Joint modeling of binary longitudinal measurement and time-to-event: An application to depression and time-to-dementia
| dc.contributor.advisor | Lim, Hyun | |
| dc.contributor.committeeMember | Li, Longhai | |
| dc.contributor.committeeMember | Feng, Cindy | |
| dc.contributor.committeeMember | Huq, Mobinul | |
| dc.contributor.committeeMember | Mondal, Prosanta | |
| dc.creator | Kabir, Md Rasel | |
| dc.date.accessioned | 2020-05-20T21:07:14Z | |
| dc.date.available | 2021-05-20T06:05:10Z | |
| dc.date.created | 2020-05 | |
| dc.date.issued | 2020-05-20 | |
| dc.date.submitted | May 2020 | |
| dc.date.updated | 2020-05-20T21:07:15Z | |
| dc.description.abstract | In recent years, the methodological development of joint models of longitudinal and time-to-event data has become one of the most popular areas of studies in clinical research and its application has increased substantially over the past decades. Joint model in this area combines both the longitudinal and survival data into a single statistical model to obtain robust estimates and draw valid inference. While most of studies concentrate on continuous longitudinal measurements, little attention has been paid to joint modeling for binary longitudinal outcome and event time data. In clinical research, patients often have binary longitudinal measurement that affects the main event of interest during the follow-up time. For example, depression, a dichotomous longitudinal measurement, might have relationship with dementia. However, no study has examined this association using a joint model. This study focuses on the joint modeling technique for binary repeated measurement and time-to-event data. This approach mainly models the longitudinal and survival processes for each individual through a shared random effect jointly, where the longitudinal part is supposed to be modeled by a generalized linear mixed model and time-to-event component is characterized by employing a parametric survival model. We applied the joint modeling technique to the Korean Health Panel Study. A generalized linear mixed model was used to model the binary repeated measurements of depression and a piecewise constant hazard model was employed for time-to-dementia. A total of 3,611 individuals aged 65 years or older were eligible for this study between 2008 and 2015. Depression and dementia were identified by the diagnosis code in medical data. In this study, 215 (6%) were diagnosed with dementia during the 8-year follow-up period. The mean age at entry was 72.2 (±5.7) years. The overall median follow-up time was 5.8 years; 3.6 years for people living with dementia compared to 5.9 years for people without dementia. Baseline depression and sex were not significantly associated with time-to-dementia. However, time-varying depression and baseline covariates including age, economic activity, education, walking frequency/week, living with other family members and diabetes were significant in multivariable joint modeling. The risk of dementia was 2.4 times (95% CI: 1.30-4.50, p-value = 0.005) higher among depressed people compared to non-depressed people. This study also found that walking not at all or less than three days a week, being older (>70 years old), having diabetes, being less educated and living in a household with multiple generations increased the risk of dementia. | |
| dc.format.mimetype | application/pdf | |
| dc.identifier.uri | http://hdl.handle.net/10388/12850 | |
| dc.subject | Joint modeling | |
| dc.subject | Binary repeated measurement | |
| dc.subject | Time-to-event | |
| dc.title | Joint modeling of binary longitudinal measurement and time-to-event: An application to depression and time-to-dementia | |
| dc.type | Thesis | |
| dc.type.material | text | |
| local.embargo.terms | 2021-05-20 | |
| thesis.degree.department | School of Public Health | |
| thesis.degree.discipline | Biostatistics | |
| thesis.degree.grantor | University of Saskatchewan | |
| thesis.degree.level | Masters | |
| thesis.degree.name | Master of Science (M.Sc.) |