Risk-adjusted surveillance of hospital-acquired infections in neonatal intensive care units: a systematic review

Summary

Comparisons of bacteraemia incidence between neonatal intensive care units (NICUs) can identify centres with effective infection control, whose practices can be shared with other units. For fair comparisons, infection incidence must be risk-adjusted to control for differences between centres in the vulnerability of babies and the intensity of invasive procedures which can introduce infection. We reviewed risk adjustment methods for between-NICU comparisons of bacteraemia incidence, both in the published literature and in regional and national NICU infection monitoring systems. PubMed and Embase were searched for studies reporting risk-adjusted bacteraemia incidence in more than one NICU. An internet search found NICU infection monitoring systems in Western industrialised countries. In all nine studies that met the inclusion criteria, risk adjustment reduced but did not eliminate variation in bacteraemia incidence between NICUs. In both the studies and the regional monitoring systems, adjustment for baby susceptibility generally involved stratification by factors measured at birth. Adjustment for length of stay and invasive procedures involved reporting incidence by days with a device, such as central venous catheter days. Methods for NICU infection monitoring lack consistency. Adjustment for factors measured at birth fails to capture changes in susceptibility throughout admission and adjustment for device days does not adequately reflect risk to babies not treated with the device. Further research should address variation in risk for all babies throughout their NICU stay.

Introduction

Between 2% and 10% of babies admitted to neonatal intensive care units (NICUs) experience at least one episode of bacteraemia, which can lead to death and other serious adverse outcomes.1, 2, 3 The majority of bacteraemia episodes are hospital-acquired, and often preventable through improvements in hygiene practices and infection control.4, 5, 6

Suggestions that bacteraemia monitoring itself might decrease infection rates⁷ are supported by evidence from a systematic review of >100 randomised controlled trials showing that audit and feedback alone produce small to moderate improvements in clinical practice.8, 9 When comparative monitoring has been used to trigger sharing of improved practices between units, substantial reductions in infection incidence appear to have been achieved.10, 11, 12

Comparisons are complicated by the fact that some of the variation observed between NICUs is attributable to factors other than quality of care, such as case mix, babies' length of stay and the invasive medical procedures carried out, all of which can influence hospital-acquired infection.¹³ To make fair and meaningful comparisons between hospitals, a multicentre monitoring system must adjust for these factors. Any residual variation may be explained, at least in part, by factors amenable to change, such as hygiene practices. To formulate a method for risk adjustment, factors must be identified which are both associated with infection and reliably recorded. These factors can then be used to stratify infection incidence, or can be included in a statistical risk adjustment model.

We performed a systematic review to determine methods used for risk adjustment in studies that compared infection incidence between NICUs, and to determine how much infection incidence varied before and after risk adjustment. We also determined the extent to which these approaches for risk adjustment are being used by regional surveillance systems for NICU-acquired infection around the world. We discuss different approaches for risk adjustment and suggest ways to improve robustness of comparisons and consistency of reporting.