Risk stratification by level of care for comparing bloodstream infection rates in neonatal intensive care units

Bloodstream infections: epidemiology and resistance

Bloodstream infections in the neonatal intensive care unit (NICU) are associated with many adverse outcomes in infants, including increased length of stay and cost, poor neurodevelopmental outcomes, and death. Attention to the insertion and maintenance of central lines, along with careful review of when the catheters can be safely discontinued, can minimize central-line-associated bloodstream infections rates. Good antibiotic stewardship can further decrease the incidence of bloodstream infections, minimize the emergence of drug-resistant organisms or Candida as pathogens in the NICU, and safeguard the use of currently available antibiotics for future infants.

A toolkit for monitoring hospital-acquired bloodstream infection in neonatal intensive care

OBJECTIVE

In neonatal intensive care units (NICUs), monitoring hospital-acquired bloodstream infection (BSI) is critical to alert clinicians to variations in the incidence of infection between units and over time. We demonstrate a toolkit of monitoring techniques that account for case mix and could be implemented using routinely available clinical data. This toolkit could enable quality of care comparisons between hospitals to facilitate the sharing of improved practices.

DESIGN

Prospective study over 4 years.

SETTING AND PATIENTS

Babies admitted to 2 tertiary London NICUs.

METHODS

We derived expected numbers of BSI episodes using a Poisson regression risk model adjusting for variations in birth weight, transfers to the NICU from other hospitals, postnatal age, and days spent at each National Health Service level of care. We compared observed and expected numbers of BSI episodes using 2 monitoring techniques: standardized infection ratios (SIRs) and the sequential probability ratio test (SPRT).

RESULTS

Using the SIR method, observed BSI incidence increased over expected incidence in 2002 at both NICUs, but this increase did not reach statistical significance at the 1% level. Using the SPRT method, neither unit showed a clinically important increase or decrease, defined as a 30% deviation from expected incidence.

CONCLUSIONS

Risk-adjusted BSI monitoring can be performed using routine hospital data. NICUs could use SIRs for an annual look back at infection incidence and SPRTs for prospective, quarterly monitoring. The SIR and the SPRT methods have different strengths, and both could help clinicians improve infection control and patient care in NICUs.