This is a comprehensive compendium of the evidence that links hospital staffing variables and risk of HAI. The majority of the studies investigated nurse staffing. This may be because nurses are the largest workforce in hospitals, and although the number of nurses has grown in the past few years, a shortage still exists and is predicted to become worse in the coming years [56
]. Also, nurses have the most direct and continuous role in performing the procedures and interventions on which the risk of infection often hinges, making them a critical component of infection prevention.
Although the limitations in the study designs make us unable to determine a specific evidence-based nurse staffing level benchmark that is associated with a decreased risk of HAI, trends are apparent from this body of research. For example, although only 2 investigators studied ventilator-associated pneumonia [23
], both reported that patients who were cared for in an intensive care unit with lower levels of nurse staffing had an increased risk of ventilator-associated pneumonia. Although the exact mechanism for this association was not studied, it is possible that when staffing levels are reduced, the nurses are unable to provide recommended care, such as keeping the head of the patient bed elevated [57
]. Additionally, studies examining organism-specific HAI with single-site study designs all found that the level and/or the use of nonpermanent staff was statistically significantly related to a patient’s risk of infection. It may seem surprising that care by float nurses versus full-time permanent staff nurses in intensive care units could put a patient at risk of HAI. However, this is consistent with Pronovost’s description of the hierarchy in intensive care units and the importance of good communication channels and strong interdisciplinary team work [58
]. Temporary staff may lack specific training and familiarity with institutional procedures and “best practices” for preventing HAIs, and they may not have the relationships needed for clear communication. Hospital administrators, nurse managers, and infection control professionals should be aware of the importance of interdisciplinary team work and the need for both consistent training and adequate nurse staffing in reducing the risk of HAI.
There is sparse evidence that examines the impact of staffing of infection control departments and availability of physicians in the prevention of HAI. More than 30 years ago, the CDC undertook the national Study on the Efficacy of Nosocomial Infection Control (SENIC) [59
]. This study established a connection between elements of infection control programs and provided strong evidence that hospitals with more staffing and more intense infection control processes had lower rates of HAI. However, the study has not been updated, and evidence to inform current practice is seriously lacking [60
]. For example, the investigators of the original study recommended that hospitals employ at least 1 full-time infection control professional for every 250 occupied beds; because of the lack of more-recent data, this outdated ratio is sometimes applied as the standard today. Although this ratio may have been adequate many years ago, there have been many changes in health care delivery, such as shorter lengths of stay, increased patient acuity, and an increased risk of HAI, including HAI due to multiple-drug resistant organisms [61
]. Although there has been a number of reports that describe infection control staffing [64
], no researchers have updated this study. The use of the outdated ratio may be contributing to the persistent nature of the HAI problem.
Throughout this body of research, there were a number of methodological flaws. In many of the larger multisite studies, the research teams used administrative data sets to examine both staffing and infection variables. These studies have a number of limitations. First, in large data sets, the staffing measures are often determined using the hospitals’ reported full-time equivalents. This method is not precise, is not unit specific, and is likely to introduce measurement error [66
]. Another limitation in many of these studies is the identification of infections with use of diagnoses and procedure codes recorded in administrative data. Previous studies have found that ICD codes and other hospital administrative databases did not accurately identify patients who had a central line–associated BSI, as defined by the CDC [67
]. Poor sensitivity (60%) and low positive predictive value (20%) for administrative data as a means of identifying HAIs was recently reported in a pediatric hospital [68
]. When these issues are considered, it is likely that variability in the measurement of both staffing and HAI each contribute to the mixed results found in these studies. Indeed, because of the limitations of ICD codes and the identification of HAI, others may not have considered these studies for review. However, we chose to include this body of evidence, because these studies are often cited and often impact policy, such as the legislated minimum nurse-to-patient ratios in California hospitals.
More-precise measurements of staffing and HAI may be obtained directly from the institution. However, the burden of data collection often limits research to smaller studies from single institutions. Single-site studies often lack sufficient sample size and have limited statistical power.
Much of the reviewed research used cross-sectional or cohort design, which limits the interpretation to association, not causality. For example, 1 multisite study had many strengths, including the fact that all hospitals used the CDC protocols to identify HAI and multiple methods were used to control for patient severity of illness and differences among settings [67
]. However, even though the researchers found significant associations between staffing and various HAIs, the interpretation was limited by the cross-sectional study design. It is possible that there were unmeasured organizational traits, besides staffing, that were responsible for the observed effect. Another group of researchers who used administrative data alone conducted a longitudinal study and analyzed variation over time within a setting using a fixed-effect statistical model [27
]. Analytically, this is a much stronger design than a cross-sectional study, because it controls for differences in the setting that are time invariant (e.g., patient population served). Although these researchers found that nurse staffing had a diminishing marginal effect on reducing inpatient mortality, they did not find a similar significant effect on HAI. This may be attributable to the use of the administrative data to identify HAI.
More-rigorous research is needed. Both the staffing and infection variables should be operationalized with the use of reliable and valid measures. The strongest design would be a multisite, randomized, controlled trial; however, this may not be feasible or pragmatic. A study that uses a cluster randomization may be possible. At the very least, researchers should use longitudinal fixed effect and/or instrumental variable research designs (both of which help control for underlying unmeasured differences in setting).
This review has limitations. Only English-language articles published after 1990 were included in the audit. The language limitation was necessary because of resources. With the changes in the health care system over the past 20 years, we believe that limiting the search to publications after 1990 is justified. Besides the SENIC study [61
], we do not know of another seminal study examining other personnel categories that was omitted from the present review. Every attempt was made to be comprehensive in the search strategy; however, we may have missed eligible articles. Different researchers may have chosen different elements to abstract from the articles and/or may have categorized the data elements differently. To ensure that the abstraction was consistent and accurate, 2 readers abstracted each data element and compared abstractions. Any discrepancies were discussed. Although we attempted to report the studies in a consistent manner (e.g., defining level of staffing as both clinician-to-patient ratio and clinical hours per patient-day) and report statistical significance of associations tested, with the varying definitions and data sources for the measurement of both the staffing and HAI variables as well as varying quality of the analyses and diverse settings, we did not report specific point estimates and SEs, and we did not conduct a meta-analysis. Although it is tempting to pool the data and compute weighted averages, with the methodological flaws we observed, we did not believe this to be prudent.
This review examining hospital staffing and HAI was comprehensive. Although nurse staffing was most thoroughly studied, additional rigorous research is warranted with use of standard CDC definitions of HAI and more-accurate and consistent measures of nurse staffing. With the increased use of hospitalists, physician assistants, and other health care providers, more research is needed that incorporates the many disciplines of staffing. Finally, the role and scope of practice of epidemiologists and infection control professionals is changing and increasing. Rigorous research characterizing infection control departments that are effective in preventing HAI in different settings and hospital types is needed.