Surveillance and monitoring of blood pressure and hypertension are critical factors in the effort to prevent and control hypertension 1. Population-based surveys that use objective blood pressure measurements, such as the US National Health and Nutrition Examination Survey and the Canadian Health Measures Survey, provide a much needed picture of the prevalence, awareness, treatment, and control of high blood pressure 2–4. In some settings, such as telephone surveys, objective blood pressure measurements are not feasible and self-reported measures of blood pressure control have been used as an alternative 5, as such reports could theoretically allow risk factors for and outcomes of uncontrolled high blood pressure to be examined.
For example, in 2009, the Canadian federal government fielded the national telephone-administered Survey on Living with Chronic Disease in Canada to assess the knowledge, attitudes, and behaviors of Canadians with hypertension 5. Among the many items measured as a part of the telephone interview 6, participants were asked whether their blood pressure was generally well-controlled, borderline, high, or low, as well as about the number of medications taken for high blood pressure and the timing of their most recent clinic blood pressure assessment. It remains unclear, however, whether these measures can be used to accurately assess blood pressure control, medication use, and recent blood pressure assessments.
To date, no study has considered the accuracy of a simple self-reported interview question in assessing blood pressure control in patients with hypertension. Although a number of studies have compared self-recorded home blood pressure readings with those recorded objectively using a home blood pressure monitor 7–9 or with those measured in a clinic 10, the results do not foster a better understanding of the extent to which individuals with hypertension can accurately report their general level of blood pressure control in an interview setting. Thus, the primary objective of the current study was to estimate the validity of self-reported blood pressure control, as assessed in the national survey protocol 6, in people with hypertension with and without diabetes (as clinically recommended blood pressure targets differ between these groups in some countries such as Canada, i.e., <130/80 vs. <140/90 mmHg 11,12). As a secondary objective, we sought to validate items describing self-reported timing of the most recent blood pressure assessment by a health professional and self-reported numbers of antihypertensive medications, as these reflect important indicators of clinical management 1,13. Validation of these measures may support or discourage their inclusion in future self-reported surveys conducted at national and other levels.
Source of study population
Participants were recruited from the patient roster of the Queen’s Family Health Team, an interdisciplinary primary care practice of family physicians, nurse practitioners, and other health professionals, serving ∼14 000 residents of Kingston, Ontario, Canada, and its surrounding areas. The study protocol was approved by the Queen’s University Health Sciences and Affiliated Teaching Hospitals Research Ethics Board. All of the family physicians who practice in this setting (n=20) consented to have their patients eligible for selection into the study. Numbers of eligible, selected, excluded, and participating patients are summarized in Fig. 1.
Participants were eligible if they were aged 20 years or older, were not currently pregnant, and presented in the electronic medical record with an ICD-9 diagnostic code for hypertension (401, 402, 403, 404, 405) on 15 November 2011 (n=1319) or 9 May 2012 (n=945, which excluded individuals selected as part of the November sample); 99% of patients had a diagnosis of essential hypertension (ICD-9 401). Patients were randomly selected from the electronic medical record in strata defined by diabetes status, sex, and age (<65 vs. ≥65 years) to try to ensure an equal distribution of people with and without diabetes and to reflect the age–sex distribution of Canadians with hypertension. Each patient’s eligibility was reviewed by his/her family physician; patients were ineligible for the study if their physician deemed them unable to provide informed consent or unable to attend the clinic.
A total of 752 eligible participants (n=375 November sample and n=378 May sample) were sent letters and an information sheet signed by their family physician inviting them to participate. Individuals who did not initially respond were sent reminder postcards 14. In wave one (November 2011 to April 2012), 68 individuals agreed to participate. In wave two (May to September 2012), 93 individuals agreed to participate. The sample size was calculated to provide estimates of sensitivity and positive predictive values that would fall within 10% of the true value 95% of the time, with emphasis on precision of these estimates rather than representativeness of the sample. As individuals with diabetes were oversampled and older men were more likely to participate, estimates were weighted to reflect the distribution of Canadians with diagnosed hypertension from the 2009 Survey on Living with Chronic Disease in Canada based on diabetes status, sex, and age (<65 vs. ≥65 years) in an effort to improve representativeness.
All participants received an information sheet and consent form to read, discuss, and sign. A 5-min in-person structured interview was conducted, using a short version (21 questions) of Statistics Canada’s 2009 Survey on Living with Chronic Disease in Canada hypertension questionnaire (available at http://www.statcan.gc.ca) 6. After the interview, blood pressure was measured using a BpTRU BP-200 device (BpTRU Medical Devices Ltd., Coquitlam, British Columbia, Canada). An appropriately sized cuff was chosen on the basis of measured upper-arm circumference: small adult (18–26 cm), regular adult (26–34 cm), large adult (32–43 cm), and extra-large adult (41–52 cm). The cuff was fastened around the participant’s arm, with the centre of the bladder over the brachial artery and the lower margin of the cuff 2–3 cm above the antecubital fossa (elbow crease). Participants were seated with both feet on the floor, and with their back and arm supported so that the elbow crease rested and the cuff was positioned in line with the heart with the palm of the hand facing downward. Respondents were left alone and asked to sit quietly, relax, and refrain from moving or talking for a 5-min rest period. After the rest period, the interviewer re-entered the room to start the BPTru, remained in the room for the first measurement in order to ensure proper functioning, and left the room for the remaining five measurements. The average systolic blood pressure and diastolic blood pressure were calculated automatically on the basis of the last five of six measurements.
After the interview and blood pressure assessment, the electronic medical records were reviewed by the interviewer using an abstraction form; thereafter, they were reviewed a second time to ensure completeness and confirm the accuracy of data abstraction. The ‘measurements’, ‘prescriptions’, and ‘disease registry’ modules of the Open Source Clinical Application Resource (OSCAR; McMaster University, Hamilton, Ontario, Canada) electronic medical record were reviewed for information on up to eight systolic and diastolic blood pressure measurements within 1 year preceding the interview (typically measured using BPTru), antihypertensive medications prescribed at the time of interview, diabetes status, and diagnosis of comorbidities for which certain antihypertensive medications are often prescribed (i.e. chronic kidney disease, edema, migraine, heart failure, arrhythmia, myocardial infarction, and angina).
Key measures and definitions
Blood pressure control
Individuals were asked: ‘In general, do you consider your blood pressure to be: (1) well-controlled (normal, fine, ok); (2) borderline; (3) high; (4) low’. Individuals were classified as controlled if they reported ‘well-controlled’ or ‘low’ blood pressure and uncontrolled if they reported ‘borderline’ or ‘high’ blood pressure. Controlled blood pressure (well-controlled or low) was compared with the blood pressures measured using BPTru on the interview day. To account for daily variation in blood pressure, self-reported blood pressure control was also compared with (i) the most recently recorded systolic and diastolic blood pressure in the electronic medical record and (ii) the average of up to eight blood pressure readings recorded in the electronic medical record in the previous year.
Both objectively measured and chart-abstracted blood pressure control were defined according to a less than 140/90 mmHg threshold for the subgroup without diabetes and according to a less than 130/80 mmHg threshold for the subgroup with diabetes 11,12. In sensitivity analyses, we explored the effects of defining blood pressure control (i) using a 140/90 mmHg threshold for all participants, including those with diabetes, (ii) using a 135/85 mmHg threshold 15, and (iii) using BPTru measurements adjusted to reflect sphygmomanometer readings, according to the following validated equations: adjusted systolic blood pressure=11.4+(0.93×BPTru systolic blood pressure) and adjusted diastolic blood pressure=15.6+(0.83×BPTru diastolic blood pressure) 16.
Number of antihypertensive medications
As part of the interview, respondents were asked how many medications they were currently taking for high blood pressure. The self-reported number (0, 1, 2, 3, 4+) was compared with active prescriptions at the time of interview, as recorded in the prescriptions module of the electronic medical record.
Timing of most recent blood pressure assessment
Participants were asked: ‘When was the last time you had your blood pressure measured by a health professional? Was it: (1) less than 1 month ago?; (2) 1 month to less than 3 months ago?; (3) 3 months to less than 6 months ago?; (4) 6 months to less than 1 year ago?; (5) 1 year to less than 2 years ago?; (6) or 2 or more years ago?’ Self-reported responses were compared with the timing of the most recently measured blood pressure recorded in the electronic medical record.
Sensitivity and specificity were calculated as measures of validity for self-reported blood pressure control, and positive and negative predictive values were calculated to determine the expected yield in a survey setting, overall and stratified according to diabetes status, sex, and age (<65 vs. ≥65 years). As previously stated, estimates were weighted to reflect the age–sex distribution and the prevalence of diagnosed diabetes in Canadians with diagnosed hypertension according to the 2009 Survey on Living with Chronic Disease in Canada. Agreement between self-reported and chart-abstracted number of antihypertensive medications and timing of the most recent blood pressure assessment was estimated using the κ-statistic.
Characteristics of the study population are shown in Table 1. Individuals with diabetes ranged in age from 46 to 88 years, with an average age of 69±10 years, 51% of whom were women. Individuals without diabetes ranged in age from 36 years to 94 years, with an average age of 68±13 years, 54% of whom were women. The majority in both groups (93 and 87%) were taking antihypertensive medication. Resistant hypertension, defined as uncontrolled high blood pressure despite concurrent use of three or more antihypertensive medications 17, affected 3% of the study population (4 and 3% of those with and without diabetes, respectively). Among people with diabetes, prevalence of blood pressure control based on self-report was 79% compared with 67% based on measured blood pressure (defined as blood pressure <130/80 mmHg). Among people without diabetes, prevalence of blood pressure control based on self-report was 70% compared with 77% based on measured blood pressure.
Validity of self-reported blood pressure control in people with and those without diabetes
Among people with diabetes, sensitivity and specificity of self-reported blood pressure control were 83% [95% confidence interval (CI): 72–94%] and 30% (95% CI: 11–49%), respectively, when compared with BPTru measured blood pressures of less than 130/80 mmHg (Table 2). When blood pressure control was defined on the basis of a 140/90 mmHg threshold, sensitivity and specificity improved to 87% (95% CI: 78–98%) and 88% (95% CI: 52–99%), respectively.
Among people without diabetes, sensitivity and specificity of self-reported blood pressure control were 78% (95% CI: 68–88%) and 58% (95% CI: 37–79%), respectively, meaning that, among people with blood pressures less than 140/90 mmHg on the day of the interview, approximately four in five reported having well-controlled or low blood pressure (controlled), and among people who had uncontrolled blood pressure (≥140/90 mmHg) on the day of the interview, nearly three in five reported that it was borderline or high (uncontrolled).
Results were generally unchanged in sensitivity analyses that used a 5 mmHg lower threshold for defining blood pressure control or BPTru measurements adjusted to reflect sphygmomanometry (data not shown). When self-reported blood pressure control was compared with the most recent blood pressure measurement in the clinic (obtained through chart abstraction), sensitivity estimates remained unchanged (83 and 82% for people with and those without diabetes), whereas specificity decreased to 24 and 50% for people with and those without diabetes. Similar findings were observed when self-reports were compared with the average of up to eight blood pressure measurements taken over the previous year (data not shown).
Validity of self-reported number of antihypertensive medications
Self-reported and chart-abstracted numbers of antihypertensive medications showed fair agreement (κ=0.7; 95% CI: 0.6–0.8), and this did not differ by diabetes status (Table 3). Seventy-seven percent of people accurately reported the number of antihypertensive medications prescribed, 14% underestimated the number of prescriptions (13% by one medication and 1% by two or more medications), and 9% overestimated the number of prescriptions (7% by one medication and 2% by two or more medications). Among participants with a chronic condition for which antihypertensive medications can be prescribed (i.e. chronic kidney disease, edema, migraine, heart failure, arrhythmia, myocardial infarction, or angina), agreement decreased (κ=0.4).
Validity of self-reported timing of the most recent clinic blood pressure assessment
Self-reported timing of the most recent blood pressure assessment showed fair agreement (κ=0.5; 95% CI: 0.4–0.6) with chart-abstracted timing. Sixty-three percent of participants accurately reported the time since the last blood pressure measurement, with 31% reporting that it had occurred more recently than it had and 6% reporting that it had occurred longer ago than it had. Ninety-eight percent of individuals reported undergoing clinic blood pressure assessment in the previous year, with 95% agreement (data not shown).
We showed, in a sample of people with hypertension from a family medicine clinic, that self-reported blood pressure control had reasonable sensitivity but low specificity. This means that in a survey setting, although four in five individuals with controlled blood pressure will report that their blood pressure is controlled, a large proportion of individuals with uncontrolled high blood pressure will also report that it is controlled (40–70%). The amount of misclassification introduced by this self-reported measure of blood pressure control suggests that it may not be useful for assessing the prevalence of hypertension control or determining associations in survey settings. The self-reported number of prescribed medications showed fair agreement with chart-abstracted information. Self-reported receipt of a clinic blood pressure assessment in the previous year was accurate for 95% of participants.
In our clinic-based sample of patient volunteers, a large proportion (40–70%) of individuals with objectively measured uncontrolled high blood pressure incorrectly reported that their blood pressure was controlled. We expect that this self-reported measure of blood pressure control would perform worse when administered by telephone to a general household sample, since our low response rate likely resulted in a volunteer sample with greater awareness of their level of blood pressure control than the general Canadian population. Compared with the general population with diagnosed hypertension, our volunteer sample (that was weighted to improve representativeness) had greater knowledge of the recommended blood pressure targets and had been diagnosed for a longer period of time. Further, as BPTru is used regularly in the clinic from which the patient sample was drawn, the study patients may have had a greater opportunity to see their own results. From this, it may be reasonable to infer that accuracy of self-reported blood pressure control would be lower when administered in other settings. If required, this could be confirmed at the population level by including the questions in a national physical measures survey. In our own setting in Canada, this might involve inclusion of the questions in a future cycle of the Canadian Health Measures Survey. In the meantime, the results of the current study suggest that the self-reported measure of blood pressure control has insufficient validity to be used in large population-based surveys and other settings.
A strength of the study is that comparisons were made with blood pressures measured on the day of interview, as well as those previously measured in the clinic, to account for daily variability in blood pressure. Further, we used an automated blood pressure monitor to assess blood pressure control; the main advantage of this method is that blood pressure measurements are made in an automated manner in the absence of an observer, thereby eliminating observer errors, digit preference, and reducing white-coat hypertension 18. Inter-rater variability was avoided as a single interviewer administered all interviews and blood pressure assessments.
A limitation of the study is that, although providing a reasonable amount of precision around estimates of sensitivity, the small sample size did not provide the same level of precision for specificity given the high prevalence of blood pressure control. The second limitation is that we did not compare the self-reported number of antihypertensive medications with the number of medications measured by recording drug identification numbers directly from medication pill bottles 19. Instead, self-reports were compared with chart-abstracted numbers of antihypertensive medications, which could have included discontinued medications. Further, timing of the most recent blood pressure assessment was compared with the timing in the medical record, which would not have included assessments made outside of the clinic, such as those made during hospital visits, cardiac rehabilitation sessions, or in other clinical settings. In addition, as the population studied was sociodemographically homogeneous, we could not explore the effect of factors such as ethnicity and language on validity 1. Finally, the information on which patients based their self-reported blood pressure assessments is unclear, considering that accuracy was not improved when self-reported blood pressure control was compared with chart-abstracted blood pressure values and that only 17% of the sample reported regularly monitoring their blood pressure at home.
We validated an existing self-reported measure of blood pressure control from the 2009 Survey on Living with Chronic Disease in Canada and found that the majority of individuals with uncontrolled high blood pressure incorrectly reported that their blood pressure is controlled. This level of misclassification suggests that a self-reported measure of blood pressure control may not be sufficiently valid (sensitive and specific) to be useful for assessing hypertension control in health surveys administered in clinical settings or large population health surveys administered by telephone.
The authors would like to acknowledge Diane Cross, Abigail Scott, Danyal Martin, and the physician group of the Queen’s Family Health Team for their help with recruitment. They also wish to thank all of the study participants for their time and contribution to this study.
M.E.G. designed the study, collected and analyzed the data, and wrote the manuscript. W.P., I.J., N.R.C., and R.B. contributed to the design of the study and reviewed, edited, and contributed to the development of the manuscript.
Conflicts of interest
M.E.G. was supported by a Canadian Institutes of Health Research (CIHR) doctoral award and is supported by a Heart and Stroke Doctoral Award. I.J. holds a Canada Research Chair in Physical Activity and Obesity. N.R.C. holds the Heart and Stroke Foundation – CIHR Research Chair in Hypertension Prevention and Control. M.E.G. is an employee of the Public Health Agency of Canada, the sponsor organization for the Survey on Living with Chronic Disease in Canada. For the remaining authors there are no conflicts of interest.
1. Campbell NR, McAlister FA, Quan H.Monitoring and evaluating efforts to control hypertension in Canada: why, how, and what it tells us needs to be done about current care gaps.Can J Cardiol2012;29:564–570.
2. McAlister FA, Wilkins K, Joffres M, Leenen FH, Fodor G, Gee M, et al..Changes in the rates of awareness, treatment and control of hypertension in Canada over the past two decades.CMAJ2011;183:1007–1013.
3. Gee ME, Bienek A, McAlister FA, Robitaille C, Joffres M, Tremblay MS, et al..Factors associated with lack of awareness and uncontrolled high blood pressure among Canadian adults with hypertension.Can J Cardiol2012;28:375–382.
4. Gee ME, Janssen I, Pickett W, McAlister FA, Bancej CM, Joffres M, et al..Prevalence, awareness, treatment, and control of hypertension among Canadian adults with diabetes, 2007 to 2009.Can J Cardiol2012;28:367–374.
5. Bienek A, Gee ME, Nolan RP, K J, Campbell NR, Bancej CM, Gwadry-Sridhar F, et al..Methodology of the 2009 Survey on Living with Chronic Diseases in Canada – hypertension component.Chronic Dis Can2012;33:267–276.
6. Statistics Canada. Survey on Living with Chronic Diseases in Canada User Guide. 2009. Available at: http://www23.statcan.gc.ca/imdb-bmdi/pub/document/5160_D5_T1_V1-eng.htm
. [Accessed 2 December 2013].
7. Cheng C, Studdiford JS, Chambers CV, Diamond JJ, Paynter N.The reliability of patient self-reported blood pressures.J Clin Hypertens (Greenwich)2002;4:259–264.
8. Mengden T, Hernandez Medina RM, Beltran B, Alvarez E, Kraft K, Vetter H.Reliability of reporting self-measured blood pressure values by hypertensive patients.Am J Hypertens1998;11:1413–1417.
9. Johnson KA, Partsch DJ, Rippole LL, McVey DM.Reliability of self-reported blood pressure measurements.Arch Intern Med1999;159:2689–2693.
10. Gould BA, Kieso HA, Hornung R, Altman DG, Cashman PM, Raftery EB.Assessment of the accuracy and role of self-recorded blood pressures in the management of hypertension.Br Med J (Clin Res Ed)1982;285:1691–1694.
11. Daskalopoulou SS, Khan NA, Quinn RR, Ruzicka M, McKay DW, Hackam DG, et al..The 2012 Canadian hypertension education program recommendations for the management of hypertension: blood pressure measurement, diagnosis, assessment of risk, and therapy.Can J Cardiol2012;28:270–287.
12. Chobanian AV, Bakris GL, Black HR, Cushman WC, Green LA, Izzo JL Jr, et al..Seventh report of the Joint National Committee on Prevention, Detection, Evaluation, and Treatment of High Blood Pressure.Hypertension2003;42:1206–1252.
13. Campbell N, Young ER, Drouin D, Legowski B, Adams MA, Farrell J, et al..A framework for discussion on how to improve prevention, management, and control of hypertension in Canada.Can J Cardiol2012;28:262–269.
14. De Leeuw ED, Hox JJ, Dillman DA.International handbook of survey methodology2008.New York:Lawrence Erlbaum Associates.
15. Godwin M, Birtwhistle R, Delva D, Lam M, Casson I, MacDonald S, et al..Manual and automated office measurements in relation to awake ambulatory blood pressure monitoring.Fam Pract2011;28:110–117.
16. Myers MG, McInnis NH, Fodor GJ, Leenen FH.Comparison between an automated and manual sphygmomanometer in a population survey.Am J Hypertens2008;21:280–283.
17. McAlister FA, Lewanczuk RZ, Teo KK.Resistant hypertension: an overview.Can J Cardiol1996;12:822–828.
18. Myers MG.Recent advances in automated blood pressure measurement.Curr Hypertens Rep2008;10:355–358.
19. Statistics Canada. Canadian Health Measures Survey (CHMS) Data User Guide: Cycle 1. 2010. Available at: http://www23.statcan.gc.ca/imdb-bmdi/pub/document/5071_D2_T1_V1-eng.htm
. [Accessed 2 December 2013].