Geographic distribution of assisted reproductive technology clinics in the USA: a multilevel sociodemographic analysis

Abstract

Purpose

To investigate whether sociodemographic characteristics of US Census tracts and counties and state-level infertility insurance policy are associated with the presence of assisted reproductive technology (ART) clinics.

Methods

Multilevel logistic regression analyses using publicly available reports of ART clinic locations (2014–2018) matched with sociodemographic data from the US Census Bureau and state infertility insurance policy information.

Results

At the tract-level, multivariate multilevel logistic regression found significant associations with the likelihood of an ART clinic in a tract and the size of the tract population (adjusted odds ratio (aOR): 1.063, SE = 0.018, p < .001), the tract median household income (aOR = 0.990, SE = 0.002, p < 0.001), and the percentages of the population who were Hispanic (aOR = 0.975, SE = 0.007, p < 0.001), women over 25 with a bachelor’s degree or higher (aOR: 1.052, SE = 0.004, p < 0.001), and foreign-born (aOR: 1.037, SE = 0.009, p < .001). At the county-level, significant associations were found with the county median household income (aOR: 1.016, SE = 0.006, p < .01) and the percentage of the population that identified as Black (aOR = 1.013, SE = 0.006, p < .05) and Hispanic (aOR = 1.028, SE = 0.009, p < .05). Multivariate models showed no associations between tract clinic counts and state infertility policy.

Conclusion

There is mixed evidence that clinic concentration is associated with expected sociodemographics. In particular, physical proximity may not principally drive racial disparities in ART access. Furthermore, insurance mandates are not associated with the presence of an ART clinics in a tract, suggesting alternative policy levers may be needed to address differential access and utilization of ART services.

Introduction

Infertility, defined as the inability to become pregnant after 12 months of regular unprotected intercourse, is estimated to impact 38–47% of women in the USA in their lifetimes [1–3]. Clinical treatments for infertility range from counseling to ovulation medications to intrauterine insemination to in vitro fertilization and related assisted reproductive technologies (ART). Note, we define ART as treatments in which both egg and sperm or an embryo are handled outside of the body [4].

As of 2017, ART births comprised 1.9% of all births in the USA [5]. However, despite growth in US ART births, access and use of ART and live births from ART are concentrated among White and Asian women, foreign-born women, and those with higher socioeconomic status (SES) [6–12]. Notably, the prevalence among infertility for Black women, Hispanic women, and women who have lower SES is higher or equivalent to that of populations that use these services more often [13–15]. Thus, it is unlikely that disparities in ART utilization are the result of differential need.

As a result, research on the mechanisms underlying disparities in ART use remains an active area of inquiry. In particular, scholars have focused on understanding the barriers to use of needed and desired infertility services. Adashi and Dean [16] have provided a useful conceptual framework and categorization of these barriers which include factors such as sociocultural, emotional, infrastructural, geographic, and economic barriers. Research on each of these barriers is developing, yet, the issue of geography has remained relatively understudied.

An important exception is the recent work of Harris and colleagues [17] that investigated the geographic distribution of ART clinics in the USA for the years of 2009–2013. They found ART clinics were concentrated in metro areas and that many women of reproductive age live far from any ART clinics. This contribution demonstrated the uneven distribution of these clinics, but the analyses did not address the issue of disparities or sociodemographic composition of areas where clinics are located. That is, in addition to being concentrated in metro areas, it is possible that ART clinics may also be concentrated in areas farther from underserved populations (e.g., Black populations and populations with lower SES). If clinics are further or less accessible to underserved populations, this uneven distribution of clinics may serve as a structural barrier to care.

In other domains of women’s healthcare utilization, such as abortion, contraception use, cervical and breast cancer screenings and treatments, and infant and maternal mortality, geographic distribution of clinics or hospitals and their linked sociodemographic characteristics and distance to care, generally, are commonly researched and have been associated with utilization disparities [18–26]. Yet, it is not clear how ART clinics are distributed at the micro-level. Similarly, little work has sought to specifically quantify the role of geographic access to ART utilization. However, given prior research, it is possible that the geographic distribution of ART clinics may contribute to the observed racial and socioeconomic disparities in US ART use. On the other hand, ART is a relatively rare, complex, and specialized treatment modality. Like other specialized treatments and as Harris and colleagues found [17], ART providers are concentrated in metropolitan areas, which also have higher concentrations of people of color. Yet, racial and ethnic and socioeconomic disparities in ART and other specialized care persist. As a result, geographic access may be less important to ART utilization and utilization disparities than observed in other domains of women’s healthcare utilization. Even still, understanding the sociodemographic characteristics of the areas where ART clinics are located has the potential to clarify the role of geographic barriers in disparities in ART use.

For ART access in the USA, states are also a particularly important geographic unit. The number of clinics in the USA varies by state [5]. However, state sociodemographics may be overly broad in attempting to specify the relationship between these factors and ART clinic distribution. At the same time, infertility insurance mandates that operate at the state-level may have a consequential impact on the distribution of ART clinics. Specifically, in the USA, some states have legislated requirements for insurance companies to include coverage of infertility care. While the impact of utilization following these mandates is mixed and tends to suggest that it increases utilization among populations who use services more often already [27–30], it is less clear whether these insurance mandates are associated with the presence of ART clinics more generally.

Thus, in this paper, there are two central aims. First, I build upon and expand Harris and colleagues’ [17] work with a multilevel analysis of the likelihood of any ART clinics across US Census tracts between 2014 and 2018. I quantify how US Census tract and county sociodemographic characteristics such as racial composition, educational attainment of women, nativity, population size, and median household income are associated with the presence of any ART clinics in US Census tracts (hereafter: tracts). Given previous findings, I also hypothesized that ART clinics would be concentrated in areas that had larger population sizes and were comprised of more advantaged or well-served populations (high income, White, and foreign-born). Second, I investigate whether state infertility insurance mandates are associated with the presence of ART clinics in tracts net of sociodemographic tract and county factors.

Materials and methods

National ART Surveillance System (NASS) data

ART clinics in the USA are required to report success rates per federal regulations [31]. These reports are aggregated and publicly available on the CDC’s website. These reports include addresses for each of the reporting clinics. Using this geographic information for clinics that reported data to the CDC between 2014 and 2018, we matched these addresses to counties and Census tracts using multiple sources including the StatsAmerica City-to-County Finder (http://www.statsamerica.org/CityCountyFinder/), the US Census Bureau Geocoder (https://geocoding.geo.census.gov/), and Geocodio (https://www.geocod.io/). Importantly, because the NASS dataset includes only the main address of ART clinics, satellite clinics and other infertility providers are excluded from these analyses. This issue is discussed in detail in the “Limitations” section below.

After matching the individual clinics with state-, county-, and tract-level identifiers, we calculated the number of clinics in each tract. Clinics that shared the same name and were in the same tract or at the same address across years were counted as one clinic in the aggregated file. That is, clinics that reported multiple times in the period were only counted once. However, clinics with multiple locations across counties or states are counted as separate clinics. Furthermore, only clinics that remained open for the entire study period were included. This decision was made to better align with the Census estimates currently available for tracts. Furthermore, due to small counts of clinics at the tract-level, our central outcome variable was a binary measure of whether there were no ART clinics or one or more clinics in the tract during the study period.

US Census Bureau’s population demographics

County and tract population and sociodemographic characteristics were collected from the US Census Bureau website. We gathered the 2014–2018 estimates of the tract and county population counts overall and by (1) race and ethnicity, including persons who identified as Hispanic, non-Hispanic (NH) Black, NH White, Native American and Alaska Native (AIAN), NH Asian, NH Native Hawaiian or other Pacific Islander (NHOPI), and NH persons of another race or multiple races, (2) educational attainment for women over 25 years of age, including the categories less than a bachelor’s degree and a bachelor’s degree or more, (3) nativity, including the categories foreign- and US-born, and (4) median household income. For race, education, and nativity composition, we calculated the percentage of the population in each of these groups across geographic levels.

These population estimates are drawn from the American Community Survey (ACS)—a nationally representative, cross-sectional survey administered by the US Census Bureau [33]. However, we did not aggregate these counts from the ACS as the publicly available ACS dataset does not include tract-level information nor information for counties with populations smaller than 100,000 persons. In addition, population estimates of all tracts and counties are only available in 5-year periods. Thus, we use the available aggregated data from the US Census Bureau that estimates the population for 2014 to 2018.

State insurance mandates

Information on the infertility insurance policy mandates was collected from RESOLVE, the National Infertility Association [34]. We measure the infertility insurance mandate policy in two ways. First, we include a binary indicator for whether the state had no infertility insurance mandate in the period or had a mandate any time within the period under study. Second, because the nature of these infertility insurance mandates varies, we also include an analysis that categorizes these mandates as (1) insurance mandate with IVF coverage, (2) insurance mandate without IVF required or unspecified, and (3) no mandate during the study period.

Data availability

The data underlying this article are publicly available online from the CDC [35] and US Census Bureau [36].

Analytic method

We matched the US Census data with the clinic data by merging the files on the basis of the geographic identifiers from the US Census Bureau. This work was aided by two user-written Stata packages [37, 38]. Next, we carried out listwise deletion for missing data; 370 tracts were excluded due to missing data from the US Census Bureau. In total, the study included 72,358 tracts, 2966 counties, and data from 348 unique clinics.

To analyze these data, we first conducted a series of t-tests that compared the population and sociodemographic characteristics in tracts without any and with at least one ART clinic. We completed these descriptive analyses at the tract- and county-levels. Next, we employed a multivariate, two-level random-intercept multilevel logistic regression model. In our model, we nested tracts (level-one) within counties (level-two). A multilevel model (also called a mixed linear model or hierarchical model) was used to address the nested nature of our data [39–42]. An LR test confirmed the presence of clustering at the county-level, which further supported our modeling approach. By contrast, we did not employ a three-level model with counties nested within states because both an LR test was insignificant and analyses using this additional level did not improve model fit. In fact, once county-level clustering was accounted for, very little variation at the state-level remained. Finally, our model included both tract-level and county-level sociodemographic variables, which allows us to estimate both within-tract and between-county effects [43, 44]. Given the use of reintroduced means at the second level, we did not center the covariates [43, 44].

Following our sociodemographic analysis at the tract- and county-level, we next investigated the role of state insurance mandates on our findings. In this analysis, we incorporate our state-policy indicator into our final sociodemographic model. Importantly, mandates primarily operate at the state-level. However, once the two-level model (tracts nested within counties) was fit, there was little residual variation. Thus, the results using a three-level model (tracts nested within counties nested within states) do not differ from the model results presented (available upon request).

All analyses were conducted in Stata 17, and all logistic regression estimates are provided as odds ratios to aid interpretation of the coefficients. The county-level map was created in Stata 17, and the tract-level maps were created in GIS.

Sensitivity analyses

Alternative specifications using the count data, rather than a binary outcome, modeled with a Poisson regression with robust standard errors and with a negative binomial regression are similar to the results presented and are available upon request. In addition, we conducted the analyses using all clinics that were open at any time between 2014 and 2018 and the results were similar to those presented below (available upon request).

IRB statement

This study was not considered human subject research based upon the Common Rule definition, and, therefore, was exempt from IRB review.

Results

Descriptive and bivariate statistics

In 2014–2018, there was an average of 0.005 (SE = 0.07), 0.11 (SE = 0.69), and 6.86 (SE = 9.50) ART clinics that remained open for the entire period in tracts, counties, and US States plus the District of Columbia, respectively. The number of clinics in each tract ranged from 0 to 2, 0 to 21 in each county, and 0 to 54 in each state. The majority of tracts and counties had no ART clinics (99.55% and 94.43%, respectively). By contrast, the majority of states had three or more ART clinics (65.07%). Figure 1 shows the concentration of ART clinics across US counties. Figure 2A–H provide tract-level maps of US counties with more than 5 ART clinics (n = 8).

Fig. 1.

Choropleth map of the number of assisted reproductive technology (ART) clinics by US county, 2014–2018. Data Source: 2014–2018 National ART Surveillance System, Author’s analysis

Fig. 2.

A–H Tract-level location of assisted reproductive technology (ART) clinics among counties with more than five ART clinics, 2014–2018.

Source: 2014–2018 National ART Surveillance System, Author’s analysis. Tract-level maps created by: Jason Glatz, Western Michigan University Libraries Mapping Service

Table 1 provides the mean values for the population and sociodemographic characteristics for tracts with no ART clinics compared with one or more ART clinics. A t-test of the differences in these means is provided for the tract- and county-levels, but such results should be treated with caution considering the issue of error clustering, which is addressed with our multilevel analyses.

Table 1.

Means, standard errors, and t-tests comparing sociodemographic characteristics of areas with and without any ART clinics by tract and county

	Tract					County
	Mean 0 ART clinics (N = 72,358)	Mean 1 + ART clinic (N = 328)	Difference	SE	Sig	Mean 0 ART clinics (N = 2966)	Mean 1 + ART clinic (N = 175)	Difference	SE	Sig
Racial composition
Hispanic	16.34	11.93	4.41	1.18	***	8.87	15.63	− 6.76	1.06	***
Black, NH	13.33	8.62	4.71	1.19	***	8.63	13.95	− 5.33	1.12	***
White, NH	61.17	64.50	− 3.33	1.69	*	77.44	60.86	16.59	1.54	***
AIAN, NH	0.74	0.28	0.47	0.25	NS	1.91	0.37	1.54	0.59	**
Asian, NH	4.82	11.35	− 6.53	0.50	***	1.07	6.18	− 5.12	0.20	***
NHOPI, NH	0.15	0.23	− 0.09	0.06	NS	0.08	0.18	− 0.10	0.05	*
Other, NH	2.57	3.10	− 0.53	0.14	**	2.01	2.83	− 0.82	0.13	***
Female educational attainment, 25 years +
More than a BA	30.15	53.79	− 23.64	1.06	***	21.97	37.63	− 15.66	0.66	***
Nativity
Foreign-born	12.48	18.77	− 6.29	0.75	***	1.16	13.76	− 12.60	0.35	***
Median household income (in $1000)	63.68	85.58	− 21.90	1.80	***	50.64	67.50	− 16.85	1.02	***
Population sizea	4.43	5.00	− 0.57	0.13	***	0.57	8.73	− 8.16	0.21	***

^*p < .05, **p < .001, ***p < .001

Notes: ^aPopulation size per 1000 for tracts and per 100,000 for counties

Abbreviations: ART, assisted reproductive technology; NS, not significant at the .05 level; SE, standard error; Sig, significance; BA, bachelor’s degree; NH, non-Hispanic; AIAN, Native American and Alaska Native; NHOPI, Native Hawaiian or other Pacific Islander

Multilevel analysis

Net of the county-level covariates and accounting for the clustering of errors, we found several significant associations between tract-level sociodemographics and the likelihood of an ART clinic (Table 2). Importantly, because we introduce the weighted means of these variables at the county-level, the coefficients for the tract-level variables can be interpreted as the tract-level effects separate from any contextual, county effects. First, we observed a negative association with percentage of the population that identified as Hispanic (adjusted odds ratio = 0.975, standard error = 0.007, p < 0.001). This odds ratio indicates that net of the other covariates, for every one percentage point increase in the Hispanic population relative to the White population, the likelihood of a tract having an ART clinic is reduced by a factor of 0.975. Second, the final model showed, for every one percentage point increase in the female population over 25 years of age with a bachelor’s degree or higher at the tract-level, the likelihood of a tract having an ART clinic increased by a factor of 1.052 (SE = 0.004, p < 0.001). Third, net of the covariates, for every one percentage point increase in the foreign-born population, the likelihood of a tract having an ART clinic increased by a factor of 1.037 (SE = 0.009, p < 0.001). Fourth, net of the covariates, for every $1000 dollar increase in the median tract-level household income, the likelihood of the tract having an ART clinic was reduced by a factor of 0.990 (SE = 0.002, p < 0.001). Finally, net of the covariates, for every 1000-person increase in the tract population, the likelihood of the tract having an ART clinic increased by a factor of 1.063 (SE = 0.018, p < 0.001).

Table 2.

Multilevel random-intercept null and multivariate logistic regression models predicting likelihood of ART clinic in a tract

	Random-intercept, null model	Random-intercept, full model
Tract-level variables
Racial composition (ref: Percent White, NH)
Percent Hispanic		0.975***
		(0.007)
Percent Black, NH		0.992
		(0.005)
Percent AIAN, NH		0.981
		(0.067)
Percent Asian, NH		0.991
		(0.008)
Percent NHOPI, NH		1.042
		(0.036)
Percent Other, NH		1.015
		(0.026)
Female educational attainment population 25 + (ref: Percent with less than a BA)
More than a BA		1.052***
		(0.004)
Nativity (ref: Percent US-born)
Foreign-born		1.037***
		(0.009)
Median household income, per $1000		0.990***
		(0.002)
Population size, per 1000		1.063***
		(0.018)
County-level variables
Racial composition (ref: Percent White)
Percent Hispanic		1.028**
		(0.009)
Percent Black, NH		1.013*
		(0.006)
Percent AIAN, NH		0.918
		(0.095)
Percent Asian, NH		0.986
		(0.017)
Percent NHOPI, NH		1.112
		(0.103)
Percent Other, NH		0.858
		(0.250)
Female educational attainment population 25 + (ref: Percent with less than a BA)
More than a BA		0.998
		(0.016)
Nativity (ref: Percent US-born)
Foreign-born		0.973
		(0.016)
Median household income, per $1000		1.016**
		(0.006)
Population size, per 100,000		1.004
		(0.003)
Constant	0.004***	0.000***
	(0.000)	(0.000)
Counties variation (not exponentiated)	0.276**	0.000
	(0.091)	(0.000)
N	72,686	72,686

Exponentiated coefficients; Standard errors in parentheses

^*p < 0.05, **p < 0.01, ***p < 0.001

Abbreviations: ART, assisted reproductive technology; BA, bachelor’s degree; NH, non-Hispanic; AIAN, Native American and Alaska Native; NHOPI, Native Hawaiian or other Pacific Islander

The coefficients on the county-level variables in the final model provide information on how county-level contextual factors are associated with the likelihood of ART clinics’ net of tract-level effects (Table 2). First, in contrast to the tract-level findings, we observed positive associations between the likelihood of a tract having an ART clinic and percentage of the county population that were Hispanic or Black (aOR = 1.028, SE = 0.009, p < 0.01 and aOR = 1.013, SE = 0.006, p < 0.05, respectively). Similarly, while the association between median household income was negative at the tract-level, we observed a positive association at the county-level (aOR = 1.016, SE = 0.006, p < 0.01). No other county-level effects were significant.

Impact of state infertility insurance mandates

Next, we assessed the impact of state infertility insurance mandates on the likelihood or presence of any ART clinics in a tract (Table 3). Net of the tract and county covariates, there is no significant impact of any insurance mandate or of specific types of insurance mandates on the likelihood of an ART clinic being in a particular tract.

Table 3.

Multilevel multivariate logistic regression predicting likelihood of ART clinic in tract including state insurance infertility mandate indicators

	Infertility mandate binary	Infertility mandate type
Tract-level
Racial composition (ref: Percent White)
Percent Hispanic	0.975***	0.975***
	(0.007)	(0.007)
Percent Black, NH	0.992	0.992
	(0.005)	(0.005)
Percent AIAN, NH	0.981	0.981
	(0.068)	(0.068)
Percent Asian, NH	0.991	0.991
	(0.008)	(0.008)
Percent NHOPI, NH	1.043	1.043
	(0.036)	(0.036)
Percent Other, NH	1.014	1.014
	(0.026)	(0.026)
Female educational attainment population 25 + (ref: Percent with less than a BA)
More than a BA	1.051***	1.051***
	(0.004)	(0.004)
Nativity (ref: Percent US-born)
Foreign-born	1.037***	1.037***
	(0.009)	(0.009)
Median household income, per $1000	0.990***	0.990***
	(0.002)	(0.002)
Population size, per 1000	1.062***	1.062***
	(0.018)	(0.018)
County-level
Racial composition (ref: Percent White)
Percent Hispanic	1.028**	1.028**
	(0.009)	(0.010)
Percent Black, NH	1.013*	1.013*
	(0.006)	(0.006)
Percent AIAN, NH	0.910	0.910
	(0.097)	(0.097)
Percent Asian, NH	0.988	0.987
	(0.018)	(0.018)
Percent NHOPI, NH	1.112	1.118
	(0.103)	(0.106)
Percent Other, NH	0.878	0.884
	(0.257)	(0.260)
Female educational attainment population 25 + (ref: Percent with less than a BA)
More than a BA	0.997	0.998
	(0.016)	(0.016)
Nativity (ref: Percent US-born)
Foreign-born	0.972	0.973
	(0.016)	(0.016)
Median household income, per $1000	1.016**	1.017**
	(0.006)	(0.006)
Population size, per 100,000	1.004	1.005
	(0.003)	(0.003)
State-level
Insurance mandate (ref: No)
Yes	0.928
Insurance mandate type (ref: No mandate)	(0.124)
IVF-specific mandate		0.902
		(0.160)
Other mandate or non-specific mandate		0.943
		(0.140)
Constant	0.0003***	0.0003***
	(0.000)	(0.000)
Variation (counties)	0.000	0.000
	(0.000)	(0.000)
N	72,686	72,686

Exponentiated coefficients; Standard errors in parentheses

^*p < 0.05, **p < 0.01, ***p < 0.001

Abbreviations: ART, assisted reproductive technology; BA, bachelor’s degree; NH, non-Hispanic; AIAN, Native American and Alaska Native; NHOPI, Native Hawaiian or other Pacific Islander

Discussion

ART utilization in the USA is patterned by race, ethnicity, SES, and nativity. This paper investigated one of the potential mechanisms underlying these disparities—geographic distribution of care. Geographic access to health care is an important social determinant of health and has been associated with disparities in utilization of care in other venues. However, while prior research on ART clinics in the USA demonstrated clinics are found predominately in metro areas [17], it was unclear whether ART clinics were concentrated in more sociodemographically advantaged areas as well. As a result, in this paper, we conducted a multilevel analysis of the association between geographic sociodemographics and the presence of any ART clinics in US Census tracts between 2014 and 2018.

The results of this paper align with prior research [17] as we found a positive association between tract population size and the presence of an ART clinic at the tract-level, and our results demonstrate many US tracts and counties were without even a single ART clinic in 2014–2018. Thus, ART clinic access in the USA is, broadly, geographically limited. In addition, the results presented expand our understanding of how geographic access may contribute to ART utilization disparities. Given previous research [9, 10, 45–49], we also expected higher income and education and more White and foreign-born populations would be associated with the presence of ART clinics.

These hypotheses were only partially supported. At the tract-level, we found the expected significant positive associations with percentage of population that were women with a bachelor’s degree or higher and were foreign-born. Similarly, at the county-level, we found a positive association with median household income and the presence of ART clinics. Interestingly, the tract- and county-level associations with median household income were in opposite directions. This finding implies that wealthier counties are associated with a higher likelihood of an ART clinic in a particular tract, but that the wealthier the individual tract is, the less likely an ART clinic is to be present. It may be that there is a threshold effect such that tracts that have a particular level of wealth will be unlikely to have an ART clinic even as clinics themselves are more likely to be in counties with higher average household incomes. Overall, however, these results provide some support for the idea that ART clinics are located in areas with greater potential demand and material access. In turn, this distribution of clinics may help to enable use of care among particular populations.

By contrast, our hypotheses related to race and ethnicity were largely unsupported with the exception of the negative association at the tract-level between the percentage of the population that were Hispanic and the likelihood of an ART clinic in a tract. By contrast, no other tract-level race variables were significantly associated with ART clinic presence. The county-level findings add further information. Specifically, we observed that net of the tract-level effects, the percentages of the population that identified as Black or Hispanic at the county-level were positively associated with the likelihood of a tract having an ART clinic.

This finding is somewhat surprising, but may be spurious and the result of omitted variables. For example, these findings may be the result of the concentration of these populations in urban areas, which are also more likely to have ART clinics [17, 50, 51]. Although our findings adjust for population size, it may be that this operationalization does not sufficiently account for urbanicity. Similarly, these results may be due to the zoning of areas where populations of color live as compared with White populations. For instance, neighborhoods with more Black and first-generation immigrant populations are more often zoned in ways that permit non-residential buildings [52, 53]. Thus, it may be the case that these populations are more likely to live in places where ART clinics are permitted to open relative to more residential areas where White populations are concentrated. Future research should investigate these and other possibilities further by operationalizing urbanicity and exploring other potential mechanisms that could explain these unexpected associations.

Limitations

This study has several limitations. First, the role of geographic access and distribution of ART clinics on ART utilization is not yet clear. Second, due to data limitations, we analyzed any clinics that reported data to the CDC between 2014 and 2018. This approach does not account for clinic closures or openings, and, thus, is less dynamic and less consistent with how clinics may proliferate. While an analysis of clinic openings and closures would be informative, it was outside the scope of the present study. Third, the present study only includes the main addresses of ART clinics reported to the CDC. A newly constructed dataset based on the ART NASS data found 586 “satellite clinics” for the 471 main ART clinics in 2021 [32]. This data suggests that satellite clinics are widespread, which may expand geographic access to care in ways that are not quantified in this study. Yet, these satellite clinics likely do not provide ART procedures, instead serving as locations for consultation and monitoring [32]. As a result, people who need ART services may eventually need to access care at the main clinics. Due to the focus on main clinics, the present study cannot and should not be used to make inferences about general access to infertility care. Future research on these satellite clinics as well as other providers who provide infertility care would provide a more complete picture of geographic access to infertility care throughout the USA. Finally, due to limited data availability, this paper treated the US Census Bureau data as population counts and does not account for measurement error in the ACS in the analyses.

Implications for practice and policy

Overall, the results of this study provide mixed evidence that ART clinics are concentrated in areas with populations that use ART more often. The findings presented have important implications for practice and policy. For practice, these results imply that geographic access may not be a central factor contributing to disparities in utilization by race or ethnicity. Thus, providers and clinics must continue to explore and address other mechanisms that contribute to differential treatment utilization. At the same time, the results also demonstrate geographic access may be a more universal barrier to care as many tracts and counties in the USA do not have even a single clinic. Thus, the findings provide more information about where providers may be able to locate clinics to improve access to care for geographically underserved populations. In addition, the results also point to the potential usefulness of telemedicine to expand availability of care as others have noted [17, 54].

For policy, these results provide three key insights. First, the results show that there may be geographic differentiation in the location of ART clinics by socioeconomic status. This finding implies interventions that incentivize or otherwise provide support in opening ART clinics in socioeconomically underserved areas could be beneficial. However, given the high costs of ART, geographic access alone is very unlikely to dramatically impact utilization disparities by SES. Thus, the results reinforce the importance of creating policies that make ART more affordable. Second, and relatedly, the results show that state infertility insurance mandates are not associated with the presence of ART clinics, which suggests ART clinic presence is not sensitive to this policy lever. However, it is possible that mandates work via other pathways such as increased utilization of services. Third, the results suggest that geographic barriers to care may not be central to explaining the widespread disparities by race and ethnicity in infertility care. Thus, the results also point to the critical need for research on why populations who may need infertility services do not access care. In order to address the well-documented disparities in infertility treatment utilization, as called for by CDC’s National Public Health Action Plan [55] and recent ethics opinions from the ASRM [45, 46], continued effort is needed to better understand the individual, social, and societal reasons why utilization of infertility treatments remains unequal in the USA.

Author contribution

The author was solely responsible for the design, execution, analysis, and manuscript drafting.

Declarations

Ethics approval

The present study conducts secondary analyses of publicly available data. The data used for ART clinic locations is not considered human subject research as the unit is clinics. The data used for population estimates comes from the US Census Bureau and is deidentified and cannot be used to identify an individual. Thus, the study is not considered human subject research based upon the OHRP and the Common Rule definition. As a result, the paper was exempt from IRB review.

Western Michigan University’s IRB policies can be found here: https://wmich.edu/policies/human-subjects-research-protection. Moreover, on 4/13/2021, I received email confirmation from Julia Mays, the Associate Direction of Research Compliance at Western Michigan University, that the present study was not considered human subject research and did not require IRB approval.

Competing interests

The author declares no competing interests.