Skip to main content

Table 4 Linear models for headache impact test 6

From: Effects of neck-exercise and health promotion on headache outcomes in office workers: secondary analysis of the NEXpro stepped wedge cluster randomised controlled trial

Outcome Variable: Headache impact test 6 (HIT-6)

Linear model

Change in HIT-6 score.

Model 1

~ Intervention

Model 2

~ Intervention + Period

Model 3

~ Intervention x Period

Model 4

~ Period + Exposure time

(Intervention nested in exposure time)

Independent variable

Estimates (95%CI)

Estimates (95%CI)

Estimates (95%CI)

Estimates (95%CI)

Intercept

48.11 (46.45 to 49.77)

48.81 (46.97 to 50.65)

48.81 (46.97 to 50.65)

48.81 (46.97 to 50.64)

Intervention

−2.23 (−3.35 to −1.12)

−1.07 (−2.83 to 0.69)

 

-

Period 1

 

−0.73 (−2.39 to 0.93)

−1.06 (−2.84 to 0.73)

−1.01 (−2.68 to 0.67)

Period 2

 

−1.89 (−3.54 to −0.23)

−1.58 (−3.36 to 0.20)

−1.66 (−3.42 to −0.10)

Period 3

 

−1.34 (−3.28 to 0.61)

−1.31 (−2.34 to 1.02)

−1.76 (−3.97 to 0.45)

Period 4

 

−2.25 (−4.60 to 0.10)

−2.20 (−5.23 to 0.83)

−2.91 (−5.88 to 0.06)

Intervention x Period 1

  

−0.09 (−2.77 to 2.58)

 

Intervention x Period 2

  

−1.99 (−4.67 to 0.68)

 

Intervention x Period 3

  

−1.12 (−3.74 to 1.50)

 

Exposure: 4 months

   

0.24 (−2.19 to 1.70)

Exposure: 8 months

   

−1.75 (−4.26 to 0.77)

Exposure: 12 months

   

−0.63 (-−3.95 to 2.69)

Exposure: 16 months

   

0.77 (−3.31 to 4.86)

Observations

600

600

600

600

Log likelihood

−2066.0

−2063

−2062

−2060

Akaike Information Criterion (AIC)

4141

4143

4146

4145

Bayesian Information Criterion (BIC)

4163

4183

4194

4198

  1. Estimates are mean differences with 95% confidence intervals. Figures in bold indicate the best fitting Model
  2. Model one showed an average reduction on the HIT-6 scale of 2.23 (95% CI: -3.35 to -1.12) points after the intervention
  3. Models two and three demonstrated effects by adding period to the model (Model two) or the interaction between intervention and period (Model three)
  4. In Model two, the intervention showed the most precise and significant effect after period two. No interaction effect was found for Model three
  5. The ANOVA did not demonstrate any better statistical fit for Model two (p= 0.18) or Model three (p= 0.52) compared to Model one
  6. Model four: No effect of the intervention nested in time (exposure time) was found. Log likelihood indicated a better explanation than Models one to three, no ANOVA could be applied, as former Models are not part of Model four