Reliability of Reporting Self-Measured Blood Pressure Values by Hypertensive Patients

Abstract

For self-measurement of blood pressure to be useful, patient reporting of test results must be reliable and accurate. Until now no study directly measured the accuracy and reliability of patients’ reporting of self-measured blood pressure values.

Thirty hypertensive patients (69 ± 11 years) were instructed to measure blood pressure at home over 14 days with the highly accurate Omron IC monitor and to keep a record of all readings in a patient logbook. To assess the reliability of the records, patients were not informed about the memory capacity of the device. We compared automatically stored blood pressure readings with the respective logbook entries to analyze deletion (under-reporting), addition (over-reporting), and precision of reporting of test results.

The prevalent pattern was under-reporting, averaging 36% ± 24% (3% to 89%), which occurred significantly more than over-reporting (9% ± 11%; 0% to 38%). The precision of reporting (identical values at corresponding times) was 76% ± 34% (0% to 100%). This observer error did not affect group comparisons of automatically stored values and logbook entries, although the estimated limits of agreement were wide. Blood pressure control, duration of hypertension, age, or previous use of self-measurement and patterns of logbook entries were not found to be predictive of the patients’ reliability.

Our results demonstrate a substantial observer error in the reporting of self-measured blood pressure values. This bias may be reduced by memory-equipped blood pressure devices.

There is convincing evidence in clinical studies that multiple blood pressure measurements obtained by self-monitoring patients at home improve the precision and reproducibility of blood pressure measurements.^1,2 Furthermore, epidemiologic and pharmacologic blood pressure studies have shown an increasing trend for self-monitoring.^3–6 The increasing trend is mainly patient-driven and electronic devices for self-monitoring are now widely available in drug stores and mail order catalogues.^7,8 Self-monitoring of blood pressure is only useful if patient-generated data are reliable and accurate. Until now, analysis of self-monitored blood pressure values depended on the accurate reporting of measurements in the patients’ logbooks. In view of many well-known observer errors in clinical blood pressure measurement,^9,10 it is astonishing that all studies performed with self-monitoring heavily relied on patient-reported values. No study directly measured the accuracy and reliability of the routine daily use of home monitoring and the quality of of the reporting of the results.⁸

Patients and Methods

Patients

Thirty consecutive patients with arterial hypertension (systolic blood pressure ≥140 mm Hg, diastolic blood pressure ≥90 mm Hg, according to the World Health Organization definition) were chosen from an outpatient clinic in the Department of Internal Medicine at the University of Bonn. Twenty-one patients were already taking antihypertensive medication (11 monotherapy and 10 combination therapy). The duration of arterial hypertension ranged from 0 to 25 years. Sixteen patients had already purchased a device for home monitoring on their own, but had not been sufficiently instructed in the technique by a physician.

Procedures

Self-measurement was performed by all patients with a successfully tested and previously calibrated fully-automatic, oscillometric device with automatic inflation and deflation and a digital display of blood pressure and heart rate (Omron Intensive Control, IC; Omron, Matsusaka, Japan). The Omron IC has two components, the Omron IC blood pressure monitor and the Omron IC software.

The Omron IC monitor looks like a standard blood pressure monitor. The system used in the present study is highly accurate and includes the same algorithm as the Omron HEM-705 CP.¹¹ As a special feature, the Omron IC is equipped with a memory chip and an interface for connecting it to a computer. The software can be used for statistical analysis as well as to prepare a schedule for the patient to set the measuring times.

After sufficient instruction in the correct technique of self-measurement by a physician in charge of the trial, patients were instructed to measure blood pressure at home once in the morning and once in the evening at preset times over a period of 2 weeks. The patients were not informed about the memory capacity of the device. They were instructed to keep a record of all blood pressure readings with the corresponding time of measurement in a patient logbook. The patients were not instructed to validate the Omron IC measurements. The instructions were standardized and always performed by the same physician (TM).

Schedules of blood pressure measurement were individually prepared according to the usual living habits and intake of antihypertensive medication (6 to 8 AM and 6 to 8 PM). At set times the patient was reminded to measure blood pressure by an acoustic signal from the blood pressure device.

Analyses

Four determinations were made for each patient: 1) the number of blood pressure recordings stored in the Omron IC system; 2) the number of blood pressure recordings in the patients’ logbook; 3) the number of blood pressure recordings in the Omron IC system with a recording of any value in the logbook at the corresponding time; and 4) the number of corresponding time recordings in the Omron IC system and logbook showing identical values. From these data three measurements were derived: 1) over-reporting, 2) under-reporting, and 3) precision.

A measure of over-reporting was developed to assess the degree to which patients added the results of tests that were not actually performed. Over-reporting was defined as the number of blood pressure reportings that were not verified at a corresponding time in the Omron IC system expressed as percentage of blood pressure recordings in the logbook. A score of 0% indicated that every logbook entry was verified by a recording of any value in the Omron IC system at the corresponding time.

To assess the degree to which patients omitted test results, an under-reporting score was developed. Under-reporting was defined as the number of readings in the Omron IC system that were not reported in the patient logbook expressed as percentage of the total number of blood pressure readings in the Omron IC system. A score of 0% would thus indicate that every recording of the Omron IC system was entered in the logbook.

We assessed the accuracy of patients’ reporting of blood pressure readings by comparing them to the Omron IC recorded readings. Precision was defined as the number of recordings in the Omron IC system and logbook that were of identical value and reported at the corresponding time expressed as percentage of the number of corresponding readings found in the Omron IC system and the logbook. A score of 100% would thus indicate that every blood pressure value written in the logbook was exactly the same as the value recorded in the Omron IC system at the corresponding time.

Data are expressed as means ± SD. The data were analyzed using paired t tests to compare mean logbook values with Omron IC system values. Over- and under-reporting scores were compared by a Wilcoxon range test for nonparametric variables.

Linear regression analysis was used to examine the correlation between the Omron IC system-derived mean blood pressure values, duration of hypertension, or age and the reliability (percentage of under-reporting, over-reporting, and precision). Linear regression analysis was also used to examine the correlation between under-reporting, over-reporting, and precision.

Results

Reporting and Precision

The total number of automatically stored monitor pressures, total number of logbook pressures, and total number of accurate logbook pressures (corresponding time recordings in the Omron IC system and logbook showing identical values) are given in Table 1.

All patients showed more or less extensive under-reporting ranging from 3% to 89% (mean 36% ± 24%). Over-reporting scores were significantly lower than under-reporting scores (P < .0001) and ranged from 0% to 35% (mean 9% ± 11%).

Nineteen of the 30 patients showed a pattern of both omitting readings from and adding phantom values to logbook entries. There was no relationship between patterns of under- and over-reporting. Precision ranged from 0% to 100% (mean 76% ± 34%). Fifteen patients showed precision indices >90%. There was no significant correlation between precision and reporting. Thus neither omission nor addition of values to logbook entries was indicative of poor precision.

The analysis of individual cases showed variable patterns of patient performance with reporting of self-measured blood pressures (Figure 1). Each of these three patients had already performed self-measurement at home before inclusion in the study and was familiar with reporting self-generated blood pressure data in the logbook. All three patients were taking antihypertensive medication during the study.

Comparison of systolic blood pressure values derived from logbook entries with the automatically stored values of the Omron IC software over the whole study period in three selected patients. The open circles with solid lines depict all readings from the Omron IC software. The closed squares represent patient-reported logbook entries. Panel A: Patient 6 showed a precision index of 0% combined with under-reporting of 26% and over-reporting of 5%. In spite of the low precision, differences between patient-reported and automatically stored values were not relevant (1.1/1.1 mm Hg). A detailed analysis revealed that this patient entered all blood pressure readings in the logbook two days later than actually performed, apparently because of mixing up dates. Panel B: Patient 30 also showed a low precision index of 18% with high under- and over-reporting. Mean logbook values were significantly lower than Omron IC values (9.5/3.9 mm Hg), because this patient omitted high blood pressure readings. Panel C: Patient 21 showed a similar degree of under-reporting (65%) compared to patient 30. The precision index was, however, 100% and there was no relevant difference between logbook entries and stored Omron IC values (0.1/0.0 mm Hg).

We then tested whether there was a relationship between adequate blood pressure control, age, duration of hypertension, or previous use of self-measurement and reliability. We could not establish a significant correlation between blood pressure control, age, or duration of hypertension and the percentage of under-, over-reporting, or precision. Furthermore, previous use of self-measurement did not predict a good reliability of patients’ reporting.

Comparison of Logbook Entries With Automatically Stored Values

This observer error did not affect group comparisons of automatically stored values and logbook entries, although the estimated limits of agreement were wide. Thus, the mean of all subjects’ average logbook pressures (155 ± 15/87 ± 9 mm Hg) were very similar compared with the automatically stored monitor pressures (154 ± 16/87 ± 8 mm Hg; P = NS). The correlation coefficients between logbook entries and Omron IC values were r = 0.95 (P < .0001) for systolic and r = 0.95 (P < .0001) for diastolic values, respectively. However, for individual patients we observed differences between logbook entries and Omron IC values ranging from −11 to +10 mm Hg for systolic blood pressure and −5 to +7 mm Hg for diastolic blood pressure.

Discussion

The widespread use of self-measurement has resulted in a marked reliance on patient-reported blood pressure data. The reliability of patient-reported logbook data has not been investigated, although observer errors of blood pressure measurement are a general finding in blood pressure studies.^9,10

The problems encountered with the use of older devices consisted of terminal digit preference, expectation bias, and false cuff inflation and deflation. The development of accurate, fully automated oscillometric devices with a digital display has substantially reduced the observer errors of blood pressure measurement at home.

The present study is the first one investigating reliability in terms of observer error in reporting self-measured blood pressure values. The system used in the present study, modified to store all blood pressure readings with the respective time of measurement, enabled us to directly address the issue of patient performance in documenting self-measured blood pressures.

We found a high degree of variability in under-reporting, over-reporting, and precision of these data. In principal, the main error consisted of omission of additional blood pressure readings performed at times different from the scheduled set points. Thus, the mean number of logbook entries was significantly lower than the total number of measurements actually performed and stored in the Omron IC system (30 ± 10 υ 49 ± 28; P < .01).

Although in individual patients the omission of values from the logbook resulted in different blood pressure values (Figure 1B), this did not affect the group comparisons of blood pressure values. The very similar results for mean logbook pressures and monitor pressures may be explained by a pattern of rejecting both low and high blood pressure values that patients chose not to write down. However, we could not detect a systematic pattern of under-reporting.

In this context it is interesting that in a study with insulin-dependent diabetic patients who monitored their blood glucose levels at home there was a tendency to report lower values than the device actually displayed. This was shown by inserting a memory chip in the device without the patients’ knowledge.¹² In this study the precision index of the reported blood glucose values was 74% and thus very similar to the index found in our study (76%). In contrast to the present study in the diabetic patients, over-reporting was the prevalent pattern of reporting and averaged 40% (under-reporting was 10%). The prevalent errors in the diabetic patients consisted of omission of high glucose readings and the addition of values lower than the true test results.¹²

We thus identified a degree of unreliability among the study patients that may have profound effects on the usefulness of self-measured blood pressure. This inaccuracy could not be explained by inadequate technique or failure of the blood pressure device. It seems unlikely that the observer bias documented in the present study was due to basic patient characteristics, such as age or blurred vision, since observer bias in blood pressure measurement is a more general finding.^9,10 Digit preference was evident in physician measurements as well as patient self-measurement in the recently published Self-Measurement for the Assessment of the Response to Trandolapril (SMART) study.³ The findings emerging from our observations have important implications for the future planning of epidemiologic and pharmacologic studies.

In summary, the degree of reliability of hypertensive patients’ reporting of self-measured blood pressure values was both variable and unpredictable. We were not able to establish a relationship between the different patterns of reporting or to associate these patterns with overall blood pressure control or previous use of self-measurement. The use of memory-equipped blood pressure devices with the patients’ knowledge and the providers’ feedback may ameliorate the problems of observer bias and contribute to improved treatment of hypertension.

We thank Frank Wilmerstädt (Omron Health Care Europe) for technical support of the study.

References