ORIGINAL ARTICLE
The extent and implications of
sphygmomanometer calibration error in
primary care
A Rouse and T Marshall
Department of Public Health and Epidemiology, University of Birmingham, Edgbaston, Birmingham B15
2TT, UK
Aim: The sphygmomanometer is an essential piece of Results: Of 1462 sphygmomanometers, 9.2% gave read-
diagnostic equipment, used in many routine consul- ings were more than 5 mm Hg inaccurate. No practice
tations in primary care. Its accuracy depends on correct had arrangements for maintenance and calibration of
maintenance and calibration. This study was designed sphygmomanometers. Nationally, one of 54 practices
to: (1) assess the maintenance and calibration of sphyg- had an arrangement for maintenance and calibration.
momanometers in use in primary care; (2) assess the True hypertension is very uncommon in women under
clinical, ethical, legal and public health implications of 35, a blood pressure which is measured as high is much
our findings. more likely to be caused by calibration error than by
Method: A researcher assessed the accuracy of mercury hypertension.
and aneroid sphygmomanometers in use in 231 English Conclusion: It is rare for sphygmomanometers used in
general practices. He also made enquires about primary care to be maintained and calibrated. Because
arrangements for the maintenance and calibration of of this women under 35 are at risk of misclassification
sphygmomanometers. We conducted a small telephone and inappropriate treatment. This has ethical and public
survey in general practices across the country to deter- health implications. Clinicians using equipment which
mine maintenance and calibration arrangements across has not been maintained and calibrated may be medi-
the country. We carried out a modelling exercise to cally negligent.
explore the clinical, ethical and public health impli- Journal of Human Hypertension (2001) 15, 587–591
cations of our findings.
Keywords: blood pressure determination/is [instrumentation]; family practice; medical audit; England; calibration
588
Subjects and methods manometers—such as those identified in our sur-
vey—were used. For each age–sex group we calcu-
Survey of general practices lated the probability of a normotensive being
Between March and July 1999 a researcher trained misclassified as hypertensive. We also calculated
in the calibration of sphygmomanometers visited all the positive predictive values of blood pressure
of the 231 practices serving Birmingham Health which is measured as hypertensive.
Authority. He offered a free assessment of the accu-
racy of all aneroid and mercury sphygmoman- Results
ometers in use by any member of the practice, using
methods which complied with British Standard Survey of general practices
(BS2743).7 A sphygmomanometer, which was tested A total of 217 (94%) practices—serving a combined
and certified according to British Standard population of 1 million—accepted the offer of cali-
(BS2743), was obtained from a manufacturer. The bration. In all, 1582 sphygmomanometers were
rubber air tubing of this sphygmomanometer and a identified: 120 electronic, 949 mercury and 513
the sphygmomanometer being tested were connec- aneroid. We were unable to assess the calibration
ted with a Y connector. The air pressure of the com- of electronic sphygmomanometers. The calibration
bined system was increased to 250 mm Hg. Pressure errors of the 1462 mercury and aneroid sphygmo-
was then slowly reduced to 20 mm Hg. Any differ- manometers are shown in Table 1. Nineteen percent
ence in pressure between the certified and tested gave readings which were inaccurate by more than
sphygmomanometers was noted. The measurement 2 mm Hg; 9.2% gave readings which were inaccur-
error was recorded for each sphygmomanometer. As ate by more than 5 mm Hg. No practice had arrange-
part of the protocol, enquiries were also made using ments for the ongoing recalibration and mainte-
a checklist, about arrangements for the calibration nance of sphygmomanometers. Nearly 100
and maintenance of sphygmomanometers. sphygmomanometers were in such a poor physical
state, for instance they had air leaks or dirt in the
mercury, that the tester suggested they be with-
Telephone survey drawn from service. There was no relationship
To establish whether similar findings with respect between the age and accuracy of sphygmoman-
to arrangements for calibration and maintenance ometers, with some new devices giving inaccurate
applied to the rest of England and Wales, we carried readings.
out a small telephone survey. We obtained a list of
randomly ordered practice telephone numbers, con- Telephone survey
tacted practices in this order, aiming to obtain
results from 50. We secured results from 54 prac- Of 54 practices contacted in the telephone survey,
tices. In each we spoke with either a doctor, practice only one (95% confidence interval (CI) 0.1 to 11.2%)
manager or nurse and asked what arrangements had a formal arrangement for servicing and cali-
were in place to ensure that sphygmomanometers bration. Thirty-four practices (95% CI 49% to 75%)
were regularly serviced and calibrated. accepted servicing and calibration by drug company
Table 2 The prevalence of hypertension, the probability of misclassification with an uncalibrated sphygmomanometer and the positive 589
predictive values of raised blood pressuresa
Men
16–24 1.0 0.2 0.0 83.6 0.0 0.1 0.0 0.0
25–34 2.0 0.2 0.0 89.3 1.0 0.2 0.1 82.2
35– 44 3.0 0.3 0.1 92.0 2.0 0.4 0.1 83.5
45–54 9.0 0.4 0.1 95.6 5.0 0.6 0.3 89.4
55–64 17.0 0.6 0.3 96.5 6.0 0.6 0.4 89.9
65–74 33.0 0.8 0.5 97.6 8.0 0.6 0.5 92.8
75+ 36.0 0.8 0.4 97.7 8.0 0.5 0.5 93.9
Women
16–24 0.0 0.1 0.0 0.0 0.0 0.1 0.0 0.0
25–34 0.0 0.1 0.0 0.0 0.0 0.1 0.0 0.0
35– 44 1.0 0.1 0.0 89.6 1.0 0.2 0.1 82.4
45–54 8.0 0.3 0.1 96.2 3.0 0.3 0.2 90.7
55–64 23.0 0.6 0.3 97.3 4.0 0.4 0.2 89.6
65–74 32.0 0.8 0.4 97.5 6.0 0.4 0.4 93.3
75+ 47.0 1.0 0.6 98.0 9.0 0.4 0.5 95.2
a
British Hypertension Society criteria: blood pressure ⬎160 mm Hg systolic or ⬎100 mm Hg diastolic.
representatives on an ad-hoc or regular basis. Nine- to day. There are two causes of this variation: bio-
teen practices (95% CI, 23% to 49%) had not ser- logical variation in blood pressure and measurement
viced or calibrated their sphygmomanometers for error. More than 30 potential sources of measure-
years, nor had a servicing or calibration arrangement ment error have been cited.10 Most are easily cat-
ever been in place. egorised as either relating to the patient (for
example, ‘white coat’ reaction to the physician, anx-
iety, pain, full bladder); the clinician (faulty blood
Modelling exercise
pressure measurement technique); or the instrument
The results of the modelling exercise are shown in (sphygmomanometer error). Hypertension is diag-
Table 2. If blood pressure is checked with an uncali- nosed after blood pressure measurements taken on
brated sphygmomanometer, a small number of three separate occasions. This three-reading policy
patients will be misclassified. In patients over 35, reduces the probability of making an erroneous diag-
this is not of great practical significance as the prob- nosis of hypertension due to biological variation in
ability of being misclassified is small in relation to blood pressure. It may also reduce errors due to
the true prevalence of hypertension. However, true patient factors. However, if a patient’s blood press-
hypertension is very infrequent under the age of 35 ure is measured with a sphygmomanometer which
and is practically never encountered in young has a systematic error, the blood pressure reading
women. A woman under 35 whose blood pressure is will be inaccurate whether it is repeated three or 300
measured with an uncalibrated sphygmomanometer times. Uncalibrated sphygmomanometer error is
has a small probability of being misclassified as therefore of great clinical importance. Since few
hypertensive. The positive predictive value of a practices have a system which ensures that sphyg-
blood pressure measured as ⬎160/100 mm Hg is 0% momanometers are accurate, it appears all patients
in such a woman. Our model therefore suggests that are at risk of this kind of measurement error.
all young women classified as hypertensive have The conclusions of our modelling exercise depend
been misclassified. on two facts: that the prevalence of hypertension in
In men under 25, the positive predictive value of young women is very low; and that only young
a diastolic blood pressure measured as ⬎100 mm Hg women whose blood pressures exceed 160 mm Hg
is 0% and 83.6% for systolic hypertension. Our will be considered for treatment.
model suggests that one in every six men under 25 Our estimate of the prevalence of hypertension is
classified as hypertensive has been misclassified. derived from the Health Survey for England. This
was conducted on a representative sample of the
Discussion population, with blood pressure estimated on the
basis of the mean of the second and third of three
Blood pressure readings taken on any individual measures taken on the same occasion.9 Guidelines
vary from minute to minute, hour to hour and day recommend clinicians estimate blood pressure on
590
the basis of repeated measures on separate fied than correctly identified as hypertensive. Since
occasions.8 The reported prevalence of hypertension a misclassified young woman is clearly below the
is lower in surveys where measurement takes place benefit-risk threshold: treating her would break the
at two or more points in time.11 The Health Survey injunction to do no harm. Checking her blood press-
for England may therefore have overestimated the ure with an uncalibrated instrument can only result
true prevalence of hypertension. A recently pub- in harm, it therefore cannot be ethical.
lished survey identified 0.2% of female university
students as hypertensive and a declining prevalence
Public health implications
of hypertension in successive cohorts.12 This con-
firms the impression that hypertension is very There are about 7 million women aged from 16 to
uncommon in women of this age. 34 in England and Wales. The great majority have
Might young women who have additional risk fac- their blood pressure checked frequently: for
tors but with blood pressure below the threshold of example following registration with a practice or
160 mm Hg also benefit from treatment? There is no when receiving contraceptive advice. Our model
plausible combination of risk factors which would suggests that about 7000 (0.1%) are at risk of mis-
justify treatment of a young woman for hyperten- classification and inappropriate treatment as a result
sion. According to the Framingham risk equation, a of sphygmomanometer calibration error. In England
diabetic woman smoker aged 34 with a systolic and Wales about 24 000 women under 35 are on
blood pressure of 159 mm Hg and a total cholesterol hypotensive treatment.15 We do not know how many
to high-density lipoprotein cholesterol of 6.0 has a of these women are receiving treatment unnecess-
5-year cardiovascular risk of 11%.13 She is therefore arily.
below the 15% 5-year risk threshold at which the
British Hypertension Society suggests treatment
Legal implications
should be offered.8 One per cent of all women under
35 have systolic blood pressures 150 to 159 mm Hg,9 In a non-clinical situation a professional practitioner
less than 0.1% have a total cholesterol to high- has clear legal responsibilities. For example, a pro-
density lipoprotein cholesterol of 6.0; only a tiny fessional surveyor may take a series of measure-
minority are diabetic smokers.14 ments to determine whether a house shows signs of
subsidence. He may advise expensive restorative
work on the basis of his measurements. What if the
Practical implications of instrument calibration
owner later finds out that the surveyor’s instruments
Calibrating instruments is not difficult or expensive. were not calibrated; that the relevant professional
The police recalibrate radar guns daily, chefs cali- body recommends annual recalibration; and that
brate thermometers in beakers of crushed ice, and because of this lack of calibration the advice is very
managers of local tyre garages calibrate wheel- likely to have been inappropriate? The professional
balancing instruments weekly. Organisation of a surveyor has failed in his duty of care, he is at fault
recalibration schedule should be within the capacity and is liable for any costs incurred by the owner.
of any practice. The clinical situation is clearly equivalent. Would
the legal interpretation be the same? To date this has
not been tested.
Clinical and ethical implications
Clinicians who use an uncalibrated sphygmoman-
ometer will classify some young women as hyper-
Conclusions
tensive. We have demonstrated that virtually all of Calibration of sphygmomanometers is within the
these young women will have been misclassified. scope of all general practices. Primary care prac-
The first duty of a doctor is to do no harm: non- titioners have a clinical, ethical and possibly legal
maleficence. Unfortunately, since all drugs have responsibility to ensure their instruments are appro-
potential side effects and all treatments incon- priately serviced and maintained. As a minimum
venience patients, preventive treatments in healthy first step, all clinicians should have a system in
patients may break this injunction. This is ethical place to ensure that their sphygmomanometers are
when it is judged that the likelihood of the patient recalibrated regularly to the standard endorsed by
benefiting from treatment exceeds the likelihood of the British Hypertension Society. Any clinician who
the treatment causing harm. That is, when a benefit- does not have such a system is in place, should stop
risk threshold—such as that specified in the British measuring blood pressure in healthy young women.
Hypertension Guidelines—is exceeded.8 There may
be specific reasons to measure a young (under 35)
woman’s blood pressure, such as pregnancy or renal
Acknowledgements
disease. However, if blood pressure estimation is Thanks are due to Mr Brian Pritchard for carrying
carried out with the aim of preventing cardiovascu- out the survey of sphygmomanometers.
lar disease the results clearly show that unselected Andrew Rouse obtained data on sphygmoman-
young women are much more likely to be misclassi- ometer calibration error, made a crude estimate of
591
the prevalence of misclassification error and contrib- an urban population. Am J Pub Health 1987; 77:
uted to writing the paper. 1459–1461.
Tom Marshall developed models for estimating 7 Specification for Aneroid and mercury non-automated
the prevalence of misclassification due to uncali- sphygmomanometers. BS2743: 1990, British Standards
Institute, London.
brated sphygmomanometer error and contributed to 8 Ramsay LE et al. British Hypertension Society guide-
writing the paper. lines for hypertension management 1999: summary.
BMJ 1999; 319: 630–635.
Conflicts of interest 9 Department of Health. Health survey for England ’96.
Stationery Office: London, 1998.
None 10 Reeves RA. Does this patient have hypertension? How
to measure blood pressure. J Am Med Assoc 1995;
273: 12118.
References 11 van de Mheen PJ, Bonneux L, Gunning-Schepers LJ.
1 Specification for aneroid and mercury non-automated Variation in reported prevalences of hypertension in
sphygmomanometers. BS2743: 1990, British Standards The Netherlands: the impact of methodological vari-
Institute, London. ables. J Epidemiol Community Health 1995; 49: 277–
2 Hussain A, Cox JG. An audit of the use of sphygmo- 280.
manometers. Br J Clin Pract 1996; 50: 136–137. 12 McCarron P et al. Changes in blood pressure among
3 Maskrey N, Parker C, Powell S. THINYCE – Treating students attending Glasgow University between 1948
Hypertension Through Clinical Effectiveness in North and 1968: analyses of cross sectional surveys. BMJ
Yorkshire, October 1998. North Yorkshire Clinical 2001; 322: 885–889.
Advisory Service: North Yorkshire Health Authority. 13 Anderson KA, Wilson PWF, Odell PM, Kannel WB. An
4 Cranney M, Barton S, Walley T. Auditing the manage- updated coronary risk profile. A statement for health
ment of hypertension in British general practice: a professionals. Circulation 1991; 83: 356–362.
critical literature review. Br J Gen Pract 1998; 48: 14 Unwin N et al. The implications of applying widely
1424 –1428. accepted cholesterol screening and management
5 Jones JS, Ramsey W, Hetrick T. Accuracy of prehospi- guidelines to a British adult population: results from
tal sphygmomanometers. J Emergency Med 1987; 5: a cross sectional study of cardiovascular disease and
23–27. risk factors. BMJ 1998; 317: 1125–1130.
6 Hahn LP, Folsom AR, Sprafka JM, Prineas RJ. Preva- 15 Department of Health. Health survey for England ’98.
lence and accuracy of home sphygmomanometers in Stationery Office: London, 2000.