Community-Level Social Determinants of Health and Well-Child Visits Among Alabama Medicaid Enrollees

Abstract

Well-child visits focus on health promotion and disease detection and are critical to the appropriate provision of care. Evidence has shown that participation in well-child visits is associated with various patient-level factors; however, there has been an increasing focus on the influence of community-level social determinants of health (SDoH). This study explored associations between well-child visits and community-level SDoH at the census tract level among children enrolled in Alabama Medicaid. Through this analysis, it is possible to understand the distribution of care among this underserved population in different geographic settings, thus identifying potential disparities and areas for targeted intervention. Using administrative data from 2015 to 2017 enrollees in Alabama Medicaid that have been geographically linked to information on urbanicity and poverty, logistic regressions (both in total and stratified by age group) were estimated with separate community-level urbanicity, poverty variables, and individual characteristics. The regressions were repeated using a combined urbanicity/poverty variable. Looking at urbanicity and poverty together, with the exception of the least urban areas, it was those living in census tracts where there was discordance in urbanicity and poverty that had the highest likelihood of receiving well-child visits compared with those in census tracts classified as medium poverty (all urbanicity levels). There is a positive effect for Medicaid enrollees in the middle tertile of urbanicity in areas of low and high poverty and in wealthier more urban areas. If poverty and urbanicity were explored separately, some of the nuances would not have been apparent.

Introduction

Well-child visits represent critical interactions between children, caregivers, and primary care providers. These visits focus on health promotion and disease detection and are critical to the appropriate provision of vaccinations and screening related to growth and development, hearing and vision, safety and injury prevention, and social and emotional well-being. Although more frequent checkups are advised for infants and young toddlers, the American Academy of Pediatrics' (AAP) guidelines for well-child care recommend yearly visits for children beginning at age 3 years.¹ Furthermore, the Patient Protection and Affordable Care Act codified a panel of preventive services that are all available for children without a copayment.² These services are most often delivered during well-child visits.

Evidence has consistently shown that disparities in participation in well-child visits are associated with various patient-level factors including age, gender, and race³; however, there has been an increasing focus on neighborhood context and community-level social determinants of health (SDoH) and the possible influence they may have on access to and uptake of all types of care.⁴ Particularly, by considering the community context in which a child and his or her caregivers exist, we can gain a more robust understanding of challenges and supports for healthy growth and development, including the receipt of well-child visits and other forms of preventive care, particularly for low-income children.

It is well established that community poverty is associated with disparities in negative child health outcomes ranging from chronic illness to increased likelihood of injury and death,⁵ and family poverty is associated with reduced adherence to well-child care recommendations.⁶ It has also been shown that rural children are less likely to have regular well-child visits across all age groups.⁷ These studies have typically looked at caregiver-reported sociodemographic data or other individual-level information. Studies have also explored either a specific social determinant such as the level of urbanicity of an individual's neighborhood or poverty (measured at either the individual level or based upon where an individual lives). These community-level SDoH are measured alone instead of in concert. We have been unable to find any work that looks at the intersection of rurality/urbanicity and poverty and its relationship with preventive care.

This study aimed to explore associations between annual well-child visits and community-level SDoH, namely urbanicity and poverty, at the hyperlocal census tract level among children of ages 3–19 years enrolled in Alabama Medicaid. We looked at the patterns of well-child visits throughout childhood and adolescence in Medicaid enrollees through the lens of where these enrollees live. We explored the relationships between urban and socioeconomically marginalized enrollees' neighborhoods, first as separate variables but then through a combined measure to see, for example, if Medicaid enrollees who live in relatively wealthy urban areas have different service patterns from those who live in equally wealthy rural areas.

Methods

Sample and study design

The study used administrative claims data from 2015 to 2017 from Alabama Medicaid that was linked to data from the 2010 American Community Survey. All children enrolled at any point in 2015–2017 who were between the ages of 3 and 19 years with the latitude and longitude of their home address listed in the enrollment data set were included in the full sample (n = 1,713,805 person years). This study was approved by the institutional review board at the University of Alabama at Birmingham and by Alabama Medicaid.

To link community-level social determinant data to the Medicaid claim data, the latitude and longitude of each enrollee's home address (included with the enrollment data) were geocoded using ArcGIS 10.6. Once geocoded, each enrollee was matched with the census tract corresponding to their location data. The social determinant data were then assigned to each enrollee based on their home census tract.

Outcome variable

Based on the AAP's guidelines, which recommend 1 well-child visit per year for children between the ages of 3 and 18 years, the outcome variable of this analysis was a binary indicator for whether an enrollee had a well-child visit in a calendar year. Medicaid medical claims data were used to determine whether each enrollee received a well-child visit in a given calendar year. The enrollment data set calculated children's ages at the end of a given calendar year; thus, for this analysis, the age cap was extended to 19 years to capture all enrollees who were 18 years at any point in the calendar years of study.

Variables of interest

For the community-level social determinants, data were collected for 18 theoretically relevant variables for all census tracts in Alabama. These variables were constructed using data from the 2010 American Community Survey administered by the United States Census Bureau. An exploratory factor analysis was then completed to identify subsets of variables that comprise meaningful constructs that differentiate communities and that may be associated with health services utilization as well as to reduce complexity of the regression models.

Factor analysis collapses multiple related variables into simpler summary scores or indices and allows for variables that may be collinear to be accounted for in analyses while also preserving statistical power. For this project, to account for the fact that variables may not be entirely independent of each other, the exploratory factor analysis was performed using a PROMAX rotation. An a priori factor loading of 0.3 was used to determine whether a covariate loaded on an extracted factor. Variables with higher loading have greater impacts on the resulting factor. Our analysis resulted in 2 factors that capture (1) urbanicity and (2) poverty or neighborhood depravation. Variables that loaded on each factor are listed in Appendix Table A1. Factor scores were saved for each census tract. These were each broken down into tertiles (low, medium, and high) based on their scores. Owing to availability of data, there was a slight variation (<1.0%) in the number of missing observations between the urbanicity and poverty variables leading to a small difference in the reported sample size—1,711,131 for the poverty variable and 1,711,157 for the urbanicity variable. This led to a sample size of 1,711,131 for the combined variable.

Because we are interested in the intersection of urbanicity and poverty as community-level social determinants, a combined variable with 9 levels was created (low urbanicity/low poverty, low urbanicity/medium poverty, low urbanicity/high poverty, medium urbanicity/low poverty, medium urbanicity/medium poverty, medium urbanicity/high poverty, high urbanicity/low poverty, high urbanicity/medium poverty, and high urbanicity/high poverty) to be included in additional models.

Covariates

Enrollee characteristics including gender, age group (3–5, 6–12, 13–19 years), and race/ethnicity (non-Hispanic Black, non-Hispanic White, Hispanic, and other) were included in the analysis. To account for access to primary care providers, the number of pediatricians per 1000 population (measured at county level, from the 2016 American Hospital Association Annual Survey) was also included in the analysis. To account for any temporal changes, binary indicators for calendar year (2015, 2016, and 2017) were also included. The Alabama Coordinated Health Networks (ACHNs) make up the Alabama Medicaid care coordination program at a regional level. Although their operation was from October 2019, to potentially target future interventions and outreach activities, the enrollee's ACHN region (based on home county) was also included for analysis.

Statistical analysis

We began by examining enrollee and community-level characteristics for the total sample and stratified by enrollees who had a well-child visit and enrollees who did not. We then performed a series of logistic regressions based on observations with complete data on all variables (n = 1,709,300). Logistic regressions (both in total and stratified by age group) were initially estimated with separate community-level urbanicity, poverty variables, and individual characteristics. The regressions were then repeated using the combined urbanicity and poverty variable. For this analysis, we focused the analysis relative to the medium poverty tertile and, thus, the reference category became “all levels of urbanicity combined with medium poverty.” We used robust standard errors to account for the potential of repeated observations for individual enrollees. The tables show results in the form of “marginal effects,” which show the estimated percentage point difference in the probability of the outcome being 1 associated with a given variable level relative to the reference category (eg, female relative to male) or with a 1-unit increase in continuous variables. We used the statistical software Stata v.16 for all estimations.

Results

Table 1 presents descriptive statistics for the full sample and by well-child visit status. Overall, approximately one-quarter of enrollees in the sample (24.71%) received at least 1 well-child visit in a given year. Although enrollees were relatively evenly distributed across high, medium, and low levels of urbanicity, 41.05% of enrollees lived in the highest tertile of poverty, whereas only one-quarter (24.85%) lived in areas representing the lowest tertile of poverty.

Table 1.

Demographics of Medicaid Enrollees of Ages 3–19 Years (2015–2017) by Receipt of Well-Child Visit, n (%)

	Total (n = 1,713,805)^a	Had a well-child visit (n = 423,396)	Did not have a well-child visit (n = 1,290,409)
Poverty
Low	425,903 (24.85)	105,046 (24.81)	320,857 (24.86)
Medium	581,703 (33.94)	137,848 (32.56)	443,855 (34.40)
High	703,525 (41.05)	180,104 (42.54)	523,421 (40.56)
Urbanicity
Low	550,649 (32.13)	128,785 (30.42)	421,864 (32.69)
Medium	612,435 (35.74)	155,756 (36.79)	456,679 (35.39)
High	548,073 (31.98)	138,458 (32.70)	409,615 (31.74)
Urbanicity/poverty
Low urbanicity/low poverty	128,377 (7.49)	29,731 (7.02)	98,646 (7.64)
Low urbanicity/medium poverty	274,519 (16.02)	62,917 (14.86)	211,602 (16.40)
Low urbanicity/high poverty	147,740 (8.62)	36,137 (8.54)	111,603 (8.65)
Medium urbanicity/low poverty	204,675 (11.94)	51,380 (12.14)	153,295 (11.88)
Medium urbanicity/medium poverty	208,456 (12.16)	51,082 (12.06)	157,374 (12.20)
Medium urbanicity/high poverty	199,291 (11.63)	53,293 (12.59)	145,998 (11.31)
High urbanicity/low poverty	92,851 (5.42)	23,935 (5.65)	68,916 (5.34)
High urbanicity/medium poverty	98,728 (5.76)	23,849 (5.63)	74,879 (5.80)
High urbanicity/high poverty	356,494 (20.80)	90,674 (21.42)	265,820 (20.60)
Pediatricians per 1000^b
Mean (SD)	0.21 (0.19)	0.20 (0.19)	0.21 (0.19)
Alabama coordinated health network
Jefferson/Shelby	269,575 (15.73)	61,615 (14.55)	207,960 (16.12)
Central	226,643 (13.22)	63,491 (15.00)	163,152 (12.64)
East	241,895 (14.11)	52,416 (12.38)	189,479 (14.68)
Northeast	246,937 (14.41)	73,345 (17.32)	173,592 (13.45)
Northwest	221,403 (12.92)	52,042 (12.29)	169,361 (13.12)
Southeast	227,275 (13.26)	54,575 (12.89)	172,700 (13.38)
Southwest	279,345 (16.30)	65,824 (15.55)	213,521 (16.55)
Gender
Male	865,308 (50.49)	215,423 (50.88)	649,885 (50.36)
Female	847,394 (49.45)	207,740 (49.07)	639,654 (49.57)
Age
3–5 years	338,031 (19.72)	121,253 (28.64)	216,778 (16.80)
6–12 years	753,606 (43.97)	185,073 (43.71)	568,533 (44.06)
13–19 years	622,168 (36.30)	117,070 (27.65)	505,098 (39.14)
Race/ethnicity
Non-Hispanic Black	727,639 (42.46)	185,410 (43.79)	542,229 (42.02)
Non-Hispanic White	702,950 (41.02)	166,141 (39.24)	536,809 (41.60)
Hispanic	144,972 (8.46)	43,653 (10.31)	101,319 (7.85)
Other	138,244 (8.07)	28,192 (6.66)	110,052 (8.53)
Year
2015	579,976 (33.84)	169,521 (40.04)	410,455 (31.81)
2016	560,529 (32.71)	128,232 (30.29)	432,297 (33.50)
2017	573,300 (33.45)	125,643 (29.68)	447,657 (34.69)

Source: Alabama Medicaid Enrollment and Medical Claim Files, 2015–2017; American Community Survey, 2010; American Hospital Association Annual Survey, 2016.

Values under each variable may not sum to the full sample size due to missing or incomplete data.

Measured at county level.

Across the levels of the combined urbanicity and poverty variable, the largest percentage of enrollees among those who received well-child visits were those who did not live in areas with high urbanicity and high poverty (well-child visit: 21.42%; no well-child visit: 20.60%). The smallest percentage of enrollees who received well-child visits lived in areas categorized as high urbanicity and medium poverty (5.63%), and the smallest percentage of those who did not receive a well-child visit lived in areas categorized as high urbanicity and low poverty (5.34%). The older age groups have smaller proportions of enrollees receiving well-child visits than the youngest age group. Across the study years, the number of enrollees was inconsistent; however, the proportion of enrollees receiving well-child visits decreased each year from 2015 to 2017.

Table 2 presents results for the first series of logistic regressions in which the urbanicity and poverty variables were considered separately. Relative to medium poverty, the marginal effects for the low and high tertiles were significantly higher in the total population as well as in each age group. Effect sizes were all <1.00 percentage point with 1 exception. In the 13–19 years age group, relative to the medium tertile, those enrollees who lived in areas of high poverty were 1.02 percentage points more likely to have had a well-child visit (95% CI 0.75–1.29). Relative to medium urbanicity, enrollees living in the low and high tertiles were significantly less likely to have received well-child visits.

Table 2.

Marginal Effects of Community-Level Social Determinants of Health on Well-Child Visits for Medicaid Enrollees of Ages 3–19 Years (2015–2017), Urbanicity and Poverty Considered Separately

	All ages		3–5 years		6–12 years		13–19 years
	Marginal effect	95% CI	Marginal effect	95% CI	Marginal effect	95% CI	Marginal effect	95% CI
Poverty
Low	0.63	0.44 to 0.82	0.93	0.05 to 1.38	0.68	0.39 to 0.01	0.50	0.21 to 0.01
Medium	Ref.		Ref.		Ref.		Ref.
High	0.84	6.54 to 1.02	0.80	3.65 to 1.23	0.51	2.36 to 0.78	1.02	0.75 to 1.29
Urbanicity
Low	−1.68	−1.87 to −1.49	−1.50	−1.94 to −1.05	−1.38	−1.65 to −1.10	−1.43	−1.71 to −1.15
Medium	Ref.		Ref.		Ref.		Ref.
High	−1.08	−1.28 to −0.89	−1.37	−1.83 to −0.91	−1.31	−1.60 to −1.02	−1.19	−1.49 to −0.89
Pediatricians per 1000^a
Alabama Coordinated Health Network	1.84	1.10 to 2.58	−1.73	−3.47 to 0.00	3.28	2.18 to 4.38	0.83	−0.28 to 1.94
Jefferson/Shelby	Ref.		Ref.		Ref.		Ref.
Central	6.24	5.85 to 6.64	9.62	8.69 to 10.54	5.66	5.08 to 6.24	4.73	4.14 to 5.32
East	0.32	−0.06 to 0.70	0.82	−0.09 to 1.72	0.30	−0.26 to 0.87	−0.09	−0.66 to 0.54
Northeast	7.99	7.62 to 8.37	9.25	8.38 to 10.12	8.83	8.27 to 9.39	6.07	5.50 to 6.65
Northwest	2.47	2.08 to 2.86	4.49	3.57 to 5.41	2.64	2.06 to 3.22	0.69	0.11 to 1.27
Southeast	2.47	2.08 to 2.86	3.05	2.13 to 3.97	2.20	1.62 to 2.78	1.95	1.37 to 2.54
Southwest	1.69	1.40 to 1.99	4.36	3.66 to 5.07	0.98	5.33 to 1.42	1.02	0.57 to 1.47
Gender
Male	Ref.		Ref.		Ref.		Ref.
Female	−0.41	−0.56 to −0.27	0.65	0.31 to 0.99	−0.10	−0.31 to 0.12	−1.01	−1.22 to −0.79
Race/ethnicity
Non-Hispanic White	Ref.		Ref.		Ref.		Ref.
Non-Hispanic Black	1.61	1.42 to 1.79	4.27	3.82 to 4.71	0.30	0.24 to 0.58	2.61	2.34 to 2.88
Hispanic	5.76	5.45 to 6.06	5.55	4.90 to 6.20	4.17	3.76 to 4.59	5.47	4.96 to 5.99
Other	−3.21	−3.48 to −2.94	−4.73	−5.35 to −4.11	−3.98	−4.38 to −3.58	−2.27	−2.67 to −1.86
Year
2015	Ref.		Ref.		Ref.		Ref.
2016	−6.36	−6.51 to −6.21	−8.86	−9.24 to −8.48	−6.52	−6.75 to −6.30	−4.68	−4.90 to −4.46
2017	−7.28	−7.43 to −7.13	−11.50	−11.89 to −11.12	−7.16	−7.38 to −6.93	−5.02	−5.23 to −4.79

Source: Alabama Medicaid Enrollment and Medical Claim Files, 2015–2017; American Community Survey 2010; American Hospital Association Annual Survey, 2016.

n = 1,709,300, all estimates are significant if not italicized.

Measured at county level.

Relative to the Jefferson/Shelby ACHN region, enrollees in all other regions, with the exception of the east region, showed a significantly larger likelihood of receiving well-child visits. For the central and northeast regions, in the youngest age group (3–5 years), the effect sizes relative to Jefferson/Shelby approached 10 percentage points (central: 9.62, 95% CI: 8.69–10.54; northeast: 9.25, 95% CI: 8.38–10.12). Relative to 2015, 2016 and 2017 both showed large drops in well-child visits in the total sample as well as across the age groups. The number of pediatricians per 1000 population at the county level was not significantly related to receipt of well-child visits, with the exception of the 6–12 years age group (marginal effect: 3.28, 95% CI: 2.18–4.38).

Relative to male enrollees, in the total sample as well as the youngest and oldest age groups, female enrollees showed significantly different likelihoods of receiving well-child visits; however, the direction of the differences in effect sizes was not consistent. Overall and in the oldest age group, female enrollees had a significantly lower likelihood, but within the 3–5 years age group, the likelihood was significantly higher than the likelihood of male enrollees. There was not a significant difference in the middle age group. Relative to non-Hispanic White enrollees, non-Hispanic Black and Hispanic enrollees were more likely to have received well-child visits. This was most pronounced in the 3–5 years age group for non-Hispanic Black enrollees (marginal effect: 4.27, 95% CI: 3.82–4.71). Relative to non-Hispanic White enrollees, Hispanic enrollees had the largest positive marginal effects, particularly in the 3–5 years age group (marginal effect: 5.55, 95% CI: 4.90–6.20) and the 13–19 years age group (marginal effect: 5.47, 95% CI: 4.96–5.99).

Table 3 presents the marginal effects for the second set of models that examined the interplay between community-level poverty and urbanicity. Relative to the comparison group (areas with medium poverty and all 3 levels of urbanicity), low urbanicity/low poverty was associated with a statistically significant reduction in well-child visits (−0.76, 95% CI: −1.20 to −0.32), whereas high urbanicity/high poverty was associated with an increase (0.46, 96% CI: 0.11–0.80) in the probability of a well-child visit.

Table 3.

Marginal Effects of Community-Level Social Determinants of Health on Well-Child Visits for Medicaid Enrollees of Ages 3–19 Years (2015–2017), Combined Urbanicity and Poverty

	All ages		3–5 years		6–12 years		13–19 years
	Marginal effect	95% CI	Marginal effect	95% CI	Marginal effect	95% CI	Marginal effect	95% CI
Urbanicity/poverty
All urbanicity/medium poverty^a	Ref.		Ref.		Ref.		Ref.
Low urbanicity/low poverty	−0.49	−0.78 to −0.19	−0.07	−0.78 to 0.64	−0.14	−0.58 to 0.30	−0.76	−1.20 to −0.32
Low urbanicity/high poverty	0.04	−0.25 to 0.33	0.49	−0.20 to 1.18	−0.53	−0.97 to −0.10	0.74	0.03 to 1.15
Medium urbanicity/low poverty	1.32	1.08 to 1.57	1.73	1.15 to 2.31	1.17	0.81 to 1.52	1.17	0.80 to 1.54
Medium urbanicity/high poverty	2.20	1.95 to 2.44	1.98	1.40 to 2.57	2.01	1.64 to 2.38	2.13	1.76 to 2.50
High urbanicity/low poverty	1.59	1.25 to 1.93	1.18	0.39 to 1.98	1.47	0.97 to 1.96	1.58	1.07 to 2.08
High urbanicity/high poverty	0.48	0.26 to 0.71	0.07	−0.47 to 0.60	−0.06	−0.40 to 0.28	0.46	0.11 to 0.80
Pediatricians per 1000 population^b	2.13	1.40 to 2.87	−1.31	−3.02 to 0.41	2.92	1.83 to 4.01	1.27	0.16 to 2.37
Alabama Coordinated Health Network
Jefferson/Shelby	Ref.		Ref.		Ref.		Ref.
Central	6.28	5.89 to 6.67	9.66	8.74 to 10.58	5.55	4.96 to 6.13	4.83	4.24 to 5.42
East	0.46	0.07 to 0.84	1.01	0.09 to 1.93	0.26	−0.31 to 0.83	0.19	−0.39 to 0.77
Northeast	8.16	7.79 to 8.53	9.45	8.58 to 10.32	8.79	8.23 to 9.35	6.33	5.76 to 6.90
Northwest	2.48	2.09 to 2.87	4.51	3.58 to 5.43	2.52	1.94 to 3.10	0.80	0.22 to 1.38
Southeast	2.53	2.14 to 2.92	3.15	2.22 to 4.08	2.06	1.48 to 2.64	2.12	1.54 to 2.71
Southwest	1.77	1.47 to 2.06	4.43	3.72 to 5.14	0.96	5.14 to 1.40	1.12	0.67 to 1.57
Gender
Male	Ref.		Ref.		Ref.		Ref.
Female	0.48	−0.56 to −0.27	0.65	0.31 to 0.99	−0.10	−0.32 to 0.12	−1.00	−1.22 to −0.79
Race/ethnicity
Non-Hispanic White	Ref.		Ref.		Ref.		Ref.
Non-Hispanic Black	1.65	1.47 to 1.84	4.28	3.84 to 4.72	0.31	0.03 to 0.58	2.59	2.32 to 2.86
Hispanic	5.74	5.44 to 6.04	5.55	4.90 to 6.19	4.09	3.67 to 4.50	5.42	4.91 to 5.94
Other	−3.22	−3.49 to −2.96	−4.75	−5.36 to −4.13	−4.03	−4.43 to −3.63	−2.33	−2.73 to −1.92
Year
2015	Ref.		Ref.		Ref.		Ref.
2016	−6.36	−6.51 to −6.21	−8.86	−9.24 to −8.48	−6.52	−6.75 to −6.30	−4.68	−4.90 to −4.46
2017	−7.28	−7.43 to −7.12	−11.50	−11.88 to −11.11	−7.15	−7.38 to −6.92	−5.02	−5.25 to −4.79

Source: Alabama Medicaid Enrollment and Medical Claim Files, 2015–2017; American Community Survey 2010; American Hospital Association Annual Survey, 2016.

n = 1,709,300, all estimates are significant if not italicized.

Includes low urbanicity/medium poverty, medium urbanicity/medium poverty, and high urbanicity/medium poverty.

Measured at county level.

Compared with areas with medium poverty, areas with low poverty and high urbanicity had significant marginal effects at the middle and oldest age groups, although the directions differed. In these low/high areas, enrollees in the 6–12 years age group had a decreased likelihood of receiving a well-child visit (−0.53, 95% CI: 0.97 to −0.10), and those in the 13–19 years age group saw an increased likelihood (0.74, 95% CI: 0.03–1.15) relative to those in medium poverty areas.

Areas with medium urbanicity/low poverty, medium urbanicity/high poverty, and high urbanicity/low poverty were all associated with significantly higher rates of well-child visits overall and in each age group.

The number of pediatricians per 1000 population showed significant positive effect sizes in all of these models with the exception of the enrollees in the 3–5 years age group (All: marginal effect: 2.13, 95% CI: 1.40–2.87; 6–12 years: marginal effect: 2.92, 95% CI: 1.83–4.01; 13–19 years: marginal effect: 1.27, 95% CI: 1.63–2.37).

Discussion and Conclusions

This study explored the relationship between community-level SDoH and well-child visits for children of ages 3–19 years enrolled in Medicaid. Although a wide array of community-level factors may impact health care utilization, this analysis focused on census tract level urbanicity and poverty.

When viewing poverty and urbanicity variables separately, some significant disparities were apparent. Relative to those who lived in areas in the middle tertile of census tracts based on poverty, enrollees who lived in both low-poverty areas and high-poverty areas were more likely to receive well-child visits. Although this finding was statistically significant, it may not be clinically meaningful due to small effect sizes. Considering urbanicity, compared with those in the middle third of census tracts, those living in the least and most urban census tracts were significantly less likely to receive a well-child visit.

When looking at urbanicity and poverty together, more granular disparities emerged. We found that, with the exception of the least urban areas, it was those living in census tracts where there was discordance in the levels of urbanicity and poverty (eg, high urbanicity and low poverty or medium urbanicity and high poverty) that had the highest likelihood of receiving well-child visits compared with those living in census tracts classified as medium poverty (and all levels of urbanicity). In the least urban census tracts, there were statistically significant small effect sizes in the negative direction.

This indicates that those living in the least urban areas, regardless of the neighborhood poverty level, are as likely or slightly less likely to have a well-child visit compared with those in any medium poverty area. Those in areas of high urbanicity and high poverty had little if any difference relative to those in medium poverty areas. There is, however, a positive effect for Medicaid enrollees living in the middle tertile of urbanicity in areas of low and high poverty and in wealthier more urban areas.

Past studies have looked at neighborhood demographic and risk-factor data in relation to asthma care^8,9 and emergency or urgent care visits,¹⁰ among other conditions and outcomes. These indices and other community-level measures of social determinants have almost unilaterally focused on poverty as the primary measure of risk. Others have looked at rural–urban differences with respect to similar acute, chronic, and preventive care outcomes.^11
–13 If these studies considered both rural–urban geography and poverty, no interactions of the 2 were considered.

If we had only looked at poverty and urbanicity separately, some of the nuances of these patterns and differences would not have been apparent. Areas of low and high poverty would have been seen as having a uniform positive association with well-child visits, whereas areas of low and high urbanicity would have been perceived as having a consistent negative association. Our analysis suggests important interactions between these community-level SDoH. Enrollees who live in areas that are the least urban with the lowest poverty still saw a negative effect on the likelihood of receiving a well-child visit overall and in the oldest age group. Enrollees in areas of high urbanicity and low poverty saw an even larger effect size in all age groups than would have been captured if we had considered them in the low-poverty group alone. In short, there is value in considering community-level social determinants in combination with each other, because keeping them separate may serve to mask important variation and patterns.

Well-child visits are critical components of preventive care for children across developmental periods; thus, it is important to understand patterns of well-child visits across populations so that health promotion efforts can be targeted geographically based on neighborhood and community factors. For preschool aged children (∼3–5 years), well-child visits place heightened emphasis on developmental milestones in the physical, language, cognitive, and social emotional domains. During the middle childhood period (∼6–12 years), well-child visits continue providing essential preventive care including immunizations, healthy nutrition promotion and weight management, and emotional and behavioral adjustment to school. For both of these age groups, there is also a focus on fostering relationships between providers and caregivers and sharing of anticipatory guidance by providers to promote healthy child and family development.¹ For adolescents (13–19 years), well-child visits become opportunities to build responsibility for their own health care and prepare for transition to adult health care. Substance use screening typically begins in these visits. At these visits, sexually active adolescents should be screened for sexually transmitted infections. At this age, anticipatory guidance begins to shift away from the caregiver(s) in favor of building skills and agency in the child.¹ Providers emphasize that well-child visits are important throughout childhood and adolescence though, as children age, there is often a decrease in their likelihood of having visits for anything outside of sick care.¹⁴

Furthermore, results show that no region of the state achieved >30% uptake of well visits among children, highlighting the critical need for promotion of pediatric preventive care and improvement of access to care across the state. This low rate of well-child visits could be due, in part, to low densities of pediatric providers in many areas of the state, which may be mitigated by incentivizing providers to establish practices in health professional shortage areas. Promoting infrastructure and coverage for high-quality telehealth care may also reduce disparities in access to pediatric care. In addition, other states have increased well-child visit rates and quality by providing additional case management support.¹⁵ In a state where need is high in all areas, the more granular findings from this analysis may help prioritize areas to target for the establishment of new or expanded practices and enhanced case management.

In conjunction with the long-standing evidence that has shown associations between individual patient characteristics and adherence to recommended care schedules, a growing literature examines relationships between neighborhood-level factors, SDoH, and health care utilization.^3,4 In fact, the evidence has been so robust that it has shifted policy and focus at the federal and state levels.^16
–18 However, most studies have thus far looked at a single dimension of a neighborhood, such as rurality or poverty.^7,14,19 This study has sought to add a richer understanding by exploring dimensions in concert with one another to create a more subtle understanding of these determinants and their relationship with disparities in care and to open up a more finely tuned method of identifying geographic areas for potential intervention.^20,21

As with any study, there are limitations to this work. Most notably, this analysis shows association, not causality, as multiple underlying and potentially confounding social and political factors may influence where people live. Furthermore, we do not know how many enrollees in each group were eligible for Medicaid due to income and how many were eligible due to a qualifying disability. Noting that smaller number of enrollees were assigned to areas of low and middle poverty hints that there may be differences in the patterns of where enrollees with disabilities and other special health care needs live relative to enrollees who qualify for Medicaid due to income. When looking at the pediatricians per 1000 population at the county level, this is a measure of all pediatricians, some of whom may not accept Medicaid. Therefore, these estimated effect sizes may be inflated. There is also a sizeable drop across years in the percentage of enrollees receiving well-child visits. We were not immediately aware of any policy or programmatic changes that would lead to reduced well-child visits from 2015 to 2017; thus, this warrants more investigation.

As the literature expands on SDoH and their role in health disparities, it will be important to understand how these determinants interact with each other and the nuanced impact that these interactions bring to the health of communities. This study shows that neither urbanicity nor poverty should be viewed as a single uniformly representative indicator of neighborhood health. These 2 measures, when viewed together, may hint at differences in the ability and willingness to access care in different diverse communities. Future research should apply more causality-focused methods to better understand the implications and interactions of housing, land development, and social safety-net policies on access and uptake of preventive health care. Qualitative studies may also explore the nuanced experiences that low-income and/or Medicaid-enrolled children face when seeking care in areas that are socioeconomically concordant and discordant to their individual families and in areas that are more and less rural.

Footnotes

Authors' Contributions

All authors helped to conceptualize ideas and reviewed drafts of the article. Dr. Brisendine drafted the initial article, Mr. Sharma conducted the analyses, and all authors approved of the final version.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

Funding for this study is part of a research contract with the Alabama Medicaid agency.

Appendix

Appendix Table A1.

Variables Loaded in Each Factor

Factor 1: Poverty

Percentage of people who are not non-Hispanic White

Percentage of single parent households with children <18 years

Percentage of people with cash public assistance or Food Stamps/SNAP

Percentage of persons below Federal Poverty Line

Percentage of people older than 16 years who are unemployed

Percentage of people older than 25 years who have no high school diploma

Percentage of people in the total population who are uninsured

Percentage of households with no vehicle available

Factor 2: Urbanicity

Population density

Percentage of people older than 25 years who have no high school diploma^a

Percentage of tract population who are beyond 1 mile from supermarket^a

Percentage of population living within a half mile of a park, 2015 data

Average percentage of developed imperviousness

Flag of urban tract, 2010 data

These factor loadings were negative, meaning an inverse relationship. Higher absolute values result in a reduction of the factor score.

1 mile = 1609.3 meters.

SNAP, Supplementary Nutritional Assistance Program.

References

Green

, Palfrey

Bright futures: guidelines for health supervision of infants, children and adolescents. Arlington VA: National Center for Education in Maternal and Child Health, 2000.

Fox

, Shaw

. Clinical preventive services coverage and the Affordable Care Act. Am J Public Health, 2015; 105:e7–e10.

Selden

TM.

Compliance with well-child visit recommendations: evidence from the Medical Expenditure Panel Survey, 2000–2002. Pediatrics, 2006; 118:e1766–78.

Marmot

, Friel

, Bell

, Houweling

, Taylor

, Health

CoSDo

. Closing the gap in a generation: health equity through action on the social determinants of health. The lancet, 2008; 372:1661–1669.

Council on Community Pediatrics. Poverty and child health in the United States. Pediatrics, 2016; 137:e20160339.

Abdus

, Selden

. Adherence with recommended well-child visits has grown, but large gaps persist among various socioeconomic groups. Health Aff, 2013; 32:508–515.

Hardy

, Liu

, Kelleher

. Contribution of social determinant of health factors to rural-urban preventive care differences among Medicaid enrollees. Acad Pediatr, 2021; 21:93–100.

Beck

, Simmons

, Huang

, Kahn

. Geomedicine: area-based socioeconomic measures for assessing risk of hospital reutilization among children admitted for asthma. Am J Public Health, 2012; 102:2308–2314.

Gjelsvik

, Rogers

, Garro

, et al. Neighborhood risk and hospital use for pediatric asthma, Rhode Island, 2005–2014. Prev Chronic Dis, 2019; 16:E68.

10.

Kersten

, Adler

, Gottlieb

, et al. Neighborhood child opportunity and individual-level pediatric acute care use and diagnoses. Pediatrics, 2018; 141:e20172309.

11.

McDaniel

, McDermott

, Martinasek

, White

. Prevalence of childhood asthma in US military and non-military families: comparisons by rural–urban residence and geographic region. Chronic Illn, 2018; 16:296–306.

12.

Yue

, Upadhyayula

, Avalos

, Cage

. Pediatric traumatic brain injury in the United States: rural-urban disparities and considerations. Brain Sci, 2020; 10:135.

13.

Unger

, Moseley

, Symington

, Chavez-MacGregor

, Ramsey

, Hershman

. Geographic distribution and survival outcomes for rural patients with cancer treated in clinical trials. JAMA Network Open, 2018; 1:e181235–e181235.

14.

Nordin

, Solberg

, Parker

. Adolescent primary care visit patterns. Ann Fam Med, 2010; 8:511–516.

15.

Smith

, Merchen

, Turner

, Vaught

, Fritz

, Mold

. Improving the rate and quality of Medicaid well child care exams in primary care practices. J Okla State Med Assoc, 2010; 103:248–53.

16.

Garg

, Homer

, Dworkin

. Addressing social determinants of health: challenges and opportunities in a value-based model. Pediatrics, 2019; 143:e20182355.

17.

Garg

, Toy

, Tripodis

, Silverstein

, Freeman

. Addressing social determinants of health at well child care visits: a cluster RCT. Pediatrics, 2015; 135:e296–e304.

18.

Spencer

, Freda

, McGinnis

, Gottlieb

Measuring social determinants of health among Medicaid beneficiaries: early state lessons. Hamilton, NJ: Center for Health Care Strategies, 2016.

19.

Jones

, Brown

, Widener

, Sucharew

, Beck

. Area-level socioeconomic factors are associated with noncompletion of pediatric preventive services. J Prim Care Community Health, 2016; 7:143–148.

20.

Gottlieb

, Francis

, Beck

. Uses and misuses of patient-and neighborhood-level social determinants of health data. Perm J, 2018; 22:18-078.

21.

Gottlieb

, Tirozzi

, Manchanda

, Burns

, Sandel

. Moving electronic medical records upstream: Incorporating social determinants of health. Am J Prev Med, 2015; 48:215–218.