The Evidence-based Practice for Improving Quality method has greater impact on improvement of outcomes than dissemination of practice change guidelines and quality improvement training in neonatal intensive care units

Abstract

OBJECTIVES:

To determine whether outcome improvements achieved by neonatal intensive care units (NICUs) in the Evidence-based Practice for Improving Quality (EPIQ) trial could be reproduced in other NICUs by providing quality improvement (QI) training and practice change guidelines developed during the EPIQ trial; and to examine whether the results of the EPIQ trial were sustained.

METHODS:

The present prospective before-after study included 5812 infants born at ≤32 weeks’ gestation and admitted to 19 level 3 NICUs in the Canadian Neonatal Network between October 1, 2005 and December 31, 2007. During a three-month baseline period, multi-disciplinary teams received general training in QI techniques at a two-day workshop, and practice change guidelines targeting nosocomial infection (NI) and bronchopulmonary dysplasia (BPD) developed during the EPIQ trial were provided to all participants. Outcome data collected during the intervention period were compared with data from the baseline period and reported quarterly.

RESULTS:

In NICUs that had not previously participated in the EPIQ trial (non-EPIQ NICUs), there were no significant changes in the incidence trends of NI or BPD. However, within NICUs that had previously participated in the EPIQ trial (EPIQ NICUs) there was a continued reduction in the incidence trend of NI and BPD among EPIQ NICUs randomized during the trial to reduce NI and BPD, respectively.

CONCLUSIONS:

Providing NICUs with QI training and practice change guidelines developed during a successful QI initiative in other units is not effective. The authors speculate that successful practice change involves organizational culture and behaviour change, and should be driven by context-specific evidence.

Quality improvement (QI) methods have been criticized as subjective and difficult to generalize because of institution-specific differences (1,2). QI studies in health care have reported positive changes in patient outcomes (3–5), but also partial success or no changes (6–8).

The Evidence-based Practice for Improving Quality (EPIQ) (9) is a QI method for neonatal intensive care units (NICUs) that tailors evidence-based practice change to each NICU. This includes identifying site-specific associations between practices and outcomes through baseline data analysis; involving NICU teams in systematic evidence reviews; identifying barriers to change; developing evidence-based, site-specific priorities and practice-change strategies; monitoring compliance; and providing tailored support for continuous QI using the Plan-Do-Study-Act model (10). EPIQ was previously tested in a cluster randomized controlled trial in which NICUs randomly assigned to target nosocomial infection (NI) (EPIQ NICUs Infection group) achieved a 32% reduction in NI, and NICUs targeting bronchopulmonary dysplasia (BPD) (EPIQ NICUs Pulmonary group) achieved a 15% reduction in BPD. A comparison group that did not participate showed no improvement for either outcome (9).

The EPIQ trial generated a significant amount of knowledge including lists of practice changes that could be generalized to other NICUs without the need to follow the EPIQ method. It was also unclear whether the outcome improvements achieved would be sustained. The present study was conducted directly after the trial and aimed to determine whether other NICUs (non-EPIQ NICUs) could achieve similar improvements if provided with practice change guidelines from the EPIQ trial and QI training, and whether the improvements achieved by NICUs in the EPIQ trial would be sustained for a further two years.

METHODS

Study design

The present before-after study, conducted immediately after the EPIQ trial concluded in September 2005, included 19 of the 25 tertiary-level NICUs participating in the Canadian Neonatal Network (CNN) at the time, including 11 from the previous EPIQ trial. Study participation was determined by self-selection following circulation of information and an invitation to join. Eligible infants were those born at ≤32 weeks’ gestation and admitted to participating NICUs between October 1, 2005 and December 31, 2007 with the exception of those who were moribund on admission (palliative care planned at birth).

Preparation:

Between October 1 and December 31, 2005 each non-EPIQ NICU established a multidisciplinary QI team (site investigator, neonatologist, nurse manager, nurse educator, QI officer, infection control nurse and others as necessary). Two team members attended a two-day ‘train-the-trainer’ workshop on general QI methods and were provided with the EPIQ trial evidence reviews and practice change guidelines (Figures 1 and 2; Appendix 1 and 2). The existing EPIQ NICU QI teams also attended the workshop, which included information on team building, aims identification, clinical process analysis, Pareto charts, histograms/control charts, benchmarking, practice change using Plan-Do-Study-Act cycles (10), communication strategies and change management. QI teams used a consensus approach to select practice changes from the guidelines to implement based on their perceived needs and the evidence reviews provided to them, but without additional site-specific data on baseline clinical characteristics and outcomes, or organizational culture. Sites were encouraged to focus on reducing NI because it is the most common avoidable neonatal morbidity and can significantly impact outcomes (11), but could also target other morbidities.

Figure 1).

Key driver diagram for reduction of nosocomial infection in the neonatal intensive care unit (NICU). CDC Centers for Disease Control and Prevention (Georgia, USA); IV Intravenous; PICC Peripherally inserted central catheter

Figure 2).

Key driver diagram for reduction of bronchopulmonary dysplasia in the NICU. BW Birth weight; CPAP Continuous positive airway pressure; NICU Neonatal intensive care unit; NIDCAP Newborn Individualized Developmental Care and Assessment Program; pCO₂ Partial pressure of CO₂; wk Weeks

Intervention:

From January 1, 2006 to December 31, 2007, QI teams implemented cycles of single or multiple practice changes (Appendixes 1 and 2). The coordinating centre provided quarterly feedback of outcomes to participating NICUs. Each NICU also reported practice changes implemented to the coordinating centre, but compliance was not monitored and ongoing guidance, troubleshooting or site visits were not provided. Quarterly regional teleconferences (six NICUs per group) provided post hoc opportunities for sharing results and resources.

Outcomes

The primary outcome was incidence of NI, defined as growth of ≥1 organisms in one blood or cerebrospinal fluid culture obtained after 48 h of admission in a symptomatic infant (12). The secondary outcome was BPD, defined as oxygen dependency at 36 weeks’ postmenstrual age or at time of transfer to a level 2 facility (13). Infants who died before 36 weeks’ postmenstrual age were excluded. These outcomes are standardized across all participating sites in the CNN (14).

Data collection, ethics and confidentiality

Data were collected prospectively from each infant admitted to participating NICUs and transmitted electronically to the CNN Coordinating Centre as previously reported (15). Data variables included population characteristics (gestational age, birth weight, Apgar score at 5 min, Score for Neonatal Acute Physiology version II [SNAP-II], maternal hypertension, Caesarean birth, etc), neonatal outcomes (mortality, BPD, NI, intraventricular hemorrhage, retinopathy of prematurity, necrotizing enterocolitis, etc), and clinical practices (steroid use, surfactant use, etc) (14). The study was approved by the institutions’ ethics review boards. Informed consent from patients was waived because there was a lack of risk or possible harm to patients, data were collected anonymously and aggregated, and obtaining consent for participation was impractical.

Data analysis

Descriptive statistics were used to summarize the study population. Infant and hospital characteristics were compared between the EPIQ and non-EPIQ NICUs using the χ² test for categorical variables and Student’s t test for continuous variables. To examine the trend in outcome incidences over time, the data were grouped into time periods starting with the preparation (baseline) period (October 1, 2005 to December 31, 2005) and subsequent consecutive quarters over the intervention period (January 1, 2006 to December 31, 2007). The outcome incidence trends were examined using the Cochran-Armitage trend test or simple linear regression as appropriate. To determine the direction and significance of the trends, multilevel logistic regression models were conducted treating time and admission time from baseline as continuous covariates in the models, adjusted for potential risk factors (eg, gestational age, sex, Apgar score at 5 min, SNAP-II, steroid use, maternal hypertension, caesarean birth, outborn status). The intercepts of the models were treated as random and compound symmetric covariance structures were used to account for the effect of individual hospitals and clustering of infants within each hospital. The multilevel logistic regression analysis was conducted separately for the non-EPIQ NICUs and EPIQ NICUs randomized to reduce NI (infection group) or BPD (pulmonary group). Data were managed and analyzed using SAS version 9.2 (SAS Institute Inc, USA) and R version 2.10.1 (www.r-project.org).

RESULTS

The study population included 5767 eligible infants, 3588 (62.2%) from the EPIQ NICUs. Compared with the EPIQ NICUs, infants in the non-EPIQ NICUs had a lower mean gestational age and birth weight, a higher proportion had a SNAP-II score >20, and there was a higher incidence of antenatal corticosteroid use (Table 1). At baseline, the incidence of BPD was significantly higher in the EPIQ NICUs compared with the non-EPIQ NICUs; the incidence of NI was lower in the EPIQ NICUs, although this was not statistically significant. The characteristics of the EPIQ and non-EPIQ NICUs were not significantly different (Table 1).

TABLE 1.

Characteristics of study population and participating hospitals

	EPIQ NICUs (n=11 hospitals)	Non-EPIQ NICUs (n=8 hospitals)	P
Infant characteristics
Over entire study period
n	3588	2179
Male sex, %	53.7	54.6	0.47
Gestational age, weeks, mean ± SD	29.0±2.5	28.7±2.5	<0.01
Birth weight, g, mean ± SD	1320±455	1263±445	<0.01
Small for gestational age, %	10.5	10.0	0.55
Apgar score <7 at 5 min, %	21.9	21.7	0.91
SNAP-II score >20, %	15.7	21.4	<0.01
Antenatal steroids, %	78.2	83.6	<0.01
Caesarean section, %	59.7	59.6	0.92
Maternal hypertension, %	17.9	18.7	0.42
During baseline period
n	401	201
Rate of nosocomial infection, %	13.7	18.6	0.11
Rate of bronchopulmonary dysplasia, %	26.0	18.3	0.04
Hospital characteristics
Beds, median (range)	19 (3–45)	15 (6–42)	0.77
Neonatologists, median (range)	7 (3–10)	7 (3–10)
Admissions ≤32 weeks during the study period, median (range)	283 (143–807)	204 (49–715)	0.22
Percent outborn, median (range)	15.2 (7.8–1.0)	13.0 (8.2–20.8)	0.57

EPIQ Evidence-based Practice for Improving Quality; NICU Neonatal intensive care unit; SNAP-II Score for Neonatal Acute Physiology, version II

During the present study there was no change in the incidence trend of NI among the non-EPIQ NICUs (Table 2). However, in the EPIQ NICUs Infection group, there was a significant downward incidence trend for NI, with a 25% decrease in NI overall. There was no change in NI among the EPIQ NICUs Pulmonary group. There was also no change in the incidence trend of BPD among the non-EPIQ NICUs. However, there was a significant downward incidence trend for BPD in the EPIQ NICUs Pulmonary group, but not the Infection group. Further analysis of the EPIQ NICUs using data spanning both the EPIQ trial and the present study (Table 2; trend analysis results not shown) indicated that the incidence trend for NI over the entire period continued in a significant downward direction in the Infection group (slope [βi] −0.03 [95% CI−0.046 to −0.017]), as did the trend for BPD in the Pulmonary group (slope [βi] −0.02 [95% CI −0.04 to −0.005]).

TABLE 2.

Incidence of nosocomial infection (NI) and bronchopulmonary dysplasia (BPD) among neonatal intensive care units (NICUs) in the group that only participated in the current study (non-Evidence-based Practice for Improving Quality [EPIQ] NICUs) and the two groups that also participated in the previous EPIQ trial (EPIQ NICUs Infection and Pulmonary groups)

Outcome	Group	EPIQ trial			Current study			P*†	Adjusted*‡ slope for current study (95% CI)
		Baseline	Year I	Year II	Baseline*	Year I*	Year II*
NI	Non-EPIQ NICUs	N/A	N/A	N/A	18.6 (39/210)	22.2 (66/297)	14.0 (47/337)	0.1	−0.024 (−0.07 to 0.021)
	EPIQ NICUs Infection group	25.4 (195/769)	21.4 (46/215)	17.4 (41/236)	20.8 (32/154)	18.5 (40/216)	13.1 (25/191)	0.05	−0.078 (−0.14 to −0.022)
	EPIQ NICUs Pulmonary group	16.0 (147/918)	10.5 (29/276)	8.8 (25/283)	9.3 (23/247)	15.3 (50/326)	14.6 (46/315)	0.08	−0.03 (−0.08 to 0.02)
BPD	Non-EPIQ NICUs	N/A	N/A	N/A	18.3 (35/191)	22.9 (62/271)	18.7 (58/310)	0.91	0.02 (−0.03 to 0.07)
	EPIQ NICUs Infection group	31.8 (223/702)	31.5 (63/200)	30.6 (67/219)	19.6 (27/138)	19.3 (40/207)	22.5 (38/169)	0.51	0.02 (−0.04 to 0.08)
	EPIQ NICUs Pulmonary group	29.4 (249/848)	25.4 (64/252)	24.9 (65/261)	29.9 (68/227)	27.8 (86/309)	22.7 (66/291)	0.058	−0.05 (−0.10 to −0.01)

Data presented as % (n/total n) unless otherwise indicated.

^*P values and adjusted slope are calculated for data collected during the current study only;

^†P values for comparison in the current study were based on the Cochran-Armitage trend test;

^‡The slopes and coefficients of time period were estimated based on a multilevel logistic regression model for the current study with random intercept to account for the hospital effect. The adjusted covariates (potential confounder or risk factors) included gestational age, small for gestational age, Score for Neonatal Acute Physiology, version II (SNAP-II), sex, Apgar score at 5 min, caesarean birth, steroid use and maternal hypertension. EPIQ trial data from Lee et al (9) shown in the EPIQ baseline, year I and year II columns used with permission. BPD Bronchopulmonary dysplasia; NI Nosocomial infection; NICU Neonatal intensive care unit; SNAP-II Score for Neonatal Acute Physiology, version II

DISCUSSION

Our results suggest that the provision of practice change guidelines along with general QI training and quarterly feedback to the non-EPIQ sites had no effect on the incidence of NI and BPD. In contrast, in the EPIQ sites there were continued improvements in the outcomes already targeted.

Our results are consistent with systematic reviews of evaluations of the dissemination of published clinical practice guidelines, which have reported improvements in care processes but very few significant improvements in outcomes, and have concluded that there is little evidence of effect (16,17). Our intervention, however, also included QI training and feedback on outcomes, activities that should provide tools and information to drive change. This approach is in contrast to the current literature, where combined education and dissemination approaches more often include guideline-specific training. However, studies show that even this approach is often ineffectual at significantly improving targeted health care outcomes (18,19), with one systematic review reporting a median improvement in performance of only 6.0% (20).

Review of the practice changes implemented by the non-EPIQ sites indicates that only a few interventions were applied. This is likely the reason for the observed results and raises the question of why this occurred. Our speculation is that the EPIQ method results in a sustained cultural change toward effective QI but provision of general QI training with practice change guidelines is not sufficient to achieve the same motivation to improve outcomes. Explanations for why practice change guidelines have limited effectiveness include barriers to physician adherence to guidelines such as lack of awareness, familiarity or agreement with the guidelines, lack of self-efficacy and outcome expectation, inertia of previous practice, and external barriers that may be patient or environment related (21). To address these issues, Hayward (22) suggested that strategies for implementing practice change guidelines should be clear and specific, and should match the goals of both targeted patients and practitioners. Kitson et al (23) broadened this approach and proposed that successful implementation of research into practice should account for the interplay between the level and nature of the evidence, the context or environment where the research will be used and the method for facilitating the process.

EPIQ is based on the model described by Kitson et al (23), and uses a mixed-methods approach to combine evidence with contextual knowledge to change organizational culture and individual behaviour, so that barriers to change can be overcome and practice change sustained (9). As such, there are several differences between EPIQ and the approach used in the present study. Here, a national research team was not formed to review baseline data, share best practices and provide intersite leadership. In addition, site-specific organizational culture and behaviour change was not addressed in the non-EPIQ NICUs, nor did they conduct and share systematic literature reviews, or use baseline data to identify site-specific practices to target. Instead, each NICU team selected practice changes based on their personal understanding of their site and the evidence reviews and lists of practice changes provided to them. Our primary aim in the present study was to determine whether information from the EPIQ trial could be generalized to other sites without following the entire EPIQ method. The lack of outcome improvement in the present study suggests that the EPIQ process itself is important rather than the information generated per se.

Engaging NICU teams in evidence generation, addressing organizational culture and local context, designing site-specific evidence-based strategies, and providing active support and learning are likely to be the most important facets of EPIQ. By engaging practitioners and encouraging them to take ownership of their site’s problems and solutions, EPIQ facilitates acceptance and maximizes impact, whereas only providing general QI training and lists of practice changes does not. This reinforces the need for health care QI initiatives to move away from a ‘one size fits all’ approach and account for the local environment using an evidence-driven method. Indeed, recent work in health care QI emphasizes the effect of local context, including leadership, organizational culture, resources and motivation to change, on the success of QI initiatives (24–26).

An issue that may also have contributed to the study outcome is the extensive nature of the guidelines and the variable strength of the supporting evidence. Each of the two lists includes >30 practice changes compiled by expert consensus based on the strength of the evidence and whether the practices were implemented during the EPIQ trial. The NI list includes four category IA, nine IB and 21 II recommendations, and the BPD list includes two category IA, six IB and 31 II recommendations (post hoc categorization [27]), which cover multiple approaches to improving outcomes (Figures 1 and 2, Appendixes 1 and 2). Selecting practice changes without understanding the local context may have resulted in implementation of inappropriate practices, despite the strength of the evidence for those practices. For example, NI in a particular site may have been related to high levels of infection from staff-to-patient contact and from the NICU environment, which would have been detected during baseline analysis using EPIQ. In the present study, the site may have decided to designate and certify staff members that handle central lines. They had already adopted alcohol-based waterless cleaner in the unit but were unaware that compliance was low. Despite reviewing the evidence and making an ‘informed’ choice, there would have been little or no change in the incidence of NI.

Limitations to the present study include the absence of a control group for which no QI attempts were made. However, it is difficult to prevent cross-contamination of sites in the same network and to recruit NICUs into such a group because sites are unwilling to do nothing to improve care. Additionally, the study results may be biased because of initial differences in infant and NICU characteristics between the EPIQ and non-EPIQ groups. For example, the incidence of BPD was significantly higher in the EPIQ NICUs at study commencement, while the use of antenatal steroids was higher in the non-EPIQ NICUs, which also had more immature and smaller infants. However, no non-EPIQ NICUs implemented an antenatal steroid treatment protocol during the study, suggesting that the higher antenatal steroid use is indicative of pre-existing practices, which is likely to have resulted in the consistently lower incidence of BPD. The non-EPIQ sites did, however, attempt to improve the incidence of BPD by implementing practice changes in other domains.

By the end of the study, the incidences of both BPD and NI were comparable in all NICU groups. It is possible that these incidences represent the level achievable using EPIQ. However, the reported incidences are mean measures and some NICUs achieved lower incidences of NI (lowest: 4.2%) and BPD (lowest: 8.3%). The incidences of NI and BPD, while slightly lower than those reported in some other American NICU networks (Vermont Oxford Network: NI, 18.4%; BPD, 27.7% [28] and National Institute of Child Health and Human Development Neonatal Research Network: NI, 36%; BPD, 42% [29]), are also not as low as those achieved in Japan (NI: 6.2%; BPD: 10.6% [30]) and Australia-New Zealand (NI: 15.1%, BPD: 16.6% [31]) over the same time period, suggesting there was room for further improvement.

CONCLUSIONS

Providing NICUs with written practice change guidelines and general QI training is not effective in decreasing NI. We speculate that successful practice change involves organizational culture and behaviour change, and should be driven by context-specific evidence, and peer-to-peer support and leadership.

APPENDIX 1. List of practice changes to address nosocomial infection (NI) provided during the study with indication of selection for implementation by participating sites

Practice changes with category of recommendation*		Sites
		NI					BPD						Non-EPIQ
		A	B	C	D	E	G	H	I	J	K	L	M	N	O	P	Q	R	S	T
1. Hand hygiene
Communicate handwashing protocols to staff	II	E	E	A	A	A	–	–	–	I	A	–	–	I	A	A	–	I	–	A
Display handwashing protocol posters at wash basins	IB	E	E	A	A	A	–	–	–	I	A	–	–	I	A	A	–	I	–	A
Handwashing coach for visitors	II	–	A	A	A	–	–	–	–	–	–	–	–	–	–	–	–	–	–	A
No wearing of hand jewellery	IB	E	E	A	A	A	–	–	–	A	A	–	–	A	A	A	–	I	–	A
Adoption of alcohol-based waterless cleanser	IA	E	E	A	A	E	–	I	–	I	–	–	–	I	I	A	–	I	–	A
Strategic placement of cleanser dispensers in NICU and at entrance	IA	E	E	A	A	E	–	I	–	–	–	–	–	I	I	A	–	I	–	A
Issue NICU staff with personal bottles of waterless handrub	IA	–	–	E	E	E	–	–	–	–	–	–	–	–	A	–	–	–	–	–
Clearly mark designated patient area on floor around patient and treat as individual room with regard to standard practices and precautions	II	–	A	A	A	E	–	–	–	–	A	–	–	A	–	A	–	–	–	–
Adopt strict sterility protocols within designated sterile areas	II	–	–	A	A	E	–	–	–	A	–	–	–	A	A	A	–	–	–	–
No movement of charts, pens and objects into/out of sterile areas	II	–	A	A	A	E	–	–	–	A	–	–	–	I	A	–	–	–	–	–
2. Gown/gloves
Discontinue routine use of gowns and gloves in the NICU	IB	E	A	A	A	A	–	A	–	–	–	–	–	A	–	A	–	–	–	–
Adopt routine use of gowns/gloves only for isolation and outbreaks	IB	E	A	A	A	A	–	–	–	–	–	–	–	A	A	–	–	–	–	A
3. Skin care
Cleanse umbilicus with antiseptic prior to line insertion	IB	A	A	A	A	A	–	I	–	A	–	–	–	A	A	–	–	–	–	A
Bathe VLBW infants only using warm sterile water, no soap	II	–	–	E	E	A	–	–	–	–	–	–	–	–	A	A	–	–	–	–
Use 2% aqueous chlorhexidine for skin antisepsis	IA/II†	E	–	E	E	E	–	–	–	I	–	–	–	I	A	A	–	–	–	A
4. Skin breaks
Routine daily skin break chart and audit for all infants	II	E	E	E	E	E	–	–	–	I	–	–	–	A	–	I	–	–	–	–
Adopt peripheral IV initiation policy	II	I	A	E	E	–	–	A	–	I	–	–	–	A	A	A	–	–	–	A
Restrict number of skin breaks per patient	II	E	E	E	E	E	–	–	–	–	–	–	–	–	–	A	–	–	–	–
Restrict procedure skin breaks to two attempts/staff person	II	E	E	E	E	–	–	A	–	–	–	–	–	A	A	A	–	–	–	–
Certify, designate and restrict staff allowed to make skin breaks	II	E	E	E	E	–	–	–	–	–	–	–	–	A	A	–	–	–	–	A
Implement algorithms for blood sampling and IV starts	II	–	E	E	E	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Restrict sampling for routine blood tests	II	–	–	E	E	E	–	–	–	–	–	–	–	–	A	–	–	A	–	–
Pool blood testing to reduce number of samplings needed	II	A	–	E	E	E	–	A	–	–	–	–	–	A	A	–	–	A	–	A
5. Central line protocols
Adopt CDC guidelines for central lines	IB	A	–	E	E	E	–	–	–	–	–	–	–	–	A	A	–	–	–	–
Certify, designate and restrict people allowed to handle central lines	IA	A	E	A	A	E	–	–	–	–	–	–	–	I	A	–	–	I	–	A
Treat percutaneously inserted central catheters (PICC) as central lines	II	A	A	A	A	A	–	–	–	I	–	–	–	A	A	A	–	A	–	–
Change PICC line dressings only as needed instead of routinely	IB	A	E	A	A	A	–	–	–	I	–	–	–	A	A	A	–	–	–	–
6. Antibiotic use
Adopt ampicillin and gentamicin as first-line antibiotics during the early neonatal period (restrict empirical use of vancomycin)	II	A	A	A	A	A	–	A	–	A	A	–	–	A	A	A	–	A	–	A
Adopt cloxacillin and gentamicin as first line antibiotics for treatment of suspected nosocomial sepsis unless there is evidence of septic shock, meningitis or necrotizing enterocolitis	II	E	–	E	E	E	–	–	–	–	–	–	–	I	–	–	–	–	–	–
Adopt algorithm for antibiotic choice	II	E	A	–	A	E	–	–	–	–	–	–	–	–	–	–	–	–	–	–
To rule out sepsis, discontinue antibiotics if 36 h to 48 h cultures negative	IB	E	A	E	E	E	–	–	–	–	–	–	–	–	–	A	–	A	–	–
7. Ventilator circuits
Change ventilator circuits when visibly contaminated, malfunctioning and between patients	II	A	E	A	A	A	–	–	–	A	–	–	–	A	A	A	–	A	–	A
8. Environment
Restrict visitors to two per infant	II	A	A	A	A	A	–	A	–	A	–	–	–	–	A	–	–	A	–	A
Encourage influenza immunization among NICU staff	IB	A	E	A	A	A	–	–	–	A	–	–	–	I	A	A	–	A	–	–

^*Category of recommendation applied post hoc for guidance and adopted from the Centers for Disease Control Healthcare Infection Control Practices Advisory Committee (27);

^†Recommendation is category IA in adults and category II in infants <2 years of age. Table adapted with permission from Appendix 3 in Lee et al (9). – Intervention not introduced; A Intervention already in place; BPD Bronchopulmonary dysplasia; E Intervention introduced during Evidence-based Practice for Improving Quality (EPIQ) trial; I Intervention introduced during current study; NICU Neonatal intensive care unit

APPENDIX 2. List of practice changes to address bronchopulmonary dysplasia (BPD) provided during the study with indication of selection for implementation by participating sites

Practice changes with category of recommendation*		Sites
		NI					BPD						Non-EPIQ
		A	B	C	D	E	G	H	I	J	K	L	M	N	O	P	Q	R	S	T
1. Antenatal management
Implement antenatal steroid treatment protocol	IA	A	A	A	–	A	A	A	A	E	A	A	–	A	A	A	–	A	–	–
Treat ureaplasma infections in pregnant women with erythromycin	II	–	–	A	–	A	–	A	–	–	–	E	–	–	–	–	–	A	–	–
Antibiotics to women with prelabour rupture of the membranes preterm	II	A	A	A	–	A	–	A	E	A	E	E	–	A	–	A	–	A	–	–
2. Delivery room and resuscitation practices
Team of neonatologist (or fellow), nurse and respiratory therapist to attend high-risk deliveries	II	A	A	A	–	A	A	A	A	A	A	A	–	–	A	A	–	–	–	–
Prophylactic surfactant treatment within 30 min of birth for infants <28 weeks’ gestation or <1250 g birth weight	IA	–	I	I	–	I	E	E	I	E	A	E	–	I	–	I	I	–	I	–
Restrict manual ventilation	II	I	I	I	–	I	E	E	–	E	A	E	–	–	A	A	I	–	–	I
Limit tidal volume used in manual ventilation	II	–	–	I	–	–	E	–	–	A	–	E	–	–	–	A	–	–	–	–
Use of Laerdal bags with positive end-expiratory pressure	II	–	–	–	A	A	–	–	–	–	–	E	–	–	–	A	–	–	–	–
Restrict exposure to oxygen by using blended gases	IB	A	I	A	A	A	–	E	–	A	A	E	–	–	A	A	–	–	–	I
3. Continuous positive airway pressure (CPAP)
Adopt algorithm for CPAP use	II	–	–	I	–	–	–	E	–	E	–	–	–	–	–	–	–	–	–	–
Early use of CPAP among infants needing supplemental oxygen	IB	–	–	A	–	I	–	–	E	A	A	E	–	–	–	A	–	–	–	A
Emphasize liberal use of CPAP instead of mechanical ventilation	II	–	–	I	–	A	–	E	E	A	–	–	–	–	–	–	–	A	–	A
Employ long-term CPAP treatment strategy	II	–	–	–	–	A	–	–	–	A	–	–	–	–	–	A	–	–	–	–
Avoid reintubation if on treatment with CPAP	II	–	–	A	–	–	–	–	–	E	–	–	–	–	–	A	–	–	–	A
4. Mechanical ventilation
Adopt algorithm for mechanical ventilation	II	–	–	I	–	–	–	E	–	E	–	–	–	–	–	A	–	–	–	–
Early use of high frequency ventilation when needed	II	–	–	I	–	A	–	–	E	A	–	–	–	–	–	–	–	–	–	–
Limit tidal volumes used	II	–	–	A	A	A	–	I	E	A	–	–	–	A	–	A	–	–	–	–
Adopt trigger ventilation and volume guarantee modes	II	–	–	A	–	–	–	–	E	A	–	A	–	A	A	A	–	–	–	–
Aggressive weaning of mechanical ventilation	II	–	–	A	–	I	–	I	–	A	–	–	–	–	–	A	–	–	–	A
Early extubation to CPAP	IB	–	–	–	–	I	–	A	–	A	–	–	–	–	–	A	–	–	–	A
5. Blood gases
Adopt algorithm for blood gas monitoring	II	–	–	–	–	–	–	E	–	–	–	–	–	–	–	A	–	–	–	–
Minimize blood sampling for gases	II	A	–	A	–	A	–	–	–	A	–	–	–	A	–	–	–	A	–	–
Adopt normocarbia policy (keep pCO2 between 40 mmHg and 55 mmHg)	IB	A	–	A	–	–	E	I	E	E	–	–	–	A	A	–	–	A	–	–
6. Oxygen and oxygen saturation
Treat oxygen as a drug	II	–	–	A	I	A	–	–	–	A	A	–	–	–	A	A	–	–	–	–
Physician orders required for use of oxygen	II	–	–	–	–	–	–	–	–	A	–	–	–	–	A	–	–	–	–	–
Goal to keep oxygen saturation between 88% and 92%	IB	–	I	I	A	A	E	A	E	E	E	E	–	I	A	I	–	–	–	–
Set alarm limits between 85% and 95%	II	–	I	A	–	A	E	E	–	A	E	E	–	–	A	A	–	–	–	–
7. Fluids and nutrition
Adopt high humidity environment	IB	A	A	A	A	A	–	A	–	A	–	–	–	A	A	A	–	–	–	A
Restrict fluid intake	II	–	–	–	A	A	–	A	–	A	–	E	–	–	–	A	–	A	–	A
Avoid volume expansion to treat hypotension	II	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	I	–	–	–
Early parenteral protein and lipid nutrition within 4 h after birth	II	A	I	I	–	I	–	E	E	I	E	E	–	I	I	A	–	–	–	I
Maximize caloric intake	II	A	–	–	A	A	–	A	E	A	–	–	–	A	I	A	–	A	–	A
Early enteral nutrition	II	–	–	A	A	I	–	E	–	A	A	E	–	I	I	A	–	A	–	–
Promote use of breast milk	II	–	I	A	A	I	–	A	A	A	–	A	–	–	I	A	–	A	–	A
8. Environment
Adopt pain management protocol	II	–	A	A	A	A	–	A	–	A	I	–	–	A	–	A	–	I	–	–
Minimize noise in NICU	II	–	A	A	A	–	–	–	A	A	E	–	–	A	–	A	–	–	–	A
Adopt NIDCAP protocol	II	A	A	A	A	I	–	–	–	–	–	–	–	A	–	–	–	–	–	–
Employ massage therapy	II	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–	–
Provide comment cards to parents and pay attention to feedback	II	–	A	A	–	–	–	–	–	E	–	–	–	A	–	–	–	–	–	A

^*Category of recommendation applied post-hoc for guidance and adopted from the Centers for Disease Control Healthcare Infection Control Practices Advisory Committee (27). Table adapted with permission from Appendix 4 in Lee et al (9). – Intervention not introduced; A Intervention already in place; BPD Bronchopulmonary dysplasia; CPAP Continuous positive airway pressure; E Intervention introduced during Evidence-based Practice for Improving Quality (EPIQ) trial; I Intervention introduced during current study; NI Nosocomial infection; NICU Neonatal intensive care unit; NIDCAP Newborn Individualized Developmental Care and Assessment Program; pCO₂ Partial pressure of CO₂