Several approved protocols for the prevention of Legionella pneumonia base the type of intervention (to disinfect or not) on the level of contamination found (cfu/L). However, if the level of contamination by Legionella spp. of a water system fluctuates in a short period of time, inadequate sampling could lead to different decisions being made. To determine if there are significant variations in the bacterial count of Legionella spp., water samples were taken at different times from the same sites. Eight wards were selected from a large hospital in Southern Italy and a water sample was taken from 21 taps in each ward at the same time each day for 5 consecutive days. A Freidman test detected statistically significant differences in average Legionella spp. load over the 5 sampling days (p value<0.001). This fluctuating load can have practical implications: the Italian Guidelines recommend disinfection only for a Legionella count>10,000 cfu/L in hospitals without documented cases of disease. In the present study, the daily average loads varied, during the 5-day sampling period, above and below this cut-off (10,000 cfu/L). This means that the decision to disinfect or not would be different depending on which day the sampling was carried out. Our data suggest that, especially in health-care facilities, a single sampling would not give a realistic estimation of risk; therefore, even at lower levels of bacterial load, measures should be taken to reduce it further.