Table 3.
Multivariable longitudinal models of the 3 CES-D summary depression measures (CES-D symptom score, CES-D symptom score ≥7, Count of CES-D symptoms)
| Linear Mixed Model of CESD Total Score |
Logistic Regression of CESD Total Score ≥ 7 |
Negative Binomial Regression of # CESD Symptoms |
|||||||
|---|---|---|---|---|---|---|---|---|---|
| Multivariable models | Estimatea | 95% CI | P-value | ORa | 95% CI | P-value | IRRa | 95% CI | P-value |
| Timeb | |||||||||
| baseline vs year 1 | 0.51 | (0.29, 072) | <0.0001 | 1.41 | (1.10, 1.82) | 0.0072 | 1.45 | (1.30, 1.61) | <0.0001 |
| 1–8 years (per 1 year) | -0.09 | (−0.12, −0.06) | <0.0001 | 0.94 | (0.91, 0.98) | 0.0066 | 0.92 | (0.90, 0.94) | <0.0001 |
| Age (per 10 years younger) | 0.21 | (0.39, 0.04) | 0.0159 | 1.24 | (1.08, 1.44) | 0.0031 | 1.12 | (1.04, 1.22) | 0.0031 |
| Gender (Female vs Male) | 0.55 | (0.17, 0.94) | 0.0052 | 1.42 | (0.97, 2.09) | 0.0726 | 1.47 | (1.20, 1.80) | 0.0002 |
| Race (vs White) | |||||||||
| Black | 0.29 | (−0.13, 0.71) | 0.1769 | 1.21 | (0.83, 1.76) | 0.3177 | 1.16 | (0.94, 1.44) | 0.1633 |
| Asian | -0.96 | (−2.43, 0.51) | 0.2019 | 0.08 | (0.01, 0.45) | 0.0046 | 0.46 | (0.25, 0.85) | 0.0130 |
| Other | 1.51 | (0.59, 2.42) | 0.0012 | 2.20 | (1.23, 3.92) | 0.0077 | 2.11 | (1.45, 3.06) | <0.0001 |
| Education (vs Graduate Education) | |||||||||
| <HS | 0.9 | (0.20, 1.61) | 0.0125 | 2.93 | (1.69, 5.10) | 0.0001 | 1.41 | (1.01, 1.98) | 0.0442 |
| HS | -0.12 | (−0.76, 0.51) | 0.7080 | 1.50 | (0.89, 2.54) | 0.1304 | 0.90 | (0.66, 1.21) | 0.4756 |
| Some College | 0.40 | (−0.24, 1.05) | 0.2167 | 2.30 | (1.29, 4.10) | 0.0045 | 1.14 | (0.83, 1.56) | 0.4247 |
| College Degree | 0.10 | (−0.61, 0.80) | 0.7854 | 1.64 | (0.85, 3.17) | 0.1381 | 1.11 | (0.77, 1.60) | 0.5726 |
| Baseline VAQ (per 1-unit increase on 1–5 scale, higher is worse) |
1.26 | (0.99, 1.54) | <0.0001 | 2.41 | (1.91, 3.04) | <0.0001 | 1.89 | (1.67, 2.15) | <0.0001 |
CESD, Center for Epidemiologic Studies Depression Scale; CI, Confidence Interval; OR, Odds Ratio; IRR, Incident Rate Ratio; HS, High School; VAQ, Visual Activities Questionnaire.
Increased risk of depression is indicated by Estimate>0, OR>1 or IRR>1.
Because the drop in depression scores from baseline to 1 year was larger than predicted by a linear relationship between 1 and 8 years in all 3 models, time since diagnosis was modeled with a linear term for all time points as well as an indicator variable for deviation from the line at the baseline time point.
Note: Linear mixed regression used the compound symmetry covariance structure to adjust for the correlation between repeated measures. Repeated measures logistic and negative binomial regressions used generalized estimating equations to adjust for the correlation between repeated measures.