An integrated latent construct modeling framework for predicting physical activity engagement and health outcomes
2014
The health and well-being of individuals is related to their activity-travel patterns. Indeed, there is increasing interest in drawing connections between activity-travel indicators and public health outcomes. Many of the indicators related to physical activity participation, sedentary activity participation (such as watching television or sitting at the computer for extended periods), and extent of bicycling and walking are measures that public health professionals would be interested in connecting to health outcomes such as body mass index (BMI), blood pressure, and overall state of health. Activity-based travel demand models are able to predict activity-travel patterns of individuals at a high degree of fidelity, thus providing rich information for transportation and public health professionals to infer health outcomes that may be experienced by individuals in various geographic and demographic market segments. However, despite the widespread recognition of the importance of attitudes and lifestyle preferences on activity engagement patterns and mode use, activity-based models fail to include such variables in the model specification. Engagement in physical activities, and the use of bicycle and walk modes, are likely to be influenced by the lifestyle preferences and attitudes of individuals. But, such lifestyle preferences and attitudes are rarely, if ever, measured in surveys rendering it difficult to explicitly include such measures in activity model specifications. This study constitutes an initial attempt to fill this gap by adopting Bhat's (2014) Generalized Heterogeneous Data Model (GHDM) model system in which latent constructs that describe an individual's health consciousness and physical activity propensity are modeled as a function of observed socio-economic and demographic characteristics. The resulting latent constructs, together with socio-economic and demographic variables, are then used to predict a number of activity engagement outcomes (describing frequency and duration of participation in various types of activities - both physically active and sedentary) and health outcomes (such as body mass index, self-reported health well-being, and blood pressure). The entire system of equations is estimated jointly through the use of the maximum approximate composite marginal likelihood (MACML) estimation approach that greatly simplifies the evaluation of the likelihood function and brings about computational efficiency in the estimation of simultaneous equations model systems that involve a mixture of dependent variable types. The model system is applied using the 2005-2006 National Health and Nutrition Examination Survey (NHANES) sample. The findings of the paper show that latent constructs, health consciousness and physical activity propensity, are related to socio-economic and demographic variables. These latent constructs play a significant role in shaping activity-travel and mode use patterns, with those who are more health conscious or inclined towards physically active lifestyles reporting higher levels of physical activity engagement and better health outcomes. Given the significance of the latent variables in explaining activity engagement and mode use, activity-based microsimulation models may be enriched in terms of the model specification through the inclusion of such latent variables that are themselves functions of observed socio-economic and demographic variables collected in travel surveys. There has been a reluctance historically to include attitudinal and lifestyle preference variables in model specifications because such variables are not typically measured in travel surveys, and more importantly, they are difficult to forecast into the future. However, the approach proposed in this paper, where latent variables are functions of observed variables and can be included in models of activity-travel behavior, offers a mechanism by which such latent attitudinal and lifestyle constructs can be included in models of activity-travel demand. The study is not without its limitations. Due to the nature of the study, the survey data set used for this modeling effort had to include both activity-travel indicators as well as health indicators. The 2005-2006 National Health and Nutrition Examination Survey (NHANES) offered such a data set, but this data set suffered from the drawback that it did not include any built environment, contextual, or network level of service variables - all of which invariably affect activity-travel indicators and possibly health outcomes as well. Future research and data collection efforts should attempt to include all of the variables of interest so that contextual variables may be accounted for in the model specification. Normal 0 false false false EN-US KO X-NONE /* Style Definitions */
table.MsoNormalTable
{mso-style-name:"Table Normal";
mso-tstyle-rowband-size:0;
mso-tstyle-colband-size:0;
mso-style-noshow:yes;
mso-style-priority:99;
mso-style-parent:"";
mso-padding-alt:0in 5.4pt 0in 5.4pt;
mso-para-margin-top:0in;
mso-para-margin-right:0in;
mso-para-margin-bottom:10.0pt;
mso-para-margin-left:0in;
line-height:115%;
mso-pagination:widow-orphan;
font-size:10.0pt;
mso-bidi-font-size:11.0pt;
font-family:"Times New Roman","serif";
mso-bidi-font-family:"Times New Roman";
mso-bidi-theme-font:minor-bidi;}
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
25
References
0
Citations
NaN
KQI