Understanding Behavioral Barriers to Demand for Mental Health Services

We provide three separate treatments embedded in a survey to adolescents and parents in our sample. Two treatments use information to encourage the use of mental health services. The first of these provides respondents with information about mental health problems, stigma, prevalence of mental health issues, efficacy of treatment, and more. The second of these provides respondents with a story about a celebrity in Nepal who struggles with mental health issues and who has sought treatment. This celebrity talks about the efficacy of treatment and encourages those with mental health issues to seek help. Our last treatment is matching the gender of enumerators who conduct our survey with respondents. This treatment allows us to look at comfort in sharing information about mental health issues by gender-match.

2021-05-01

2021-08-01

Following are our main outcomes of interest, that are also collected during the baseline survey:
Domain 1: Gender-based Comfort and Trust
1. Willingness to reveal/share personal information by the gender match of enumerator
a. We ask participants if they believe they have a mental health problem—we are interested in participant willingness to reveal this by gender match. We examine whether participants are more likely to report having a mental health issue if they are matched with an enumerator of the same gender.
b. This variable will be coded as a dummy that equals 0 if a participant says they do not have a mental health problem, and 1 if they say they do.
2. Level of comfort/trust by the gender match of enumerator
a. We ask participants their level of comfort in answering questions at the end of the survey. We then measure whether there are differences by same and different gender match enumerators.
b. We ask participants to rate this question on a scale of 1-5. In our regressions, we will code this variable as a standard normal index.
Domain 2: Attitudes Towards Mental Health Problems
3. Willingness to hire/pay someone with a mental health issue
a. Parents are asked in their survey about their willingness to hire someone who has a mental health problem. In particular, they are presented with a vignette with either a man or woman’s name (this is randomized) and told that this person has mental health problems and is seeking work. We ask if they are willing to hire such a person, and, if so, what sort of wages they will pay this person. We are interested in the effects of the gender match treatment as well as the information treatments on this outcome.
b. We ask participants to answer the question about willingness to hire and pay on a scale of 1-5. In our regressions, we will code this variable as a standard normal index.
5. Willingness to seek own/child counseling
a. We ask both parents and children their willingness and preferences for different types of counseling. For instance, we ask if they would rather see in person or online counselors, if they would be willing to seek help if they had a serious emotional problem, and whether they would prefer group or individual counseling.
b. We also specifically ask parents how willing they are to seek help for their children. We ask them to imagine their child had a mental health problem, and then ask them their willingness to seek help for their child, how much money and time they would spend on this, and what types of counseling they would prefer for their child.
c. Preferences for types of counseling will be coded as dummies equal to 0 or 1 depending on preference. For example, we will code the question about face-to-face or online counseling equal to 0 if participants prefer online and 1 if participants prefer face to face. Our question about willingness to seek help for children is answered on a scale of 1-5, which we will code this variable as a standard normal index.

We are primarily interested in two domains: gender-based comfort and trust and attitudes towards mental health problems. The first domain provides information about how gender-matches can be used to increase help seeking behavior, comfort, and trust. For instance, participants may be more comfortable revealing information or talking about mental health with enumerators of the same gender; this information is very important for policy making around mental health. The second domain provides information on how to reduce stigma towards those with mental health problems and how to increase the likelihood that people seek help when struggling with mental health issues.

Secondary Outcomes
Our secondary outcomes of interest are:
1. Attitudes and gender norms about emotions and mental health issues
a. We ask participants questions about masculinity and their beliefs about whether men should be able to express and share their emotions. We are interested in whether our gender matched treatment affects participants’ beliefs about masculinity and men seeking help for mental health problems.
2. Changes in depression and anxiety
a. We measure participants’ levels of anxiety and depression using the PHQ and GHQ-12. We are interested in whether participants report these levels differentially by enumerator match. For instance, are participants more likely to report depressive or anxiety symptoms if matched to a same gender enumerator?
3. Correlation of parent and child mental health levels
a. We are interested in how child and parent mental health may correlate or affect one another. So, we will look at how parent and child mental health status correlate and if they are negatively or positively correlated.
4. Calling of a mental health hotline number we provide
a. At the end of the survey, we present the following text to participants:
“We want to give you a phone number to call to find out more about CMC Nepal an organization specializing in mental health awareness and treatment. You can call this number anytime this month from Monday to Friday 9-5 p.m. Your phone call will be picked up by a trained counselor and he/she will inform you about the mental health information, mental health symptoms and treatment available to you or your children near you. The session will provide you with relevant information related to resources that you could use for mental health issues. In addition, when you call in, you can also mention that you need personalized counselling services and a district counselor will follow-up with you and provide you counselling services. If you are struggling with any mental and emotional issues, please call in and get help. Phone number is:……… We will also send you this phone number via SMS.”
b. This variable will be coded as a dummy variable that is equal to 0 if the participant did not call the hotline and 1 if they did.

We collect a sample of around 3,000 households. These households are randomized into evenly sized cells across the treatment with the information about stigma, the treatment with the information about celebrity role models, a control group (where no information is provided), and a cross-randomization with the gender match treatment. For the gender-match treatment, we stratify by enumerator gender and randomize household assignment to enumerators. Ultimately, we have six different groups.
1. A group that receives no information and male enumerators.
2. A group that receives no information and female enumerators.
3. A group that receives the stigma information and male enumerators.
4. A group that receives the stigma information and female enumerators.
5. A group that receives the celebrity information and male enumerators.
6. A group that receives the celebrity information and female enumerators.

Each of these six groups contains around 500 households.

Of our approximately 3,000 households, we stratify by municipality and then use Stata to randomly assign households into treatments. We select the oldest adolescent in each household to interview. In each household, we ask about preferences for which adult is interviewed and abide by these preferences. For households with no preference, we randomly select whether a mother or father is interviewed using Stata. For households with no father or mother present, we interview the main caregiver of the adolescent.

Our unit of randomization is the household level. For instance, a household (which includes the adolescent being interviewed and the adult being interviewed) is randomized amongst the information interventions as well as the gender intervention.

No

We plan to cluster at the ward level, with around 200 plus wards in our sample.

We plan to observe approximately 3,000 adolescents and 3,000 adults amongst 3,000 households.

We have approximately 1,000 control households and 2,000 treatment households across the information treatments.

Impact sizes for studies that look at issues of mental health tend to be on the order of .15 standard deviations or lower. Haushofer, using a strong intervention that utilizes personal therapy, finds an impact size of .23 SD (2019). Studies that use light touch interventions often find smaller effect sizes closer to .1-.15 SD (Bursztyn et al., 2020). Power calculations suggest that we can detect an MDE of .1 to .13 with a sample range of 2,000-3,000 households.