Skip to main content

Table 4 Interpretation of the treatment coefficient from the ten statistical methods using the SF-36 MH score at 6-month follow-up in the LM trial as an example

From: Comparison of statistical methods for the analysis of patient-reported outcomes (PROs), particularly the Short-Form 36 (SF-36), in randomised controlled trials (RCTs) using standardised effect size (SES): an empirical analysis

(a) Statistical methods that used the untransformed scale

Statistical methods

Coefficient

Interpretation

MLR

2.31

The mean of the MH score at 6 months for the treatment group is 2.31 points higher than the mean for the control group, after adjusting for baseline MH score

Tobit

2.02

The mean of the uncensored MH score at 6 months for the treatment group is 2.02 points higher than the mean for the control group, after adjusting for baseline MH score

Median

5.00

The median of the MH score at 6 months for the treatment group is 5.00 points higher than the median for the control group, after adjusting for baseline MH score

CLAD

5.00

The median of the uncensored MH score at 6 months for the treatment group is 5.00 points higher than the median for the control group, after adjusting for baseline MH score

(b) Statistical methods that used the transformed scales

Statistical methods

Coefficient

Odds

Interpretation

BB

0.11

1.12

The odds of the MH score being in a given category at 6 months in the treatment group is 1.12 times that of the odds for the control group, after adjusting for baseline MH score

BLN

0.12

1.13

The odds of the MH score being in a given category at 6 months in the treatment group is 1.13 times that of the odds for the control group, after adjusting for baseline MH score

OL

0.29

1.34

The odds of the MH score being in a given category or less at 6 months in the treatment group is 1.34 times that of the odds for the control group, after adjusting for baseline MH score

OP

0.16

NA

(Marginal effects need to be calculated to generate the probability)

The probability of scoring 90.0 at 6 months is 21.2% for the treatment group and 20.1% for the usual care group, after adjusting for baseline MH score

Frac

0.14

1.15

The odds of the MH score being in a given category at 6 months in the treatment group is 1.15 times that of the odds for the control group, after adjusting for baseline MH score

BR

0.04

1.05

The odds of the MH score being in a given category at 6 months in the treatment group is 1.05 times that of the odds for the control group, after adjusting for baseline MH score

  1. BB beta-binomial regression, BLN binomial-logit-normal regression, BR beta regression, CLAD censored absolute least deviation regression, Frac fractional logistic regression, OL ordered logit model, OP ordered probit, Median median regression, MH mental health, MLR multiple linear regression, NA not applicable, SF-36 Short-Form 36, Tobit Tobit regression
Back to article page

Health and Quality of Life Outcomes

ISSN: 1477-7525

Contact us