Support to HAI-Net coordination and analysis of HAI-Net data ECDC collaborative project

A common and standardized approach to the surveillance of health-care associated infections (HAI), as insured by ECDC HAI-Net, is key for benchmarking and improving the overall standard of infection control across European hospitals.

Intensive care unit-acquired nosocomial infections (ICU-NI) are thought to increase patient mortality. However, the magnitude of this effect remains controversial and depends on study design, type of infection and target population. Previous investigations have reported death occurrence related to ICU-NI in excess of 4% and 50%. The corresponding relative risks (RR) of death due to ICU-NI ranged from 1.4 to 4.0 and from 1.7 to 3.2 when odds ratios (OR) were used. Different models have been used to estimate attributable mortality related to HAI.

We propose to assess different statistical approaches currently used with emphasize on the most suitable one for data collected by observational studies i.e. by surveillance networks of HAI.

The objective of this collaborative project between ECDC and the Department of Epidemiology and Public Health is to use the HAI- ECDC data to estimate the potential excess of hospital stay for patients with HAI compared to those without HAI. Nested case-control studies will be set-up to assess the excess of hospital stay. Multistate modelling and adjusted survival model will be used, with HAI as the time-dependent variable and taking into account the informative censure at discharge (survivors, death).

The work will be based on two majors HAI, i.e., nosocomial pneumonia (NP) and blood stream infection (BSI) in order to:

                - Estimate the attributable mortality of Intensive Care Unit (ICU)-acquired pneumonia or BSI   in Europe i.e. estimate of the total effect of NP or BSI on mortality including direct and indirect effects after the onset of these infections. This means that the effects of NP or BSI acquired at time t on further Death mediated by some time-varying covariates after t must be taken into account;

                - Estimate of the extra length of stay (LOS) attributable to NP or to BSI;

Analysis will be performed by stratifying on the type of nosocomial infection, age, clinical characteristics, and severity scores at hospital admission. Descriptive analysis of the population, calculation of incidence density by type of nosocomial infection, and mortality rate will be carried out. Statistical analyses will include i) the mean LOS in each stratum in both cases and controls; ii) LOS between the onset of infection and discharge/death in cases; iii) LOS between the onset of infection (cases) and discharge (controls); differences between mean duration of hospital stay (cases and controls); iv) mean of excess of hospital stay for all cases and v) excess of hospital stay according to the date of the occurrence of infection.