Coagulase-Negative Staphylococcal Bacteremia Among Very Low Birth Weight Infants: Relation to Admission Illness Severity, Resource Use, and Outcome

(1)

Coagulase-Negative

Staphylococcal

Bacteremia

Among

Very

Low

Birth

Weight

Infants:

Relation

to Admission

Illness

Severity,

Resource

Use,

and

Outcome

James E. Gray, MD, SM*; Douglas K. Richardson, MD, MBA*;

Marie C. McCormick, MD, ScD*; and Donald A. Goldmann, MD

ABSTRACT. Objective. To examine the impact of

ad-mission-day illness severity on nosocomial bacteremia

risk after consideration of traditional risk determinants such as birth weight and length of stay.

Methods. The hospital courses for 302 consecutive

very low birth weight (less than 1500 g) infants admitted to two neonatal intensive care units were examined for the occurrence of nosocomial coagulase-negative staphy-lococcal bacteremia. Using both cumulative incidence and incidence density as measures of bacteremia risk, we explored the relation between illness severity (as mea-sured by the Score for Neonatal Acute Physiology [SNAP]) and bacteremia both before and after birth weight adjustment. In addition, the effect of bacteremia on hospital resource use was estimated.

Results. Coagulase-negative staphylococcus was the

most common pathogen noted in blood cultures drawn at

48 hours after admission or later. It was isolated on at

least one occasion in 53 patients (cumulative incidence of

17.5 first episodes per 100 patients). These episodes

oc-curred during 7652 days at risk, giving an incidence

den-sity of 6.9 initial bacteremias per 1000 patient-days at

risk. As expected, when compared with the

nonbacter-emic group, bacteremic patients were of lower birth

weight (888 ± 231 vs 1127 ± 258 g P < .01) and gestational

age (26.4 ± 2.1 vs 28.9 ± 2.8 weeks; P < .01). In addition, these patients were more severely ill on admission (SNAP 17.3 ± 6.5 vs 12.2 ± 5.8; P < .01). Even after birth

weight stratification, the risk of bacteremia by both mea-sures increased with higher SNAP scores. For example, among infants with birth weights greater than 1 kg, 25%

of the most severely ill patients (SNAP 20 and higher)

experienced at least one bacteremic episode, whereas the rates seen in infants with intermediate (SNAP 10 to 19) and low illness severity (SNAP 0 to 9) were 8.6% and

3.0%, respectively ( for trend = 7.25; P < .01).

Multivar-iate linear regression showed that bacteremia was asso-ciated with a prolongation of neonatal intensive care unit stay of 14.0 ± 4.0 days (P < .01) and an increase in hospital

charges of $25 090 ± 12 051 (P < .05), even after adjust-ment for birth weight and admission-day SNAP.

From the *Joint Program in Neonatology (Brigham and Women’s Hospital, Children’s Hospital, Beth Israel Hospital, and Harvard Medical School); tthe Department of Maternal and Child Health, Harvard School of Public Health; and §the Hospital Epidemiology Program (Children’s Hospital and Harvard Medical School), Boston, MA.

Presented, in part, at the Society for Pediatric Research Annual Meeting, Washington, May 1993.

Received for publication Mar 3, 1994; accepted Jun 13, 1994.

Reprint requests to (J.E.G.) Joint Program in Neonatology, 221 Longwood Avenue, Boston, MA 02115.

Conclusions. Nosocomial coagulase-negative

bacter-emia is an important complication among very low birth

weight infants. Assessment of illness severity with SNAP

provides information regarding nosocomial infection

risk beyond that available from birth weight alone.

Pediatrics 1995;95:225-230; nosocomial infection,

ill-ness severity index, neonatal intensive care, quality of

care, quality improvement.

ABBREVIATIONS. VLBW, very low birth weight; NICU, neonatal intensive care unit; SNAP, Score for Neonatal Acute Physiology;

NTISS, Neonatal Therapeutic Intervention Scoring System; CI,

confidence interval; NNIS, National Nosocomial Infection Survey.

Very low birth weight (VLBW) (less than 1500 g)

infants are at high risk for nosocomial bacteremia

because of both their immature host defenses and

their frequent exposure to invasive diagnostic and

therapeutic procedures. Birth weight, as a marker of

both immaturity and the need for such procedures, is

an important, well-studied determinant of risk.1

However, differences in birth weight alone are

insuf-ficient to explain much of the impressive variation in

nosocomial infection rates observed among neonatal

intensive care units (NICUs).2 These differing

infec-tion rates can be reconciled further by adjusting for

the use of central venous catheters in each NICU, as

exposure to central catheters is a major risk factor for

nosocomial bacteremia in neonates. Such

standard-ized infection rates usually are expressed as

infec-tions per 1000 catheter days. Although useful for

inter-NICU comparisons, these rates do not adjust

fully for the impact of the local practice style and

potential over- or under-utilization of invasive

ther-apies. A more unbiased estimate could be achieved

by stratifying for the underlying severity of illness of

a NICU population.

For comparative data to serve as a valid basis for

quality improvement efforts, neonatologists require

better measures of severity of illness. Until recently,

objective measures of neonatal illness severity have

not been available.3 The development of validated

measures, such as the Score for Neonatal Acute

Physiology (SNAP)4 and the Clinical Risk Index for

Babies,5 has proven useful in examining other

ad-verse outcomes of NICU care. Such measures may

provide an opportunity to assess more accurately the

relation between illness severity and nosocomial

bac-teremia. Accordingly, we used SNAP to study

at Viet Nam:AAP Sponsored on September 1, 2020

www.aappublications.org/news

(2)

ulase-negative staphylococcal bacteremia in a cohort

of VLBW infants admitted to two NICUs.

METHODS 36 (11.9%) 78 (25.8%) 188 (62.3%) 46 (15.2%) 115 (38.1%) 112 (37.1%) 29 (9.6%) 150 (49.7%) 208 (68.9%) 47 (15.6%) 47 (15.6%) 13.1±6.2 16.9 ±7.4 Patient Population

The study population consisted of all VLBW infants admitted

to two NICUs between November 1, 1989 and September 30, 1990. The Brigham and Women’s Hospital, a large high-risk perinatal center, cares for inborn patients, whereas Children’s Hospital, a large general pediatric hospital, maintains a mixed medical/sur-gical NICU and cares for outborn infants. Medical care in both NICUs is provided by a single group of neonatologists and neo-natal fellows. These infants were enrolled in a larger parent study which developed and validated SNAP4 and the Neonatal Therapeutic Intervention Scoring System (NTISS).6

Data Collection

The following information was collected for each patient.

Patient Characteristics

Sex, race, birth weight, and gestational age were recorded from the medical record.

Hospital Course

Length of stay and discharge disposition (coded as alive and discharged home, alive and transferred to community hospital, or died) were recorded.

Admission-Day SNAP and NTISS Scores

Information needed for calculating these scores was collected as described previously.4-6 The SNAP is a physiology-based sever-ity of illness measure that examines a number of routinely ob-tamed laboratory and clinical parameters to quantify the degree of a patient’s illness, The NTISS measures illness severity indirectly by quantifying the intensity of therapy received by a patient.

Microbiology Information

Blood cultures were ordered at the discretion of the primary care team. Generally, this was done for indications such as apnea and bradycardia, temperature instability, feeding intolerance, or the need for increased respiratory support. The timing and results of blood cultures were obtained from the hospital’s microbiology reporting and billing systems. Nosocomial coagulase-negative staphylococcal bacteremia was defined according to Freeman et al.7 Accordingly, isolation of coagulase-negative staphylococci alone from a single blood culture was considered evidence of bacteremia due to that organism. We considered patients to be at risk for initial nosocomial coagulase-negative staphylococcal bac-teremia from the third hospital day until hospital discharge or the date of the first positive culture. We defined the risk period for recurrent coagulase-negative staphylococcal bacteremia as begin-ning 7 days after a previous infection and extending until the patient’s discharge from the NICU or next positive culture. We chose this duration of non-risk after a positive culture because I week is the usual duration of antibiotic therapy for coagulase-negative staphylococcal bacteremia in these NICUs. The total days at risk for each patient were calculated as the sum of days at risk for initial and recurrent bacteremia.

Hospital Resource Use

Total hospital charges were obtained from each hospital’s uni-form discharge abstract. These charges included per diem and ancillary charges (laboratory, radiology, pharmacy, respiratory

care, physical/occupational therapy, and social work charges), but not physician charges. For patients transferred between the two study hospitals, total charges were calculated as the arithmetic sum of the individual hospital charges.

Analysis

To examine the occurrence of initial coagulase-negative staph-ylococcal bacteremia, we calculated crude and hirth-weight-spe-cific cumulative incidences as the number of patients experiencing bacteremia divided by the number at risk. To control for the

confounding influence of length of stay on cumulative incidence, we also calculated crude and birth-weight-specific incidence den-sities. Incidence density for bacteremic episodes was calculated as the number of episodes divided by the number of days at risk, as defined above. The univariate relations between patient charac-teristics, therapies administered, and the occurrence of bacteremia were analyzed using Student’s f-test and testing, whereas mul-tivanate associations between these were examined using multi-ple logistic regression. Poisson regression was used to examine the relations between SNAP, birth weight, and incidence density. Multiple linear regression was used to describe the relations

be-tween hospital charges, patient characteristics, and nosocomial bacteremia. In these multivariate models, SNAP was entered as a continuous term, whereas birth weight was represented by cate-goncal variables (ie, birth weight less than 750 g, 750 to 999 g,and 1000 to 1499 g). Analyses were performed using the Statistical Analysis System version 6.07 (SAS Institute, Cary, NC) and STATA Release 3.01 (STATA Corporation, College Station, TX). Data are presented as mean ± standard deviation.

Human Subjects

The study was approved by the human subjects committees at both hospitals.

RESULTS

Study Population

During the study period, 344 VLBW infants were cared for in the study NICUs. Patient characteristics

are presented in Table I. The in-hospital mortality

rate was 21 .5% (74 deaths). As expected, mortality

rates increased with decreasing birth weight. Among

infants with birth weights less than 750 g, the rate

was 62.1 %, whereas the rate was 24.4% in patients

with birth weights between 750 and 1000 g and 8.1%

among those born weighing more than 1000 g. A

total of 42 infants were excluded from the analysis as

not being at risk for nosocomial infections (38 infants

died before 48 hours and 4 were transferred to

an-other institution before 48 hours). Thus, the final

number of infants included in this study was 302.

Blood Culture Results

A total of 522 cultures were obtained at least 48

hours after admission in study patients. The average

number of blood cultures per patient was 1 .7 ± 2.8

TABLE 1. Population Characteristics*

Patients at Risk for Nosocomial

Bacteremiat Birth weight, g

<750 750-999

1000

Gestational age, wk

<26 26-28 29-32 >32 Male gender Race White Black Other/unknown Admission-day SNAP Admission-day NTISS

* Abbreviations: SNAP, Score for Neonatal Acute Physiology; NTISS, Neonatal Therapeutic Intervention Scoring System. t Excludes patients remaining in the neonatal intensive care unit for less than 48 hours because of either death or transfer to another facility.

(3)

29-42

Day ofhospitalization

Fig 1. Risk of bacteremia by day of hospitalization.

A

40%

8

13

I

20%

0 10%

0% 0-9 10-19

Mmission day SNAP

B

Incidence density (cases/bOO days at risk) 20

15

10

0-9

TABLE 2. Organisms Isolated From Blood Cultures Drawn at Least 48 Hours After Admission

20+

20+

(range, 0 to 19). Organisms were isolated from 130 of

these 522 cultures (24.9% positive rate). The

organ-isms and their frequencies are noted in Table 2.

Co-agulase-negative staphylococcus was the most

com-mon pathogen, representing 77.6% of positive

cultures. All subsequent results will pertain to these

101 cultures that grew coagulase-negative

staphy-lococci and were drawn at least 48 hours after

admission.

Coagulase-Negative Staphylococcal Bacteremia

Coagulase-negative staphylococci were isolated on

at least one occasion from 53 patients, giving a

cu-mulative incidence of 17.5%. There was no difference

in cumulative incidence between inborn and outborn

infants (17.8% vs 15.8%;

P

= .76). Sixty-two of the 101

positive cultures from these 53 patients occurred

during at-risk periods and were therefore considered

to represent distinct episodes of bacteremia. The

oth-ers represented duplicate positive cultures during

the same episodes. The average age at initial

bacter-emia was 22.0 ± 15.6 days. The initial bacteremic

episodes occurred during 7652 at-risk days (6.9

epi-sodes per 1000 patient-days at risk). The risk of initial

bacteremia varied by period of hospitalization, as

seen in Fig I. This risk decreased over the course of

hospitalization

(

for trend = 5.60,

P

< .01).

Recurrent bacteremia affected only 8 of 53

bacter-emic patients (15.1 %). Seven patients experienced a

single recurrence, whereas one infant experienced

two. These recurrent episodes arose during 1976

at-risk days (4.0 episodes per 1000 days at risk). The

combined risk of all bacteremias (ie, initial and

recurrent) was 6.3 per 1000 patient-risk days.

Relation of Initial Bacteremia Risk to Patient Characteristics,

Therapies, and Illness Severity

Univariate Analyses. Patients with at least one

epi-sode of coagulase-negative staphylococcal

bactere-mia and their nonbacteremic counterparts did not

differ with respect to gender, race, or presence of

1-or 5-minute Apgar scores less than 7. As expected,

bacteremic patients had lower birth weights (888 ±

231 vs 1127 ± 258 g; P < .01) and gestational ages (26.4

± 2.1 vs 28.9 ± 2.8 weeks; P < .01). The bacteremic

group was more severely ill (admission-day SNAP

17.3 ± 6.5 vs 12.2 ± 5.8; P < .01) and received more

intense therapy (admission day NTISS 21.1 ± 7.0 vs

Organism No. of Isolates

Bacillus sp 1

Candida sp 3

Coagulase-negative staphylococci 101

Escherichia coli 5

Enterobacter cloacae 3

Group B streptococci 2

Klebsiella pneumoniae I

Micrococcus sp 1

Mucor sp I

Pseudomonas aeruginosa 2

Staphylococcus aureus 8

Streptococcus faecalis 2

16.0 ± 7.2; P < .01). On the first hospital day,

bacter-emic infants were more likely to have required

me-chanical ventilation (62.3% vs 44.8%; P < .05), colloid

support (54.7% vs 25.3%; P < .001), vasopressor

sup-port (37.7% vs 14.1%; P < .01), arterial pressure

mon-itoring (62.3% vs 48.6%; P < .01), central line

place-ment (30.2% vs 14.5%; P < .01), packed red-cell

transfusion (62.3% vs 33.3%; P < .01), frequent blood

drawing (90.6% vs 74.7%; P < .01), and phototherapy

(79.3% vs 42.2%; P < .01).

As seen in Fig 2A, the risk of initial bacteremia, as

measured by attack rate, rose steadily with

increas-ing illness severity (x2 for trend = 23.1, P < .01). The

sickest infants experienced a bacteremia rate almost

15 times that seen in the lowest severity group. To

determine whether this association resulted simply

10-19

Mmissbn day SNAP

Fig 2. A, Relation between admission-day Score for Neonatal Acute Physiology (SNAP) and coagulase-negative staphylococcal bacteremia among VLBW infants, based on cumulative incidence. B, Relation between admission-day SNAP and coagulase-negative staphylococcal bacteremia among VLBW infants, based on mci-dence density.

(4)

0-9 10-19 20+

Admis.bn day SNAP

B

Incidence density (cases/bOO days at risk)

25

20

15

10

5

n 0-9 20+

Birth weight

01000-1499

U750-999

. <750

from the fact that more severely ill infants (ie, higher

SNAP) remained in the NICU and therefore at risk

for bacteremia for a longer period, we repeated these

analyses using incidence density (episodes per 1000

patient-days) as the measure of bacteremia risk.

Again, as seen in Fig 2B, higher SNAP scores were still

associated with an increased risk of initial bacteremia.

Multivariate Analyses. Even after consideration of

birth weight, there remained a gradient of risk

asso-ciated with illness severity, as seen in Fig 3A and 3B.

Using multiple regression, we found that each

5-point increment in SNAP was associated with a

53.9% increase (95% confidence interval [CII: 33.0%,

77.2%) in a patient’s risk of experiencing at least one

nosocomial bacteremia episode while in the NICU.

We found very similar results using Poisson

regres-sion, in which a 5-point SNAP increment was

asso-ciated with a 56% increase (95% CI: 23%, 97%) in

episodes per 1000 days at risk. For example, consider

two infants born weighing 1000 g, but one with an

admission SNAP of 10 and the second with a score of

15. According to the results of the multiple logistic

regression models, the first infant has a 6% chance of

developing at least one nosocomial bacteremia

epi-sode while in the NICU, whereas the risk for the

latter is 54% higher, or 9.3%. Similarly, infants such

as the first infant would be expected to experience 5.8

episodes of initial bacteremias per 1000 days at risk,

whereas the rate for the second infant would be 9.0

episodes per 1000 days at risk.

We next examined the association between

admis-sion-day therapies and bacteremia attack rate after

A Birth weight

Io1000-1499

I9750-999

60%

I

40% 30%

c 20%

10%

0%

10-19

Mmis.bn day SNAP

Fig 3. A, Coagulase-negative staphylococcal bacteremia risk strat-ified by admission-day Score for Neonatal Acute Physiology (SNAP) and birth weight, based on cumulative incidence. B, Co-agulase-negative staphylococcal bacteremia risk stratified by ad-mission-day SNAP and birth weight, based on incidence density.

adjusting for birth weight and illness severity. Those

NTISS items that exceeded a threshold level of

uni-variate significance (P < .15) as suggested by Hosmer

and Lemeshow were entered into separate multiple

logistic models along with birth weight and SNAP.

After this adjustment, only phototherapy and blood

pressure support with colloid or vasopressor

admin-istration demonstrated significant multivariate

asso-ciations with nosocomial bacteremia (odds ratios 2.9

and 2.2, respectively). It should be noted that we

examined only those therapies administered on

ad-mission day, and did not include later interventions

such as central line placement or intralipid infusions.

Relation of Bacteremia to Resource Use and Patient Outcomes

Univariate Analyses. The NICU length of stay was

significantly greater for the bacteremic group (66.4

± 36.4 vs 28.6 ± 29.6 days; P < .01), as were total

hospital charges ($128 158 ± 95 259 vs $52 270 ±

66 533; P < .01). A trend toward increased

in-hos-pital mortality was seen among bacteremic infants

(18.9% vs 10.8%;

P

= .11).

Multivariate Analyses. The independent

contribu-tions of patient characteristics, illness severity, and

the occurrence of nosocomial bacteremia to NICU

length of stay in survivors were investigated using

multiple linear regression. In addition to these three

factors, community retrotransport was included

be-cause of the high frequency of its use in these two

NICUs. The results of these model computations

ex-plained 60.1 % of the variation in NICU hospital

length of stay, and are shown in Table 3. Birth

weight, admission-day SNAP, and retrotransport

carried the greatest explanatory power, but even

af-ter consideration of these predictors, the presence of

coagulase-negative staphylococcal bacteremia was

significantly associated with prolongation of hospital

stay. The adjusted increment in length of stay

attrib-utable to bacteremia was 14.0 ± 4.0 days. Similar

models predicting total hospital charges explained

39% of the variation in charges, with each bacteremic

patient accruing $25 090 ± 12 051 in additional

charges.

Next, we examined the multivariate association of

nosocomial bacteremia with in-hospital mortality.

After adjustment for birth weight and admission-day

SNAP, the odds ratio for in-hospital death associated

with bacteremia changed from its univariate value of

1.89 (95% CI: 0.86, 4.24) to 0.50 (95% CI: 0.18, 1.35).

Although these results could be interpreted as

re-flecting a direct protective effect of bacteremia, this is

unlikely to be the case. Rather, nosocomial infection

usually occurs in the third week of hospitalization,

and therefore is a marker for survival past the initial

period of highest mortality risk. In this population,

62 of 75 (83%) of deaths occurred before the mean

age at the time of initial bacteremia.

DISCUSSION

The 17.5% cumulative incidence of nosocomial

bacteremia seen in the current study is similar to

rates reported by other investigators. For example,

Baumgart et al9 found that 12% of infants admitted to

(5)

TABLE 3. Hospital Length of Stay (LOS) Prediction Model for Neonatal Intensive Care Unit Very Low Birth Weight Survivors

Step Variable Entered Parameter Estimate F-Test P Value Partial R2 Model R2 Predicted Change in LOS, d

Intercept 26.8 53.7 .0091

1) Birth weight <750 g* 43.5 54.6 .0001 .37 .37 +43.5

2) SNAP (per 5 points)t 7.9 31.51 .0001 .090 .46 +7.9

3) Retrotransport* -24.5 49.1 .0001 .10 .56 -24.5

4) Birth weight 750-999 g* -19.4 1.8 .0005 .03 .59 -19.4

5) Nosocomial bacteremia* 14.0 18.9 .0005 .02 .61 +14.0

*Predicted change associated with dichotomous variable is equal to change associated with presence of characteristic.

t Predicted change in charges associated with a one-unit change in predictor variable. Abbreviation: SNAP, Score for Neonatal Acute Physiology.

their NICU experienced at least one bacteremic

epi-sode, and Freeman et aP#{176}reported that rates among

VLBW infants ranged from 5% to 30%, depending on

birth weight. To adjust for exposure to the NICU

environment, we calculated the incidence density of

bacteremia, or the number of bacteremic episodes

per 1000 patient-days at risk. Comparable incidence

density data are not available in the literature. The

Centers for Disease Control in the National

Nosoco-mial Infection Survey (NNIS) adjusts bacteremia

rates by exposure to central venous lines, expressed

as cases of bacteremia per 1000 central-line days;

detailed information regarding exposure to central

lines was not available in our data set. As in a

pre-vious study,9 we detected no increase in mortality

attributable to bacteremia, presumably because

co-agulase-negative staphylococcal bacteremia occurs

principally in neonates who have survived the

pe-nod of highest mortality risk during the first days of

life. It should be noted that we considered only

co-agulase-negative staphylococcal bacteremia in our

analyses of mortality. Extrapolation of these findings

to nosocomial infections caused by other organisms

is not warranted.

In addition to reconfirming the importance of

co-agulase-negative staphylococcal bacteremia as a

complication of NICU care, we have demonstrated

that the admission-day SNAP is a valuable predictor

of subsequent nosocomial infection risk. The

associ-ation between increased bacteremia rates and higher

SNAP persists even after consideration of two potent

confounders of infection risk, birth weight and

hos-pital length of stay. This finding has important

implications for the study of NICU nosocomial

in-fection, as it provides a method for improved

char-acterization of bacteremia risk in NICU populations.

Measures of illness severity other than birth

weight have not been applied previously to the study

of neonatal nosocomial bacteremia risk, but

rela-tively crude severity analyses have been performed

in pediatric intensive care units. For example,

Pol-lock et al’1 categorized 480 pediatric intensive care

unit patients into low and high severity groups using

the Pediatric Risk of Mortality score and found more

than a threefold increase in the cumulative incidence

of nosocomial infection associated with high

sever-ity. However, when length of stay was considered in

these analyses, illness severity was not associated

with infection rates

(P

= .10). In the NNIS, pediatric

intensive care unit patients are categorized using the

Average Severity of Illness Score, a scoring system

that has not been validated thoroughly.

Notably, the neonatal component of the NNIS

stratifies patients only on birth weight, with no

at-tempt to define illness severity. Data from the NNIS

has been valuable in documenting significant

van-ability in nosocomial infection rates among NICUs.

However, the lack of adequate illness-severity

ad-justment in the NNIS analyses hampers further

def-inition of the variation’s cause. Although NNIS

ad-justs for exposure to central venous catheters, a

major risk factor for bacteremia, this does not control

for bias of indication. Our results suggest that

com-parative infection rate data, such as those provided

by the NNIS, would be more informative if

severity-of-illness adjustment were performed using SNAP.

Calculation of severity- and birth-weight-adjusted

infection rates would provide neonatologists with

external benchmarks to examine their own practice

style and to focus their quality improvement efforts.

Our finding of a 14-day prolongation of NICU stay

attributable to nosocomial bactenemia is comparable

to that in a previous study from the same NICUs.

Using matched cohort methodology, matching on

birth weight as a proxy for severity of illness,

Free-man et aP2 found a 19.8-day increase in hospital stay

among bactenemic patients hospitalized in 1982. The

similarity of the results from these two studies is

interesting given the former study’s lack of severity

stratification. The concordance is likely to be the

result of two factors. First, the birth-weight-stratified

random selection process used for controls by

Free-man et al should result in bacteremic/nonbactenemic

groups with similar illness severity. Differences in

length of stay or in-hospital mortality between

groups were therefore adjusted for illness severity by

the matching process. Second, the referral patterns of the two NICUs have changed little in the intervening decade, and therefore the population characteristics (eg, admission illness-severity distribution) are likely

to be similar in the two studies. Therefore, although

such techniques are useful in examining the

cone-lates of infection in a single NICU environment, they

are insufficient for analyzing nosocomial infection

rates among units that may serve vastly different

populations and may use differing approaches to

care.

It should be noted that our current results are

limited to the examination of admission-day SNAP

and therapies. As such, we did not consider

inter-ventions such as intralipid administration and

cen-tral line placement, which are known to be potent

modifiers of bacteremia risk.6 Future studies of

inter-ventions administered later in the hospital stay

(6)

coupled with serial assessment of SNAP are likely to

provide useful information regarding the role of

NICU practice styles in the occurrence of nosocomial

bacteremia. The relation between the receipt of

pho-totherapy on admission day and the subsequent

de-velopment of nosocomial bacteremia was

unex-pected. Rather than denoting a causal relation

between the two, this association most likely reflects

that the need (or perceived need) for admission-day

phototherapy serves as a marker for the most

imma-ture infants. As such, this therapy may provide

ad-ditional information beyond that available from birth

weight and SNAP.

In summary, we have shown that SNAP is a

valu-able adjunct to birth weight in the assessment of

nosocomial bacteremia risk among VLBW NICU

pa-tients. This finding parallels our previous analyses of

the relations of SNAP to other NICU outcomes,

in-cluding in-hospital mortality,’3 intraventricular

hem-orrhage,14 and chronic lung disease.15 Together, these

results demonstrate that SNAP is a robust tool for the

examination of NICU care.

ACKNOWLEDGMENTS

This project received support from the Agency for Health Care Policy and Research (ROI HS 06123). We gratefully acknowledge the work of our research assistants Sonya Stevens, Yvonne Reis, and Stephanie Fischer along with the help of Sandy Gabai in manuscript preparation. We thank the members of our consultant panel (William Edwards, MD; Michael Epstein, MD; Alan

Fleis-chman, MD; Rita Gibes-Grossman, RN; Ronald Poland, MD; and

Murray Pollack, MD) for their help in construction ofthe SNAP scale.

REFERENCES

I. Freeman J, Platt R, Epstein MF, Smith NE, Sidebottom DC, Goldmann DA. Birth weight and length of stay as determinants of nosocomial coagulase-negative staphylococcal bacteremia in neonatal intensive care unit populations: potential for confounding. Am IEpidemiol. 1990;132: 1130-1140

2. Gaynes Rl, Martone WJ, Culver DH, et al. Comparison of rates of

nosocomial infection in neonatal intensive care units in the United States. Am IMed. 1991;91:I92-1965

3. Richardson DK, Tarnow-Mordi WO. Measuring illness severity in newborn intensive care. Intensive Care Med. 1994 (in press)

4. Richardson DK, Gray JE, McCormick MC, Workman-Daniels K, Gold-mann DA. Score for Neonatal Acute Physiology (SNAP): validation of a new physiology-based severity of illness index. Pediatrics. 199391: 617-623

5. International Neonatal Network. The CRIB (Clinical Risk Index for

Babies) score: a tool for assessing initial neonatal risk and comparing performance of neonatal intensive care units. Lancet. 1993;342:193-198 6. Gray JE, Richardson DK, McCormick MC, Workman-Daniels K,

Gold-mann D. Neonatal Therapeutic Intervention Scoring System (NTISS): a therapy-based severity of illness assessment tool. Pediatrics. 199290: 561-567

7. Freeman J, Goldmann DA, Smith NE, Sidebottom DG, Epstein MF, Platt R. Association of intravenous lipid emulsion and coagulase-negative staphylococcal bacteremia in neonatal intensive care units. N

Engi JMed. 1990;323:301-308

8. Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: John Wiley & Sons, Inc.; 1989

9. Baumgart 5, Hall SE, Campos JM, Polin RA. Sepsis with coagulase-negative staphylococci in critically ill newborns. Am IDis Child. 1983; 137:461-463

10. Freeman J, Platt R, Sidebottom DG, LeclairjM, Epstein MF, Goldmann DA. Coagulase negative staphylococcal bacteremia in the changing neonatal intensive care unit population. JAMA. 1987;258:2548-2552 11. Pollock E, Ford-Jones EL, Corey M, et al. Use of the Pediatric Risk of

Mortality score to predict nosocomial infection risk in a pediatric

inten-sive care unit. Crit Care Med. 1991;19:160-165

12. Freeman J, Epstein MF, Smith NE, Platt R, Sidebottom DG, Goldmann DA. Extra hospital stay and antibiotic usage with nosocomial coagulase-negative staphylococcal bacteremia in two neonatal intensive care unit populations. Am IDis Child. 1990;144:324-329

13. Richardson DK, Phibbs CS, Gray JE, McCormick MC, Workman-Daniels K, Goldmann DA. Birth weight and illness severity: indepen-dent predictors of neonatal mortality. Pediatrics. 1993;9I :969-975 14. GrayJE, Richardson DK, McCormick MC, Workman K, Goldmann DA.

Score for Neonatal Acute Physiology (SNAP) and risk of intraventric-ular hemorrhage (IVH). Pediatr Res. 1992;31:249A

15. Richardson DK, Gray JE, Goldmann DA, Ungarelli RA, McCormick MC. Admission illness severity predicts VLBW prolonged ventilator dependence. Pediatr Res. 1993;33:273A

SINE QUA NON

[Reports of interventions should providel busy clinical readers with three sorts

of information (complete with their confidence intervals): at least one absolute

measure of efficacy (such as the number of patients who would need to be treated

to prevent an event), the susceptibility of control patients to the target outcome (as

a starting point for extrapolation to their own patients), and (although they could

calculate it from the former two) some relative measure of efficacy (such as the

relative risk reduction).

REFERENCE

1. Sacrett DL, Cook RJ. Understanding clinical trials. Br Med J.1994;309:755-756

Submitted by Student

(7)

1995;95;225

Pediatrics

James E. Gray, Douglas K. Richardson, Marie C. McCormick and Donald A. Goldmann

Infants: Relation to Admission Illness Severity, Resource Use, and Outcome

Coagulase-Negative Staphylococcal Bacteremia Among Very Low Birth Weight

Services

Updated Information &

http://pediatrics.aappublications.org/content/95/2/225

including high resolution figures, can be found at:

Permissions & Licensing

http://www.aappublications.org/site/misc/Permissions.xhtml

entirety can be found online at:

Information about reproducing this article in parts (figures, tables) or in its

Reprints

http://www.aappublications.org/site/misc/reprints.xhtml

Information about ordering reprints can be found online:

(8)

1995;95;225

Pediatrics

James E. Gray, Douglas K. Richardson, Marie C. McCormick and Donald A. Goldmann

Infants: Relation to Admission Illness Severity, Resource Use, and Outcome

Coagulase-Negative Staphylococcal Bacteremia Among Very Low Birth Weight

http://pediatrics.aappublications.org/content/95/2/225

the World Wide Web at:

The online version of this article, along with updated information and services, is located on

been published continuously since 1948. Pediatrics is owned, published, and trademarked by the

Pediatrics is the official journal of the American Academy of Pediatrics. A monthly publication, it has

http://pediatrics.aappublications.org/content/95/2/225

http://www.aappublications.org/site/misc/Permissions.xhtml

http://www.aappublications.org/site/misc/reprints.xhtml

http://pediatrics.aappublications.org/content/95/2/225