Skip to main content
. 2021 Apr 16;38(4):395–402. doi: 10.1093/fampra/cmaa141

Table 2.

Bivariate Cox proportional hazard model: relationship between overall mortality and explanatory variables in older people receiving physician-led home visits in Japan, 2013–17

Pairwise deletion Multiple imputation (n = 825)
n Hazard ratio [95% CI] P-value Hazard ratio [95% CI] P-value
Sex (female/male = 1/0) 825 0.61 [0.50–0.75] <0.01
Age (year) 825 0.99 [0.98–1.00] 0.18
CCI 815
 0–1 Reference
 2 1.33 [0.97–1.82] 0.07 1.34 [0.98–1.84] 0.07
 3–4 1.60 [1.19–2.15] <0.01 1.63 [1.21–2.18] <0.01
 ≥5 5.82 [4.46–7.59] <0.01 5.83 [4.47–7.61] <0.01
Serum albumin level (g/dl) 770
 0–3 Reference
 3.1–3.5 0.48 [0.37–0.62] <0.01 0.48 [0.37–0.62] <0.01
 3.6–3.9 0.37 [0.27–0.50] <0.01 0.36 [0.26–0.48] <0.01
 ≥4 0.19 [0.14–0.26] <0.01 0.19 [0.13–0.26] <0.01
BI 819
 0–24 Reference
 25–59 0.68 [0.52–0.88] <0.01 0.68 [0.52–0.89] <0.01
 60–84 0.43 [0.32–0.57] <0.01 0.43 [0.32–0.57] <0.01
 ≥85 0.41 [0.31–0.55] <0.01 0.41 [0.31–0.55] <0.01
Pressure ulcer treatment (recipient/non-recipient = 1/0) 825 1.52 [0.89–2.59] 0.12
Oxygen therapy (recipient/non-recipient = 1/0) 825 2.91 [2.10–4.04] <0.01
CSDD 801
 0 Reference
 1–2 1.24 [0.93–1.64] 0.14 1.24 [0.94–1.65] 0.13
 3–4 1.19 [0.87–1.64] 0.28 1.22 [0.89–1.67] 0.23
 ≥5 2.05 [1.58–2.67] <0.01 2.07 [1.59–2.70] <0.01
MMSE-J 699
 0–14 Reference
 15–19 0.81 [0.58–1.13] 0.20 0.80 [0.58–1.10] 0.17
 20–23 0.70 [0.50–0.97] 0.03 0.69 [0.51–0.95] 0.02
 ≥24 0.79 [0.58–1.07] 0.12 0.77 [0.58–1.03] 0.08
Nursing home resident (yes/no = 1/0) 825 0.31 [0.17–0.54] <0.01
Public assistance recipient (yes/no = 1/0) 825 0.54 [0.38–0.76] <0.01
Full-time caregiver (present/not present = 1/0) 825 1.79 [1.42–2.26] <0.01
Living alone (yes/no = 1/0) 825 0.77 [0.60–0.99] 0.04

The pairwise deletion column shows results of the analyses using pairwise deletion (i.e. each row represents the results from analysis of the available data for that variable). Hazard ratios for the variables with no missing values are shown only for the pairwise deletion analyses, because there were no differences in hazard ratios between the multiple imputation analyses and the pairwise deletion analyses. Little’s missing completely at random test was significant at P < 0.01, indicating that the missing data were not missing completely at random. We, therefore, used multiple imputation for interpretation.