Pharmacy claims data versus patient self-report to measure contraceptive method continuation

Abstract

Objective

To compare self-reported 12-month continuation of oral contraceptive pills (OCPs), patch, and ring versus continuation by pharmacy claims data.

Study Design

Women in the Contraceptive CHOICE Project who chose OCPs, the patch, or ring as their initial method were included. Continuation was assessed by periodic telephone survey and by obtaining prescription claims data. Continuation was defined as no gap of more than 30 days. Kaplan-Meier survival functions were used to estimate continuation rates and cumulative unintended pregnancy rates. Kappa statistic assessed the level of agreement between self-report and claims data.

Results

We analyzed 1,510 women who initiated use by 3 months and provided information on discontinuation. Of OCP users, 59% continued their method at 12 months by self-report versus 38% by pharmacy claims. Patch and ring users had self-reported/pharmacy continuation of 45%/28%, and 57%/37%, respectively. Kappa coefficients and their 95% confidence intervals between the two measurements were 0.46 [0.40–0.52], 0.54 [0.39–0.68], and 0.54 [0.47–0.61] for OCP, patch, and ring, respectively. Among women who self-reported continuation, unintended pregnancy rates were 0.4% in those who continued by pharmacy claims versus 4.9% in those who discontinued according to claims data.

Conclusion

Contraceptive continuation rates differ by self-report versus pharmacy claims with women overestimating their continuation by self-report.

1. Introduction

Half of all unintended pregnancies in the United States are the result of contraceptive failure or misuse [1]. While combined hormonal contraceptives have high efficacy rates in clinical trials, failure rates with typical use are approximately 9% per year [2]. Many of these failures are related to method discontinuation [3]. Additionally, women who discontinue combined hormonal contraceptives are more likely to switch to less-effective methods of contraception, including no method, putting them at even greater increased risk of unintended pregnancy [4,5].

Although sequelae of contraceptive discontinuation are well known, current methods of assessing discontinuation patterns may be inaccurate. Most studies have relied on patient self-report, with 12-month continuation rates ranging from 53–67% for oral contraceptive pills (OCPs), and continuation rates of 42–54% and 26–49% reported for contraceptive ring and patch, respectively [5–9]. A few studies have compared self-reported measurement of OCP use to more objective measures of usage including electronic pill monitoring and blood hormone levels, but these methods are difficult to reproduce in large cohorts [10,11]. Analysis of a large US pharmaceutical database demonstrated low rates (26–29%) of uninterrupted contraceptive use for OCPs, patch, and ring at one year [12]. However, large database studies are limited by the lack of correlated clinical data. Studies have additionally been limited by lack of consistently defined terminology [7].

To further understand patterns of contraception continuation, we analyzed continuation at 12 months among Contraceptive CHOICE Project participants using OCP, patch, or ring. We compared self-reported continuation rates to pharmacy claims data. Given lower continuation rates in pharmacy claims studies compared to survey-based studies, we hypothesized that women in our study would overestimate their continuation compared to pharmacy records [5–9,12]. We further analyzed the subgroup of women who self-reported continuation and compared rates of unintended pregnancy between women whose pharmacy records indicated continuation versus those whose records indicated discontinuation.

2. Materials and Methods

The Contraceptive Choice Project (CHOICE) is a prospective cohort study that sought to decrease unintended pregnancy rates in the St. Louis region by providing no-cost reversible contraception as described previously [13]. Women in the study were 14–45 years of age, sexually active with a male partner in the past 6 months or planning to be sexually active in the next 6 months; at risk of unintended pregnancy (no permanent sterilization or hysterectomy) with no plans for pregnancy in the next 12 months; residents of the St. Louis region; and willing to start a new form of reversible contraception. Baseline demographic characteristics, reproductive history, and sexual behaviors were captured at time of enrollment. All women who enrolled received standardized contraceptive counseling on reversible contraceptive methods and baseline sexually transmitted infection (STI) testing. Follow-up telephone interviews were conducted at 3, 6, and 12 months after enrollment. The study was approved by the Washington University in St. Louis School of Medicine Human Research Protection Office and all participants provided written informed consent.

Of the 1,686 women enrolled in CHOICE who chose OCPs, patch, or ring at baseline, 1,510 women (90%) initiated use by 3 months and provided information on discontinuation. To mimic many insurance plans available in Missouri, participants were required to fill their prescriptions on a monthly basis without charge or copay at a local family planning clinic or one local grocery chain (82 locations) by presenting CHOICE ID cards. The grocery chain and the clinic sent claims data documenting the dispense date for contraceptive method refill to study staff at the end of each month.

To assess continuation from survey data, participants were asked, “Are you still using the method?” and, “Did you ever stop using the method?” The start month and year were recorded. If the participants reported stopping the method, the stop month and year were recorded, and also restart month and year if any. Continuation was defined as contraceptive use throughout the 12-month interval with no gap of use of one month or longer. It is considered a method stop if participant switched to a different contraceptive method. To measure continuation by claims data we obtained contraception start and stop dates from participants’ pharmacy records. The contraceptive start date represented the dispense date, the stop date was coded as 28 days after the dispense date. Patients were counted as continuing their contraceptive method if the gap between a stop date and subsequent start date was 30 days or less. We censored participants who were lost to follow-up at the time of their last contact.

We used ANOVA for continuous variables and χ² for categorical variables to compare baseline characteristics among OCP, patch, and ring users. Kaplan-Meier survival functions were used to estimate the probability of surviving the event (method discontinuation or unintended pregnancy) by certain time point, from which we calculated continuation rates and cumulative unintended pregnancy rates. We constructed 2x2 tables comparing continuation by survey versus pharmacy data and calculated Cohen’s kappa coefficient to measure the agreement between survey and pharmacy claims data on continuation status. Level of agreement by kappa statistic was based on guidelines by Landis and Koch [14]. Baseline characteristics of patients who were continuers by survey were stratified by whether or not they were also continuers by pharmacy claims data. We used Poisson regression to evaluate the relative risk of discontinuation because discontinuation was a common outcome. We used a stepwise selection process to identify factors associated with a discrepancy in continuation status among women who self-reported continuation yet their pharmacy claims data suggested discontinuation.

3. Results

Baseline characteristics of OCP, (n=769), patch (n=153), and ring (n=588) users are presented in Table 1. There were significant differences between the groups in terms of race, socioeconomic status, and past reproductive history. Patch users were significantly more likely than OCP and ring users to require public assistance (p<0.01) and to be uninsured (p<0.01). Patch users had more previous unintended pregnancies (p<0.01), and were significantly more likely to have had an abortion (p<0.01).

Table 1.

Baseline demographic and reproductive characteristics by chosen method among women in the Contraceptive CHOICE project

Characteristics	OCP (n=769)	Patch (n=153)	Ring (n=588)	P
Age	23.8±5.2	24.2±4.6	24.4±4.5	0.12
Race				<0.01
Black	336 (43.7)	89 (58.2)	223 (38.0)
White	370 (48.1)	53 (34.6)	310 (52.8)
Others	63 (8.2)	11 (7.2)	54 (9.2)
Education				<0.01
High school or less	199 (25.9)	45 (29.4)	108 (18.4)
Some college	345 (44.9)	73 (47.7)	256 (43.5)
College or graduate degree	225 (29.3)	35 (22.9)	224 (38.1)
Marital Status				0.06
Single/Never married	556 (72.3)	109 (71.2)	411 (69.9)
Married/Living with partner	183 (23.8)	32 (20.9)	158 (26.9)
Separated/Divorced/Widowed	30 (3.9)	12 (7.8)	19 (3.2)
Receives public assistancea	157 (20.4)	50 (32.9)	108 (18.4)	<0.01
Insurance				<0.01
None	336 (44.0)	83 (54.6)	221 (38.2)
Private	382 (50.0)	56 (36.8)	334 (57.7)
Public	46 (6.0)	13 (8.6)	24 (4.1)
Gravidity				<0.01
0	403 (52.4)	53 (34.6)	303 (51.5)
1	168 (21.8)	46 (30.1)	126 (21.4)
2	109 (14.2)	19 (12.4)	70 (11.9)
3 or higher	89 (11.6)	35 (22.9)	89 (15.1)
Parity				<0.01
0	570 (74.1)	88 (57.5)	415 (70.6)
1	120 (15.6)	39 (25.5)	107 (18.2)
2	57 (7.4)	16 (10.5)	44 (7.5)
3 or higher	22 (2.9)	10 (6.5)	22 (3.7)
No. of unintended pregnancies				<0.01
0	433 (56.5)	58 (37.9)	320 (54.4)
1	189 (24.7)	51 (33.3)	145 (24.7)
2	86 (11.2)	18 (11.8)	64 (10.9)
3 or higher	58 (7.6)	26 (17.0)	59 (10.0)
History of abortion	221 (28.7)	65 (42.5)	168 (28.6)	<0.01
History of STIb	222 (28.9)	66 (43.1)	226 (38.4)	<0.01
STI at baselinec	51 (7.1)	12 (8.4)	24 (4.3)	0.05

OCP, oral contraceptive pill; STI, sexually transmitted infection.

Data are mean±standard deviation or n (%).

^a“Self-reported current receipt of food stamps; vouchers from the supplementary nutritional program for women, infants, and children (WIC); welfare; or unemployment benefits.

^bReported a history of Chlamydia trachomatis, Neisseria gonorrhea, Trichomonas vaginalis, syphilis, herpes, or human immunodeficiency virus at study enrollment

^cTested positive for C. trachomatis, N. gonorrhea, or T. vaginalis at study enrollment.

All values in bold are statistically significant at p<0.05.

Continuation rates at 12 months by self-report were 59%, 45%, and 57% for OCP, patch, and ring, respectively. Using pharmacy claims, continuation rates were lower at 38%, 28%, and 37%, respectively. For all three methods, there was only moderate agreement between self-report and pharmacy claims (kappa and 95% confidence interval for OCP, patch, and ring respectively = 0.46 [0.40–0.52], 0.54 [0.39–0.68], and 0.54 [0.47–0.61]).

Table 2 shows the characteristics of participants who were continuers by self-reported survey data, stratified by whether or not they were continuers by pharmacy claims data. Reproductive and demographic characteristics associated with disagreement in continuation by self-report and pharmacy claims data include public insurance, higher parity, and history of STI. Conversely, higher educational attainment was associated with agreement in continuation status. Among women who continued their method by self-report and by pharmacy claims, the unintended pregnancy rate was 0.4%; however, those who discontinued by pharmacy claims had an unintended pregnancy rate of 4.9% (p<0.01).

Table 2.

Characteristics associated with discrepancy in continuation status among those who self-reported continuing their baseline method at 12 months

Characteristics	Continuer by Pharmacy Data		Discontinuation by Pharmacy Claims
	Yes (n=514)	No (n=363)	Relative Risk (95% CI)	Adjusted Relative Risk (95% CI)
Age	24.4±4.6	24.3±5.2	1.00 (0.98, 1.01)	-
Race
Black	180 (35.0)	177 (48.8)	1.39 (1.18, 1.63)	-
White	294 (57.2)	163 (44.9)	reference	-
Others	40 (7.8)	23 (6.3)	1.02 (0.72, 1.45)	-
Education
High school or less	91 (17.7)	96 (26.4)	reference	-
Some college	223 (43.4)	173 (47.7)	0.85 (0.71, 1.02)	0.93(0.77, 1.12)
College or graduate degree	200 (38.9)	94 (25.9)	0.62 (0.50, 0.77)	0.72(0.56, 0.91)
Marital Status
Single/Never married	342 (66.5)	260 (71.6)	reference	-
Married/Living with partner	155 (30.2)	92 (25.3)	0.86 (0.72, 1.04)	-
Separated/Divorced/Widowed	17 (3.3)	11 (3.0)	0.91 (0.57, 1.46)	-
Receives public assistancea	68 (13.3)	98 (27.0)	1.58 (1.35, 1.85)	-
Insurance
None	210 (41.3)	158 (44.3)	1.16 (0.98, 1.37)	1.00(0.84, 1.19)
Private	286 (56.2)	169 (47.3)	Reference	Reference
Public	13 (2.6)	30 (8.4)	1.88 (1.49, 2.36)	1.46(1. 15, 1.86)
Gravidity
0	284 (55.3)	160 (44.1)	reference	-
1	114 (22.2)	79 (21.8)	1.14 (0.92, 1.40)	-
2	59 (11.5)	58 (16.0)	1.38 (1.10, 1.72)	-
3 or higher	57 (11.1)	66 (18.2)	1.49 (1.21, 1.83)	-
Parity
0	400 (77.8)	227 (62.5)	Reference	Reference
1	71 (13.8)	81 (22.3)	1.47 (1.23, 1.77)	1.26 (1.04, 1.53)
2	31 (6.0)	36 (9.9)	1.48 (1.16, 1.90)	1.26 (0.96, 1.64)
3 or higher	12 (2.3)	19 (5.2)	1.69 (1.26, 2.28)	1.43 (1.05, 1.95)
Number of unintended pregnancies
0	300 (58.5)	183 (50.6)	Reference	-
1	124 (24.2)	91 (25.1)	1.12 (0.92, 1.36)	-
2	49 (9.6)	45 (12.4)	1.26 (0.99, 1.61)	-
3 or higher	40 (7.8)	43 (11.9)	1.37 (1.08, 1.73)	-
History of abortion	149 (29.0)	116 (32.0)	1.08 (0.92, 1.28)	-
History of STIb	140 (27.2)	136 (37.6)	1.31 (1.12, 1.53)	1.20(1.02, 1.42)
Any current STIc	20 (4.1)	27 (8.0)	1.43 (1.10, 1.85)	-

^a“Self-reported current receipt of food stamps; vouchers from the supplementary nutritional program for women, infants, and children (WIC); welfare; or unemployment benefits.

^bReported a history of Chlamydia trachomatis, Neisseria gonorrhea, Trichomonas vaginalis, syphilis, herpes, or human immunodeficiency virus at study enrollment

^cTested positive for C. trachomatis, N. gonorrhea, or T. vaginalis at study enrollment.

4. Discussion

Our study suggests that women overestimate their 12-month continuation of OCP, patch, and ring increasing their risk of unintended pregnancy. Self-reported continuation rates were comparable to previous reports [5–9]. Continuation rates by pharmacy claims in this study were similar to rates of “timely refills” at 12 months as reported from a large pharmacy database study, despite the previous study using a 14-day timely refill threshold [12]. However, our study is unique in directly comparing these rates. For all three contraceptive methods, the agreement between self-reported continuation rates and pharmacy-verified continuation rates was moderate.

The differences seen between self-report and pharmacy claims are driven by several factors. Social desirability bias, which is common in studies related to contraception and family planning, may lead to inflated continuation rates by self-report [15]. Women in our study may be particularly at risk of social desirability bias as the survey was conducted by the same research group that was providing their no-cost contraception. Additionally, women may not have interpreted gaps in their contraceptive use as discontinuation. For example, a woman who does not have a current partner may temporarily suspend use of her contraception without considering that she has discontinued the method. It is also possible that women received their contraceptives at a non-study pharmacy; thus their survey data would reflect continuation while pharmacy data would not. However, we believe this was uncommon. There were many participating pharmacy locations in the area (n=82), and women would have to cover the cost of their contraceptive at non-study pharmacies.

We defined discontinuation as non-use of more than 30 days. We selected this period for several reasons. First, when women reported starting or stopping a method by survey data, only month and year were recorded. Month level pharmacy data, including a 30-day gap constituting discontinuation, could then be readily compared to survey data. Second, brief interruptions in use, which many women would not consider method discontinuation, should not have affected survey or pharmacy continuation status.

Our analysis included detailed clinical information collected from participants. This allowed us to compare differences between subpopulations of women who reported continuation and did or did not fill their prescriptions as reported. We identified certain subgroups with discrepancy between continuation status by self-report versus pharmacy claims. Specifically, having public health insurance, higher parity, and history of an STI were all associated with discrepancy in continuation status according to claims data among self-reported continuers. Having public insurance and higher parity have all previously been linked to increased risk of contraceptive discontinuation [6]. These women may have impediments to accessing their contraception on a timely basis, or lack of understanding that gaps in usage can affect contraceptive effectiveness.

A limitation of our findings is that our study reflects a selected population of women at high-risk for unintended pregnancy willing to start a new method of contraception. Women who wanted to continue their current contraception were not eligible, which may have weighted our sample population toward lower continuation rates. However, our results are generalizable to women starting a new form of contraception or who desire to change methods. Our analysis includes 21 (1.4%) women who discontinued their method to become pregnant. We included these participants because our objective was to measure agreement between continuation rates regardless of reason for discontinuation. Reason for discontinuation was purposefully not assessed as part of this study, as it had previously been reported elsewhere [5].

We found many women report refillable contraceptive continuation despite data from prescription refills that do no support continued use. This disagreement was relatively constant for women who chose OCPs, the patch, or the ring. These results suggest that previously published continuation rates obtained by participant self-report may actually overestimate contraceptive continuation, especially among certain high-risk women. These discrepant continuation rates also highlight the importance of broad education on using refillable methods effectively and offering emergency contraception routinely as many women will not have “perfect use” of their method.

Implications.

This article directly compares contraception continuation rates by self-report and by pharmacy claims data. The study suggests that previously reported continuation rates from survey data overestimate specific method use.