Strengths and weaknesses of the study
Loss to follow-up in NOIS is lower than that reported in many other studies (12) – (15). The high proportion of infections discovered after discharge emphasises the importance of proper follow-up after hospitalisation. Without such follow-up the infection rate will be underestimated, and thereby also the patient-related and economic disadvantages of such infections. In our study, 90 % of the women are completely follow-up and the hospitals made a considerable effort to contact as many patients as possible after discharge.
It has been common to include all operated patients in the denominator not only those who have been followed up after discharge an approach which assumes that hospitals receive information about all patients who develop infections after discharge (probably not the case). An Australian study showed that 32 % additional infections were identified when patients who had not answered letters from the hospital were contacted. The researchers concluded that comparison of incidence rates between studies requires a definition of the denominator, i.e. whether it includes patients who were not followed up after discharge or not (16). In this presentation of NOIS data, the analyses only include patients that have been completely followed up.
In the European surveillance protocol it is not required to follow up patients after hospital discharge. In European countries, the incidence of infections after caesarean section (before hospital discharge) varies between 0.1 % and 3.7 % (17), in our study it is 1.2 %. Patients in the European hospitals stayed for an average of 7 days after the operation, in our study they stayed for 5 days. In European hospitals that followed up patients after discharge (30 % to 89 % of patients were followed up), the incidence rates for infection varied between 7.7 % and 17.0 % (15, 16, 18) – (21); in our study the incidence rate was 8.3 % and 90 % of patients were followed up. Because conditions such as average postoperative hospital stay and methods to detect patients with infections vary between studies, comparisons must be made with caution. However, the findings can still indicate the magnitude of the problem. The incidence rate in Norway seems to lie in the lower segment of reported rates from comparable studies in other European countries.
In studies from other countries, several of the variables that are included in NOIS; such as age, emergency intervention, lack of prophylactic antibiotics and high ASA score, were significantly correlated with infection (7, 15, 19, 20, 22, 23). In our study, age above 29 years and SSI grade 3 were significantly correlated with development of postoperative SSI. It was difficult to use NNIS (the US risk index) to identify caesarean sections with a high risk in NOIS. This may be explained by the low number of caesarean sections studied, incorrect interpretation or coding of risk variables used in the index, or that the nearly 20-year-old US risk index is not applicable in Norway today. In future surveillance periods we will include the variables height and weight (and thereby also body mass index), as well as diabetes, to see if these factors influence the risk of infections and are suitable for inclusion in a risk index classification. This may improve the basis for comparisons between hospitals and ease the identification of risk patients, so that prevention of infections can be targeted better.
The risk of infection varied considerably between hospitals, also when other variables where taken into account, even though few hospitals had statistically significant lower or higher incidence rates than a hospital with an incidence rate equal to the country average (tab 4). However, the study had low power to detect such differences. We can assume that local factors at the hospitals and individual factors among the surgeons contribute to the variation in incidence of infections. Such factors are not included in the national dataset, but can be analysed locally at each hospital.
It is also possible that such local factors may confound the correlation between some of the measured variables and infection. For example, a surgical team may have a combination of long duration of surgery and poor technique. In this case it may be the technique, and not the duration, that causes the infection. In the national dataset we can only analyse at a hospital level, not lower levels. Taking hospitals into consideration had marginal effect on the other variables. The validity of determination of diagnoses and understanding of the variables have not been studied. However, we see examples of obvious mistakes, such as an operation with a duration of 3 minutes. It is uncertain whether, and to what degree, misinterpretation of these conditions may have influenced the results. We are working continuously to improve data quality and investigate the sensitivity of infection diagnostics.