Fig 1.
A flow chart of data linkage, data processing, feature selection, principal component analysis and index generation with the Medical Expenditure Panel Survey (MEPS) 1996 to 2012.
Table 1.
Proposed publication cycle for weighted indices.
Table 2.
The characteristics of the interviewees in the first to 16th Medical Expenditure Panel Survey.
Fig 2.
The Kaplan-Meier survival curves of the interviewees in the second years of the MEPS panels.
(a) The Kaplan-Meier survival curves by sex. Chi-square = 12.23, p < 0.001. (b) The Kaplan-Meier survival curves by races. Chi-square = 27.46, p < 0.001.
Fig 3.
The distributions of those dying in the second years of the MEPS panels by principal components.
(a) The distribution of those dying by first and second principal components (PC1 and PC2). (b) The distribution of those dying by first and third principal components (PC1 and PC3). (c) The distribution of those dying by first and fourth principal components (PC1 and PC4). (d) The distribution of those dying by first and fifth principal components (PC1 and PC5). Note: red circles: those dying in the second years of the MEPS panels; gray circles: those surviving throughout the MEPS panels.
Table 3.
Coefficients of the first principal component to predict mortality in the second years of the MEPS panels.
Table 4.
Coefficients of the first principal component and demographics to predict mortality in the second years of the MEPS panels.
Fig 4.
The p values of all PCA-based weighted indices regarding the prediction of mortality risk.
(a) P values for 134689 PCA-based indices regarding mortality risk in models that take time (in quarters) and interactions between indices and time. (b) P values for 134689 PCA-based indices regarding mortality risk in models that take age, sex, races, time (in quarters) and interactions between indices and time.
Table 5.
Summaries of the significance (p<0.05) of all PCA-based weighted indices.
Fig 5.
Flowchart of the process of index review and evaluation.
Note: PCA: principal component analysis; PLS: partial least squares.