Table 2.
GEE models for the comparison of primary and secondary outcomes across time between control and tai chi qigong groups
| Outcomes | Regression coefficients of the GEE models
|
|||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Group
|
T1
|
T2
|
Group by T1
|
Group by T2
|
||||||
| B (95% CI) | P-value | B (95% CI) | P-value | B (95% CI) | P-value | B (95% CI) | P-value | B (95% CI) | P-value | |
| Sleep quality (CPSQI) | ||||||||||
| CPSQI total score (range: 0–21) | 0.97 (−0.86, 2.80) | 0.300 | −0.23 (−1.74, 1.28) | 0.764 | 0.13 (−1.23, 1.49) | 0.854 | −1.37 (−3.19, 0.44) | 0.138 | −2.67 (−4.51, −0.83) | 0.004 |
| CPSQI componentsψ | ||||||||||
| Subjective sleep quality (range: 0–3) | 0.40 (−0.75, 1.55) | 0.493 | 0.00 (−1.39, 1.38) | 0.996 | −0.06 (−1.19, 1.08) | 0.924 | −0.36 (−1.97, 1.25) | 0.660 | −0.55 (−2.08, 0.99) | 0.487 |
| Sleep latency (range: 0–3) | −0.10 (−1.29, 1.09) | 0.869 | −1.03 (−2.12, 0.05) | 0.062 | 0.07 (−0.92, 1.07) | 0.884 | 1.13 (−0.21, 2.47) | 0.097 | −0.66 (−1.99, 0.67) | 0.330 |
| Sleep duration (range: 0–3) | 1.16 (0.09, 2.23) | 0.034 | 0.71 (−0.39, 1.81) | 0.205 | 1.11 (−0.31, 2.53) | 0.126 | −1.89 (−3.29, −0.49) | 0.008 | −2.69 (−4.53, −0.84) | 0.004 |
| Habitual sleep efficiency (range: 0–3) | 0.95 (−0.10, 2.01) | 0.075 | 0.08 (−1.00, 1.15) | 0.891 | 0.78 (−0.38, 1.94) | 0.190 | −1.15 (−2.51, 0.21) | 0.096 | −2.58 (−4.24, −0.91) | 0.002 |
| Sleep disturbance (range: 0–3) | 0.32 (−0.94, 1.58) | 0.615 | −0.48 (−1.64, 0.67) | 0.411 | −1.50 (−2.67, −0.33) | 0.012 | −1.38 (−2.97, 0.21) | 0.089 | 0.31 (−1.09, 1.70) | 0.669 |
| Use of sleeping medication (range: 0–3)# | – | – | – | – | – | – | – | – | – | – |
| Daytime dysfunction (range: 0–3) | −0.17 (−1.33, 0.99) | 0.772 | 0.37 (−0.79, 1.53) | 0.533 | −1.17 (−2.33, 0.00) | 0.050 | −0.77 (−2.36, 0.82) | 0.342 | 0.80 (−0.75, 2.36) | 0.310 |
| Cognitive functions | ||||||||||
| MMSE (range: 0–30) | 0.94 (−0.72, 2.60) | 0.268 | 1.79 (0.54, 3.04) | 0.005 | 2.36 (1.20, 3.52) | <0.001 | −0.75 (−2.46, 0.97) | 0.394 | −1.00 (−2.60, 0.59) | 0.219 |
| MIC (range: 0–108)* | 0.11 (−1.06, 1.28) | 0.854 | 0.45 (−0.29, 1.18) | 0.231 | −0.02 (−0.75, 0.72) | 0.965 | −0.82 (−1.72, 0.08) | 0.074 | −0.34 (−1.30, 0.62) | 0.486 |
| Health-related quality of life (SF-12) | ||||||||||
| PCS | 9.20 (1.77, 16.62) | 0.015 | 1.39 (−3.22, 6.00) | 0.555 | 3.92 (−0.08, 7.91) | 0.055 | −3.66 (−9.42, 2.09) | 0.212 | −2.91 (−8.93, 3.11) | 0.343 |
| MCS | −4.49 (−9.53, 0.55) | 0.081 | −6.04 (−10.20, −1.88) | 0.004 | −0.78 (−5.19, 3.62) | 0.727 | 11.53 (6.68, 16.39) | <0.001 | 0.70 (−6.18, 7.58) | 0.842 |
Notes: Only the model estimates of regression coefficients (B) of the dummy variables for the group (group: 0= control [reference]; 1= [tai chi]), time points (T1 and T2 with the baseline [T0] as reference), time points and group interaction terms (group by T1 and group by T2) are shown for the GEE models.
The CPSQI component scores were all treated as ordinal data, and the ordinal logistic link function was used in the GEE models, the coefficients reported were in log-odds ratios.
Square root-transformed before being entered into the GEE model.
The GEE analysis was not performed for this variable owing to sparseness of data other than 0. Effect sizes: CPSQI (T2 vs T0), Cohen’s d=0.70; CPSQI – Sleep duration component (T1 vs T0), odds ratio=6.6; CPSQI – Sleep duration component (T2 vs T0), odds ratio=14.7; CPSQI – Habitual sleep efficiency (T2 vs T0), odds ratio=13.2.
Abbreviations: CI, confidence interval; CPSQI, Chinese Pittsburgh sleep quality index; GEE, generalized estimating equation; MIC, memory inventory for Chinese; MCS, mental health component score; MMSE, mini-mental state examination; PCS, physical component score; SF-12, Short-form 12.