Objective
Oregon and federal laws prohibit giving informed consent for permanent contraception when presenting for an abortion. The primary objective of this study was to estimate the number of unintended pregnancies associated with this barrier to obtaining concurrent tubal occlusion and abortion, compared with the current policy, which limits women to obtaining interval tubal occlusion after abortion. The secondary objectives were to compare the financial costs, quality-adjusted life years, and the cost-effectiveness of these policies.
Study Design
We designed a decision-analytic model examining a theoretical population of women who requested tubal occlusion at time of abortion. Model inputs came from the literature. We examined the primary and secondary outcomes stratified by maternal age (>30 and <30 years). A Markov model incorporated the possibility of multiple pregnancies. Sensitivity analyses were performed on all variables and a Monte Carlo simulation was conducted.
Results
For every 1000 women age <30 years in Oregon who did not receive requested tubal occlusion at the time of abortion, over 5 years there would be 1274 additional unintended pregnancies and an additional $4,152,373 in direct medical costs. Allowing women to receive tubal occlusion at time of abortion was the dominant strategy. It resulted in both lower costs and greater quality-adjusted life years compared to allowing only interval tubal occlusion after abortion.
Conclusion
Prohibiting tubal occlusion at time of abortion resulted in an increased incidence of unintended pregnancy and increased public costs.
See related editorial, page 3
In 1978, the federal government enacted 42 Code of Federal Regulations (CFR) §§ 50.201-210, which placed restrictions on female permanent contraception (ie, tubal occlusion) performed through federally assisted family planning programs. These laws were enacted to safeguard women from involuntary sterilization, as a response to decades of coercive tactics perpetrated by state eugenics boards. These coercive practices included: forced sterilization of institutionalized people, rape victims, those deemed to have low intelligence quotients (IQ), and, at times, women requesting an abortion. Several of these well-known federal restrictions include waiting periods, an age limit, and the need to have a government-informed consent form (“tubal papers”) present in the operating room. An additional key restriction in 42 CFR § 50.204 is the prohibition against giving informed consent for female permanent contraception at time of abortion. These federal restrictions were adopted more broadly by the state of Oregon in 1995, when Oregon Revised Statute § 436.225 took effect. This statute included the same prohibition against giving informed consent for permanent contraception at the time of an abortion for all women in the state, regardless of payer status.
Women who present for an abortion often have an unmet need for contraception. Delaying provision of contraception to a follow-up visit is known to result in high rates of unfulfilled requests for contraception, which in turn perpetuates the risk of unintended pregnancy. Women with unfulfilled requests for permanent contraception may have an even higher risk of having an unintended pregnancy within 1 year, compared to women who request reversible methods. A prior decision analysis found that reducing barriers to Medicaid-funded postpartum tubal occlusion resulted in 19,000 fewer unintended births and 10,000 fewer abortions, saving $215 million nationally over 1 year.
Tubal occlusion is the second most popular form of contraception in the United States and the most popular form of permanent contraception. The reason for its popularity is likely several-fold, including that tubal occlusion is safe and effective. Procedure-related mortality and serious morbidity are rare, and complication rates are similar whether performed concurrently with an abortion or as an interval procedure. Tubal occlusion is >99% effective in preventing unintended pregnancy.
Although the current restriction on giving informed consent for permanent contraception at time of abortion was intended to protect women, it is problematic in terms of the key ethical principles of autonomy, beneficence, and justice. To deny a woman the right to consent to a popular form of contraception in the setting of a pregnancy termination deems her incompetent to choose the best method to protect herself against further unintended pregnancy. Furthermore, delaying access to desired contraception creates multiple risks to a woman’s health.
Given the risk of unintended pregnancy, the popularity and effectiveness of tubal occlusion, and the possible negative repercussions on patient autonomy and ability to obtain desired health services, the question arises as to whether the current prohibition limiting concurrent permanent contraception with abortion is beneficial to women. This study compared a policy of allowing concurrent tubal occlusion with abortion with the current policy of offering only interval tubal occlusion. We specifically considered unintended pregnancy, quality-adjusted life years (QALYs), and direct Medicaid costs as our outcomes.
Materials and Methods
We designed a decision-analytic model using TreeAge Pro 2011 (TreeAge Software, Williamstown, MA). Decision analysis is a tool that allows stepwise comparison of probabilities and outcomes of different options. Our theoretical cohort consisted of women with Medicaid insurance, who requested tubal occlusion when presenting for abortion. Given differences in fertility and permanent contraception regret, we considered 2 cohorts of women: one <30 years old, and the other ≥30 years old.
Our model compared number of unintended pregnancies, costs, and QALYs for a policy of offering concurrent tubal occlusion compared with the current practice of offering interval tubal occlusion only ( Figure 1 ). We used a Markov, or recursive, model to analyze the probability of pregnancy over 5- and 10-year time periods, accounting for changes in contraceptive use and method failure. Such a model assumed that, in each cycle of the model, a woman could either become pregnant or not. If not, she could become pregnant in the next cycle. If she did become pregnant, she could experience a miscarriage, an elective pregnancy termination, an ectopic pregnancy, or a viable pregnancy. The Institutional Review Board at Oregon Health and Science University deemed the research protocol exempt from review.
The model included partial interval salpingectomy as the method of tubal occlusion. This method was chosen as it is common, and long-term data on risk of failure are available from the Collaborative Review of Sterilization Workgroup (CREST) study. Women who did not receive requested tubal occlusion had the option of using short- or long-acting reversible contraception (LARC). For women who became pregnant and did not undergo termination of pregnancy, or did not experience spontaneous abortion or ectopic pregnancy, the outcomes were conservatively assumed to be uncomplicated pregnancies.
We included the probability of permanent contraception regret to evaluate QALYs associated with undesired infertility. The model accounted for the probability of a woman with regret undergoing a fertility procedure–in vitro fertilization (IVF) or tubal reversal–and for the success of these procedures. However, because Oregon Medicaid does not cover fertility procedures, we assumed at baseline that no women would undergo IVF or a tubal reversal. Variation in the probability of receiving these procedures was accounted for in the sensitivity analyses, including a Monte Carlo simulation.
Markov process
The Markov process included both 5- and 10-year time horizons, and each cycle length was 1 year. A 10-year time horizon was chosen to account for the possibility of regret several years after a woman received permanent contraception, and because it was the interval of action for the longest-duration LARC. The model began by comparing 2 policies: tubal occlusion concurrent with abortion, and interval tubal occlusion only. In both branches reversible contraception was also offered. In the branch where concurrent tubal occlusions occurred, all tubal occlusion procedures occurred at time of abortion. The subsequent probability of pregnancy was due to method failure and was accounted for by the Markov process. In the alternate branch, where only interval tubal occlusion occurred, the probability of pregnancy was dependent upon method failure, as well as ability to access an interval procedure. These factors were reflected in the Markov process.
The Markov nodes were depicted by the letter “M” within a circle ( Figure 1 ). In the first cycle of the Markov process, all women who underwent either a concurrent tubal occlusion with abortion or interval tubal occlusion entered the model through the tubal occlusion branch of the Markov node. All other women entered the model through the no tubal occlusion branch of the respective Markov node.
In subsequent cycles of the Markov process, a woman could reenter any of the branches of the Markov node. All deliveries, except for those obtained through IVF, ended with the possibility of a postpartum tubal occlusion. By the end of each cycle, if a woman had a tubal occlusion at the time of her abortion, at an interval visit, or postpartum, she then reentered the Markov process in the branch labeled “tubal occlusion” in the next cycle. We assumed the same probability of becoming pregnant after a tubal occlusion no matter where in the model the tubal occlusion occurred. All women who received LARC and did not become pregnant reentered the model through the LARC branch of the respective Markov node in the next cycle. A woman reentered the Markov process in her respective no tubal occlusion branch if she had not undergone a tubal occlusion or if she had become pregnant and did not have a postpartum tubal occlusion.
Probabilities
Probability inputs came from the literature ( Table 1 ). The annual probability of becoming pregnant using a given type of contraception was the sum of the failure rate of that method plus half of the discontinuation rate. The discontinuation rate was adjusted to simplify the model and to lead to a conservative bias of the probability of becoming pregnant by assuming that half of women who discontinue a method will switch to an equally effective contraceptive method. The failure rate for tubal occlusion was based off of the first year of the CREST data to be conservative; the data for our younger and older cohorts came from the 27- to 33-year-old and the 34- to 44-year-old CREST groups, respectively. The failure rate for all long-acting contraceptives (implant and intrauterine device [IUD]) was assumed equivalent to the copper IUD, which has the same duration of action as our longest time horizon. The failure rate of all short-acting methods (injectable, pill, patch, ring, and condom) was assumed equivalent to the failure rate of oral contraceptive pills.
| Variable | Probability, % | SD | Reference |
|---|---|---|---|
| Obtain interval TO | 21.0 | 7.0 | |
| Become pregnant after: | |||
| TO | |||
| Age <30 y | 0.75 | 0.74 | |
| Age ≥30 y | 1.23 | 0.85 | |
| LARC a | 11.8 | 0.54 | |
| Short-acting contraception | 25.5 | 0.83 | |
| No contraception | 85 | 0.23 | |
| IVF | 30.9 | 1.6 | |
| Tubal reversal | 73.0 | 4.7 | |
| Use contraception b | 56.9 | 3.3 | |
| Use LARC | 19.5 | 1.9 | |
| Regret sterilization c | |||
| Age <30 y | |||
| 5 y | 1.5 | 0.16 | |
| 10 y | 1.5 | 0.16 | |
| Age ≥30 y | |||
| 5 y | 0.7 | 0.01 | |
| 10 y | 0.4 | 0.08 | |
| Undergo fertility procedure d | |||
| Age <30 y | 4.1 | 0.26 | |
| Age ≥30 y | 0.4 | 0.083 | |
| Choose IVF if undergoing fertility procedure | 48.5 | 3.9 | |
| Have miscarriage | 12.0 | 0.17 | |
| Have abortion after: | |||
| Fertility procedure e | 2.0 | 0.16 | |
| All others | 35.0 | 0.25 | |
| Have ectopic pregnancy after: | |||
| TO | 43.0 | 2.4 | |
| LARC | 3.7 | 2.6 | |
| IVF | 2.8 | 0.078 | |
| Tubal reversal | 3.9 | 1.7 | |
| All others | 1.0 | 0.0014 | |
| Have term delivery after f : | |||
| IVF | 55.8 | 1.3 | |
| Tubal reversal | 50.6 | 5.3 | |
| All others | 53.0 | 0.27 | |
| Have vaginal delivery | 71.0 | 0.21 | |
| Obtain postpartum TO | 54.0 | 1.9 | |
| Unintended pregnancy | 99.2 g | 1.0 | |
| Undesired infertility | 98.9 g | 0.1 |
a Mean value, including discontinuation, for copper intrauterine device between both references
b Average of values found in both references
c Per yearly cycle of Markov process
d Baseline model used 0% as probability for both age groups–values in table were used for Monte Carlo simulation
e Half of value for intended pregnancies ending in abortion in reference–this assumes that half of abortions of intended pregnancies are for maternal health and half are for fetal anomalies–we assumed that abortion in this population would be for anomalies
f Delivery after TO was calculated as [1 – (sum of probabilities for miscarriage, ectopic pregnancy, and abortion)] but in all other scenarios, miscarriage was calculated as [1 – (sum of probabilities for delivery, ectopic pregnancy, and abortion)]
Costs
Costs were calculated from the perspective of Oregon’s Medicaid plan and were in 2013 US dollars ( Table 2 ). The cost of LARC came from the state’s family planning program (CCare) billing records and the literature. It was an average of the costs for the levonorgestrel and copper IUD. The cost of short-acting contraception was the average cost of depot medroxyprogesterone acetate, oral contraceptive pills, ring contraceptive, and condoms from CCare billing data. Short-acting contraception included an additional $100 office visit each year. The costs for IVF and tubal reversal were set at $15,000 based on estimates from the Portland, OR, medical community. The remainder of the costs came from the literature. As global billing for both prenatal care and delivery is the common in pregnancy care coverage, particularly Medicaid, the delivery costs were assumed to account for prenatal care as well. Additionally, as a conservative assumption, newborn and childhood care were not included in the model.
| Variable | Cost, $ | Reference |
|---|---|---|
| TO | 1308.00 | |
| LARC | 350.00 | |
| Short-acting contraception | 173.00 | |
| IVF | 15,000.00 | |
| Tubal reversal | 15,000.00 | |
| Miscarriage | 1407.00 | |
| Abortion | 2208.00 | |
| Ectopic pregnancy | 6448.00 | |
| Vaginal delivery | 4520.00 | |
| Cesarean delivery | 10,295.00 |
QALYs
QALYs, the product of both life expectancy and utility, allow for comparisons between different types of outcomes. Utility, a measure of satisfaction or value for a particular health state, was defined in this case as avoiding an unintended pregnancy and infertility. By convention, utility values range from 0-1, with 0 representing death and 1 indicating perfect health. The utility of an unintended pregnancy was derived from the literature and, based on the time tradeoff metric, was 0.992. The QALY for undesired infertility was derived from an international study estimating the disutility of secondary infertility.
Analysis
We stratified the analyses for women <30 years old and women at least 30 years old. Costs and QALYs were calculated to determine the incremental cost-effectiveness ratio of each strategy. Incremental cost-effectiveness ratio is a measure of cost-effectiveness, which compares the differences between the costs and health outcomes of 2 competing interventions. The willingness-to-pay threshold was set at $100,000 per QALY. This meant that a strategy was considered cost-effective until Oregon Medicaid would have to pay >$100,000 for an additional QALY gained. Costs and QALYs were discounted at a standard 3% rate. We discounted these outcomes because the value of money and QALYs is often perceived to be worth less if the outcome is not realized immediately.
Sensitivity analysis is a statistical tool that allows estimation of how a change in one of our model parameters would affect outcomes. Univariate and multivariate sensitivity analyses were performed to test model assumptions and determine consistency of findings. Univariate sensitivity analyses were performed on all variables to test for threshold values. A threshold value marked where a change in the input resulted in a different outcome. All variables were tested broadly. Costs were tested from one half to twice their base value. If the 2 strategies trended toward an intersection for cost, then the upper cost value was increased until a threshold value was determined. Two-way sensitivity analyses were performed on all variables with a threshold value as well as other key variables.
We performed a Monte Carlo simulation using 1000 trials to evaluate how simultaneous multivariable changes could affect outcomes. The Monte Carlo simulation enabled variation of all probability estimates concurrently by sampling distributions around the baseline estimate. Each probability, cost, and QALY was randomly assigned from the range according to Tables 1 and 2 , and these values were reassigned in each trial and each stage of the Markov process. Probabilities and utilities used beta distributions with SDs listed in Table 1 . Costs used gamma distributions with a SD of 50%. Scatter plots were developed to represent uncertainty in results, and a 95% confidence ellipse was generated.
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
