Table 2. Mental well-being outcomes at endline.
| Indicator | Cronbach Alpha | Intervention | Control | ||||
|---|---|---|---|---|---|---|---|
| Scores | N | Mean (SD) | N | Mean (SD) | - | difference (95% CI) a e | |
| Anxiety score b | 0.89 | 742 | 1.46 (0.61) | 729 | 1.52 (0.64) | - | ‒0.06 (‒0.14, 0.02) |
| Depression score b | 0.87 | 742 | 1.35 (0.48) | 728 | 1.39 (0.52) | - | ‒0.04 (‒0.08, 0.01) |
| Emotional distress score b | 0.93 | 741 | 1.29 (0.46) | 728 | 1.33 (0.49) | - | ‒0.04 (‒0.09, 0.01) |
| Well-being score c | 0.97 | 749 | 17.6 (6.8) | 728 | 17.0 (6.7) | - | 0.50 (‒0.19, 1.28) |
| Prevalence | N | % | N | % | PR (95% CI) d | PD (95% CI) e | |
| High Anxiety f | - | 742 | 22.2 | 729 | 24.8 | 0.90 (0.72, 1.11) | ‒0.03 (‒0.08, 0.03) |
| High Depression f | - | 742 | 14.0 | 728 | 16.9 | 0.83 (0.64, 1.07) | ‒0.03 (‒0.07, 0.01) |
| High Emotional distress f | - | 741 | 14.0 | 728 | 16.4 | 0.86 (0.67, 1.09) | ‒0.02 (‒0.06, 0.01) |
| Poor well-being g | - | 749 | 25.2 | 728 | 27.8 | 0.90 (0.74, 1.10) | ‒0.03 (‒0.08, 0.02) |
Notes.
a We used linear regression models to estimate the difference in the outcomes comparing the intervention and control arms. Models accounted the stratified design by including woreda indicator variables [58], and accounted for clustering within kebeles by using generalized estimating equations with robust standard errors.
b We asked respondents to indicate how much the symptoms bothered them in the previous week with four potential response options (not at all (1) to extremely (4)). The first ten symptoms assess anxiety (i.e., ‘suddenly scared for no reason’, ‘nervousness or shakiness inside’), the next 13 assess depression (i.e. ‘feeling low in energy’, ‘feeling hopeless about the future’), and the 23 collectively assess non-specific emotional distress. For each outcome, the score is the sum of the responses divided by the number of items.
c We asked respondents about well-being, and responses ranged from ‘(0) At no time’ to (5) All of the time’. Scores were summed, and range from 0–25; the higher the score, the better the well-being.
d We used similar log-linear binomial regression models to compare the prevalence of the outcomes between the intervention and control arms.
e Prevalence differences (PD) were calculated using post-estimation commands to estimate the average marginal effects.
f Each of the above scores was dichotomized, with scores greater than 1.75 indicating a positive status for any of the three outcomes.
g The above score was dichotomized with scores below 13 indicating poor well-being.