Abstract
Objective:
To evaluate whether part-year or year-round uninsurance is associated with reduced likelihood of using prescription contraception methods rather than using nonprescription methods or using no contraceptive methods.
Methods:
We identified nonpregnant and sexually active female respondents participating in the National Longitudinal Survey of Youth, 1997 cohort between 2007 and 2019. At each interview, we classified the contraceptive method used most frequently as prescription, nonprescription, or none, and used mixed-effects multinomial logistic regression to predict contraceptive method based on health insurance coverage over the past year (classified as continuous private, continuous public, part-year uninsured, or year-round uninsured).
Results:
Our sample included 3,738 respondents and 18,678 observations (person-years). In the most recent interview, 35% of respondents used prescription contraception, 16% used nonprescription methods only, and 49% used no method. On multivariable analysis using all available years of data, respondents with part-year uninsurance were 20% less likely to use prescription rather than nonprescription methods, as compared to respondents with continuous private insurance (95% confidence interval: −31%, −6%; p = 0.007), but did not differ on the likelihood of using prescription methods rather than no method.
Conclusions:
Part-year uninsurance was associated with lower use of prescription contraceptive methods rather than nonprescription methods when compared with continuous private insurance coverage. Use of prescription contraceptives was lowest among people with year-round uninsurance. Policy efforts ensuring continuous insurance coverage with greater flexibility of eligibility and enrollment periods may promote greater access to prescription contraceptives.
Introduction
Contraception is widely recognized for its role in reducing unintended pregnancy and expanding reproductive autonomy. 1 There is ample data that demonstrates that comprehensive contraceptive access has empowered women to participate in the labor force if they so choose, continue their education, and prioritize their health and well-being. 2 Despite improvements in insurance coverage and access to contraception over the past decade, nearly half of all pregnancies in the United States are unintended. 3 Prior studies have evaluated factors such as cost, insurance coverage, access to a usual source of care, and proximity to health care, and how these elements influence contraceptive access and selection. 4 –9 However, further investigation into barriers impacting contraceptive access and choice is essential in the continuing effort to expand reproductive health care and promote reproductive justice.
Although consistent contraceptive use can reduce the likelihood of unintended pregnancy, failure rates of individual methods vary. Nonprescription methods of contraception—including spermicides, the male and female condom, cervical cap, sponge, and diaphragm—are inexpensive, widely accessible, and easy to use. However, these methods have high failure rates with typical use, ranging from 12% to 28%. 10 Prescription contraceptives are more difficult to access but have lower failure rates compared with nonprescription methods. The failure rates of the oral contraceptive pill, injectable contraceptive, vaginal ring, and transdermal patch range from 6% to 9% with typical use. 10 Intrauterine devices (IUD) and subdermal contraceptive implants—known generally as long-acting, reversible contraceptives (LARC)—are prescription contraceptives that have a much lower failure rate (<1%), are quickly reversible, and require no effort on the part of the patient after placement. 10,11 However, access to these methods can be cost-prohibitive, and both insertion and removal require considerable time and effort on the part of the patient. 10
Concerns surrounding cost and access may partly explain the choice to use no method or a nonprescription method of contraception. For uninsured and underinsured women, out-of-pocket costs may represent a significant barrier to obtaining prescription contraception, especially LARCs. 6,12 Furthermore, the availability of certain contraceptive methods may be limited by what insurers are willing to cover. 12,13 Prior research has focused on contemporaneous type of insurance coverage and choice of contraceptive method. 4,5 In one study, women with Medicaid had significantly higher odds of receiving any prescription contraceptive when compared to women with commercial insurance, and women with self-pay or other insurance had significantly lower odds of receiving LARC when compared to women with commercial insurance. 5 However, the impact of past coverage gaps on contraceptive access and choice remains unknown.
A lapse in insurance coverage may impact women needing a prescription renewal, or may make a prescription method unaffordable altogether. Specifically, methods of contraception requiring refill or redosing (pills, injectable contraceptive, vaginal ring, and transdermal patch) may become temporarily unaffordable, and because consistent use is key to their efficacy, inability to refill would render them ineffective. For patients using LARC, gaps in insurance coverage may interfere with discontinuation when desired, because of the associated office fee. In both cases, reproductive autonomy is adversely affected by coverage instability, which may discourage women from using prescription contraception methods, even after regaining coverage. In a recent study, young adults experiencing part-year uninsurance reported subsequently reduced utilization of preventive care services up to 4–6 years later, although that study did not specifically examine reproductive health care. 14 In this study, we used data from a longitudinal birth cohort study to evaluate whether experiencing part-year uninsurance (as compared to continuous coverage by private or public insurance) is associated with a reduced likelihood of using prescription contraception methods rather than using nonprescription methods or using no contraceptive method at all.
Methods
Study sample
This study used deidentified, publicly available data from the National Longitudinal Survey of Youth, 1997 cohort (NLSY97) and is not considered human subjects research. 15 The NLSY97 enrolled a nationally representative cohort of Americans born between 1980 and 1984. Subjects were first interviewed in 1997 and followed annually through 2011 when follow-ups switched to a biennial schedule during odd-numbered years. Surveys were conducted using in-person or telephone interviews, with questions about sexual activity and contraception completed via confidential, self-administered questionnaires. 16 For this study, we included female respondents who contributed one or more interviews between 2007 (when detailed questions about insurance coverage gaps were added to the survey) and 2019, the latest year of available data. Respondents ages ranged from 23 to 27 in 2007 and 35 to 39 in 2019. Interviews were excluded if the respondent was currently pregnant or had not been sexually active within the past 12 months (or since the last interview, for data collected before 2013). Between 2013 and 2015, participants were asked about the gender identity of their sexual partners, and interviews from these years were excluded if the participant reported having exclusively female sexual partners in the past 12 months. Each interview meeting these inclusion criteria and having complete data on study variables was included in our analysis as a separate person-year observation.
Measures
The primary outcome assessed in this study was the contraceptive method, classified into mutually exclusive groups of prescription methods, nonprescription methods, or none. Respondents reporting any form of birth control were coded as using prescription contraception if they reported their most frequently used method to be oral contraceptive pills, the injectable contraceptive, vaginal ring, transdermal patch, or a LARC method such as an IUD or subdermal implant. Respondents were coded as using nonprescription contraception if they reported their most frequently used method of birth control to be spermicide, the cap, sponge, diaphragm, or other barrier method, or the morning-after pill. The use of condoms was not included in the question about the most frequent method of birth control, but condom use was ascertained in a separate question about how frequently respondents had used condoms over the past 12 months (recoded to any vs. none). Therefore, respondents who reported any condom use, and did not report using a prescription contraceptive as their most frequently used method of birth control, were also included in the nonprescription group. Lastly, respondents who reported never using birth control (in a separate question) or respondents who did not use condoms and reported anything other than the prescription or non-prescription methods listed above as their most frequently used method of birth control (e.g., withdrawal, fertility awareness, or no other method) were classified as using no method of contraception. Data on this outcome variable were considered missing if the respondent reported that their most commonly used contraceptive method was tubal ligation or a vasectomy.
The primary independent variable was respondent health insurance coverage over the past year, classified as continuous private (employer-sponsored, privately purchased, or marketplace plans), continuous public (Medicaid or Medicare), part-year uninsured, or year-round uninsured. 14 Respondents were initially asked if they had insurance coverage, and if so, what kind of insurance coverage they currently had. Respondents without current coverage were asked if they had insurance at any point over the past 12 months. Respondents with current coverage were asked if they had a gap in coverage at any point over the past 12 months. Respondents were classified as having continuous private coverage if they currently had private insurance and had no gaps in coverage over the past 12 months. Similarly, respondents were classified as having continuous public coverage if they had public insurance at the time of the interview, with no gaps in coverage in the past 12 months. Respondents with current coverage, but a gap within the past 12 months, and respondents without current coverage, but with a history of recent coverage, were classified as experiencing part-year uninsurance. Lastly, respondents without current coverage and without coverage over the past 12 months were classified as year-round uninsured.
Demographic covariates for each person-year observation included age at the time of interview, race, and ethnicity (non-Hispanic White; non-Hispanic Black; Hispanic or Latino; or none of the above), and residence in a rural versus urban or suburban area (derived by NLSY97 staff from respondents’ address). Our analysis also controlled for socioeconomic characteristics including highest degree received (high school/equivalent or less; some college or a 2-year degree; or 4-year college degree), and respondents’ employment status at the beginning of the month in which they were interviewed (employed, unemployed, or out of the labor force). 17 We also included controls for parenthood and partnership-related measures to account for their likely impacts on contraceptive use. Marital status at the current interview was categorized as never married, currently married, separated/divorced/widowed, or currently cohabiting (regardless of whether the respondent was never married or formerly married). The presence of biological children in the home was categorized as any versus none at the time of the interview. Lastly, we controlled for whether participants had seen a doctor or health care provider within the past 12 months. 9
Statistical analysis
Data were summarized as medians with interquartile ranges (IQR) or counts with percentages. Using data from the most recent interview available for each individual respondent (2019, or the year in which the respondent was censored, whichever occurred first), we compared the study outcome and study covariates based on the insurance coverage categories listed above, using chi-square tests, Fisher’s exact tests, or Kruskal–Wallis tests, as applicable. Descriptive analysis was conducted on a person-level file (one observation per person) to accommodate bivariate tests assuming independence of observations. A mixed effects multinomial logistic regression was fitted to the entire dataset of person-years (where each completed interview was treated as a separate observation) using the “gsem” command in Stata/SE 16.1 (College Station, TX: StataCorp, LP). Our presentation of model results focused on the relative risk ratios (RRRs) for the use of prescription versus nonprescription methods, and for the use of prescription methods versus no method. The model was adjusted for all covariates, and included a participant-level random intercept and cluster-robust standard errors, to account for nonindependence of data from repeated interviews with the same participant. This specification of model variance was not compatible with the use of survey weights, thus we did not incorporate survey weights in the analysis.
Results
The dataset included 4,126 female respondents who participated in the survey at least once between 2007 and 2019. After excluding interviews collected when respondents were pregnant, not sexually active, or sexually active exclusively with female partners (ascertained in 2013 and 2015 only), we retained 3,969 respondents in the sample. We then excluded a further 231 respondents with missing data on study variables, leading to a final analytic sample of 3,738 unique respondents and 18,678 person-year interviews (observations). Based on the most recent interview of each respondent, 35% (n = 1,317) used prescription contraception, 16% (n = 586) used nonprescription methods only, and 49% (n = 1,835) used no method. At the time of the most recent interview, 57% of respondents (n = 2,143) had continuous private coverage, 18% (n = 675) had continuous public coverage, 15% (n = 549) had part-year uninsurance, and 10% (n = 371) were considered to be uninsured year-round.
Additional descriptive statistics, stratified by pattern of health insurance coverage, are shown in Table 1. On this bivariate analysis, respondents with continuous private insurance were more likely to use prescription contraceptive methods (40%) than those with continuous public insurance or part-year uninsurance (32% in both groups) or those who were uninsured year-round (21%). Considering socioeconomic characteristics, respondents with part-year uninsurance were less likely to complete college or to be employed than respondents with year-round private coverage, but tended to have higher levels of educational attainment and higher employment rates than respondents with year-round public coverage and respondents who were uninsured year-round. However, respondents experiencing part-year uninsurance were much less likely to have seen a doctor in the past year (65%) when compared with respondents with either private or public year-round coverage (82% and 83%, respectively).
Study Outcomes and Covariates According to Pattern of Health Insurance Coverage
IQR, interquartile range.
Table 2 summarizes results from the multivariable model, contrasting the use of prescription contraception with the use of nonprescription contraception methods only. Respondents with part-year uninsurance were 20% less likely to use prescription rather than nonprescription methods, as compared to respondents with private insurance (RRR: 0.80; 95% confidence interval [CI]: 0.69, 0.94; p = 0.007). However, part-year uninsurance was not associated with significantly lower use of prescription rather than nonprescription methods when compared to continuous public insurance (RRR: 0.90; 95% CI: 0.78, 1.04; p = 0.141). Continuous lack of insurance was associated with a significantly lower likelihood of using prescription rather than nonprescription methods, whether in comparison to continuous private coverage (RRR: 0.57; 95% CI: 0.47, 0.71; p < 0.001) or continuous public coverage (RRR: 0.64; 95% CI: 0.51, 0.81; p < 0.001).
Multivariable Multinomial Logistic Regression Predicting Use of Prescription Versus Nonprescription Contraceptive Method
CI, Confidence interval; RRR, relative risk ratio; Ref, Reference.
Based on Table 2, the use of prescription rather than nonprescription methods was also less likely among older respondents, Black or Hispanic respondents (compared to non-Hispanic White), and respondents who were out of the labor force (compared to employed respondents). By contrast, the use of prescription rather than nonprescription methods was more likely among respondents with higher educational attainment. The use of prescription methods was strongly associated with seeing a doctor in the past year, which nearly doubled the relative risk of using prescription versus nonprescription method. In this model, the variance of the random intercept was highly statistically significant (p < 0.001), indicating that the choice of contraceptive method was strongly correlated across multiple interviews contributed by the same respondent.
Table 3 presents model results comparing the use of prescription contraception to no contraception at all. Respondents with part-year uninsurance did not differ from respondents with continuous private coverage in their likelihood of using prescription methods rather than no method (RRR: 0.90; 95% CI: 0.78, 1.04; p = 0.141). Continuously uninsured respondents were least likely to use prescription contraception rather than no contraception method (RRR vs. private insurance: 0.57; 95% CI: 0.47, 0.69; p < 0.001; RRR vs. public insurance: 0.67; 95% CI: 0.54, 0.84; p = 0.001).
Multivariable Multinomial Logistic Regression Predicting Use of Prescription Contraceptive Method Versus No Contraceptive Method
CI, Confidence interval; RRR, relative risk ratio; Ref, Reference.
Considering other variables in our analysis, Table 3 indicates that the use of prescription methods rather than no method was less likely among older respondents, respondents who were unemployed or out of the labor force (compared to employed respondents), and respondents who were married or cohabiting (compared to never-married respondents). The use of prescription methods rather than no method was less likely among Black and Hispanic respondents, respondents with higher levels of educational attainment, and respondents who had seen a doctor in the past year.
Discussion
Despite abundant evidence that contraceptives play a crucial role in the autonomy over one’s reproductive choices, barriers to contraceptive access persist, particularly among minority and socioeconomically disadvantaged patients, and these barriers contribute to the morbidity and mortality crisis among pregnant patients in the United States. 18 Apart from their significance for reproductive autonomy, many contraceptive methods have benefits such as lightening of menstrual cycle flow and relief from dysmenorrhea, treatment of polycystic ovary syndrome and endometriosis, and improvement in hormonal acne. 19 In this study, we expand upon prior evidence on disparities in contraception access by demonstrating that young women with continuous private insurance were more likely to use prescription contraceptive methods compared to their peers experiencing part-year or year-round uninsurance. These findings align with prior studies that demonstrate gaps in health insurance coverage are associated with poorer health and persistently lower preventive health care use. 14,20 To address these disparities, policy efforts should focus on increasing flexibility of eligibility and enrollment periods for health insurance and supporting continued coverage for preventive care services during times when adults might otherwise experience coverage gaps.
Prescription contraceptive access generally requires consistent access to clinical care, made easier and more affordable by stable insurance coverage. In our study, the use of prescription contraceptive methods was strongly associated with seeing a doctor in the past year. Starting, stopping, or switching contraceptives is common as an individual’s preferences and priorities surrounding contraceptives change, family planning goals evolve, or unwanted side effects surface. 19,21 The oral contraceptive pill, vaginal ring, and transdermal patch require prescription refills at variable intervals, and the injectable contraceptive must be administered at 90-day intervals to be effective. 10 Continuing to use these prescription methods may be difficult for women not able to access follow-up care, or who are unsure if they will have access in the foreseeable future. Furthermore, LARC methods, such as IUDs and subdermal contraceptive implants, require visits for both insertion and discontinuation; a visit for removal may discourage the use of these methods among patients anticipating a future gap in coverage. 22 While completion of a doctor’s visit was included as a potential confounder in our analysis, we acknowledge that it could also mediate the effect of health insurance coverage. This strengthens our conclusions about the unique contribution of insurance coverage patterns to choice of contraception method since these associations are evident even when controlling for a measure of health care utilization.
Financial barriers are often cited as a concern by women seeking prescription contraception, particularly among uninsured and publicly insured women. For insured women, costs associated with contraception can be attributed to policy deductibles, co-pays, or services not covered by insurance. 6 For women who are uninsured or experiencing temporary loss of insurance, and who are seeking LARC placement, these financial barriers represent the entire cost of the device and placement or over $1,000 in out-of-pocket costs. 6 In addition, women experiencing part-year or year-round uninsurance may be more likely to utilize safety net health care settings for their primary care needs, such as emergency departments (EDs), urgent care facilities, and local health departments. These facilities often lack sufficient preventive care services, and may not provide adequate contraceptive options. 8,9 In one study, young women who reported not having a usual source of care and women who used the ED as their usual source of care had higher likelihood of using no contraception or a less effective method of contraception rather than highly effective prescription methods. 9 Young adults are also at an age range when they are highly likely to experience coverage gaps, as they age out of eligibility for federally funded public health insurance programs or out of eligibility for coverage by a parental insurance plan. 14
The association of part-year uninsurance with lower rates of prescription contraceptive use (relative to the use of nonprescription methods only) underscores the need for interventions to address coverage instability. While the Affordable Care Act guarantees coverage of women’s preventive care services, including free birth control and contraceptive counseling for individuals and covered dependents, coverage instability is a common reality for patients, particularly those with Medicaid. 12,13 This is of increasing concern now that pandemic era continuous Medicaid eligibility has expired. Extending coverage for preventive care services for times when adults might otherwise be uninsured could mitigate the consequences of coverage loss, particularly for services such as contraception. Meanwhile, online services for obtaining contraception, specifically the oral contraceptive pill, have grown substantially, with nine different online platforms offering these services as of 2018. 23 In addition, disparities in contraceptive access may be addressed through publicly supported sites offering contraceptives, including Title X funded sites such as Planned Parenthood, and non-Title X funded sites such as federally qualified health centers. 24 These sites provide low-cost or free family planning services for patients who are frequently uninsured and may lack any other usual source of health care. 25
This study has several limitations related to the source data and analytic approach. Respondents were born between 1980 and 1984 and may not be representative of women currently reaching reproductive age. The NLSY97 also relies on self-reported data, introducing the potential for recall bias. Contraceptive method was classified based on the “most frequently” used method, thus introducing the potential for classification error. In addition, the use of any birth control methods and the use of condoms specifically were ascertained in two separate questions, further contributing to potential classification bias. If the respondent reported only having female partners in the 12 months before the interview the observation was excluded, but this information was only available in 2013 and 2015 and did not account for the sex of partners in other years or between interviews. Although respondents who were currently pregnant were excluded, the data set did not include information on whether respondents were currently attempting to conceive at the time of interview, meaning we were unable to account for individual reproductive goals. This resulted in a grouping of unalike populations (those trying to conceive and those not using contraceptives for other reasons) but was unlikely to bias our analysis comparing the use of prescription contraceptives to the use of nonprescription contraceptives. Furthermore, we excluded data on surgical sterilization owing to its irreversible nature and the inability to distinguish between female and male sterilization in the NLSY97 data (presumably, sterilization of a male partner would be less affected by the female respondent’s health insurance status than the respondent’s own choice of contraception method). Lastly, we were unable to account for the role of contraception dissatisfaction in contraceptive choice. 21
Despite these limitations, our study contributes to a growing body of evidence that identifies how part-year uninsurance is associated with poorer health and reduced use of preventive care services. Given the association of stable insurance coverage with increased use of prescription contraceptives, further efforts are needed to ensure the stability of insurance coverage among US adults of reproductive age. Policy interventions can also target flexibility of eligibility and enrollment periods, and continued coverage for certain preventive services during periods of coverage loss or change. Expanding existing interventions, including funding for Title X sites and other publicly funded contraceptive providers, is also crucial. Meanwhile, further research should identify specific mechanisms explaining the association between coverage continuity and prescription contraceptive use. Specific directions may focus on subsets of prescription contraceptives (LARC vs. more readily accessible oral contraceptive pills), whether longer coverage gaps are associated with greater impact on contraceptive choice, and how geographic access to publicly funded contraceptive providers can modify the significance of individual insurance coverage.
Footnotes
Authors’ Contributions
S.M.: Conceptualization and design, data interpretation, and drafting of the article. L.E.: Conceptualization and design, data acquisition, data analysis, and drafting of the article. A.F.: Conceptualization and design, data acquisition, data interpretation, and critical revision of the article. K.C.: Conceptualization and design, data interpretation, and critical revision of the article. D.T.: Conceptualization and design, data acquisition, data analysis, and critical revision of the article. All authors reviewed and approved the final article for submission.
Author Disclosure Statement
D.T. discloses salary support from the Kate B. Reynolds Charitable Trust Foundation and Lilly and Co. Inc. for research and quality improvement projects unrelated to this work. All other authors have no potential conflicts of interest to disclose.
Funding Information
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.