Sir: We read with interest the article by Eggimann et al. [1]. Their results contribute to the growing understanding of the risk of biased risk assessment and jeopardized benchmarking between units. Results of our own unit support and add to this theory. We evaluated the incidence of ventilator associated pneumonia (VAP), definitions were based on the recommendations by the Centers for Disease Control, and the adherence of the staff to local prevention protocols in our 12-bed ICU during a 3 month follow up on a yearly basis. The ICU has a predominantly medical patient population, a substantial number of immune-compromised patients, and a large caseload of critically ill patients transferred from other hospitals. We identified one definite and one probable VAP/1000 ventilator days (Table 1).

Table 1 Patient characteristics and reasons for admission during a 3-month observation period
Full size table

Endotracheal mechanical ventilation carries a low event rate of VAP in the described population. Considering that a substantial number of our patients are immunocompromised the results may be even more impressive. Whether our preventive strategies are perhaps ‘the best’ is open for discussion. Our study could have suffered from several limitations. Firstly, in our experience, most pulmonary infiltrates are not caused by pneumonia but rather by oedema or atelectasis. A low clinical likelihood of VAP encourages and reinforces a policy to withhold diagnostic procedures in patients with ‘little’ clinical evidence of infection [2]. Secondly, seasonal variation or the applied study design (3 month follow up from April to June) may have influenced the low incidence of VAP. Thirdly, although our nursing staff has specific knowledge of non-pharmacological VAP preventive measures, it is unclear whether any of our non-pharmacological measures or combination of measures really impact on outcome.

Despite the low infection rate, our findings do not add anything new to the field of nosocomial infections. In fact, any given result is due to chance alone. It might well be that there is a difference in preventive measures but that this difference cannot be detected with statistical certainty.

Longitudinal surveys with continuous risk estimation are often initiated to allow targeting treatments to those with the highest risk tailored to the characteristics of the individual ICU [3]. But due to methodological weakness, almost every single unit observational study on VAP is underpowered, such a study does not demonstrate the value of local measures and distinguishes associations. In our view the emphasis of quality control programs should be on structural and process indicators instead of on outcome parameters. Only prospective, randomised trials conducted to compare preventive measures with other strategies with clearly defined patient populations and disease stages can identify valuable preventive measures. Such trials should be well powered, and therefore multi-centered by design, with extreme attention and strict adherence to the protocols that would be compared, their results may help in formulating guidelines to be generally applied.

In conclusion, discrepancies between ICUs may reflect: effectiveness of local VAP diagnosis and prevention strategies, the prevalence of resistant isolates, patient to staff ratio, the patient population examined, the way of reporting VAP rates, and other unknown factors. Data of local surveillance studies should be read as process indicators. Due to limited sample size most of these studies are insufficiently powered and the results do not permit a definitive conclusion. Failure to recognize these issues will lead to compromised benchmarking between units [1].