Skip to main content
. 2016 Feb 10;11(2):e0148414. doi: 10.1371/journal.pone.0148414

Table 3. Multivariate Analysis of the Predictors of OT Referral by Fitting Multiple Logistic Regression Model: IADL and BADL Models.

IADL Model BADL Model
OR P-value 95% CI OR P-value 95% CI
Lawton Scale 0.41 < 0.001 0.28–0.60 - - -
Barthel Index - - - 0.77 < 0.01 0.64–0.94
Dementia 61.49 < 0.05 2.31–1636.29 48.26 < 0.05 2.09–1114.66
SPMSQ (2–9) 14.60 < 0.01 2.93–72.79 8.87 < 0.01 2.10–37.50
ROM Limitation 13.76 < 0.01 2.01–94.24 13.97 < 0.01 2.08–93.87
Psychiatric Disorders 11.04 < 0.01 2.17–56.09 13.95 < 0.001 3.12–62.34
≥ 2 (for ADL model) or ≥ 3 (for IADL model) Behavioral and Emotional Measures 9.87 < 0.01 1.79–54.35 5.78 < 0.05 1.25–26.81
Fear of Falling 11.45 < 0.01 2.15–61.11 13. 93 < 0.01 2.71–71.75
Unable to Express Oneself - - - 6.85 < 0.05 1.23–38.04

OT: occupational therapy; IADL: instrumental activities of daily living; BADL: basic activities of daily living; OR: odds ratio; CI: confidence interval; SPMSQ: Short Portable Mental Status Questionnaire, range: 0–10; ROM: range of motion. IADL Model: Sample size (n) = 221 and the number of estimated parameters (p) = 8. Goodness-of-fit assessment: n = 221, adjusted generalized R2 = 0.800 > 0.3, the estimated area under the Receiver Operating Characteristic (ROC) curve = 0.977 > 0.7 (se = 0.008), the modified Hosmer-Lemeshow goodness-of-fit F test p = 0.5001 > 0.05 (df = 9, 211), and the ratio of residual deviance and its degrees of freedom, 71.5356/213 = 0.3358 ≤ 1.0, which indicated an excellent fit. BADL Model: Sample size (n) = 221 and the number of estimated parameters (p) = 9. Goodness-of-fit assessment: n = 221, adjusted generalized R2 = 0.778 > 0.3, the estimated area under the Receiver Operating Characteristic (ROC) curve = 0.972 > 0.7 (se = 0.009), the modified Hosmer-Lemeshow goodness-of-fit F test p = 0.6234 > 0.05 (df = 9, 211), and the ratio of residual deviance and its degrees of freedom, 78.1694/212 = 0.3687 ≤ 1.0, which indicated an excellent fit. In both logistic regression models, np ≥ 200, which indicated that our sample size was not too small.