Aplication of Hierarchical Linear Model for Predicting Systolic Blood Pressure (Study on The Brides-to-be in Probolinggo District, East Java Province, Indonesia)
Mahmudah*, Sri Sumarmi* and Nunik Puspitasari*
*Faculty of Public Health, Airlangga University, Indonesia
*Faculty of Public Health, Airlangga University, Indonesia
Introduction :
Health research is often conducted in a wide geographical area of study. The study area is usually divided into several levels or hierarchical structure. This condition causes relatively homogeneous characteristics at the same level and relatively heterogeneous at different levels. Therefore it is important to use statistical analysis that takes into account the variation between levels. Hierarchical Linear Models (HLM) or Multilevel Linear Models is a useful statistical method for data analysis in a hierarchical structure. The objective of this study is to apply Hierarchical Linear Models (HLM) in health research.
Methods :
The dependent variable is systolic blood pressure, whereas independent variables are Body Mass Index (BMI) and Waist to Hip Ratio (WHR). The data were taken from 642 brides-to-be in Probolinggo district, which include systolic blood pressure, BMI and WHR. HLM is employed in 2- level. Level-1is the individual level obtained from the measurement of each bride-to-be. Level-2 is a sub-district level taken from the average BMI and WHR of individual measurement for each subdistrict. The best model is determined by comparing the deviance, AIC and BIC. The source of data analyzed are taken from the other research (Sri Sumarmi et al, 2014)
Results :
The results showed that the ICC was 47.96%. This ICC value is quite large. There is indication that district variation has to be controlled in model building, as well as the influence of variables at district level. The results indicate that systolic blood pressure is influenced by BMI (individual level) and WHR (sub-district level).
Conclusion :
The best model is systolic blood pressure = 0.142358 (BMI) +127.703406 (average WHR).
Keywords : Hierarchical Linear Models, Intraclass Correlation, Systolic Blood Pressure, Body Mass Index, Waist to Hip Ratio.
Paper presented at APACPH 46th, Kuala Lumpur, Malaysia, October 2014.
Health research is often conducted in a wide geographical area of study. The study area is usually divided into several levels or hierarchical structure. This condition causes relatively homogeneous characteristics at the same level and relatively heterogeneous at different levels. Therefore it is important to use statistical analysis that takes into account the variation between levels. Hierarchical Linear Models (HLM) or Multilevel Linear Models is a useful statistical method for data analysis in a hierarchical structure. The objective of this study is to apply Hierarchical Linear Models (HLM) in health research.
Methods :
The dependent variable is systolic blood pressure, whereas independent variables are Body Mass Index (BMI) and Waist to Hip Ratio (WHR). The data were taken from 642 brides-to-be in Probolinggo district, which include systolic blood pressure, BMI and WHR. HLM is employed in 2- level. Level-1is the individual level obtained from the measurement of each bride-to-be. Level-2 is a sub-district level taken from the average BMI and WHR of individual measurement for each subdistrict. The best model is determined by comparing the deviance, AIC and BIC. The source of data analyzed are taken from the other research (Sri Sumarmi et al, 2014)
Results :
The results showed that the ICC was 47.96%. This ICC value is quite large. There is indication that district variation has to be controlled in model building, as well as the influence of variables at district level. The results indicate that systolic blood pressure is influenced by BMI (individual level) and WHR (sub-district level).
Conclusion :
The best model is systolic blood pressure = 0.142358 (BMI) +127.703406 (average WHR).
Keywords : Hierarchical Linear Models, Intraclass Correlation, Systolic Blood Pressure, Body Mass Index, Waist to Hip Ratio.
Paper presented at APACPH 46th, Kuala Lumpur, Malaysia, October 2014.
