Write a single spaced 2 paragraph critique of the article. One paragraph should be a brief summary and the other should be a critique and providing suggestions on how the study could be improved. No

Methods.We used the National Survey of Family Growth and pooled pregnancyfiles from 2002 through the 2015–2017 surveys. We report baseline rural–suburban–urban disparities infirst pregnancy intention and outcomes. We used multinomial logistic regression to estimate these disparities, accounting for sociodemographic background, religious upbringing, and other factors.

Results.Thefirst adolescent pregnancies of rural women were more likely to be unintended and end in live birth relative to their urban counterparts. Disparities were most striking among Black adolescents, with about 60% offirst adolescent pregnancies among rural Black women being unintended and ending in live birth (urban: 51%). Newly collected state health department data on rural and urban adolescent births and abortions corroborate thefindings from the National Survey of Family Growth.

Conclusions.Rural–urban differences in the share offirst adolescent pregnancies ending in live births are not accounted for by pregnancy intention or confounding individual-level characteristics. Future research should explore the role of structural bar- riers, including access to family planning and abortion services. (Am J Public Health.2019; 109:1762–1769. doi:10.2105/AJPH.2019.305318) I n 1967, President Lyndon B. Johnson’s National Advisory Commission on Rural Poverty produced a report entitledThe People Left Behind. The rural United States was then characterized by low education, chronic poverty, high unemployment, dilapidated housing, and food insecurity. Although the life chances of rural people have improved substantially, 1rural–urban differences in life expectancy have recently widened. 2Rural people face significant deficits in health be- haviors and outcomes, 3,4 limited access to community-based health care, 5,6 and expo- sure to unhealthy living conditions at home and work. 7High rates of rural mortality from drug abuse, alcoholism, and suicide 8,9 high- light a new“geography of despair”and give urgency to theHealthy People 2020report’s goal of eliminating persistent spatial in- equalities (http://bit.ly/2kV3vMP). Un- fortunately, rural women today are often“leftbehind”in such discussions, especially on matters affecting them most, such as re- productive health and childbearing.

Indeed, a 2016 Centers for Disease Con- trol and Prevention (CDC) brief marked the first government report of rural–urban dis- parities in adolescent birth rates. 10Following the 2016 presidential election, the report received widespread news coverage, with the Los Angeles Timespublishing an article titled “There’s Another Type of Rural–Urban Divide in America: Teens Having Babies.” 11 The National Campaign to Prevent Teen andUnintended Pregnancy 12first documented this trend in 2013, reporting greater recent declines in adolescent birth rates in urban than rural counties. Adolescent birth rates in rural areas are about one third higher than in urban areas. 10,12 These reports placed the national spotlight on large and persistent rural–urban disparities in adolescent birth rates. Rural adolescents may be vulnerable to local conditions that limit unintended preg- nancy management options.

Although about onefifth of US adolescent pregnancies are intended, 13adolescent pregnancies and births are usually assumed to be unintended when information on preg- nancy intentions is unavailable. Whether rural–urban disparities in adolescent births in part reflect differences in pregnancy in- tentions or in how unintended pregnancies are resolved is unclear but important for public policy and family planning. In addi- tion, assessing unmet need in early adulthood can be especially challenging without reli- able information on pregnancy intentions.

Women in their early 20s have higher rates of unintended pregnancies than do adolescents, 13and early adulthood is a critical life stage. Unintended pregnancies can derail educational plans and restrict employment options 14among rural women already facing limited economic prospects.

Another important unanswered question is whether rural–urban disparities in women’s unintended pregnancy outcomes—a live birth or not—mostly reflect well- documented geographical differences in ABOUT THE AUTHORSApril Sutton is with the Department of Sociology, University of California, San Diego, La Jolla. Daniel T. Lichter and Sharon Sassler are with the Department of Policy Analysis and Management, Cornell University, Ithaca, NY. Daniel T. Lichter is also with the Department of Sociology, Cornell University.

Correspondence should be sent to April Sutton, Department of Sociology, 9500 Gilman Drive, MC 0533, UC San Diego, La Jolla, CA 92093 (e-mail: [email protected]). Reprints can be ordered at http://www.ajph.org by clicking the“Reprints”link.

This article was accepted July 30, 2019.

doi: 10.2105/AJPH.2019.305318 1762ResearchPeer ReviewedSutton et al.AJPHDecember 2019, Vol 109, No. 12 AJPHOPEN-THEMED RESEARCH family socioeconomic status and religious upbringing 1,15,16 or result from spatial dis- parities in local conditions. For example, rural women face large geographic barriers to abortion providers. Research shows that 95% of women in California lived in counties with an abortion clinic in 2014 compared with only 9%, 4%, and 23% in the rural states of Mississippi, Montana, and South Dakota, respectively. 17 Rural–urban disparities in educational and economic opportunities may also shape rural–urban disparities in whether unintended preg- nancies end in a live birth or abortion. In- deed, women in poor communities may perceive lower opportunity costs to having a child, 18and rural women benefitlessfroma college degree and often have fewer em- ployment options than do urban women. 19 We used recently released data from the National Survey of Family Growth (NSFG) to investigate these unresolved questions and provide new estimates of rural–urban dis- parities in pregnancy intentions and outcomes among female adolescents (aged 15–19 years) and young women (aged 20–24 years). These individual-level data allow us to assess unmet need by examining the resolutions of un- intended pregnancies. We also highlight rural and urban disparities across racial/ethnic minority populations. Unlike previous studies that used conventional population or other highly aggregated areal data, the NSFG en- ables us to show rural–urban disparities that “net out”key individual-level sociodemo- graphic and cultural differences. Residual gaps may reflect structural disadvantages in rural areas, including distance from abortion clinics. Comprehensive sex education and access to contraception also may be less available to rural adolescents. Sensitivity analyses using newly collected state health department data on rural, suburban, and urban adolescent births and abortion cor- roborate our main conclusions. METHODS We used data from the NSFG, a National Center for Health Statistics (NCHS) survey of a nationally representative sample of women (civilian and noninstitutionalized) aged 15 to 44 years. We pooled pregnancyfiles from the 2002, 2006–2010, 2011–2013, 2013–2015,and the newly released 2015–2017 NSFG surveys (n = 60 768). We restricted our analysis tofirst pregnancies occurring be- tween the ages of 15 to 19 years and 20 to 24 years that were not in progress at the time of interview (age 15–19 years: n = 9134; age 20–24 years: n = 5977). We only examined first pregnancies since 1995 (age 15–19 years:

n = 5256; age 20–24 years: n = 4078). This sample restriction excluded pregnancies among women who lacked legal access to abortion beforeRoe v. Wadeand ensured the timeliness of our results while maintaining adequate cell sizes. We also eliminated cases with missing or NCHS-imputed data for measures used to construct our dependent variable (age 15–19 years: n = 5233; age 20–24 years: n = 4053). For women aged 20 to 24 years, we excluded observations with missing data on our high school diploma measure (n = 4033).

The NSFG includes a measure of preg- nancy intention (i.e.,“wantresp”) with 6 mutually exclusive categories based on sur- vey items that ask respondents to recall the wantedness and timing of each pregnancy right before it occurred. Consistent with convention, 13,20 we defined intended preg- nancies as those reported as occurring at the right time or later than desired (“overdue”), and those for which the woman was in- different or was not sure about the timing.

Unintended pregnancies included those re- ported as unwanted or occurring too soon (mistimed). Our dependent variable com- bined information on pregnancy intention and pregnancy outcome and consisted of the following categories: (1) intended pregnancy ended in live birth, (2) unintended preg- nancy ended in live birth, (3) unintended pregnancy ended in abortion, and (4) a residual category (else), which included pregnancies ended in miscarriage or stillbirth, or intended pregnancies ended in abortion (0.55% of an- alytic sample). Data from the NSFG is subject to abortion underreporting, 13but sensitivity analyses of abortion data from state health departments (see Results section) provide reassurance that abortion underreporting cannot explain the rural disadvantage.

Finally, the NSFG assigns US Office of Management and Budget (OMB)–defined metropolitan statistical area (MSA) status based on respondents’address at the time of interview. We refer to women living inprincipal cities of MSAs as“urban,”in other parts of MSAs as“suburban,”and outside MSAs as“rural.”We defined race/ethnicity as non-Hispanic White, non-Hispanic Black, Hispanic, and other. Our control variables included age at interview, year of conception, whether the conception occurred within 48 months of the interview, respondent’s mother’s age atfirst birth and educational attainment, whether the respondent lived with both biological or adoptive parents since birth, the religion (and Christian de- nomination) in which the respondent was raised, whether the respondent held a high- school diploma at the time of conception (for women aged 20–24 years), and marital status at conception. Descriptive statistics for controls are shown in Table A1 (available as a supplement to the online version of this article at http://www.ajph.org).

Wefirst present weighted descriptive statistics for pregnancy intentions and reso- lutions by urbanicity and racial/ethnic background. Given the potential for resi- dential relocation and recall bias for older conceptions, we present the distribution of our dependent variable amongfirst concep- tions occurring within 48 months of the survey interview (cutoff guided by NSFG cell size recommendations) for comparison.

Next, we report average marginal effects estimated from multinomial logistic re- gressions predicting the relationship between metropolitan status and pregnancy intentions and resolutions, adjusting for controls. For example, the average marginal effect of rural residence is the computed difference between the rural and urban (the referent) average predicted probabilities of the out- come, holding all other covariates at observed values.

Next, we report the results of our calcu- lation of adjusted average predicted proba- bilities for urbanicity by race/ethnicity among adolescent conceptions (where cell sizes permit), holding all other controls at observed values. We did not include interactions in the model used to predict these probabilities.

A Wald test indicated that the overall in- teraction between urbanicity and race was not statistically significant. Moreover, the re- lationship between urbanicity and pregnancy intention and resolution for Black and His- panic adolescent conceptions was not statis- tically significantly different compared with AJPHOPEN-THEMED RESEARCH December 2019, Vol 109, No. 12AJPHSutton et al.Peer ReviewedResearch1763 that of White adolescent conceptions. We used NSFG-imputed variables for missing data and applied appropriate weights to ac- count for the survey design.

RESULTS Table 1 shows that, relative to urban women’sfirst pregnancies, a larger share of rural women’sfirst pregnancies were inten- ded (age 15–19 years: 25% vs 21%; age 20–24 years: 55% vs 47%). For women aged 20 to 24 years, the share of unintended pregnancies ending in live birth was about 19 percentage points higher for rural relative to urban women (73% vs 54%) while the share of unintended pregnancies ending in abortion was about 14 percentage points lower for rural relative to urban women (rural: 13%; urban:

27%). We saw a consistent pattern among adolescent unintended pregnancies.

Moving to our dependent variable, a larger share of both intended and unintended pregnancies among rural women ended in live births. Smaller shares of pregnancies among rural women were unintended and ended in abortion.

A striking share of rural Black and Hispanic women’s adolescent pregnancies were un- intended and ended as births (60%). Notably, the share of adolescent pregnancies that were un- intended and ended in abortion was 4% among Black rural women compared with 20% among Black urban women. We observed a 15-per- centage-point difference between urban White women (45%) and rural Black women (60%) in the share of adolescent pregnancies that were unintended and ended in birth. Conversely, the share of adolescent pregnancies that were un- intended and ended in birth was 1 percentage point lower for urban Black women (51%) than for rural White women (52%), on average.

With the exceptions offirst pregnancy intentions among women aged 15 to 19 years, conceptions among Hispanic women aged 15 to 19 and 20 to 24 years, and conceptions among Black women aged 20 to 24 years, the c 2test rejected the null hypothesis of in- dependence for the previously mentioned relationships (P<.05). The patterns in un- intended births among suburban women paralleled those in big cities.

Among conceptions occurring within 48 months of the survey interview, 13% of rural TABLE 1—Weighted Percentages for Intention and Resolution of First Pregnancies by Urbanicity: United States, 1995–2017 Ages 15–19 Years (n = 5233), % Ages 20–24 Years (n = 4033), % Urban (n = 2325)Suburban (n = 2006)Rural (n = 902)Urban (n = 1569)Suburban (n = 1818)Rural (n = 646) First pregnancy intentions a Intended 21 20 25 47 48 55 Unintended 79 80 75 53 52 45 First pregnancy resolutions a,b Birth 67 67 76 69 73 81 Abortion 16 16 7 16 11 6 Else 17 17 17 15 16 13 Unintendedfirst pregnancies a,b Birth 63 62 74 54 61 73 Abortion 20 19 9 27 20 13 Else 18 19 17 19 18 14 Pregnancy intention and resolution a,b Intended, birth 17 18 21 41 41 48 Unintended, birth 49 50 55 29 32 33 Unintended, abortion 16 15 7 15 11 6 Else 18 18 17 16 16 14 Conceptions within 48 mo of interview c Intended, birth 13 16 15 35 43 42 Unintended, birth 46 44 55 28 29 35 Unintended, abortion 19 18 9 18 11 6 Else 22 22 21 19 17 17 Race/ethnicity Non-Hispanic White a,b Intended, birth 15 15 22 37 42 50 Unintended, birth 45 46 52 27 30 33 Unintended, abortion 17 16 8 19 11 6 Else 23 22 18 17 17 12 Non-Hispanic Black b Intended, birth 11 13 19 29 27 22 Unintended, birth 51 52 60 40 41 40 Unintended, abortion 20 20 4 13 14 15 Else 19 15 17 18 19 22 Hispanic Intended, birth 25 24 21 55 49 . . .

Unintended, birth 53 54 60 25 33 . . .

Unintended, abortion 9 9 9 8 6 . . .

Else 13 13 11 12 13 . . .

Note. Sample restricted tofirst pregnancies occurring between 1995 and 2017. Because of small cell sizes, estimates for women from other races/ethnicities and rural Hispanic women aged 20–24 years are not shown.

Source.National Survey of Family Growth, pooled pregnancyfiles from 2002, 2006–2010, 2011–2015, and 2015–2017.

aP<.05 fromc 2test of relationship between urbanicity and the outcome for women aged 20–24 years. bP<.05 fromc 2test of relationship between urbanicity and the outcome for women aged 15–19 years. cn = 1070 among those aged 15–19 years and n = 969 among those aged 20–24 years. AJPHOPEN-THEMED RESEARCH 1764ResearchPeer ReviewedSutton et al.AJPHDecember 2019, Vol 109, No. 12 and 15% of urban women’s adolescent con- ceptions were intended and ended via live birth. We saw a larger rural–urban disparity in the share of these more recent adolescent pregnancies ending in unintended birth (55% in rural vs 46% in urban) compared with adolescent conceptions irrespective of elapsed time (55% in rural vs 49% in urban). Rural– urban disparities in unintended births among women aged 20 to 24 years were also larger among more recent conceptions.

Table 2 presents selected average marginal effects of metropolitan status onfirst pregnancy intentions and resolutions, with adjustment for family socioeconomic and religious back- ground, marital status at conception, and other factors (full set of average marginal effects shown in Table A2, available as a supplement to the online version of this article at http:// www.ajph.org). Compared with urban women, on average, rural women’s adolescent pregnancies are about 7.6 percentage points more likely to be unintended and end in live birth (P<.01) but are about 8.4 percentage points less likely to be unintended and end in abortion (P<.001). Although the overall ru- ral–urban patterns among young women’s conceptions are similar to those among ado- lescents, the rural–urban disparity in un- intended births is weaker and nonsignificant.In Table 3, we present the adjusted average predicted probabilities for urbanicity by race/ ethnicity, holding all other covariates at their observed values. We focused on adolescent conceptions given the previously outlined results and because of small cell sizes among women aged 20 to 24 years. The rural–urban patterns among White, Hispanic, and Black women are consistent with the aforemen- tioned results. Sensitivity Analyses Abortion underreporting may threaten our main results if rural women are more likely to underreport abortions than are urban women. This would result in larger shares of unobserved rural abortions (and unintended pregnancies) and bias our estimates of rural–urban disparities. However, rural underreporting is unlikely to account for the rural–urban disparities we observed. Other highly reliable data from the Guttmacher Institute and elsewhere indicate that women from more rural areas are less likely to have an abortion than are their more urban coun- terparts. 21,22 We also collected health de- partment data from 25 states that mandate provider abortion reporting and identifiy maternal county of residence. Significantly,our sensitivity analysis with these new data aligns with recent recommendations for research using survey reports of abortion 23 and reveals metro–nonmetro patterns con- sistent with our NSFG results.

To illustrate, we compared abortion ratios based on the NSFG to those calculated from state health department data (Table 4). These state health department data are provided at the county level, which allowed us to distinguish between all metropolitan residents (i.e., sub- urban and urban) and nonmetropolitan resi- dents (i.e., rural). We followed others 17and defined abortion ratios as the proportion of pregnancies resolved through abortion, exclud- ing miscarriages and stillbirths.

For comparison, the NSFG revealed persistent metro–nonmetro disparities in abortion ratios amongfirst pregnancies and all pregnancies, and among pregnancies across 1995 to 2017 and 2000 to 2017. Among all adolescent pregnancies between 2000 and 2017, the NSFG nonmetro abortion ratios were about 47% lower than overall metropolitan abortion ratios (0.10 vs 0.19). Nonmetro abortion ratios from state health department data across these same years were about 52% lower than metro abortion ratios (0.14 vs 0.29). High numbers of abortions in New York disproportionately influenced these TABLE 2—Difference in Probability of First Pregnancy Intention and Resolution by Metropolitan Status and Race/Ethnicity: United States, 1995–2017 Age 15–19 Years (n = 5233), Difference a(95% CI) Age 20–24 Years (n = 4033), Difference a(95% CI) Intended, BirthUnintended, BirthUnintended, Abortion ElseIntended, BirthUnintended, BirthUnintended, Abortion Else Metro status (Ref: Urban) Suburban–0.003 (–0.040, 0.034)0.012 (–0.030, 0.054)–0.005 (–0.038, 0.027)–0.003 (–0.042, 0.035)–0.010 (–0.052, 0.033)0.041 (–0.001, 0.084)–0.037 (–0.070,–0.004)0.005 (–0.029, 0.039) Rural 0.027 (–0.024, 0.077)0.076 (0.021, 0.131)–0.084 (–0.120,–0.048)–0.019 (–0.062, 0.024)0.063 (–0.005, 0.131)0.050 (–0.016, 0.115)–0.087 (–0.126,–0.048)–0.025 (–0.068, 0.017) Race/ethnicity (Ref: Non-Hispanic White) Non-Hispanic Black–0.023 (–0.059, 0.014)0.029 ( –0.025, 0.083)0.040 (0.000, 0.080)–0.046 (–0.090,–0.003)–0.056 (–0.118, 0.005)0.039 (–0.023, 0.101)–0.007 (–0.043, 0.029)0.025 (–0.021, 0.071) Hispanic 0.027 (–0.012, 0.067)0.070 (0.019, 0.122)–0.035 (–0.069,–0.001)–0.063 (–0.108,–0.018)0.059 (–0.005, 0.122)0.003 (–0.067, 0.073)–0.035 (–0.074, 0.004)–0.027 (–0.069, 0.016) Other–0.034 (–0.111, 0.043)0.096 (–0.020, 0.213)0.041 (–0.034, 0.117)–0.104 (–0.168,–0.040)–0.048 (–0.161, 0.064)–0.084 (–0.170, 0.001)0.061 (– 0.023, 0.144)0.072 (–0.037, 0.181) Note.CI = confidence interval. Sample restricted tofirst pregnancies occurring between 1995 and 2017. Model also controlled for education level of respondent’s mother, age of respondent’s mother atfirst birth, whether lived with both biological or adoptive parents since birth, religious affiliation in which respondent was raised, whether respondent was unmarried atfirst conception, age at interview, year of conception, whether pregnancy occurred within 48 months of interview, and whether respondent had a high school degree at conception (among women aged 20–24 years).

Source.National Survey of Family Growth, pooled pregnancyfiles from 2002, 2006–2010, 2011–2015, and 2015–2017.

aAverage marginal effects from multinomial logistic regression. AJPHOPEN-THEMED RESEARCH December 2019, Vol 109, No. 12AJPHSutton et al.Peer ReviewedResearch1765 population-weighted statistics and may be less accurate if judged by observed discrepancies in the numbers of abortions reported to CDC and Guttmacher. Yet, even when New York is excluded, nonmetro abortion ratios are about 46% lower than metro abortion ratios.

Although cell sizes prohibit comparison with the NSFG for recent years, we presented abortion ratios from state health departments for the years 2011 to 2017 and across racial/ ethnic groups from selected states providing such breakdowns. Nonmetro abortion ratios were about 69% and 48% lower than those of large central and other metro counties, respectively (nonmetro: 0.11; large central:

0.34; other metro: 0.21). We observed this overall pattern across states. Compared with large central metro ratios, nonmetro abortion ratios were about 69%, 54%, and 50% lower for non-Hispanic White, non-Hispanic Black, and Hispanic women, respectively (among selected states).

Nationally, most adolescent pregnancies are unintended, 13and Table 1 shows anonsignificant relationship between metro residence and adolescentfirst pregnancy inten- tion (unintended: urban = 79%; rural = 75%).

These state patterns likely indicate larger shares of nonmetro adolescent unintended preg- nancies ending in live birth (vs abortion).

State health department data arguably provide the most appropriate and reliable abortion data for this sensitivity analysis because the Guttmacher Institute does not provide abortion data by county of residence.

Still, health departments do not fully capture out-of-state abortions, and abortion providers may not report all abortions to state health departments. These potential reporting problems, however, cannot explain patterns shown in Table 4. Specifically, sensitivity analyses indicated disparities in rural–urban abortion ratios among (1) states with abortion occurrences that closely match those reported by the Guttmacher Institute and (2) counties for which the nearest abortion provider was located within the state (where out-of-state travel would be least likely).We also conducted a variety of other sen- sitivity analyses. For example, rural, urban, and suburban residents at the time of survey may have lived in areas with different urbanicity classifications from those of the areas at the time of conception. However, residential relocation is an unlikely source of bias. Indeed, we ob- served consistent patterns with state health department data, which provided maternal county of residence both at the time of birth and abortion. Moreover, rural–urban baseline disparities in unintended childbearing among adolescent conceptions occurring within 18 months of the survey interview—women with lower chances of relocation—are larger than those reported here (Table B1, available as a supplement to the online version of this article at http://www.ajph.org).

Given that many states require parental consent for an abortion, we estimated models for women aged 19 to 21 years—legal adults who were also less likely to be under direct parental supervision. Results were consistent (Table B2, available as a supplement to the online version of this article at http:// www.ajph.org). Analyses also demonstrated a higher share of adolescent pregnancies among rural women ended in live birth when re- stricted to unintendedfirst pregnancies, net of controls (Table B3, available as a supplement to the online version of this article at http:// www.ajph.org). Furthermore, rural–urban disparities in the outcomes of both un- wanted and mistimed recent conceptions were consistent with patterns shown in Table 1 (Table B4, available as a supplement to the online version of this article at http:// www.ajph.org). DISCUSSION This article highlighted patterns of preg- nancy intentions and childbearing in the rural United States at a time when rural people and places have been“left behind”in a globalizing economy and urban-centric policies. Our goal was to identify geographic and racial disparities infirst pregnancy intentions and resolutions, focusing in particular on the resolution of unintended pregnancies. Data from the NSFG (2002–2017) provide a singular conclusion: rural adolescents are significantly more likely than are their sub- urban and big-city counterparts to have TABLE 3—Adjusted Average Predicted Probabilities of First Pregnancy Intention and Resolution for Urbanicity by Race/Ethnicity: United States, 1995–2017 Ages 15–19 Years (n = 5233), PP (95% CI) Intended, Birth Unintended, Birth Unintended, Abortion Else Non-Hispanic White Urban 0.176 (0.140, 0.212) 0.460 (0.419, 0.500) 0.152 (0.121, 0.182) 0.213 (0.171, 0.255) Suburban 0.173 (0.142, 0.203) 0.471 (0.432, 0.511) 0.146 (0.120, 0.173) 0.209 (0.175, 0.243) Rural 0.204 (0.161, 0.247) 0.535 a(0.488, 0.582) 0.069 a(0.040, 0.098) 0.192 (0.155, 0.228) Non-Hispanic Black Urban 0.153 (0.122, 0.183) 0.485 (0.440, 0.530) 0.197 (0.154, 0.240) 0.165 (0.125, 0.205) Suburban 0.150 (0.118, 0.183) 0.497 (0.446, 0.549) 0.190 (0.145, 0.235) 0.162 (0.125, 0.199) Rural 0.180 (0.132, 0.228) 0.576 a,b(0.517, 0.635) 0.092 a,c(0.050, 0.135) 0.151 (0.110, 0.192) Hispanic Urban 0.204 (0.165, 0.244) 0.533 (0.486, 0.580) 0.112 (0.082, 0.143) 0.150 (0.113, 0.187) Suburban 0.201 (0.164, 0.237) 0.545 (0.498, 0.592) 0.108 (0.077, 0.139) 0.146 (0.108, 0.185) Rural 0.229 (0.172, 0.285) 0.594 a(0.530, 0.658) 0.049 a(0.024, 0.073) 0.128 (0.088, 0.169) Note. CI = confidence interval; PP = predicted probability. Adjusted average predicted probabilities estimated from the multivariable regression presented in Table 2. Model controlled for education level of respondent’s mother, age of respondent’s mother atfirst birth, whether lived with both biological or adoptive parents since birth, religious affiliation in which respondent was raised, whether respondent was unmarried at conception, age at interview, year of conception, and whether pregnancy occurred within 48 months of interview.

Source.National Survey of Family Growth, pooled pregnancyfiles from 2002, 2006–2010, 2011–2015, and 2015–2017.

aStatistically significant (2-tailedP<.05) differences relative to pregnancies of urban women within the same race/ethnicity category.

bRural–urban gap in the probability of unintended birth is statistically significantly different from the rural–urban gap among White and Hispanic women.

cRural–urban gap in the probability of abortion is statistically significantly different from the rural–urban gap among White women.

AJPHOPEN-THEMED RESEARCH 1766ResearchPeer ReviewedSutton et al.AJPHDecember 2019, Vol 109, No. 12 unintendedfirst pregnancies that end in a live birth.

Our study makes several specific contri- butions. First, it provides baseline estimates of rural–urban disparities in pregnancy in- tentions and their resolution. Significantly, rural–urban disparities in the share offirst adolescent pregnancies ending in unintended live birth cannot be explained by differences in key cultural factors—such as early marriage or religion—or by family background.

Second, the adolescent pregnancies of rural women were not more likely to end in live birth simply because they were more likely to be intended. Rather, compared with those in big cities and suburbs, a larger proportion of ruralfirst pregnan- cies were unintended and ended in live birth.

Third, we examined pregnancy outcomes among women aged 20 to 24 years. We did not observe statistically significant rural– urban disparities in the probability of having a first conception that ended in unintended live birth, net of controls, among these young women. Yet, when we considered resolu- tions of more recent conceptions and reso- lutions among unintended conceptions only (Table 1), we saw sizeable baseline disparities.

Notably, 73% of rural women’s unintended conceptions ended in live birth compared with only 54% of urban women’s unintended conceptions. Research should further ex- amine rural–urban disparities in unintended pregnancy resolution among women in early adulthood.

Fourth, we presented rural–urban dispar- ities by race and ethnicity. Rural women of color are often neglected in the academy and public health sphere. However, our results revealed large baseline shares of His- panic and Black adolescent conceptions ending in unintended births and striking rural–urban disparities among Black adoles- cents. In addition, we saw comparable shares of urban Black and rural White adolescent conceptions ending in unintended births, further underscoring the importance of considering both urbanicity and racial/ethnic background when studying pregnancy in- tention and resolutions.

Using up-to-date survey data and the most reliable source of county-level data on abortion, our analyses of unintended preg- nancies—and their resolution—placed the TABLE 4—Adolescent Abortion Ratios by Metropolitan Status, National Survey of Family Growth and State Health Department Data: United States, 1995–2017 Metro Total Large Central Other Metro Nonmetro NSFG First pregnancies, 1995–2017 (n = 4401) 0.19 0.08 First pregnancies, 2000–2017 (n = 2849) 0.18 0.10 All pregnancies, 2000–2017 (n = 4349) 0.19 0.10 State health department data 2000–2017 a 0.29 0.14 2011–2017 b 0.34 0.21 0.11 Race/ethnicity (select states; 2011–2017) c Non-Hispanic White 0.32 0.19 0.10 Non-Hispanic Black 0.28 0.25 0.13 Hispanic 0.14 0.11 0.07 States (2011–2017) Arkansas NA 0.12 0.07 Colorado 0.23 0.21 0.16 Delaware NA 0.32 NA Georgia 0.37 0.23 0.10 Idaho NA 0.17 0.12 Kansas NA 0.16 0.08 Maine NA 0.29 0.20 Michigan 0.29 0.27 0.14 Minnesota 0.32 0.27 0.11 Mississippi NA 0.15 0.10 Missouri 0.19 0.17 0.06 Montana NA 0.26 0.18 Nevada 0.22 0.20 0.15 New York (2011–2016) 0.58 0.41 0.28 North Carolina 0.31 0.20 0.13 North Dakota NA 0.19 0.13 Ohio 0.28 0.19 0.10 Oregon 0.38 0.27 0.18 Pennsylvania 0.42 0.25 0.11 South Carolina NA 0.20 0.15 South Dakota NA 0.10 0.04 Utah 0.18 0.11 0.06 Texas (2011–2016) 0.20 0.12 0.08 Virginia 0.39 0.25 0.09 Washington (2011–2016) 0.50 0.30 0.23 Note. NA = not applicable; NSFG = National Survey of Family Growth. Abortion ratios were calculated as the proportion of pregnancies (excluding miscarriages and stillbirths) that ended in abortion. 17The National Center for Health Statistics rural–urban classification schemes (based on 2000 and 2010 Office of Management and Budget delineations of metropolitan statistical areas) 24were applied to county- level data from state health departments.“Total”metro includes large central metro and other metro counties. Additional data sources available in Appendix C (available as a supplement to the online version of this article at http://www.ajph.org). aCalculated from all states listed except Arkansas, Maine, Ohio, and South Dakota for which multiple years could not be obtained. Ratios calculated without New York: metro: 0.24, nonmetro: 0.13.

bCalculated from all states listed. Ratios calculated without New York: large central: 0.27, other metro:

0.20, nonmetro: 0.11.

cCalculated from the following states that provided custom data by both race and ethnicity: Arkansas, Colorado, Georgia, Idaho, Kansas, Michigan, Minnesota, Missouri, Montana, Nevada, North Dakota, Oregon, and Texas. Rates without Texas were White—large central: 0.31, other metro: 0.20, nonmetro:

0.10; Black—large central: 0.28, other metro: 0.26, nonmetro: 0.13; Hispanic—large central: 0.14, other metro: 0.15, nonmetro: 0.09. AJPHOPEN-THEMED RESEARCH December 2019, Vol 109, No. 12AJPHSutton et al.Peer ReviewedResearch1767 spotlight on rural women, who are often invisible in the empirical literature. Our sensitivity analyses indicated that limitations of the NSFG, especially on the under- reporting of abortions, are unlikely to affect our main conclusions. Indeed, rural–urban patterns in the resolution of adolescent pregnancies from state health data corrobo- rate ourfindings from the NSFG.

Our study represents a national baseline for further study, not thefinal answers to a neglected topic that clearly deserves more attention. For example, we were only able to speculate about possible mechanisms un- derlying the rural–urban disparities. Perhaps the most obvious potential mechanism is more limited access of rural women to family planning and abortion clinics. 25In fact, new state legislative restrictions on abortion access 26likely affect isolated rural women most. 27,28 Rural adolescents also may perceive fewer opportunity costs of an unintended birth because of limited employment options. 19 Local cultural scripts also may discourage rural adolescents—regardless of family background and religious upbringing—from terminating unintended pregnancies. Our estimates provide an empirical benchmark for future research to explore how restricted abortion access, local economic disadvantages, and cultural norms shape unintended pregnancy management in the rural United States.

Public Health Implications The conceptualization and measurement of pregnancy intentions are increasingly fraught with controversy. 29,30 As a public health issue, reducing unintended pregnan- cies is usually viewed as a matter of either insuring reproductive autonomy or of im- proving birth outcomes. 13In the rural South, for example, more than 50% of all African American babies are born to poor mothers, 31 which is likely linked to failures of the health care system and unintended childbearing. 32 Limited use of or access to reproductive health services and poor birth outcomes is presumably linked to intergenerational poverty. A reproductive justice framework, however, does not problematize or stigmatize women’s pregnancy intentions but instead emphasizes the barriers or constraints to family planning services, including abortion, that limit reproductive autonomy. 30The factthatfirst pregnancies of rural adolescents are more likely to end as unintended births— even independent of individual-level risk factors—suggests that living in rural areas may restrict reproductive autonomy and women’s choices for managing unintended pregnan- cies. By contrast, urban adolescents—even those with an unintended pregnancy—are presumably better able to exercise their reproductive autonomy, especially if abortion is regarded as an“accepted, legitimate and accessible means of fertility control.” 30(p1) Conclusions Rural Americans continue to be“left behind”despite significant economic gains following the publication ofThe People Left Behindmore than 50 years ago. Our research shines a light on arguably the most neglected of the left behind. Urban–rural disparities are perhaps less tied today than in the past to individual sociodemographic factors. Issues such as geographic access and isolation in health care—including reproductive health care—may instead represent an important dimension of the new geography of ex- clusion. For rural women, the long-run costs of unintended pregnancies seem most ger- mane, especially because unintended child- bearing has negative consequences for educational attainment, earnings, and ma- ternal and child health. 14,33–35 Now is not the time to restrict access to family planning services in rural areas or ignore other con- straints on rural women’s reproductive autonomy. CONTRIBUTORSA. Sutton and D. T. Lichter wrote the article. A. Sutton collected data, conducted the analyses, managed the revisions and responses to reviewers, and contributed to the conceptualization and design of the study. D. T.

Lichter conceptualized the study and contributed to its design. S. Sassler contributed to the conceptualization and design of the study and to writing and editing the article.

All authors critically reviewed the study.

ACKNOWLEDGMENTSWe are thankful for the support of the Frank H. T. Rhodes Postdoctoral Fellowship Program and the Cornell Population Center at Cornell University. This article benefitted from presentations at the annual meetings of the Rural Sociological Society and Population Association of America. We appreciate Bolun Zhang’s assistance compiling the data from state health departments.

We are also grateful for the helpful feedback of our anonymous reviewers.

CONFLICTS OF INTERESTThere are no funding or other conflicts of interest. HUMAN PARTICIPANT PROTECTIONThis study reports results from analysis of de-identified, publicly released data and is exempt from institutional review board review as per section 46.101(b) of National Institutes of Health document 45 CFR 46.

REFERENCES1. Lichter DT, Ziliak JP. The rural–urban interface:

new patterns of spatial interdependence and inequality in America.Ann Am Acad Pol Soc Sci. 2017;672(1):

6–25.

2. James W, Cossman JS. Long-term trends in Black and White mortality in the rural United States: evidence of a race-specific rural mortality penalty.J Rural Health.

2017;33(1):21–31.

3. Roberts ME, Doogan NJ, Stanton CA, et al. Rural versus urban use of traditional and emerging tobacco products in the United States, 2013–2014.Am J Public Health. 2017;107(10):1554–1559.

4. Meit M, Knudson A, Gilbert T, et al. The 2014 update of the rural–urban chartbook. Rural Health Policy and Research Centers. Available at: https://www-cdc-gov.

ezproxy.lib.utexas.edu/nchs/data/data_acces_files/2014- rural-urban-chartbook-update.pdf. Accessed September 12, 2019.

5. Hart LG, Larson EH, Lishner DM. Rural definitions for health policy and research.Am J Public Health. 2005; 95(7):1149–1155.

6. The obstetrician-gynecologist distribution atlas.

Washington, DC: American Congress of Obstetricians and Gynecologists; 2013.

7. Marré A. Rural education at a glance, 2017 edition. EIB 171. Washington, DC: US Department of Agriculture, Economic Research Service; April 2017.

8. Nestadt PS, Triplett P, Fowler DR, Mojtabai R.

Urban–rural differences in suicide in the state of Mary- land: the role offirearms.Am J Public Health. 2017; 107(10):1548–1553.

9. Monnat SM, Rigg KK. Examining rural/urban dif- ferences in prescription opioid misuse among US ado- lescents.J Rural Health. 2016;32(2):204–218.

10. Hamilton BE, Rossen LM, Branum AM. Teen birth rates for urban and rural areas in the United States, 2007–2015. National Center for Health Statistics Data Brief No. 264. Atlanta, GA: Centers for Disease Control and Prevention; November 2016.

11. Kaplan K. There’s another type of rural–urban divide in America: teens having babies.Los Angeles Times.

November 16, 2016. Available at: https://www.latimes.

com/science/sciencenow/la-sci-sn-teen-birth-rate- rural-urban-20161116-story.html. Accessed September 12, 2019.

12. Ng AS, Kaye K. Science says 47: Teen childbearing in rural America. The National Campaign to Prevent Teen and Unplanned Pregnancy. 2013. Available at: http:// youthtoday.org/wp-content/uploads/sites/13/hotdocs/ teen_child_bearing_in_rural_america.pdf. Accessed September 12, 2019.

13. Finer LB, Zolna MR. Unintended pregnancy in the United States: incidence and disparities, 2006.Contraception.

2011;84(5):478–485.

14. Fletcher JM, Wolfe BL. Education and labor market consequences of teenage childbearing: evidence using the timing of pregnancy outcomes and communityfixed effects.J Hum Resour. 2009;44(2):303 –325.

15. Housing Assistance Council. Poverty in rural America: rural poverty and income at a glance. September AJPHOPEN-THEMED RESEARCH 1768ResearchPeer ReviewedSutton et al.AJPHDecember 2019, Vol 109, No. 12 2011. Available at: http://www.ruralhome.org/storage/ documents/info_sheets/povertyamerica.pdf. Accessed December 1, 2018.

16. Kelly P, Lobao L. The social bases of rural–urban political divides. Social status, work, and sociocultural beliefs. Rural Sociol . In press.

17. Jones RK, Jerman J. Abortion incidence and service availability in the United States, 2014. Perspect Sex Reprod Health . 2017;49(1):17 –27.

18. South SJ, Baumer EP. Community effects on the resolution of adolescent premarital pregnancy. JFam Issues . 2001;22(8):1025 –1043.

19. Smith KE, Glauber R. Exploring the spatial wage penalty for women. Does it matter where you live? Soc Sci Res . 2013;42(5):1390 –1401.

20. Guttmacher Institute. Unintended pregnancy in the United States. September 2016. Available at: https:// www.guttmacher.org/sites/default/ files/factsheet/ fb-unintended-pregnancy-us_0.pdf. Accessed July 26, 2018.

21. Finer LB, Henshaw SK. Abortion incidence and services in the United States, 2000. Perspect Sex Reprod Health . 2003;35(1):6 –15.

22. Bennett T, Skatrud JD, Guild P, Loda F, Klerman LV.

Rural adolescent pregnancy. A view from the South. Fam Plann Perspect . 1997;29(6):256 –260, 267.

23. Tierney KI. Abortion underreporting in Add Health:

fi ndings and implications. Popul Res Policy Rev. 2019; 38(3):417 –428.

24. National Center for Health Statistics. NCHS Rural – Urban Classi fication Scheme for Counties data file doc- umentation. Centers for Disease Control and Prevention.

Available at: https://www.cdc.gov/nchs/data/oae/ nchsurbrural filedocumentation.pdf. Accessed November 10, 2018.

25. Jones RK, Jerman J. How far did US women travel for abortion services in 2008? J Womens Health (Larchmt).

2013;22(8):706 –713.

26. Aiken AR, Scott JG. Family planning policy in the United States: the converging politics of abortion and contraception. Contraception. 2016;93(5):412 –420.

27. Bearak JM, Burke KL, Jones RK. Disparities and change over time in distance women would need to travel to have an abortion in the USA: a spatial analysis. Lancet Public Health . 2017;2(11):e493 –e500.

28. Guttmacher Institute. Targeted Regulation of Abortion Providers (TRAP) laws. February 2018.

Available at: https://www.guttmacher.org/evidence- you-can-use/targeted-regulation-abortion-providers- trap-laws. Accessed December 5, 2018.

29. Aiken AR, Borrero S, Callegari LS, Dehlendorf C.

Rethinking the pregnancy planning paradigm: un- intended conceptions or unrepresentative concepts?

Perspect Sex Reprod Health . 2016;48(3):147 –151.

30. Potter JE, Stevenson AJ, Coleman-Minahan K, et al.

Challenging unintended pregnancy as an indicator of reproductive autonomy. Contraception. 2019;100(1):1 –4.

31. Thiede BC, Sanders SR, Lichter DT. Born poor?

Racial diversity, inequality, and the American pipeline.

Sociol Race Ethn (Thousand Oaks) . 2017;4(2):206–228.

32. Coles MS, Makino KK, Stanwood NL, Dozier A, Klein JD. How are restrictive abortion statutes associated with unintended teen birth? J Adolesc Health. 2010;47(2):

160 –167.

33. Herd P, Higgins J, Sicinski K, Merkurieva I. The implications of unintended pregnancies for mental health in later life. Am J Public Health . 2016;106(3):421–429. 34. Barber JS, Axinn WG, Thornton A. Unwanted childbearing, health, and mother –child relationships.

J Health Soc Behav . 1999;40(3):231–257.

35. Kost K, Lindberg L. Pregnancy intentions, maternal behaviors, and infant health: investigating relationships with new measures and propensity score analysis. De- mography . 2015;52(1):83 –111. AJPH OPEN-THEMED RESEARCH December 2019, Vol 109, No. 12 AJPH Sutton et al.Peer Reviewed Research1769 Reproduced with permission of copyright owner. Further reproductionprohibited without permission.