Skip to main content
. 2005 Nov 1;55(520):831–837.

Table 2.

Effect of education programme on primary and secondary outcomes: baseline and 4-month values. Intention to treat analysis.

Outcome Baseline 4 months Effect size (95% CI) P-value

Control Intervention Control Intervention
Self efficacy (scale 5–30,↑ better) 16.39 16.31 17.62 18.22 0.67a (0.08 to 1.25) 0.025

Self care behaviour (scale 5–25, ↑ better) 7.16 7.23 8.31 8.95 0.53a (0.01 to 1.06) 0.047

Communication with physician (scale 5–20, ↑ better) 6.68 6.49 7.49 7.60 0.15 a (−0.23 to 0.54) 0.435

Health status

 Depression (scale 0–21, ↓ better) 8.64 8.84 8.49 8.55 −0.22a (−0.71 to 0.28) 0.392

 Anxiety (scale 0–21, ↓ better) 8.41 8.66 8.08 8.23 −0.09a (−0.55 to 0.35) 0.682

 Pain (scale 1–5, ↓ better) 3.49 3.36 3.23 3.07 −0.13b (−0.48 to 0.22) 0.462

 Fatigue (scale 1–5, ↓ better) 3.33 3.40 3.17 3.25 0.06b (−0.22 to 0.12) 0.721

 Shortness breath (scale 1–5, ↓ better) 2.29 2.45 2.23 2.26 −0.22b (−0.60 to 0.16) 0.249

Health status EQ5D 0.59 0.61 0.65 0.66 0.005a (−0.044 to 0.054) 0.827

Healthcare use

 Visits to GP/practice nurse in last 3 months 4.12 3.97 3.71 3.53 0.99c (0.90 to 1.10) 0.893

In all analyses, age, sex, main condition at baseline and baseline values of outcome are adjusted for. In the Poisson regression baseline value of outcome is included as an offset. The coefficients presented are:

a

adjusted difference in means,

b

coefficient on logistic scale,

c

incidence rate ratio. The coefficient on the logistic scale presented for the proportional odds ordered logistic regression model is similar to the coefficient from a logistic regression model. For a logistic model the coefficient can be directly converted to an easily interpretable odds ratio. This is not the case for the coefficients from the proportional odds ordered logistic model. Negative coefficients for the proportional odds ordered model indicate that symptom scores are lower in the intervention group.