Knowledge
Of 46 individuals with complete pre-and post-intervention data, a total of 29 (63.04%) improved in accuracy of diagnosis for the case vignettes, 9 (19.57%) had no change in knowledge, and knowledge got worse in 8 (17.39%). Amongst those who improved
, there was approximately a 100% increase in diagnostic accuracy (from mean accurate diagnoses (SD) of 2.03 (1.27) before the intervention, to 4 (0.96) after the intervention). Only pre-intervention knowledge (Odds Ratio (OR) = 0.217
,95% CI = 0.085–0.552,
p = 0.0013) and highest level of education (OR = 2.159
,95% CI = 1.081–4.315,
p = 0.029)
, were significant predictors of gains in accuracy in diagnosing disorders following the intervention. An increase in the highest Grade of education by one year was associated with more than doubling the odds of showing an improvement in correctly diagnosing the disorder in the vignettes, after adjusting for the other covariates. There was a marginally larger proportion of improvers at OTSM than Masincedane (64.3% versus 61.1%, Deviance score = 3.666
, p = 0.0555), with the site-effect probably due to the higher pre-intervention knowledge at OTSM. These differences in site were significant for pre-intervention knowledge only (Mann-Whitney Z = − 2.4325
, p-value = 0.015). Quantitative outcomes are presented Table
3.
Table 3
CHW Training Quantitative Outcomes
Knowledge |
(MAKS) | 41.48 (5.85), N = 58 | 45.57 (4.25) N = 56 | t = −4.523 df = 55 | < 0.001 | 45.67 (4.59) N = 54 | t = −5.0 df = 53 | p < 0.001 |
Confidence |
(MHNCCS) | 45.25 (9.97), N = 58 | 61.75 (7.42) N = 54 | t = −8.749 df = 54 | < 0.001 | |
| Pre-training (mean, SD) N = 45 | Post-training (mean, SD) N = 45 | |
Attitudes (CAMI) |
Authoritarianism
| 27.87 (2.97) | 26.38 (4.1) | t = 2.720 | 0.99 | |
Benevolence
| 37.67 (4.46) | 38.82 (3.79) | t = −1.818 | 0.04 | |
Social Restrictiveness
| 24.73 (4.28) | 22.4 (5.3) | t = 2.96 | 0.002 | |
Tolerance to rehabilitation in the community
| 36.49 (5.11) | 38.09 (4.22) | t = −2.18 | 0.02 | |
There was a significant increase in the average scores on the MAKS pre- to post-training for 56 participants (t = − 4.523, df = 55, p < 0.001). This difference remained when comparing the post-assessment and 3-month assessment scores (mean 45.67, SD 4.59, t = − 5.0, df = 53, p < 0.001, N = 54). Highest level of education was significantly associated with change in knowledge on the MAKS pre-and post the training (coeff = 1.106, t = 2.433, p = 0.019). An increase in the highest Grade of education by one year was associated with a relative increase by more than 1 point on changes in MAKS scores after adjusting for the other covariates.
Confidence
There was a significant increase in the average confidence scores pre-to post training for 54 participants (t = − 8.749, df = 54, p < 0.001) after 4 subjects were removed due to missing data. A qqplot indicated that distributional assumptions of normality were upheld. The only variable that was significantly associated with change in the confidence score pre-and post the training was the pre-training confidence score (coeff − 1.150, t = − 10.981, p < 0.001). A CHW with 1 extra point on the confidence scale at baseline demonstrated a reduction in change in confidence after the training by more than 1 point after adjusting for differences in age, education and years of service. A statistical trend was also observed for age (coeff = − 0.177, t = − 1.875, p = 0.067), with each additional year associated with a reduction in gains in confidence post training. A trend towards a significant effect on changes in confidence pre-post training was observed for having a medical condition (coeff = − 4.845, t = − 1.969, p = 0.055). Having a medical condition was associated with a decreased gain relative to someone without a condition in the effect of the training on confidence.
Attitudes
The CAMI data from 13 participants were removed, as a substantial amount of data was missing for the pre-training variables in particular. The total sample for the CAMI analyses was therefore 45. There was no evidence of a change in authoritarian cluster scores following the training (t = 2.720, p-value = 0.995), however there was some significant improvement in the rest of the attitude clusters.
Benevolence cluster scores increased significantly following training (t = − 1.818, df = 44, p-value = 0.0379). Only the corresponding pre-training benevolence score was significantly associated with change in scores pre-and post the training (coeff = − 0.597, t = − 4.919 p < 0.001), after adjusting for differences in age, education and years of service. An increase of one point at baseline was associated with a decrease by more than half a point in changes associated with the training. CHWs with their own medical conditions displayed a lower average change in scores than those without medical conditions (coeff = − 4.273, t = − 4.620, p < 0.001), after adjusting for pre-training scores and differences between groups in change scores. This was due to an increase in scores on this cluster for those without medical conditions only (mean = 2.19, SD = 4.25, N = 31), with minimal change observed in those with medical conditions (mean = − 1.14, SD = 3.39, N = 14). This was observed even though most of the people with medical conditions (13 of 14) were in the OTSM group, which actually demonstrated increased changes pre-post training. The change in score is substantially higher in the OTSM group (mean = 3.89, SD = 4.2, N = 18) than the Masincedane group (mean = − 0.15, SD = 3.16, N = 13) when people with medical conditions are removed.
There was a decrease in socially restrictive attitudes following training (t = 2.960, df = 44, p = 0.002). Only the corresponding pre-training score for this cluster was significantly associated with a pre- and post-training change in scores. An increase of one point at baseline was associated with a decrease by close to half a point in changes associated with the training (coeff = − 0.437, t = − 2.525, p = 0.016), after adjusting for differences in age, education and years of service. We noted a trend towards a reduction in scores post-training in CHW’s with more years of service (coeff = − 0.684, t = − 1.852, p = 0.071). Having a medical condition was also associated with a change in attitudes. CHW with medical problems changed their responses to a greater extent following the training (coeff = 3.2624, t = 2.083, p = 0.0436). A substantial reduction in score was only observed in CHWs without medical conditions (mean, SD = − 3.35, 5.17), compared to those with medical conditions (mean, SD = − 0.07, 4.98).
CHW’s were more tolerant to rehabilitation of mental health service users in the community following the training (t = − 2.176, df = 44, p-value = 0.018). The only variable that was significantly associated with change in this cluster of attitudes was the corresponding pre-training score, with an increase of 1 point at baseline being associated with a lesser increase in tolerance associated with the training (coeff = − 0.621, t = − 5.478, p < 0.001) after adjusting for differences in age, education and years of service. There was a sub-threshold effect for having a medical condition, with those with a condition demonstrating a reduced change in scores post training (coeff = − 2.105, t = − 1.760, p = 0.086).