ARTICLES
Practical risk-adjusted quality control charts for infection control*

https://doi.org/10.1067/mic.2000.109883Get rights and content

Abstract

Background: Control chart methodology has been widely touted for monitoring and improving quality in the health care setting. P charts and U charts are frequently recommended for rate and ratio statistics, but their practical value in infection control may be limited because they (1) are not risk-adjusted, and (2) perform poorly with small denominators. The Standardized Infection Ratio is a statistic that overcomes both these obstacles. It is risk-adjusted, and it effectively increases denominators by combining data from multiple risk strata into a single value. Setting: The AICE National Database Initiative is a voluntary consortium of US hospitals ranging in size from 50 to 900 beds. The infection control professional submits monthly risk-stratified data for surgical site infections, ventilator-associated pneumonia, and central line–associated bacteremia. Methods: Run charts were constructed for 51 hospitals submitting data between 1996 and 1998. Traditional hypothesis tests (P values <.05) flagged 128 suspicious points, and participating infection control professionals investigated and categorized each flag as a “real problem” or “background variation.” This gold standard was used to compare the performance of 5 unadjusted and 11 risk-adjusted control charts. Results: Unadjusted control charts (C, P, and U charts) performed poorly. Flags based on traditional 3-sigma limits suffered from sensitivity <50%, whereas 2-sigma limits suffered from specificity <50%. Risk-adjusted charts based on the Standardized Infection Ratio performed much better. The most consistent and useful control chart was the mXmR chart. Under optimal conditions, this chart achieved a sensitivity and specificity >80%, and a receiver operating characteristic area of 0.84 (P <.00001). Conclusions: These findings suggest a specific statistic (the Standardized Infection Ratio) and specific techniques that could make control charts valuable and practical tools for infection control. (AJIC Am J Infect Control 2000;28:406-14)

Section snippets

Surveillance system

The AICE National Database Initiative is a voluntary national consortium of US hospitals ranging in size from 50 to 900 beds. ICPs collect data on 3 types of nosocomial infections: surgical site (SSI), ventilator-associated pneumonia (VAP) and central line–associated bacteremia (CLB). All hospitals in the AICE National Database Initiative project that submitted data for at least 12 consecutive months between 1996 and 1998 were included in this study. The sample included 51 hospitals in 29

Results

Of the 128 flags analyzed, 84 were flagged for suspiciously high rates, 28 for suspiciously low rates, and 16 for excessive month-to-month variability. Among high rates, 28 of 84 (33%) lasted more than 1 month. Among low rates, 18 of 28 (64%) lasted more than 1 month. If the 16 variability flags are included, then 62 of 128 flags (48%) involved more than 1 month of data.

The primary or root causes of all 128 flags were grouped into 5 major categories as shown in Table 2.Sixty-two flags (48%)

Discussion

No control chart or statistic can perfectly predict which nosocomial infections have common causes and which infections have special causes. Thus control charts only benefit hospitals with trained ICPs who know how investigate suspicious clusters once they are identified. Still, professionals at any level of experience should welcome a tool that helps focus their search for preventable special causes.

Whereas classic statistical tests like χ2, Fisher exact, or incidence density are valuable

Acknowledgements

I thank Kathryn Miller, AICE National Database Initiative Project Manager, for her tireless data collection efforts, and Dr Robert Haley for his review and suggestions on the manuscript.

Glossary

GLOSSARY

Background variation:
See common cause variation.
C Chart.
Control chart that plots counts (infections) over time.
Common cause variation.
Normal background variation related to many interacting factors that are difficult to prevent or control.
CUSUM chart.
Cumulative Sum control chart plots the cumulative sum of deviations from the mean instead of the continuous variable itself.
EWMA chart.
Exponentially Weighted Moving Average control chart plots a moving average of a continuous variable, giving more

References (17)

  • DJ Wheeler et al.

    Understanding statistical process control

  • RW Haley et al.

    The efficacy of infection surveillance and control programs in preventing nosocomial infections in US hospitals

    Am J Epidemiol

    (1985)
  • JC Benneyan

    Statistical quality control methods in infection control and hospital epidemiology, part I: introduction and basic theory

    Infect Control Hosp Epidemiol

    (1998)
  • JC Benneyan

    Statistical quality control methods in infection control and hospital epidemiology, part ii: chart use, statistical properties, and research issues

    Infect Control Hosp Epidemiol

    (1998)
  • RG Carey et al.

    Measuring quality improvement in healthcare: a guide to statistical process control applications

  • LJ Finison et al.

    The use of control charts to improve healthcare quality

    J Healthcare Quality

    (1993)
  • JA Sellick

    The use of statistical process control charts in hospital epidemiology

    Infect Control Hosp Epidemiol

    (1993)
  • Joint Commission on Accreditation of Healthcare Organizations

    ORYX pre-survey report clarification

    ORYX Gazelle

    (May 2000)
There are more references available in the full text version of this article.

Cited by (0)

*

Reprint requests: Infection Control and Prevention Analysts, Inc, 4425 S MoPac Expressway, Suite 205, Austin, TX 78735.

View full text