NEJM 2018 appendix

Supplementary Appendix
This appendix has been provided by the authors to give readers additional information about their work.
Supplement to: Magill SS, O’Leary E, Janelle SJ, et al. Changes in prevalence of health care–associated infections
in U.S. hospitals. N Engl J Med 2018;379:1732-44. DOI: 10.1056/NEJMoa1801550

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table of Contents
Emerging Infections Program Hospital Prevalence Survey Team members................................................. 1
Funding source and author roles .................................................................................................................. 2
Methods: hospital and patient selection ...................................................................................................... 3
Methods: training and data collection.......................................................................................................... 4
Methods: National Healthcare Safety Network surveillance definitions ..................................................... 5
Methods: modeling and national burden estimates .................................................................................... 6
Results: comparison of prevalence of health care-associated infections .................................................... 8
Discussion: limitations .................................................................................................................................. 9
Figure S1...................................................................................................................................................... 11
Table S1 ....................................................................................................................................................... 12
Table S2 ....................................................................................................................................................... 13
Table S3 ....................................................................................................................................................... 15
Table S4 ....................................................................................................................................................... 18
Table S5 ....................................................................................................................................................... 21
Table S6 ....................................................................................................................................................... 22
Table S7 ....................................................................................................................................................... 24
Table S8 ....................................................................................................................................................... 26
Table S9 ....................................................................................................................................................... 28
Table S10 ..................................................................................................................................................... 31
References .................................................................................................................................................. 33

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Emerging Infections Program Hospital Survey Team members
The following individuals were members of the Emerging Infections Program Hospital Survey
Team and non-author contributors:
California Emerging Infections Program, Oakland, CA: Deborah Godine, RN, CIC; Linda Frank, RN, BSN;
Lauren Pasutti, MPH; Erin Parker, MPH; Brittany Martin, MPH; Karen Click
Colorado Department of Public Health and Environment, Denver, CO: Helen Johnston, MPH; Sarabeth
Friedman, CNM, MSN; Annika Jones, MPH; Tabetha Kosmicki, MPH
Connecticut Emerging Infections Program, New Haven and Hartford, CT: James Meek, MPH; Richard
Melchreit, MD; James Fisher, MPH; Amber Maslar
Division of Healthcare Quality Promotion, Centers for Disease Control and Prevention, Atlanta, GA:
Katherine Allen-Bridson, RN, BSN, MScPH, CIC; Angela Anttila, PhD, MSN, NPC, CIC (CACI, Inc.); Henrietta
Smith, RN, MSN, CIC (Northrop Grumman); Anthony Fiore, MD, MPH
Georgia Emerging Infections Program, Decatur, GA: Susan L. Morabit, MSN, RN, PHCNS-BC, CIC; Lewis
Perry, DrPH, MPH, RN; Scott K. Fridkin, MD
Maryland Department of Health, Baltimore, MD: Elisabeth Vaeth, MPH; Rebecca Perlmutter, MPH, CIC
Minnesota Department of Health, St. Paul, MN: Jane Harper, BSN, MS, CIC; Annastasia Gross, MPH,
MT(ASCP); Nabeelah Rahmathullah, MBBS, MPH; Brittany Von Bank, MPH
New Mexico Department of Health, Santa Fe, NM: Lourdes M. Irizarry, MD; Joan Baumbach, MD, MS,
MPH
New York Emerging Infections Program and University of Rochester Medical Center, Rochester, NY: Gail
Quinlan, RN, CIC; Anita Gellert, RN
Oregon Health Authority, Portland, OR: Alexia Zhang, MPH
Tennessee Department of Health, Nashville, TN: Patricia Lawson, RN, MS, MPH; Raphaelle H. Beard,
MPH; Vicky P. Reed, RN; Daniel Muleta, MD, MPH
1

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Funding source and author roles
This work was supported through the Centers for Disease Control and Prevention’s Emerging
Infections Program Cooperative Agreement with funds from the CDC’s Division of Healthcare Quality
Promotion in the National Center for Emerging and Zoonotic Infectious Diseases. Author roles are as
follows:

Project concept or design: Shelley S. Magill, Joelle Nadle, Sarah Janelle, Wendy Bamberg, Susan M. Ray,
Lucy E. Wilson, Katherine Richards, Ruth Lynfield, Linn Warnke, Ghinwa Dumyati, Zintars Beldavs,
Marion A. Kainer, Jonathan R. Edwards

Data acquisition: Joelle Nadle, Sarah Janelle, Tolulope Oyewumi, Samantha Greissman, Meghan
Maloney, Nicolai Buhr, Katherine Richards, Linn Warnke, Jean Rainbow, Deborah L. Thompson, Marla
Sievers, Shamima Sharmin, Emily B. Hancock, Cathleen Concannon, Valerie Ocampo, Monika Samper,
Ruby M. Phelps, Cindy Gross, Denise Leaptrot, Janet Brooks, Eileen Scalise, Farzana Badrun

Data analysis: Shelley S. Magill, Erin O’Leary, Jonathan R. Edwards

Data interpretation: Shelley S. Magill, Erin O’Leary, Joelle Nadle, Susan M. Ray, Lucy E. Wilson, Katherine
Richards, Nicolai Buhr, Ruth Lynfield, Shamima Sharmin, Ghinwa Dumyati, Zintars Beldavs, Marion A.
Kainer, Cindy Gross, Denise Leaptrot, Janet Brooks, Eileen Scalise, Jonathan R. Edwards

Shelley S. Magill wrote the first draft of the manuscript. All of the authors vouch for the completeness
and accuracy of the data, and all authors decided to submit the manuscript.

2

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Methods: hospital and patient selection
To engage additional hospitals beyond those that participated in the 2011 survey, Emerging
Infections Program sites used the same approach employed in the 2011 survey.1 Each site recruited
additional hospitals using randomly sorted hospital lists stratified by bed size, with the following goals in
each bed size stratum: 13 small (<150 beds), 9 medium (150–399 beds), and 3 large (≥400 beds)
hospitals. Participation was voluntary.
The numbers of randomly selected acute care inpatients to be included in the survey were
determined by hospital bed size category, as in 2011. For small and medium hospitals, the sample goal
was 75 patients; if the hospital had < 75 patients on the survey date then all patients were to be
included. For large hospitals, the sample goal was 100 patients.

3

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Methods: training and data collection
Hospital staff participating in 2015 survey activities were asked to view recorded survey
operations training or join a live training session prior to their hospitals’ survey dates. Emerging
Infections Program staff participated in live training or viewed recorded training on survey operations,
health care-associated infection (HAI) definitions, and data collection. Training provided for the 2015
survey was similar to training provided for the 2011 survey, except for the option of viewing recorded
training sessions. Emerging Infections Program data collectors also received training for expanded data
collection activities in the 2015 survey, including HAI data collection using two different sets of National
Healthcare Safety Network HAI surveillance definitions (the 2011 definitions and the 2015 definitions).
Hospitals in the 2015 survey were asked to complete a questionnaire that included information
on hospital characteristics and infection control and antimicrobial stewardship policies and practices.
Emerging Infections Program staff also gathered limited information on selected hospital characteristics.
Hospital data were entered into a Research Electronic Data Capture (REDCap)2 database hosted at CDC,
and were included with patient data in the analysis. Emerging Infections Program sites had the option of
utilizing their data collectors for all aspects of patient data collection, or engaging hospital staff to collect
a limited amount of demographic and clinical information for each surveyed patient in their facility. In
addition, in the 2015 survey Emerging Infections Program sites and hospitals were given the option of
collecting the initial, limited demographic and clinical data on the survey date or retrospectively. If these
data were collected retrospectively, data collectors were instructed only to report information present
in the medical record up until 17:00 hours on the survey date. Emerging Infections Program staff
reviewed medical records to collect detailed information on antimicrobial use and HAIs; hospital staff
did not participate in these reviews. CDC staff provided support to Emerging Infections Program data
collectors for questions regarding National Healthcare Safety Network HAI definitions, HAI
determinations, or other aspects of data collection.
4

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Methods: National Healthcare Safety Network surveillance definitions
The Emerging Infections Program hospital prevalence surveys of HAIs and antimicrobial use are
conducted using the National Healthcare Safety Network’s acute care hospital HAI surveillance
definitions. Each year CDC updates these surveillance definitions to improve the objectivity, usability or
clinical credibility of the definitions. In 2015, major revisions to the National Healthcare Safety Network
definitions were implemented. Therefore, in the 2015 survey, we opted to collect HAI data using two
different sets of National Healthcare Safety Network HAI definitions. Data were collected using the same
HAI definitions used in the 2011 survey3 for the purposes of comparing HAI prevalence and distribution
in the 2011 and 2015 surveys, which is the focus of this manuscript. Data were also collected using the
2015 definitions for the five most common HAI types (pneumonia, surgical-site infections,
gastrointestinal infections, bloodstream infections, and urinary tract infections) and for ventilatorassociated events.4 A detailed description of the National Healthcare Safety Network HAI definition
changes implemented in 2015 is beyond the scope of this appendix; in general, changes were designed
to reduce the subjectivity of the surveillance definitions by providing specific time periods within which
HAI definition criteria must be met, and update HAI definition criteria to reflect current practices in
diagnostic testing (see https://www.cdc.gov/nhsn/pdfs/newsletters/vol9-3-eNL-Sept-2014.pdf). A
“repeat infection timeframe” was also implemented in the National Healthcare Safety Network in 2015,
specifying a duration of 14 days for most HAIs, but this timeframe was not strictly implemented in the
2015 survey due to its cross-sectional design. We did not have the data to be able to apply the 2015
definitions retroactively to patients in the 2011 survey.

5

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Methods: modeling and national burden estimates
We developed approaches to handling missing or unknown data to maximize the numbers of
patients whose data could be included in the modeling process. For patients who had surgical-site
infections (SSIs) with onset before hospitalization but for whom a specific onset date was unknown, we
created a proxy onset date. First, we determined the median number of days from the operative
procedure date to SSI onset date in patients with known SSI onset dates before admission. We added
the median number of days from procedure to SSI onset to the operative procedure dates of patients
with unknown SSI onset dates before admission to create proxy SSI onset dates. For one patient with
pneumonia for whom onset date was unknown, but onset was before hospitalization (such infections
could be deemed HAIs if related to a prior, recent hospitalization), we set the onset date equal to the
admission date. There were also patients with missing hospital length of stay data. Of these 9 patients, 8
were still in the hospital 6 months after the survey date when follow up for discharge and outcome
information ended. For these patients, a proxy for hospital length of stay was considered the time from
admission to last follow up date. For the ninth patient, hospital discharge date was unknown.
We developed national burden estimates for 2015 using a process similar to the method used in
2011.1 First, we used logistic and log-binomial regression models to identify patient and hospital factors
associated with HAIs, and we assessed model fit using the likelihood ratio test and Akaike Information
Criterion score. Log-binomial regression models were compared and verified for robustness using
Poisson regression in a Generalized Estimating Equations framework. Second, we used factors
independently associated with HAIs to partition survey data and 2014 National Inpatient Sample (NIS)
data (Healthcare Cost and Utilization Project, Agency for Healthcare Research and Quality) into patient
strata.5 We predicted HAI prevalence within each stratum using the final log-binomial regression model.
We calculated HAI incidence in each stratum with the formula of Rhame and Sudderth,6 using the
predicted prevalence and stratum-specific data from the prevalence survey on hospital length of stay
6

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
and time to HAI onset. Finally, we generated national burden estimates of hospital patients with HAIs by
multiplying HAI incidence by the total number of discharges in each NIS stratum and summing across
strata. The point estimate of the total number of patients with HAIs and the upper and lower bounds of
the 95% confidence interval (CI) were rounded to the nearest hundred.

7

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Results: comparison of HAI prevalence
Among patients in the 2015 survey, we compared the percentage of patients with HAIs detected
by the 2011 definitions vs. the 2015 definitions. When the 2011 definitions were applied, 342 of 12,299
patients had pneumonia, SSIs, bloodstream infections, urinary tract infections, or gastrointestinal
infections (2.8%; 95% confidence interval [CI], 2.5 to 3.1). When the 2015 definitions were applied, 345
patients had ≥1 of these 5 HAI types (2.8%; 95% CI, 2.5 to 3.1). A comparison of the distribution of HAI
types using the 2011 definitions vs. the 2015 definitions is shown in Table S5.

8

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Discussion: limitations
Additional limitations of our prevalence survey include its restriction to only those HAIs that
were active at the time of the survey, defined as HAIs with signs or symptoms on the survey date, or
HAIs still being treated with antimicrobial agents. Although we used the same definitions in 2011 and in
2015, practice changes could have affected detection of active HAIs. For example, substantial changes in
medical record documentation of signs and symptoms or antimicrobial prescribing could have affected
our ability to detect HAIs. Similar findings were observed in the subset of hospitals that participated in
both surveys and in the subset of patients who received antimicrobial agents and met our HAI review
criterion, suggesting that changes in documentation and prescribing likely do not account for the
observed decrease in prevalence.
Point prevalence surveys have the potential to over-represent HAIs of longer duration, such as
SSIs, since on any given day patients with such infections are more likely to have signs or symptoms or
be receiving antibiotics than patients with shorter-duration infections, such as urinary tract infections.7
Although this prevalence survey bias could influence the distribution of HAI types detected in the
survey, it would not be expected to affect substantially the comparison of overall prevalence in 2011
compared with 2015.
We used the Rhame and Sudderth formula for converting HAI prevalence to incidence,4 which is
a method with well-described limitations.8-12 The formula was published almost 40 years ago, and its
components may not fully account for the complexities of present-day health care delivery. For
example, the formula incorporates a term representing the time from hospital admission to HAI onset,
which may present challenges for HAIs that begin prior to the prevalence survey hospitalization. As an
example, most SSIs have their onset outside the hospital, following discharge from the hospitalization
during which the operative procedure occurred, and some investigators have reported a poor
correlation between observed SSI incidence and SSI incidence calculated using the Rhame and Sudderth
9

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
formula.12 Similarly, patients may be readmitted for Clostridioides difficile infections that begin in the
outpatient setting but are related to a prior hospitalization. Although the HAI surveillance definitions we
used allow for detection of certain HAIs that are present on admission, whether the timing of these
infections is adequately accounted for in the conversion of prevalence to incidence is unclear.
The formula is intended to capture active and cured infections and uses the time from
admission to the first HAI in patients with multiple HAIs;6 because of our survey methods, we detected
only active HAIs, and we cannot assume that all infections active at the time of the survey were in fact
patients’ first HAIs during the hospitalization. For patients with multiple HAIs active at the time of the
survey, we used time from admission to onset of the first infection in our analysis.
Finally, Rhame and Sudderth recommended using the average daily census and average daily
admissions from the survey month to approximate the average length of stay of all hospital patients in
their formula. They cautioned that using the average length of stay of all patients on the survey date
would result in an artificially inflated length of stay, since prevalence surveys are biased toward longerstay patients.6 Although we asked hospitals to provide data on average daily census from a recent year,
we did not have data on average daily census or daily admissions at the time of the survey, and
therefore we used the average length of stay of patients included in the survey. It is unlikely that this or
the other limitations discussed above affected the results of our analysis comparing 2011 and 2015 HAI
prevalence, since we used the same approach in both surveys.

10

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Figure S1. Numbers of patients surveyed by month, 2011 vs. 2015.
6000

5196
(46%)

5000

4236
(38%)
3422
(28%)

4000
3000
2000
1000
0

2952
(24%)

2917
(24%)
1960
(16%)
1048
667 (9%)
(6%)

May

801
(7%)

June

July
2011

2015

11

August

382
(3%)
September

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S1. Catchment areas, hospitals and patients included in the survey, by Emerging Infections
Program site.
No. of

No. of

Hospitals (%)

Patients (%)

Site

Survey Catchment Areaa

California

3-county San Francisco Bay area

14 (7.0)

919 (7.5)

Colorado

11-county Front Range areab

16 (8.0)

1078 (8.8)

Connecticut

Entire state

14 (7.0)

1049 (8.5)

Georgia

20-county metropolitan Atlanta area

22 (11.1)

1525 (12.4)

Maryland

Entire state

22 (11.1)

1437 (11.7)

Minnesota

Entire state

25 (12.6)

1377 (11.2)

New Mexico

Entire state

18 (9.0)

876 (7.1)

New York

10-county western New York areac

22 (11.1)

1312 (10.7)

Oregon

10-county metropolitan Portland and Eugene area

22 (11.1)

1370 (11.1)

Tennessee

Entire state

24 (12.1)

1356 (11.0)

199 (100)

12,299 (100)

Total
Percentages may not total 100 due to rounding.
a

Catchment areas for the 2015 survey were the same as for the 2011 survey unless otherwise specified.
Catchment area for the 2011 survey consisted of 5 Front Range counties.

b

c

Catchment area for the 2011 survey consisted of 9 western New York counties.

12

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S2. Characteristics of hospitals participating in the 2015 survey.
Hospitals
(N=199)

Characteristic
Region — no. (%)
Midwest

25 (12.6)

Northeast

36 (18.1)

South

68 (34.2)

West

70 (35.2)

Locationa — no. (%)
Rural

22 (11.1)

Urban

177 (88.9)

Teaching hospitalb — no. (%)
Yes

88 (44.2)

No

111 (55.8)

Infection preventionist staffingc — no. (%)
At least 1 full-time equivalent

171 (85.9)

Less than 1 full-time equivalent

28 (14.1)

Hospital epidemiologist staffingd — no. (%)
At least 1 full-time equivalent

42 (21.1)

Less than 1 full-time equivalent

157 (78.9)

Percentages may not total 100 due to rounding.
a

Urban vs. rural location was determined based on 2010 U.S. Census data. Hospitals located in counties

that are part of metropolitan statistical areas were considered urban. Hospitals located in counties in
micropolitan statistical areas or rural areas were considered rural.
13

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Teaching hospitals were defined on the basis of membership in the Council of Teaching Hospitals, or

b

having an American Medical Association-approved residency program, or a self-reported or calculated
intern/resident to bed ratio of ≥0.25. This is similar to how teaching status was defined in the 2014
National Inpatient Sample.13 Teaching status was initially missing for one hospital; this hospital was
subsequently categorized as a teaching hospital based on information submitted by Emerging Infections
Program staff.
c

Hospitals were asked to submit staffing data from the most recent year for which data were available:

2015 (51 hospitals, 26%); 2014 (146, 73%); or 2013 (2, 1%). In one instance where data were reported in
aggregate for >1 hospital in the same system, Emerging Infections Program site staff were consulted,
and aggregated data were apportioned to each hospital.
Hospitals were asked to submit staffing data from the most recent year for which data were available:

d

2015 (52 hospitals, 26%); 2014 (145, 73%); 2013 (2, 1%). In one instance where data were reported in
aggregate for >1 hospital in the same system, Emerging Infections Program site staff were consulted,
and aggregated data were apportioned to each hospital.

14

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S3. Additional demographic and clinical characteristics of surveyed patients, 2015.

Characteristic

Patients

Patients

All Patients

without HAIs

with HAIs

P

(N=12,299)

(N=11,905)

(N=394)

Valuea,b

Sex — no. (%)

0.01

Female

6822 (55.5)

6628 (55.7)

194 (49.2)

Male

5476 (44.5)

5276 (44.3)

200 (50.8)

1 (<0.1)

1 (<0.1)

0

Missing data
Age category — no. (%)
<1 year

<0.001
1339 (10.9)

1319 (11.1)

20 (5.1)

1-17 years

527 (4.3)

514 (4.3)

13 (3.3)

18-24 years

457 (3.7)

444 (3.7)

13 (3.3)

25-44 years

1951 (15.9)

1910 (16.0)

41 (10.4)

45-64 years

3211 (26.1)

3056 (25.7)

155 (39.3)

65-84 years

3756 (30.5)

3634 (30.5)

122 (31.0)

≥85 years

1058 (8.6)

1028 (8.6)

30 (7.6)

Race — no. (%)

0.33

American Indian or Alaska Native

142 (1.2)

140 (1.2)

2 (0.5)

Asian

312 (2.5)

307 (2.6)

5 (1.3)

2007 (16.3)

1939 (16.3)

68 (17.3)

615 (5.0)

598 (5.0)

17 (4.3)

Black or African-American
Multiple races or other unspecified
race

15

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Patients

Patients

All Patients

without HAIs

with HAIs

P

(N=12,299)

(N=11,905)

(N=394)

Valuea,b

41 (0.3)

40 (0.3)

1 (0.3)

White

8161 (66.4)

7895 (66.3)

266 (67.5)

Missing data

1021 (8.3)

986 (8.3)

35 (8.9)

Characteristic
Native Hawaiian or other Pacific
Islander

Ethnicity — no. (%)
Hispanic or Latino

0.79
977 (7.9)

944 (7.9)

33 (8.4)

Not Hispanic or Latino

7991 (65.0)

7734 (65.0)

257 (65.2)

Missing data

3331 (27.1)

3227 (27.1)

104 (26.4)

Primary payer — no. (%)

0.39

Medicaid

2446 (19.9)

2377 (20.0)

69 (17.5)

Medicare

4952 (40.3)

4781 (40.2)

171 (43.4)

No charge

11 (<0.1)

10 (<0.1)

1 (0.3)

Other

309 (2.5)

300 (2.5)

9 (2.3)

Private

3850 (31.3)

3724 (31.3)

126 (32.0)

Self-pay

430 (3.5)

421 (3.5)

9 (2.3)

Missing data

301 (2.5)

292 (2.5)

9 (2.3)

Normal

3601 (29.3)

3464 (29.1)

137 (34.8)

0.12

Overweight

2887 (23.5)

2789 (23.4)

98 (24.9)

0.91

Obese

3846 (31.3)

3727 (31.3)

119 (30.2)

0.15

Body mass index categoryc — no. (%)

16

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”

Characteristic
Missing data

Patients

Patients

All Patients

without HAIs

with HAIs

P

(N=12,299)

(N=11,905)

(N=394)

Valuea,b

1965 (16.0)

1925 (16.2)

40 (10.2)

0.001

Outcome of hospitalization — no. (%)
Died

<0.001d
358 (2.9)

313 (2.6)

45 (11.4)

11,927 (97.0)

11,579 (97.3)

348 (88.3)

Still in hospital 6 months after survey

8 (<0.1)

7 (<0.1)

1 (0.3)

Missing data

6 (<0.1)

6 (<0.1)

0

Survived

Percentages may not total 100 due to rounding.
a

Chi-square test, unless otherwise indicated.
Comparison excludes patients with missing data, unless otherwise indicated.

b

c

Body mass index (BMI) categories were generated using reported or calculated body mass index for

patients ≥2 years of age. BMI was considered missing for children <2 years of age, even if BMI was
reported in the medical record. For adults (≥20 years), normal weight was BMI <25; overweight
25≤BMI<30; and obese BMI ≥30. For children (2–19 years), normal weight was BMI <85th percentile for
age and sex; overweight BMI between the 85th and 95th percentile for age and sex; and obese BMI ≥95th
percentile for age and sex.
Comparison includes patients who were known to have survived or died during the hospitalization;

d

patients still in the hospital and those with unknown outcome were excluded.

17

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S4. Comparison of additional, selected patient characteristics, 2011 vs. 2015 survey.

Characteristic

2011 Survey

2015 Survey

Patients

Patients

(N=11,282)

(N=12,299)

Sex — no. (%)

0.83

Female

6236 (55.3)

6822 (55.5)

Male

5034 (44.6)

5476 (44.5)

12 (0.1)

1 (<0.1)

Missing data
Age category — no. (%)
<1 year

0.08
1151 (10.2)

1339 (10.9)

1–17 years

479 (4.3)

527 (4.3)

18–24 years

462 (4.1)

457 (3.7)

25–44 years

1686 (15.0)

1951 (15.9)

45–64 years

3060 (27.1)

3211 (26.1)

65–84 years

3429 (30.4)

3756 (30.5)

≥85 years

1014 (9.0)

1058 (8.6)

1 (<0.1)

0

Missing data
Race — no. (%)

<0.001c

American Indian or Alaska Native

119 (1.1)

142 (1.2)

Asian

254 (2.3)

312 (2.5)

1905 (16.9)

2007 (16.3)

254 (2.3)

615 (5.0)

Black or African-American
Multiple races or other unspecified

P Valuea,b

race

18

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
2011 Survey

2015 Survey

Patients

Patients

(N=11,282)

(N=12,299)

20 (0.2)

41 (0.3)

White

7537 (66.8)

8161 (66.4)

Missing data

1193 (10.6)

1021 (8.3)

Characteristic
Native Hawaiian or other Pacific

P Valuea,b

Islander

Ethnicity — no. (%)

<0.001c

Hispanic or Latino

846 (7.5)

977 (7.9)

Not Hispanic or Latino

3715 (32.9)

7991 (65.0)

Missing data

6721 (59.6)

3331 (27.1)

Ventilator in place on survey date — no.

0.71

(%)
Yes

527 (4.7)

586 (4.8)

No

10,748 (95.3)

11,683 (95.0)

7 (<0.1)

30 (0.2)

6 (3–13)d

6 (3–13)d

Missing data
Median hospital length of stay among
patients who received antimicrobial
therapy at the time of the survey (or
information not available) (IQR)

Percentages may not total 100 due to rounding. IQR denotes interquartile range.
a

Chi-square test, unless otherwise indicated.
Comparison excludes patients with missing data, unless otherwise indicated.

b

c

Comparison includes patients with missing data.
19

0.15e

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Hospital length of stay data were missing for 53 patients in the 2011 survey and 2 patients in the 2015

d

survey. Excludes patients in the 2011 survey who were screen-positive for antimicrobial therapy at the
time of the survey based on a special criterion for dialysis patients. This criterion was not implemented
in the 2015 survey.
e

Median 2-sample test

20

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S5. Distribution of common HAI types in the 2015 survey, 2011 definitions vs. 2015 definitions.
No. of HAIs (%), 2011 HAI

No. of HAIs (%), 2015 HAI

Definitions (N=361)

Definitions (N=370)

Pneumonia

110 (30.5)

97 (26.2)

Gastrointestinal infection

91 (25.2)

95 (25.7)

Surgical site infection

69 (19.1)

88 (23.8)

Bloodstream infection

52 (14.4)

55 (14.9)

Urinary tract infection

39 (10.8)

35 (9.5)

Percentages may not total 100 due to rounding.

21

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S6. Multivariable log binomial regression model to identify variables associated with health careassociated infections (HAIs) in the subset of patients meeting the HAI review criterion, combined 2011
and 2015 survey populations (N=9118).
No. of
Total

Patients

Adjusted

95%

No. of

with

Risk

Confidence

Patients

HAIs

Ratio

Interval

P Value

Survey year 2015

4614

394

0.84

0.75–0.94

0.003

Ventilator on the survey datea

700

176

1.28

1.09–1.52

0.003

Survey date in May or Juneb

3662

310

0.88

0.78–1.00

0.04

Large hospital

1744

280

1.25

1.11–1.41

<0.001

Critical care unit on the survey date

1597

271

1.28

1.10–1.49

0.002

≤1 day

1881

27

Ref

--

--

2–4 days

3501

81

1.62

1.07–2.54

0.03

5–6 days

1144

76

4.59

3.02–7.19

<0.001

7–9 days

942

127

8.95

6.06–13.74

<0.001

10–12 days

480

122

16.17

10.98–24.76

<0.001

13–20 days

606

174

18.38

12.62–27.93

<0.001

≥21 days

564

239

24.15

16.66–36.57

<0.001

Variable*

Time from admission to survey

*

Other variables that were tested but found not to be statistically significant predictors of HAI risk were

age and presence of a central line or urinary catheter on the survey date.
a

Ventilator presence was unknown for 17 patients without HAIs and 1 patient with HAI (patients with

unknown ventilator status were grouped with patients without ventilators for analysis).
22

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Survey dates were categorized as being in May–June versus July–September.

b

23

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S7. Multivariable log binomial regression model to identify variables associated with health careassociated infections (HAIs) in the subset of patients in 148 hospitals that participated in both the 2011
and 2015 surveys, combined 2011 and 2015 survey populations (N=18,451).
No. of
Total No.

Patients

Adjusted

95%

of

with

Risk

Confidence

Patients

HAIs

Ratio

Interval

P Value

Survey year 2015

9169

297

0.78

0.68–0.90

<0.001

Ventilator on the survey datea

877

139

1.69

1.40–2.02

<0.001

Central line on the survey dateb

3371

382

1.87

1.59–2.20

<0.001

Urinary catheter on the survey datec

3875

241

1.18

1.01–1.39

0.04

Large hospital

4310

255

1.24

1.07–1.43

0.004

≤1 day

5408

20

Ref

Ref

—

2–4 days

7043

69

2.43

1.51–4.10

<0.001

5–6 days

1688

56

6.93

4.24–11.81

<0.001

7–9 days

1480

105

13.68

8.68–22.71

<0.001

≥10 days

2832

430

26.52

17.30–43.14

<0.001

6389

166

Ref

Ref

—

10,448

456

1.49

1.26–1.78

<0.001

1614

58

1.76

1.31–2.31

<0.001

Variable*

Time from admission to survey

Aged
<45 years
45–84 years
≥85 years
*

Other variables that were tested but found not to be statistically significant predictors of HAI risk were

survey month (May–June versus July–September) and location in a critical care unit on the survey date.
24

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
a

Ventilator presence was unknown for 26 patients without HAIs and 0 patients with HAI (patients with

unknown ventilator status were grouped with patients without ventilators for analysis).
Central line presence was unknown for 51 patients without HAIs and 0 patients with HAI (patients with

b

unknown central line status were grouped with patients without central lines for analysis).
c

Urinary catheter presence was unknown for 45 patients without HAIs and 3 patients with HAI (patients

with unknown catheter status were grouped with patients without urinary catheters for analysis).
Model excluded 1 patient without HAIs in the 2011 survey for whom age was unknown.

d

25

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S8. Multivariable log binomial regression model to identify variables associated with health careassociated infections (HAIs), excluding the presence of devices, in the subset of patients in 148 hospitals
that participated in both the 2011 and 2015 surveys, combined 2011 and 2015 survey populations
(N=18,451).
No. of
Total

Patients

Adjusted

95%

No. of

with

Risk

Confidence

Patients

HAIs

Ratio

Interval

P Value

Survey year 2015

9169

297

0.76

0.66–0.87

<0.001

Critical care unit on the survey date

2790

212

1.58

1.35–1.85

<0.001

Large hospital

4310

255

1.28

1.11–1.49

<0.001

≤1 day

5408

20

Ref

Ref

--

2–4 days

7043

69

2.51

1.56–4.24

<0.001

5–6 days

1688

56

7.74

4.74–13.17

<0.001

7–9 days

1480

105

16.39

10.43–27.14

<0.001

≥10 days

2832

430

35.90

23.59–58.08

<0.001

<40 years

5739

143

Ref

Ref

--

40–50 years

1708

63

1.76

1.32–2.33

<0.001

51–65 years

4179

208

2.13

1.73–2.62

<0.001

66–69 years

1203

43

1.66

1.19–2.28

0.002

≥70 years

5622

223

2.15

1.75–2.65

<0.001

Variable*

Time from admission to survey

Agea

26

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
*

Survey month (May–June versus July–September) was also tested but was not found to be a statistically

significant predictor of HAI risk.
a

Model excluded 1 patient without HAIs in the 2011 survey for whom age was unknown.

27

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S9. Log-binomial regression model to identify factors associated with HAIs among patients surveyed in 2015 (N=12,299).
Full, Final Model

Final Model* For Burden Estimation

Total

No. of

No. of

Patients

Adjusted

Confidence

Patients

with HAIs

Risk Ratio

Interval

P Value

≤1 year

1388

22

Ref

Ref

—

2–26 years

1119

26

2.33

1.34–4.05

0.003

27–51 years

2574

72

2.94

1.84–4.70

52–64 years

2404

122

4.10

65–77 years

2607

82

≥78 years

2207

≤4 days
5–6 days

Factor

95%

95%
Adjusted

Confidence

Risk Ratio

Interval

P Value

Ref

Ref

—

<0.001

2.26

1.58–3.22

<0.001

2.62–6.42

<0.001

3.21

2.32–4.44

<0.001

2.89

1.82–4.61

<0.001

2.19

1.54–3.11

<0.001

70

3.77

2.35–6.04

<0.001

2.71

1.88–3.90

<0.001

5861

20

Ref

Ref

—

1427

15

2.66

1.36–5.19

0.004

Ref

Ref

—

Agea

Hospital length of
stayb

28

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Full, Final Model
Total

No. of

No. of

Patients

Adjusted

Confidence

Patients

with HAIs

Risk Ratio

Interval

7–8 days

1064

26

5.84

9–14 days

1543

75

15–23 days

992

≥24 days

Final Model* For Burden Estimation

95%

95%
Adjusted

Confidence

P Value

Risk Ratio

Interval

P Value

3.26–10.44

<0.001

4.69

2.83–7.75

<0.001

11.22

6.83–18.42

<0.001

9.43

6.34–14.04

<0.001

85

17.70

10.79–29.04

<0.001

16.98

11.51–25.03

<0.001

1412

173

26.90

16.58–43.67

<0.001

28.83

20.12–41.32

<0.001

Ventilatorc

586

81

1.53

1.21–1.93

<0.001

Not included in final model

Central lined

2081

213

1.88

1.52–2.32

<0.001

Not included in final model

11,719

381

1.88

1.11–3.19

0.02

Not included in final model

3557

176

1.37

1.11–1.70

0.004

Not included in final model

1474

56

0.60

0.45–0.81

<0.001

Not included in final model

Factor

Rural hospitale
Hospital with >400
licensed bedsf
Hospital with 500–
800 licensed bedsf
*

The final model for burden estimation included factors significant in multivariable models and available in the prevalence survey dataset and in

the 2014 National Inpatient Sample.
29

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
a

Age categories ≤1 year and 2–26 years were collapsed in the final model for burden estimation due to no health care-associated infection

events in certain categories of age and length of stay.
Hospital length of stay was available for 12,290 patients; length of stay was unknown for 1 patient, and 8 patients were still in the hospital at

b

least 6 months after the survey date. Time from admission to the date of follow-up (≥6 months following the survey date) was used as a proxy
for hospital length of stay in patients who remained in the hospital for more than 6 months after the survey date. Hospital length of stay
categories ≤4 days and 5–6 days were collapsed in the final model for burden estimation due to there being no health care-associated infection
events in certain categories of age and length of stay.
c

Ventilator presence was unknown for 29 patients without HAIs and 1 patient with HAI (patients with unknown ventilator status were grouped

with patients without ventilators for analysis).
Central line presence was unknown for 42 patients without HAIs and 1 patient with HAI (patients with unknown central line status were

d

grouped with patients without central lines for analysis).
Hospitals were categorized as urban versus rural based on U.S. Census data; hospitals located in a metropolitan county were considered urban,

e

and hospitals located in a micropolitan or rural county were considered rural.
f

Hospitals were asked to submit licensed bed data from the most recent year for which data were available: 2015 (39 hospitals, 20%); 2014 (157,

79%); 2013 (3, 2%). In one instance where data were reported in aggregate for >1 hospital in the same system, Emerging Infections Program site
staff were consulted, and aggregated data were apportioned to each hospital.

30

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
Table S10. Estimated numbers of health care-associated infections in the United States in 2015.
Percentage of Patients

Estimated Infectionsb in the

with Infection Typea

United States

No. of Infections

(95% Confidence Interval)

(95% Confidence Interval)

Pneumonia

110

27.9 (23.7–32.5)

176,700 (51,200–621,600)

Gastrointestinal infection

91

23.1 (19.1–27.5)

146,300 (41,300–526,000)

Surgical-site infection

69

17.5 (14.0–21.5)

110,800 (30,200–411,200)

Bloodstream infection

52c

13.2 (10.1–16.8)

83,600 (21,800–321,300)

Urinary tract infection

39

9.9 (7.2–13.2)

62,700 (15,600–252,500)

Skin and soft tissue infection

22

5.6 (3.6–8.2)

35,500 (7,800–156,800)

Eye, ear, nose throat and mouth infection

21d

5.3 (3.4–7.9)

33,600 (7,300–151,100)

Lower respiratory infection

18

4.6 (2.8–7.0)

29,100 (6,000–133,900)

Bone and joint infection

2

0.5 (0.08–1.7)

3,200 (200–32,500)

Central nervous system infection

1

0.3 (0.01–1.2)

1,900 (0–23,000)

Cardiovascular infection

1

0.3 (0.01–1.2)

1,900 (0–23,000)

Reproductive tract infection

1

0.3 (0.01–1.2)

1,900 (0–23,000)

Systemic infection

0

0 (0–0.8)

0 (0–15,300)

Infection Type

31

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”

Infection Type

No. of Infections

Total
a

Percentage of Patients

Estimated Infectionsb in the

with Infection Typea

United States

(95% Confidence Interval)

(95% Confidence Interval)
687,200 (181,400–2,691,200)

Among the 394 surveyed patients with health care-associated infections, the percentage with each infection type.
Estimates are based on the total number of patients with health care-associated infections (and upper and lower bounds of the 95% CI), prior to

b

rounding to the nearest hundred, multiplied by the rounded proportions (and upper and lower bounds of the 95% CIs) of patients with each type
of infection. These products were then rounded to the nearest hundred to estimate the total numbers of each HAI. The rounded products were
added together to determine the total number of all HAIs. For the purposes of burden estimation, we assumed each infection occurred in a
unique patient.
c

One patient had 2 separate bloodstream infections. For the purposes of burden estimation, we assumed that each of these 52 infections

occurred in a unique patient.
One patient had 2 separate eye, ear, nose, throat and mouth infections. For the purposes of burden estimation, we assumed that each of these

d

21 infections occurred in a unique patient.

32

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
References
1) Magill SS, Edwards JR, Bamberg W, et al. Multistate point-prevalence survey of health careassociated infections. N Engl J Med 2014;370:1198-208.
2) Harris PA, Taylor R, Thielke R, Payne J, Gonzalez N, Conde JG. Research electronic data capture
(REDCap) – A metadata-driven methodology and workflow process for providing translational
research informatics support. J Biomed Inform 2009;42:377-81.
3) Horan TC, Andrus M, Dudeck MA. March 2010 update to: CDC/NHSN surveillance definition of
healthcare-associated infection and criteria for specific types of infections in the acute care
setting. Am J Infect Control 2008;36:309-32
(https://www.cdc.gov/nhsn/pdfs/archive/17pscNosInfDef_NOTcurrent.pdf).
4) National Healthcare Safety Network 2015 validation manual. Atlanta: Centers for Disease
Control and Prevention (https://www.cdc.gov/nhsn/pdfs/validation/2015/2015-validationmanual.pdf).
5) HCUP National Inpatient Sample (NIS). Healthcare Cost and Utilization Project (HCUP). 2014.
Rockville, MD: Agency for Healthcare Research and Quality (www.hcupus.ahrq.gov/nisoverview.jsp).
6) Rhame FS, Sudderth WD. Incidence and prevalence as used in the analysis of the occurrence of
nosocomial infections Am J Epidemiol 1981;113: 1-11.
7) Zing W, Huttner BD, Sax H, Pittet D. Assessing the burden of healthcare-associated infections
through prevalence studies: what is the best method? Infect Control Hosp Epidemiol
2014;35:674-84.
8) Gastmeier P, Bräuer H, Sohr D, et al. Converting incidence and prevalence data of nosocomial
infections: results from eight hospitals. Infect Control Hosp Epidemiol 2001;22:31-4.

33

Supplementary Appendix, “Changes in Prevalence of Health Care-Associated Infections in U.S. Hospitals”
9) Berthelot P, Garnier M, Fascia P, Guyomarch S, et al. Conversion of prevalence survey data on
nosocomial infections to incidence estimates: a simplified tool for surveillance? Infect Control
Hosp Epidemiol 2007;28:633-6.
10) Kanerva M, Ollgren J, Virtanen MJ, Lyytikäinen O; Prevalence Survey Study Group. Estimating
the annual burden of health care-associated infections in Finnish adult acute care hospitals. Am
J Infect Control 2009;37:227-30.
11) Ustun C, Hosoglu S, Geyik MF, Parlak Z, Ayaz C. The accuracy and validity of a weekly pointprevalence survey for evaluating the trend of hospital-acquired infections in a university hospital
in Turkey. Int J Infect Dis 2011;15:e684-7.
12) Meijs AP, Ferreira JA, De Greeff SC, Vos MC, Koek MB. Incidence of surgical site infections
cannot be derived reliably from point prevalence survey data in Dutch hospitals. Epidemiol
Infect 2017;145:970-80.
13) Introduction to the HCUP National Inpatient Sample (NIS) 2014. Rockville, MD: Agency for
Healthcare Research and Quality, November 2016 (https://www.hcupus.ahrq.gov/db/nation/nis/NIS_Introduction_2014.jsp).

34