Disparities in access to appointments for contraceptive services among Black, Hispanic, White, and recently incarcerated women in Alabama, Louisiana, and Mississippi

Abstract

Objective

To measure differences in access to contraceptive services based on history of incarceration and its intersections with race/ethnicity and insurance status.

Data Sources and Study Setting

Primary data were collected from telephone calls to physician offices in Alabama, Louisiana, and Mississippi in 2021.

Study Design

We deployed a field experiment. The outcome variables were appointment offers, wait days, and questions asked of the caller. The independent variables were callers' incarceration history, race/ethnicity, and insurance.

Data Collection Methods

Using standardized scripts, Black, Hispanic, and White female research assistants called actively licensed primary care physicians and Obstetrician/Gynecologists asking for the next available appointment for a contraception prescription. Physicians were randomly selected and randomly assigned to callers. In half of calls, callers mentioned recent incarceration. We also varied insurance status.

Principal Findings

Appointment offer rates were five percentage points lower (95% CI: −0.10 to 0.01) for patients with a history of incarceration and 11 percentage points lower (95% CI: −0.15 to −0.06) for those with Medicaid. We did not find significant differences in appointment offer rates or wait days when incarceration status was interacted with race or insurance. Schedulers asked questions about insurance significantly more often to recently incarcerated Black patients and recently incarcerated patients who had Medicaid.

Conclusions

Women with a history of incarceration have less access to medical appointments; this access did not vary by race or insurance status among women with a history of incarceration.

What is known on this topic

• Many incarcerated women have neglected reproductive health needs that persist after release.

• Nearly one in three people who have been recently released from prison self‐report that they have been discriminated against by the health service providers.

What this study adds

• Using a field experiment methodology, we find that offer rates for appointments for contraceptive services were five percentage points lower for patients with a history of incarceration.

• Schedulers asked questions about insurance significantly more often to recently incarcerated Black patients and recently incarcerated patients who had Medicaid.

1. INTRODUCTION

Reproductive autonomy is an important aspect of the transition out of incarceration. As women establish or continue romantic relationships, care for their children, work toward economic security, and recover mentally and physically from incarceration, it is critical that they should be able to control whether and when to bear a child. However, many incarcerated women have neglected reproductive health needs, and these needs often persist after release. ¹ , ² , ³ , ⁴

The formerly incarcerated are not a protected class in the United States, meaning that providers are legally allowed to discriminate against that group. A survey of people with criminal records in New York found that 42% reported experiencing discrimination based on their incarceration history in some facet of their life. ⁵ While some states have enacted “ban the box” laws to reduce employer and housing‐based discrimination, there are no such laws governing healthcare services. A more recent study found that 27% of those recently released from prison thought that they had been discriminated against by the healthcare service providers. ⁶ Research that directly measures the degree to which the healthcare system discriminates based on criminal history is scarce; regardless, the perception of discrimination may discourage those with a history of incarceration from seeking healthcare services.

Race‐based differences in reproductive health services may impact discrimination based on incarceration history; however, the relationship between race/ethnicity and contraceptive services is complicated. Practitioner‐level biases related to race/ethnicity and socioeconomic status have been shown to impact the recommendations of healthcare providers for long‐acting reversible contraceptive methods to women at risk of unintended pregnancies. ⁷ Hispanic and Native American women are more likely to report an inability to conceive desired children due to sterilization than non‐Hispanic White women. ⁸ Among findings that provide potentially contradictory evidence of disparate behavior by providers, Black and Hispanic women report higher levels of counseling to encourage the use of birth control and Hispanic women have reported higher levels of being counseled to consider sterilization compared with White women, regardless of whether they were proactively seeking family planning services. ⁹

Although they receive higher rates of counseling for contraception, a Mississippi study found that nulliparous Black women were nearly three times as likely as nulliparous White women to use the least effective methods of contraception (e.g., withdrawal, condoms, or natural family planning as opposed to “moderately effective” pills, hormonal injections, and contraceptive rings or “highly effective” implants and IUDs), although overall, more nulliparous Black women (84%) reported current contraceptive usage compared to nulliparous White women (58%). ¹⁰ This suggests that Black women may still face barriers to access to highly effective contraceptive services, despite using contraception at equal or higher rates, while White women may be less likely to be offered contraception.

Health insurance coverage may pose another barrier to healthcare for recently incarcerated women. The majority of incarcerated people in the United States were Medicaid‐eligible prior to incarceration, but the Medicaid Inmate Exclusion Policy mandates that imprisoned individuals lose that coverage, which means that most people are uninsured at the time of their release. ¹¹ Although policies such as the Affordable Care Act have increased the numbers of people who are reenrolled in Medicaid at the time of release, people with recent contact with the criminal justice system are still 16 percentage points more likely to be uninsured than those without such a history. ¹² , ¹³ Research on the general population has found appointment access to be higher for patients who do not have insurance compared with Medicaid. ¹⁴ However, this may be due to a preference for those who are perceived to be self‐employed/self‐insured and may not carry over to those who are recently incarcerated.

In this study, we employ a field experiment to measure differences in access to appointments for contraception among women with and without a recent history of incarceration in three states on the southern Gulf Coast: Alabama, Louisiana, and Mississippi. We chose the study locations because of the high rates of incarceration, poverty, and poor reproductive health outcomes in these three states. ¹⁵ , ¹⁶ , ¹⁷ We hypothesized that recently incarcerated women would be offered fewer and later appointments compared with those without this history. Furthermore, we expected that among recently incarcerated women, those who do not have insurance and those who are White and would be additionally disadvantaged when it came to access; this was based on the literature on racial minorities being prescribed contraception at higher rates. We also assessed differences in the ways that groups of women are treated during the scheduling call (i.e., the types of questions asked) in order to better understand the mechanisms of discrimination.

2. METHODS

2.1. Field experiment

We recruited female undergraduate research assistants (RAs) with similar age and education levels who self‐identified as Black (non‐Hispanic), White (non‐Hispanic), or Hispanic. Using a random number generator, we selected a representative sample of actively licensed medical doctors (MDs) and doctors of osteopathy (DOs) from the pooled medical board physician rosters of Alabama, Louisiana, and Mississippi. Selected physician specialties were limited to family medicine, general medicine, internal medicine, preventive medicine, and obstetrics and gynecology.

Research assistants called practices asking to be scheduled for the next available appointment as a new patient seeking a new prescription for contraception. Callers used pseudonyms (instead of their real names) that signaled their race/ethnicity.

Throughout data collection, RAs alternated mentioning that they were seeking a new provider because they had recently been released from prison and not mentioning prior incarceration. We also alternated insurance status across calls to either Medicaid or uninsured. Historically, a vast majority of recently incarcerated women of childbearing age have been either uninsured or covered by Medicaid, with fewer than 5% reporting private insurance coverage in one study. ¹⁸ Since the implementation of the Affordable Care Act, Medicaid coverage has increased, and uninsurance rates have decreased, among the recently incarcerated in states that have expanded Medicaid. ¹⁹ , ²⁰

A Google voice number with an area code corresponding to the state being called was utilized. Calls were made during normal business hours (9:00 a.m–4:30 p.m). Callers left on hold for more than 5 min at one time abandoned the attempt. Data collection occurred between January 3, 2021 and December 15, 2021.

The date and time of appointments offered were recorded, as well as any other information that the scheduler requested, which included details about the caller's insurance, questions about their health status, and invasive questions surrounding parity and sexual activity. Callers were instructed not to schedule an actual appointment, if offered, and end the call by saying they “had to check their schedule.”

2.2. Salience of signal

People can often correctly guess an individual's race by their voice, accent, and name. ²¹ , ²² , ²³ To assess the strength of the race/ethnicity signal through voice and/or name in our study, we recorded test calls and had respondents listen to them, in random order, using the survey platform, Lucid, which recruits research participants from the general public. ²⁴ , ²⁵ , ²⁶ For this study, we specified that respondents should be adults residing in the United States. Lucid's publicly reported respondent demographics indicate that its respondents skew younger but otherwise align with the demographic characteristics of the general US population (Table S1).

Using the standardized script from the field experiment, each RA recorded otherwise identical opening statements (i.e., introducing themselves using their pseudonyms and requesting the next available appointment). Based on the recordings, respondents to the Lucid survey indicated the race/ethnicity that they perceived the caller to be from the following list: Non‐Hispanic White or Euro–American; Black, Afro–Caribbean, or African American; Latino or Hispanic American; East Asian or Asian American; South Asian or Indian American; Middle Eastern or Arab American; Native American or Alaskan Native; or Other. These results were used to create a variable for the strength of the race/ethnicity signal, defined as the percentage of Lucid survey respondents who correctly identified an individual caller's race/ethnicity.

The outcome variables of interest were appointment offer, wait days, and questions about insurance, ability to pay, health status, and invasive questions. The variable appointment offer was a binary variable indicating whether a caller was offered an appointment. Wait days was a continuous variable representing the difference between the date and time of a call and the date and time of an offered appointment (wait days was conditional on an appointment being offered.) The remaining outcome variables were binary indications of whether the caller was asked at least one question about a specific topic prior to an appointment offer/rejection. Questions about insurance were, for example, whether the caller had insurance and if so, what type. Questions about the ability to pay could be inquiries about the patient's out‐of‐pocket charges. Health status questions were questions about the caller's health in general, and invasive questions were questions about sensitive topics such as sexual history or relationship status, that were unnecessary for scheduling an appointment. We included variables about questions asked in the leadup to an appointment decision to better understand the schedulers' thought processes and the access barriers that patients may face. Control variables included physician license type (MD or DO) and physician specialty (family medicine, general medicine, internal medicine, preventive medicine, and obstetrics and gynecology).

2.3. Analysis

We conducted separate linear regression analyses for each of the outcome variables. The independent variables of interest were binary signals of Black, Hispanic, Medicaid, and recent incarceration (RI) that we varied at the level of the caller/provider combination. Each regression controlled for caller fixed effects, physician license type, physician specialty, fixed effects for state and three‐digit ZIP code, and month, day of week, and hour of call fixed effects. Standard errors were clustered on three‐digit ZIP code. We additionally interacted race/ethnicity with recent incarceration status using the same basic equation. As a robustness check, we ran the models interacting racial/ethnic saliency with self‐reported (or “true”) race/ethnicity.

We also ran the analysis using logit models, using average marginal effects, and including all probability‐based outcomes that were statistically significant from the linear specifications (i.e., appointment offer, ability to pay, and health status). We included estimates using the full sample and, owing to sample loss using the nonlinear models, estimates using identical samples.

This study was approved by the Institutional Review Board at the institution of the first author.

3. RESULTS

Overall, 1285 qualifying practices were reached (Table 1). Patients were offered an appointment in 32% (SD = 0.47) of the cases. Those who were offered an appointment had a mean wait time of 13.2 days (SD = 20.2). In 29% of calls (SD = 0.46), patients were asked questions about their insurance. Questions about ability to pay, health status, and invasive topics were rare (1%–2% of calls).

TABLE 1.

Summary statistics.

		Mean (SD)	N
Primary outcomes	Appointment offer	0.32 (0.47)	1285
	Wait days	13.2 (20.2)	397
Secondary outcomes	Insurance details	0.29 (0.46)	1285
Patient asked about…	Ability to pay	0.02 (0.16)	1285
	Health status	0.01 (0.11)	1285
	Invasive topics	0.01 (0.11)	1285
Patient characteristics	Incarceration status
	Recently incarcerated	0.49 (0.50)	1285
	Race/ethnicity
	Black	0.16 (0.37)	1285
	Hispanic	0.29 (0.45)	1285
	White	0.55 (0.50)	1285
	Insurance status
	Medicaid	0.49 (0.50)	1285
	Uninsured	0.51 (0.50)	1285

Signaling a history of incarceration was significantly associated with lower rates of appointment offer (linear effect (LE) = −0.05; p = 0.08), as was having Medicaid (Table 2). We did not find overall differences in offer rates based on race or ethnicity, nor did we find significant differences in wait days by race, insurance, or incarceration status. In the conversation preceding an appointment decision, schedulers were significantly more likely to ask patients with a history of incarceration about their ability to pay (LE = 0.01; p = 0.03) and less likely to ask about their health status (LE = ‐0.01, p = 0.08) compared with patients with no history of incarceration. Hispanic patients were more likely to be asked invasive questions (LE = 0.03; p = 0.02) and about their ability to pay (LE = 0.06; p = 0.02).

TABLE 2.

Effects of incarceration, race/ethnicity, and insurance on access to reproductive health appointments.

	Appointment				Questions asked of patients during call
	Appointment offer		Wait days		Insurance		Ability to pay		Health status		Invasive question
	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value
Recently incarcerated	−0.05*	0.08	−0.89	0.66	0.05	0.11	0.01**	0.03	−0.01*	0.08	<0.01	0.43
Black	<0.01	1.00	0.84	0.90	0.07	0.25	0.02	0.42	<0.01	0.85	0.01	0.69
Hispanic	−0.08	0.36	0.71	0.92	0.05	0.59	0.03**	0.02	0.02	0.32	0.06**	0.02
Medicaid	−0.11***	<0.01	0.21	0.91	−0.04	0.16	−0.03***	0.01	−0.01	0.11	<0.01	0.39
Overall sample mean (SE)	0.32 (0.47)		13.21 (20.20)		0.29 (0.46)		0.02 (0.16)		0.01 (0.11)		0.01 (0.11)
N	1285		397		1285		1285		1285		1285

Note: Each regression controlled for caller fixed effects, physician license type (MD or DO), physician specialty type (family medicine, general medicine, internal medicine, preventive medicine, and obstetrics and gynecology), fixed effects for state and 3‐digit ZIP code, and month, day of week, and hour of call fixed effects. Standard errors (SEs) were clustered on three‐digit ZIP code. Significance is considered at *p < 0.10; **p < 0.05; ***p < 0.01.

We did not find any significant differences in appointment offer rates or wait days when incarceration status was interacted with race or insurance status (Table 3). Schedulers asked questions about insurance significantly more often of recently incarcerated Black patients (LE = 0.12; p = 0.08) and recently incarcerated patients who had Medicaid (LE = 0.15; p < 0.01). Recently incarcerated Hispanic patients were asked about insurance significantly less frequently (LE = −0.12; p = 0.02).

TABLE 3.

Recently incarcerated by race/ethnicity and Medicaid interactions.

	Appointment				Questions asked of patients during call
	Appointment offer		Wait days		Insurance		Ability to pay		Health status		Invasive question
	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value	Linear effect	p‐Value
Race/ethnicity interactions
RI	−0.04	0.20	−0.06	0.99	0.06*	0.07	0.02**	0.04	−0.01	0.30	0.01	0.20
Black	−0.03	0.72	0.71	0.90	0.02	0.82	0.03	0.32	0.01	0.83	0.01	0.86
Hispanic	−0.06	0.53	2.04	0.77	0.10	0.30	0.05***	0.01	0.02	0.41	0.06**	0.02
RI × Black	0.07	0.38	0.30	0.95	0.12*	0.08	−0.01	0.56	<0.01	0.87	0.01	0.73
RI × Hispanic	−0.05	0.48	−3.57	0.51	−0.12**	0.02	−0.03	0.26	<0.01	0.81	−0.01	0.36
Medicaid interactions
RI	−0.07	0.11	−0.12	0.97	−0.03	0.40	0.01	0.41	−0.02*	0.05	<0.01	0.89
Medicaid	−0.13***	<0.01	1.11	0.52	−0.11***	<0.01	−0.03***	0.01	−0.02*	0.07	<0.01	0.92
RI × Medicaid	0.05	0.40	−1.89	0.66	0.15***	<0.01	0.01	0.48	0.02	0.14	0.01	0.29
Overall sample mean (SE)	0.32 (0.47)		13.21 (20.20)		0.29 (0.46)		0.02 (0.16)		0.01 (0.11)		0.01 (0.11)

Note: Each regression controlled for caller fixed effects, physician license type (MD or DO), physician specialty type (family medicine, general medicine, internal medicine, preventive medicine, and obstetrics and gynecology), fixed effects for state and three‐digit ZIP code, and month, day of week, and hour of call fixed effects. Standard errors (SEs) were clustered on three‐digit ZIP code. We additionally interacted race/ethnicity with recent incarceration status using the same basic equation. Significance is considered at *p < 0.10; **p < 0.05, ***p < 0.01.

Abbreviation: RI, recently incarcerated.

Overall, for recently incarcerated patients, the results for appointment offer and questions about ability to pay are consistent when using a logit model compared with a linear model (Tables S4 and S5). The result for questions related to insurance, health status, and invasive questions, however, differ between the two models. The effect size remains the same (LE = 0.05) but becomes statistically significant in the logit model (p = 0.07). The effect on questions related to health status loses statistical significance in the logit model (p = 0.08 vs. p = 0.44), whereas the effect on invasive questions gains statistical significance in the logit model (p = 0.43 vs. p < 0.01).

There were 1148 responses to the Lucid saliency survey. Notably, across the data, callers had on average a high level of racial/ethnic saliency (Table S2). Only one caller had less than a majority of respondents correctly identify their race or ethnicity. That caller self‐identified as Black, and 34% of respondents identified her correctly, whereas 49% identified her as White, and 2% identified her as Hispanic. The models that examined the impact of the saliency of the racial signal showed that Black (β = 0.008; p = 0.02) and Hispanic patients (β = 0.018; p = 0.07) were more likely to be offered an appointment if they strongly signaled their race/ethnicity through their voice and pseudonyms (Table S3). The same was not true for White patients.

4. DISCUSSION

In this study, we used a field experiment to measure discrimination in access to reproductive health services, specifically, contraception, against women who signaled a recent history of incarceration. We found evidence that recently incarcerated women are offered fewer appointments compared with those without this history. This may subsequently mean that recently incarcerated women have to travel further for the appointments that they are offered, which may be more expensive, time consuming, and complicated than if the provider was close by. It may also lead to missed appointments, which could result in increased stress, missed opportunities to receive other health services at the contraception appointment, and in some cases, unintended pregnancy.

We also found that schedulers asked questions about insurance significantly more often of recently incarcerated Black patients and recently incarcerated patients who had Medicaid. The additional scrutiny around insurance status and ability to pay that recently incarcerated women receive about may constitute another barrier to care. Although our callers were trained in how to respond to these questions, women who are recently incarcerated may be less versed in how to use their insurance compared with women with no history of incarceration. They may also be less likely to have an income source and have less confidence navigating the healthcare system than their counterparts with no history of incarceration.

Findings related to race/ethnicity were unexpected in that we found no significant differences in appointment offer rates or wait days for Black or Hispanic patients compared with their White peers in the main analysis. These results were also observed when race/ethnicity was interacted with incarceration status. Although prior field experiments have detected unequal access to medical appointments based on race/ethnicity, they have been conducted in a general medicine setting. It may be the case that racial dynamics operate differently in the reproductive health setting, where research has shown that Black and Hispanic women are more likely to be offered contraception during a visit. As this is, to our knowledge, the first such field experiment focused on the Gulf coast region (Alabama, Louisiana, and Mississippi), our findings could also reflect regional differences in healthcare access.

When we accounted for racial saliency in the analysis, we found that the strength of the racial signal was positively associated with appointment offer rates for Black and Hispanic patients. The reasons for these findings are beyond the scope of this study; however, our study design may offer some insight. Although our callers identified as different race/ethnicities, they were all students at a competitive, private university. It is likely that in addition to race, their voices signaled a relatively high social class. It may be that our findings reflect preferential treatment of higher class patients who are perceived to be Black or Hispanic compared with those whose race is perceived to be White or whose race is ambiguous. ²¹ , ²² , ²³ Social class may also be driving the null findings on race/ethnicity differences in access. Future research should disentangle the impacts of self‐reported, signaled race/ethnicity and social class. Furthermore, future field experiments should not exclusively use self‐reported race/ethnicity but should account for the salience of the signal as well.

Overall, having Medicaid was more disadvantageous than having a history of incarceration. This reflects trends commonly observed in the general population. However, among recently incarcerated women, here were no differences in appointment access between those with Medicaid versus those with no insurance. This seems to support efforts to enroll incarcerated people in Medicaid prior to release, as it will likely improve their financial status while not impacting the appointments that they are offered.

4.1. Limitations

This study was conducted in three southern states (Louisiana, Mississippi, and Alabama) and may not be generalizable outside of that region. Furthermore, it reflects the earliest available appointment and may overstate access for women with transportation challenges or other constraints. Patients proactively disclosed a history of incarceration, and so results may overstate disparities. Although incarceration status was stated directly during the calls, race/ethnicity was communicated indirectly via name signal and voice, and so we cannot be certain that the schedulers accurately perceived them. However, the results of our robustness check indicate that callers' race/ethnicity signal was correctly understood by the majority of survey respondents, and that the strength of the signal was associated with appointment offer rates in similar patterns to other studies on racial disparities in access.

Estimates generated for our main results in Table 2 using nonlinear probability models are generally, but not always, similar to those generated using nonlinear models. The results hold for appointment offer and ability to pay, but they do not hold for questions about insurance, health status, or invasive questions (Tables S4 and S5). Generally, we included estimates derived from linear probability models in our main tables given challenges that can arise when interpreting estimates on interaction variables. ²⁷ , ²⁸

Another limitation is that the saliency analysis used salience measures that were generated from Lucid respondents, a population that was not the underlying physician office population, potentially introducing bias in the salience measures.

Lastly, this study only addresses one aspect of reproductive autonomy (access to contraception) and cannot be generalized to other aspects such as fertility or abortion services.

4.2. Conclusions

Women who disclose a recent history of incarceration are less likely to be offered appointments for contraceptive services. In addition to increasing the number of providers who are open to recently incarcerated patients, coaching patients not to proactively disclose incarceration history may help reduce these disparities. Laws prohibiting clinic staff from asking about incarceration history except when clinically necessary should also be considered.

FUNDING INFORMATION

This research was supported with grant funding from the Newcomb Institute at Tulane University.

Supporting information

Data S1: Supporting information.

Wisniewski JM, Walker B, Patlola I, Sharma R, Tinkler S. Disparities in access to appointments for contraceptive services among Black, Hispanic, White, and recently incarcerated women in Alabama, Louisiana, and Mississippi. Health Serv Res. 2024;59(2):e14275. doi: 10.1111/1475-6773.14275