The n e w e ng l a n d j o u r na l of
Original Article
From the Department of Obstetrics and
Gynecology, Centre Hospitalier Universi-
taire (CHU) de Sherbrooke, Sherbrooke,
QC (N.C., J.-C.P., E.D., W.D.F.), Depart-
ment of Epidemiology and Biostatistics,
McGill University (M.A., R.B.), Depart-
ment of Obstetrics and Gynecology, Uni-
versity of Montreal, Centre Hospitalier
Universitaire Sainte-Justine (F.A.), De-
partment of Obstetrics and Gynecology,
McGill University, Royal Victoria Hospi-
tal (P.M.), and Department of Obstetrics
and Gynecology, McGill University, Jew-
ish Hospital (H.A.A.), Montreal, and the
Population Health and Optimal Health
Practices Research Unit, CHU de Qubec
Research Centre, Quebec, QC (M.D.)
all in Canada; and the Research Institute
for Development, Universit Paris Des-
cartes, Sorbonne Paris Cit, UMR 216,
Paris (A.D.). Address reprint requests to
Dr. Chaillet at the Departments of Ob-
stetrics and Gynecology and Family Med-
icine, University of Sherbrooke, Faculty
of Medicine and Health Sciences CHUS,
3001, 12e Ave. Nord, Centre de Recher-
che Clinique, Local 2921, Sherbrooke, QC
J1H 5N4, Canada, or at n ils
.
chaillet@
usherbrooke.ca.
*A complete list of the members of the
Quality of Care, Obstetrics Risk Manage-
ment, and Mode of Delivery (QUARISMA)
trial research group is provided in the
Supplementary Appendix, available at
NEJM.org.
N Engl J Med 2015;372:1710-21.
DOI: 10.1056/NEJMoa1407120
Copyright 2015 Massachusetts Medical Society.
1710
Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
m e dic i n e
A Cluster-Randomized Trial to Reduce
Cesarean Delivery Rates in Quebec
Nils Chaillet, Ph.D., Alexandre Dumont, M.D., Ph.D., Michal Abrahamowicz, Ph.D.,
JeanCharles Pasquier, M.D., Ph.D., Francois Audibert, M.D.,
Patricia Monnier, M.D., Ph.D., HaimA. Abenhaim, M.D., M.P.H., Eric Dub, M.Sc.,
Marylne Dugas, Ph.D., Rebecca Burne, M.Sc., and WilliamD. Fraser, M.D.,
for the QUARISMA Trial Research Group*
A BS T R AC T
BACKGROUND
In Canada, cesarean delivery rates have increased substantially over the past decade.
Effective, safe strategies are needed to reduce these rates.
METHODS
We conducted a cluster-randomized, controlled trial of a multifaceted 1.5-year inter-
vention at 32 hospitals in Quebec. The intervention involved audits of indications
for cesarean delivery, provision of feedback to health professionals, and implemen-
tation of best practices. The primary outcome was the cesarean delivery rate in the
1-year postintervention period.
RESULTS
Among the 184,952 participants, 53,086 women delivered in the year before the
intervention and 52,265 women delivered in the year following the intervention.
There was a significant but small reduction in the rate of cesarean delivery from
the preintervention period to the postintervention period in the intervention group
as compared with the control group (change, 22.5% to 21.8% in the intervention
group and 23.2% to 23.5% in the control group; odds ratio for incremental change
over time, adjusted for hospital and patient characteristics, 0.90; 95% confidence
interval [CI], 0.80 to 0.99; P=0.04; adjusted risk difference, 1.8%; 95% CI, 3.8 to
0.2). The cesarean delivery rate was significantly reduced among women with
low-risk pregnancies (adjusted risk difference, 1.7%; 95% CI, 3.0 to 0.3; P=0.03)
but not among those with high-risk pregnancies (P=0.35; P = 0.03 for interaction).
The intervention group also had a reduction in major neonatal morbidity as com-
pared with the control group (adjusted risk difference, 0.7%; 95% CI, 1.3 to 0.1;
P=0.03) and a smaller increase in minor neonatal morbidity (adjusted risk differ-
ence, 1.7%; 95% CI, 2.6 to 0.9; P<0.001). Changes in minor and major maternal
morbidity did not differ significantly between the groups.
CONCLUSIONS
Audits of indications for cesarean delivery, feedback for health professionals, and
implementation of best practices, as compared with usual care, resulted in a signifi-
cant but small reduction in the rate of cesarean delivery, without adverse effects
on maternal or neonatal outcomes. The benefit was driven by the effect of the
intervention in low-risk pregnancies. (Funded by the Canadian Institutes of Health
Research; QUARISMA Current Controlled Trials number, ISRCTN95086407.)
n engl j med 372;18nejm.org April 30, 2015
The New England Journal of Medicine
Copyright 2015 Massachusetts Medical Society. All rights reserved.
 A Trial to Reduce Cesarean Delivery R ates
R
ates of cesarean delivery are high
in developed countries.1-3 In Canada, these
rates increased from 21.2% to 28.0% be-
tween 2000 and 2008 and remained stable until
2011.4-6 High rates of cesarean delivery are of
substantial concern owing to the potential harm
to the mother and her baby associated with a
medically unnecessary cesarean delivery and to
the related costs of health care.7-15 Providing ev-
idence-based guidance to health professionals
regarding the appropriate selection of women
who could benefit from cesarean delivery is now
a priority.
Systematic reviews of strategies designed to
reduce cesarean delivery rates and to improve the
quality of intrapartum care suggest that the dis-
semination of clinical practice guidelines alone
is generally ineffective; that educational strategies,
opinion leaders, quality- improvement strategies,
and academic detailing (university or noncommer-
cial educational outreach) have mixed effects; that
audits of indications for cesarean delivery asso-
ciated with the provision of feedback and remind-
ers about clinical practices are generally effective;
and that multifaceted strategies may be more
effective than individual strategies.16-21 However,
data from large randomized trials have not been
available to assess the effects of a multifaceted
strategy involving audits and feedback on cesar-
ean delivery rates and measures of quality of care.
We designed this trial to assess whether a
multifaceted intervention to promote professional
onsite training with audits and feedback would
reduce the rate of cesarean delivery. Secondary
objectives included assessment of the effects of
this strategy on other obstetrical interventions
and on maternal and neonatal outcomes.
Me thods
Hospitals and Participants
We conducted the QUARISMA (Quality of Care,
Obstetrics Risk Management, and Mode of De-
livery) trial at 32 public hospitals in Quebec prov-
ince from April 1, 2008, to October 31, 2011. To
be eligible to join the trial, hospitals were required
to have at least 300 deliveries in the year before
initiation of the study, a rate of cesarean delivery
higher than 17% (i.e., in the 10th percentile of the
distribution of cesarean deliveries in Quebec from
April 2006 to March 2007),22 and at the time of
recruitment, no recent or ongoing quality-improve-
n engl j med 372;18nejm.org April 30, 2015
The New England Journal of Medicine
Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.
ment programs specifically designed to reduce
the rate of cesarean delivery. All women who de-
livered at participating centers and whose new-
borns had a gestational age of at least 24 weeks
and weighed at least 500 g at delivery were in-
cluded in the analysis.
Study Design and Oversight
The QUARISMA trial was a stratified, cluster-
randomized, parallel-group trial in which hospi-
tals were the units of randomization and women
were the units of analysis. By designating hospi-
tals as the units of randomization, we ensured that
all women within a given maternity unit were as-
signed to the same trial group, thereby reducing
the risk of contamination of the intervention ef-
fect. Randomization was stratified according to
level of care (community, regional, or tertiary
hospital). The study included a 1-year preinter-
vention (baseline) period, a 1.5-year intervention
period, and a 1-year postintervention period. Af-
ter the baseline period, hospitals were randomly
assigned to the intervention group or the control
group. To avoid imbalance in the size of the two
groups, we used computer-generated, blocked ran-
domization within each stratum, with blocks
consisting of four centers or, for strata with fewer
than eight hospitals, two centers. Local investi-
gators at each hospital were then immediately in-
formed of the assignment status of their hospital.
In-hospital data were abstracted by trained
research nurses or medical archivists from the
medical records of mothers and newborns
3 months after delivery; data were abstracted in
the same way at both intervention and control
hospitals. Data completeness and quality were
assessed every 3 months through onsite visits
and queries sent to onsite data collectors to re-
solve discrepancies identified by the data-man-
agement team. Data collectors were aware of the
randomization assignments but were not in-
volved in outcomes assessment. Until the end of
the trial, access to the database was restricted to
the data manager. All steps involved in the man-
agement of clinical data were monitored annu-
ally and validated by an independent data and
safety monitoring board.
The trial received approval from the institu-
tional review board at each participating hospital.
The first author assumes responsibility for the
completeness and integrity of the data and the
fidelity of the report to the study protocol, which
1711
 The n e w e ng l a n d j o u r na l of m e dic i n e

is available with the full text of this article at mendations, and to provide feedback and en-
NEJM.org. sure the implementation of the recommendations
(through regular staff meetings, training ses-
Intervention sions, and informal discussions) was approxi-
The QUARISMA program, which was imple- mately 2 days per 3-month cycle.
mented at the hospital level in the intervention No intervention from the QUARISMA team
group, targeted physicians and nurses involved was planned for the control group. In order to
in the decision-making process for cesarean de- assess contamination bias, quality-improvement
liveries. The program consisted of initial onsite programs were reviewed annually in control
training in evidence-based clinical practices by hospitals.
instructors from the Society of Obstetricians and
Gynecologists of Canada, clinical audits, and im- Outcomes
plementation of best practices (for details see The primary outcome was the overall rate of
Section 2 in the Supplementary Appendix, avail- cesarean delivery. Secondary outcomes included
able at NEJM.org). A local opinion leader acted rates of planned and intrapartum cesarean deliv-
as the facilitator at each site. No financial incen- ery, vaginal delivery with the use of instruments
tive was provided. (i.e., forceps or vacuum), pharmacologic induction
The first 6 months of the 1.5-year interven- of labor, artificial rupture of membranes, aug-
tion period focused on identifying the opinion mentation with oxytocin during labor, epidural
leader in each intervention hospital (with the use analgesia, and episiotomy; composite risks of mi-
of surveys) and selecting the local audit commit- nor and major maternal complications; and com-
tee (which consisted of one or two obstetrician posite risks of minor and major neonatal complica-
gynecologists, one or two general practitioners, tions, excluding lethal congenital abnormalities.
and one nurse), developing local expertise in Composite morbidity outcomes were prespeci-
conducting audits and providing feedback (1-day fied on the basis of literature reviews and the
training),23 and improving the performance of consensus of experts from the QUARISMA re-
health professionals in monitoring indications search team (for details see Section 3 in the
for cesarean delivery and managing intrapartum Supplementary Appendix).24-27
care (1-day training). During the year after the
training period, four 3-month audit cycles were Statistical Analysis
implemented by audit committees, with the sup- The sample size was calculated to maximize
port of external facilitators who made quarterly statistical power while minimizing the number
educational outreach visits. Each cycle included of clusters.28 To account for clustering by hospi-
five standardized steps: the identification of wom- tal, we assumed an intraclass correlation coef-
en who had cesarean deliveries during the first ficient of 0.0065, estimated on the basis of the
month of each cycle; the collection of data, with number of deliveries in Quebec in the year
the use of standardized forms, regarding the man- 20062007.22 We calculated that we would have
agement of labor and delivery; the assessment by to enroll 32 hospitals, with a total of 34,848
the local audit committee, with the use of clinical expected deliveries per year, for the study to have
algorithms, of the relevance of the indications for 90% power to detect a 20% relative reduction
cesarean delivery; the formulation of recommen- with the intervention in the rate of cesarean
dations for best practices and the evaluation of deliveries, assuming a baseline rate of 23.5%, at
previous recommendations, both performed by a two-sided alpha significance level of 0.05.29
the committee; and the provision of informal and In the primary intention-to-treat analyses, we
formal feedback to health professionals. During assessed the effect of the intervention on the
the 1-year postintervention period, health profes- rate of cesarean delivery using the multivariable
sionals in the intervention group were encour- generalized-estimating-equations extension of lo-
aged to continue performing clinical audits, but gistic regression, with an exchangeable covari-
without supervision, in order to assess the pro- ance matrix, to account for the clustering of
grams sustainability. The mean time required by women within hospitals.30 Changes in the risk of
the audit committee members to conduct each cesarean delivery in the two study groups be-
audit session, to formulate and produce recom- tween the 1-year baseline (preintervention) peri-

1712 n engl j med 372;18nejm.org April 30, 2015

The New England Journal of Medicine

Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.
 A Trial to Reduce Cesarean Delivery R ates

od and the 1-year postintervention period were
compared with the use of an adjusted odds ratio
(with 95% confidence intervals) for the interac-
tion between group (intervention vs. control)
and time period (postintervention vs. baseline).28
The adjusted odds ratio for interaction was esti-
mated with the use of data on women who deliv-
ered during the baseline period or the postinter-
vention period and measured the intervention effect
with the difference-in-differences approach,31,32
which was adapted for generalized-estimating-
equations analyses of clustered binary outcomes
(see Section 6 in the Supplementary Appendix).28
Two-tailed P values of less than 0.05 were con-
sidered to indicate statistical significance.
All primary analyses included adjustments for
prespecified potential risk factors associated with
cesarean delivery according to individual hospital
and patient. These included adjustments for hos-
pital academic status and level of care; maternal
age, parity, previous cesarean delivery, any patho-
logic condition during pregnancy, smoking dur-
ing pregnancy, and the use of assisted reproduc-
tive technology; and the babys gestational age,
birth weight, and fetal presentation at delivery.
To conform to the intention-to-treat approach,
all women who delivered at participating hospi-
tals were included in the analyses. Consequently,
we used random imputation (performed on the
basis of the observed distributions of the im-
puted variable and of highly correlated covari-
ates) for adjustment variables for which less than
1% of the data were missing. The only character-
istic for which missing data accounted for more
than 1% of the data was body-mass index before
pregnancy, for which data were missing for 27.2%
of the women in the study; data on body-mass
index were excluded from the analyses. No data
were missing for cesarean delivery or obstetrical
interventions. To assess whether the interven-
tion effect varied according to hospital type or
level of risk associated with the pregnancy (Sec-
tion 4 in the Supplementary Appendix), we tested
the corresponding three-way interactions: level
of careinterventiontime and risk levelinter- similar among hospitals and among women, with
ventiontime. Subgroup-specific intervention ef-
fects were reported for outcomes with significant
three-way interactions (tests were two-tailed, with
P<0.05 considered to indicate statistical signifi-
cance).31,32
Prespecified secondary outcomes were analyzed
by means of methods similar to those used for
the primary outcome. Analyses of obstetrical in-
terventions and maternal and neonatal morbidity
were based on all deliveries; analyses of intrapar-
tum maternal morbidity were restricted to women
who attempted labor. If the generalized-estimat-
ing-equations model did not converge because
there were small numbers of outcomes in some
hospitals, the intervention effect was estimated
with the use of a multivariable logistic model,
which did not account for within-hospital clus-
tering; to correct for the resulting underestima-
tion of the standard errors, a conservative P value
of less than 0.001 was used.30,32
Immediately after the intervention period,
adherence to the protocol was assessed in inter-
vention hospitals through analyses of audit reports
and onsite visits. Hospitals were considered to
have adhered to the program if they met the fol-
lowing prespecified criteria: conduct of at least
three 3-month audit sessions annually, review of
more than 80% of eligible cesarean cases, and
formulation of recommendations and formal
feedback to health personnel in maternity units
within 2 months after each audit cycle. Interven-
tion hospitals that did not meet all these criteria
were excluded from the intervention group for
per-protocol analyses. All analyses were performed
with the use of SAS software, version 9.3 (SAS
Institute) by an independent team whose mem-
bers were unaware of the group assignments.
R e sult s
Primary Outcome
Among the 40 eligible hospitals, 38 agreed to
participate and 32 were randomly selected for
inclusion in the trial in April 2008 (4 community,
22 regional, and 6 tertiary hospitals). Among the
184,952 women who delivered during the overall
study period, 105,351 women delivered during
the pre- or postintervention period and contrib-
uted directly to the estimation of the interven-
tion effect. No hospital or woman was lost to
follow-up (Fig.1). Baseline characteristics were
the exception of a marginally significant between-
group difference in maternal parity (Table1), for
which adjustments were made in multivariable
analyses. The baseline rate of cesarean delivery
was slightly higher in the control group than in
the intervention group (23.2% vs. 22.5%), and the
postintervention rate increased to 23.5% in the

n engl j med 372;18nejm.org April 30, 2015 1713

The New England Journal of Medicine
Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.
 The n e w e ng l a n d j o u r na l of m e dic i n e

whereas no significant effect was observed among

32 Hospitals underwent randomization at end
of baseline (preintervention) period, with
stratification according to hospital type
16 Were assigned to intervention group
Implementation (6 mo) and
intervention with supervision (1 yr)
16 Hospitals were included in follow-up
analysis
84,227 Women were included
in analysis (none lost to
follow-up)
24,388 Women (24,823 new-
borns) were included in
baseline period (1 yr)
36,355 Women (37,045 new-
borns) were included
in intervention period
(1.5 yr)
23,484 Women (23,902 new-
borns) were included in
postintervention period
(1 yr)
Figure 1. Randomization, Intervention, and Follow-up.
16 Were assigned to control group
No intervention
16 Hospitals were included in follow-up
analysis
100,725 Women were included
in analysis (none lost to
follow-up)
28,698 Women (29,107 new-
borns) were included in
baseline period (1 yr)
43,246 Women (43,942 new-
borns) were included
in intervention period
(1.5 yr)
28,781 Women (29,211 new-
borns) were included in
postintervention period
(1 yr)
women with high-risk pregnancies (P=0.35). In
both groups, previous cesarean delivery, nonce-
phalic presentation, prolonged labor, and an ab-
normal pattern in fetal heart rate were the lead-
ing indications for cesarean delivery (Table S1 in
the Supplementary Appendix). The reduction in
the rate of cesarean delivery was driven mostly
by the effect of the intervention on indications
for elective repeat cesarean deliveries and cesar-
ean deliveries performed after prolonged labor
(Table S1 in the Supplementary Appendix).
Among the 16 hospitals in the intervention
group, 12 (75%) met the prespecified criteria for
adherence to the protocol. The main reason for
nonadherence was the absence of formal feed-
back provided by the audit committee to local
health professionals. The per-protocol analysis,
which excluded the 4 nonadherent intervention
hospitals, yielded a more marked intervention ef-
fect than the intention-to-treat analysis (adjusted
odds ratio, 0.82; 95% CI, 0.75 to 0.89; P<0.001;
adjusted risk difference, 3.4%; 95% CI, 4.8 to
2.0) (Table2); no significant heterogeneity across
hospitals was found (P=0.38). Center-specific
results are shown in Section 5 in the Supplemen-
tary Appendix.

control group but decreased to 21.8% in the in-
tervention group (Table2).
In a between-group comparison of the change
from the preintervention period to the postinter-
vention period, there was significant but small
reduction in the rate of cesarean delivery in the
intervention group as compared with the control
group, with an adjusted odds ratio of 0.90 (95%
confidence interval [CI], 0.80 to 0.99; P=0.04)
and an adjusted absolute risk difference of 1.8%
(95% CI, 3.8 to 0.2) (Table2). The interven-
tion effect did not vary significantly across hos-
pitals with different levels of care (P=0.86 for
the three-way interaction). In contrast, the inter-
action with the level of risk associated with preg-
nancy was significant (P=0.03). Among low-risk
pregnancies, the cesarean rate in the interven-
tion group as compared with the control group
decreased significantly from the preintervention
period to the postintervention period (adjusted risk
difference, 1.7%; 95% CI, 3.0 to 0.3; adjusted
odds ratio, 0.80; 95% CI, 0.65 to 0.97; P=0.03),
Secondary Outcomes
There was a significant but small difference be-
tween the intervention group and the control
group with respect to the change in the rate of
assisted vaginal delivery from the preintervention
period to the postintervention period (adjusted
odds ratio [intervention vs. control], 0.88; 95%
CI, 0.77 to 0.99; P=0.04; adjusted risk differ-
ence, 1.1%; 95% CI, 2.2 to 0.1), and although
the rate of labor induction increased in both
groups, there was a lesser increase in the inter-
vention group (adjusted odds ratio, 0.82; 95% CI,
0.76 to 0.87; P<0.001; adjusted risk difference,
3.8%; 95% CI, 5.1 to 2.7) (Table3). Oxytocin
use during labor declined in both groups, but
the decline was greater in the control hospitals
(adjusted odds ratio, 1.16; 95% CI, 1.09 to 1.23;
P<0.001; adjusted risk difference, 3.2%; 95% CI,
1.9 to 4.6).
The intervention did not significantly affect
the risks of minor and major maternal complica-
tions (Table4). Among maternal complications
(Table S2 in the Supplementary Appendix), only
the rate of blood transfusion increased signifi-

1714 n engl j med 372;18nejm.org April 30, 2015

The New England Journal of Medicine

Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.
 A Trial to Reduce Cesarean Delivery R ates

Table 1. Baseline Characteristics of Hospitals and Patients.*

Characteristic Intervention Control

Hospitals Patients Hospitals Patients

(N=16) (N=24,388) (N=16) (N=28,698)
Hospitals
Type of hospital no. (%)
Community 2 1,325 (5.4) 2 803 (2.8)
Regional 11 16,045 (65.8) 11 18,684 (65.1)
Tertiary care 3 7,018 (28.8) 3 9,211 (32.1)
Academic hospital 5 8,977 (36.8) 4 16,159 (56.3)
Physicians per hospital
Obstetriciangynecologist 5.64.3 6.24.8
Family physician 9.48.9 7.45.3
Patients
Maternal age at delivery
Mean yr 29.45.1 29.84.9
Range no. (%)
17 yr 255 (1.0) 122 (0.4)
1834 yr 20,777 (85.2) 24,370 (84.9)
35 yr 3,356 (13.8) 4,206 (14.7)
Parity no. (%)
0 10,727 (44.0) 13,165 (45.9)
1 8,893 (36.5) 10,607 (37.0)
2 4,768 (19.6) 4,926 (17.2)
Gestational age at delivery no. (%)
<37 wk 2,069 (8.5) 1,982 (6.9)
3741 wk 22,269 (91.3) 26,675 (93.0)
42 wk 50 (0.2) 41 (0.1)
Previous cesarean delivery no. (%) 2,782 (11.4) 3,306 (11.5)
High-risk pregnancy no. (%) 12,910 (52.9) 13,981 (48.7)
Delivery by family physician no. (%) 9,715 (39.8) 11,805 (41.1)
Presentation of infant no./total no. (%)
Cephalic 23,190/24,823 (93.4) 27,285/29,107 (93.7)
Breech 1,436/24,823 (5.8) 1,522/29,107 (5.2)
Transverse 197/24,823 (0.8) 300/29,107 (1.0)
Neonatal birth weight no./total no. (%)
<1500 g 351/24,823 (1.4) 245/29,107 (0.8)
15002499 g 1,367/24,823 (5.5) 1,417/29,107 (4.9)
25003999 g 20,563/24,823 (82.8) 24,650/29,107 (84.7)
4000 g 2,542/24,823 (10.2) 2,79/29,1075 (9.6)
Stillbirths no./total no. (%) 121/24,823 (0.5) 109/29,107 (0.4)

* Plusminus values are means SD. There were no significant between-group differences at baseline except with regard to parity.
Percentages may not sum to 100 because of rounding.
P<0.05 for the difference between groups. The P value was calculated by means of a univariate model with the use of generalized estimating
equations in which the structure for patient characteristics was exchangeable or independent.
A pregnancy was considered to be low risk if the woman gave birth to a single baby in cephalic presentation, had not used assisted repro-
ductive technology, was between 18 and 39 years of age, had a body-mass index before pregnancy between 17 and 29, had no prior cesarean
delivery, no prior or current stillbirth, no transfer to another hospital during preganancy, and no other pathologic condition or complication
during the current pregnancy or a prior pregnancy, and if the gestational age was between 37 and 42 weeks. A pregnancy was considered to
be at risk if any of these conditions was not met. See Section 4 of the Supplementary Appendix for further details.

n engl j med 372;18nejm.org April 30, 2015 1715

The New England Journal of Medicine
Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.
 1716
Table 2. Rates of Cesarean Delivery.

Crude Difference
in Rate Change
Factor Intervention Group Control Group (95% CI)* Effect of Intervention

Post- Post- Adjusted Absolute Adjusted

Baseline intervention Baseline intervention Risk Difference Odds Ratio
(N=24,388) (N=23,484) Difference (N=28,698) (N=28,781) Difference (95% CI) (95% CI) P Value

number (percent) percent number (percent) percent

Intention-to-treat analysis
The

Cesarean delivery 5484 (22.5) 5128 (21.8) 0.7 6671 (23.2) 6767 (23.5) 0.3 0.9 (1.9 to 0.1) 1.8 (3.8 to 0.2) 0.90 (0.80 to 0.99) 0.04
no. (%)
Risk level of pregnancy
Low
Total no. 11,478 10,067 14,717 13,019
Cesarean delivery 971 (8.5) 763 (7.6) 0.9 1256 (8.5) 1172 (9.0) 0.5 1.3 (2.3 to 0.4) 1.7 (3.0 to 0.3) 0.80 (0.65 to 0.97) 0.03
no. (%)
High
Total no. 12,910 13,417 13,981 15,762
Cesarean delivery 4513 (35.0) 4365 (32.5) 2.4 5415 (38.7) 5595 (35.5) 3.2 0.8 (0.8 to 2.4) 0.9 (3.1 to 1.1) 0.96 (0.87 to 1.05) 0.35
n e w e ng l a n d j o u r na l

no. (%)

The New England Journal of Medicine

of

Per-protocol analysis
Total no. 16,802 16,144 28,698 28,781

n engl j med 372;18nejm.org April 30, 2015

Cesarean delivery 3669 (21.8) 3268 (20.2) 1.6 6671 (23.2) 6767 (23.5) 0.3 1.9 (3.0 to 0.7) 3.4 (4.8 to 2.0) 0.82 (0.75 to 0.89) <0.001
no. (%)

Copyright 2015 Massachusetts Medical Society. All rights reserved.

m e dic i n e

* The unadjusted crude difference in rate change was calculated as follows: (postintervention rate baseline rate in intervention group) (postintervention rate baseline rate in control
group).
The adjusted absolute risk difference represents adjusted differences between group-specific changes over time and was estimated with the use of the generalized-estimating-equations
(GEE) model (see Section 5 in the Supplementary Appendix).
The adjusted odds ratios for the interaction between groups (intervention vs. control) and time (postintervention period vs. baseline period) were estimated with the use of the GEE
model.

Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
In the GEE model, P values of less than 0.05 were considered to indicate statistical significance, and P values of less than 0.06 were considered to indicate marginal significance.
Subgroup-specific effects were reported when a significant interaction with the hospital type or the pregnancy risk level was detected.
A logistic model was used because the calculations for the GEE model did not converge. For this model, P values of less than 0.001 were considered to indicate statistical significance
and P values of less than 0.003 were considered to indicate marginal significance
For the per-protocol analysis, four hospitals with a low level of adherence to the protocol were excluded from the intervention group.
 Table 3. Rates of Obstetrical Interventions Other Than Unplanned Cesarean Delivery.

Crude Difference
in Rate Change
Intervention Intervention Group Control Group (95% CI)* Effect of Intervention

Adjusted
Post- Post- Absolute Risk Adjusted
Baseline intervention Baseline intervention Difference Odds Ratio
(N=24,388) (N=23,484) Difference (N=28,698) (N=28,781) Difference (95% CI) (95% CI) P Value

number (percent) percent number (percent) percent

All deliveries
Planned cesarean 2,939 (12.1) 2,872 (12.2) 0.1 3,701 (12.9) 3,907 (13.6) 0.7 0.5 1.0 0.92 0.42
delivery (1.3 to 0.3) (3.2 to 2.0) (0.74 to 1.13)
Pharmacologic 4,345 (17.8) 5,501 (23.4) 5.6 5,235 (18.2) 7,872 (27.4) 9.2 3.5 3.8 0.82 <0.001
induction of (4.5 to 2.5) (5.1 to 2.7) (0.76 to 0.87)
labor
Artificial rupture 11,563 (47.4) 11,972 (51.0) 3.6 13,495 (47.0) 14,534 (50.5) 3.5 0.1 0.5 1.02 0.87
of membranes (1.1 to 1.3) (4.4 to 5.0) (0.84 to 1.22)
Women who attempted
labor
Total no. 21,449 20,612 24,997 24,874
Intrapartum 2,545 (11.9) 2,256 (10.9) 1.0 2,970 (11.9) 2,860 (11.5) 0.4 0.5 0.9 0.91 0.14
cesarean delivery (1.4 to 0.3) (2.1 to 0.3) (0.80 to 1.03)

n engl j med 372;18nejm.org April 30, 2015

Assisted vaginal 2,535 (11.8) 2,223 (10.8) 1.0 2,574 (10.3) 2,605 (10.5) 0.2 1.2 1.1 0.88 0.04

The New England Journal of Medicine

delivery (2.0 to 0.4) (2.2 to 0.1) (0.77 to 0.99)
A Trial to Reduce Cesarean Delivery R ates

Use of oxytocin 7,652 (35.7) 6,205 (30.1) 5.6 9,932 (39.7) 7,572 (30.4) 9.3 3.7 3.2 1.16 <0.001
during labor (2.5 to 4.9) (1.9 to 4.6) (1.09 to 1.23)
Epidural analgesia 14,416 (67.2) 14,004 (67.9) 0.7 18,364 (73.5) 18,339 (73.7) 0.2 0.5 0.4 1.02 0.75

Copyright 2015 Massachusetts Medical Society. All rights reserved.

(0.7 to 1.7) (2.3 to 2.9) (0.89 to 1.17)
Episiotomy 3,762 (17.5) 2,953 (14.3) 3.2 4,777 (19.1) 3,871 (15.6) 3.5 0.3 0.1 1.01 0.87
(0.6 to 1.3) (2.0 to 2.7) (0.85 to 1.21)

* The unadjusted crude difference in rate change was calculated as follows: (postintervention rate baseline rate in intervention group) (postintervention rate baseline rate in control
group).
The adjusted absolute risk difference represents adjusted differences between group-specific changes over time and was estimated with the use of the generalized-estimating-equations

Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
(GEE) model (see Section 5 in the Supplementary Appendix).
The adjusted odds ratios for the interaction between groups (intervention vs. control) and time (postintervention period vs. baseline) were estimated with the use of the GEE model.
A logistic model was used to calculate the values for the crude difference in rate change for high-risk pregnancies because the calculations for the GEE model did not converge.
According to the GEE model, P values of less than 0.05 were considered to indicate statistical significance (for the logistic model, P<0.001 was considered to indicate significance), and
P values of less than 0.06 were considered to indicate marginal significance (for the logistic model, P<0.003 was considered to indicate marginal significance). Subgroup-specific effects
were reported when a significant interaction with the hospital type or the pregnancy risk level was detected.

1717
 The n e w e ng l a n d j o u r na l
cantly in the intervention group relative to the
control group (adjusted odds ratio, 1.70; 95% CI,
1.18 to 2.43; P=0.004).
Minor neonatal morbidity increased in both
groups (Table4), but the increases were smaller
in the intervention group (adjusted odds ratio,
0.88; 95% CI, 0.82 to 0.94; P<0.001; adjusted risk
difference, 1.7%; 95% CI, 2.6 to 0.9). Major
neonatal morbidity decreased significantly in
the intervention group as compared with the
control group (adjusted odds ratio, 0.81; 95% CI,
0.66 to 0.98; P=0.03; adjusted risk difference,
0.7%; 95% CI, 1.3 to 0.1) (Table S4 in the
Supplementary Appendix); effects included small
absolute reductions in major trauma (adjusted
risk difference, 0.23%; 95% CI, 0.40 to 0.01;
P=0.046), the use of invasive mechanical ventila-
tion (adjusted risk difference, 0.38%; 95% CI,
0.60 to 0.09; P=0.01); and intrapartum and
neonatal deaths (adjusted risk difference, 0.06%;
95% CI, 0.08 to 0.03; P<0.001). The effect of
the intervention on major neonatal morbidity re-
mained significant after exclusion of preterm
births (adjusted odds ratio, 0.82; 95% CI, 0.70 to
0.96; P=0.02) and appeared to be similar in wom-
en with low-risk pregnancies and those with high-
risk pregnancies (Table S5 in the Supplementary
Appendix), but the effect was not significant in
women with low-risk pregnancies (P=0.11).
Discussion
This multifaceted intervention, which involved on-
site professional training in evidence-based man-
agement of labor and delivery and was designed to
promote clinical audits, feedback, and implemen-
tation of best practices, led to a statistically sig-
nificant but clinically small reduction in the rate
of cesarean deliveries. The reduction was observed
among women with low-risk pregnancies but not
among those with high-risk pregnancies.
Furthermore, the intervention was associated
with a significant reduction in minor and major
neonatal morbidity among babies born to women
with low-risk pregnancies and among those born
to women with high-risk pregnancies. This re-
sult may reflect improvements in the standard of
care implemented in individual hospitals in the
intervention group. However, these results must
be interpreted with caution because of an unex-
pected increase in neonatal morbidity in the con-
trol group, which was presumably due to random
1718 n engl j med 372;18nejm.org April 30, 2015
The New England Journal of Medicine
Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.
of m e dic i n e
variation. Overall, these results suggest that the
QUARISMA program resulted in a significant but
small reduction in the total cesarean delivery rate,
without increasing neonatal and maternal mor-
bidity and mortality.
Among 10 prior studies randomized and
nonrandomized that assessed the effects of
similar interventions targeting health care pro-
viders (obtaining second opinions, conducting
audits with feedback, and conducting peer re-
views), 3 showed significant reductions in rates of
cesarean delivery.33 However, these results were
ambiguous, since only 1 of the 4 randomized
trials showed a significant, albeit small, reduc-
tion (adjusted risk difference, 1.9%).21,33 Our
trial yielded very similar results to that trial and
confirmed the benefits of a multifaceted strategy
involving audits and feedback and targeting health
care providers, as suggested in previous nonran-
domized studies and systematic reviews.17-21,33-40
Our trial also had one of the largest samples,
including more than 65% of all deliveries in
Quebec province during the study period. In ad-
dition, our results indicated that the reduction in
the rate of cesarean delivery was driven by the
effect of the intervention in low-risk pregnancies.
The QUARISMA program was designed to al-
low health professionals to assess care relative
to operational standards (algorithms), to detect
cases in which care could have been improved
and an unnecessary cesarean delivery avoided, and
to standardize clinical practice. The interventions
in the trial included identification of opinion
leaders, the use of onsite training, and the intro-
duction of standardized internal clinical audits
and feedback. The safe reduction, albeit modest,
in the rates of cesarean delivery observed in this
trial and the moderate efforts required to maintain
the program (approximately 2 days per 3-month
cycle to conduct an audit session, develop recom-
mendations, provide feedback, and review the
implementation of the recommendations) sug-
gest that a similar intervention may be benefi-
cial in other countries or regions in which the
rates of cesarean delivery are similar or higher.
Our study had some limitations. Since hospi-
tals were the unit of randomization, and the
number of deliveries varied across hospitals, there
were differences in the distribution of certain
hospital characteristics across groups at baseline.
These differences were adjusted a priori in multi-
variable analyses. Furthermore, the intervention
 Table 4. Maternal and Neonatal Morbidity.*

Crude Difference
in Rate Change
Factor Intervention Group Control Group (95% CI) Effect of Intervention

Post- Post Adjusted Absolute Adjusted

Baseline intervention Baseline intervention Risk Difference Odds Ratio
(N=24,388) (N=23,484) Difference (N=28,698) (N=28,781) Difference (95% CI) (95% CI) P Value

number (percent) percent number (percent) percent

Composite risk of maternal
morbidity
Minor morbidity 3293 (13.5) 3576 (15.2) 1.7 3869 (13.5) 4244 (14.7) 1.3 0.5 (0.4 to 1.3) 0.3 (1.2 to 1.8) 1.02 (0.91 to 1.15) 0.76
Major morbidity 161 (0.66) 167 (0.71) 0.05 138 (0.48) 141 (0.49) 0.01 0.04 (0.14 to 0.23) 0.03 (0.11 to 0.23) 1.06 (0.77 to 1.48) 0.71
Composite risk of neonatal
morbidity
Total no. of births 24,823 23,902 29,107 29,211
Minor morbidity 3936 (15.9) 4261 (17.8) 2.0 3947 (13.6) 5002 (17.1) 3.6 1.6 (2.5 to 0.7) 1.7 (2.6 to 0.9) 0.88 (0.82 to 0.94) <0.001
Major morbidity 1172 (4.7) 1070 (4.5) 0.2 1,018 (3.5) 1,156 (4.0) 0.5 0.7 (1.2 to 0.2) 0.7 (1.3 to 0.1) 0.81 (0.66 to 0.98) 0.03

* Events that determined minor maternal morbidity included blood transfusion, perineal tear (grade 34), puerperal infection or sepsis, gastrointestinal complications, complications
from analgesia, postpartum hospital stay of 7 days or more, admission to the intensive care unit, and readmission to the hospital. Events that determined major maternal morbidity in-
cluded maternal death, hysterectomy, symptomatic uterine rupture, thromboembolic disease, and injury to internal organs. Events that determined minor neonatal morbidity included
cardiopulmonary morbidity, an Apgar score between 4 and 6 at 5 minutes after birth, moderate acidosis, minor trauma, noninvasive mechanical ventilation, blood transfusion, and neo-
natal infection or sepsis. Events that determined major neonatal morbidity included intrapartum or neonatal death, an Apgar score of less than 4 at 5 minutes after birth, major acido-

n engl j med 372;18nejm.org April 30, 2015

The New England Journal of Medicine

sis, major trauma, intraventricular hemorrhage, seizure, neurologic damages, invasive mechanical ventilation, necrotizing enterocolitis, and hypoxicischemic encephalopathy. For com-
plete definitions of minor and major maternal and fetal morbidity, see Section 3 in the Supplementary Appendix.
A Trial to Reduce Cesarean Delivery R ates

The unadjusted crude difference in rate change was calculated as follows: (postintervention rate baseline rate in intervention group) (postintervention rate baseline rate in control
group).
The adjusted absolute risk difference represents adjusted differences between group-specific changes over time and was estimated with the use of the generalized-estimating-equations
(GEE) model (see Appendix 5 in the Supplementary Appendix).

Copyright 2015 Massachusetts Medical Society. All rights reserved.

The adjusted odds ratios for the interaction between groups (intervention vs. control) and time (postintervention period vs. baseline) were estimated with the use of the GEE model.
Included are infants born at a gestational age of at least 24 weeks and with a birth weight of at least 500 g at delivery.
According to the GEE model, P values of less than 0.05 were considered to indicate statistical significance, and P values of less than 0.06 were considered to indicate marginal signifi-
cance. Subgroup-specific effects were reported when a significant interaction with the hospital type or the pregnancy risk level was detected.

Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
1719
 The n e w e ng l a n d j o u r na l of m e dic i n e

was not fully implemented at four intervention vention significantly reduced the rate of cesarean
hospitals. Finally, because we tested a complex, delivery among women with low-risk pregnancies
multifaceted intervention, it is not possible to de- but not among those with high-risk pregnancies.
termine which of its components were primarily Supported by grants (200702MCT-171307-RFA-CFCF-153236
responsible for the observed effect. and MOP 81275) from the Canadian Institutes of Health Re-
search.
In summary, a program in which audits and Disclosure forms provided by the authors are available with
best practices were implemented resulted in a the full text of this article at NEJM.org.
significant but small reduction in the rate of We thank all the medical and administrative staff at the 32
participating hospitals for their contributions to this trial and the
cesarean deliveries without increasing neonatal data collectors, research nurses, and medical archivists in each
and maternal morbidity and mortality. The inter- hospital whose assistance helped to ensure the quality of the data.

References
1. Martin JA, Hamilton BE, Sutton PD, et 12. Allen VM, OConnell CM, Liston RM,
al. Births: final data for 2003. Natl Vital Baskett TF. Maternal morbidity associat-
Stat Rep 2005;54:1-116. ed with cesarean delivery without labor
2. Thomas J, Paranjothy S. Royal College compared with spontaneous onset of la-
of Obstetricians and Gynecologists Clini- bor at term. Obstet Gynecol 2003;102:477-
cal Effectiveness Support Unit:national 82.
sentinel caesarean section audit report. 13. Zanardo V, Simbi AK, Franzoi M, Sol-
London:RCOG Press, 2001 (https://w ww d G, Salvadori A, Trevisanuto D. Neona-
.rcog.org.uk/globalassets/documents/ tal respiratory morbidity risk and mode of
guidelines/research--audit/nscs_audit delivery at term: influence of timing of
.pdf). elective caesarean delivery. Acta Paediatr
3. Blondel B, Lelong N, Kermarrec M, 2004;93:643-7.
Goffinet F. Trends in perinatal health in 14. Levine EM, Ghai V, Barton JJ, Strom
France between 1995 and 2010: results CM. Mode of delivery and risk of respira-
from the National Perinatal Surveys. J Gy- tory diseases in newborns. Obstet Gyne-
necol Obstet Biol Reprod (Paris) 2012;41: col 2001;97:439-42.
151-66. (In French.) 15. Giving birth in Canada: the costs. Ot-
4. Canadian perinatal health report, tawa:Canadian Institute for Health Infor-
2008. Ottawa:Public Health Agency of mation, 2007 (https://secure.cihi.ca/free_
Canada, 2008 (http://www.phac-aspc.gc products/Costs_Report_06_Eng.pdf).
.ca/publicat/2008/cphr-rspc/index-eng 16. Rowe AK, de Savigny D, Lanata CF,
.php). Victora CG. How can we achieve and
5. Health system performance: health maintain high-quality performance of
indicators. Catalogue no. 82-221-XIE. Vol. health workers in low-resource settings?
2005, No. 3. Ottawa:Statistics Canada, Lancet 2005;366:1026-35.
2005 (http://www.statcan.gc.ca/pub/82- 17. Chaillet N, Dub E, Dugas M, et al.
221-x/82-221-x2005002-eng.htm). Evidence-based strategies for implement-
6. Giving birth in Canada: regional ing guidelines in obstetrics: a systematic
trends from 2001-2002 to 2005-2006. Ot- review. Obstet Gynecol 2006;108:1234-45.
tawa:Canadian Institute for Health Infor- 18. Chaillet N, Dumont A. Evidence-
mation, 2007 (https://secure.cihi.ca/free_ based strategies for reducing cesarean
products/childbirth_aib_070725_e.pdf). section rates: a meta-analysis. Birth 2007;
7. National Collaborating Centre for 34:53-64.
Womens and Childrens Health. Caesar- 19. Lomas J, Anderson GM, Domnick-
ean section: clinical guideline. London: Pierre K, Vayda E, Enkin MW, Hannah WJ.
RCOG Press, 2004. Do practice guidelines guide practice?
8. Liu S, Heaman M, Joseph KS, et al. The effect of a consensus statement on
Risk of maternal postpartum readmis- the practice of physicians. N Engl J Med
sion associated with mode of delivery. 1989;321:1306-11.
Obstet Gynecol 2005;105:836-42. 20. Lomas J, Enkin M, Anderson GM,
9. Hall MH, Bewley S. Maternal mortal- Hannah WJ, Vayda E, Singer J. Opinion
ity and mode of delivery. Lancet 1999;354: leaders vs audit and feedback to imple-
776. ment practice guidelines: delivery after
10. Alexander S, Wildman K, Zhang W, previous cesarean section. JAMA 1991;
Langer M, Vutuc C, Lindmark G. Maternal 265:2202-7.
health outcomes in Europe. Eur J Obstet 21. Althabe F, Belizn JM, Villar J, et al.
Gynecol Reprod Biol 2003; 111:
Suppl 1: Mandatory second opinion to reduce rates
S78-S87. of unnecessary caesarean sections in Lat-
11. Lydon-Rochelle M, Holt VL, Martin in America: a cluster randomised con-
DP, Easterling TR. Association between trolled trial. Lancet 2004;363:1934-40.
method of delivery and maternal rehospi- 22. Ministry of Health and Social Ser-
talization. JAMA 2000;283:2411-6. vices (Quebec). Administrative database
MED-ECHO (Quebecs hospital discharg-
es). January 2008.
23. Beyond the numbers: reviewing ma-
ternal deaths and complications to make
pregnancy safer. Geneva:World Health
Organization, 2004 (http://www.who.int/
maternal_child_adolescent/documents/
9241591838/en).
24. McMahon MJ, Luther ER, Bowes WA
Jr, Olshan AF. Comparison of a trial of
labor with an elective second cesarean
section. N Engl J Med 1996;335:689-95.
25. Hannah ME, Hannah WJ, Hewson
SA, Hodnett ED, Saigal S, Willan AR.
Planned caesarean section versus planned
vaginal birth for breech presentation at
term: a randomised multicentre trial.
Lancet 2000;356:1375-83.
26. Landon MB, Hauth JC, Leveno KJ, et
al. Maternal and perinatal outcomes as-
sociated with a trial of labor after prior
cesarean delivery. N Engl J Med 2004;351:
2581-9.
27. American College of Obstetricians
and Gynecologists. ACOG practice bulle-
tin: vaginal birth after previous cesarean
delivery. Number 5, July 1999 (replaces
practice bulletin number 2, October
1998): clinical management guidelines
for obstetrician-gynecologists. Int J Gyn-
aecol Obstet 1999;66:197-204.
28. Donner A, Klar N. Design and analy-
sis of cluster randomization trials in
health research. New York:Oxford Uni-
versity Press, 2000.
29. Pinol APY, Piaggio G. ACluster: de-
sign and analysis of cluster randomiza-
tion trials. 2nd ed. Geneva:World Health
Organization, 2000.
30. Zeger SL, Liang KY, Albert PS. Models
for longitudinal data: a generalized esti-
mating equation approach. Biometrics
1988;44:1049-60.
31. Abrahamowicz M, Beauchamp ME,
Fournier P, Dumont A. Evidence of sub-
group-specific treatment effect in the ab-
sence of an overall effect: is there really a
contradiction? Pharmacoepidemiol Drug
Saf 2013;22:1178-88.
32. Dumont A, Fournier P, Abrahamow-
icz M, Traor M, Haddad S, Fraser WD.
Quality of care, risk management, and

1720 n engl j med 372;18nejm.org April 30, 2015

The New England Journal of Medicine

Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.
 A Trial to Reduce Cesarean Delivery R ates

technology in obstetrics to reduce hospi- SM. Using the medical audit cycle to re- education: do conferences, workshops,
tal-based maternal mortality in Senegal duce cesarean section rates. Am J Obstet rounds, and other traditional continuing
and Mali (QUARITE): a cluster-ran- Gynecol 1996;174:199-205. education activities change physician be-
domised trial. Lancet 2013;382:146-57. 36. Bickell NA, Zdeb MS, Applegate MS, havior or health care outcomes? JAMA
33. Khunpradit S, Tavender E, Lumbiga- Roohan PJ, Sui AL. Effect of external peer 1999;282:867-74.
non P, Laopaiboon M, Wasiak J, Gruen review on cesarean delivery rates: a state- 39. Lagrew DC Jr, Morgan MA. Decreas-
RL. Non-clinical interventions for reduc- wide program. Obstet Gynecol 1996;87: ing the cesarean section rate in a private
ing unnecessary caesarean section. Co- 664-7. hospital: success without mandated clini-
chrane Database Syst Rev 2011; 6: 37. Main EK. Reducing cesarean birth cal changes. Am J Obstet Gynecol 1996;
CD005528. rates with data-driven quality improve- 174:184-91.
34. Liang WH, Yuan CC, Hung JH, et al. ment activities. Pediatrics 1999;103:Suppl 40. Myers SA, Gleicher N. A successful
Effect of peer review and trial of labor on E:374-83. program to lower cesarean-section rates.
lowering cesarean section rates. J Chin 38. Davis D, OBrien MA, Freemantle N, N Engl J Med 1988;319:1511-6.
Med Assoc 2004;67:281-6. Wolf FM, Mazmanian P, Taylor-Vaisey A. Copyright 2015 Massachusetts Medical Society.
35. Robson MS, Scudamore IW, Walsh Impact of formal continuing medical

n engl j med 372;18nejm.org April 30, 2015 1721

The New England Journal of Medicine
Downloaded from nejm.org on September 22, 2015. For personal use only. No other uses without permission.
Copyright 2015 Massachusetts Medical Society. All rights reserved.