American Journal of Epidemiology
ª The Author 2007. Published by the Johns Hopkins Bloomberg School of Public Health.
All rights reserved. For permissions, please e-mail: [email protected].
Vol. 166, No. 2
DOI: 10.1093/aje/kwm139
Advance Access publication June 7, 2007
Practice of Epidemiology
Validity of Preeclampsia-related Diagnoses Recorded in a National Hospital
Registry and in a Postpartum Interview of the Women
Åse K. Klemmensen1,2, Sjurdur F. Olsen1, Marie Louise Østerdal1, and Ann Tabor3
1
Maternal Nutrition Group, Danish Epidemiology Science Centre, Statens Serum Institut, Copenhagen, Denmark.
Department of Obstetrics and Gynecology, H:S Hvidovre Hospital, University of Copenhagen, Hvidovre, Denmark.
3
Ultrasound Clinic, Juliane Marie Center, H:S Rigshospitalet, Copenhagen, Denmark.
2
Received for publication June 2, 2005; accepted for publication June 8, 2006.
In a population-based sample, the authors examined the validity of preeclampsia and related diagnoses recorded in a mandatory Danish national hospital discharge registry and in a standardized telephone interview of
women who gave birth between 1998 and 2002. Using a ‘‘gold standard’’ for preeclampsia defined in accordance
with the guidelines from the American College of Obstetricians and Gynecologists, the authors reviewed hospital
charts of 3,039 women and found that 61of 88 preeclampsia cases (69.3%) and 24 of 55 cases of serious subtypes
of preeclampsia (43.6%) were recorded as such by the registry. A total of 21 of 2,951 women without preeclampsia
(0.71%) had a preeclampsia diagnosis in the registry. All registrations of serious subtypes of preeclampsia reflected true cases. The positive predictive value of a preeclampsia diagnosis in the registry was 74.4%. Including
interview data reduced the sample size to 2,307 women. In this sample, of 62 women with preeclampsia, 45
(72.6%) reported in the interview to have had preeclampsia. Of 2,245 women with no preeclampsia, 31 (1.4%)
reported to have had preeclampsia. The positive predictive value of the women’s own report on preeclampsia was
59.2%. The authors conclude, for the purpose of etiologic studies, that the registry had acceptable validity, whereas
the usefulness of self-reported information may be limited.
interviews; postpartum period; pre-eclampsia; registries; reproducibility of results; women
Abbreviation: ICD-10, International Classification of Diseases, Tenth Revision.
Editor’s note: An invited commentary on this article
appears on page 125, and the authors’ response is
published on page 128.
Identifying the causes and describing the consequences of
preeclampsia require large, prospectively designed studies.
When the number of study participants gets above a certain
limit, however, it may become an insurmountable task to
undertake detailed clinical chart reviews for all participants,
and other methods for case ascertainment may be required.
An option increasingly available in many countries is to
make linkage to national or private administrative registries
with information on diagnoses on individuals.
Very limited experience exists, however, regarding the
validity of preeclampsia-related diagnoses collected in registries (9). In the present paper, we examine the validity of
Preeclampsia remains a major problem in modern obstetric practice. It occurs in 2–6 percent of all pregnancies and is
responsible for a large proportion of perinatal and maternal
morbidity and mortality (1, 2). Preeclampsia has also been
associated with diseases appearing later in the woman’s or the
offspring’s life, such as cancer (3, 4), cardiovascular diseases (5–7), and the metabolic syndrome (8).
Correspondence to Åse K. Klemmensen, Department of Obstetrics and Gynecology, H:S Hvidovre Hospital, University of Copenhagen,
Kettegård Allé 30, DK-2650 Hvidovre, Denmark (e-mail: [email protected]).
117
Am J Epidemiol 2007;166:117–124
118 Klemmensen et al.
FIGURE 1. Criteria for various degrees of preeclampsia, according to the definitions of the American College of Obstetricians and Gynecologists.
such diagnoses in a mandatory national hospital discharge
registry, that is, the Danish National Patient Registry, which
for many years has recorded virtually all discharge diagnoses for hospitalizations in Denmark (10). Additionally, we
examined the validity of the women’s own report on whether
they had had preeclampsia or not. The latter information
was obtained in a standardized, computer-assisted telephone
interview undertaken 6 months after delivery as part of
a large national survey (11). An evaluation of the diagnoses
based on these two sources was performed against diagnoses based on a detailed review of all relevant clinical charts.
The study included 3,084 pregnancies, making it the largest
detailed validation study of registry-based and interviewbased recordings of preeclampsia of which we are aware.
MATERIALS AND METHODS
Sources used
The validation study comprised in principle all those
women in the Danish National Birth Cohort who gave birth
in the years 1998–2000 at three hospitals: Hvidovre University Hospital, Hilleroed Central Hospital, and the local hospital in Helsingoer. These three hospitals represented the
three categories of hospitals in Denmark with obstetric service. If analyses were made separately for each of the three
hospitals, they tended to give very similar estimates of sen-
sitivity, specificity, positive predictive value, and kappa
(data not shown).
The chart review. All pregnant women in Denmark receive a ‘‘personal record’’ to carry with them at all visits to
the midwife, general practitioner, or hospital. Blood pressure and urine dipstick (checking for proteinuria) measurements are noted at every contact; a copy of this personal
record is made at the maternity ward and kept in the hospital
archives together with documents related to the hospitalization. Registrations related to diagnoses of preeclampsia and
hypertension were performed by one of the authors (Å. K.
K.) and entered into a Microsoft Access database (Microsoft
Corporation, Redmond, Washington). Each day that
a woman had an abnormal blood pressure/urine dipstick
measurement in either her chart or personal record, all data
on blood pressure/urine test on that date were recorded, including the interval between recordings (figure 1). We refer
to this source as the chart review or the gold standard.
The National Patient Registry. The Danish National
Patient Registry introduced the International Classification of
Diseases, Tenth Revision (ICD-10), in 1994 (10). For
validation, we explored the following ICD-10 codes:
DO10–DO119 (covering hypertension before pregnancy);
DO12–DO122 (pregnancy-induced edema or proteinuria without hypertension); DO13–DO139 (pregnancy-induced hypertension without proteinuria); and DO14–DO159 (the range
of levels of preeclampsia) (Web table 1). (The information is
described in the first of two supplementary tables; each is
Am J Epidemiol 2007;166:117–124
Validation of Preeclampsia-related Diagnoses
referred to as ‘‘Web table’’ in the text and is posted on the
Journal’s website (http://aje.oxfordjournals.org/).). The
Danish National Patient Registry registrations are based
on patient contacts, which means that every time a patient
is admitted to a hospital as an inpatient or outpatient, registration of an ICD-10 diagnosis is made, although the same
diagnosis may have been registered earlier in the pregnancy.
We refer to this source as the registry.
The Danish National Birth Cohort. The Danish National
Birth Cohort started enrolment in 1996 and ended in October
2002, with 101,039 women included (11). The women were
recruited at the first antenatal visit with their general practitioner and were interviewed four times by telephone: twice
during and twice after pregnancy. Approximately 35 percent
of all pregnant women in Denmark were included. A total of
70,296 women (70 percent) participated in the first postpartum interview, where information on elevation of blood
pressure and preeclampsia was gathered. These questions
were phrased as follows: ‘‘Did you have elevated blood
pressure during pregnancy?’’; ‘‘Did you have preeclampsia
during pregnancy, including elevated blood pressure and
urine protein loss?’’; and ‘‘Did you take antihypertensive
medication during pregnancy?’’ We refer to this source as
the interview. Furthermore, information on parity, smoking,
body mass index, and height was available from the two
interviews during pregnancy.
Method of linkage between sources. All citizens in
Denmark receive a unique 10-digit personal identification
number. The personal identification number is used as
identification for the chart review, the registry, and the
interview.
The study sample
A file with the relevant personal identification numbers
was extracted from the Danish National Birth Cohort database on deliveries at the three hospitals for 3 consecutive
years giving 4,911 deliveries, corresponding to 2.46 percent
(4,911/199,479) of the total Danish deliveries during this
3-year period. At the university hospital, the patient files
(n ¼ 1,840) were kept as paper documents, whereas at the
other two hospitals, the files (n ¼ 3,071) were partly electronic (n ¼ 1,244). The electronic obstetric chart was introduced in October 1999 at Frederiksborg County, where
the two hospitals were located. We decided to use only the
electronic charts, as they contained the same information as
the paper documents. Our study group thus included 3,084
deliveries (figure 2).
Definition of gold standard
We compared the diagnostic criteria used in Denmark, in
the United Kingdom, and in the United States in a previous
paper (12). For the gold standard, we decided to use the
recommended guidelines from the American College of Obstetricians and Gynecologists (figure 1), which are almost
identical with the most recent Danish guidelines (13). These
criteria are not absolute for eclampsia and were reported
absent in 38 percent of the cases with eclampsia (14, 15).
Elevation of the blood pressure should return to normal
Am J Epidemiol 2007;166:117–124
119
levels before 8 weeks postpartum (16). These criteria are a
minimum, and many other symptoms or measurements can
be added (17).
Statistical methods
We initially examined the marginal distribution of the
diagnoses according to the registry, the interview, and the
chart review, respectively. We proceeded with 2 3 2 crosstables calculating specificity, sensitivity, and positive and
negative predictive values for each of the registry and the
interview, regarding the chart review as the gold standard.
Moreover, Cohen’s kappa (18) was estimated as a measure
of agreement between the chart review and the diagnosis obtained from the registry/interview. Moreover, we examined
whether the validity of preeclampsia diagnosis depended on
known risk factors, by stratifying the analyses upon these.
Finally, measures of association between preeclampsia diagnoses and known risk factors were assessed by means of
logistic regression.
RESULTS
Among the 3,084 women, files were available and had
information on the specific pregnancy in 3,039 (98.5 percent).
In this group, 2,347 (77.2 percent) participated in the
6-month postpartum interview.
According to the gold standard, hypertensive disorders of
pregnancy occurred in 5.86 percent, whereas the registry
had less (3.27 percent) and the interview substantially more
(12.01 percent) registrations (table 1). The gold standard
incidence for preeclampsia (2.90 percent) was similar to that
in the registry (2.72 percent), while the interview-based
estimate was higher (3.36 percent). The specific subgroups of
preeclampsia, not distinguishable in the interview, showed
different values for the incidences in the gold standard and
the registry, except for the most severe forms, such as
the hemolysis elevated liver enzymes and low platelet
count and eclampsia, which were nearly identical. The registry severely underestimated the incidence of gestational
hypertension (0.55 percent vs. gold standard 2.96 percent),
whereas the interview overestimated the incidence
(8.64 percent).
When ascertainment was based on the registry, the sensitivity for all types of preeclampsia (69.32 percent) was
moderate, whereas it was lower for serious preeclampsia
(43.64 percent) and hypertensive disorders of pregnancy
(48.88 percent) and much lower for gestational hypertension
(10.00 percent) (table 2). The positive predictive value
ranged from 56.25 to 100 percent and was highest for serious types of preeclampsia. Specificity was above 99.0
percent for all four diagnoses (100 percent for serious preeclampsia). For all diagnoses, kappa was moderate except
for gestational hypertension (kappa ¼ 0.16).
Implementation of data from the interview (table 3) reduced our sample by 23 percent as mentioned earlier. The
sensitivities related to preeclampsia were similar in the registry and the interview (70.97 percent vs. 72.58 percent), but
specificity was higher in the registry (table 3). Including
a question about medication in the interview substantially
120 Klemmensen et al.
FIGURE 2. Flowchart illustrating the women included in the subcohort with files to be reviewed, representing three types of Danish hospitals
during 3 consecutive years (1998–2000).
increased specificity (from 98.62 to 99.69 percent) but simultaneously reduced sensitivity drastically (from 72.58 to
27.42 percent). Regarding gestational hypertension, the sensitivity was 58.46 percent in the interview compared with
12.31 percent in the registry; on the other hand, the specificity in the interview was poor (92.64 percent) compared
with the registry (99.82 percent) (table 3).
For the preeclampsia diagnosis, we examined to what
extent the validity depended on the mother’s parity, height,
body mass index, smoking, and previous diseases (table 4);
each of these variables was dichotomized for the purpose of
the analyses. Validity figures were quite similar across
strata defined by these variables (table 4). We also examined
to what extent the estimate of association between
TABLE 1. Incidence of various types of hypertension related to pregnancy in an
extracted subcohort of 3,084 Danish women who delivered during a 3-year period
(1998–2000) and had files available for review*
Gold standard
No. of files
No hypertension in pregnancy
Proteinuria
Essential hypertension
Registry
No.
%
No.
3,039
98.5
3,084
2,861
94.1
2,983
2
0.07
%
1
100
96.7
Interview
No.
%
2,349
76.2
2,067
88.0
12.0
0.03
17
0.59
2
0.06
178
5.86
101
3.27
282
Gestational hypertension
90
2.96
17
0.55
203
8.64
All types of preeclampsia
88
2.90
84
2.72
79
3.36
Mild preeclampsia
33
1.09
59
1.91
Serious preeclampsia
55
1.81
25
0.81
45
1.48
16
0.52
9
0.30
8
0.26
1
0.03
1
0.03
45
1.46
0
0
735
3,084
100
3,084
Hypertensive disorders of pregnancy
Severe preeclampsia
HELLPy
Eclampsia
Missing files
Total
3,084
100
23.8
100
* Files included records from three registries: the ‘‘gold standard,’’ the Danish National Patient
Registry, and the Danish National Birth Cohort. The gold standard was obtained by systematic
registration from patient files of both cases and noncases.
y HELLP, hemolysis elevated liver enzymes and low platelet count.
Am J Epidemiol 2007;166:117–124
Validation of Preeclampsia-related Diagnoses
121
TABLE 2. Validity of hypertension-related diagnoses in the Danish National Patient Registry (n ¼ 3,039) during 1998–2000*
Sensitivity
(%)
Specificity
(%)
Positive predictive
value (%)
100
Gold standard (þ)
Gold standard (–)
Kappa
100
Registry (þ)
Registry (–)
Registry (þ)
Registry (–)
Serious preeclampsiay
43.64
0.60
24
31
0
2,984
All types of preeclampsiaz
69.32
99.29
74.39
0.71
61
27
21
2,930
Gestational hypertension§
10.00
99.76
56.25
0.16
9
81
7
2,942
Hypertensive disorders
of pregnancy{
48.88
99.62
88.78
0.61
87
91
11
2,850
* The last four columns show the absolute number of women with (þ) or without (–) a diagnosis according to the ‘‘gold standard’’ and the
registry, respectively.
y International Classification of Diseases, Tenth Revision (ICD-10), codes DO141, DO142, and DO150.
z ICD-10 codes DO140, DO141, DO142, DO149, and DO150.
§ ICD-10 code DO139.
{ ICD-10 codes DO139, DO140, DO141, DO142, DO149, and DO150.
preeclampsia and these variables, expressed as odds ratios,
depended on whether the preeclampsia diagnosis was based
on the gold standard, the registry, the interview, or a combination (table 5). For all other variables than smoking, the
estimates did not seem to depend much on the method of
ascertaining preeclampsia.
DISCUSSION
Our study addressed a number of different validity issues
pertaining to the use of information from a mandatory hospital discharge registry and from a postpartum interview to
ascertain cases with preeclampsia for research purposes.
The implications of our findings depend on the aim of the
research.
If the aim is to determine the disease incidence in a population, some misclassification may be acceptable. This is
particularly true if the number of cases wrongly classified to
be noncases is similar to the number of noncases wrongly
classified to be cases. We found that the incidences of preeclampsia estimated by the registry (2.72 percent) and the
interview (3.36 percent) were both close to that based on our
chart review (2.90 percent). Moreover, the registry-based
incidences of the more serious diagnoses, the hemolysis
elevated liver enzymes and low platelet count (0.26 percent)
TABLE 3. Validity of hypertension-related diagnoses in the Danish National Patient Registry and the first postpartum interview of the
Danish National Birth Cohort for women who gave birth during 1998–2000*
Gold standard (þ)
Gold standard (–)
Sensitivity
(%)
Specificity
(%)
Positive predictive
value (%)
Kappa
Registry
70.97
99.47
78.57
0.74
44
18
12
Interviewy
72.58
98.62
59.21
0.64
45
17
31
2,214
Interviewz
27.42
99.69
70.83
0.39
17
45
7
2,238
Source
Source (þ)
Source (–)
Source (þ)
Source (–)
All types of preeclampsia
(n ¼ 2,307)
2,233
Gestational hypertension
(n ¼ 2,308)
Registry
12.31
99.82
66.67
0.20
8
57
4
2,239
Interview§
58.46
92.64
18.72
0.25
38
27
165
2,078
Registry
47.24
99.59
86.96
0.60
60
67
9
2,176
Interview{
85.04
92.22
38.85
0.50
108
19
170
2,015
Interview#
22.83
99.68
80.56
0.34
29
98
7
2,178
Hypertensive disorders
of pregnancy (n ¼ 2,312)
* The last four columns show the absolute number of women with (þ) or without (–) a diagnosis, according to the ‘‘gold standard’’ and the
source in question, respectively. Included were subsets of women participating in relevant parts of the interview.
y ‘‘Yes’’ to the question, ‘‘Did you have preeclampsia in pregnancy?’’
z ‘‘Yes’’ to the question, ‘‘Did you have preeclampsia in pregnancy?’’ and ‘‘yes’’ to the question, ‘‘Did you, in your pregnancy, take medication
due to hypertension?’’
§ ‘‘Yes’’ to the question, ‘‘Did you have hypertension in pregnancy?’’ and ‘‘no’’ to the question, ‘‘Did you have preeclampsia in pregnancy?’’
{ ‘‘Yes’’ to the question, ‘‘Did you have hypertension in pregnancy?’’
# ‘‘Yes’’ to the question, ‘‘Did you, in your pregnancy, take medication due to hypertension?’’
Am J Epidemiol 2007;166:117–124
122 Klemmensen et al.
TABLE 4. Sensitivity, specificity, and kappa values for the preeclampsia diagnosis in strata defined by well-recognized determinants
of preeclampsia (parity, height, body mass index, smoking, or previous disease (hypertension, preeclampsia, or renal disease)),
among Danish women who gave birth during 1998–2000
No. of women
in strata
Sensitivity
(%)
Specificity
(%)
Positive predictive
value (%)
Kappa
Test for equal kappas
across strata (p value)
Nullipara
1,579
74.60
98.94
74.60
0.74
0.28
Multipara
1,457
56.00
99.65
73.68
0.63
Height of 1.68 m
1,500
73.68
99.04
66.67
0.69
Preeclampsia according to the
registry (n ¼ 3,039)*
Height of 1.69 m
1,532
66.00
99.53
82.50
0.73
Body mass index of <25 kg/m2
2,309
76.60
99.43
73.47
0.74
Body mass index of 25 kg/m2
650
53.33
98.87
69.57
0.59
Smoker
876
52.17
99.53
75.00
0.61
2,158
75.38
99.19
74.24
0.74
51
53.85
97.37
87.50
0.59
2,920
71.62
99.33
73.61
0.72
Nullipara
1,183
76.74
98.25
62.26
0.67
Multipara
1,123
63.16
99.00
52.17
0.56
Height of 1.68 m
1,157
78.57
98.23
52.38
0.62
Height of 1.69 m
1,146
67.65
99.01
67.65
0.67
1,742
70.00
98.89
52.50
0.59
Nonsmoker
Previous disease
No previous disease
0.68
0.10
0.19
0.35
Preeclampsia according to the
interview (n ¼ 2,309)y
Body mass index of <25 kg/m2
2
Body mass index of 25 kg/m
509
75.00
97.53
60.00
0.65
Smoker
649
60.00
99.05
66.67
0.62
1,656
78.57
98.45
56.90
0.65
42
75.00
91.18
66.67
0.63
2,216
71.70
98.71
57.58
0.63
Nonsmoker
Previous disease
No previous disease
0.30
0.61
0.59
0.79
0.99
* Results for diagnoses based solely on the Danish National Patient Registry.
y Results for diagnoses based solely on the postpartum interview.
and eclampsia (0.03 percent), were close to the incidences
based on the chart review. On the other hand, gestational
hypertension was severely underascertained in the registry
(0.55 percent vs. 2.96 percent according to the chart review), whereas the interview overestimated the incidence
(8.64 percent).
If the aim is to identify as many as possible of the cases
that occur in a population (e.g., in order to make a more
detailed investigation of all potential preeclampsia cases),
one may want to maximize sensitivity. The sensitivity of
the registry was 70.97 percent for preeclampsia, whereas
the interview had a sensitivity of 72.58 percent. It is
TABLE 5. Odds ratios for preeclampsia risk factors, with preeclampsia diagnosis based on the ‘‘gold
standard,’’ the National Patient Registry, the postpartum interview, or a combination of the registry and the
interview (n ¼ 2,200), among Danish women who gave birth during 1998–2000
Source of preeclampsia diagnosis
Gold standard
Odds
ratio
Registry
95% confidence
interval
Odds
ratio
95% confidence
interval
Interview
Odds
ratio
95% confidence
interval
Multipara vs. nullipara
0.38
0.20, 0.70
0.27
0.14, 0.54
0.35
0.20, 0.60
Height of 1.68 vs. 1.69 m
0.74
0.42, 1.30
1.18
0.66, 2.12
1.30
0.79, 2.13
Body mass index of
25 vs. <25 kg/m2
2.75
1.56, 4.85
1.82
0.98, 3.36
2.61
1.58, 4.30
Smoker vs. nonsmoker
1.55
0.86, 2.79
0.88
0.45, 1.71
0.86
0.49, 1.52
4.58, 31.0
9.57
3.30, 27.7
Previous disease—yes vs. no
11.9
10.0
4.13, 24.3
Am J Epidemiol 2007;166:117–124
Validation of Preeclampsia-related Diagnoses
noteworthy that, with a strategy based on both sources,
requesting that the diagnosis should be present in at least
one of the two, the sensitivity did not get above 86 percent,
leaving 14 percent of preeclampsia cases unidentified
(Web table 2).
We are aware that use of antihypertensive medication is
different in North America compared with Europe. Our data
on self-reported use of antihypertensive medication may
therefore not be directly applicable to a US population.
If the aim is to investigate disease etiologies (be they the
consequences of particular exposures for preeclampsia risk
or the consequences of preeclampsia for diseases later in life
in the mother (5, 6, 8, 19) or offspring (3, 20)), the situation
gets more complex, as both sensitivity and specificity of the
diagnosis may be critical. A low sensitivity will reduce the
number of detected cases, which will lead to a reduced statistical power in the analysis. On the other hand, a low specificity will attenuate the observed odds ratio, compared with
the underlying true odds ratio. In addition to the problem of
a biased estimate, however, attenuation in the observed odds
ratio will also in itself (everything else being equal) lead to
a reduced statistical power to detect any true association,
which may exist in our population. Numerical examples to
illustrate these points are available from the authors.
An important validity concern in any etiologic analysis is
whether the misclassification is differential (whether the degree of misclassification in the diagnosis depends on the
exposure status). Whereas nondifferential misclassification
can lead only to attenuation of the odds ratio, differential
misclassification can potentially create biases in any direction (it can both exaggerate and attenuate the observed odds
ratio). Differential misclassification is therefore generally
less desirable. For parity, prepregnant weight, and previous
diseases, no conspicuous differences were detected across
strata in measures of positive predictive value, sensitivity,
specificity, and kappa.
We also examined to what extent the observed associations between preeclampsia and known risk factors depended on the sources of the diagnosis. Again, for parity,
prepregnant weight, and previous disease, it was reassuring
that the odds ratios with preeclampsia were generally similar, and in several cases remarkably similar, irrespective of
whether the diagnosis was based on the chart review, the
registry, or the interview. Smoking behaved differently. Very
unexpectedly, and contrary to expectations from earlier
studies (21–23), the gold standard-based estimate tended
to exhibit an elevated risk of preeclampsia among smokers.
However, the registry- and interview-based estimates exhibited the expected (albeit insignificantly) reduced risk
among smokers. We have no other explanation for this
unexpected finding, than that it may be due to random
fluctuations.
Our study had several strengths. To our knowledge, it is
the largest of its kind concerning preeclampsia and related
diagnoses. According to the clinical chart review, the incidence of preeclampsia was 2.90 percent in our study population, which is very similar to a previously published
incidence (2.7 percent) for a Nordic population (24). This
suggests that our study sample, when it comes to preeclampsia incidence, is representative for such populations. Our
Am J Epidemiol 2007;166:117–124
123
detailed chart review was, unlike some other studies (9),
not restricted to only those pregnancies that had been identified as recognized cases according to the registry. This
enabled us to estimate specificity and sensitivity, in addition
to positive predictive value. Moreover, from the interviews,
we had information on the woman’s smoking, body mass
index, and height, enabling us to consider such factors in our
analyses.
Few earlier studies have evaluated the validity of registrybased preeclampsia diagnoses. Geller et al. (9) found a
positive predictive value of 54 percent for preeclampsia,
whereas it was 79 percent in our registry. A comparison of
the Danish National Patient Registry and the Danish Medical Birth Registry (since 1994 an integrated part of the
Danish National Patient Registry) was performed for
1,921 births (25). These were restricted to preterm births
and therefore were not representative of the general population. Among these, 100 medical records were reviewed
with respect to the validity of hypertensive disorders of
pregnancy. A sensitivity of 75 percent and a kappa value
of 0.7 were reported (25), which were higher values than in
our data, possibly because their analysis was based on a preterm population.
The Danish National Patient Registry allows for registrations of diagnoses during multiple hospitalizations throughout pregnancy. This may be unusual for most comparable
registries in other countries, which rely more heavily on the
hospital admission related to delivery. When we restricted
our analysis to take into account only the preeclampsia diagnosis recorded at the delivery hospitalization, we reduced
the total number of registered cases by 13, from 82 to 69
cases. This led only to a slight decline in sensitivity and
kappa and even to a slight increase in specificity and positive
predictive value (sensitivity, specificity, positive predictive
value, and kappa changed from 69.32 percent, 99.29 percent,
74.39 percent, and 0.71 (table 2) to 60.23 percent,
99.46 percent, 76.81 percent, and 0.67, respectively).
Despite our finding of a relatively high diagnostic validity
of the Danish National Patient Registry, we believe that
there is scope for substantial improvement in the quality
of these registrations. Klemmensen et al. (12) recently reported a lack of consistency in guidelines regarding preeclampsia and difficulties in handling the criteria among
the hospital personnel, who are responsible for the reporting
to the registry. It is essential to have comprehensible national guidelines agreed upon by the obstetricians and their
adherence to them. National guidelines also need to be similar to the international ones, to enable comparison of studies across borders.
We know of no other study that has evaluated the validity
of women’s own reports (shortly after pregnancy) on
whether or not they had had preeclampsia or gestational
hypertension. Not surprisingly, the interview overestimated
incidences. Many of these women probably experienced one
or more occasions with elevated blood pressure or other
symptoms related to preeclampsia and, while they were under observation, got the impression of having preeclampsia
and reported it as such. However, including a question about
whether or not they received medication for the ailment increased specificity to acceptable levels (99.69 percent) but
124 Klemmensen et al.
led also to a drastic reduction in sensitivity (27.42 percent).
When preeclampsia is ascertained this way, the odds ratio is
likely to be estimated with limited room for bias but with
a drastically reduced power.
In conclusion, we found that the Danish National Patient
Registry estimated incidences of preeclampsia correctly;
this included other serious diseases, such as the hemolysis elevated liver enzymes and low platelet count and
eclampsia. On the other hand, we found that gestational
hypertension was severely underascertained by this source.
For etiologic studies, we found that specificity for preeclampsia was very high in the Danish National Patient
Registry.
As actual chart extraction is very time consuming, this
might be restricted to those cases where women in the
interview reported the disease to be present. This would
dramatically reduce the number of charts to be assessed
and, in our study, detect 70–85 percent of cases while
maintaining a high specificity. Whether this would also be
the case in other studies on self-reported diagnoses remains
to be shown.
ACKNOWLEDGMENTS
The study was supported by grants from the Faculty of
Health Sciences, University of Copenhagen; the Danish
Hospital Foundation for Medical Research in the regions
of Copenhagen, the Faeroe Islands, and Greenland; the
Copenhagen Medical Society; H:S Direktionens forskningspulje; and the Foundation of Desirée & Niels Yde.
Financial support for the Danish National Birth Cohort
was obtained from the March of Dimes Birth Defects
Foundation, Health Foundation, European Union (QLK12000-00083), Danish Medical Research Foundation, Heart
Foundation, Danish National Research Foundation, Pharmacy Foundation, Egmont Foundation, and Augustinus
Foundation.
The managerial team of the Danish National Birth
Cohort consisted of Jørn Olsen (Chair), Mads Melbye,
Anne Marie Nybo Andersen, Sjurdur F. Olsen, Thorkild I.
A. Sørensen, and Peter Aabye.
Conflict of interest: none declared.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
REFERENCES
1. Report of the National High Blood Pressure Education Program Working Group on High Blood Pressure in Pregnancy.
Am J Obstet Gynecol 2000;183(suppl):S1–22.
2. Roberts JM, Speer P. Antioxidant therapy to prevent preeclampsia. Semin Nephrol 2004;24:557–64.
3. Ekbom A, Hsieh CC, Lipworth L, et al. Intrauterine environment and breast cancer risk in women: a population-based
study. J Natl Cancer Inst 1997;89:71–6.
4. Vatten LJ, Romundstad PR, Trichopoulos D, et al. Preeclampsia in pregnancy and subsequent risk for breast cancer.
Br J Cancer 2002;87:971–3.
5. Wilson BJ, Watson MS, Prescott GJ, et al. Hypertensive
diseases of pregnancy and risk of hypertension and
22.
23.
24.
25.
stroke in later life: results from cohort study. BMJ
2003;326:845–51.
Wolf M, Hubel CA, Lam C, et al. Preeclampsia and future
cardiovascular disease: potential role of altered angiogenesis
and insulin resistance. J Clin Endocrinol Metab 2004;89:
6239–43.
Basso O, Wilcox AJ, Weinberg CR, et al. Height and risk of
severe pre-eclampsia. A study within the Danish National
Birth Cohort. Int J Epidemiol 2004;33:858–63.
Pouta A, Hartikainen AL, Sovio U, et al. Manifestations of
metabolic syndrome after hypertensive pregnancy. Hypertension 2004;43:825–31.
Geller SE, Ahmed S, Brown ML, et al. International classification of diseases–9th revision coding for preeclampsia:
how accurate is it? Am J Obstet Gynecol 2004;190:
1629–33.
Sorensen HT, Sabroe S, Olsen J. A framework for evaluation
of secondary data sources for epidemiological research. Int J
Epidemiol 1996;25:435–42.
Olsen J, Melbye M, Olsen SF, et al. The Danish National Birth
Cohort—its background, structure and aim. Scand J Public
Health 2001;29:300–7.
Klemmensen AK, Olsen SF, Wengel CM, et al. Diagnostic
criteria and reporting procedures for pre-eclampsia: a national
survey among obstetrical departments in Denmark. Eur J
Obstet Gynecol Reprod Biol 2005;123:41–5.
Johansen M, Kjaergaard N, Legarth J, et al. Hypertension in
pregnancy. In: Sandbjerg guidelines. (In Danish). Copenhagen, Denmark: Danish Society for Neuroscience, 2001.
Douglas KA, Redman CW. Eclampsia in the United Kingdom.
BMJ 1994;309:1395–400.
Sibai B, Dekker G, Kupferminc M. Pre-eclampsia. Lancet
2005;365:785–99.
ACOG practice bulletin. Diagnosis and management of
preeclampsia and eclampsia. Int J Gynaecol Obstet 2002;
77:67–75.
Roberts JM, Pearson GD, Cutler JA, et al. Summary of the
NHLBI Working Group on Research on Hypertension during
Pregnancy. Hypertens Pregnancy 2003;22:109–27.
Altman DG. Some common problems in medical research. In:
Practical statistics for medical research. London, United
Kingdom: Chapman & Hall/CRC, 1991:396–439.
Irgens HU, Reisaeter L, Irgens LM, et al. Long term mortality
of mothers and fathers after pre-eclampsia: population based
cohort study. BMJ 2001;323:1213–17.
Barker DJ, Gluckman PD, Godfrey KM, et al. Fetal nutrition
and cardiovascular disease in adult life. Lancet 1993;341:
938–41.
Cnattingius S, Mills JL, Yuen J, et al. The paradoxical effect of
smoking in preeclamptic pregnancies: smoking reduces the
incidence but increases the rates of perinatal mortality,
abruptio placentae, and intrauterine growth restriction. Am J
Obstet Gynecol 1997;177:156–61.
Newman MG, Lindsay MK, Graves W. Cigarette smoking and
pre-eclampsia: their association and effects on clinical outcomes. J Matern Fetal Med 2001;10:166–70.
Xiong X, Wang FL, Davidge ST, et al. Maternal smoking and
preeclampsia. J Reprod Med 2000;45:727–32.
Clausen T, Slott M, Solvoll K, et al. High intake of energy,
sucrose, and polyunsaturated fatty acids is associated with
increased risk of preeclampsia. Am J Obstet Gynecol 2001;
185:451–8.
Kristensen J, Langhoff-Roos J, Skovgaard LT, et al. Validation
of the Danish birth registration. J Clin Epidemiol 1996;49:
893–7.
Am J Epidemiol 2007;166:117–124