07_129_00 12/07/2001 1:02 pm Page 383
(Black plate)
Family Practice
© Oxford University Press 2001
Vol. 18, No. 4
Printed in Great Britain
New-onset palpitations in general practice:
assessing the discriminant value of items within
the clinical history
N Summerton, S Mann, A Rigby, S Petkar and J Dhawan
Summerton N, Mann S, Rigby A, Petkar S and Dhawan J. New-onset palpitations in general
practice: assessing the discriminant value of items within the clinical history. Family Practice
2001; 18: 383–392.
Background. Palpitations are non-specific, with less than half of patients experiencing palpitations having a cardiac arrhythmia. Currently it seems that there is little evidence available to
assist GPs in discriminating between patients complaining of palpitations who have significant
cardiac arrhythmias and those who do not.
Objectives. Our aim was to estimate discriminant functions for specific items of clinical information in relation to the categorization of a patient (aged over 18 years) with a symptom of
new-onset palpitations presenting to primary care.
Methods. A network of 62 GPs spread amongst 36 practices agreed to recruit patients with
new-onset palpitations over the course of a 9-month study period. Patients consenting to be
involved in the study were asked a number of questions, focusing particularly on the medical
history, and were requested to complete a Hospital Anxiety and Depression Scale. Each patient
was also provided with a RhythmCard cardiac event recorder for up to 2 weeks and was asked
to record their heart rhythm if they experienced palpitations. Odds ratios (adjusted for age and
sex) were used to compare the clinical information obtained from patients with the final diagnosis.
Results. Of the 139 patients with palpitations presenting to GPs, it would appear that males
[odds ratio = 2.1 (1.0–4.5)], those with regular palpitations [odds ratio = 2.5 (1.0–5.8)], those
experiencing palpitations at work [odds ratio = 3.0 (1.3–7.2)] and those experiencing palpitations
affected by sleeping (odds ratio = 3.3 (1.4–7.7)] were more likely to have a cardiac cause for their
palpitations. Similar findings were made in an analysis focusing solely on the 81 patients with
a RhythmCard result. Furthermore, amongst this group, it is interesting to note that patients
with regular palpitations were more than twice as likely to have a ‘significant’ cardiac arrhythmia
as a cause for their palpitations. There were suggestions of dose–response effects between the
rate of the palpitation, the duration of the palpitation and the likelihood of it being a ‘significant’
arrhythmia.
Conclusions. This study provides some information on the characteristics of patients reporting palpitations to GPs who may have ‘significant’ cardiac arrhythmias. Based on this work, we
believe that a larger community-based study would be worthwhile and would provide useful
and useable clinical discriminant information for GPs in the settings where they work and
amongst the types of patients they encounter.
Keywords. Arrhythmia, diagnosis, family practice, palpitations, medical history taking.
reported by as many as 16% of patients.1 It can be
estimated that a GP with an average list size of 2000
patients will see about six patients with new-onset
palpitations per year.2
In most cases, palpitations are not associated with
structural heart disease. They may be provoked by a host
of factors including exercise, emotional stress, fever,
caffeine, nicotine and alcohol. Situations in which awareness of the heartbeat may lead an individual to seek
Introduction
Palpitations are non-specific and represent one of the
most common symptoms in general medical settings,
Received 13 June 2000; Revised 28 November 2000; Accepted
12 March 2001.
Winterton Medical Practice, The Surgery, Manlake Avenue,
Winterton, Scunthorpe DN15 9TA, UK.
383
07_129_00 12/07/2001 1:02 pm Page 384
384
(Black plate)
Family Practice—an international journal
medical attention include hyperthyroidism, anaemia and
sick sinus syndrome. Other causes of palpitations are
panic attacks, depression and somatization disorders.3 In
a recent study of secondary care patients in New York
State, the aetiology of the palpitations was stated to be
cardiac in 43% and psychiatric in 31%.4
Currently, it seems that it is mainly clinical experience
combined with information derived from studies within
referred populations that guides GPs in determining
what action to take when confronted by a patient with
new-onset palpitations. Despite an extensive literature
search of MEDLINE (1966–1999) using free text and
MeSH terms combined with a hand search of key primary
care-orientated journals over the last 2 years, we have
been able to identify only one study that describes the
usefulness of either specific items or clusters of items
of clinical information in the evaluation of new-onset
palpitations within primary care.5 Unfortunately, in applying the results from this paper into routine general practice, difficulties are encountered in relation to the patient
inclusion criteria. The authors involved not only patients
consulting their GP with palpitations but also patients
with dyspnoea, faintness, angina, fatigue, collapse and
“other complaints . . . if the GP considered an arrhythmia
as a possible cause of the complaints”. In addition, patients
also had to present a diagnostic problem for the GP who
they were consulting. Concerns must also be expressed
about the validity of the gold standard (a usual ECG
recording) used by Zwietering et al. because of the paroxysmal nature of some arrhythmias. Even supplementing this by asking certain patients to return for a further
ECG recording during symptoms is clearly open to bias
due to time constraints affecting both the participating
GPs and the patients. All this detail is important because,
in making any decisions based on probabilities, especially
in low prevalence populations, it is essential to be certain
of the adequacy of the diagnostic knowledge base. If the
internal or external validity is dubious, the information
may become inapplicable in relation to the amount of
risk GPs are willing to accommodate.
In the current study, we have sought to restrict our
recruitment to patients clearly describing palpitations in
accordance with an explicit case definition. In addition,
we have elected to use a transtelephonic post-event
recorder (RhythmCard) as the gold standard. These
hand-held credit card-sized devices are given to patients
and are applied to the chest when symptoms occur. The
patient presses a button to record ~30 seconds of the
cardiac rhythm, which is stored in the memory of
the device. The recording is transmitted later over the
telephone for printing and interpretation.
Both the published evidence and unpublished information from the manufacturers indicate that the RhythmCard is a valid and reliable assessment of the likelihood
of the patient having a significant cardiac arrhythmia.6
According to Kinley et al., in a small randomized controlled
trial, cardiac event recorders were demonstrated to yield
more diagnoses than 48-h Holter monitoring in patients
with palpitations.7 If patients are provided with a
RhythmCard for 2 weeks, it will detect 94% of clinically
relevant arrhythmias; a further 2 weeks will pick up an
additional 3%.
Using the RhythmCard as the gold standard, the
purpose of the present study was to seek to estimate the
discriminant functions for specific items of clinical
information in relation to the categorization of a patient
(aged over 18 years) with a symptom of new-onset palpitations presenting to primary care. The classification
was dichotomous (cardiac versus non-cardiac) to reflect
the realities of general practice decision making. Furthermore, cardiac diagnoses were divided into ‘significant’
arrythmias and ‘insignificant’ arrhythmias.
Methods
As a result of a positive response to a postal invitation, a
network of 62 GPs spread amongst 36 practices within
Yorkshire, Lincolnshire, Leicestershire, Cumbria and
County Durham agreed to recruit patients with newonset palpitations over the course of a 9-month study
period. Palpitations were defined as one or more of the
following complaints: fast heartbeats, skipped heart beats,
irregular heart rate and heart fluttering, racing or pounding. Any patient remarking on one or more of these
complaints during the course of any consultation with a
GP participating in the study was eligible for inclusion.
In order to be included, the patient’s symptoms had to
have occurred within 3 months preceding the visit to the
GP. The following were ineligible: those under the age of
18 years, patients with fever .39°C, patients with aphasia
or dementia, and pregnant women. Patients who had a history of previous episodes of palpitations were also excluded.
Patients consenting to be involved in the study were
asked a number of questions, focusing particularly on
the medical history, and were requested to complete a
Hospital Anxiety and Depression Scale.8 These forms
were then sent to a research co-ordinating centre at
Winterton where it was arranged for a RhythmCard to
be despatched to every patient. In addition, the information from the forms was entered onto an SPSS computerized research database at Winterton.
Each patient was provided with a RhythmCard for up
to 2 weeks and asked to use it to record their heart
rhythm if they experienced palpitations. The RhythmCard
had a capacity to store up to three 30-seconds readings
and, at the end of the recording period, any recordings
were transmitted to CardioAnalytics Ltd. Here the recording was printed out and interpreted by a CardioAnalytics
technician. An independent and separate interpretation
was also provided by SP under the supervision of JD. The
results from the RhythmCards were then classified by
NS and SP as ‘normal‘, ‘significant cardiac arrhythmia’ or
‘insignificant cardiac arrhythmia’. A ‘significant cardiac
07_129_00 12/07/2001 1:02 pm Page 385
(Black plate)
385
New-onset palpitations in general practice
There were 58 patients with no RhythmCard result; 39
with a RhythmCard result had a normal (‘non-cardiac’
diagnosis). Of the 42 patients with a RhythmCardgenerated ‘cardiac’ diagnosis, 16 warranted no further
action (‘insignificant’ arrythmias) while 26 patients had
‘significant’ arrhythmias.
arrhythmia’ constituted one of the following: ventricular
tachycardia (including torsade de pointes), paroxysmal
supraventricular tachycardia, atrial fibrillation (including
paroxysmal atrial fibrillation), atrial flutter, atrial tachycardia, junctional tachycardia or ventricular ectopics occurring in salvoes, in couplets or having a multifocal origin.
Such diagnoses warranted further investigation. ‘Insignificant cardiac arrhythmias’ encompassed patients with nonpathological ventricular ectopics (e.g. sporadic, occasional
unifocal ectopics), sinus tachycardia and atrial ectopics.
Five randomly selected practices were approached on
completion of the study with a request to check their
computerized records in order to assess the degree of
selection bias in relation to the recruitment of patients
into the study.
The data were analysed by calculating odds ratios
using the GLIM statistical package,9 odds ratios being
used as an approximation to the true relative risk.10 The
95% confidence intervals were calculated according to
the methods outlined in Morris and Gardner.11
Three comparisons were made: (i) normal RhythmCard results/no RhythmCard results compared with
RhythmCard results indicating a ‘cardiac’ diagnosis; (ii)
normal RhythmCard results compared with RhythmCard
results indicating a ‘cardiac’ diagnosis; and (iii) normal/
‘insignificant’ arrhythmias on RhythmCards compared
with ‘significant’ arrhythmias on RhythmCards.
Demographic details: patients and practices
Amongst the 36 participating practices, the mean number of partners was 4.7 and the mean number of patients
9880. One-third of practices were predominantly rural,
one-third urban and the remainder mixed.
Seventy-nine per cent of the participating GPs were
male, 66% were members of the Royal College of
General Practitioners and the mean age was 43.5 years.
Selection bias
By checking against their computerized records, the
five randomly selected practices were able to provide
reassurance that selection bias did not occur in relation
to their recruitment.
Statistical analysis
The detailed comparisons are presented in Tables 1–8.
Odds ratios are adjusted for age and sex.
Discussion
Within this study, we have made a number of interesting
findings. Amongst all the 139 patients with palpitations
presenting to GPs, it would appear that males [odds
ratio = 2.1, 95% confidence interval (CI) 1.0–4.5],
those with regular palpitations (odds ratio = 2.5, 95%
CI 1.0–5.8), those experiencing palpitations at work
Results
Demographic details: patients
There were 139 patients of whom 93 (67%) were female:
the mean (median) age of all patients was 46.4 (45) years
with a range of 19–77 years.
Variable
TABLE 1
Personal characteristics
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
Age (years)
<39
40–59
60+
34
44
19
17
11
14
1.0
0.5 (0.2, 1.2)
1.5 (0.6, 3.6)
12
18
9
17
11
14
1.0
0.4 (0.2, 1.2)
1.1 (0.4, 3.4)
19
23
13
10
6
10
1.0
0.5 (0.2, 1.6)
1.5 (0.5, 4.5)
Sex
Female
Male
70
27
23
19
1.0
2.1 (1.0, 4.5)
27
12
23
19
1.0
1.9 (0.7, 5.0)
34
21
16
10
1.0
1.0 (0.4, 2.6)
HAD score
,9
10–19
20+
26
55
19
9
28
5
1.0
1.4 (0.6, 3.5)
0.6 (0.2, 2.1)
9
22
8
9
28
5
1.0
1.3 (0.4, 4.2)
0.5 (0.1, 2.2)
13
33
9
5
17
4
1.0
1.3 (0.4, 4.5)
1.1 (0.2, 5.6)
07_129_00 12/07/2001 1:02 pm Page 386
(Black plate)
386
Family Practice—an international journal
TABLE 2
Variable
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Palpitations: nature
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
Regular
palpitation
No
Yes
DK
37
39
21
10
28
4
1.0
2.5 (1.0, 5.8)
0.6 (0.2, 2.3)
14
15
10
10
28
4
1.0
2.6 (0.9, 7.2)
0.6 (0.1, 2.5)
18
25
12
6
18
2
1.0
2.2 (0.7, 6.7)
0.5 (0.08, 2.8)
Continuous
palpitation
No
Yes
DK
14
70
13
4
32
6
1.0
1.5 (0.5, 6.2)
1.7 (0.4, 7.5)
6
27
6
4
32
6
1.0
1.7 (0.4, 6.9)
1.3 (0.2, 7.7)
7
41
7
3
18
5
1.0
1.1 (0.2, 4.4)
1.8 (0.3, 11.0)
Rate
(taps/min)
,100
100–120
120+
DK
24
37
14
22
16
14
6
6
1.0
0.6 (0.3, 1.6)
0.4 (0.2, 2.5)
0.4 (0.1, 1.3)
11
12
5
11
16
14
6
6
1.0
0.8 (0.3, 2.3)
0.8 (0.2, 3.5)
0.4 (0.1, 1.3)
20
16
6
13
7
10
5
4
1.0
1.8 (0.4, 8.0)
2.5 (0.6, 11.1)
0.9 (0.2, 3.5)
Palpitation
experienced:
constantly
No
Yes
DK
83
13
1
37
4
1
1.0
0.8 (0.2, 2.6)
2.1 (0.1, 40.3)
36
3
0
37
4
1
1.0
1.2 (0.2, 7.6)
***
51
4
0
22
3
1
1.0
1.8 (0.4, 8.9)
***
Palpitation
experienced:
at work
No
Yes
DK
79
16
2
26
15
1
1.0
3.0 (1.3, 7.2)
1.6 (0.1, 20.8)
33
6
0
26
15
1
1.0
3.2 (1.1, 9.8)
***
42
13
0
17
8
1
1.0
1.6 (0.5, 4.8)
***
Palpitation
No
experienced:
Yes
during holidays DK
84
12
1
33
8
1
1.0
2.0 (0.7, 5.5)
2.4 (0.1, 49.6)
36
3
0
33
8
1
1.0
3.0 (0.7, 12.9)
***
47
8
0
22
3
1
1.0
0.8 (0.2, 3.5)
***
Palpitation
experienced:
lying in bed
No
Yes
DK
42
54
1
11
30
1
1.0
2.1 (0.9, 4.7)
3.5 (0.2, 69.1)
16
23
0
11
30
1
1.0
1.9 (0.7, 5.0)
***
18
37
0
9
16
1
1.0
0.9 (0.3, 2.3)
***
Palpitation
experienced:
at weekends
No
Yes
DK
78
18
1
30
11
1
1.0
1.7 (0.7, 4.1)
2.5 (0.1, 49.1)
32
7
0
30
11
1
1.0
1.6 (0.5, 4.9)
***
41
14
0
21
4
1
1.0
0.6 (0.2, 1.9)
***
31
31
33
2
9
11
20
2
1.0
1.1 (0.4, 3.1)
2.0 (0.8, 5.1)
3.4 (0.4, 28.5)
10
13
14
2
9
11
20
2
1.0
0.9 (0.3, 3.3)
1.6 (0.5, 5.2)
1.1 (0.1, 11.2)
14
17
21
2
5
7
13
2
1.0
1.1 (0.3, 4.4)
1.7 (0.5, 5.9)
0.9 (0.07, 11.1)
Duration
,60
(max seconds) 60–600
600+
DK
***Not possible to estimate the odds ratio due to zero in cells; DK = don’t know/no data; example interpretation: Males are 2.1 times as likely as
females to have an arrhythmia.
(odds ratio = 3.0, 95% CI 1.3–7.2) and those experiencing
palpitations affected by sleeping (odds ratio = 3.3, 95%
CI 1.4–7.7) were significantly more likely to have a
cardiac cause for their palpitations. Although many
of the other associations are non-significant in a strict
statistical sense, it would appear that treatment with
a β-blocker, diuretic, angiotensin-converting enzyme
(ACE) inhibitor or an anti-hypertensive was correlated
with an increased risk of cardiac arrhythmia. The use
of antidepressants/benzodiazepines was associated
with a decreased risk of cardiac arrhythmia. There also
appeared to be a dose–response effect with duration; the
longer the duration of the palpitation the more likely an
arrhythmia.
Unfortunately, a RhythmCard result was not available
for 58 patients, as these patients did not, apparently,
experience an arrhythmia during the time in which they
had the RhythmCard. The first-level analysis assumes
that these 58 patients’ palpitations were non-cardiac.
However, in order to avoid either verification bias or
misclassification bias, it is important to repeat the analysis only on those patients with definite gold standard
results. Amongst the 81 patients with a RhythmCard
result, 41 subsequently were classified as ‘cardiac’. In a
second-level analysis, odds ratios .1.9 were again identified amongst males, those with regular palpitations,
those experiencing palpitations at work and those experiencing palpitations affected by sleeping. However, in
07_129_00 12/07/2001 1:02 pm Page 387
(Black plate)
387
New-onset palpitations in general practice
Variable
TABLE 3
Palpitations: modifiers
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
Palpitation
affected by:
breathing
No
Yes
DK
65
31
1
35
7
0
1.0
0.4 (0.2, 1.1)
***
28
11
35
7
1.0
0.5 (0.2, 1.5)
41
14
22
4
1.0
0.5 (0.2, 1.9)
Palpitation
affected by:
exercise
No
Yes
DK
72
25
0
34
8
0
1.0
0.7 (0.3, 1.8)
***
27
12
34
8
1.0
0.5 (0.2, 1.6)
40
15
21
5
1.0
0.6 (0.2, 2.0)
Palpitation
affected by:
caffeine
No
Yes
DK
88
9
0
34
7
1
1.0
1.8 (0.6, 5.3)
***
35
4
0
34
7
1
1.0
1.6 (0.4, 6.0)
***
49
6
0
20
5
1
1.0
2.0 (0.6, 7.5)
***
Palpitation
affected by:
alcohol
No
Yes
DK
85
10
2
36
6
0
1.0
1.4 (0.5, 4.3)
***
36
2
1
36
6
0
1.0
2.9 (0.5, 15.3)
51
3
1
21
5
0
1.0
4.2 (0.9, 19.8)
***
Palpitation
affected by:
sleeping
No
Yes
DK
78
18
1
24
18
0
1.0
3.3 (1.4, 7.7)
***
29
10
24
18
1.0
2.5 (0.9, 6.9)
40
15
13
13
1.0
2.8 (1.0, 7.7)
***Not possible to estimate the odds ratio due to zero in cells; DK = don’t know/no data; example interpretation: Males are 2.1 times as likely as
females to have an arrhythmia.
view of the small numbers, the confidence intervals are
broad and only the ‘palpitations at work’ reached
significance at the 5% level.
For the GP, the bottom line is not whether the patient
has an arrhythmia but rather whether the arrhythmia
matters. Hence, a third-level analysis seeks to compare
patients with normal/‘insignificant’ cardiac arrhythmias
on a RhythmCard with those with ‘significant’ cardiac
arrhythmias requiring further investigation or treatment.
Amongst the 81 patients with a RhythmCard result, 26
were classified as having ‘significant’ cardiac arrhythmias.
Although only one of the numerical associations identified reached statistical significance (palpitations affected
by sleeping), it is interesting to note that patients with
regular palpitations were more than twice as likely to
have a ‘significant’ cardiac arrhythmia as a cause for their
palpitations. There were also suggestions of dose–
response effects between the rate of the palpitation, the
duration of the palpitation and the likelihood of it being
a ‘significant’ cardiac arrhythmia.
In considering the broader applicability of the results,
it is reassuring to note that the characteristics of the 36 participating practices were comparable with the national
picture in terms of numbers of partners, patient list size
and urban/rural classification.12 The 62 participating
GPs were also similar to their peers as judged by their
average age, but contained a higher proportion of men
and MRCGP holders than the national average.
Unfortunately, some concerns must be raised that the
patients recruited into this study were not typical of all
general practice patients presenting with palpitations.
GPs may selectively enlist those with intermediate
probabilities of illness or those without significant
co-morbidities, thereby altering the spectrum of patients
being examined. However, by checking against their computerized records, five practices were able to provide
reassurance that selection bias did not occur in relation
to their recruitment. In addition, there does seem to
be some consistency between our results and those of
others5 in terms of the proportion of palpitation patients
within primary care (one-third) exhibiting cardiac
arrhythmias. However, if a GP with an average list size
of 2000 patients would expect to see six patients with
palpitations per year, we should have expected the 62
GPs involved in the study over a 9-month period to have
recruited 279 patients and yet only 139 were entered into
the study (with 81 of these providing a RhythmCard
recording).
Based on our work, we now feel able to suggest a more
focused examination of some specific discriminators
within the history, i.e. patient sex, regularity of palpitations, rate and duration of palpitations, and whether
07_129_00 12/07/2001 1:02 pm Page 388
(Black plate)
388
Family Practice—an international journal
TABLE 4
Variable
Palpitations: associations
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
Associated
with:
flushes
No
Yes
DK
61
34
2
28
14
0
1.0
1.1 (0.5, 2.5)
***
23
16
28
14
1.0
0.9 (0.3, 2.3)
38
17
13
13
1.0
2.6 (0.9, 7.3)
Associated
with: sweats
(generalized)
No
Yes
DK
59
37
1
28
14
0
1.0
0.9 (0.4, 1.9)
***
23
16
38
14
1.0
0.8 (0.3, 2.0)
38
18
14
12
1.0
1.8 (0.7, 4.8)
Associated
with:
sweaty palms
No
Yes
DK
72
21
4
32
9
1
1.0
1.1 (0.5, 2.9)
0.8 (0.08, 7.3)
28
10
1
32
9
1
1.0
0.8 (0.3, 2.4)
1.1 (0.06, 19.3)
42
11
2
18
8
0
1.0
1.7 (0.6, 5.0)
***
Associated
with:
chest pain
No
Yes
DK
70
25
2
33
9
0
1.0
0.8 (0.3, 1.9)
***
26
12
1
33
9
0
1.0
0.6 (0.2, 1.6)
***
39
15
1
20
6
0
1.0
0.8 (0.3, 2.4)
***
Associated
with:
syncope
No
Yes
DK
83
10
4
36
5
1
1.0
1.3 (0.4, 4.2)
0.6 (0.06, 5.5)
34
4
1
36
5
1
1.0
1.3 (0.3, 5.6)
0.9 (0.05, 15.3)
50
4
1
20
5
1
1.0
3.4 (0.8, 14.4)
2.6 (0.3, 43.1)
Associated
with:
dizzy spells
No
Yes
DK
44
51
2
21
21
0
1.0
0.9 (0.4, 1.8)
***
18
20
1
21
21
0
1.0
0.9 (0.4, 2.2)
***
30
24
1
9
17
0
1.0
2.4 (0.9, 6.4)
***
Associated
with:
dyspnoea
No
Yes
DK
16
15
66
13
2
27
1.0
0.2 (0.03, 0.9)
0.5 (0.2, 1.3)
5
6
28
13
2
27
1.0
0.07 (0.01, 0.6)
0.4 (0.1, 1.2)
10
7
38
8
1
17
1.0
0.2 (0.01, 1.8)
0.6 (0.2, 1.7)
***Not possible to estimate the odds ratio due to zero in cells; DK = don’t know/no data; example interpretation: Males are 2.1 times as likely as
females to have an arrhythmia.
palpitations occur at work or were affected by sleeping.
However, we do not believe that our study is sufficiently
powerful or free of bias in order to encourage the use of
these discriminators within routine clinical practice at
present. A larger study would need to address selection
bias more formally and also to assess both the reliability13
and the validity of the suggested discriminators.
One approach to the problem of selection is to consent
a random proportion of patients to be re-interviewed by
phone in order to validate the eligibility criteria being
applied by the recruiting GPs. Another way is to contact
all patients directly: this was the method adopted by
Stoffers et al. when they sought to assess the diagnostic
value of signs and symptoms associated with peripheral
arterial occlusive disease seen in general practice. A list
was obtained of all the patients aged 40–75 years registered with 18 GPs in Limburg, The Netherlands. These
26 620 subjects then received a simple postal questionnaire enquiring about ‘leg complaints on walking’ and
those reporting in the affirmative were invited to attend
for assessment. However, in adopting such an approach,
care must be taken to enquire about whether such
patients do consult or would consider consulting their
GP about such symptoms as we clearly need to understand both the significance of community symptoms
as well as identifying those patients that may present
to the GP. Clearly some symptoms such as chest pain
have a greater ‘iatrotrophic stimulus’15 than, perhaps,
palpitations.
Although we have reasonable confidence in the sensitivity of the RhythmCard, it remains possible that some
short duration episodes of cardiac arrhythmia might
have been missed as it clearly takes time for an individual
to respond and to activate the recorder. Therefore, in
order to enhance the validity of any larger study, we
would propose that a proportion of patients could also
undergo continuous loop recording.16 Unfortunately, we
have been unable to identify any information on the
specificity of the RhythmCard; in particular the number
of false positives.
Finally, some statistical issues warrant further examination. In our analysis, we have made a large number
07_129_00 12/07/2001 1:02 pm Page 389
(Black plate)
389
New-onset palpitations in general practice
TABLE 5
Variable
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Previous
history of:
hypertension
Medical history
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
No
Yes
DK
80
15
2
36
6
0
1.0
0.6 (0.2, 2.0)
***
31
8
36
6
1.0
0.5 (0.2, 2.9)
45
10
22
4
1.0
0.7 (0.2, 2.9)
Previous history No
of: coronary
Yes
heart disease
DK
91
2
4
41
1
0
1.0
0.7 (0.06, 8.5)
***
37
1
1
41
1
0
1.0
0.8 (0.05, 14.8)
***
53
1
1
25
1
0
1.0
2.0 (0.1, 34.9)
***
Previous history No
of: valvular
Yes
heart disease
DK
90
2
5
42
0
0
1.0
***
***
37
2
42
0
1.0
53
2
26
0
1.0
Previous history No
of: congenital
Yes
heart disease
DK
88
4
5
42
0
0
1.0
***
***
37
2
42
0
1.0
53
26
1.0
***
2
0
***
Previous history No
of: pericarditis DK
92
5
42
0
1.0
***
37
2
42
0
1.0
***
53
2
26
0
1.0
***
Previous
No
history of:
DK
cardiomyopathy
90
41
41
1
1.0
***
36
3
41
1
1.0
0.2 (0.02, 2.7)
52
3
25
1
1.0
0.6 (0.06, 6.7)
Current
medical
problems:
thyroid disease
No
Yes
DK
86
6
5
41
1
0
1.0
0.4 (0.04, 3.2)
***
34
3
2
41
1
0
1.0
0.3 (0.03, 3.1)
***
49
4
2
26
0
0
1.0
***
***
Current
medical
problems:
asthma
No
Yes
DK
88
2
5
40
2
0
1.0
1.1 (0.2, 6.0)
***
34
3
2
40
2
0
1.0
0.6 (0.09, 3.6)
***
48
5
2
26
0
0
1.0
***
***
Current
medical
problems:
anaemia
No
Yes
DK
89
4
4
40
2
0
1.0
1.5 (0.2, 8.6)
***
35
2
2
40
2
0
1.0
1.1 (0.1, 8.5)
***
51
2
2
24
2
0
1.0
2.2 (0.3, 4.8)
***
Current
medical
problems:
anxiety
No
Yes
DK
60
32
5
27
14
1
1.0
1.0 (0.4, 2.2)
0.6 (0.06, 5.3)
24
13
2
27
14
1
1.0
1.1 (0.4, 2.8)
0.6 (0.05, 7.0)
34
19
2
17
8
1
1.0
0.8 (0.3, 2.3)
1.0 (0.4, 12.2)
Current
medical
problems:
depression
No
Yes
DK
81
11
5
37
4
1
1.0
0.8 (0.2, 2.8)
0.4 (0.05, 3.9)
33
5
1
37
4
1
1.0
0.7 (0.2, 3.0)
0.8 (0.05, 14.5)
47
7
1
23
2
1
1.0
0.6 (0.1, 3.0)
2.2 (0.1, 37.2)
Current
No
medical
Yes
problems:
DK
perimenopausal
15
0
4
19
4
0
1.0
***
***
17
7
3
19
4
0
1.0
0.5 (0.1, 2.1)
***
23
8
3
13
3
0
1.0
0.7 (0.2, 3.4)
***
***
***
***Not possible to estimate the odds ratio due to zero in cells; DK = don’t know/no data; example interpretation: Males are 2.1 times as likely as
females to have an arrhythmia.
of comparisons and it could be argued that some of
our results could have come about by chance. However,
the need to take account of this and to undertake a
Bonferroni adjustment is the subject of much discussion
in the literature.17 Perneger contends that adjusting for
the perceived problem of multiple comparisons creates
07_129_00 12/07/2001 1:02 pm Page 390
(Black plate)
390
Family Practice—an international journal
TABLE 6
Variable
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Family history
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
Family
history of:
thyroid disease
No
Yes
DK
83
11
3
34
7
1
1.0
1.7 (0.6, 4.9)
0.7 (0.07, 7.9)
35
3
1
34
7
1
1.0
2.6 (0.6, 11.1)
1.0 (0.06, 16.9)
46
7
2
23
3
0
1.0
0.9 (0.3, 3.7)
***
Family
history of:
palpitations
No
Yes
DK
71
21
5
33
8
1
1.0
0.9 (0.4, 2.3)
0.4 (0.05, 4.0)
26
12
1
33
8
1
1.0
0.6 (0.2, 1.6)
0.7 (0.04, 13.1)
39
14
2
20
6
0
1.0
0.8 (0.3, 2.5)
***
***Not possible to estimate the odds ratio due to zero in cells; DK = don’t know/no data; example interpretation: Males are 2.1 times as likely as
females to have an arrhythmia.
TABLE 7
Variable
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Consumptions
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
No. of
cigarettes/day
None
1–10
11+
DK
73
8
12
4
33
3
4
2
1.0
0.9 (0.2, 3.8)
0.6 (0.2, 2.2)
0.7 (0.1, 4.2)
33
1
4
1
33
3
4
2
1.0
2.3 (0.3, 33.3)
0.9 (0.2, 4.0)
2.0 (0.1, 18.5)
44
4
5
2
22
0
30
1
1.0
***
1.1 (0.3, 5.7)
0.9 (0.07, 13.0)
No. cups of
coffee/day
None
1–4
5+
DK
34
47
11
5
19
14
6
3
1.0
0.5 (0.2, 1.1)
0.8 (0.3, 2.7)
0.6 (0.1, 3.2)
11
19
5
4
19
14
6
3
1.0
0.4 (0.2, 1.2)
0.6 (0.1, 2.4)
0.3 (0.04, 1.8)
17
27
5
6
13
6
6
1
1.0
0.2 (0.07, 0.8)
1.8 (0.4, 7.6)
0.2 (0.01, 1.6)
No. cups of
tea/day
None
1–4
5+
DK
23
35
34
5
7
22
10
3
1.0
2.1 (0.8, 6.0)
0.9 (0.3, 2.7)
1.2 (0.2, 7.1)
9
15
22
3
7
22
10
3
1.0
2.1 (0.6, 7.0)
1.0 (0.3, 3.9)
0.8 (0.1, 6.3)
10
23
17
5
6
14
5
1
1.0
1.1 (0.3, 3.6)
0.5 (0.1, 1.9)
0.2 (0.02, 3.0)
No. of units
None
of alcohol/week 1–9
10+
DK
30
46
16
5
13
20
5
4
1.0
1.0 (0.4, 2.3)
0.6 (0.2, 2.2)
1.3 (0.3, 6.3)
14
21
2
2
13
20
5
4
1.0
1.0 (0.4, 2.7)
2.3 (0.4, 14.8)
1.6 (0.2, 11.4)
17
30
3
5
10
11
4
1
1.0
0.6 (0.2, 1.9)
2.4 (0.4, 5.7)
0.3 (0.03, 3.3)
***Not possible to estimate the odds ratio due to zero in cells; DK = don’t know/no data; example interpretation: Males are 2.1 times as likely as
females to have an arrhythmia.
more difficulties than it actually solves. For example,
Bonferroni methods suggest that all null hypotheses are
either true or false simultaneously, and we do not feel
that such a general null hypothesis should be applied to
our study. Although we could have derived a multiple
regression model from our data, it is clear that the
07_129_00 12/07/2001 1:02 pm Page 391
(Black plate)
391
New-onset palpitations in general practice
Medication
TABLE 8
Variable
Level Normal RhythmCard results/ Odds ratio
no RhythmCard results
(95% CI)
compared with RhythmCard
results indicating a ‘cardiac’
diagnosis
(n = 139)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 97)
Normal RhythmCard
results compared with
RhythmCard results
indicating a ‘cardiac’
diagnosis
(n = 81)
Odds ratio
(95% CI)
No
Arrhythmia
arrhythmia
(n = 42)
(n = 39)
Normal/insignificant
Odds ratio
arrhythmias on
(95% CI)
RhythmCards compared
with significant arrhythmias
on RhythmCards
(n = 81)
Insignificant
or no
arrhythmia
(n = 55)
Significant
arrhythmia
(n = 26)
Current
No
medication:
Yes
gastrointestinal DK
87
9
1
38
4
0
1.0
0.7 (0.2, 2.6)
***
33
6
38
4
1.0
0.4 (0.08, 1.8)
47
8
24
2
1.0
0.4 (0.06, 2.2)
Current
medication:
β-blockers
No
Yes
DK
91
4
2
38
4
0
1.0
2.0 (0.5, 9.1)
***
35
3
1
38
4
0
1.0
1.2 (0.2, 6.0)
***
51
3
1
22
4
0
1.0
3.1 (0.6, 15.4)
***
Current
No
medication:
Yes
benzodiazepines DK
92
4
1
40
2
0
1.0
0.9 (0.2, 5.4)
***
37
2
40
2
1.0
1.0 (0.1, 7.9)
52
3
25
1
1.0
0.7 (0.06, 6.8)
Current
medication:
diuretic
89
7
1
38
4
0
1.0
1.0 (0.3, 4.2)
***
35
4
38
4
1.0
0.9 (0.2, 4.5)
49
6
24
2
1.0
0.6 (0.1, 3.5)
Current
No
medication:
Yes
diabetic (oral
DK
hypoglycaemics/insulin)
94
2
1
40
2
0
1.0
2.5 (0.3, 18.8)
***
37
2
40
2
1.0
1.0 (0.1, 7.8)
53
2
24
2
1.0
2.2 (0.3, 6.7)
Current
medication:
thyroxine
No
Yes
DK
91
5
1
41
1
0
1.0
0.4 (0.04, 3.8)
***
36
3
41
1
1.0
0.3 (0.03, 3.2)
51
4
26
2
1.0
***
Current
No
medication:
Yes
ACE inhibitors DK
95
1
1
40
2
0
1.0
4.1 (0.3, 51.0)
***
38
1
40
2
1.0
2.1 (0.2, 26.6)
53
2
25
1
1.0
1.0 (0.08, 12.1)
Current
No
medication:
Yes
other antiDK
hypertensive
therapy (i.e. not
β-blockers,
diuretics or
ACE inhibitors)
89
7
1
38
4
0
1.0
1.0 (0.2, 3.9)
***
34
5
38
4
1.0
0.6 (0.1, 2.9)
48
7
24
2
1.0
0.5 (0.09, 2.8)
Current
No
medication:
Yes
other antiDK
angina therapy
(i.e. not β-blockers)
94
2
1
41
1
0
1.0
0.9 (0.1, 11.0)
***
37
2
41
1
1.0
0.4 (0.04, 5.5)
53
2
25
1
1.0
1.0 (0.08, 12.0)
Current
No
medication:
Yes
antidepressants DK
87
9
1
42
0
0
1.0
***
***
36
3
42
0
1.0
***
52
0
26
0
1.0
***
Current
No
medication:
Yes
any anginal,
antihypertensive,
diuretic, β-blocker
or ACE inhibitor
85
12
33
9
1.0
1.7 (0.6, 5.1)
31
8
33
9
1.0
1.1 (0.3, 3.6)
44
11
20
6
1.0
1.1 (0.3, 4.2)
Current
No
medication:
Yes
benzodiazepines/
antidepressants
85
12
40
2
1.0
0.4 (0.08, 1.8)
35
4
40
2
1.0
0.5 (0.08, 2.5)
50
5
25
1
1.0
0.4 (0.04, 3.5)
No
Yes
DK
***Not possible to estimate the odds ratio due to zero in cells; DK = don’t know/no data; example interpretation: Males are 2.1 times as likely as
females to have an arrhythmia.
07_129_00 12/07/2001 1:02 pm Page 392
392
(Black plate)
Family Practice—an international journal
numbers are not large enough to produce reliable
estimates. However, we can use the information from
our study to calculate an appropriate sample size in
order to power a larger study. The sample size depends
on the background frequency of the risk factor in the ‘no
arrhythmia‘ group and the odds ratio to be detected. For
example, to show a 2-fold difference between patients
with/without significant arrhythmia assuming a
background frequency of 27.3% (i.e. 15/55 from Table 3;
palpitations affected by sleeping) would require a total
of 420 patients with 90% power and 5% significance
(two-tailed). Such a larger study would also provide
sufficient power to enable combinations of clinical
information to be examined using the statistical
techniques advocated by Speigelhalter.18 These clusters
may be particularly appropriate within low prevalence
primary care populations to assist in rapid problem
solving by increasingly busy GPs.19
3
4
5
6
7
8
9
10
11
Acknowledgements
12
We would like to acknowledge the assistance of CardioAnalytics for supplying the RhythmCards and for interpreting the results. We are also grateful for the advice
received from colleagues at 3M Pharmaceuticals, in particular Sue Ellerby and Steve Foster. In re-drafting the
paper, we should also like to acknowledge the helpful
comments made by the anonymous referee. Funding was
provided by 3M Pharmaceuticals (RhythmCards) and
NHS Executive (Research and Development, Trent).
13
References
17
1
2
Barsky AJ, Ahern DK, Bailey D, Delamater BA. Predictors of
persistent palpitations and continued medical utilization. J Fam
Pract 1996; 42: 465–472.
Zwietering PJ, Knottnerus A, Gorgels T, Rinkens P. Occurrence of
arrhythmias in general practice. Scand J Prim Health Care 1996;
14: 244–250.
14
15
16
18
19
Weitz HH, Weinstock PJ. Approach to the patient with palpitations.
Med Clin North Am 1995; 79: 449–456.
Weber BE, Kapoor WN. Evaluation and outcomes of patients with
palpitations. Am J Med 1996; 100: 138–148.
Zweitering PJ, Knottnerus JA, Rinkens PELM, Kleijne MAWJ,
Gorgels APM. Arrhythmias in general practice: diagnostic
value of patient characteristics, medical history and symptoms.
Fam Pract 1998; 15: 343–353.
Zimetbaum PJ, Josephson ME. The evolving role of ambulatory
arrhythmia monitoring in general clinical practice. Ann Intern
Med 1999; 130: 848–856.
Kinlay S, Leitch JQ, Neil A, Chapman BL, Hardy DB, Fletcher PJ.
Cardiac event recorders yield more diagnoses and are more
cost-effective than 48-hour Holter monitoring in patients with
palpitations. A controlled clinical trial. Ann Intern Med 1996;
124: 16–20.
Zigmond AS, Snaith RP. The Hospital Anxiety and Depression
Scale. Acta Psychol Scand 193; 67: 361–370.
Francis B, Green M, Payne C. The GLIM System. Release 4 Manual.
Oxford: Numerical Algorithms Group, 1994.
Rigby AS. Statistical methods in epidemiology. III The odds ratio as
an approximation to the relative risk. Disabil Rehab 1999; 21:
145–151.
Morris JA, Gardner MJ. Calculating confidence intervals for
relative risks (odds ratios) and standardised ratios and rates.
Br Med J 1990; 296: 1313–1316.
Jones R, Menzies S. General Practice—Essential Facts. Oxford:
Radcliffe Medical Press, 1999.
Barsky AJ, Cleary PD, Barnett MC, Christiansen CL, Ruskin JN.
The accuracy of symptom reporting in patients complaining of
palpitations. Am J Med 1994; 97: 214–221.
Stoffers HEJH, Kester ADM, Kaiser V, Rinkens PELM,
Knottnernus JA. Diagnostic value of signs and symptoms associated with peripheral arterial occlusive disease seen in general
practice: a multivariable approach. Med Decis Making 1997; 17:
61–70.
Feinstein AR. Clinical Judgement. Baltimore: Williams & Wilkins,
1967.
Linzer M, Pritchett ELC, Pontinen M, McCarthy E, Divine GW.
Incremental diagnostic yield of loop electrocardiographic
recorders in unexplained syncope. Am J Cardiol 1990; 66:
214–219.
Perneger TV. What’s wrong with Bonferroni adjustments? Br Med J
1998; 316: 1236–1238.
Spiegelhalter DJ. Statistical methodology for evaluating gastrointestinal symptoms. Clin Gastroenterol 1985; 14: 489–515.
Muris JWM, Starmans R, Fijten GH, Crebolder HFJM, Schouten
HJA, Knottnernus JA. Non-acute abdominal complaints in
general practice: diagnostic value of signs and symptoms. Br J
Gen Pract 1995; 45: 313–316.