There are many active research projects accessing and applying shared ADNI data. Use the search above to find specific research focuses on the active ADNI investigations. This information is requested annually as a requirement for data access.
| Principal Investigator | |
| Principal Investigator's Name: | David Madigan |
| Institution: | Columbia University |
| Department: | Statistics |
| Country: | |
| Proposed Analysis: | Hypothesis and Specific Aims: Hypothesis: The δ homolog for dementia case finding described in Royall and Palmer (2019) can be replicated in the open source R software system. Specific Aim 1. Replicate the structural equation model in Royall and Palmer using the open source “sem” package in R. Specific Aim 2. Use the “JAGS” software package in R to develop a Bayesian version of the Royall and Palmer model. Background and Significance: Royall and Palmer (2019) developed the “dT2A” measure in ADNI. It is indicated by a battery of six cognitive measures, i.e., the Boston Naming Test (BNT), Category Fluency (Animals), Logical Memory I (LMI) and II (LMII), the Mini-Mental Status Examination (MMSE), and Trial-Making Part B (TrailsB). dT2A’s target indicator in ADNI is the Functional Activities Questionnaire (FAQ). The dT2A composite is strongly associated with CDR-SB in ADNI and achieves a high AUC for dementia’s diagnosis (i.e., 0.988 (0.983–0.993) in ADNI). dT2A has tremendous potential importance as a simple predictor of dementia. This has many potential applications including identification of subjects for clinical trials. dT2A was developed using the commercial software AMOS. This presents a significant barrier to other researchers in terms of both critiquing (and potentially improving) dT2A, as well as its use by practitioners. We propose to replicate Royall and Palmer’s approach and make freely available an open source implementation of dT2A for others to use. In addition, we propose to implement a Bayesian version of the same model using a Bayesian latent variable approach and estimated using Markov chain Monte Carlo as implemented in JAGS. We will make this freely available also. The key advantage of the Bayesian approach is that it provides outputs that are simple to interpret. I have discussed this with Professors Royall and Palmer and they are supportive. Preliminary Studies: None Experimental Plan: BNT, Animals, LMI-I, LMI-II, MMSE, TrailsB, FAQ, and CDR-SB are the only data elements to which we require access. The PI has decades of experience working with R and, more recently, developing simple “apps” that provide an easy interface to models developed in R. See, for example the “ShinyApps” associated with our recent Lancet and Harvard Data Science Review articles. We will describe our work in a publication and make the software freely available via a Columbia University website. References: Royall DR, Palmer RF. A δ homolog for dementia case finding with replication in Alzheimer's Disease Neuroimaging Initiative. Journal of Alzheimer’s Disease. 2019;67:67-79. Schuemie, M.J., Cepeda, M.S., Suchard, M.A., Yang, J., Tian, Y., Schuler, A., Ryan, P.B., Madigan, D., and Hripcsak, G. (2020). How Confident Are We About Observational Findings in Healthcare: A Benchmark Study. Harvard Data Science Review, 2.1, DOI: 10.1162/99608f92.147cc28e Suchard, M.A., Schuemie, M.J., Krumholz, H.M., You, S., Chen, R., Pratt, N., Reich, C.G., Duke, J., Madigan, D., Hripcsak, G., and Ryan, P.B. (2019). Comprehensive comparative effectiveness and safety of first-line antihypertensive drug classes. The Lancet, DOI:https://doi.org/10.1016/S0140-6736(19)32317-7. |
| Additional Investigators |
