A Decision Analysis Model of 1-Year Effectiveness of Intended Postplacental Compared With Intended Delayed Postpartum Intrauterine Device Insertion

Abstract

Objective: To compare, using decision analysis methodology, the 1-year probability of pregnancy after intended postplacental intrauterine device (IUD) insertion with intended delayed insertion at an outpatient postpartum visit (delayed postpartum placement).

Methods: We developed an evidence-based decision model with the primary outcome of 1-year probability of pregnancy. We compared 1-year probability of pregnancy after intended postplacental or intended delayed postpartum IUD placement. We obtained estimates from the literature for the proportions of the following: mode of delivery, successful IUD placement, IUD type, postpartum visit attendance, IUD expulsion, IUD discontinuation, and contraceptive use, choice, and efficacy after IUD discontinuation. We performed sensitivity analyses and a Monte Carlo simulation to account for variations in proportion estimates.

Results: One-year probabilities of pregnancy among a theoretical cohort of 2,500,000 women intending to receive a postplacental IUD after vaginal birth and 1,250,000 women intending to receive a postplacental IUD after cesarean birth were 17.3% and 11.2%, respectively; the 1-year probability of pregnancy among a theoretical cohort of 2,500,000 women intending to receive a delayed postpartum IUD was 24.6%. For delayed postpartum IUD placement to have effectiveness equal to postplacental placement, 91.4% of women delivering vaginally and 99.7% of women delivering by cesarean would need to attend postpartum care. Once placed, the effectiveness of postplacental IUDs was lower than that of delayed postpartum IUDs: 1-year probabilities of pregnancy after IUD placement at a vaginal birth, cesarean birth, and an outpatient postpartum visit were 15.4%, 6.6%, and 3.9%, respectively.

Conclusion: After accounting for factors that affect successful IUD placement and retention, this decision model indicates that intended postplacental IUD insertion results in a lower 1-year probability of pregnancy as compared with intended delayed postpartum IUD insertion.

Figure 1.

Decision tree model of postplacental intrauterine device (IUD) placement, used for both vaginal and cesarean birth. Probabilities at each decision node add up to 100% and are listed with their references in Table 1. The modeled one-year probability of pregnancy after intended vaginal postplacental IUD is 17.3%. The modeled one-year probability of pregnancy after intended cesarean postplacental IUD is 11.2%. LNG-IUD, levonorgestrel-releasing intrauterine device; LARC, long-acting reversible contraceptives.

Figure 2.

Decision tree model of delayed postpartum intrauterine device (IUD) placement. The delayed postpartum IUD model contains an additional modeling the probability of a patient returning for a postpartum follow-up visit. Probabilities at each decision node add up to 100% and are listed with their references in Table 1. The modeled one-year probability of pregnancy after intended delayed postpartum IUD is 24.6%. LNG-IUD, levonorgestrel-releasing intrauterine device; LARC, long-acting reversible contraceptives.

Figure 3.

One-way deterministic sensitivity analysis for the postpartum intrauterine device (IUD) decision model. A. Postplacental IUD insertion after vaginal delivery. B. Postplacental IUD insertion after cesarean delivery. The Pprobabilities of events in the model were varied individually from the lowest to the highest reasonable value for method failure to identify which variables had the largest effect on the difference in contraceptive outcome. The base case model is shown by the vertical line. Each horizontal bar represents the change in effectiveness between immediate postplacental insertion and delayed postpartum insertion when varied across the range of probabilities tested (ranges indicated in parentheses). LNG-IUD, levonorgestrel-releasing intrauterine device.

Figure 4.

Probabilistic sensitivity analysis for the postpartum intrauterine device (IUD) decision model. For 10,000 theoretical scenarios, the probabilities for each decision (input) were varied across the ranges displayed in Table 1. Each vertical bar represents the percent of the 10,000 scenarios that showed improved effectiveness of postplacental as compared with delayed postpartum insertion (output). The vertical line represents the threshold for effectiveness of postplacental insertion (bars to the right favor postplacental insertion). A total of 9,957 vaginal model scenarios favor immediate postplacental insertion. All 10,000 cesarean model scenarios favor immediate postplacental insertion. The additional percentage increase in efficacy expected with immediate postplacental insertion are as follows: vaginal: 7.30% (2.55%, 7.19%, 12.41%); cesarean: 13.38% (8.35%, 13.33%, 18.65%). Values reported are mean (5^th percentile, 50^th percentile, 95^th percentile, respectively).