Program Schedule

1005
Empirical Performance of Statistical Process Control Methods for Regional Hospital-Acquired Infection Surveillance: A 10-Year Multi-State Study

Session: Poster Abstract Session: Surgical Site Infections
Friday, October 10, 2014
Room: The Pennsylvania Convention Center: IDExpo Hall BC

Background:

Use of statistical process control (SPC) in healthcare is increasing but remains less common than in other industries. Healthcare-associated infections (HAIs) are important applications that can benefit from SPC methods, including use within existing surveillance programs to trigger investigation and intervention. The empirical value of SPC, however, typically is anecdotal and remains unclear to many practitioners.

Methods:

We retrospectively applied Shewhart and EWMA SPC charts to 10 years (2003 to 2013) of de-identified surgical site infection (SSI) data, including 8 known SSI outbreaks, from the Duke Infection Control Outreach Network of 40 community hospitals. For both methods, we computed the number of outbreaks detected, months of earlier detection versus traditional surveillance, and total signals produced (TS) over the entire study period as a workload measure. To distinguish between minor unsustained versus major sustained HAI rate increases, we additionally calculated the number of months with signals during the year prior (PS) to each outbreak, monthly consecutive signals (CS) before each outbreak, and signals during each outbreak period (DS).

Results:

All known outbreaks were detected by all charts 0 to 12 months earlier than known start dates, on average by 5.8 (Shewhart) and 6.8 (EWMA) months (Table 1). 62.5% of these earlier detections included more than one signal. The total number of signals over the study period (outbreaks, uncertain, or false alarms) averaged 5.5 (Shewhart) and 16.5 (EWMA) per outbreak hospital.

Conclusion:

SPC methods appear useful and practical for augmenting current HAI surveillance programs with empirical ability to detect outbreaks earlier. EWMA charts exhibited fastest detection, agreeing with theoretic comparisons in the literature. Alarm rates appear manageable in terms of investigation burden on health systems.

Table 1: Performance of Shewhart and EWMA control charts to detect known outbreaks (TS: total, PS: previous year, CS: consecutive, DS: during).

 

 

Method

Number of signals

Early detection (months prior to traditional detection)

Total (TS)

Previous year (PS)

Consecutive (CS)

During (DS)

First signal

First consecutive signal

Shewhart

5.5

1.6

1.1

0.9

5.8

0.6

EWMA

16.5

3.1

1.9

3.3

6.8

1.5

 

Arthur W. Baker, MD1, Nicholas Andrianas Jr.2, Salah Haridy2, Deverick Anderson, MD, MPH1, Daniel J. Sexton, MD, FIDSA1 and James Benneyan, PhD.2, (1)Division of Infectious Diseases, Duke University Medical Center, Durham, NC, (2)Northeastern University Healthcare Systems Engineering Institute, Boston, MA

Disclosures:

A. W. Baker, None

N. Andrianas Jr., None

S. Haridy, None

D. Anderson, None

D. J. Sexton, UpToDate: Editor, Royalties
National Football League: Consultant, Consulting fee and Educational grant
Cubist: Grant Investigator, Grant recipient
Johnson and Johnson: Consultant, Consulting fee

J. Benneyan, PhD., None

Findings in the abstracts are embargoed until 12:01 a.m. EDT, Oct. 8th with the exception of research findings presented at the IDWeek press conferences.

Sponsoring Societies:

© 2014, idweek.org. All Rights Reserved.

Follow IDWeek