Computational technology of daily blood pressure monitoring results comparison in patients with arterial hypertension

Material and methods. The open controlled study included 10 patients with AH at the third stage in long term after a cerebral stroke. Patients were treated according to the regulatory basis. DBPM was conducted using device АВРМ-01 (Meditech, Hungary) twice (before the treatment and after 1 year). Results comparison of the first and repeated DMBP was carried out by means of proposed computational technology and its own software ArtHyper in which the technology was implemented. The technology is based on the comparison of DMBP indices distribution functions using the two-sample Kolmogorov test and the Wilcoxon test.

Daily monitoring of blood pressure (DМBP) is a modern and effective method used for diagnosis and evaluation the effectiveness of arterial hypertension treatment (AH) [2,[4][5][6]. Currently, the methods and software have been developed for analysis of DМBP results with the purpose of arterial hypertension diagnosis [3,4,7,8]. However, the treatment effectiveness and the further patient's condition assessment require blood pressure re-monitoring and comparison of these results with previously obtained ones. Usually the comparison is limited to DМBP indices values comparison. But differences in the values can be insignificant and caused by chance. So, such kind of comparison cannot be considered as acceptable. This work is aimed at developing a more effective and objective technology in order to compare a patient's two DМBP results.

Purpose
Purpose of the work involves development, software implementation and practical approval of computational technology in order to compare the results of patient's two daily monitoring of blood pressure.

Materials and methods
10 patients were examined in the State Institution "Ukrainian State Scientific and Research Institute of Medical and Social Problems of Disability". Inclusion criterion for study was essential hypertension at the 3 rd stage in patients aged 30-59 years. Exclusion criterions were secondary form of arterial hypertension, renal artery stenosis, permanent atrial fibrillation, severe disturbances of liver and kidneys functions, myocardial infarction, cardiosurgical interventions, uncompensated diabetes, presence of comorbidities that could affect the survival and quality of patients life within 3 months, treatment refusal. All patients received information for patients and signed an agreement on consent to participate in the study.
Arterial DMBP was conducted in all patients by means of the conventional method using device АВРМ-01 (Meditech, Hungary) [4][5][6]. During the day five parameters were recorded: systolic blood pressure (SBP), diastolic blood pressure (DBP), pulse blood pressure (PBP), average blood pressure (BP) and heart rate (HR). The measurements were carried out in the following way: during the day -every 15 minutes and at night (from 22.00 to 6.00) -every 30 minutes. Patients led a normal lifestyle and experienced regular physical and psycho-emotional stresses [4][5][6].
One year later DBPM was carried out using the same device.
The question arises as to whether patients' condition has changed in a year, and whether medical treatment was effective.
The computational technology to compare the first and second DМBP patient's results was suggested in order to have this question answered.
The following considerations were used in the technology development. DМBP indices for processing are not constant throughout the day. Firstly, they fluctuate due to physiological factors, for example, blood pressure at night is always lower, and secondly, they are influenced by random external factors such as stress, weather changes and so on. In view of the random effects, DМBP indices can be recognized as random values and hereby it is possible to consider their probability distribution. Then, in order to check the changes in DМBP indices of the patient after re-examination, it is advisable to check whether the distribution function changes. These considerations make it possible to suggest the computational technology to compare the distribution functions of the DМBP indices.
For each of the five DМBP indices the computational technology involves the following steps: 1. To build an empirical distribution functions [1,9] for the selected index according to the information of the 1 st and 2 nd blood pressure monitoring.
2. To compare the distribution functions of the indices during the 1 st and 2 nd monitoring by means of the two-sample Kolmogorov test [1,9]. If provided by the given significance level a inequality P < a holds true, it can be argued that distribution function of the index during the 2 nd monitoring Оригинальные исследования significantly differs from the function that took place during the 1 st monitoring. In this case one shall proceed to step 3. Otherwise, it is assumed that the distribution function doesn't change.
3. To check whether the differences in the distribution functions of indices are explained by the shift and, if so, the distribution function for the 2 nd monitoring is shifted left or right relative to the distribution function for the 1 st monitoring. For this purpose it is possible to apply the non-parametric statistical test such as the Wilcoxon test [1,9].
The proposed computational technology is implemented in its own software ArtHyper written in the C# programming language in the environment Microsoft Visual Studio 2012.

Results and discussion
By means of ArtHyper software the practical testing of the proposed technology on real data was conducted. As an example, the results are provided below for two patients -one, who followed the prescribed medical treatment, and the other, who didn't follow it completely. During the application of all the tests, a = 0.05 was used. The empirical distribution functions of the SBP, DBP, PBP, BP, HR were constructed for the patient who followed the prescribed medical treatment in accordance with the 1 st and repeated DМBP, see Fig. 1.
According to the two-sample Kolmogorov test results, the distribution functions of PBP for two monitoring processes differed significantly and for the other indices Original research they were equal ( Table 1). According to the Wilcoxon test results ( Table 2), the distribution function of PBP obtained during the second DMAT shifted left as compared to another that took place during the first patient's examination. This indicated that the PBP level had declined in a year. Consequently, most of the DMAT indices distribution for this patient did not change in a year, in other words, the blood pressure and heart rate of the patient remained stable due to the treatment. The level of PBP slightly decreased, but in this case it was a positive factor.
For the patient who didn't follow the complete course of treatment the empirical distribution functions of the SBP, DBP, PBP, BP, HR were constructed according to the 1 st and repeated DMAT, see Fig. 2.
Results of the indices distribution comparison by means of the two-sample Kolmogorov test suggest that the distribution of SBP in a year remained unchanged and the distribution of four other indices changed ( Table  3). Herewith in accordance with the Wilcoxon test results ( Table 4), we might as well say that the distribution functions of DBP, BP and HR in a year turned out to be shifted right in comparison with those that occurred at the beginning. Consequently, the level of these indices increased in a year. The distribution function of PBP turned out to be shifted left, that sounds logically in this case.
Thus in a year the patient's condition became worse. The levels of diastolic blood pressure and heart rate increased.

Оригинальные исследования
A similar study was conducted for another 8 patients. The conclusions regarding the condition of all 10 patients derived from the application of the proposed technology of the 1 st and repeated DMBP results comparison were confirmed by an additional medical patients' examination. The conclusions were confirmed to be correct and justified. This suggests that the technology may be effective for the evaluation of rehabilitation and treatment efficiency.

Conclusions
The computational technology of two patient's DMBP results comparison was developed and implemented in its own software ArtHyper. The technology is based on distribution functions of the DMBP indices comparison using the twosample Kolmogorov test and the Wilcoxon test. Its practical testing was carried out on real data and demonstrated the effectiveness of technology to detect changes in the second results monitoring and the possibility of its application for the evaluation of rehabilitation and treatment.
Further researches can be directed to carry out computational technology practical testing for a large number of patients.