This is a cross-sectional descriptive study for the validation of a diagnostic test.
Population
We worked with hypertensive patients registered in the computer system at the Vistalegre-La Flota Urban Health Centre in Murcia, Spain and those receiving routine care.
Study period
December 2011 to December 2012.
Criteria for inclusion and exclusion
Inclusion criteria included the diagnosis of complicated and uncomplicated hypertension, an age of 18–80 years, sufficient vision and hearing to perform the self-measurement, and adequate intellectual capacity to obtain the measurements or the oversight of a responsible caregiver for doing so. The exclusion criteria that were considered are those considered valid in various international HT guidelines [1, 5]: Immobilized patients without a responsible caregiver and hypertensive patients diagnosed with obsessive-compulsive disorder.
Selection mechanism
Duplicates, diagnostic errors, patients in the computerized clinical history, and those with their last visit to the health centre (to see a doctor or nurse) in the year prior to the start of the patient selection (N = 2,245) were filtered from anonymized lists. The sample was calculated with an accuracy of 5%, a confidence level of 95%, and a sensitivity and specificity of 85.7 and 75%, respectively. An estimated prevalence of patients with uncontrolled BP of 35% (n = 141) was determined. From this figure, we calculated a percentage of expected losses of 15%, leading to a sample size of n = 153. The calculation of the sample size was performed using the online calculator by Fisterra [12].
Sampling method
A systematic random sampling procedure was carried out. The first subject was chosen at random, and the sampling fraction used was 1/10 patients. The patients were recruited through telephone contact or through their doctors, for those who went to the health centre during the patient selection period. In the case of negative or no contact after several attempts, the next patient was chosen from the list. The recruitment flowchart is shown in Fig. 1.
After giving consent, each selected patient who agreed to participate was scheduled for an appointment at 8:30 a.m. to perform the ABPM. The perimeter of the patient’s arm was measured, and the appropriate cuff was provided. If the arm circumference was greater than 32 cm, a large cuff was provided. The blood pressure was measured in both arms, and the non-dominant arm was chosen as the measuring arm. If both measurements were equal, the cuff was placed on the left arm for right-handed patients or on the right arm for left-handed patients. After the placement of the ABPM device, a forced measurement was immediately carried out. The recording began in the morning at the time the device was placed.
The patients were instructed to perform normal daily activities, except that when the cuff warned that a measurement was about to begin, the arm should be kept in a relaxed position. The patients were asked to return the next day at the same time.
The programming of the device was as follows: Frequency of readings: every 15 min during daytime and every 30 min while sleeping. For this, the patient was questioned when placing the device about their sleep schedule, which was confirmed and readjusted the next day when the device was removed. With these data, the program calculated the beginning of the night or sleep period and the day or activity period for the purposes of the analysis. Types of recordings: Measurement of SBP, DBP and HR over 24 h, daytime and evening. Measurement range: HR: 40 to 180 beats per minute. Pressure: 70 to 285 mmHg for systolic; 40 to 200 mmHg for diastolic and 60 to 240 mmHg for mean blood pressure values. Valid record criteria: 70% valid measurements on the ABPM, more than 14 valid measurements of systolic and diastolic BPs during the day and more than 7 measurements of systolic and diastolic BPs during the night.
The next day, the ABPM device was removed and the patient was shown how to use the validated semi-automatic tensiometer to obtain SBPM records. The subject was seated, and after a period of 5 min of rest, 2 measurements were taken from the dominant arm with a 1-min interval between them. A third measurement was taken if the first 2 had a difference of greater than 5 mmHg.
After the demonstration, the patient was instructed in the proper handling of the device, specifically that the measurements should be taken while sitting, at rest, with the cuff placed on the arm that showed the highest BP. The patient was also instructed on how to place the cuff, which was to be 2–3 cm above the flexure of the elbow, its fit on the arm, and the position at which to take the pressure reading.
The correct performance of the first 2 measurements on the first day was verified by the patient in front of the researcher. Then, the blood pressure monitor was given to the patient, who was instructed in writing to take 2 measurements in the morning (between 7:00 and 10:00 a.m.) and 2 before going to bed (between 21:00 and 23:00 p.m.) with 1–2 min between measurements for 3 consecutive days. A total of 12 measurements were obtained for each patient (the first day measurements were later discarded), establishing the 3-day SBPM pattern. The instruments used in this study included 2 ABPM devices Microlife Watch BP 03 (Microlife, Widnau, Switzerland) [13] and 10 automatic arm blood pressure monitors for SMBP (Microlife Watchbp Home) that were validated according to the standards of the Spanish Hypertension Society (SEH) and the British Hypertension Society (BHSOC) (accessible on January 11, 2016 at: https://www.seh-lelha.org/microlife-watchbp-o3/ and https://bihsoc.org/bp-monitors/for-home-use/).
Regarding the definitions of the variables applied in this study, we defined poor HT control as when the mean BP measured by ABPM over 24 h was greater than 130/80 mmHg. For the 3-day SMBP, we considered uncontrolled average blood pressure as greater than 135/85 mmHg.
This descriptive study was carried out with the demographic variables of age, sex and work status, taking into account the records of spanish administrative document (TSI) that accredits access to citizens to health care benefits. If the patient was not active: TSI < 002 (Without income, retired) or active: TSI 003 (income level < € 18,000/year); TSI > 4 (income level > € 18,000/year). Clinical variables included associated comorbidities such as a diagnoses of dyslipidaemia, diabetes mellitus, chronic kidney disease, atrial fibrillation, and stroke.
Statistical analysis
The statistical programs SPSS version 22 and Epidat version 3.1 were used to analyse the data.
Categorical variables are presented as absolute frequencies (%) and quantitative variables as the means and standard deviations. Values of p < 0.05 were considered statistically significant, and 95% confidence intervals (95%CI) were calculated. The sources of the data were the BP records collected by SBPM and ABPM as well as the electronic health records of the patients in the OMI-AP program.
Test validation analysis
A comparison of the data collected using the 3-day SBPM (discarding measurements from the first day) was performed using the figures obtained by the 24-h ABPM as the reference standard. Given the results obtained by De León-Robert et al. [10], the 3-day pattern was chosen (SBPM-DAYS-2&3) and an optimal cut-off point was obtained. The calculations of the predictive capacity of the 3-day protocol with the new cut-off point were performed using a statistical calculator. With the data obtained, the cases with misclassifications (false positives and negatives) in the SBPM-DAYS-2&3 data were identified at the cut-off point where the sensitivity and specificity of the diagnosis of HT were optimized, using the 24-h ABPM as the gold standard (130/80 mmHg). An analysis of how sociodemographic and clinical factors related to errors in diagnostic classifications was performed. Calculations were performed for the crude and adjusted odds ratios, according to sex, of the diagnostic classification errors for both systolic and diastolic HT. These were performed using logistic regression models that included age, income level, and clinical comorbidities such as diabetes, dyslipidaemia, chronic kidney disease, ictus, and atrial fibrillation as predictor variables. The estimating of the agreement between ABPM and SBPM readings were perfomed using the Bland-Altman plot analysis and the concordance was determinated using intraclass correlation coefficient (ICC).