Skip main navigation
Close Drawer MenuOpen Drawer Menu

The American Psychiatric Association (APA) has updated its Privacy Policy and Terms of Use, including with new information specifically addressed to individuals in the European Economic Area. As described in the Privacy Policy and Terms of Use, this website utilizes cookies, including for the purpose of offering an optimal online experience and services tailored to your preferences.

Please read the entire Privacy Policy and Terms of Use. By closing this message, browsing this website, continuing the navigation, or otherwise continuing to use the APA's websites, you confirm that you understand and accept the terms of the Privacy Policy and Terms of Use, including the utilization of cookies.

×
Back to table of contents
ArticlesFull Access

Pretraining Skills as Predictors of Competence of Nonspecialists in Delivery of Mental Health Services

Published Online:10 Jan 2023https://doi.org/10.1176/appi.ps.202100691

Abstract

Objective:

Task-shared delivery of mental health care, which includes training people who are not mental health specialists to deliver components of care, has been identified as a core strategy for increasing access to mental health care globally. However, after standard training, nonspecialists attain variable and sometimes poor competence in task-shared mental health care. This study examined whether pretraining interpersonal skills (nonverbal communication, verbal communication, rapport building, and empathy-warmth) are related to posttraining competence in task-shared mental health care among nonspecialists in Nepal.

Methods:

Nonspecialists (e.g., auxiliary health workers and health assistants) (N=185) were assessed at pretraining and posttraining (4 months after training and supervision) in a task-shared mental health care program in Nepal. This study employed both a classification algorithm and a logistic regression model to examine the relationship between pretraining interpersonal skills and posttraining competence.

Results:

The classification model predicted posttraining competence at above-chance levels on the basis of pretraining interpersonal skills. In particular, pretraining nonverbal communication skill distinguished participants whose posttraining competence was rated as acceptable from those whose rating was not acceptable. Nonverbal communication was also a significant predictor in the regression model. No other interpersonal skills were significantly related to posttraining competence outcomes in the regression model.

Conclusions:

Some pretraining interpersonal skills of nonspecialists may predict overall competence outcomes in task-shared mental health care. Future studies confirming the relationship between pretraining interpersonal skills and posttraining competence in care delivery could improve staff selection and training strategies in task-shared mental health care programs.

HIGHLIGHTS

  • Task-shared mental health care, which includes training of nonspecialists in delivery of mental health services, can increase access to services, although posttraining competence among trainees is variable and sometimes poor.

  • Little is known about the relationship between interpersonal skills of nonspecialists and their competence in the delivery of task-shared mental health care.

  • Among nonspecialists in Nepal, pretraining interpersonal skills predicted competence in task-shared mental health care after training and supervision.

  • Further examination of pretraining interpersonal skills may help improve training programs for the delivery of task-shared mental health care by nonspecialists.

A substantial global gap exists between the number of individuals needing and the number of those receiving mental health services, referred to as the mental health treatment gap (1). A worldwide shortage of trained mental health specialists contributes to this gap and has led to recommendations for the provision of task-shared mental health services—or the distribution of mental health care tasks to nonspecialist workers (2). Strong evidence exists in support of the effectiveness, feasibility, and acceptability of task-shared mental health services in a wide range of settings, particularly in low- and middle-income countries (LMICs) (3).

However, evidence also exists that nonspecialists vary in their competence for carrying out task-shared mental health care. In a study in India, 12 of 31 (39%) nonspecialists were evaluated as competent after training and advanced to providing services without an expert supervisor (4). In a study in South Africa, around half of the 12 nonspecialists who completed training were not certified as competent (5). In a study conducted across Liberia, Uganda, and Nepal, 65% of 206 nonspecialists were evaluated as competent (6). This variability in competence further stretches already limited resources for training in task-shared mental health care and potentially undermines the quality and safety of services. Strategies to increase nonspecialist competence in task-shared mental health care represent an important and undeveloped area of research within global mental health services (7). A greater understanding of factors that predict competence after training could increase the scalability of task-shared mental health care, but this area is not yet well studied.

Prior evidence from both high-income countries (HICs) and LMICs suggests that pretraining interpersonal skills may be related to posttraining competence in task-shared mental health care. Research in the United States and Germany has shown that baseline interpersonal skills of trainee mental health specialists, such as nonverbal skills or expressed empathy, are key predictors of competence and improved patient symptoms (8, 9). In a qualitative systematic review representing studies from the United Kingdom, Scotland, Pakistan, India, Nepal, and Zimbabwe, interpersonal skills were identified as critical to the development of competence in task-shared mental health care (10). In a study in Ethiopia aiming to improve nonspecialists’ competence, improvement in interpersonal skills preceded improvement in mental health care competencies (11). In addition, many studies of task-shared mental health care have reported that researchers assessed interpersonal skills or that stakeholders valued strong interpersonal skills in selecting trainees, although those skills and methods for measurement were typically not further specified (12, 13).

To date, however, no study that we are aware of has examined how interpersonal skills predict posttraining competence in task-shared mental health care among nonspecialists. The purpose of this study was to explore whether pretraining interpersonal skills predict posttraining competence among a sample of nonspecialists, with the intention of informing much-needed future efforts to improve nonspecialist competence in task-shared mental health care.

Methods

Setting

Data were drawn from a pilot cluster-randomized controlled trial conducted in Chitwan, Nepal, that aimed to examine the feasibility and acceptability of collaborating with people with lived experience of mental illness when training primary health care workers in the delivery of mental health services (14, 15). Participants with prescribing rights (e.g., health assistants, auxiliary health workers) were trained for 10 days. Five days of training covered psychosocial concepts, verbal and nonverbal communication skills, psychoeducation, emotional support, and case management, and the remaining time focused on World Health Organization (WHO) Mental Health Gap Action Programme (mhGAP) material for mental, neurological, and substance use disorders (16, 17). Participants without prescribing rights (e.g., auxiliary nurse midwives) were trained for only the first 5 days. For the larger study, participants were also allocated to two groups: the intervention group interacted with patients with lived experiences of mental illness during mhGAP training, and the control arm received standard mhGAP training (14). Pretraining data were collected on the first day of training, and posttraining assessment was conducted 4 months after training and supervision.

Approval for the study was given by the Nepal Health Research Council, the Duke University Health System Institutional Review Board, and the George Washington University Committee on Human Research. After participants were given a full explanation of procedures, written informed consent was obtained.

Participants

Nonspecialists were recruited from primary care facilities in Chitwan district. All health care workers, with and without prescribing privileges, at all facilities in which mental health services had not been integrated at the time of the study were eligible.

Enhancing Assessment of Common Therapeutic Factors (ENACT) Rating Scale

This study used an ENACT data set collected pre- and posttraining. The ENACT rating scale is a tool for assessing nonspecialists’ competence in task-shared mental health care, developed in Nepal (18, 19). It has since been adapted across settings for diverse types of task-shared mental health programs (11, 20). The version used in this study includes 18 items, four of which evaluate interpersonal skills (nonverbal communication, verbal communication, rapport building, and empathy-warmth). The other 14 items assess competence in skills taught in task-shared mental health care training sessions (e.g., suicide risk assessment, exploration of prior coping, and explanation of confidentiality). Each item is rated on a scale from 1 to 3: 1, needs improvement; 2, done partially; and 3, done well. The complete measure was used at both time points (pre- and posttraining), with items assessed via standardized 10-minute role-plays. Using locally contextualized vignettes, trained counselors portraying patients with mental health concerns conducted role-plays and rated trainees. Counselors were not involved in providing training and achieved strong interrater reliability prior to conducting ratings (single-measures intraclass correlation coefficient=0.88, 95% CI=0.81–0.93, N=7) (18).

Sample and Analytic Approach

Of 205 participants, complete pretraining data (i.e., interpersonal skills and demographic information) were available for 204, and complete data (i.e., pretraining data plus posttraining competence data) were available for 185—the analytic sample. Pretraining interpersonal skills were operationalized as the scores on the four interpersonal skills items of the ENACT rating scale at pretraining. Posttraining competence was operationalized as the scores on the 14 remaining ENACT rating scale items (regarding mental health care delivery) at posttraining. We dichotomized our outcome, because such an approach would make it easier for trainers to understand which trainees attained minimally acceptable competence (rather than having to interpret quantitative scores along a continuum). The highest possible rating on the 14 posttraining items was 42, and we selected a threshold of 75% (rating ≥32) as indicating “acceptable” competence (a rating <32 indicated “not-acceptable” competence). This threshold was informed by prior work and corresponds to all items done partially plus some items done well (6).

First, we calculated descriptive statistics. To explore possible confounders, we examined the relation between each demographic variable and our outcome by examining unadjusted ORs and by using correlation tests and a Bonferroni correction for multiple comparisons (21, 22). Although some demographic variables were associated with the outcome, no demographic variables were associated with pretraining interpersonal skills; they were therefore not included in further analyses. To investigate systematic differences between the participants who were present versus those who were missing at follow-up, we examined bivariate associations between pretraining interpersonal skills and posttraining missingness. Pretraining interpersonal skills were not related to posttraining missingness.

Next, we used a random-forest approach to examine a model’s ability to predict posttraining class membership (acceptable or not acceptable) and the importance of each of the four interpersonal skills in driving that prediction. Random-forest models are nonparametric methods stemming from machine learning, in which decision trees are independently trained on random samples of a data set and results are pooled (23). Although machine learning is known for analysis of large data sets, a random-forest approach can be used on relatively small data sets (24, 25), and this approach is appropriate for identifying predictors when the relation between predictors and an outcome is nonlinear (26), which we expected could be true in this study. Random-forest models are sensitive to class imbalance in outcomes (27). Therefore, upon examining our outcome, we used the smote function to randomly oversample the less common outcome of acceptable competence, synthetically increasing the model sample size to 350 (28). We used variable importance, determined by combining mean decrease on the Gini index, a measure of node impurity within random forest trees, and mean decrease in accuracy, a measure of the loss of a model’s prediction performance when a particular variable is excluded, with higher variable importance values indicating greater importance, to assess predictive power of variables (29). We used area under the curve, which measures a model’s capability of distinguishing between outcomes, to assess model fit (30).

We also built a multivariate logistic regression model, in which the four interpersonal skills were used as predictors and competence (acceptable versus not acceptable) was our outcome. Each interpersonal skill was dummy coded with a score of 1 (needs improvement) as reference (31). Again, we used area under the curve as a measure of model fit. This analysis used the sample of 185 participants for whom we had complete data.

As a sensitivity analysis, we examined two alternative thresholds of acceptable competence—65% and 85%. To maintain consistency with our approach for the 75% threshold, we again used the smote function before analysis at the 65% threshold, resulting in a sample size of 574. Because of extreme imbalance in outcomes at the 85% threshold, we did not examine the 85% threshold sample beyond the descriptive stage. Otherwise, all methods were the same as for the 75% threshold.

Finally, because prior evidence suggests that improvement in interpersonal skills precedes improvement in competence in task-shared mental health care (11), we examined whether change in interpersonal skills is associated with posttraining competence. We regressed the acceptable competence outcomes on pretraining-to-posttraining change in interpersonal skills by using a logistic regression model and the sample of 185 participants for whom we had complete data. All analyses were conducted in R, version 4.1.0.

Results

Demographic Characteristics

Baseline demographic characteristics are presented in Table 1. Of the 205 participants, 47% were women, 63% were between the ages of 21 and 39, and 70% were from a high-caste group. Five percent had worked in health care for less than 1 year, 30% for 1–5 years, 12% for 6–10 years, and 52% for more than 10 years. Forty-six percent had prescribing privileges. At follow-up, 50% (N=10 of 20) of participants who were missing data had been transferred, had retired, or had their contract end (full posttraining demographic information is reported in the online supplement to this article).

TABLE 1. Baseline characteristics of nonspecialist primary care providers who received training in task-shared mental health care (N=205)

CharacteristicN%
Gender
 Female9647
 Male10953
Age (years)
 21–297436
 30–395627
 40–495225
 ≥502311
Caste or ethnicitya
 High-caste group14470
 Lower-caste group or ethnic minority group6130
Primary care provider group’s prescribing rights
 Without rights (e.g., auxiliary nurse midwife)11054
 With rights (e.g., health assistant)9546
Years working in health care servicesb
 <1105
 1–56230
 6–102512
 >1010752

aCaste is a social stratification that in Nepal is closely correlated with ethnicity.

bData were missing for one participant.

TABLE 1. Baseline characteristics of nonspecialist primary care providers who received training in task-shared mental health care (N=205)

Enlarge table

Pretraining Interpersonal Skills

At pretraining, 22% of participants attained a rating of 1 (needs improvement) on nonverbal communication, 55% attained a 2 (done partially), and 23% attained a 3 (done well) (Table 2). For verbal communication, 52% attained a 1, 45% attained a 2, and 2% attained a 3. For rapport building, 65% attained a 1, 32% attained a 2, and 3% attained a 3. For empathy-warmth, 34% attained a 1, 61% attained a 2, and 5% attained a 3.

TABLE 2. Participants’ ENACT rating scale scores at pre- and posttraininga

ENACT domain, itemPretraining score (N=204)Posttraining score (N=185)
123123
N%N%N%N%N%N%
Interpersonal skills
 Nonverbal communication452211255472321553012869
 Verbal communication1075292455212694517943
 Rapport building13365653263804382442312
 Empathy-warmth70341246110553119646133
Competence in tasks taught in training sessions
 Normalization of feelings84411125584105105577038
 Assesses functioning135666632314826101553619
 Explores explanatory models1607843211<1137150812212
 Assesses coping139686230314926107582916
 Assesses recent stressors13767643131412298534625
 Assesses additional mental health problems, substance use, and general medical health147725426314122108583619
 Appropriately involves important others11556834163341887476435
 Collaborative goal setting1497354261<14524110593016
 Promotes realistic hope for change57281286319912687478646
 Psychoeducation84411065214784136744122
 Problem-solving1527552250—2514138742212
 Elicits feedback80391226021148124674725
 Explains confidentiality1587745221<112869482695
 Assesses suicide risk1748529141<18646904995

aENACT=Enhancing Assessment of Common Therapeutic factors rating scale. Scores: 1, needs improvement; 2, done partially; 3, done well. Scores range from 18 to 54, with higher scores indicating a higher level of interpersonal skill and greater competence in delivery of task-shared mental health care.

TABLE 2. Participants’ ENACT rating scale scores at pre- and posttraininga

Enlarge table

Pretraining and Posttraining Competence

At pretraining, one trainee (<1%) attained acceptable competence on the 14 ENACT rating scale items at the 75% threshold, seven trainees (3%) attained competence at the 65% threshold, and no trainees attained competence at the 85% threshold.

At posttraining, 50 (27%) of the 185 participants attained acceptable competence at the 75% threshold. A total of 103 (56%) attained acceptable competence at the 65% threshold, and seven (4%) attained acceptable competence at the 85% threshold.

Interpersonal Skills as Predictors of Competence: Random-Forest and Logistic Regression Approaches

Our random-forest model had an area under the curve of 0.72 with 75% as the threshold and 0.64 with 65% as the threshold. Importance ratings for the four interpersonal skill variables are reported in Table 3; nonverbal communication was ranked as the most important variable at both thresholds, followed by empathy-warmth.

TABLE 3. Importance of four interpersonal skills as predictors of posttraining competence of nonspecialists (N=185)a

75% threshold65% threshold
RankSkillMean decrease in accuracybMean decrease in Gini indexcMean decrease in accuracybMean decrease in Gini indexc
1Nonverbal communication40.917.952.127.1
2Empathy-warmth32.414.818.211.6
3Rapport building17.89.512.38.7
4Verbal communication17.26.154.07.9

aA 75% threshold in the random-forest model indicated acceptable competence as measured by a score of ≥32 on the Enhancing Assessment of Common Therapeutic factors rating scale, for which the highest (best) possible task competence score is 42 (range 14–42). An alternative threshold of 65% in the random-forest model (score of ≥28) was also tested. The algorithm artificially oversampled the underrepresented outcome (acceptable competence) to balance the data.

bMean decrease in accuracy is a measure of the loss of a model’s prediction performance when a particular variable is excluded (29). Higher variable importance values indicate greater importance of the predictor.

cGini index is a measure of node impurity within random forest trees (29). Higher variable importance values indicate greater importance of the predictor.

TABLE 3. Importance of four interpersonal skills as predictors of posttraining competence of nonspecialists (N=185)a

Enlarge table

Table 4 shows results from the logistic regression model. The model had an area under the curve of 0.70 with 75% as the threshold and 0.65 with 65% as the threshold. At the 75% threshold, participants with a score of 3 (done well) on pretraining nonverbal communication skills had significantly higher odds of posttraining competence (β=1.50, adjusted OR [aOR]=4.50, p=0.04) compared with those with a score of 1 (needs improvement), holding all other predictors constant. At the 65% threshold, participants with a score of 2 (done partially) or a score of 3 (done well) on pretraining nonverbal communication skills had significantly higher odds of posttraining competence (score of 2: β=0.98, aOR=2.67, p=0.03; score of 3: β=1.71, aOR=5.52, p=0.01) compared with those with a score of 1 (needs improvement). Holding all else constant, the odds of achieving posttraining competence with a pretraining score of 3 on nonverbal communication skills were approximately twice the odds with a score of 2. No other variables were found to predict competence.

TABLE 4. Logistic regression model examining pretraining interpersonal skills of nonspecialists as predictors of posttraining acceptable competence (N=185)a

75% threshold65% threshold
Skill and scoreβSEpOR95% CIβSEpOR95% CI
Interceptb−2.41.54<.01——−.57.33.09——
Nonverbal communication
 Score of 2.98.63.122.67.83−10.33.98.45.032.671.13−6.55
 Score of 31.50.72.044.501.20−20.501.71.61.015.521.73−18.88
Verbal communication
 Score of 2.12.37.751.12.54−2.40−.36.35.29.66.35−1.36
 Score of 31.011.30.442.75.23−65.93−.951.15.41.39.04−4.31
Rapport building
 Score of 2.29.38.451.33.63−2.80.36.37.321.44.70−2.98
 Score of 3c−16.701,171.30.99——−.981.50.512.66.20−94.41
Empathy-warmth
 Score of 2.49.46.291.63.68−4.11.04.38.921.04.49−2.17
 Score of 3.33.90.711.40.22−7.92−1.26.84.13.28.05−1.45

aA 75% threshold indicated acceptable competence as measured by a score of ≥32 on the Enhancing Assessment of Common Therapeutic factors rating scale, for which the highest (best) possible task competence score is 42 (range 14–42). An alternative threshold of 65% (score of ≥28) was also tested. ENACT scores: 1, needs improvement; 2, done partially; 3, done well. The reference group for all skills was a score of 1.

bScore of 1 on each of the four skills (verbal communication, rapport building, empathy-warmth, and nonverbal communication).

cThe OR and 95% CI for a score of 3 on rapport building at the 75% threshold could not be fully calculated by the model because so few people attained this score.

TABLE 4. Logistic regression model examining pretraining interpersonal skills of nonspecialists as predictors of posttraining acceptable competence (N=185)a

Enlarge table

Association Between Improvement in Interpersonal Skills and Competence: Logistic Regression

At both the 75% and 65% thresholds, change in interpersonal skills was significantly positively associated with posttraining competence, such that improvement in interpersonal skills was associated with increased odds of posttraining competence (75% threshold: β=0.28, OR=1.32, 95% CI=1.1–1.6, p=0.002; 65% threshold: β=0.34, OR=1.40, 95% CI=1.2–1.7, p<0.001) (for full results, see online supplement).

Discussion

We found that the pretraining interpersonal skills of nonspecialists may predict competence after training and supervision in task-shared mental health care. This result is consistent with prior findings from HICs among more specialized providers and builds on prior work in LMICs that has described the importance of interpersonal skills in successfully taking on task-shared mental health work (8–11). Meaningfully, this result suggests that skills assessment of nonspecialists prior to training could provide valuable information for shaping training and supervision—for example, guiding selection criteria for nonspecialists, determining training length on the basis of pretraining levels of interpersonal skills, or highlighting areas on which training and supervision should focus on the basis of trainees’ baseline skill levels. Tailoring training in this way could help improve nonspecialists’ competence in task-shared mental health care.

Nonverbal communication was ranked as most important in the random-forest model and was the only significant predictor of posttraining competence in the logistic regression model, indicating that it may be the most useful skill to screen at pretraining. Nonverbal skills are also among the easiest interpersonal skill items to assess, requiring raters to examine body language, facial expressions, and eye contact rather than the content of conversation (19). Therefore, even short role-plays of a few minutes could be used to evaluate nonverbal communication. Importantly, all interpersonal skills are culturally defined, meaning that valid rating of nonverbal communication must be culturally informed (32, 33).

Several avenues of future research can follow from this work. First, it is necessary to further understand the predictive validity of nonverbal communication skills and other interpersonal skills with regard to posttraining competence, as well as the predictive validity of other constructs such as mental health knowledge and stigma, and to what degree improvement in these areas predicts improvement in competence. We conducted this analysis by using an exploratory approach with secondary data, but prospective studies of this relationship with larger samples designed to answer these questions are needed. In such research, it will be critical to improve model accuracy, because all models indicated relatively poor fit (34). More recent versions of the ENACT rating scale provide increased guidance for rating interpersonal skills, which could improve model accuracy. Moreover, increased accuracy could be accomplished by increasing the sample size. Larger data sets would also help disentangle possible confounders. Patient-provider hierarchy, which may vary by gender, caste, or race-ethnicity of the individuals in an interaction, may influence interpersonal interactions (35). Other barriers, such as stigma, may also contribute to poor interpersonal skills during interactions with patients who have mental health concerns. Stigma is well documented as being related to poorer care in both HICs and LMICs (6, 36, 37); in prior research on the sample in this study, higher stigma levels were associated with poorer competence (15, 38). It will also be important to examine how years of experience or education level moderate the relationship between interpersonal skills and competence. Turnover may partially explain missing data in our sample; however, future studies should aim to better understand which nonspecialists do not complete training or supervision or fail to implement mental health services.

The relationship between interpersonal skills and posttraining competence should be tested in other settings. Of note, nonverbal and other interpersonal skills may be harder to assess or may present differently over videoconferencing, meaning that future research in this area may not be as applicable to increasingly common virtual training programs (39). Fortunately, there is ongoing work on an ENACT-Remote tool that can be used to assess competence in telephone and videoconferencing formats (40, 41).

Another important area for future work is understanding the feasibility and acceptability of pretraining assessments of nonspecialists—specifically, determining whether nongovernmental organizations and other institutions view pretraining assessment of nonspecialists as a scalable practice in their contexts. Establishing the degree to which pretraining skills inform training approaches is of limited practical use if programs find pretraining assessments unacceptable or impractical. Communication that such assessments are intended to support optimal training and not to exclude individuals from training may be important.

Our findings highlight that it is critical to develop and test effective strategies to support trainees with lower pretraining abilities. This support is especially important in settings with limited pools of health workers or in programs in which all health workers are trained. When those with low skill levels cannot be diverted from training and service delivery, training and supervision must be designed to meet a wider range of needs. Predicting and improving nonspecialist competence are both understudied areas. This knowledge gap could be narrowed by the piloting of interpersonal communication training as an add-on to training in task-shared mental health care, longer courses of training, or different methods of teaching to examine whether these strategies reduce variability in competence outcomes. Making resources for competency-based training and supervision more accessible to implementers is one of the goals of the WHO Ensuring Quality in Psychological Support initiative (42) and associated resources, such as the modular Foundational Helping Skills training (43). Ongoing supervision is also known to be essential in task-shared mental health care (44, 45); future prospective studies should aim to identify the degree to which competence can be improved via supervision after training.

This study was an initial examination of an understudied, critical implementation issue in global mental health, and findings must be interpreted in light of some limitations, including a small sample and data that may not have been missing at random, although missing data were not related to predictors. We were unable to examine some possible confounders with the available data, such as patient-provider hierarchies. Because role-plays lasted 10 minutes, some participants may not have been able to demonstrate all skills within that time. However, each participant received the same amount of time. Because nonverbal skills may be easiest to observe, they may have been more likely to be rated higher in such a time frame and thus may have appeared more influential in analyses. Regardless, our results are consistent with prior work on the relationship between interpersonal skills and competence in mental health care, both in HICs and LMICs. We also used multiple methods, with similar results. However, further studies are essential before formal methods are developed for using interpersonal skills to predict competence among trainees in task-shared mental health care.

Conclusions

Nonspecialists’ interpersonal skills, specifically nonverbal communication, evaluated prior to training in task-shared mental health care may predict posttraining competence. Further studies are needed to replicate these findings and to determine whether pretraining assessment of interpersonal skills can be used to identify which nonspecialists, following training and supervision, are most likely to achieve competence in task-shared mental health care.

Department of Psychology (Rose, Magidson) and Measurement, Statistics, and Evaluation Program (Feng), University of Maryland, College Park; Transcultural Psychosocial Organization Nepal, Kathmandu, Nepal (Rai, Shrestha, Kohrt); Department of Psychiatry and Behavioral Sciences, George Washington University, Washington, D.C. (Rai, Kohrt).
Send correspondence to Ms. Rose ().

This research was supported by grants F31MH123020 (principal investigator [PI], Ms. Rose) and K01MH104310 (PI, Dr. Kohrt) from NIMH and by the Programme for Improving Mental Health Care (PRIME), which is funded by aid from the government of the United Kingdom for the benefit of developing countries.

The authors report no financial relationships with commercial interests.

The funders had no role in study design, data collection and analysis, the decision to publish, or preparation of the manuscript.

The authors thank Anup Adhikari, M.P.H., Manoj Dhakal, M.P.H., Mark Jordans, Ph.D., Nagendra Luitel, Ph.D., Pooja Pokharel, M.P.H., and the PRIME staff of Transcultural Psychosocial Organization Nepal.

References

1. Thornicroft G, Chatterji S, Evans-Lacko S, et al.: Undertreatment of people with major depressive disorder in 21 countries. Br J Psychiatry 2017; 210:119–124Crossref, Medline, Google Scholar

2. Hoeft TJ, Fortney JC, Patel V, et al.: Task-sharing approaches to improve mental health care in rural and other low-resource settings: a systematic review. J Rural Health 2018; 34:48–62Crossref, Medline, Google Scholar

3. Raviola G, Naslund JA, Smith SL, et al.: Innovative models in mental health delivery systems: task sharing care with non-specialist providers to close the mental health treatment gap. Curr Psychiatry Rep 2019; 21:44Crossref, Medline, Google Scholar

4. Singla DR, Weobong B, Nadkarni A, et al.: Improving the scalability of psychological treatments in developing countries: an evaluation of peer-led therapy quality assessment in Goa, India. Behav Res Ther 2014; 60:53–59Crossref, Medline, Google Scholar

5. Rotheram-Borus MJ, Tomlinson M, Roux IL, et al.: Alcohol use, partner violence, and depression: a cluster randomized controlled trial among urban South African mothers over 3 years. Am J Prev Med 2015; 49:715–725Crossref, Medline, Google Scholar

6. Kohrt BA, Mutamba BB, Luitel NP, et al.: How competent are non-specialists trained to integrate mental health services in primary care? Global health perspectives from Uganda, Liberia, and Nepal. Int Rev Psychiatry 2018; 30:182–198Crossref, Medline, Google Scholar

7. Davies T, Lund C: Integrating mental health care into primary care systems in low- and middle-income countries: lessons from PRIME and AFFIRM. Glob Ment Health 2017; 4:e7Crossref, Medline, Google Scholar

8. Schöttke H, Flückiger C, Goldberg SB, et al.: Predicting psychotherapy outcome based on therapist interpersonal skills: a five-year longitudinal study of a therapist assessment protocol. Psychother Res 2017; 27:642–652Crossref, Medline, Google Scholar

9. Anderson T, Ogles BM, Patterson CL, et al.: Therapist effects: facilitative interpersonal skills as a predictor of therapist success. J Clin Psychol 2009; 65:755–768Crossref, Medline, Google Scholar

10. Shahmalak U, Blakemore A, Waheed MW, et al.: The experiences of lay health workers trained in task-shifting psychological interventions: a qualitative systematic review. Int J Ment Health Syst 2019; 13:64Crossref, Medline, Google Scholar

11. Asher L, Birhane R, Teferra S, et al.: “Like a doctor, like a brother”: achieving competence amongst lay health workers delivering community-based rehabilitation for people with schizophrenia in Ethiopia. PloS One 2021; 16:e0246158Crossref, Medline, Google Scholar

12. Murray LK, Dorsey S, Haroz E, et al.: A common elements treatment approach for adult mental health problems in low- and middle-income countries. Cogn Behav Pract 2014; 21:111–123Crossref, Medline, Google Scholar

13. Surjaningrum ER, Jorm AF, Minas H, et al.: Personal attributes and competencies required by community health workers for a role in integrated mental health care for perinatal depression: voices of primary health care stakeholders from Surabaya, Indonesia. Int J Ment Health Syst 2018; 12:46Crossref, Medline, Google Scholar

14. Kohrt BA, Jordans MJD, Turner EL, et al.: Reducing stigma among healthcare providers to improve mental health services (RESHAPE): protocol for a pilot cluster randomized controlled trial of a stigma reduction intervention for training primary healthcare workers in Nepal. Pilot Feasibility Stud 2018; 4:36Crossref, Medline, Google Scholar

15. Kohrt BA, Jordans MJD, Turner EL, et al.: Collaboration with people with lived experience of mental illness to reduce stigma and improve primary care services: a pilot cluster randomized clinical trial. JAMA Netw Open 2021; 4:e2131475Crossref, Medline, Google Scholar

16. Jordans MJD, Luitel NP, Kohrt BA, et al.: Community-, facility-, and individual-level outcomes of a district mental healthcare plan in a low-resource setting in Nepal: a population-based evaluation. PloS Med 2019; 16:e1002748Crossref, Medline, Google Scholar

17. Jordans MJ, Luitel NP, Tomlinson M, et al.: Setting priorities for mental health care in Nepal: a formative study. BMC Psychiatry 2013; 13:332Crossref, Medline, Google Scholar

18. Kohrt BA, Jordans MJD, Rai S, et al.: Therapist competence in global mental health: development of the ENhancing Assessment of Common Therapeutic factors (ENACT) rating scale. Behav Res Ther 2015; 69:11–21Crossref, Medline, Google Scholar

19. Kohrt BA, Ramaiya MK, Rai S, et al.: Development of a scoring system for non-specialist ratings of clinical competence in global mental health: a qualitative process evaluation of the Enhancing Assessment of Common Therapeutic factors (ENACT) scale. Glob Ment Health 2015; 2:e23Crossref, Medline, Google Scholar

20. Pedersen GA, Gebrekristos F, Eloul L, et al.: Development of a tool to assess competencies of Problem Management Plus facilitators using observed standardised role plays: the EQUIP competency rating scale for Problem Management Plus. Intervention 2021; 19:107–117 Google Scholar

21. Armstrong RA: When to use the Bonferroni correction. Ophthalmic Physiol Opt 2014; 34:502–508Crossref, Medline, Google Scholar

22. Khamis H: Measures of association: how to choose? J Diagn Med Sonogr 2008; 24:155–162 Crossref, Google Scholar

23. Bannerman-Thompson H, Bhaskara Rao M, Kasala S: Bagging, boosting, and random forests using R; in Handbook of Statistics 31—Machine Learning: Theory and Applications. Edited by Rao CR, Govindaraju V. Amsterdam, Elsevier, 2013 Crossref, Google Scholar

24. Guo Y, Graber A, McBurney RN, et al.: Sample size and statistical power considerations in high-dimensionality data settings: a comparative study of classification algorithms. BMC Bioinformatics 2010; 11:447Crossref, Medline, Google Scholar

25. Han S, Kim H: On the optimal size of candidate feature set in random forest. Appl Sci 2019; 9:898 Crossref, Google Scholar

26. Breiman L: Random forests. Mach Learn 2001; 45:5–32 Crossref, Google Scholar

27. Bader-El-Den M, Teitei E, Perry T: Biased random forest for dealing with the class imbalance problem. IEEE Trans Neural Netw Learn Syst 2019; 30:2163–2172Crossref, Medline, Google Scholar

28. Torgo L: smote: SMOTE Algorithm for Unbalanced Classification Problems. Gold Coast, Australia, rdrr, 2019. https://rdrr.io/cran/performanceEstimation/man/smote.html. Accessed Apr 11, 2022 Google Scholar

29. Han H, Guo X, Yu H: Variable selection using mean decrease accuracy and mean decrease Gini based on random forest; in 7th IEEE International Conference on Software Engineering and Service Science, Aug 26–28, 2016, Beijing. Edited by Babu MSP, Li W. New York, IEEE, 2016 Google Scholar

30. Melo F: Area under the ROC curve; in Encyclopedia of Systems Biology. Edited by Dubitzky W, Wolkenhauer O, Cho KH, et al.. New York, Springer, 2013 Crossref, Google Scholar

31. Chen J: Dummy coding; in Encyclopedia of Research Design. Edited by Salkind NJ. Thousand Oaks, CA, SAGE, 2010 Google Scholar

32. Joo E, Hill CE, Kim YH: Using helping skills with Korean clients: the perspectives of Korean counselors. Psychother Res 2019; 29:812–823Crossref, Medline, Google Scholar

33. Baugh AD, Vanderbilt AA, Baugh RF: Communication training is inadequate: the role of deception, non-verbal communication, and cultural proficiency. Med Educ Online 2020; 25:1820228Crossref, Medline, Google Scholar

34. Metz CE: Basic principles of ROC analysis. Semin Nucl Med 1978; 8:283–298Crossref, Medline, Google Scholar

35. Moore M: What do Nepalese medical students and doctors think about patient-centred communication? Patient Educ Couns 2009; 76:38–43Crossref, Medline, Google Scholar

36. Knaak S, Mantler E, Szeto A: Mental illness-related stigma in healthcare: barriers to access and care and evidence-based solutions. Healthc Manage Forum 2017; 30:111–116Crossref, Medline, Google Scholar

37. Kane JC, Elafros MA, Murray SM, et al.: A scoping review of health-related stigma outcomes for high-burden diseases in low- and middle-income countries. BMC Med 2019; 17:17Crossref, Medline, Google Scholar

38. Kohrt BA, Turner EL, Rai S, et al.: Reducing mental illness stigma in healthcare settings: proof of concept for a social contact intervention to address what matters most for primary care providers. Soc Sci Med 2020; 250:112852Crossref, Medline, Google Scholar

39. Naslund JA, Shidhaye R, Patel V: Digital technology for building capacity of non-specialist health workers for task-sharing and scaling up mental health care globally. Harv Rev Psychiatry 2019; 27:181–192Crossref, Medline, Google Scholar

40. McBride K, Harrison S, Sudeshna M, et al.: Building mental health and psychosocial support capacity during a pandemic: the process of adapting Problem Management Plus for remote training and implementation during COVID-19 in New York City, Europe and East Africa. Intervention 2021; 19:37–47 Google Scholar

41. Pedersen GA, Pfeffer KA, Brown AD, et al.: Identifying core competencies for remote delivery of psychological interventions: a rapid review. Psychiatr Serv (in press) Google Scholar

42. Kohrt BA, Schafer A, Willhoite A, et al.: Ensuring Quality in Psychological Support (WHO EQUIP): developing a competent global workforce [letter]. World Psychiatry 2020; 19:115–116Crossref, Medline, Google Scholar

43. Watts S, Hall J, Pedersen GA, et al.: The WHO EQUIP Foundational Helping Skills Trainer’s Curriculum [letter]. World Psychiatry 2021; 20:449–450Crossref, Medline, Google Scholar

44. Kemp CG, Petersen I, Bhana A, et al.: Supervision of task-shared mental health care in low-resource settings: a commentary on programmatic experience. Glob Health Sci Pract 2019; 7:150–159Crossref, Medline, Google Scholar

45. Jacobs Y, Myers B, van der Westhuizen C, et al.: Task sharing or task dumping: counsellors experiences of delivering a psychosocial intervention for mental health problems in South Africa. Community Ment Health J 2021; 57:1082–1093Crossref, Medline, Google Scholar