Surgical Volume in Community Hospitals and the Risk of Surgical Site Infection

1
Surgical Volume in Community Hospitals and the
Risk of Surgical Site Infection
521
Contact information D J Anderson, M.D. DUMC Box
3824 Durham, NC 27710 Phone [phone] Fax
[phone] dja7_at_duke.edu

• DJ Anderson, KS Kaye, ZA Kanafani, DJ Sexton
• Duke Infection Control Outreach Network and Duke
  Univ Med Ctr, Durham, NC, USA

Results
Background
ABSTRACT (REVISED) Background SSIs lead to
adverse patient outcomes. Increasing hospital
size is associated with increasing risk of SSI in
hospitals that participate in the National
Nosocomial Infection Surveillance System (NNIS).
As few community hospitals participate in NNIS,
this association may not be applicable to small,
community hospitals. Objective To determine
the association between the size of community
hospitals and the risk of SSI. Methods The Duke
Infection Control Outreach Network (DICON) is a
network of 35 community hospitals in the
southeastern US. The DICON surgical database
consists of validated prospectively collected
data including patient demographics, date of
surgery, length of procedure, type of procedure
(by NNIS code), and NNIS Risk Index variables for
all procedures. Data was analyzed for the time
period 1/1/2004 to 12/31/2005. SSI was defined
using standard CDC definitions only deep and
organ-space infections were included in the
analysis. Surveillance for SSI was standardized
among all hospitals. Hospitals were grouped into
three categories small ( 1,500 procedures
annually), medium (1,500 lt annual procedures
4,000), and large (gt 4,000 annual procedures).
Poisson regression was used to determine
prevalence rates and ratios. Results Eighteen
hospitals participated in the DICON surgical
database during the study period (median bed size
220, range 102-537). Surveillance was performed
on 132,111 surgical procedures. A median of
3,421 procedures (IQR 1,479 5,392) was
performed at each hospital. 1,434 SSIs were
identified (prevalence rate of SSI 1.09/100
procedures). S. aureus was the most common
infecting organism and was isolated from 468 SSIs
(33). The prevalence rate of SSI at small
hospitals was 1.24/100 procedures (95 confidence
interval (CI) 1.18 1.29) the prevalence rate
of SSI at medium hospitals was 0.65/100
procedures (95 CI 0.61 0.65) and the
prevalence rate of SSI at large hospitals was
1.16/100 procedures (95 CI 1.14 1.18).
Compared to medium hospitals (referent), the
prevalence rate ratio for small hospitals was
1.91 (95 CI 1.78-2.05) and the prevalence rate
ratio for large hospitals was 1.79 (95 CI 1.70
1.90). After adjusting for confounders and
important effect modifiers, the prevalence rate
ratio for small hospitals was 1.62 (95 CI
1.51-1.74) and the prevalence rate ratio for
large hospitals was 1.40 (95 CI
1.33-1.48). Conclusions The association between
size of community hospital and risk of SSI is
complex and different than in NNIS hospitals.
SSIs were more prevalent at small and large
community hospitals compared to medium-sized
community hospitals.

• Overall, 1,434 SSIs were identified after
  132,111 procedures (PR1.09/100 procedures)
• - S. aureus was the most common organism
  (33)
• Hospital characteristics are summarized in Table
  1
• PR and PRR for hospitals are summarized in Table
  2 and Figure

• Sensitivity analyses controlling for NNIS RI,
  general surgical, high risk procedures, and
  procedures performed only at large hospitals
  showed no change in PRR
• Multivariable analysis controlling for all of
  these factors together showed essentially same
  results
• - Adjusted PRR for small hosp1.62 (95 CI
  1.51-1.74)
• - Adjusted PRR for large hosp1.40 (95 CI
  1.33-1.48)

• SSIs lead to adverse patient outcomes including
  increased length of stay, mortality, and cost
• In NHSN hospitals, risk of SSI decreases with
  increasing surgical volume
• The relationship between surgical volume and
  risk of SSI has not been studied in community
  hospitals
• The NHSN hospitals do not represent the majority
  of hospitals in the US
• - 83 have academic affiliations
• - Median bed size360

Table 1. General Description of Study Hospitals
Table 2. Risk of SSI Stratified by NNIS RI
Methods

• The DICON Surgical Database contains
  prospectively collected surveillance data from a
  total of 26 DICON hospitals 18 hospitals (bed
  range 102-537) were included in this analysis
• Data for all operative procedures include
  patient demographics, date of surgery, length of
  procedure, type of procedure, and NNIS risk index
  (RI) variables
• Hospitals were separated into 3 categories based
  on annual surgical volume small (lt 1,500),
  medium (1,500-4,000), and large ( 4,000).
• A high-risk procedure was defined as a
  procedure for which the mean rate of SSI
  published by NNIS was gt 3 for patients with a
  risk index of 1.
• A surgeon experience score was calculated by
  determining the average number of procedures
  performed annually by each surgeon at each
  hospital during the study period.
• Prevalence rates (PR) and prevalence rate ratios
  (PRR) were determined using Poisson regression.
• Poisson regression was also used to create a
  multivariable model to estimate PRR after
  adjusting for important covariates, potential
  confounders, and effect measure modifiers.

Figure. Relationship between PR of SSI and
Surgical Volume
plt0.0001 for comparison with medium-sized
hospitals
Conclusions

• The relationship between surgical volume and SSI
  risk in community hospitals is complex and
  different than in NHSN hospitals
• Even after adjusting for case-mix in sensitivity
  and multivariable analyses, small and large
  community hospitals had higher rates of SSI
• Methods for risk adjustment should include
  measures of surgical volume to ensure adequate
  intra-hospital comparisons, particularly with the
  increase in mandates for public reporting,

Dr. Anderson is a recipient of the Pfizer
fellowship in Infectious Disease Dr. Kaye was
supported by K23 AG23621-01A1 from the National
Institute of Aging and by the John A. Hartford
Foundation