Skip to main content

Table 2 Comparing Predictors of Death using Inverse Probability Weighting from Tracing and Imputationa

From: Treating loss-to-follow-up as a missing data problem: a case study using a longitudinal cohort of HIV-infected patients in Haiti

 

Complete Case

OR (95% CI)

IPW

OR (95% CI)

MICE

OR (95% CI)

MICE with deletion

OR (95% CI)

N=735b

N=7992

N = 910

N = 799

Female

0.83 (0.61, 1.13)

0.79 (0.57, 1.10)

0.83 (0.62, 1.12)

0.81 (0.60, 1.10)

Age (for 10-yr difference)

1.17 (1.00, 1.37)

1.26 (1.06, 1.51)

1.20 (1.03, 1.39)

1.20 (1.03, 1.40)

Residence

1.13 (0.81, 1.56)

1.28 (0.90, 1.82)

1.18 (0.87, 1.60)

1.17 (0.85, 1.61)

Severe poverty

1.78 (1.2, 2.44)

1.59 (1.14, 2.24)

1.74 (1.27, 2.37)

1.71 (1.25, 2.34)

CD4 (for 100-cell difference)

0.86 (0.74, 1.00)

0.86 (0.75, 0.99)

0.85 (0.73, 0.98)

0.84 (0.72, 0.98)

Baseline Weight (for 10-kg difference)

0.67 (0.58, 0.79)

0.69 (0.56, 0.83)

0.65 (0.56, 0.76)

0.66 (0.57, 0.77)

WHO stage

2.19 (1.60, 3.00)

2.19 (1.56, 3.06)

2.18 (1.62, 2.95)

2.19 (1.61, 2.98)

Baseline Tuberculosis

1.97 (1.20, 3.22)

1.92 (1.13, 3.25)

1.98 (1.23, 3.18)

1.95 (1.21, 3.15)

  1. aModels were constructed with only one covariate
  2. bFor the CD4 only model N = 671; Baseline weight model N = 671
  3. 2For the CD4 only model N = 710
  4. p -value ≤ 0.05