Table 1.
Serial indirect models of combat exposure with health-related behaviors: model coefficients and bias-corrected bootstrapped (k = 1000 bootstraps) 95% confidence intervals
| 95% CI | n | |||||
|---|---|---|---|---|---|---|
| Criterion | Mediator(s) | b | SE | LL | UL | |
| Hypothesis model | ||||||
| Risk Behaviors | ||||||
| Sleep Disturbance → PCL-C† | .0003* | .0001 | .0001 | .0005 | 2354 | |
| Sleep Disturbance† | .0009* | .0004 | .0002 | .0017 | 2354 | |
| PCL-C† | .0041* | .0008 | .0027 | .0057 | 2354 | |
| Sleep Disturbance → PHQ-8‡ | .0001 | .0001 | −.0001 | .0003 | 2354 | |
| Sleep Disturbance‡ | .0001 | .0002 | −.0001 | .0006 | 2354 | |
| PHQ-8‡ | −.0010* | .0004 | −.0019 | −.0003 | 2354 | |
| Sleep Disturbance → GAD-7§ | .0000 | .0000 | −.0001 | .0000 | 2354 | |
| Sleep Disturbance | .0003 | .0003 | .0000 | .0010 | 2354 | |
| GAD-7§ | −.0004 | .0003 | −.0011 | .0002 | 2354 | |
| Aggression | ||||||
| Sleep Disturbance → PCL-C† | .0019* | .0005 | .0010 | .0031 | 2344 | |
| Sleep Disturbance† | .0022* | .0011 | .0005 | .0050 | 2344 | |
| PCL-C† | .0261* | .0031 | .0201 | .0323 | 2344 | |
| Sleep Disturbance → PHQ-8‡ | .0003 | .0005 | −.0006 | .0013 | 2344 | |
| Sleep Disturbance‡ | .0006 | .0008 | −.0009 | .0025 | 2344 | |
| PHQ-8‡ | −.0050* | .0017 | −.0088 | −.0018 | 2344 | |
| Sleep Disturbance → GAD-7§ | .0000 | .0001 | −.0002 | .0001 | 2344 | |
| Sleep Disturbance§ | .0011 | .0008 | −.0002 | .0030 | 2344 | |
| GAD-7§ | −.0021 | .0020 | −.0059 | .0018 | 2344 | |
| Alcohol Use | ||||||
| Sleep Disturbance → PCL-C† | .0005* | .0002 | .0002 | .0011 | 2360 | |
| Sleep Disturbance† | .0017 | .0011 | −.0002 | .0042 | 2360 | |
| PCL-C† | .0073* | .0020 | .0035 | .0115 | 2360 | |
| Sleep Disturbance → PHQ-8‡ | .0001 | .0002 | −.0002 | .0006 | 2360 | |
| Sleep Disturbance‡ | .0003 | .0005 | −.0003 | .0017 | 2360 | |
| PHQ-8‡ | −.0017* | .0008 | −.0037 | −.0005 | 2360 | |
| Sleep Disturbance → GAD-7§ | .0000 | .0000 | −.0001 | .0000 | 2360 | |
| Sleep Disturbance§ | .0007 | .0006 | .0000 | .0024 | 2360 | |
| GAD-7§ | −.0006 | .0006 | −.0021 | .0003 | 2360 | |
| Opioid Use | ||||||
| Sleep Disturbance → PCL-C† | .0002 | .0001 | .0000 | .0005 | 2384 | |
| Sleep† | .0047* | .0014 | .0023 | .0079 | 2384 | |
| PCL-C† | .0032* | .0015 | .0003 | .0065 | 2384 | |
| Sleep Disturbance → PHQ-8‡ | .0000 | .0001 | .0000 | .0003 | 2384 | |
| Sleep Disturbance‡ | .0012 | .0012 | −.0010 | .0037 | 2384 | |
| PHQ-8‡ | −.0005 | .0004 | −.0015 | .0002 | 2384 | |
| Sleep Disturbance → GAD-7§ | .0000 | .0000 | .0000 | .0001 | 2384 | |
| Sleep Disturbance§ | .0018 | .0011 | .0000 | .0043 | 2384 | |
| GAD-7§ | −.0004 | .0004 | −.0014 | .0001 | 2384 | |
| Reverse-order model | ||||||
| Risk Behaviors | ||||||
| PCL-C→ Sleep Disturbance† | .0003 | .0002 | −.0002 | .0008 | 2354 | |
| Sleep Disturbance† | .0002 | .0002 | −.0001 | .0006 | 2354 | |
| PCL-C† | .0015 | .0011 | −.0007 | .0037 | 2354 | |
| Aggression | ||||||
| PCL-C→ Sleep Disturbance† | .0009 | .0007 | −.0004 | .0022 | 2344 | |
| Sleep Disturbance† | .0005 | .0005 | −.0003 | .0017 | 2344 | |
| PCL-C† | .0161* | .0039 | .0089 | .0243 | 2344 | |
| Alcohol Use | ||||||
| PCL-C→ Sleep Disturbance† | .0007 | .0007 | −.0006 | .0021 | 2360 | |
| Sleep Disturbance† | .0004 | .0005 | −.0004 | .0017 | 2360 | |
| PCL-C† | .0032 | .0032 | −.0030 | .0094 | 2360 |
Coefficients (b) are unstandardized betas.
†Controlling for PHQ-8 (Patient Health Questionnaire) and GAD-7 (Generalized Anxiety Disorder).
‡Controlling for PCL-C (Post-traumatic Stress Disorder (PTSD) Checklist) and GAD-7.
§Controlling for PCL-C and PHQ-8; CE (Combat exposure): Sum of 28-item combat exposure checklist; Poor Disturbance: Sum of three sleep disturbance items; PCL-C: PTSD checklist sum. Risk Behaviors: Sum of four yes/no risk behavior questions; Aggression: Sum of four aggression items; Alcohol Misuse: Recoded AUDIT-C (Alcohol Use Disorder Identification Test- C). Opioid use is logistic regression on single yes/no item on use of opiates. Confidence intervals were used to determine the statistical significance of indirect effects with bootstrapping. t-test statistics are not appropriate as the indirect tests do not conform to normality assumptions. Confidence intervals (CI): 95% not crossing “0” are considered statistically significant.
SE = standard error; LL = lower limit; UL = upper limit; n = sample size.