Analysis of Pregnancy and Birth Rates among Black and White Medicaid-Enrolled Teens

Abstract

Purpose

In the U.S. black teens overall have higher pregnancy and birth rates than whites, and it is commonly believed that minority race and low-income account for this disparity. We examined racial differences in pregnancy and birth rates among teens from low-income households, using Medicaid-enrollment as a proxy for low income.

Methods

This was a retrospective study of Louisiana Medicaid claims data for females ages 15–17 years in 2014 (n = 66,069). Pregnancy and pregnancy outcome codes were identified (n = 2,276) and analyzed for differences by black and white race. We conducted validity analyses with different rate definitions and teens’ claims status.

Results

The cohort was 36% white and 54% black. More black teens than whites lacked any claims data (15.6% vs 12.6%; P < .001). Rates calculated as events per 1,000 person-years of Medicaid coverage showed no difference in live birth rates between white and black teens (24.6 vs 25.8; relative incidence ratio [RIR] 1.05; 95% confidence interval [CI] 0.93—1.18; P = .43), however, pregnancy rates for whites were higher than for blacks (42.7 vs 36.1; RIR 0.85; 95% CI 0.77—0.93; P < .001).

Conclusion

In contrast to national trends, which include teens from diverse racial and socioeconomic backgrounds, Louisiana Medicaid-enrolled teens ages 15–17 years had equal birth rates regardless of black or white race, and whites had higher pregnancy rates. Decreased racial disparities in pregnancy and birth rates among these adolescents highlights socioeconomic influences in sexual health behavior, and a need to examine the interplay of risk factors contributing to racial disparities seen among adolescents nationally.

Introduction

U.S. teen birth rates are declining, but disparities in teen births persist between different groups and seem to be related to factors such as race, geographic location, and socioeconomic status. For example, from 2007 to 2015 birth rates among 15–19-year-olds fell by 50% in large urban counties but only 37% in rural counties [1]. This disparity was reflected in Louisiana, a rural state with a relatively high percentage of racial minority youth, which had one of the smallest declines in teen birth rates since 2007 [1]. The majority of all teen births occur among 18 and 19-year-olds [2], however, so teen birth rates among 15–17-year-olds are much lower across the country. For example, in 2016 the birth rates among teens ages 15–17years ranged from 3.8 per 1,000 in Connecticut to 15.1 per 1,000 in New Mexico and Texas. Even among this lower age group, however, Louisiana ranked 8^th highest in the nation with 12.7 births per 1,000 teens ages 15–17 years [3]. In addition, although pregnancy rates are always higher than birth rates, since not all pregnancies result in birth, reported pregnancy rates mirror disparities seen in birth rates. For example, in 2013 among white Louisianans, the estimated pregnancy rate was 42 per 1,000 girls aged 15–19 and the birth rate was 31, compared to the estimated pregnancy and birth rates among same-aged black girls, which were 73 and 50, respectively [2].

Several approaches have been used to examine the effect of race on teen birth risk, resulting in evidence to explain some of the racial disparities. For example, analyses of data from the National Longitudinal Study of Adolescent Health, as well as the Youth Risk Behavior Surveillance System, show black teens are more likely than whites to have ever had sexual intercourse [4], [5]. Moreover, factors such as family structure and household income [6], neighborhood disadvantage [7], and parental monitoring [8] contribute to black-white disparities in sexual behaviors, and subsequent pregnancy risk. Given Medicaid is the primary source of payment for over 75% of deliveries to mothers under age 20, an analysis of Medicaid-enrolled teens may essentially “control” for the influence of at least one social factor—household income—in examining racial disparities in teen births [3].

There is currently a gap in the literature regarding racial differences in teen pregnancy and birth rates when controlling for specific socioeconomic factors. Therefore, we aimed to examine differences in white and black teen pregnancy and birth rates among a group of 15–17-year-old teens enrolled in Louisiana Medicaid [1]. This group of teens may have similar exposures to pregnancy risk factors due to similarities in location and household income, assuming Medicaid is a proxy for low income. It bears noting that teen pregnancy rates are typically extrapolated from various data sources that estimate not only live births, but also ectopic pregnancies, spontaneous abortions, elective abortions, and fetal demise [2]. Therefore, this study also provided a novel measure of pregnancy rates within a subgroup of teens. Lastly, we demonstrated the feasibility of using state-level Medicaid claims data to calculate these pregnancy rates and outcomes among teens enrolled in that insurance program.

Methods

Study Design and Data Selection

This was a retrospective study of pregnancy and birth-related diagnosis and procedure billing codes from the Louisiana Medicaid database. Our cohort consisted of females who were 15–19 years old at any time in 2014 with at least 2 years of previous Medicaid enrollment. We chose 2014 to avail several years of subsequent billing code data on study teens for future examination, and to allow future comparisons between the 2014 cohort and cohorts representing that same age group in subsequent years. In Louisiana in 2014, children ages 6–18 years with household incomes up to 217% of the federal poverty level (or $76,873 annual income for a 4-person household) were eligible for Medicaid (inclusive of Children’s Health Insurance Program recipients) [9], and approximately 45% of all children in the state who were under 18 were enrolled [10]. Louisiana did not expand Medicaid for adults until 2016. Institutional Review Board approvals were obtained from Louisiana State University Health Sciences Center New Orleans, Pennington Biomedical Research Center, and the Louisiana Department of Health. A Data Use Agreement was completed between these institutions prior to transfer of masked data.

Data obtained included: randomly generated identification numbers for each teen; single random date offsets (integer between −30 and +30 days, excluding 0) for teens’ birthdates and event dates (the same integer applied to every date relating to an individual); corresponding claims data spanning 2012–2016; and teens’ race/ethnicity. Claims data included: International Classification of Diseases, 9th Revision (ICD-9), International Classification of Diseases, 10th Revision (ICD-10), and Current Procedural Terminology (CPT) codes. Race/ethnicity was self-assigned by teens or their guardians at the time of enrollment in Medicaid from a combined set of race and ethnicity categories that included Asian, Black, Hispanic, Native American, Alaskan Native, Hawaiian, White, multiple, and unknown. The dataset did not differentiate race and ethnicity, but because about 95% of Louisiana’s population identifies as either non-Hispanic white or non-Hispanic black/African American [11], we limited our analysis of race to white and black.

There was a disproportionately lower number of 18 and 19-year-old enrollees in this cohort compared to younger ages, consistent with declining Medicaid enrollment among adolescents who are approaching age cutoffs for children’s Medicaid eligibility [12], [13]. We were unable to account for variables affecting dates that enrollments ceased or follow their claims data for reliable study outcomes. Therefore, we applied our analyses to an age-limited sample of 15–17-year-olds from the overall cohort. This is consistent with national reports that also categorize 15–17-year-olds as a distinct group; implicit in this age restriction is that unplanned pregnancies occur more often among these younger teens than their older peers [14]. We also identified teens who did not have Medicaid claims at any time during 2012–2016 to determine if any bias was introduced from teens who may have obtained healthcare through another payer.

Algorithm for Assigning Pregnancy and Pregnancy Outcome Codes

The data contained 2,276 distinct diagnosis and procedure codes that indicated either a pregnancy or pregnancy outcome event (e.g., a childbirth or miscarriage) and are summarized in Table 1. Figure 1 further illustrates how study teens and claims data was categorized. We identified teens’ distinct pregnancy episode(s) and established a conception date and pregnancy outcome date for each. The algorithm used to identify pregnancy and pregnancy outcome events as categorized in Table 1 was as follows:

Table 1:

Diagnosis and Procedure Codes Used to Identify Pregnancies and Pregnancy Outcomes

Pregnancy Events
ICD-9a	630—679, 760—779, V22—V24.2, V27—V29
ICD-10b	O00—P96.9, Z33—Z3A.49
CPTc	01960—01961, 01967—01968, 0502F, 58611, 59000—59001
Non-Live Birth Outcomes
ICD-9a	630—639.9, 656.40—656.43, 768.0, 779.6, V27.1, V27.4—V27.7
ICD-10b	O00—O04.89, O08—O08.9, O36.4, P95, Z33.2, Z37.1, Z37.4
CPTc	59812—59856
Live Birth Outcomes
ICD-9a	649.01—649.02, 649.11, 649.21—649.22, 649.31—649.32, 649.41, 649.51, 649.61, 649.71, 649.81—650.00, 651.01, 651.11, 651.21, 651.31, 651.91, 652.01, 652.10, 652.11, 652.21, 652.31, 652.41, 652.51, 652.61, 652.71, 652.81, 652.91, 653.01, 653.11, 653.21, 653.31, 653.41, 653.51, 653.61, 653.71, 653.81, 653.91, 654.01, 654.11—654.12, 654.21, 654.31, 654.41, 654.44, 654.51, 654.61, 654.71, 654.74, 654.81, 654.91—654.92, 655.01, 655.11, 655.21, 655.31, 655.41, 655.51, 655.71, 655.81, 655.91, 656.01, 656.11, 656.21, 656.31, 656.51, 656.61, 656.71, 656.81, 656.91, 657.01, 658.01, 658.11, 658.21, 658.31, 658.41, 658.81, 658.91, 664.04, 664.14, 664.24, 659.01, 659.11, 659.21, 659.31, 659.41—659.51, 659.61, 659.71, 659.81, 659.91, 660.01, 660.11, 660.21, 660.31, 660.41—660.51, 660.61, 660.71—660.81, 660.91, 661.01, 661.11, 661.21, 661.31, 661.41, 661.91, 662.01, 662.11, 662.21, 663.01, 663.11, 663.21, 663.31, 663.41, 663.61, 663.81, 663.91, 664—664.0, 664.01, 664.11, 664.21, 664.31, 664.34, 664.4—664.41, 664.44, 664.51, 664.54, 664.81, 664.84, 664.91, 665.01, 665.11, 665.31, 665.41, 665.44, 665.51, 665.71, 665.74, 665.81, 665.91, 665.44, 666—667.14, 668, 668.81—668.82, 668.84, 669.01, 669.12, 669.21, 669.32—669.41—669.42, 669.44, 669.5—669.71, 669.81—669.82, 669.91, 670—670.84, 671.01—671.02, 671.11, 671.21, 671.24, 671.31, 671.42, 671.44—671.52, 671.54, 671.81—671.82, 671.84—672.04, 673.01, 673.12, 673.22, 673.24, 673.82, 674—674.01, 674.12, 674.22, 674.14, 674.24, 674.32—674.51, 674.54—674.94, 675.11—675.12, 675.14, 675.21—675.22, 675.24, 675.82, 676—676.12, 676.21—676.22, 676.24, 676.31, 676.34—676.61, 676.64—676.91, 763.0—763.89, 765.10—767.2, 768.4—777.9, 779.31, 779.81—779.9, V24—V24.2, V27.0, V27.2—V27.3, V27.9, V29—V29.9
ICD-10b	O10.03, O10.13, O10.43, O10.93, O15.2, O24.03, O24.13, O24.430—O24.439, O24.93, O26.63, O26.73, O60.1—O60.14X9, O67, O69—O70.9, O72—O73.1, O74—O75.1, O75.5—O75.81, O75.89—O84, O85—O87.2, O88.13—O91, O91.12—O91.13, O91.22—O92.79, O98.13, O98.23, O98.33, O98.43, O98.53, O98.73, O98.83, O99.03, O99.13, O99.215, O99.285, O99.325, O99.335, O99.345, O99.355, O99.43, O99.53, O99.63, O99.73, O99.825, O9A.23, P00.2—P78.9, P92.9, P96.89—P96.9, Z37.0, Z37.2—Z37.3, Z37.9—Z38.2, Z39—Z39.2
CPTc	01960—01961, 01967—01968, 58611, 59160, 59300, 59409—59410, 59430, 99460—99462

^aInternational Classification of Diseases, 9th Revision

^bInternational Classification of Diseases, 10th Revision

^cCurrent Procedural Terminology

Figure 1:

Flowchart of study teens and claims composition.

1. “Pregnancy event” codes identified pregnancy-related healthcare encounter dates.

2. “Non-live-birth outcome” codes defined outcome dates for non-live births (i.e., any code that implies the end of a pregnancy other than a live birth).

3. “Live birth outcome” codes defined outcome dates for live births, and preferentially used CPT codes when available. Most live birth outcome dates were assigned based on a code indicating childbirth explicitly (e.g., V27.0 codes a vaginal delivery). However, our algorithm recognized other codes that only implied childbirth (e.g., V24.2 indicates routine postpartum follow-up). This latter type of code was required to assign outcome dates on only 4 occasions when a more explicit delivery code was unavailable.

4. For “live birth outcomes”, a conception date was established by subtracting days from the live birth date. That number of days was determined by specific diagnosis and procedure codes in earlier portions of the pregnancy episode (e.g., codes that specified gestational weeks) or, in the absence of such information, by subtracting 280 days from the live birth date (average length of a term pregnancy).

5. Each “live birth outcome” implicitly established an initial “time to first episode encounter” defined as time from conception to first pregnancy-related encounter in the episode. This “time to first episode encounter” varied systematically with age of the mother and was imputed using data from live birth episodes. Therefore, for each “non-live birth outcome,” a conception date was established by subtraction from the first pregnancy-related encounter date in the episode (not from the outcome date) using the number of days that was the age-specific median “time to first episode encounter.”

6. For a small percentage of pregnancy episodes, no outcomes codes were found. Therefore, conception dates were imputed using the same approach described in step 5 above.

Validation of Algorithm

Defined pregnancy episodes were validated by comparing algorithm results to conclusions from expert review. Experts consulted were physicians who routinely review obstetric and infant records. One expert reviewed several probabilistic samples of teens to identify problems and revise the algorithm. Final review involved two experts who independently reviewed the same final sample of teens and pregnancy episodes. The algorithm was deemed correct if both reviewers agreed with the algorithm regarding the (1) calendar year of conception, (2) type of outcome, and (3) year of outcome. All needed algorithm revisions identified were made before generating final results. However, the final two-reviewer agreements exercise estimated that (before final corrections) the algorithm generated errors in live birth results for 3.2% of teens. Assuming that level of error remains and is independent of teen race, there is an approximately 0.05 probability it would bias estimated racial relative risks of live births by more than one percent.

Data Analysis

Data was analyzed to determine rates of pregnancy and pregnancy outcomes in 2014 among Medicaid-enrolled teens ages 15–17 years in Louisiana. Rates were calculated as rates per person-year of Medicaid coverage in 2014. Unless otherwise noted, reported levels of pregnancy and pregnancy outcomes for Louisiana Medicaid enrollees were age-standardized to the 2014 female 15–17-year-old Louisiana Medicaid population. In particular, comparisons of black and white pregnancy and pregnancy outcome rates were made after age standardization. This age-based standardization had only a negligible effect on results, however, because the age distribution of female Medicaid enrollees was nearly uniform over the ages of 15, 16, and 17, both overall and within racial groups.

Total rates reported below were unadjusted for age or race. However, all reported race-specific rates, as well as the relative incidence ratios (RIRs) used to compare racial groups, were standardized to the combined age distribution of black plus white Medicaid enrollees. Using data weighting to maintain the standardization, most confidence intervals and p-values for racial RIRs were then estimated using negative binomial regression models. However, bootstrapping was used to derive confidence intervals and p-values for analysis in which negative binomial modeling could not be used.

Several validation analyses were conducted to check for various possibilities that might distort the estimated rates and, in particular, bias the racial rate comparisons. It is possible that some of the individuals without Medicaid claims in our available data were only nominally Medicaid enrollees and actually received healthcare through other payers. To check if this would introduce bias in racial comparisons, the pregnancy and live birth rates were calculated and compared using only those individuals who had at least one claim represented in the available data. Another possibility is that some individuals became Medicaid enrollees after they became pregnant and/or because they were pregnant. In particular, some may have become pregnant in 2013 and enrolled in 2013 as a result. To assess for possible bias in racial comparisons arising from that possibility, rates were calculated using only those individuals who had claims from before 2013. All analyses were performed use Stata 14.2 and 15.1 statistical software (Stata Corp, College Station, TX).

Results

There were 66,069 teens who were 15–17 years old at some point during 2014 with continuous Medicaid-enrollment, of whom 36% were white and 54% were black. Table 2 details our sample’s characteristics by age, race, and claims availability. Overall, we found that black teens were significantly more likely than white teens to be enrolled in Medicaid but not have any claims data, which supported our need to consider claims status when comparing pregnancy and birth rates by race. Table 3 shows pregnancy and pregnancy outcome rates for the sample. Rates calculated as events per 1,000 person-years of Medicaid coverage showed no difference in live birth rates between white and black teens (24.6 vs 25.8; relative incidence ratio [RIR] 1.05; 95% confidence interval [CI] 0.93—1.18; P = .43). In fact, there was no pregnancy outcome (live birth, non-live birth, or unknown outcome) RIR that was statistically significantly different from 1.0. In contrast, the pregnancy rate was slightly higher in white teens (42.7 per 1,000) than black teens (36.1 per 1,000) and this difference was statistically significant. Table 3 reports rates both by person-years and by the number per 1,000 who were in that age group as of July 1, 2014. We included the latter definition because it is often used in governmental reports of birth rates, but our analysis showed that it did not differ from person-year rates. Table 3 also shows rates separately for black and white enrollees; however, the “total” rates include all racial and ethnic groups in the sample. RIRs were calculated as black rates divided by white rates.

Table 2:

Ages and Claims Status of Total Cohort by Race (N = 66,069)

	All subjects		White subjects		Black Subjects
Agea (years)	Total	No Claims, n (%)	Total	No Claims, n (%)	Total	No Claims, n (%)	Pb
14	8,857	1,855 (20.9)	2,788	513(18.4)	4,834	1,100(22.8)	<.001
15	17,016	2,930 (17.2)	6,183	861 (13.9)	9,373	1,707 (18.2)	<.001
16	16,524	2,256 (13.7)	5,998	686 (11.4)	8,962	1,245 (13.9)	<.001
17	16,359	2,012 (12.3)	5,836	612 (10.5)	8,728	1,067 (12.2)	0.007
18	7,313	914 (12.5)	2,606	287 (11.0)	3,924	452 (11.5)	0.98
Total	66,069	9,967 (15.1)	23,411	2,959 (12.6)	35,821	5,571 (15.6)	<.001
Pc		<.001		<.001		<.001

^aBased on a subject’s age as of July 1, 2014.

^bCompares white and black subjects on “no claims” by age. Calculated by using logistic regression likelihood ratio tests, adjusted for multiple comparisons.

^cTests of relationship of “no claims” to age for the total sample in each race group.

Table 3:

Rates of Pregnancy and Pregnancy Outcomes Among 15–17-year-old Medicaid Enrollees in Louisiana in 2014, by Race

	Totala	White	Black	RIRb	95% CI for RIRc,d	Pc,d
Per 1,000 person-years of Medicaid coverage in 2014
Live Birth	24.6	24.6	25.8	1.05	0.93—1.18	.43
Episode concluded with non-live birth	4.4	5.1	4.2	0.83	0.63—1.09	.17
Episodes with unknown outcome	5.2	4.7	5.6	1.20	0.91—1.56	.20
Total Episodes with outcomes in 2014	34.2	34.3	35.5	1.04	0.94—1.15	.49
Pregnancy Rate for 2014	38.6	42.7	36.1	0.85	0.77—0.93	.00
Per 1,000 subjects 15–17 years as of July 1,2014
Live Birth	24.6	24.4	25.7	1.05	0.88—1.27	.58
Episode concluded with non-live birth	4.4	5.0	4.2	0.83	0.54—1.28	.40
Episodes with unknown outcome	5.2	4.7	5.6	1.20	0.82—1.75	.35
Total Episodes with outcomes in 2014	34.1	34.1	35.5	1.04	0.90—1.21	.60
Pregnancy Rate for 2014	38.5	42.5	36.0	0.85	0.73—0.98	.03

^aIncludes all racial and ethnic groups in the sample.

^bRelative incidence ratio of black divided by white rates.

^cP-values and confidence intervals for per person-year results from negative binomial regression.

^dBootstrapping used to derive p-values and confidence intervals based on subject age as of July 1, 2014.

As shown in Table 4, Medicaid enrollees with claims occurring at any time in the available data had slightly higher calculated birth rates, but the racial comparisons (as reflected in the RIR values) changed very little.

Table 4:

Birth and Pregnancy Rates in 2014 (per person year of Medicaid Coverage) Among 15–17-year-old Medicaid Enrollees in Louisiana, by Race and Claims Status

	White	Black	RIRa	95% CI for RIRb	Pb
Live Birth Rates
From Primary Analysis	24.6	25.8	1.05	0.93—1.18	.43
Requiring at least one Claim	28.1	30.2	1.08	0.96—1.21	.21
Requiring Claim before 2013	27.6	30.2	1.09	0.94—1.28	.26
Pregnancy Rates
From Primary Analysis	42.7	36.1	0.85	0.77—0.93	.00
Requiring at least one Claim	48.7	42.3	0.87	0.79—0.96	.00
Requiring Claim before 2013	46.3	44.0	0.95	0.84—1.08	.41

^aRelative incidence ratio.

^bConfidence intervals and p-values calculated using negative binomial regression.

Discussion

To our knowledge, this is the first study to examine racial differences in pregnancy and birth rates among Medicaid-enrolled teens. We found no significant difference in birth rates between black and white Medicaid-enrolled teens ages 15–17 years in Louisiana. There was also no difference between black and white rates of non-live births or unknown pregnancy outcomes. This held true in validity analyses that considered differences in birth rate definitions and teens’ claims status. The lack of a racial difference in our data stands in stark contrast to national trends for which it is 66% more likely for black adolescents ages 15–17 years to give birth compared to whites in the same age group [15]. However, we did note a significant difference in pregnancy rates between groups, with white teens ages 15–17 years having higher pregnancy rates in most analyses. This may reflect varied possible scenarios that are not reflected in this data, such as higher contraception use by black teens. Also notable, however, is that the pregnancy rate among the sample is considerably higher than the aggregate outcome rate. This is partially because the aggregate outcomes pertain to pregnancies for which conception occurred at an earlier age and, among adolescents, conception rates increase rapidly with age. In addition, not all pregnancies eventuate in live births.

As previous studies have shown [16], low socioeconomic status is associated with teen childbearing. Our sample not only contained teens from low-income households but may have also contained a high percentage of teens from rural locations, both of which may be proxies for lower socioeconomic status. About 27% of Louisiana’s population lives in a rural area [17] and that may have impacted our teens’ access to healthcare and contraception. Other factors may also be playing a role, such as beliefs about contraception [18], [19] relationship patterns [18], and religious beliefs [20] that make low-income adolescents in Louisiana more similar to each other in their sexual and reproductive health behaviors, regardless of race, than they are to adolescents in the rest of the state or country.

Our study was limited in several ways. First, race and ethnic categories attached to each teen were mutually exclusive, and thus could not capture other ethnicities such as Hispanic black or Hispanic white. Our data was also limited in geographic specificity, and we were unable to explore effects of rural or urban residence. Additionally, claims codes provided an incomplete clinical picture regarding pregnancies and pregnancy outcomes, and an absence of a pregnancy code does not mean a pregnancy never occurred. For example, many pregnant teens may never present to care and receive a pregnancy diagnosis that is reflected in claims data, especially if they have an early miscarriage or abortion. Therefore, our use of pregnancy diagnosis codes may introduce bias that could be related to race. In addition, elective abortions are not covered by Medicaid and so these procedures rarely appeared in the data. Given the unreliable representation of abortion rates in the dataset, we did not attempt to measure these outcomes. Also, the overall quality of claims data reflects the consistency and accuracy of physician and nonphysician billing practices, which are prone to human error. Lastly, and perhaps most importantly, although Medicaid enrollment is a useful tool for analysis, it does not reflect different levels of poverty within that group of teens, nor circumstances related to poverty that may affect black and white teens differently, such as systemic racism and access to healthcare.

Despite these limitations, our study provides additional support for the argument that racial disparities in teen pregnancy and birth rates are more complicated than differences in race alone. Additional years of data are needed to determine if trends among Medicaid enrollees follow national trends regarding overall decreased teen birth rates over time and a narrowing of racial differences in teen pregnancies and births. An analysis of Medicaid data for evidence of contraceptive use will also be helpful in understanding potential points of intervention for these young teens at increased risk of pregnancy.

Adolescent mothers and their children are at increased risk of several medical [21], [22], [23], mental health [24], [25], educational [26], and socioeconomic difficulties [27]. Yet adolescent pregnancy is preventable. By recognizing that socioeconomic factors outside of race are important to consider when thinking about who is most at risk for teenage pregnancy, we can more precisely identify barriers teens face in pregnancy prevention, and develop feasible interventions tailored to their needs.

Implications and Contribution:

Nationally, black teens have much higher birth rates than whites. However, black and white teens ages 15–17 years had equal birth rates in analyses of Louisiana Medicaid data. Broader racial disparities in teen births may reflect geographic and socioeconomic influences in sexual behaviors that impact black and white teens differently.

Abbreviations:

RIR

relative incidence ratio

CI

confidence interval

ICD- 9

International Classification of Diseases, 9th Revision

ICD- 10

International Classification of Diseases, 10th Revision

CPT

Current Procedural Terminology