Correspondence
Catheter-Related Bloodstream Infections
N Engl J Med 2007; 356:1267-1268March 22, 2007
- Article
To the Editor:
The report by Pronovost et al. (Dec. 28 issue)1 on an intervention to decrease catheter-related bloodstream infections in the intensive care unit (ICU) would have been more convincing had the authors substantiated that the decline in the rate of these infections was matched by a decline in the number of positive blood cultures or in the use of antibiotics. A culture of safety in which the rates of catheter-related bloodstream infections were an explicit benchmark of quality could have biased clinicians against attributing positive blood cultures to such infections. The seemingly straightforward criteria for catheter-related bloodstream infections of the National Nosocomial Infections Surveillance (NNIS) system of the Centers for Disease Control and Prevention are deceptively ambiguous. For example, even though coagulase-negative staphylococcus is the most common cause of catheter-related bloodstream infections and the most common contaminant of blood cultures, the NNIS does not define a method to distinguish between the two possibilities. Clinical practice — whether positive blood cultures are repeated or automatically trigger treatment — affects the subsequent adjudication of the culture result as indicating contamination or a true bloodstream infection. Furthermore, attributing gram-negative bloodstream infections or candidemia to a catheter, when they occur in association with other potential sources of infection, is inherently subjective. The reduction in the rate of catheter-related bloodstream infections reported by Pronovost and colleagues may have been due to a collective bias against attributing bacteremia to catheters.
1 ReferencesElizabeth R. Jenny-Avital, M.D.
Jacobi Medical Center, Bronx, NY 10461
jennyavita@earthlink.net 1
Pronovost P ,Needham D ,Berenholtz S , et al. An intervention to decrease catheter-related bloodstream infections in the ICU. N Engl J Med 2006;355:2725-2732
Full Text | Web of Science | Medline
To the Editor:
In the study reported by Pronovost and colleagues, the handling of missing data is of concern. The scale of the missing data is difficult to determine. Complete data were obtained for only 53 of the 108 ICUs. A potential maximum of 324 ICU-months of data exist for each 3-month study period for these 108 ICUs. However, the ICU-months of data actually obtained for each 3-month period are not included in Table 3 of the article. Another report on this cohort by the same group1 documents 193 ICU-months of data obtained for the crucial postintervention period from 0 to 3 months. This value represents only 60% of the potential ICU-months of data. Calculations of low infection rates based on incomplete data are worrying. Subgroup analysis of the 49% of ICUs with complete data does not eliminate the problem. The paucity of information regarding the scale of the missing data affects the appraisal of the potential bias and internal validity of the study.
1 ReferencesMark R. Daley, B.Med.
Royal Prince Alfred Hospital, Sydney 2050, Australia
mark.daley@email. cs. nsw. gov. au 1
Pronovost PJ ,Berenholtz SM ,Goeschel CA , et al. Creating high reliability in health care organizations. Health Serv Res 2006;41:1599-1617
CrossRef | Web of Science | Medline
Author/Editor Response
Large-scale quality-improvement studies are challenging to conduct, and they receive substantially less funding than randomized trials of comparable size. In our study, staff in the 103 participating ICUs did not receive funding to support data collection. Thus, research intended to improve the quality of care must carefully balance the collection of data that are scientifically sound, feasible to collect, and focused on the specific aims of the study.1
We chose to limit the quantity but not the quality of data collected. It was not feasible to collect data on the organisms cultured or the antibiotics used, as Jenny-Avital suggests. We used standardized though somewhat subjective surveillance rather than clinical definitions of catheter-related bloodstream infections.2 However, we believe that potential bias was minimized, because infection-control practitioners who were independent of the ICU teams performed all measurements in their routine manner.
Daley's comments highlight the importance of minimizing and reporting missing data in quality-improvement studies. We made great efforts to minimize information bias and missing data. To garner participation, the ICUs were permitted to choose when to implement the study intervention. The ICUs provided data at varying times, depending on when they joined the project and started data collection. Forty hospitals implemented the intervention immediately on initiation of the study, which precluded the collection of baseline data. Thus, data were not missing for these ICUs. The results of our sensitivity analysis, which excluded these hospitals, were similar to those of the primary analysis. During the period when each ICU reported data, 30 of 103 ICUs (29%) did not report completely, resulting in missing data for 113 of 2216 potential ICU-months (5%). This level of missing data represents an improvement over a preliminary report conducted before all ICUs had reported their data.3
The need to improve the quality of care is too great and the resources devoted to this effort are too limited to be uncertain about whether quality-improvement interventions actually work. Quality-improvement studies will require more rigor and resources. Since funding for such research has lagged behind funding for basic and clinical research by a factor of more than 100, additional funding will be needed to achieve the goal of improved care. Though our study has helped to advance the science of rigorous quality-improvement studies, more work is needed.
3 ReferencesPeter J. Pronovost, M.D., Ph.D.
Dale M. Needham, M.D., Ph.D.
Sean Berenholtz, M.D.
Johns Hopkins University, Baltimore, MD 212051
Pronovost P ,Wachter R . Proposed standards for quality improvement research and publication: one step forward and two steps back. Qual Saf Health Care 2006;15:152-153
CrossRef | Web of Science | Medline2
National Nosocomial Infections Surveillance System
CrossRef | Web of Science | Medline3
Pronovost PJ ,Berenholtz SM ,Goeschel CA , et al. Creating high reliability in health care organizations. Health Serv Res 2006;41:1599-1617
CrossRef | Web of Science | Medline
- Citing Articles (2)
Citing Articles
1
D. M. Needham, D. J. Sinopoli, V. D. Dinglas, S. M. Berenholtz, R. Korupolu, S. R. Watson, L. Lubomski, C. Goeschel, P. J. Pronovost. (2009) Improving data quality control in quality improvement projects. International Journal for Quality in Health Care 21:2, 145-150
CrossRef2
Marta Rodríguez-Créixems, Luis Alcalá, Patricia Muñoz, Emilia Cercenado, Teresa Vicente, Emilio Bouza. (2008) Bloodstream Infections. Medicine 87:4, 234-249
CrossRef
