Power limited by small number of transmission events ? Chris Mansell, Waikato Hospital, New Zealand 24 March 2010 Thank you to the authors for describing an elegant mathematical approach to this "difficult to evidence" issue. Some more explanation of the data augmentation step would be of interest. Number of patients followed: 8552 Mean length of stay: 3.4 - 5.8 days (my rough point estimate: 4.5 days) Median length of stay: 1 or 2 days Mean number of swabs per patient: 1.2 - 1.5 Estimated colonisation rate on admission: 4.5 - 20.6% Estimated incidence: 5.8 - 18.2% per month (my rough point estimate: 10% / 30d = 0.3% per day) Note also that culture based MRSA screen methods typically take 2 days to yield presumptive positive results. Therefore: 1) Most patients will have left the ICU long before an unexpected positive result would have been available. 2) Most patients wouldn't have had a second swab during their short stay in the ICU. 3) The approximate number of transmission events would have been about 8552 x 4.5 x 0.003 = 115 ? Most observed transmission events were presumably in the small number of long stay patients, who would have been biologically different from the majority. The low number of events observed may have provided only weak constraints on the parameters modelled. This may be why the confidence intervals are so large, especially when stratified among each of the 8 wards. I also suspect that random events in the stochaistic modelling process have blurred some of the transmission into being attributed to the "background" (beta 0) and "isolated" (beta 2) groups. I hope that this is why we have the shocking picture seen in Figure 2, where isolation precautions are calculated to have an insignificant effect on MRSA transmission. The confidence intervals obtained are too wide to address the provacative trends reported, which suggest that MRSA acquisition from environmental contamination and staff carriers predominates and that failure of isolation precautions is also a major threat, comparable to that due to undetected carriers. The mathematical method appears intuitively plausible; its' empiric robustness might be better demonstrated using a smaller patient sample followed over a longer exposure and screening period, for example the complete hospital stay from admission to discharge. Yours, Dr Chris Mansell MB, ChB FRCPA Clinical Microbiologist Competing interests No competing interests.