|
Predictor
|
Category
|
Logistic regression
|
Linear regression
|
|---|
|
OR
|
95% CI
|
p-value
|
Coefficient
|
95% CI
|
p-value
|
|---|
|
Age
|
0–17 years
|
1
| |
0.006
|
0
| |
0.004
|
|
≥18 years
|
0.30
|
0.12, 0.70
|
−0.52
|
−0.88, −0.16
|
|
Experience with IgG prior to SCIG
|
Ig-naïve
|
1
| |
0.01
| | | |
|
Ig-experienced
|
2.36
|
1.22, 4.54
|
|
Time on treatment
|
< 2 years
| | | |
0
| |
0.06
|
|
≥2 years
|
0.27
|
−0.01, −0.54
|
|
Confidence after traininga
|
1–5
|
1
| |
0.03
| | | |
|
6–7
|
2.18
|
1.07, 4.44
|
|
Number of sites
|
1–3
|
1
| |
0.01
|
0
| |
0.04
|
|
≥4
|
0.44
|
0.24, 0.84
|
−0.24
|
−0.47, −0.02
|
|
TSQM effectiveness scoreb
|
T2 + T3, ≤75
|
1
| |
0.001
|
0.13
|
0.07, 0.18
|
< 0.001
|
|
T1, ≥76 (best)
|
2.73
|
1.50, 4.80
|
|
PROMIS Fatiguec
|
T2 + T3, ≥54
|
1
| |
< 0.001
|
−0.33
|
−0.39, −0.26
|
< 0.001
|
|
T1, ≤53 (best)
|
8.26
|
4.56, 15.0
|
- Multivariate logistic regression and linear regression models calculated predictors for being in the best tertile of GHP scores. GHP was measured on an anchored numeric 1–7 scale (1 = poor health and 7 = excellent health), where respondents were grouped in T2 + T3 (intermediate/worst) if they had a score of ≤ 5 and in T1 (best) if they scored 6 or 7. PROMIS Fatigue T-scores are obtained from published raw score to T-score concordance tables of the PROMIS Fatigue Short Form 7a. With 5 levels on each of the 7 items, the raw scores vary from 7 to 35 and are converted to corresponding T-scores in the range of 29.4 (least fatigue) to 83.2 (most fatigue). TSQM transformed scores (T-scores) were measured on a 0–100 scale (0 = worst satisfaction and 100 = perfect satisfaction)
- CI confidence interval, GHP general health perception, IgG immunoglobulin G, OR odds ratio, PROMIS Patient-Reported Outcome Management Information System, SCIG subcutaneous immunoglobulin, SD standard deviation, TSQM Treatment Satisfaction Questionnaire for Medication
- aPredictor on an anchored numeric scale from 1 to 7 (1 = not very confident and 7 = very confident). The logistic regression yields an OR which predicts the likelihood of each category achieving the desired best tertile, and a significant OR > 1 implies higher odds than with the reference category. The least squares regression models score on a continuous linear scale using the original 1–7 scale, where a higher coefficient implies a higher GHP
- bRegression coefficient reported for a 0.5 SD increase in score (equivalent to 10 units)
- cRegression coefficient reported for a 0.5 SD increase in score (equivalent to 5 units). The model had an R2 = 36.2%, suggesting that over a third of scores can be explained by the factors in the final model
