Assessing the impact of two statewide family planning initiatives on birth rates: outcomes from the Colorado Family Planning Initiative and the Iowa initiative to reduce unintended pregnancies

(1)

ASSESSING THE IMPACT OF TWO STATEWIDE FAMILY PLANNING INITIATIVES ON BIRTH RATES: OUTCOMES FROM THE COLORADO FAMILY PLANNING INITIATIVE

AND THE IOWA INITIATIVE TO REDUCE UNINTENDED PREGNANCIES by

TARYN QUINLAN B.S., Syracuse University, 2008

A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment

of the requirements for the degree of Master of Science

Health Services Research, Policy and Administration 2018

(2)

This thesis for the Master of Science

Health Services Research, Policy and Administration degree by Taryn Quinlan

has been approved for the Health Systems, Management & Policy

by

Richard Lindrooth, Chair Beth McManus Marcelo Coca Perraillon

(3)

Quinlan, Taryn (MS, Health Services Research, Policy, and Administration)

Assessing the Impact of Two Statewide Family Planning Initiatives on Birth Rates: Outcomes from the Colorado Family Planning Initiative and the Iowa Initiative to Reduce Unintended Pregnancies

Thesis directed by Interim Chair and Professor Richard Lindrooth ABSTRACT

The Title X family planning program has provided low-income women and teens access to family planning services for over 40 years through its network of about 4,000 clinics

nationwide. Expanding access to long-acting reversible contraceptives (LARCs) at Title X clinics could have a measurable impact on the rates of unintended pregnancies across the country. To understand this impact, two states—Iowa and Colorado—received funding from an anonymous donor to address barriers to LARCs via their respective Title X networks. Using county birth rates from the CDC National Center for Health Statistics and county socioeconomic data from the Health Resources and Services Administration Area Resource File, I estimated the impact of the Colorado Family Planning Initiative (CFPI) and the Iowa Initiative to Reduce Unintended Pregnancies on the birth rates of young women and teens. I used propensity-score weighted difference-in-difference models to compare birth rate trends between Iowa counties and counties not exposed to the Iowa Initiative in the pre and post periods; and Colorado

counties and counties not exposed to the CFPI in the pre and post periods. The CFPI resulted in a statistically significant decline of 5.4 births per 1,000 women among aged 15-19. The birth rate among 20-24-year-old women declined by 6.6 births per 1,000 women, but the decrease was not statistically significant. The Iowa Initiative resulted in a small decline in birth rates among 15-19- and 20-24-year-old women (less than 1 birth per 1,000 women), but the decline was not statistically significant in either age group. The varying outcomes of the two initiatives suggests

(4)

that successfully expanding access to LARCs at Title X clinics to reduce unintended pregnancies requires a coordinated effort involving patients, providers, healthcare administrators, legislators, and the community.

The form and content of this abstract are approved. I recommend its publication. Approved: Richard Lindrooth

(5)

TABLE OF CONTENTS

I. INTRODUCTION ... 1

II. BACKGROUND ... 2

Iowa Initiative to Reduce Unintended Pregnancies ... 4

Colorado Family Planning Initiative ... 4

III. LITERATURE REVIEW ... 6

IV. DATA ... 9 V. METHODS ... 10 VI. RESULTS ... 14 Iowa Analysis ... 14 Colorado Analysis ... 15 VII. DISCUSSION ... 22 VIII. CONCLUSION ... 24 REFERENCES ... 25

(6)

CHAPTER I

INTRODUCTION

This paper evaluates the impact of two statewide family planning initiatives on birth rates: the Colorado Family Planning Initiative (CFPI) and the Iowa Initiative to Reduce Unintended Pregnancies. The CFPI provided low-income women with long-acting reversible contraceptives (LARCs) at Title X clinics throughout Colorado from 2009 to 2015. The Iowa Initiative increased access to LARCs to low-income women receiving care at Title X clinics throughout Iowa from 2007-2013. Both states received funding from the same anonymous donor to implement these initiatives. For each state, I perform a propensity-score weighted difference-in-difference analysis comparing birth rates before and after each initiative in the participating state counties with birth rates in other US counties over the same period of time. The results help to understand what might occur if funding were increased to expand access to LARCs at Title X family planning clinics throughout the US.

(7)

CHAPTER II

BACKGROUND

LARC methods—intrauterine devices (IUDs) and subdermal hormonal implants—are the most effective form of reversible contraception (more than 99% effective) and have the highest continuation and satisfaction rates among users. Once correctly inserted, LARC methods are effective for 3 to 12 years (depending on the type) and work without user action. LARCs are an ideal contraceptive option for many young women and teens looking to prevent pregnancy. In fact, in 2012, the American College of Obstetricians and Gynecologists (ACOG) revised its practice guidelines on LARCs to advise that sexually active teens who are at high risk of unintended pregnancy be encouraged to consider LARCs for contraception.1

While LARCs are a safe, effective, reversible contraception option that requires little to no maintenance, many young women and teens are unable to obtain these methods due to barriers such as cost, lack of knowledge or interest, provider interest, and lack of provider experience and training.2_{Cost is a significant barrier for both patients and the clinics providing} care as LARC methods can cost anywhere from $500 to $1000. Kumar and Brown3_summarized existing access barriers to LARC methods for adolescents through a review of US studies published after 2000. Identified barriers included costs for patients and the organizations providing family planning care; consent and confidentiality concerns; providers’ attitudes,

misconceptions and inadequate training; and patients’ lack of awareness. Most likely due to these barriers, only 6% of women using contraception in the US used a LARC method in 2007.4

The low rate of LARC usage is concerning considering the high rate of unintended pregnancy in the US. In 2011, the rate of unintended pregnancy was 45 per 1,000 for women aged 15-44.5_{The rates of unintended pregnancy are disproportionately high among teens, young,} unmarried, uneducated, poor, and minority women.6,7_{The rate of unintended pregnancy among}

(8)

poor women (those with incomes below the federal poverty level) was 112 per 1,000 women in 2011.5_{This rate is more than five times the rate of women with incomes of at least 200% FPL} (20 per 1,000).5_{Women who did not complete high school had the highest rate of unintended} pregnancy among all educational levels in 2011 (73 per 1,000).5_{These are groups of women for} whom an unintended pregnancy can have particularly negative health, social and economic consequences for both mother and child.6,8_{The rate of unintended pregnancy generally decreases} with age. In 2011, the age group to experience the largest unintended pregnancy rate was women aged 20-24 (81 per 1,000 women).5_{LARC methods have the potential to reduce the rates of} unintended pregnancy, especially among low-income women and teens.

In the US, the Title X family planning program has provided low-income, underinsured and uninsured women access to birth control for over 40 years through its network of about 4,000 clinics nationwide. Title X is the only federal program dedicated solely to family planning and related preventive health care services. This national effort provides the structural resources to help providers deliver high-quality contraceptive and related care to the most vulnerable communities. In 2015, clinics funded by Title X served 3.8 million women.9_{These clinics are} located where women need them most: in 2015, 64% of US counties had at least one Title X-supported safety-net family planning center, and 90% of women in need of publicly funded contraception services lived in those counties.10

Expanding access to LARC methods at Title X clinics could have a significant impact on the rates of unintended births across the country. To understand this impact, two US states— Iowa and Colorado—received funding from an anonymous donor to address identified barriers to LARC methods for low-income women and teens via their respective Title X networks. These were the first statewide policy interventions to improve access to LARCs at federally-funded family planning clinics.

(9)

Iowa Initiative to Reduce Unintended Pregnancies

In 2007, the Iowa Initiative to Reduce Unintended Pregnancies was launched with

increased funding for 17 Title X agencies providing family planning services in 81 clinics across the state. The Iowa Initiative used the additional funding support to provide low- or no-cost LARCs, increase capacity at existing clinics, expand hours, add staff and clinic locations, train clinical nurse practitioners and physicians on the benefits and use of LARCs, and increase marketing and outreach efforts. Iowa was considered an ideal state for examining strategies to reduce unintended pregnancy due to an unintended pregnancy rate that closely mirrored national estimates and having collected data on unintended pregnancies for a few decades. In 2006, the rate of unintended pregnancy in Iowa was 42 per 1,000 women aged 15-44 (44% of all

pregnancies).11_{The initiative was carried out in partnership with the Iowa Department of Public} Health, the Family Planning Council of Iowa, Planned Parenthood of the Heartland and

researchers at the University of Northern Iowa, the University of Iowa, and the University of Alabama-Birmingham.

Colorado Family Planning Initiative

The Colorado Department of Public Health and the Environment (CDPHE) received funding to implement the Colorado Family Planning Initiative (CFPI)—an expansion of the state’s Family Planning Program, which is run through family planning clinics that receive operational support through county appropriations, insurance reimbursement, client fees, and Title X. CDPHE launched the CFPI in 2009 using the $27 million anonymous donation to provide training, operational and outreach support, and low- or no-cost LARCs to low-income women statewide. 28 Title-X funded family planning agencies added sites and increased hours, hired additional staff, dedicated health care providers to LARC insertion, upgraded equipment and billing procedures, and conducted outreach efforts at local schools and other community

(10)

organizations. In 2006, the rate of unintended pregnancy in Colorado was 48 per 1,000 women aged 15-44 (48% of all pregnancies).11

The implementation of these family planning initiatives provides a unique opportunity to examine the impact of expanding access to LARCs at Title X clinics on unintended births among young women and teens. Lindo and Packham12_{have established the impact of the CFPI on teen} birth rates (see Literature Review), but there is no published evidence on the impact of the CFPI on young women aged 20 to 24, and no evidence of the impact of the Iowa Initiative on birth rates among teens and young women. This paper seeks to fill this evidence gap to present a broader picture of how a more fully funded Title X network could impact the rates of unintended pregnancy across the US.

(11)

CHAPTER III

LITERATURE REVIEW

The barriers to LARC methods are well documented,13-15_{and removing these barriers can} increase LARC uptake among women in need of contraceptive care. Harper et al.16_{conducted a} cluster randomized trial to assess the impact of an educational intervention to increase patients’ access to LARCs on pregnancy rates among 18 to 25-year old women receiving family planning or care after an abortion. Forty reproductive health clinics across the US were involved in the intervention between 2011-2013. Twenty clinics were randomly assigned to receive evidence-based training on providing counseling and insertion of LARC methods and 20 to provide standard care. The usual costs for contraception were maintained at all sites to specifically focus on provider-driven barriers to LARC access. The rate of unintended pregnancy was substantially reduced (by nearly half) among women who attended family planning visits, although not among those who attended abortion care visits. At the control clinics, fewer than half of women reported receiving LARC counseling. Thirty-eight percent of women in the study reported they had no insurance, which highlights the importance of providing funding coverage for the high up-front costs of LARC methods in low-income populations. Despite the strong study design—the cluster randomized designed avoids potential contamination effects of the educational intervention among providers—the study focused only on women looking prevent to pregnancy after an abortion, which limits generalizability of the results.

The Contraceptive CHOICE Project (CHOICE) showed the positive impact of removing the cost barrier on LARC uptake. CHOICE aimed to reduce unintended pregnancies in the St. Louis area by removing cost, education and access barriers to LARC methods. Conducted between 2007 and 2011, a cohort of over 9,000 women 14-45 years of age received tiered

(12)

contraceptive counseling and were offered birth control of their choice at no cost for 2-3 years.17 The study found that LARC methods were 20 times more effective than non-LARC methods and 75 percent of study participants chose LARCs as their method of choice. The overall 12-month continuation rate of the LARC methods was 86 percent. The researchers projected that if the CHOICE model were adopted among all sexually active teens in the US, the pregnancy rate of 67.8 per 1,000 women (in 2008) could be reduced to 29.6.18,19_{This projection was confirmed in a} recent paper showing reductions in teenage pregnancy and abortion rates in England as LARC usage increases.20

Frost et al.21_{estimated the number of unintended pregnancies prevented by all US}

publicly-funded family planning clinics in 2004, nationally and on the state level. They estimated that women who received care from family planning clinics were able to avoid over 1.4 million unplanned pregnancies. Frost et al. based their expectations of contraceptive use in the absence of access to publicly funded family planning services on the behavior of similar women but did not apply an experimental design. Their basic methodology is similar to that of prior cost-benefit analyses looking at the impact of publicly-funded family planning clinics on unintended births. 22-24_{These results provide further evidence that public investment in family planning services} yields significant benefits for women. However, the evidence on improving access to LARC methods at publicly-funded family planning clinics is still emerging.

Ricketts et al.25_{assessed the effectiveness of the CFPI on both the program and the} population levels. They examined the impact of the CFPI on fertility rates; however, their analysis was descriptive. Differences in fertility rates among low-income women in Colorado counties with Title X-funded clinics were compared to low-income women in counties without Title X-funded clinics before and after the intervention by computing z tests for proportional differences. Expected fertility rates for 2010 and 2011 were calculated from linear trend lines.

(13)

Statistically significant differences between observed and expected birth rate trends were observed in women aged 15-19, 20-24, and 15-24. Goldthwaite et al.26_{evaluated the CFPI on} adverse birth outcomes. They found that better access to family planning services and increased use of LARCs resulted in a lower risk of preterm birth but had no effect on low birth weight.

Lindo and Packham12_{used a county-level difference-in-difference approach to assess the} impact of the CFPI on teen birth rates. They included only those Colorado counties with a Title X clinic and compared to all counties outside Colorado with Title X clinics. Their analysis found that the CFPI reduced the teen birth rate by 6.4% over 5 years with the effects of the program concentrated in the second through fifth years and in Colorado counties with high poverty rates. Their models included county-specific linear trends to address concerns of differences in

preexisting trends between treatment and control group counties; however, the linear trends likely underestimated the effect of the CFPI.

There is one published paper on the impact of the Iowa Initiative on the rate of abortions. Biggs et al.27_{conducted a longitudinal analysis of family planning visit and vital statistics data} controlling for population density, the number of abortion facilities and change in percentage of people living in poverty. The study found that an increase of one new LARC user per 100 women was associated with a 4 percent reduction in abortions each year.

The literature establishes the positive impact of improving access to LARC methods for low-income women and teens. A causal evaluation of the CFPI and Iowa Initiative will expand the evidence base on the impact of increasing funding for and access to LARCs at Title X family planning clinics. My hypotheses are that the Iowa Initiative to Reduce Unintended Pregnancies and the Colorado Family Planning Initiative reduced the birth rate among women aged 15-19 and 20-24 women in their respective states.

(14)

CHAPTER IV

DATA

The primary outcome of my analyses is the birth rate per 1,000 women aged 15-19 and 20-24. The birth rates for all US counties were obtained from the CDC National Center for Health Statistics. These data were supplemented with county socioeconomic information, including demographics, educational attainment, income, and unemployment rates from the Health Resources and Services Administration Area Resource File. For the Iowa analysis, the sample period included two years of pre-initiative data (2005-2006) and six years of post-initiative data (2008-2013). The Colorado analysis included four years of pre-CFPI data (2005-2008) and five years of post-CFPI data (2010-2014).

(15)

CHAPTER V

METHODS

To test my hypotheses, I use propensity-score weighted difference-in-difference specifications to compare birth rate trends between Iowa counties and counties not exposed to the Iowa Initiative in the pre and post periods; and Colorado counties and counties not exposed to the CFPI in the pre and post periods. A variable indicating whether observations are in the post period is interacted with a variable indicating the treatment group. An association between the initiatives and birth rates is evident if this interaction term is significantly different from zero.

Counties were included in the comparison groups when they closely matched treatment counties on key demographic characteristics. At the outset, Missouri counties were excluded due to a similar family planning intervention during the same time period as the Iowa and Colorado initiatives. I excluded Colorado from the Iowa analysis and vice versa. I identified counties to include in my comparison groups when they closely matched my treatment counties on the following variables: total population; percent of population that was female in each age range of interest (15-19 and 20-24); percent of females in each range by ethnicity; percent of civilian females in the labor force; and unemployment rate. I identified the minimum and maximum values for each of the aforementioned variables in Iowa/Colorado counties in the pre-periods and only included unexposed counties in my comparison groups that had values between the

minimum and maximum of Iowa/Colorado counties.

The summary statistics for the variables used in my county-level analyses of birth rates are shown in Tables 1 and 2. Table 1 shows the weighted means for key demographic statistics of Iowa counties and counties outside of Iowa included in my comparison group in the pre-period. All 99 counties in Iowa were included in my treatment group. I identified 356 counties similar to Iowa to include in my comparison group. Table 2 shows the weighted means for key

(16)

demographic statistics of Colorado counties and counties outside of Colorado included in my comparison group in the pre-period. All 64 counties in Colorado were included in my treatment group. I identified 79 counties similar to Colorado to include in my comparison group.

Propensity score weights were used to further control for selection bias and give more importance to those counties most similar to the treatment counties. With propensity scores, the resulting treatment and control groups will have similar characteristics to those created through random assignment. The propensity score is the probability a county is in the treatment group (Colorado or Iowa) before their respective initiatives were launched controlling for observed county demographic and economic characteristics. Average treatment effect on the treated (ATET) weighting keeps all observations in the dataset and weights them according to the propensity score. Observations that received treatment are given a weight of 1 and those that did not receive treatment are given a weight of p/1-p. County demographic and economic variables included in my propensity score calculation included: percent of female civilians in the labor force, percent population with less than high school education, median family income, rural and urban indicator variables, unemployment rate, percent females Medicaid eligible, percent population foreign born, and percent females in each range (15-19 or 20-24) by ethnicity.

I use a difference-in-difference approach to estimate the effect of the Iowa and Colorado initiatives by comparing the changes in outcomes over time between the treatment counties (Iowa and Colorado) and my comparison counties. To ensure internal validity of this approach, there must not be any underlying time-dependent trends in birth rates unrelated to the family planning initiatives. The two main assumptions of the difference-in-difference approach are parallel trends and common shocks. The parallel trends assumption requires that birth rate trends between the treated and comparison groups be the same prior to the intervention. I validate this assumption usual visual inspection of difference-in-difference graphs and with a regression

(17)

model to assess the significance of the interaction term between time and treatment exposure in the pre-period. The common shocks assumption states that any events occurring over the time period of analysis will equally affect the treatment and comparison groups.

While the difference-in-difference approach controls for temporal trends, I use county-level fixed effects to control for all time-invariant factors that may differ between treatment and comparison group counties. Fixed effects can account for those factors related to pregnancy outcomes that are not observed in my dataset, such as religiosity, welfare policies and income inequality. My primary outcome is based on the following fixed effects model:

Birth Rate = β0 + β1Postt + β2Postt*Treati + γXit + αi + εit (Equation 1)

The time-varying factors that I adjust for in my specifications include those variables related to pregnancy rates: unemployment rate; median family income; female civilians in the labor force; education; percent of females in each age range (15-19 and 20-24) by ethnicity; and total

(18)

Table 1: Summary Statistics—Weighted Means of Iowa and Comparison Group Counties

Iowa Counties (n=792) Comparison Counties (n=2848)

Weighted Standardized

Differences

Unemployment Rate 5.12 5.70 -0.313

Percent Female Civilians in Labor Force 5 year

24.93 24.53 0.233

Percent Population with Less than HS Education

10.47 10.31 0.047

Median Family Income 58,631 59,143 -0.058

Percent Females aged 15-19 3.34 3.26 0.182

Percent White 92.17 91.65 0.085

Percent Black 1.44 1.32 0.071

Percent Hispanic 4.81 4.71 0.022

Percent Asian 1.00 0.77 0.234

Percent NHPI 0.06 0.06 -0.009

Percent White 91.04 90.63 0.067 Percent Black 1.37 1.26 0.071 Percent Hispanic 4.87 4.38 0.108 Percent Asian 1.22 0.92 0.228 Percent NHPI 0.08 0.06 0.124 Total Population 30,282 40,425 -0.191

Table 2: Summary Statistics—Weighted Means of Colorado and Comparison Group Counties

Colorado Counties (n=576) Comparison Counties (n=711) Weighted Standardized Differences Unemployment Rate 5.99 5.59 0.159

Percent Female Civilians in Labor Force 5yr

23.43 21.43 0.573

Percent Population with Less than HS Education

11.59 13.81 -0.357

Median Family Income 61,669 58,998 0.124

Percent Females aged 15-19 2.95 3.10 -0.244

Percent White 71.34 69.60 0.097

Percent Black 1.33 1.51 -0.071

Percent Hispanic 23.78 25.01 -0.073

Percent Asian 1.13 0.83 0.249

Percent NHPI 0.12 0.10 0.076

Percent White 71.39 71.39 0.000 Percent Black 1.20 1.53 -0.146 Percent Hispanic 23.15 22.32 0.051 Percent Asian 1.23 0.92 0.237 Percent NHPI 0.14 0.12 0.053 Total Population 77,524 58,301 0.142

(19)

CHAPTER VI

RESULTS

Iowa Analysis

Before presenting model-based estimates, I present the balance and overlap analyses from my propensity scores and graphical analyses corresponding to my difference-in-difference

approach. I constructed propensity scores for each age group of women (15-19 and 20-24) starting with a parsimonious specification and adding interaction and quadratic terms of continuous variables until I identified the specification with the best overlap (Figures 1 and 2) and smallest standardized differences between Iowa and control counties. The inverse probability weights of each county were constructed using 2006 values. My propensity scores passed

overidentification tests of covariate balance, but I did not achieve acceptable ranges of

standardized differences on some covariates. Regression adjustment was used to control for the imbalance of covariates in my fixed effects models.

In order for my difference-in-difference approach to be valid, the birth rate trends among Iowa counties and my comparison group counties must be parallel prior to implementation of the initiative. The unadjusted trends in birth rates among Iowa and comparison counties can be found in Figures 3 and 4. Figures 5 and 6 plot the propensity-score weighted adjusted trends, which visually confirm the parallel trends assumption. I confirmed this visual inspection with a regression model by interacting time and treatment exposure in the pre-intervention period.

Tables 3 and 4 present model-based estimates based on the fixed effects model described in equation 1. Column 1 shoes the estimated effects from the unweighted, unadjusted baseline model. These estimates indicate that the Iowa Initiative reduced the birth rate among 15-19-year-old women by 0.2 births per 1,000 women; and by 1.7 births per 1,000 women among 20-24-year-old women. Column 2 shows unweighted estimates after controlling for time-varying

(20)

factors that may affect birth rates in Iowa and control counties. The unweighted estimates indicate that the Iowa Initiative reduced the birth rate per 1,000 women by 0.2 births among 15-19-year-old women, and 1.6 births among 20-24-year-old women. Finally, Column 3 shows the estimates for my propensity-score weighted, adjusted fixed effects models. These estimates indicate that the Iowa Initiative reduced the birth rate among 15-19-year-old women by 0.7 births per 1,000 women, and by 0.6 births per 1,000 women among 20-24-year-old women. None of the fixed effects estimates were statistically significant for either age group of women.

Colorado Analysis

I followed the same approach to my propensity score and difference-in-difference specification for Colorado as with my Iowa analysis. Figures 7 and 8 show the overlap of my propensity scores for each age group of women. The inverse probability weights of each county were constructed using the 2008 values. These propensity scores passed statistical tests for covariate balance; however, I did not achieve acceptable ranges of standardized differences on some covariates. As with the Iowa analysis, regression adjustment was used to control for the imbalance of covariates.

Figures 9 and 10 show the unadjusted trends in birth rates in both age groups of women, and Figures 11 and 12 show the weighted, adjusted trends. The propensity-score weighted

adjusted trends visually confirm the parallel trends assumption. I confirmed this visual inspection with a regression model by interacting time and treatment exposure in the pre-intervention

period.

Tables 5 and 6 present model-based estimates based on the fixed effects model described in equation 1. Column 1 shoes the estimated effects from the unweighted, unadjusted baseline model. These estimates indicate that the CFPI reduced the birth rate among 15-19-year-old women by 3.6 births per 1,000 women; and by 2.3 births per 1,000 women among

(21)

20-24-year-old women. Column 2 shows unweighted estimates after controlling for time-varying factors that may affect birth rates in Colorado and control counties. The unweighted estimates indicate that the CFPI reduced the birth rate per 1,000 women by 5.4 births among 15-19-year-old women, and 3.8 births among 20-24-year-old women. Finally, Column 3 shows the estimates for my propensity-score weighted, adjusted fixed effects models. These estimates indicate that the CFPI reduced the birth rate among 15-19-year-old women by 5.4 births per 1,000 women, and by 6.6 births per 1,000 women among 20-24-year-old women.

(22)

Figure 1. Overlap of Propensity Score for Iowa versus Comparison Counties: Birth Rate among Women Aged 15-19

Figure 2. Overlap of Propensity Score for Iowa versus Comparison Counties: Birth Rate among Women Aged 20-24

Figure 3. Unadjusted Trends in Birth Rate per 1,000 Women Aged 15-19 in Iowa versus Comparison Counties Notes: Produced using Table 3 Column 1 estimates.

Figure 4. Unadjusted Trends in Birth Rate per 1,000 Women Aged 20-24 in Iowa versus Comparison Counties

Notes: Produced using Table 4 Column 1 estimates.

0 2 4 6 d e n si ty .4 .6 .8 1

Propensity score, IA=0 IA=0 IA=1 0 1 2 3 4 5 d e n si ty .2 .4 .6 .8 1

Propensity score, IA=0 IA=0 IA=1 Pre-Intervention Washout Period Post-Intervention 15 20 25 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e d 1 5 -1 9 2004 2006 2008 2010 2012 2014 Year Iowa Control

Unadjusted Trends in Birth Rate, Women aged 15-19

Pre-Intervention Washout Period Post-Intervention 0 20 40 60 80 100 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e d 2 0 -2 4 2004 2006 2008 2010 2012 2014 Year Iowa Control

(23)

Figure 5. Propensity Score Weighted Adjusted Trends in Birth Rate per 1,000 Women Aged 15-19 in Iowa versus Comparison Counties

Figure 6. Propensity Score Weighted Adjusted Trends in Birth Rate per 1,000 Women Aged 20-24 in Iowa versus Comparison Counties

Notes: Produced using Table 4 Column 3 estimates.

Table 3. Fixed Effects Estimates of the Effect of the Iowa Initiative on Birth Rate per 1,000 Women Aged 15-19, Difference-in-Difference Using Counties Outside of Iowa for

Comparison

(1) (2) (3)

Effect of the Iowa Initiative in the Post Period

p-value (test average effect in post period = 0) -0.228 (0.980) 0.816 -0.183 (0.973) 0.851 -0.684 (0.943) 0.469 Counties Observations 455 3,640 455 3,640 455 3,640

Propensity Score Weighted No No Yes

Adjusted for Time-varying Factors

No Yes Yes

Notes: Time-varying county-level factors adjusted for in Columns (2) and (3) include median family income,

unemployment rate, percent of civilian females in the labor force, percent of population with less than high school education, percent females aged 15-19 by ethnicity, and total population. Robust standard errors clustered at the county level are shown in parentheses.

15 20 25 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e d 1 5 -1 9 2004 2006 2008 2010 2012 2014 Year

Iowa without Intervention Iowa with Intervention Control

Adjusted Trends in Birth Rate, Women Aged 15-19

65 70 75 80 85 90 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e d 2 0 -2 4 2004 2006 2008 2010 2012 2014 Year

Iowa without Intervention Iowa with Intervention Control

(24)

Table 4. Fixed Effects Estimates of the Effect of the Iowa Initiative on Birth Rate per 1,000 Women Aged 20-24, Difference-in-Difference Using Counties Outside of Iowa for

Comparison

(1) (2) (3)

Effect of the Iowa Initiative in the Post Period

No Yes Yes

Figure 7. Overlap of Propensity Score for Colorado versus Comparison Counties: Birth Rate among women aged 15-19

Figure 8. Overlap of Propensity Score for Colorado versus Comparison Counties: Birth Rate among women aged 20-24

0 .5 1 1.5 2 2.5 d e n si ty 0 .2 .4 .6 .8 1

Propensity score, co=0

co=0 co=1 0 .5 1 1.5 2 2.5 d e n si ty 0 .2 .4 .6 .8 1

Propensity score, co=0

(25)

Figure 9. Unadjusted Trends in Birth Rate per 1,000 Women aged 15-19 in Colorado versus Comparison Counties

Figure 10. Unadjusted Trends in Birth Rate per 1,000 Women aged 20-24 in Colorado versus Comparison Counties

Figure 11. Propensity Score Weighted Adjusted Trends in Birth Rate per 1,000 Women Aged 15-19 in Colorado versus Comparison Counties

Figure 12. Propensity Score Weighted Adjusted Trends in Birth Rate per 1,000 Women aged 20-24 in Colorado versus Comparison Counties Notes: Produced using Table 6 Column 3 estimates.

Pre-CFPI Washout Period Post-CFPI 10 15 20 25 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e 1 5 -1 9 2004 2006 2008 2010 2012 2014 Year Colorado Control Unadjusted Pre-CFPI Washout Period Post-CFPI 0 20 40 60 80 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e 2 0 -2 4 2004 2006 2008 2010 2012 2014 Year Colorado Control Unadjusted 10 15 20 25 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e 1 5 -1 9 2004 2006 2008 2010 2012 2014 Year

Colorado without CFPI Colorado with CFPI Control

Adjusted Trends in Birth Rate, Women Age 15-19

Pre-CFPI Washout Period Post-CFPI 40 50 60 70 80 Me a n Bi rt h R a te p e r 1 0 0 0 W o me n Ag e 2 0 -2 4 2004 2006 2008 2010 2012 2014 Year

Colorado without CFPI Colorado with CFPI Control

(26)

Table 5. Fixed Effects Estimates of the Effect of the CFPI on Birth Rate per 1,000 Women Aged 15-19, Difference-in-Difference Using Counties Outside of Colorado for Comparison

(1) (2) (3)

Effect of the CFPI in the Post Period

No Yes Yes

Table 6. Fixed Effects Estimates of the Effect of the CFPI on Birth Rate per 1,000 Women Aged 20-24, Difference-in-Difference Using Counties Outside of Colorado for Comparison

(1) (2) (3)

Effect of the CFPI in the Post Period

No Yes Yes

(27)

CHAPTER VII

DISCUSSION

The results of my analysis confirm prior findings by Lindo and Packham12_{that the CFPI} resulted in a significant decline in the birth rate among women aged 15-19. The birth rate among women aged 20-24 also declined, but the results were not statistically significant. My analysis of the Iowa Initiative found the initiative resulted in a small decline in the birth rates among 15-19- and 20-24-year-old women (less than 1 birth per 1,000 women), but the decline was not

statistically significant in either age group. My model estimates were based on births to all women, which may understate the effects on women who use Title X clinics and women receiving LARCs through the Colorado and Iowa initiatives. Additionally, there are likely spillover effects from the Colorado and Iowa initiatives. Both interventions had strong educational components and during the time of the programs there was a general increased awareness and usage of LARC methods. The spillover is likely to be higher in Iowa and

Colorado due to the promotion of LARCs through peer networks of women using that method of birth control. Spillovers generally bias the results toward no effect.

A limitation of my approach is that I only had county-level data to calculate my primary outcome. Consequently, the estimates from my difference-in-difference specifications combined the effects of the Colorado and Iowa initiatives and any other factors impacting birth rates in Colorado/Iowa that occurred at the same time. Causal interpretation of propensity-score

weighted estimators relies on the Conditional Independence Assumption and balance/overlap of the treatment and control groups. Outcomes are independent of assignment to treatment

conditional on pretreatment covariates implying that all variables that influence treatment assignment and potential outcomes must be observed. My dataset included most of the variables related to treatment and outcomes; however, I did not have variables on pregnancy intention,

(28)

relationship status, religion, income inequality, or welfare policies. An incorrect propensity score specification will cause bias, which could go in either direction, depending on the imbalance of covariate distributions and the unmeasured confounding.

Variation in demographic and socioeconomic factors are likely contributors to differences in the impact of the initiatives in Iowa and Colorado. Existing funding, infrastructure, availability of family planning services, and state policies (e.g., sex education, contraception coverage) are also important factors. Data from the 2013 Title X Family Planning Annual Report shows that Colorado was more successful than Iowa in getting women to choose LARCs as their

contraceptive method: 21.3% of Colorado Title X female patients were using a LARC method at exit from encounter versus 14.7% of Iowa patients.28_{Since the CFPI resulted in a greater decline} in birth rates compared to the Iowa Initiative, future research efforts should seek to identify the differences between Iowa and Colorado that caused the CFPI to be more successful.

(29)

CHAPTER VIII

CONCLUSION

The Title X family planning program advances the health and wellbeing of women and their children by supporting low-income women and teens in family planning. Women have reported that contraception allows them to better care for themselves and their families, complete their education, and support themselves financially.29_{Accessible and highly effective}

contraception, such as LARC methods, is essential for women of childbearing age. However, the challenges to increase LARC uptake are many and include decreasing costs, insuring easy access to training, promoting increased patient and provider knowledge, and encouraging patient

interest.

The varying outcomes of the Colorado Family Planning Initiative and the Iowa Initiative to Reduce Unintended Pregnancy suggest that successfully expanding access to LARCs at Title X family planning clinics requires a coordinated effort involving patients, providers, healthcare administrators, legislators, and the community. The CFPI should be further examined to

(30)

REFERENCES

1. Adolescents and Long-Acting Reversible Contraception: Implants and Intrauterine Devices [press release]. ACOG, October 2012.

2. Long-Acting Reversible Contraception: Implants and Intrauterine devices [press release]. ACOG, October 2015.

3. Kumar B, Brown JD. Access barriers to long-acting reversible contraceptives for adolescents. Journal of Adolescent Health. 2016;59(3):248-253.

4. Kavanaugh ML, Jerman J, Finer LB. Changes in use of long-acting reversible contraceptive methods among U.S. women, 2009–2012. Obstetrics & Gynecology. 2015;126(5):917-927.

5. Finer LB, Zolna MR. Declines in unintended pregnancy in the United States, 2008–2011. New England Journal of Medicine. 2016;374(9):843-852.

6. Gipson J, Koenig M, Hindin M. The effects of unintended pregnancy on infant, child, and parental health: a review of the literature. Studies in Family Planning. 2008;39:18-38. 7. Finer LB, Zolna MR. Unintended pregnancy in the United States: incidence and

disparities, 2006. Contraception. 2011;84(5):478-485.

8. Brown S, Eisenberg L. The Best Intentions: Unintended Pregnancy and the Well-Being of Children and Families. Washington, DC1995.

9. Frost JJ, Frohwirth LF, Blades N, et al. Publicly funded contraceptive services at U.S. Clinics, 2015. New York Guttmacher Instititute April 2017 2017.

10. Frost JJ, Zolna MR. Response to inquiry concerning the availability of publicly funded contraceptive care to U.S. women, memo to U.S. Senator Patty Murray, Senate Health, Education, Labor and Pensions Committee, New York. In: Guttmacher Institute; 2017. 11. Finer LB, Kost K. Unintended pregnancy rates at the state level. Perspectives on Sexual

and Reproductive Health. 2011;43(2):78-87.

12. Lindo JM, Packham A. How much can expanding access to long-acting reversible contraceptives reduce teen births rates? . American Economic Journal: Economic Policy 2017;9(3):348-376.

13. Pritt NM, Norris AH, Berlan ED. Barriers and Facilitators to Adolescents' Use of Long-Acting Reversible Contraceptives. J Pediatr Adolesc Gynecol. 2017;30(1):18-22.

14. Beeson T, Wood S, Bruen B, Goldberg DB, et al. Accessibility of long-acting reversible contraceptives (LARCs) in Federally Qualified Health Centers (FQHCs). Contraception. 2013;89(2):91-96.

(31)

16. Harper C, Rocca C, et al. Reductions in pregnancy rates in the USA with long-acting reversible contraception: a cluster randomized trial. Lancet. 2015;386:562-568. 17. Birgisson NE, Zhao Q, et al. . Preventing unintended pregnancy: The contraceptive

CHOICE project in review. J Womens Health (Larchmt). 2015;24(5):348-353.

18. Secura GM, Allsworth J, Madden T, et al. The Contraceptive CHOICE Project: reducing barriers to long-acting reversible contraception. American Journal of Obstetrics & Gynecology. 2010;203(2):115.e111-117.

19. Mestad R, Secura G, Allsworth JE. Acceptance of long-acting reversible contraceptive methods by adolescent participants in the Contraceptive CHOICE Project. Contraception. 2011;84:493-498.

20. Connolly A, Pietri G, Yu J, et al. Association between long-acting reversible contraceptive use, teenage pregnancy, and abortion rates in England. Int J Women’s Health. 2014;6:961-974.

21. Frost JJ, Finer LB, Tapales A. The impact of publicly funded family planning clinic services on unintended pregnancies and government cost savings. Journal of Health Care for the Poor and Underserved. 2008;19(3):778-796.

22. Chamie M, Henshaw SK. The costs and benefits of government expenditures for family planning programs. Fam Plann Perspect. 1981;13(3):117-118; 120-124.

23. Forrest JD, Singh S. Public-sector savings resulting from expenditures for contraceptive services. Fam Plann Perspect. 1990;22(1):6-15.

24. Forrest JD, Samara R. Impact of publicly funded contraceptive services on unintended pregnancies and implications for Medicaid expenditures. Fam Plann Perspect.

1996;28(5):188-195.

25. Ricketts S, Klingler G, Schwalberg R. Game change in Colorado: widespread use of long-acting reversible contraceptives and rapid decline in births among young, low-income women. Perspectives on Sexual and Reproductive Health. 2014;46(3):125-132. 26. Goldthwaite LM, Duca L, Johnson RK, et al. Adverse birth outcomes in Colorado:

assessing the impact of a statewide initiative to prevent unintended pregnancy. American Journal of Public Health. 2015;105(9):e60-66.

27. Biggs MA, Rocca CH, Brindis CD, et al. Did increasing use of highly effective

contraceptive contribute to declining abortions in Iowa? Contraception. 2015;91(2):167-173.

28. Office of Population Affairs. Title X Family Planning Annual Report: 2013 National Summary. Research Triangle Park, NC: US Department of Health and Human Services; November 2013.

29. Frost JJ, Lindberg LD. Reasons for using contraception: perspectives of US women seeking care at specialized family planning clinics. Contraception. 2012;87(4):465-472.