BACKGROUND: Analytical statistics revealed a variety of risk factors for hypertension, but the complex interplay between different factors remains to be determined by more powerful statistical techniques. METHODS: Analytical as well as new, explorative statistical methods such as natural segmentation (k-means) and predictive modelling algorithms (C4.5) were used to classify the interactions of the individual risk factors for arterial hypertension in a large cohort of subjects. Fifty-five attributes (subject base, sociodemographic, medical history, laboratory data) were obtained from each of the 3547 participants of a community-based health survey. The study subjects, mean age of 41 years, were free of major comorbidity. RESULTS: Twenty-five percent of the subjects had at least stage 1 hypertension. No clear linear dependency of risk factors with the diagnosis hypertension could be derived by the analytical statistics. In particular, the mutual amplification of different risk factors towards hypertension could not be revealed by these techniques. Explorative analytics however, uncovered body mass index (BMI) as the main single risk factor associated with hypertension. High predictive accuracy was achieved when combinations of certain risk factors including male gender and age were used. CONCLUSIONS: In summary, the survey of risk factors for hypertension using explorative analytics yielded high increases for the correct prediction of arterial hypertension. In this cohort, BMI was the single strongest parameter associated with arterial hypertension.
BACKGROUND: Analytical statistics revealed a variety of risk factors for hypertension, but the complex interplay between different factors remains to be determined by more powerful statistical techniques. METHODS: Analytical as well as new, explorative statistical methods such as natural segmentation (k-means) and predictive modelling algorithms (C4.5) were used to classify the interactions of the individual risk factors for arterial hypertension in a large cohort of subjects. Fifty-five attributes (subject base, sociodemographic, medical history, laboratory data) were obtained from each of the 3547 participants of a community-based health survey. The study subjects, mean age of 41 years, were free of major comorbidity. RESULTS: Twenty-five percent of the subjects had at least stage 1 hypertension. No clear linear dependency of risk factors with the diagnosis hypertension could be derived by the analytical statistics. In particular, the mutual amplification of different risk factors towards hypertension could not be revealed by these techniques. Explorative analytics however, uncovered body mass index (BMI) as the main single risk factor associated with hypertension. High predictive accuracy was achieved when combinations of certain risk factors including male gender and age were used. CONCLUSIONS: In summary, the survey of risk factors for hypertension using explorative analytics yielded high increases for the correct prediction of arterial hypertension. In this cohort, BMI was the single strongest parameter associated with arterial hypertension.
Authors: Jimeng Sun; Candace D McNaughton; Ping Zhang; Adam Perer; Aris Gkoulalas-Divanis; Joshua C Denny; Jacqueline Kirby; Thomas Lasko; Alexander Saip; Bradley A Malin Journal: J Am Med Inform Assoc Date: 2013-09-17 Impact factor: 4.497