Serratia sp., formerly considered to be a benign commensal only, is now recognized as an important cause of outbreaks, especially in neonatal intensive care units 16,22. According to our knowledge this is the largest published outbreak in terms of case numbers and duration.
Our study showed that gestational age, antimicrobial therapy and arterial lines are significantly associated with colonization and infection with Serratia marcescens.
The development of the intestinal microflora starts at birth and is influenced by various factors such as gestational age, mode of delivery, local environment, type of feeding and antibiotic treatment [13, 14].
The role of antibiotics as a risk factor has been investigated in several similar case–control studies [5, 6, 9, 10]. Statistically significant associations between previous antibiotic treatment of mothers [5, 6] and neonates  have been found by some authors. It is known that antibiotic treatment may affect the colonization with gram-negative bacteria. Antibiotic therapy suppresses formation of normal flora, reduces remarkably colonization resistance and increases risk of acquiring a nosocomial strain . It is difficult to estimate to what extent antibiotic treatment influenced our outbreak. We suppose that our empirical antibiotic treatment didn`t influence the acquisition of Serratia significantly, the major factor was rather the poor compliance to infection control measures. We did not change the guidelines of empirical antibacterial treatment at a time of the outbreak, because in similar studies its significance in terminating the outbreak has remained unclear .
The second finding was the presence of an arterial line as an independent risk factor for colonization and infection with Serratia marcescens. Other studies have also shown invasive intravascular devices as risk factors for colonization and infection with Serratia sp.[5, 17, 18]. However, Maragakis et al.  suppose that these are rather markers of severity of illness and susceptibility of a patient to infection than independent risk factors. We also agree with this statement and presume that the usage of these devices has probably no direct causal role in colonization with Serratia marcescens. On the other hand, invasive device use increases the number of contacts with personnel and thus, in case of suboptimal hand hygiene, increases the possibility of transmission. In order to improve insertion and care of intravascular catheters, including arterial catheters, a guideline was drawn up and implemented.
As in most of the other outbreak studies, the reservoir and mode of transmission remained unknown also in our study. We suppose that asymptomatic patients are the most important reservoir of Serratia sp., because after the implementation of screening policy (included samples from nasopharynx and rectum) and isolation precautions the number of new cases decreased significantly. As in our study the gastrointestinal tracts as well as respiratory tracts of neonates were frequently colonized, we agree with Giles et al.  that during Serratia sp. outbreak both respiratory and gastrointestinal samples should be collected in order to maximize the identification of colonized infants.
There is a likelihood that the organisms may have been transmitted via the hands of health care workers and mothers. This hypothesis is supported by the evidence that no other cultures taken from the environment were positive. However, cultures taken from the hands of health care workers were negative. In 2001–2003 a study conducted in New York Presbyterian Hospital in New York City aimed to determine the relative frequency of potential horizontal transmission between patients and health care workers in the NICU. This large prospective study indicated a potentially higher probability of cross-transmission with certain gram-negative bacilli, such as Klebsiella pneumoniae and Serratia marcescens via the hands of health care workers . We speculate that in addition to personnel, mothers may have had a role in transmission of Serratia sp. in the ward. Hand hygiene rules were explained to mothers, but their compliance was not followed routinely. Unfortunately no samples were cultured from mothers’ hands during the outbreak.
Closing the ward temporarily has been an efficient measure to terminate an outbreak [6, 21]. This option was under discussion in our case, but it was not possible to transfer intensive care patients to another hospital (due to long distance) or to open an additional ward (lack of qualified personnel). It was also impossible to employ additional personnel.
Limitation of our study concerns the aspect that among control patients there could be patients colonized with other bacteria due to the fact that a selective culture for Serratia sp. was used for screening. It is possible that previous colonization with other gram-negative bacteria causes colonization resistance of the gastrointestinal tract and reduces the opportunity to colonize with potentially pathogenic microorganisms such as Serratia marcescens. On the other hand, as the length of stay in PICU was 12.4 days till the first positive culture and in NU 11.8 days, there is a possibility that case-patients were colonized with other gram-negative bacteria also by that time.
In conclusion, we can say that as the implementation of infection control measures was delayed, the outbreak lasted a long time. In order to terminate an outbreak caused by Serratia sp. promptly, early implementation of thoroughly considered aggressive infection control measures involving patients and mothers as well as the personnel is of utmost importance. Continued attention should be paid on optimizing antibiotic usage.