Patient selection: age >= 65 years and admitted to a VA hospital
Factors associated with mortality in multivariate analysis:
(1) BMI <= 20 kg per meter square
(2) weight as percent of usual weight <= 85%
Other factors correlated with mortality in univariate (but not multivariate) analysis:
(3) > 10% weight loss in the past year
|
Risk Factor
|
Adjusted Relative Risk
|
95% CI
|
|
BMI <= 20
|
1.83
|
1.17 to 2.85
|
|
weight <= 85% of usual body weight
|
1.78
|
1.14 to 2.77
|
|
> 10% weight loss
|
2.31
|
1.35 to 3.94
|
from Table 3, page M744
where:
• The relative risk adjusted for Katz index of ADL and APACHE II scores.
The authors performed a Cox regression model for survival but did not give details. I have tried to capture the concepts described in the paper using the following model:
|
Parameter
|
Finding
|
Points
|
|
BMI
|
> 20 kg per square meter
|
0
|
|
|
<= 20
|
1
|
|
weight as percent usual body weight
|
> 85%
|
0
|
|
|
<= 85%
|
1
|
number of risk factors =
= SUM(points for parameters)
Interpretation:
• minimum score: 0
• maximum score: 2
|
Total Score
|
Risk for Mortality within 1 Year
|
|
0
|
low
|
|
1
|
intermediate
|
|
2
|
high
|
Limitations:
• Since the patients were admitted to a VA hospital, most patients must have been male.
• The mortality rate for a low BMI was 17% at 1 year, while the mortality for a BMI > 20 kg per square meter was 8%, suggesting an appreciable mortality rate in the "low" risk group.
