was not randomized by hospital. A randomized intervention time would have further reduced any correlation and therefore would have further reduced the potential bias of time-varying confounders.
Changes to infection control practice and policy occurred in most of the hospitals during the study period. The timing of these changes varied between hospitals, and via interviews with infection control staff, we found no evidence of these changes occurring concurrently with the introduction of the hand hygiene intervention. Overall, we believe that such potential changes are unlikely to confound the observed associations between the NHHI and monthly SAB infection rates.
The more proximal question of whether the intervention improved hand hygiene compliance rates is not part of this article. This is because the next stage of our research is to estimate the cost-effectiveness of the intervention, and because the major costs are attributable to infections, we need to know whether and by how much the intervention reduced infections. Large amounts of money have been invested in the NHHI, so it is important to evaluate whether this money has been well spent, because this evidence should determine its continued funding.
Source: ...