International Journal for Quality in Health Care 1999; Volume 11, Number 4: pp. 283–291
Definition and adjustment of Cesarean
section rates and assessments of hospital
performance
STEPHEN B. KRITCHEVSKY1, BARBARA I. BRAUN2, PETER A. GROSS3, CAROL S. NEWCOMB2,
CAROL ANN KELLEHER2 AND BRYAN P. SIMMONS4
1
Department of Preventive Medicine, University of Tennessee, Memphis, 2Department of Research and Evaluation, Joint Commission
on Accreditation of Healthcare Organizations, Oakbrook Terrace, Illinois, 3Department of Internal Medicine, Hackensack University
Medical Center, New Jersey, and 4Quality Management, Methodist Health System, Memphis, Tennessee, USA
Abstract
Background. Demand is growing for comparative data such as Cesarean section rates, but little effort has been made to
develop either standardized definitions or risk adjustment approaches.
Objective. To determine to what extent a seemingly straightforward indicator like Cesarean section rate will vary when
calculated according to differing definitions used by various performance measurement systems.
Design. Retrospective data abstraction of 200 deliveries per hospital.
Setting. Fifteen acute care hospitals including two from outside the USA.
Measurements. Four widely-used performance measurement systems provided specifications for their Cesarean section
indicators. Indicator specifications varied on inclusion criteria (whether the population was defined using Diagnostic Related
Groups or ICD-9-CM procedure codes or ICD-9-CM diagnosis codes) and risk-adjustment methods and factors. Rates and
rankings were compared across hospitals using different Cesarean section indicator definitions and indicators with and
without risk adjustment.
Results. Calculated Cesarean section rates changed substantially depending on how the numerator and denominator cases
were identified. Relative performance based on Cesarean section rankings is affected less by differing indicator definitions
than by whether and how risk adjustment is performed.
Conclusions. Judgments about organizational performance should only be made when the comparisons are based upon
identical indicators. Research leading to a uniform indicator definition and standard risk adjustment methodology is needed.
Keywords: Cesarean section, hospitals, risk, statistics
Nearly every health care organization is asked by payers,
purchasers, business coalitions, consumer groups, accrediting
bodies, and/or government agencies to provide clinical performance measure data [1]. Hundreds of performance
measurement systems exist to help process this data into
information used to make inter-hospital comparisons. Although many different systems support indicators that ostensibly measure the same clinical occurrence, they often use
different specifications for event definition, data collection,
analysis and reporting [2,3]. One of the most commonly
reported performance measures is the rate of Cesarean sections. In the USA, the Cesarean section rate rose precipitously
during the 1980s and remains much higher in this country
than in others [4]. Large variations in individual physicians’
rates of performance suggest that a percentage of Cesarean
sections are done for reasons other than medical necessity.
Cesarean sections are relatively more expensive than vaginal
deliveries and safely reducing the rate would be expected to
Address correspondence to Stephen B. Kritchevsky, Department of Preventive Medicine, University of Tennessee, 66 North
Pauline, Suite 633, Memphis TN 38105, USA. Tel: +1 901 448 8757. Fax: +1 901 448 7641. E-mail: [email protected]. Address requests for reprints to Barbara I. Braun, Department of Research and Evaluation,
Joint Commission on Accreditation of Healthcare Organizations, One Renaissance Blvd, Oakbrook Terrace IL 60181, USA.
Tel: +1 630 792 5928. Fax: +1 630 792 4928. E-mail: [email protected]
 1999 International Society for Quality in Health Care and Oxford University Press
283
S. B. Kritchevsky et al.
yield substantial cost savings [5]. A Cesarean section rate has
high face validity and is considered easy to measure as its
determinants can be derived from administrative data. Based
on the premise that release of Cesarean section rate information will help make providers accountable for the quality
of care and allow users of information to compare quality
and cost across providers, comparative Cesarean section data
has been released by organizations such as the Public Citizen’s
Health Research Group and the New England HEDIS Coalition [6,7].
The apparent simplicity of the Cesarean section rate,
however, can be deceptive. Though many performance
measurement systems include Cesarean section rates in their
list of indicators, there is little consistency across these systems
in the specifications of how to calculate the rate. There are
differences in how the population is defined (i.e. who is
included and excluded) and in the application of risk adjustment methodologies. For example, the overall rate reported by the National Center for Health Statistics is not
risk adjusted. On the other hand, several investigators recommend the use of sophisticated risk-adjusted models which
explain a high percentage of the variation in Cesarean section
rates using patient factors [8–10]. Aron et al. have recently
used a risk adjustment algorithm developed for their study
to compare hospital performance in a sample of 21 Cleveland
area hospitals [10]; risk adjustment led to marked differences
in hospital rankings.
The impact of differing definitions and risk adjustment
strategies on Cesarean section rates has not been formally
evaluated. The objective of this study was to determine
whether Cesarean section rates as defined by different comparative measurement systems would lead to similar rates
and rankings among hospitals. If currently used definitions
are inconsistent, then judgments concerning hospital and
health plan performance may be unreliable based on currently
available measures.
Methods
This study is part of a ongoing collaboration between the
Society for Healthcare Epidemiology of America (SHEA)
and the Joint Commission on Accreditation of Healthcare
Organizations intended to support the effective use, development, understanding and continuous improvement of
clinical quality indicators [11,12]. A mailed survey was sent
to a volunteer sample of SHEA hospital epidemiologists in
April of 1995 asking which indicator focus areas they would
prefer to study based on salience to their institution. Based
on the results of this survey, three clinical areas were identified:
Cesarean section, peri-operative mortality and mortality after
coronary artery bypass graft surgery. This paper describes the
findings related to the Cesarean section indicators; information on the other two clinical areas is forthcoming.
Before the project began, performance measurement systems with indicators of interest to the study were identified.
Five had Cesarean section indicators in current use and
consented to cooperate with the study. Two of the five
284
indicators were identical, thus four indicators are compared
herein. Measurement systems agreed to participate under
the condition of anonymity: therefore, the systems are not
specifically identified. The sponsors of these systems included
the United States government, state-hospital associations, and
a private system. Each of these systems provided specifications for their indicator definitions and algorithms. Two
of the systems agreed to apply their risk adjustment models
directly to the study data. Indicator specifications and risk
adjustment models used in this study may not be identical
to those currently used by the measurement system because
the systems may have revised their specifications since the
study was conducted.
Data collection
Indicator specifications from each system were consolidated
into a single data collection form with instructions for a data
collection process that would accommodate the analyses
needed for each system. Most of the data elements were
available from administrative data except for parity and a
history of Cesarean section. Sites were instructed to collect
the most recent 200 deliveries, or for sites where fewer than
200 deliveries occurred in a year, to collect the total number
of deliveries over the course of the study year (September
1994 to August 1995). Two sites used a sampling approach
that the project had used in its study of peri-operative
mortality. In this approach, all Cesarean section cases were
sampled as was a random sample of non-cases. The sampling
fractions for non-Cesarean section deliveries were 19% and
68% for the two hospitals.
Data analysis
Separate programs were written for the indicator algorithms
according to the performance measurement system’s specifications using statistical software (SAS, SAS Institute, Inc.,
Cary NC, USA). Rates at the hospitals that employed sampling
were calculated after weighting records by the inverse of their
sampling fraction.
Unadjusted rates were calculated for each system. The risk
adjusted rate for systems B and D used logistic regression
models to calculate the ratio of the hospital’s observed rate
to the hospital’s predicted rate (O/P) multiplied by the overall
rate for that system’s measure. The system’s overall rate was
based on the hospitals that routinely provided data to the
system, not the hospitals participating in this study. The
consistency of rankings across systems was assessed using
Spearman’s rank correlation coefficient.
For unadjusted rates, outlier hospitals were identified after
constructing 95% confidence intervals (CI) around the systems average rate. If the hospital’s overall rate was outside
of these limits, the hospital was identified as an outlier. The
formula used for calculating the 95% CI for proportions (p)
was
CI=p±(1.96 × SE)
where the standard error (SE)=([p(1–p)/n] [13].
Cesarean section rate variation
Table 1 Characteristics of the study hospitals
For adjusted rates, performance measurement systems B and
D designated hospitals as outliers as part of their processing
of the study data.
associated with deliveries (370–375). System B used only
procedure codes to identify cases, System C used only V
codes (V27.0–V27.9) to identify deliveries and System D
used both ICD-9-CM diagnosis codes and V codes to identify
deliveries. V codes are defined as a supplementary classification of factors influencing health status and contact with
health services to deal with occasions when circumstances
other than a disease or injury classifiable to categories 001–999
are recorded as ‘diagnoses’ or ‘problems’ [15].
The net effect of the differing numerator and denominator
definitions on Cesarean section rates for the aggregate of the
15 study hospitals is shown in Table 3. System D’s definition
was more inclusive that the other systems and identified
the greatest number of cases for both the numerator and
denominator. Compared with Systems B and A, the number
of cases had little or no net overall effect on the mean
Cesarean section rate across hospitals. The last column in
Table 3 shows the rates after adjustments by Systems B and
D. Both B and D employed logistic regression models that
accounted for selected diagnosis codes, age of the mother,
and payer. System D also included parity, race of the mother,
and history of Cesarean section as adjusting variables. After
risk adjustment the mean rates for Systems B and D showed
different patterns. The overall adjusted rate rose with System
B and fell with System D.
Five hospitals coded fewer than 30 cases using V codes
to indicate normal deliveries; therefore their rates for System
C could not be calculated. When the same subset of the
other 10 hospitals is used to compare the other systems to
System C, System C has 13–14% fewer cases in the population
than other systems. Relatively more Cesarean sections were
excluded compared with vaginal births. The net effect was a
slightly lower Cesarean section rate in System C (20.9%
overall) compared with the other systems (A, 21.7%; B,
21.3%; D, 21.5%) in the subset.
Results
Hospital rates and rank order
Characteristics of participating hospitals
Table 4 presents the unadjusted Cesarean section rates calculated using the algorithms specified by the different performance measurement systems using the same raw data.
Though the systems’ rates were correlated with one another
[all Spearman rank correlations (rs) between 0.91 and 0.98],
there were differences in rates that could be attributed to
differences in indicator specifications. The last column in the
table shows the maximum percentage difference (MPD)
between the system that yields the lowest rate and the one
that yields the highest rate. The MPD’s ranged from 0 to
47.2%; the median was 4.9%. It did not appear that one
particular system was consistently discrepant with the other
systems. The largest MPD’s involved System A’s indicator
three times, System B’s indicator twice, and Systems C and
D’s indicators once each.
The relative rankings of hospitals within indicator systems
are shown in parentheses in Table 4. As only 10 hospitals
were included in System C, its rankings are not directly
comparable with those of the other systems. When a common
subset of just these 10 hospitals was examined, System
Characteristic
n
%
............................................................................................................
Bed size
< 249
1
6.7
250–499
6 40.0
500–749
5 33.3
750–1000
3 20.0
Location
Urban
13 86.7
Rural
2 13.3
Level of obstetrical care
1
2 13.3
2
5 33.3
3
8 53.3
Teaching hospital
Member of Council of Teaching Hospitals
6 40.0
Non-member with education
1
6.7
Not a teaching hospital
8 53.3
Ownership
Government, non-federal
2 13.3
Non-government, not-for-profit
9 60.0
Investor-owned/for profit
2 13.3
Government, Federal
1
6.7
Location
East
4 26.7
Midwest
2 13.3
South/Southeast
7 46.7
Other countries
2 13.3
Fifteen of 26 participating sites gathered Cesarean section
data. Table 1 shows the demographic characteristics of the
participating hospitals. The average hospital size was 537
beds (SD=262) with only one considered small and most
considered medium to large. Participating hospitals tended
to be larger than the mean for the USA hospital population
[14]. They also provided more tertiary care; approximately
half the hospitals had neonatal intensive care units. The
hospitals were located predominantly in the Eastern and
Southern USA. Two of the hospitals were located overseas.
Overall system rates
Each system used a different approach to define Cesarean
section rates (see Table 2). Systems B, C, D specified the
numerator – Cesarean sections – using ICD-9-CM procedure
codes, while system A used Diagnostic Related Group (DRG)
categories 370 or 371 only. There was greater variety in the
specification of the denominator. System A used the DRGs
285
S. B. Kritchevsky et al.
Table 2 Numerator and denominator specifications for four Cesarean section rate indicators
Performance
measurement
Patient population
system
Numerator
(denominator)
............................................................................................................................................................................................
A
DRG 370, 371
DRG 370–375
B
Procedure codes:
Procedure codes: 72.0, 72.1, 72.21, 72.29, 72.31, 72.39,
74.0, 74.1, 74.2,
72.4, 72.51, 72.53, 72.54, 72.6, 72.71, 72.79, 72.8, 72.9;
74.4, 74.99
73.22, 73.51, 73.59, 73.6; 74.0, 74.1, 74.4, 74.99
C
Procedure codes: 74.0,
Diagnosis codes: V27.0–V27.9
74.1, 74.2, 74.4, 74.99
D
Procedure codes:
Diagnosis codes: 640.81–669.92 or V27.0–V27.9
74.0, 74.1, 74.2,
74.4, 74.99
Table 3 Mean Cesarean section rates by performance measurement system
Performance
Unadjusted
Adjusted
measurement
overall rates
overall rates
system
Numerator
Denominator
Mean (range)
Mean (range)
.................................................................................................................................................................
A
B
C
D
790
789
455
804
3392
3372
2181
3436
23.3
23.4
20.9
23.4
(9.6–36.6)
(9.6–36.6)
(9.6–29.6)
(9.6–33.7)
Not applicable
25.0 (18.3–35.3)
Not applicable
21.4 (18.6–38.5)
Table 4 Comparison of unadjusted hospital Cesarean section rates and rankings (in
parentheses) as calculated by four different Cesarean section indicators
Hospital
System A
System B
System C
System D
MPD1
............................................................................................................................................................
16.1 (3)
15.0
103
15.1 (3)
14.0 (2)
—2
105
19.1 (5)
20.1 (6)
20.0 (4)
20.0 (5)
5.2
106
14.9 (2)
14.9 (3)
14.2 (3)
14.2 (2)
4.9
107
32.5 (13)
33.3 (13)
—
33.2 (14)
2.5
108
18.7 (4)
18.1 (4)
12.7 (2)
18.2 (4)
47.2
109
33.7 (14)
33.7 (14)
—
33.7 (15)
0
110
29.7 (12)
31.2 (12)
29.6 (10)
31.0 (12)
5.4
111
23.0 (6)
26.5 (9)
—
26.4 (9)
15.2
115
23.0 (7)
19.4 (5)
23.4 (7)
23.5 (8)
21.1
116
23.1 (8)
23.4 (8)
23.2 (6)
23.1 (7)
1.3
117
25.1 (9)
21.8 (7)
22.2 (5)
21.6 (6)
16.2
119
26.4 (10)
27.2 (10)
27.2 (8)
26.6 (10)
3.0
120
28.3 (11)
28.1 (11)
28.9 (9)
28.1 (11)
2.8
123
9.6 (1)
9.6 (1)
9.6 (1)
9.6 (1)
0
126
36.6 (15)
36.6 (15)
—
31.7 (13)
15.5
Overall rate
23.3
23.4
20.9
23.4
12.0
1
Maximum percentage difference was calculated as the highest rate minus the lowest rate divided
by the lowest rate for each hospital.
2
Rate not calculated because there were fewer than 30 cases in the denominator.
C’s ranks were identical to those derived from System D
(unadjusted), except that hospitals 106 and 108 were reversed.
In general, there was a fair amount of consistency in relative
ranking of hospitals across indicator systems. Excluding
286
System C, the maximum number of differences in ranks
between systems was three (hospitals 111, 115 and 117) and
five hospitals were ranked identically across the three systems.
Figure 1 shows the comparison of the unadjusted and risk-
Cesarean section rate variation
adjustment between the two systems. For example, the hospitals that experienced the biggest change in rank due to risk
adjustment in System B were not affected much by risk
adjustment in System D. Conversely, the hospital with the
largest change in rank in System D was unaffected by
adjustment in System B.
Outlier status
Table 5 shows which hospitals were identified as low or high
outliers on the calculated rates. For two of the four systems
(A and B), there was very good consistency in determining
outlier status using the unadjusted data. System C flagged
two additional hospitals as outliers due to the lower overall
Cesarean section rate calculated. System D failed to flag two
hospitals that were flagged by systems A and B, and also
flagged one of the additional hospitals flagged by system C
(hospital 120). More than one-half of the hospitals were
flagged by at least one system when using unadjusted data.
Using adjusted data, System B identified two high outliers
and three low outliers. Four of these had similar status using
unadjusted data, but one hospital (105) was identified as a
low outlier only after risk adjustment. The risk-adjusted data
for system D flagged four high outliers and no low outliers.
Three of the four high outliers had been flagged in the
unadjusted data as well. Using risk-adjusted data, Systems B
and D identified only two hospitals in common out of the
seven flagged by either system.
Figure 1 A comparison of the effect of risk adjustment on
reported Cesarean section rates between two performance
measurement systems. Solid lines connect unadjusted to
adjusted rates for a hospital. Dashed lines connect the unadjusted rates between the systems.
adjusted rates for the two systems, B and D, that provided
risk-adjusted Cesarean section rates. Within systems the
adjusted rates were moderately correlated with the unadjusted
rates (rs=0.69 and 0.65 for Systems B and D, respectively).
In both systems, more than 25% of the hospitals changed
at least 4 ranks following adjustment. For System B, two
hospitals changed 7 ranks and one changed 6. For System
D, risk adjustment caused one hospital to change 9 ranks
and another to change 7 ranks.
Because adjusted rates are calculated relative to the average
Cesarean section rate for all the hospitals submitting data to
that system and not to the hospitals in this study, the absolute
values of the risk-adjusted rates cannot be strictly compared
between systems. However, given the similarities in risk
adjustment methodologies used by Systems B and D, it is
interesting to compare the rankings after risk adjustment
between the two systems. Overall, the adjusted rates from
the two systems were moderately correlated (rs=0.61), but
there was more disparity in relative rankings after risk adjustment than before. Eight hospitals differed by 1 rank or
less, five hospitals differed by 3–5 ranks and two hospitals
differed by 8 or more ranks after risk adjustment. Important
inconsistencies were noted in the relative effect of risk
Discussion
Our study calculated Cesarean section rates for 15 hospitals.
We processed the same data through the numerator and
denominator specifications used by five different performance
measurement systems. The differences in specifications led
to differences in rates of up to 47.2%. Risk adjustment used
by two of the five systems led to larger differences in both
rates and relative rankings. Risk adjustment did not affect all
hospitals in the same direction or to the same degree. The
overall study Cesarean section rate varied between 21.0%
and 25.0%, depending on definition. These rates are slightly
higher than the national statistics. The Centers for Disease
Control and Prevention report the national rate for 1995 to
be 20.8 per 100 deliveries from the 1995 National Hospital
Discharge Survey [16]. The national rate is based on primary
medical record abstraction of a nationally representative
sample of 29 000 inpatients discharged from 466 participating
hospitals.
The differences in unadjusted rates can be attributed to
differences in numerator and denominator definitions and to
differences in coding practices by the individual hospitals.
According to the U.S. Department of Health and Humans
Services, guidelines for coding and reporting using the International Classification of Diseases, 9th Revision, Clinical Modification
(ICD-9-CM), codes in the chapter ‘Complications of Pregnancy, Childbirth and the Puerperium’ (630–677) are required
for every delivery. A V code for the outcome, V27.0–V27.9,
287
S. B. Kritchevsky et al.
Table 5 Comparison of outlier status across Cesarean section indicators
Before risk adjustment
After risk adjustment
............................................................................................................. ......................................................
Hospital
System A
System B
System C
System D
System B
System D
.......................................................................................................................................................................................................
103
L
L
N
I
I
I
105
I
I
I
I
L
I
106
L
L
L
L
I
I
107
H
H
N
H
I
I
108
L
L
L
I
L
I
109
H
H
N
H
I
H
110
H
H
H
H
H
H
111
I
I
N
I
I
I
115
I
I
I
I
I
I
116
I
I
I
I
I
I
117
I
I
I
I
I
I
119
I
I
H
I
I
I
120
I
I
H
H
I
H
123
L
L
L
L
L
I
126
H
H
N
H
H
H
H, High outlier; L, low outlier; I, inlier; N, insufficient data.
should also be included on every maternal record when a
delivery has occurred [15]. Our findings, albeit from a small
group of hospitals, suggest that this coding convention is
not universally followed.
The disparity between adjusted and unadjusted rates and
resultant rankings was expected and has been observed by
others looking at a variety of patient outcomes including
Cesarean section rates [10]. Iezzoni et al. demonstrated that
the application of differing algorithms for risk adjustment
also effects rankings [17–19]. Hartz et al. [20] found that
inaccurate coding practices can artificially raise risk-adjusted
mortality rates. Romano and Mark found that errors in the
fields of admission source and type biased the estimation of
risk adjusted mortality more than underreporting of comorbidities [21]. Coding issues are a particular problem for
public hospitals whose reimbursement may be minimally
affected by coding practices.
The fact that variation in indicator specifications leads to
differences in calculated rates suggests that standardization
of definitions for commonly used performance indicators
should be a high priority. To our knowledge, no national
standardized specifications for calculating a Cesarean section
rate exist. The National Center for Health Statistics reports
their methodology for calculating Cesarean section rates as
including procedure codes 74.0–74.2, 74.4 and 74.99 in the
numerator and the denominator as V27.0–V27.9, though the
codes and algorithm have not been widely distributed to date
[22].
The need for standardization is even more urgent when
one realizes that current clinical performance comparisons
are based on any number of definitions of Cesarean section
rates. For example, in at least one comparative indicator
288
system, the specifications for the Cesarean section rate indicator provide different options for identifying the denominator population, thus leaving the approach to the
discretion of health care organizations. Since the choice of
denominator can lead to noticeable differences in reported
rates, consumers and other users of this information cannot
be assured of fair, meaningful comparisons.
There remains controversy regarding the scientific validity
and usefulness of report cards in general [23–26]. One popular
feature of many report cards is the identification of statistical
outliers. The theory behind outliers is that the outlying rates
are unlikely to be due to random variation, and therefore
reflect some real difference in practice of the outlying organization compared with other institutions. By implication,
hospitals and the physicians practicing at them are held
accountable. Risk adjustment is intended to make this process
‘fairer’ by allowing for differences in patient populations that
both determine the Cesarean section risk and cannot be
controlled by the organization. In the current study, risk
adjustment led to the identification of fewer outliers than
did unadjusted rates. However, there were inconsistencies
between the hospitals flagged by the two risk-adjusted indicators.
The findings of this project also demonstrate that the
decision to compare performance based on outliers as opposed to rankings will affect the judgments about hospital
performance. The use of statistically significant outlier status
is affected both by sample size and by effect size (e.g. how
different the rate was from the overall mean). Rankings, on
the other hand, may overestimate the magnitude of differences
between organizations. For example, hospitals with similar
rates (e.g. 20.0, 20.3, 20.4 and 20.5) may be ranked 6, 7, 8
Cesarean section rate variation
Figure 2 Examples of factors that affect indicator rates.
and 9 with a difference in rank of 3 but a difference in rate
of 0.5.
It is important to remember that this study was not
designed to judge the participating performance measurement
systems, their indicators or the hospitals participating in the
study. Since there are no consensus-based external criteria
for the validity of indicators or performance measurement
systems, one cannot conclude that one indicator is superior
to another (except perhaps, to the extent that one is more
in concert with coding guidelines) or that certain hospitals
were good or poor performers. The data further suggest that
using an ‘outlier’ criteria based on unadjusted data may be
of little use in identifying improvement opportunities. Thus,
the findings support the need for additional research and
consensus on criteria for establishing the validity of indicators
in order to judge which measures are best. For example, a
study could be designed to test which indicator specifications
best identify organizations or individual patient records in
which the care needs to be improved. This may be where
the benefits of risk adjustment on patient factors are most
apparent, e.g. not flagging cases which received appropriate
care or those in which practitioners and organizations could
not have influenced the mode of delivery.
Strengths of this demonstration project include the involvement of hospital epidemiologists in the data collection
process and the variety of hospital sizes and locations included. This study is unique in being able to disentangle
differences due to indicator definition versus those due to
risk adjustment. A limitation of the project includes the fact
that the data came from a relatively small number of hospitals.
Differences among hospitals in data collection procedures
may have affected indicator rates. Future studies should
evaluate the additional contribution to the indicator variation
introduced by data collection practices.
This study focused on variation in indicator specifications,
but there are many other factors that influence a given
indicator rate. These include organization-related factors (e.g.
equipment, systems of care, practitioner skill, completeness
and accuracy of data collection) and external factors such as
severity of illness and random variation (Figure 2). For an
indicator to be a useful guide in quality improvement activities,
it must reliably index organizational factors, i.e. those that
can be controlled by the organizations being compared.
Additional research needs to be done to examine both the
organizational factors and external factors that influence
Cesarean section rates.
In summary, there is a need for standardization of the
specifications for calculating Cesarean section rates, particularly when these rates are used for comparative purposes.
It is essential to define carefully how to identify cases for
the numerator and denominator and whether or not risk
adjustment is required. If risk adjustment is required, it will
be important to establish which factors are appropriate to
include in risk adjustment models. Despite the inconsistencies
between measurement systems demonstrated here, the indicators as currently defined may well be useful to organizations for monitoring and improving their own
performance over time [27].
289
S. B. Kritchevsky et al.
Our results suggest that health care organizations should
carefully consider indicator-related factors when selecting a
performance measurement system and when comparing
results across organizations. The indicator specifications for
something as simple as a Cesarean section rate need to be
articulated and carefully implemented before the results can
be used appropriately for making comparative judgments of
health care provider performance. Given the widespread
demand for external release of outcome data from hospitals
by insurers, employers, legislators, consumer advocates, regulatory agencies, accrediting bodies and many others, there is
a serious need for further education on factors that influence
and potentially confound the reported rates.
11. Kritchevsky SB, Simmons BP, Braun BI. The project to monitor
indicators: a collaborative effort between the Joint Commission
on Accreditation of Healthcare Organizations and the Society
for Healthcare Epidemiology of America. Infect Control Hosp
Epidemiol 1995; 16: 33–35.
Acknowledgments
15. Illustrated ICD-9-CM Code Book, Volumes 1,2,3, 1998. Reston VA:
St. Anthony Publishing, Inc., 1997.
This study received financial support from the Methodist
Hospitals Foundation, Memphis, TN, USA. The authors
gratefully acknowledge PMI executive committee members
Jerod Loeb PhD, Alfred Buck MD, Paul Schyve MD and
Ronald Shorr MD for their advice during the study and for
manuscript review. Thanks also to Mary Ellen Baruch for
assistance with algorithm programming and data analysis.
16. Curtin SC, Kozak LJ. Cesarean delivery rates in 1995 continue
to decline in the United States. Birth 1997; 24: 194–196.
References
1. Newman L, Simon K, Beck D et al. (eds) 1998 Medical Outcomes
and Guidelines Sourcebook. New York: Faulkner and Gray, Inc.,
1997.
2. Reichard J, Simon K, Youngs MT, Vibbert S. (eds) 1997 Comparative Performance Data Sourcebook. New York: Faulkner and
Gray, Inc.,1996.
3. Joint Commission on Accreditation of Healthcare Organizations. National Library of Health Care Indicators: Network
Edition. Oakbrook Terrace IL: JCAHO, 1997.
4. Eskew PN, Saywell RM, Zollinger TW et al. Trends in the
frequency of Cesarean delivery: A 21-year experience, 1970–
1990. J Reprod Med 1994; 39: 809–817.
5. Goyert GL, Bottoms SF, Treadwell MC, Nehra PC. The physician factor in Cesarean birth rates. N Engl J Med 1989; 320:
706–709.
6. Gambay M, Wolfe SM. Unnecessary Cesarean sections: curing
a national epidemic. Washington, DC: Public Citizen Health
Research Group, 1994.
7. New England HEDIS Coalition. New England HEDIS Coalition
1993 Baseline Performance Profile. Boston: New England HEDIS
Coalition; 1994.
8. Keeler EB, Park RE, Bell RM et al. Adjusting Cesarean delivery
rates for case mix. Health Serv Res 1997; 32: 511–528.
9. Burns LR, Geller SE, Wholey DR. The Effect of physician
factors on the Cesarean section decision. Med Care 1995; 33:
365–382.
10. Aron DC, Harper DL, Shepardson LB, Rosenthal GE. Impact
of risk-adjusting cesarean delivery rates when reporting hospital
performance. J Am Med Assoc 1998; 279: 1968–1972.
290
12. Simmons BP, Schyve PM. SHEA and JCAHO: Partners in
science. Infect Control Hosp Epidemiol 1995; 15: 5–6.
13. Sempos CT, Kahn HA. Statistical Methods in Epidemiology. New
York: Oxford University Press, 1989.
14. American Hospital Association. Hospital Statistics: Emerging Trends
in Hospitals, 1995–96 Edition. Chicago: American Hospital Association, 1995.
17. Iezzoni LI, Ash AS, Schwartz M et al. Judging hospitals by
severity-adjusted mortality rates: the influence of the severityadjustment method Am J Public Health 1996; 86: 1379–1387.
18. Iezzoni LI, Schwartz M, Ash AS et al. Using severity-adjusted
stroke mortality rates to judge hospitals. Int J Qual Health Care
1995; 7: 81–94.
19. Landon B, Iezzoni LI, Ash AS et al. Judging hospitals by
severity-adjusted mortality rates: the case of CABG surgery.
Inquiry 1996; 33: 155–166.
20. Hartz AJ, Kuhn EM, Yuan Z et al. Adequacy of National
mortality data for internal hospital use: a case study. Int J Qual
Health Care 1995; 7: 109–118.
21. Romano PS, Mark DH. Bias in coding of hospital discharge
data and its implications for quality assessment. Med Care 1994;
32: 81–90.
22. Gregory KD, Curtin SC, Taffel SM, Notzon FC. Changes in
indication for Cesarean delivery, 1985 and 1994. Am J Public
Health 1998; 9: 1384–1387.
23. Health Care Reform, Report Cards Are Useful But Significant Issues
Need to Be Addressed. Washington DC: United States General
Accounting Office, 1994. Publication GAO/HEHS-94–219.
24. Epstein A. Performance reports in quality- prototypes, problems
and prospects. N Engl J Med 1995; 333: 57–61.
25. Iezzoni LI. The risks of risk adjustment. J Am Med Assoc 1997;
278: 1600–1607.
26. Gross, PA, Cataruozolo P, DeMauro P et al. Severity adjustment
for length of stay: is it always necessary? Clin Perform Qual Health
Care 1997; 5: 169–172.
27. Longo DR, Land G, Schramm W et al. Consumer reports in
health care, do they make a difference? J Am Med Assoc 1997;
278: 1579–1584.
Cesarean section rate variation
Appendix
Society for Healthcare Epidemiology of America (SHEA)
member epidemiologists from hospitals participating in the
PMI Study Group included: Brian Cooper MD, Maureen
Theroux EdD, RN, James Steinberg MD, Louis Katz MD,
Sharon Welbel MD, August Valenti MD, Mark Keroack MD,
Jo Wilson MD, Peter Gross MD, Isabel Guererro MD, Larry
Strausbaugh MD, James Bross MD, Bruce Ribner MD MPH,
J. John Weems Jr. MD, Richard Rose III, MD, John Adams
MD, Fred Barrett MD, William Scheckler MD, Michael Climo
MD, Kenji Kono MD, Ziad Memish MD and Z. Ahmed
Quraishi, PhD.
Accepted for publication 7 April 1999
291