Ordinal logistic regression analysis of hypertension data and identifying factors that influence the incidence of hypertension in Gondar University Teaching Hospital, North- west Ethiopia

Hypertension is the force of blood pushing up against the walls of blood vessels and it is defined as a systolic blood pressure ≥ 140 mm Hg and/or diastolic blood pressure ≥ 90 mm Hg. It is estimated to cause 9.4 million deaths every year. This study aims to conduct ordinal logistic regression analysis of hypertension severity and identifying associated factors for severity of hypertension, using the data of Gondar Univeristy Referral Hospital, North west Ethiopia. The cumulative logit or the proportional odds regression model is one of the popular choices used to study covariate effects on ordinal responses. The proportional odds model was used to model the severity of hypertension and to identify associated factors for its severity. The fit of the model was illustrated with 420 hypertension records obtained from Gondar University Referral Teaching Hospital (in the chronic illness medication and follow up clinic). Findings of this study show that age of patients (OR= 2.195, P = 0.0051), residence (OR= 1.954, P = 0.0008), alcohol use (OR= 2.329, P = 0.0027), using salty foods (OR= 1.970, P = 0.0007), habit of using vegetable/fruit (OR= 0.516, P = 0.0010), doing physical exercise (OR= 0.571, P = 0.0062) and BMI (OR= 3.293, P <0.0001) had significant association with severity of hypertension. Polices and plans have to be put in place to motivate vegetable/fruit users, to decrease the consumption of salty foods, non-alcohol drinkers, normal body mass index and concerned bodies need to implement hypertension prevention strategies.


INTRODUCTION
Hypertension is the force of blood pushing up against the walls of blood vessels and it is defined as a systolic blood pressure ≥ 140 mm Hg and/or diastolic blood pressure ≥ 90 mm Hg. The risk of dying from high blood pressure in low-and middle-income countries is more than double that in high-income countries. In the high income countries, only 7% of deaths caused by high blood pressure occur under age 60; in the African Region, this has increased to 25% (WHO, 2010). Non-Communicable Diseases (including hypertension) are becoming an increasingly important public health problem as epidemiological transition is progressing in Ethiopia (Awoke et al, 2014).World Health Organization reported that the prevalence of hypertension in Ethiopia was 35.2% for the crude adjusted estimates of both sexes (WHO, 2010).
Hypertension is the growing public health problem in many developing countries including Ethiopia. However, its prevention, early diagnose and control have not received greater attention. Hypertension does not have its own signs and symptoms like other chronic disease and its treatments are not given on time. It causes many deaths before it is treated, and it is usually known as silent killer disease. Hypertension increases the risk of ischemic heart disease, stroke, peripheral vascular disease and other cardiovascular diseases, and is also a risk factor for cognitive impairment and dementia and chronic kidney disease (Assefa et al., 2014). In this study, by classifying the severity of hypertension as severe, moderate and mild ordinal logistic regression model was applied to assess the effect of multiple covariates in different levels of hypertension severity.

Description of the hypertension data
The target population of this study was hypertensive patients at Gondar University Hospital chronic illness medication and follow up clinic; their age is age ≥18 years. Theyw came to this clinic before April 30, (2016) for hypertension treatment (continuation or checkup). So the data were collected from February 1 to April 30, 2016 for three months. The total numbers of patients that started hypertension follow up treatment were 482. From the total patients, only 420 patients had a complete outcome and were available at this three month of data collection period. 29 patients did not come to the clinic during their appointment (check-up) time and the charts of 33 patients were not recorded in the data properly; hence they are excluded from the study.

Inclusion and exclusion criteria
The study considers all hypertensive diseased patients in chronic illness medication and follow up clinic; their age is ≥18 years, and they started hypertensive follow up treatment before April 30, 2016. They came for check-up during their appointment time for three months. However, the study excludes those patients on hypertensive follow up treatment whose age < 18 years and patients who did not come to the clinic during these three months; patients who had an appointment time of more than three months. Disabled hypertensive patients were also excluded from the study.

Variables of the study
The dependent variable for this study is the severity of hypertension which is reviewed from their medical charts. This includes mild, moderate and severe level of hypertension. The severity (level) of hypertension was classified from the blood pressure measurement of two numbers systolic and diastolic reading in mmHg and according to the World Health Organization criteria for severity of hypertension classified as mild (systolic 140-159, diastolic 90-99), moderate (systolic 150-179, diastolic 100-109) and severe (systolic ≥180, diastolic ≥100) [WHO, 1999].
This study considers many explanatory variables, including socio-demographic variables (Age (in year), sex (M, F), marital status (Single, Married, Others), level of education (Illiterate, Informal education, Formal education)), lifestyle factors (Diet, alcohol use, smoking, physical activity), co-morbidity (Obesity, kidney Disease, Diabetes mellitus) and genetic and family related factors (Family history Hypertension). The duration that these respondents have been hypertensive was not taken into consideration as independent variable for this analysis.

Ordinal logistic regression model
SAS statistical software was used for the analysis. The issue of correlation among independent variables (multicolinearity) has been checked by using tolerance and variance inflation factor (VIF) method. Logistic regression model can be classified as multinomial, ordinal and binary. In this investigation ordinal logistic regression model was used.The ordinal logistic regression procedure empowers one to select the predictive model for ordered dependent variables. Logistic regression model can be classified as multinomial, ordinal and binary. In this investigation ordinal logistic regression model was used. The ordinal logistic regression procedure empowers one to select the predictive model for ordered dependent variables. The ordinal logistic regression model describes the relationship of an ordered response variable and a set of explanatory variables. The explanatory variables may be continuous or discrete (or any type). The explanatory variables may be continuous or discrete (or any type).
Ordinal response models have major importance in social sciences as well as demography and many social phenomena. The responses are discrete or qualitative rather than continuous or quantitative in nature. Hypertension is the continuous variable, but here it has been changed of ordinal by classifying as severe, moderate and mild. Many such analyses involve an outcome or dependent variable that is ordinal and in this study the logistic regression model was the statistical model of choice. The most popular model in ordinal logistic is the proportional odds model.

Proportional Odds (PO) model
Proportional odds model is used as a tool to model the ordinal nature of a dependent variable by defining the cumulative probabilities differently instead of considering the probability of an individual event. It considers the probability of that event and all events that are ordered before it. When response categories are ordered, logits can directly incorporate the ordering.

Uni-variable ordinal logistic regression analysis
Nineteen variables were selected and 19 uni-variable ordinal logistic regression models were developed to assess the relationship between severity of hypertension and the selected variables. Any variable whose uni-variable test has a p-value≤ 0.25 will be a candidate for Multiple Ordinal Logistic Regression Analysis.

Multiple ordinals logistic regression analysis
Once the variables were identified by uni-variable ordinal logistic regression analysis, we began with a model containing all of the

RESULTS
This study used hypertension data collected from hypertension diseased patients linked to Gondar University Hospital chronic illness medication and follow up clinic. Four hundred twenty hypertension diseased patients were included. From the total hypertension diseased patients' data 44.5% (187) were severe, 35.2% (148) were moderate and 20.2% (85) were mild (Table 1).   The result reveals that the goodness of fit test is fulfilled, which is confirmed by both the deviance and Pearson with P-values of 0.2124 and 0.6325, respectively. Consequently, the observed and expected cell counts  are similar and the value of each statistic is small ( Table  7). As shown in Table 8, the tolerance of all the variables greater than 0.1 implies there is no a problem of collinearity between the dependent and independent variables. Moreover, the VIF values are greater than 10, which indicates there is no problem of mulitcollinearity. As shown in Table 9, the tolerance of all the variables greater than 0.1 implies there is no a problem of multicollinearity between age and other independent variables.Moreover, the VIF values are greater than 10, which indicates there is no a problem of mulitcollinearity between age and other independ variables.Similarily, we can continoue for other independent variables.

DISCUSSION
The study shows that there is significant association between age of patients and the severity of hypertension. Older age group patients (that is, 60 and above), who develop hypertension belong to "severe hypertension" category and they are also liable to high prevalence of hypertension. The result is consistent with Mulugeta (2015), Assefa et al. (2014), Kassawmar et al. (2015), Mungati et al. (2014), Erhun et al. (2005), Marwiro (2012) and Kumar et al. (2002) that age [being aged] has significant effect on the occurrence of hypertension. This study also agrees with national academy on an aging society.
There is a significant difference in developing hypertension between urban and rural residences. This shows that people living in urban areas have a high risk of developing hypertension than people living in ruralareas. The study result is consistent with Iwelunmor et al. (2014). The study shows that there is significant association between alcohol drinking habit of patients and the severity of hypertension. Current alcohol drinker individuals are more likely to develop severe hypertension than never drinker individuals. This implies most current alcohol drinker individuals who develop hypertension belong to "severe HTN" category and they are also liable to high prevalence of hypertension. This result of the study agrees with that of Mulugeta (2015), Erhun et al. (2005), Birhanu (2014) and Kumar et al (2002); alcohol drinking was a significant factor of hypertension. There is high significant difference between individuals with excessive salty food user and non-users in terms of developing hypertension. That is, individuals who are salty food users are more likely to develop "severe hypertension" than those non-users. In addition, the result reveals that the overall prevalence of hypertension is high with those individuals who are salty food users. This result of the study agrees with that of Seid (2014), Mulugeta (2015) and Mizumoto (2004); use of excessive salty foods was the significant factor of hypertension.
There is high significant difference in developing hypertension between vegetable/ fruit users and nonusers. That is, individuals who are vegetable or fruit users are less likely to develop "severe hypertension" than nonusers. In addition, the result reveals that the overall prevalence of hypertension is high with those individuals who are non-users.The study result is consistent with the result of Mulugeta (2015) that use of fruit or vegetable was significantly associated with hypertension. Consequently, vegetable or fruit users are less susceptible to develop severe hypertension.
There is a high significant difference in developing hypertension between individuals who are physically active and not physically active individuals. Among all hypertension developed in physically inactive individuals most of them belong to "severe hypertension" category and it is also responsible for high incidence of hypertension. This shows people who are physically inactive are at a high risk of hypertension than physically active individuals. The study result is consistent with Seid (2014), Mulugeta (2015), Akilew et al. (2012) and Kumar et al. (2002) that not doing physical activity is significantly associated with hypertension.
The study shows that there is significant association between body mass index of patients and the severity of hypertension. Obese individuals are more likely to develop severe hypertension than normal/underweight body mass index individuals. This implies most "obese body mass index" individuals who develop hypertension belong to "severe hypertension" category and they are also liable to high incidence of hypertension. The study result is consistent with Seid (2014), . Mulugeta (2015), Henok (2014), Assefa et al. (2014), Kassawmar et al. (2015), Jenei et al. (2002), Erhun et al. (2005), Akilew et al. (2012), Marwiro (2012), Mizumoto (2004), Birhanu (2014), Jasmine et al. (2013), Kumar et al. (2002) that body mass index was a significant factor of hypertension.

Conclusions
The study examined the socio-demographic and risk factor determinants of hypertension in GoUTH. Results of proportional odds model show that socio-demographic and risk factor variables are very important for determinant of hypertension outcome. The finding of the study shows that different factors such as patients' age, Residence, Alcohol use, using salty foods, habit of using vegetable/fruit, Doing physical exercise and BMI have statistically significant effect on the outcome of hypertension.

RECOMMENDATION
Based on the findings of this study, the following issues should be considered and promoted for improving the prevention of hypertensive condition in community people. The government and concerned bodies should work on the preventive strategies of hypertension including promoting and advocating healthy lifestyles, by improvement of dietary habits, eradication of alcohol consumption, decreasing the consumption of salty foods, increasing physical activity and encouraging vegetable or fruit users.

Limitation of the study
One of the limitations of this study is all patients who were proposed to be included in the study did not come to the clinic during the three months of data collection time and hence the proposed sample size decreased. The other major limitation of this study is related to the problem of the availability of complete secondary data. Patients did not volunteer to give response during primary data collection. Some variables were not included in the study due to lack of equipment and limitation of resource or budget (like waist-to-hip ratio).