Race, Ethnicity and Gender Identity Discrimination in Access to Mental Health Care

2020-01-28

2021-12-31

We record MHP responses to our email inquiries, as well as record and code any voicemails or text messages we receive. We categorize these responses into eight mutually exclusive categories, as well collapse these outcomes in a binary outcome variable. We consider all of these outcomes to be primary outcomes of interest

Mutually exclusive categories:
Appointment Offer (A): The MHP explicitly offer an appointment.
Call/Consultation Offer (C): The MHP offers to speak on the phone but does not explicitly offer an appointment.
Screening Questions (Q): The MHP requests additional information but does not explicitly offer an appointment.
Waitlist (W): The MHP offers to put the prospective patient on a waitlist.
Referral (R): The MHP gives a referral to an alternative provider but does not explicitly offer an appointment.
Rejection (X): The MHP rejects the client
No Voicemail (NV): We receive a phone call from the MHP’s phone number, but the MHP did not leave a voicemail.
No Response (NR): We do not receive a response from the MHP within one week.

These categories are collapsed in a binary variable that equals one if the MHP responded with an Appointment Offer or Call/Consultation Offer, and zero otherwise.

We conduct a matched pair email correspondence audit in which each MHP receives an email indicating the client is transgender and an email indicating the client is cisgender. Someone transgender identifies with a gender identity that does not exclusively match one’s assigned at birth. Someone who is transgender may identify as the “opposite” to one assigned at birth (e.g., a transgender woman was assigned male at birth), with both genders, or no gender, i.e., both transgender and non-binary (Liszewski et al., 2018). We signal transgender identity using some version of the following phrase: “I am a transgender woman and am looking for a trans-friendly therapist.” or “I am transgender man and am looking for a trans-friendly therapist.” A transgender woman has a feminine name whereas a transgender man has a masculine name. To signal non-binary identity, a client will reveal they are non-binary (i.e. “I am non-binary and am looking for a trans-friendly therapist.”) and including a signature line that signals non-binary status.

We then randomly assign names to signal race, ethnicity and gender. We randomly assign gender, through feminine or masculine names, so that an MHP has an equal probability of receiving an inquiry from a male or female (50/50). Gender is disclosed also through preferred pronouns after the signature at the end of the email, about 50% of the times. For cisgender and transgender people we used pronouns in accordance with the name (e.g., “Pronouns: she/her”), while for non-binary people we used they/them pronouns (e.g., “Preferred gender pronouns: they/them”). We randomly assign names that signal race and ethnicity such that a MHP has approximately a 21.5 percent probability of receiving a name that signals that the prospective client is African American, approximately a 35 percent probability of receiving a name that signals that the prospective client is white, approximately a 22 percent probability of receiving a name that signals that the prospective client is Hispanic, and approximately a 21.5 percent probability of receiving a name that signals that the prospective client is Chinese. Please note that these probabilities are approximates. The final distribution may not reflect this distribution due to random sampling.

Next, we randomly assign insurance status so that an MHP has a 10 percent probability of receiving an inquiry in which insurance is not mentioned, a 16 percent probability of receiving an email in which self-pay with no reference to a slide scale is mentioned, a 14 percent probability of receiving an email in which self-pay with a reference to a slide scale is mentioned, a 30 percent probability of receiving an inquiry in which Medicaid is mentioned, and a 30 percent probability of receiving an email in which private insurance is referenced.

To collect our sample of auditable mental healthcare providers, we use a popular online therapist search database. In order to be included in our sample, an MHP must: (1) treat adults; (2) practice in the U.S., and (3) indicate they treat one of the following mental health disorders: anxiety, depression, or stress. After accounting for these characteristics, we select MHPs proportionately to state populations; specifically, we divide states in terciles, based on their population. Within states, we select MHPs proportionally to the population zip code, conditionally on a minimum population zip code—10,000 residents.

Randomization done in office by a computer.

Individual mental health practitioner

Yes

1,500

3,000

1,500 mental health practitioners with treatment, 1,500 messages without treatment

Hypothesis One: MHPs are less likely to respond to racial minorities than non-Hispanic white individuals. We will compare the relevant outcome for non-white individuals (Chinese, African American, and non-white Hispanics) to Non-Hispanic whites. Minimal Detectable Effect: 0.5 percentage point, based on total n = 3,000 with a baseline response rate of 67.4 (estimate from pilot study) Hypothesis Two: MHPs are less likely to respond to transgender and non-binary individuals than cisgender individuals. We will compare the relevant outcome for cisgender individuals to transgender and non-binary individuals. Minimal Detectable Effect: 0.5 percentage point, based on total n = 3,000 with a baseline response rate of 67.4 (estimate from pilot study) Hypothesis Three: MHPs are more likely to respond to individuals that reference private insurance or self-pay compared to those that reference Medicaid. We will compare the outcome for the inquiries that reference private insurance or self-pay to inquiries that reference Medicaid. Minimal Detectable Effect: 0.8 percentage point, based on total n = 3,000 with a baseline response rate of 67.4 (estimate from pilot study)