Does access to free contraception change the method choice of people who say they have difficulty paying for health-related services? | BMC Women’s Health
HER Salt Lake is a prospective, quasi-experimental, observational cohort study that enrolled eligible women at four participating Planned Parenthood Association of Utah (PPAU) clinics in Salt Lake County between September 2015 and March 2017 . Prior to the beginning of the control period, staff at all PPAU health centers across the state were trained to provide clients with standardized, patient-centric counseling; The Method Effectiveness Chart and Counseling Interview Guide used during these counseling sessions are available as Supplementary Files 1 and 2. FPL) received inexpensive but not free care. During this time, the copper IUD was available on a sliding cost scale going up to no charge, and the levonorgestrel 52 mg IUS, Liletta® (sold by Allergan/Medicines360) was available through 340b, the government-sponsored medication assistance program, for a total device cost of $50 . The contraceptive implant was not available at a reduced price. After the control period, in a 12-month intervention period, all out-of-pocket contraceptive costs were eliminated for three years from a patient’s admission visit, regardless of the patient’s income. Those treated during the free intervention period could return as many times as they wished and switch to any other method free of charge for three years. Details on the suitability and methodology of the HER Salt Lake study are reported elsewhere .
To be eligible for the prospective arm of the study, patients had to be (1) between 18 and 45 years of age; (2) speak fluent English or Spanish; (3) desire to prevent pregnancy for at least one year; (4) have a working cell phone; and (5) have income below 300% FPL. During the 18-month study period, 4,425 patients consented to participate in the prospective study and agreed to complete detailed questionnaires at enrollment and at eight subsequent time points (1st, 3rd, 6th, 12th, 18th, 24th, 30th – and 36 months after enrollment). This secondary analysis is limited to prospective HER Salt Lake study participants who answered “yes” to the question “In the past 12 months, have you had any problems paying for medical care or medication?” in the enrollment survey. One participating abortion clinic offered low-cost and free contraception prior to the introduction of HER Salt Lake; We therefore excluded participants treated at this clinic from this analytical sample. The selection of our study sample is shown in detail in Fig. 1. We used both survey data and medical record data. We have collected and stored survey data via secure, web-based Research Electronic Data Capture (REDCap). We extracted the participants’ health records, including the contraceptive method chosen at enrollment and changes in contraceptive method over the course of the study, from the PPAU electronic medical record system and linked this data to the enrollment data.
Statistical Methods and Analysis
Our primary endpoint assessed changes in method choice between the control and intervention phases in HER participants who reported having difficulty paying for health care, when asked a yes/no question. Our secondary outcome compared one-year satisfaction with the method in contraceptive continuators. We compared baseline differences in participant demographics between the control and intervention periods using chi-square tests and two-sample t-tests. Our baseline comparison includes the full list of race and ethnicity categories found on the HER Salt Lake study enrollment form; Due to the small number, we combined these categories in our primary and secondary analyzes using methodology consistent with previously published literature from the HER Salt Lake study . Although we have included all methods in our denominator, both the primary and secondary analyzes only rate the six most popular birth control methods (copper and hormonal IUD, contraceptive implant, birth control injection, vaginal ring, and oral contraceptives) due to the small selection of other methods (2.6 %).
We assessed our primary outcome by running simultaneous multivariate logistic regression models comparing differences in method uptake by study period and between methods. To develop our full model, we first performed unadjusted regression analyzes on all variables proposed for inclusion in the final multivariable models. We used a cutoff of 0.25 to determine covariate inclusion in the final models, as supported by the literature [8, 9]. Our covariates included variables known to influence contraceptive choice, including age, race and ethnicity, education, employment status, insurance type, federal poverty level, and parity. In addition, we controlled for health center admission location, use of LARC, and history of abortions as these were significant in unadjusted analyses. After determining our final covariates, we ran six multivariable logistic models to assess predictors of method choice for each of the six most popular birth control methods. Accordingly, we applied the Benjamini-Hochberg procedure as a test correction for multiple comparisons.
For our secondary outcome, which examines satisfaction with the method among birth control continuers at one year, we defined “continuers” as those who reported continuing the same method they chose at enrollment in their three, six, and 12-month follow-up surveys had. We made an exception when participants reported male or female condom use, fertility awareness-based methods, withdrawal, or emergency contraception in these follow-up surveys: when a participant reported using one of these methods but later reported the same method she used had selected registration, we categorized them as continuators with supplementary method use.
To assess the predictors of method satisfaction within this cohort, we used reported method satisfaction at the 12-month survey (measured on a Likert scale with choices of “completely satisfied”, “somewhat satisfied”, “neutral”, “somewhat satisfied”) dissatisfied” and “completely dissatisfied”). To ensure sufficient numbers for analysis, we aggregated responses into three categories: fully satisfied, somewhat satisfied/neutral, somewhat/completely dissatisfied, and compared the distribution of responses in the original categories and our aggregated categories. We performed a single multivariable model that assessed predictors of method satisfaction among continuers. We hypothesized that users using the same method for a year would be “completely satisfied” with that method, so we used this category as a reference. We performed all analysis in Stata 15.0 or later (StataCorp LP, College Station, TX). The University of Utah IRB approved this study.