Increased contraceptive knowledge is associated with reduced risk of unplanned pregnancy by more consistent use of highly effective contraceptives.1 However, overall contraceptive knowledge among U.S. women is low; at least half underestimate the effectiveness of contraceptives for pregnancy prevention.1–4 Decision aids such as posters are one tool that could be used to educate women about contraception; a Cochrane review found that decision aids such as posters could increase knowledge, help patients make decisions, and help them experience less conflict about those decisions.5
The Centers for Disease Control and Prevention (CDC) recommends that clinicians educate patients about contraceptive effectiveness and suggests their own contraceptive effectiveness poster as a potential tool.6 However, the CDC's poster may not improve knowledge of the risk of pregnancy with unprotected sex, which is an important risk factor for inconsistent or nonuse of contraceptives,3,7,8 because it does not include this information. Furthermore, the CDC poster's design may be difficult to interpret for women with low health literacy or numeracy (ie, facility with mathematics). The Institute of Medicine has declared designing educational materials for low health literacy and numeracy populations a key public health priority.9
We designed a patient-centered poster that is appropriate for women with low numeracy that includes information about the risk of pregnancy with unprotected sex. Our main hypotheses are that women who view the patient-centered poster will immediately show greater increases in their contraceptive knowledge, greater accuracy in their perceived pregnancy risk, and greater effectiveness in their contraceptive intentions than women who view the CDC poster.
Our intervention compared exposure to either the CDC10 or the patient-centered (Fig. 1) poster for as long as desired with a minimum of 1 minute (average 1.96 minutes for CDC, 1.79 minutes for patient-centered). The patient-centered poster was developed through cognitive interviews with 26 women aged 18–44 years living in North Carolina who spoke and read English and had ever had sex.11 In that study, the final version of the patient-centered poster was preferred over the CDC poster by women overall based on its ease of comprehension, relevance to their decision-making needs, and visual appeal.
For this study, we used Amazon Mechanical Turk to select a convenience sample of U.S. women aged 18–44 years who spoke and read English, were not pregnant or trying to conceive, and who had engaged in vaginal intercourse with a man in the past 3 months. Amazon Mechanical Turk is an online service that allows individuals to post surveys to be completed online for a fee.12 Data from Amazon Mechanical Turk users have been found to be as reliable or more reliable than data from other sources: workers have been consistently found to be attentive, their answers to questions consistent, and their answers no more or less truthful than in high-quality probability samples of the general population.13
We first screened for eligibility using a short online survey, for which participants were reimbursed $0.05. Eligible participants were invited to complete the full study survey online and reimbursed $3.60 on completion, equivalent to the federal minimum wage for their time. The survey was implemented in Qualtrics, which automatically randomized women to equal-sized groups. The baseline data collection, intervention implementation, and outcome assessment were all conducted within one survey and the researchers were blind to assignment. The study was approved by the University of North Carolina at Chapel Hill institutional review board (number 17-2955).
This study measured change in the mean scores for three primary outcomes: contraceptive knowledge, effectiveness of most likely contraceptive intended in the next year, and accuracy of perceived pregnancy risk. We gathered baseline and follow-up measures for each of these outcomes immediately before and after the intervention.
Contraceptive knowledge was measured objectively using the 25-item Contraceptive Knowledge Assessment.14 This produced a score between 0 (0% correct) and 25 (100% correct). Our contraceptive knowledge outcome was the change in this score between baseline and follow-up.
“Effectiveness of most likely contraceptive” was operationalized using a woman’s intention to continue using her current contraceptive and the contraceptive she reported being most likely to switch to were she to change methods in the next year. We first asked women at both baseline and follow-up: “Do you intend to use the same birth control method(s) that you are currently using for the next year?” If the woman said she intended to keep her contraceptive(s), the effectiveness of the most effective method she used in the past 3 months was used as her most likely contraceptive. The effectiveness of contraceptives was scored using the following World Health Organization-defined categories2: intrauterine devices, implants, and sterilization were considered highly effective (score=3, 0–1% annual failure rate); the pill, patch, ring, and injection were considered effective (2, 2–9% annual failure rate); condoms, withdrawal, fertility tracking, and other methods were considered less effective (1, 10–30% annual failure rate); and no method was its own category (0, 85% annual failure rate). If a woman said she did not intend to keep her current contraceptive, we used the effectiveness of the contraceptive she intended to use. We measured this with the question, “If you had to change to a new birth control method in the next year, which of the following methods would you consider using?” Participants selected each method they would consider and then ranked the selected methods from most to least likely method. Our “effectiveness of most likely contraceptive intended in next year” outcome was the difference between a woman’s score at baseline and follow-up.
Finally, accuracy of perceived pregnancy risk was assessed by comparing a woman's current contraceptive with her perceived pregnancy risk. Perceived pregnancy risk was measured using the following question: “What is your chance of getting pregnant this year?” with possible responses being very high (score=5, annual pregnancy risk greater than 50%), high (4, risk 25–50%), moderate (3, risk 5–25%), low (2, risk 1–5%), and very low (1, risk 1% or less). We assessed the accuracy of perceived risk based on the most effective birth control method a woman used in the past 3 months. In accordance with the World Health Organization categories,2 for highly effective methods, we coded an accurate perception to be very low risk; for effective methods, an accurate perception was low or moderate risk; for less effective methods, an accurate perception was moderate or high risk; for no method, an accurate perception was very high risk. An accurate perception was assigned a score of 1 and an inaccurate perception, 0. Our accuracy in perceived pregnancy risk outcome was the change in this score between baseline and follow-up.
Baseline data were collected on factors that might influence these outcomes. We measured prospective pregnancy intentions with the question, “Are you currently trying to get pregnant or avoid pregnancy?” with the response options: trying to get pregnant, would not mind getting pregnant, would not mind avoiding pregnancy, trying to avoid pregnancy, and do not know.15 We measured past pregnancy scares by asking: “Have you ever had a pregnancy scare; that is, thought you were pregnant when you did not want to be, but later discovered that you were not pregnant after all?” We measured numeracy using the Berlin single-item numeracy scale.16 This scale has been tested and validated to show that people who answer this question correctly are in the top 50% of the population in numeracy.16 We measured whether there were any contraceptives the woman could not use as a result of health or safety reasons using two questions. First, we asked the yes or no question: “Are there any types of birth control that you cannot use for health or safety reasons?”. If the woman responded “yes” to this question, she was asked the follow-up question: “If yes, which forms of birth control are you prevented from using? Mark all that apply.” Her options included all of the methods except withdrawal and no method. Data were also collected on the sexes of the woman’s past sex partners, whether she had ever seen the poster before, and whether there were any types of birth control the woman could not use for cost reasons. The following variables were measured using questions from the National Survey of Family Growth: biological sex, age, whether the participant was pregnant or trying to conceive, sexual intercourse in the past 3 months, education, time since first sex, and marital status. Finally, the following variables were measured using questions from the National Longitudinal Survey of Adolescent to Adult Health (Add Health): race and ethnicity (Wave V), income (Wave IV), relationship status (Wave IV), and health insurance type (Wave IV).
We first tested whether the demographic and other factors were balanced between our randomized groups using two-sample t tests and likelihood ratio tests as appropriate. We did not find any statistically significant imbalances for any of the variables. We conducted two-sample t tests on the change in the mean score for each of our outcomes to test whether each poster improved the three primary outcomes relative to baseline and in comparison with the other poster. We used the Bonferroni correction to account for multiple comparisons. Using the same methods, we also tested the hypotheses that the three prespecified subgroups (low numeracy, pregnancy scares, and no birth control) had greater increases in their mean scores for the patient-centered poster compared with the CDC poster. We chose these subgroups because the patient-centered poster was designed to appeal to the needs of these groups. Finally, because correct answers to some of the questions on the Contraceptive Knowledge Assessment were not given by either poster, we could determine the proportion of the change in contraceptive knowledge that was attributable to the posters. We did this by analyzing the change in contraceptive knowledge separately for questions that did and did not have the correct answer provided by either poster. All analyses were conducted in Stata 15.
For our power calculations, we assumed an α of 1% and a power of 80%. For our final analysis sample of N=936, comparing the two posters, we can detect a 3 percentage point difference in mean change in contraceptive knowledge (SD 0.1814), a 0.8 percentage point difference in accuracy of perceived pregnancy risk (SD 0.05), and a 6 percentage point difference in the mean change in effectiveness of most likely contraceptive (SD 0.3517).
Participants were enrolled between January 26 and February 13, 2018 (Fig. 2). Enrollment ended when our target enrollment goals were met.
To evaluate the representativeness of our sample, we descriptively compare the distributions of baseline factors in our study sample with their distribution in the 2013–2015 National Survey of Family Growth survey, weighted to represent the national population of U.S. women eligible for our study (Table 1). We found no significant differences between the randomized groups on any of these baseline characteristics, but there are differences between the study population and the National Survey of Family Growth sample. The study sample appears to be more educated, more white, more middle income, more likely to be cohabiting, less likely to be monogamous, more likely to have had female sexual partners, and less likely to be using effective contraceptives.
Table 2 shows descriptive results for our outcomes. Both groups started with a score of approximately 66% correct on the Contraceptive Knowledge Assessment. At baseline, the majority of women believed they were at very low risk of getting pregnant and only 23–24% of women had an accurate pregnancy risk perception. The majority of women (72%) using no method believed they had a low or very low chance of getting pregnant in the next year, despite the fact that 85 of 100 sexually active nonusers of contraceptives (or 164 of the 194 nonusers of contraceptives in our study) will conceive over the course of a year.18 High percentages of women at baseline in both poster groups (64% CDC, 63% patient-centered) reported they were likely to use no or less effective contraceptives.
Table 3 shows the results of our main hypothesis tests. Of our three main hypotheses, we found that the patient-centered poster was only significantly more effective than the CDC poster at improving contraceptive knowledge (P<.001). Both posters significantly improved contraceptive knowledge relative to baseline (P<.001). The patient-centered poster improved contraceptive knowledge scores by 6.4 percentage points, or 1.6 additional correct questions, and the CDC poster improved scores by 3.6 percentage points, or 0.9 additional correct questions on average.
The results for the analyses testing the change in contraceptive knowledge for questions that were and were not addressed by the posters can also be found in Table 3. We found a smaller increase in the mean percent correct for questions that were not addressed by either poster (1.8 percentage points for CDC and 2.1 percentage points for patient-centered) as compared with questions that were addressed by the posters (5.8 percentage points for CDC and 11.9 percentage points for the patient-centered poster). The magnitude of the change in the mean score for questions that were not addressed by either poster did not significantly differ between the posters.
Although neither poster performed significantly better than the other at improving the score measuring the effectiveness of the most likely contraceptive intended in the next year, both posters improved this score compared with baseline by 3 percentage points (P<.001) (Table 3). This increase corresponds to between 1 and 17 out of 100 women increasing the effectiveness of their most likely contraceptive by one category (ie, moving from no method to a less effective method).
The results in our subgroup analyses of women with pregnancy scares, low numeracy, or no current contraceptive were similar for all outcomes (results available from corresponding author). Participants reported no harms or unintended effects.
Of our three primary outcomes, we found that the patient-centered poster was only significantly more effective than the CDC poster at improving contraceptive knowledge. There were no statistically significant differences between the CDC and patient-centered posters' effects on perceived risk of pregnancy and the score measuring effectiveness of the most likely contraceptive intended for the next year. However, relative to baseline, we found that both the CDC and patient-centered posters significantly improved contraceptive knowledge and a score measuring effectiveness of the most likely contraceptive intended for the next year. We also found that the increases in contraceptive knowledge were attributable to the posters themselves.
A Cochrane review19 identified interventions that increased contraceptive knowledge, including two that tested educational tables2 or charts.20 These two studies reported 14–37 percentage point increases, depending on the chart, for two questions asking participants to select the more effective contraceptive from a pair of methods.2,20 However, compared with past studies that only assessed a small number of items tailored to the intervention,19 ours comprehensively assessed the effect of these educational posters on contraceptive knowledge. Our study also found significant effects on women's intended contraceptive, which the Health Belief Model21 suggests is likely to be more strongly associated with contraceptive behavior than contraceptive knowledge. Our results held in subgroups of participants who had low numeracy, prior pregnancy scares, and who do not use birth control, who may have greater challenges understanding information about contraception. We also saw these results despite participants only being exposed to the poster passively and for a very short period of time, similar to what they might experience if viewing the posters while waiting in a clinician’s office.
Our results are not generalizable to the general population of U.S. women because participants were recruited online using Amazon Mechanical Turk, meaning that in addition to our other inclusion criteria, participants also were by definition internet users. It is notable that our sample is more white, educated, and wealthy than the U.S. female population. However, the differences between our study sample and the National Survey of Family Growth sample are similar to the differences between Americans who use the internet and the general U.S. population.22 In the United States 99% of 18–29 year olds and 96% of 30–45 year olds use the internet.23 Our study sample also appears to be more knowledgeable about contraception than the general population14; because of this, it is possible that our findings underestimate the effect of posters on contraceptive knowledge. On the other hand, because our sample is better educated than the general U.S. population, it is possible that they were more capable of learning from the posters, and our findings may overestimate the effect of posters on contraceptive knowledge. Finally, our study does not assess the effect of these posters on actual behaviors. Because of this limitation and the lack of previous studies using the Contraceptive Knowledge Assessment in a clinical population, we are unable to comment on the clinical significance of the 3 percentage point difference in the increase in contraceptive knowledge that we observed between the CDC and patient-centered posters. However, we did measure contraceptive intentions, which have been shown to be a good predictor of behavior.24,25 The effect of these posters on actual contraceptive choices in clinical practice should be studied in future research.
Clinicians often struggle to educate their patients about the multitude of important health topics in the limited amount of time they have during appointments.26 This study tested two inexpensive tools to educate patients about contraception independently from a health care provider and found that they effectively increase contraceptive knowledge and intentions to use more effective contraceptives. Using these posters in practice could allow doctors to spend more of their time answering questions about the patient’s specific contraceptive needs rather than educating them on the basics of how each method works and how effective it is.
1. Frost JJ, Lindberg LD, Finer LB. Young adults' contraceptive knowledge, norms and attitudes: associations with risk of unintended pregnancy. Perspect Sex Reprod Health 2012;44:107–16.
2. Steiner MJ, Dalebout S, Condon S, Dominik R, Trussell J. Understanding risk: a randomized controlled trial of communicating contraceptive effectiveness. Obstet Gynecol 2003;102:709–17.
3. Kaye K, Suellentrop K, Sloup C. The fog zone: how misperceptions, magical thinking, and ambivalence put young adults at risk for unplanned pregnancy. Washington, DC: The National Campaign to Prevent Teen and Unplanned Pregnancy; 2009.
4. Biggs MA, Foster DG. Misunderstanding the risk of conception from unprotected and protected sex. Womens Health Issues 2013;23:e47–53.
5. O'Connor AM, Bennett CL, Stacey D, Barry M, Col NF, Eden KB, et al. Decision aids for people facing health treatment or screening decisions. The Cochrane Database of Systematic Reviews 2009, Issue 3. Art. No.: CD001431. DOI: 10.1002/14651858.CD001431.pub2.
6. Gavin L, Moskosky S, Carter M, Curtis K, Glass E, Godfrey E, et al. Providing quality family planning services: recommendations of CDC and the U.S. Office of Population Affairs. MMWR Morb Mortal Wkly Rep 2014;63:1–54.
7. Polis CB, Zabin LS. Missed conceptions or misconceptions: perceived infertility among unmarried young adults in the United States. Perspect Sex Reprod Health 2012;44:30–8.
8. Eisenberg DL, Secura GM, Madden TE, Allsworth JE, Zhao Q, Peipert JF. Knowledge of contraceptive effectiveness. Am J Obstet Gynecol 2012;206:479.e1–9.
9. Kindig DA, Panzer AM, Nielsen-Bohlman L, editors. Health literacy: a prescription to end confusion. Washington, DC: National Academies Press; 2004.
11. Anderson S, Barry M, Frerichs L, Wheeler SB, Halpern CT, Kaysin A, et al. Cognitive interviews to improve a patient-centered contraceptive effectiveness poster. Contraception 2018;98:528–34.
12. Paolacci G, Chandler J, Ipeirotis PG. Running experiments on Amazon Mechanical Turk. Judgm Decis Mak 2010;5:411–9.
13. Chandler J, Shapiro D. Conducting clinical research using crowdsourced convenience samples. Annu Rev Clin Psychol 2016;12:53–81.
14. Haynes MC, Ryan N, Saleh M, Winkel AF, Ades V. Contraceptive Knowledge Assessment: validity and reliability of a novel contraceptive research tool. Contraception 2017;95:190–7.
15. Schwarz EB, Lohr PA, Gold MA, Gerbert B. Prevalence and correlates of ambivalence towards pregnancy among nonpregnant women. Contraception 2007;75:305–10.
16. Cokely ET, Galesic M, Schulz E, Ghazal S, Garcia-Retamero R. Measuring risk literacy: the Berlin numeracy test. Judgm Decis Mak 2012;7:25–47.
17. Antonishak J, Kaye K, Swiader L. Impact of an online birth control support network on unintended pregnancy. Soc Mar Q 2015;21:23–36.
18. Trussell J. Contraceptive failure in the United States. Contraception 2011;83:397–404.
19. Lopez LM, Steiner M, Grimes DA, Schulz KF. Strategies for communicating contraceptive effectiveness. The Cochrane Database of Systematic Reviews 2008, Issue 2. Art. No.: CD006964. DOI: 10.1002/14651858.CD006964.pub2.
20. Steiner MJ, Trussell J, Mehta N, Condon S, Subramaniam S, Bourne D. Communicating contraceptive effectiveness: a randomized controlled trial to inform a World Health Organization family planning handbook. Am J Obstet Gynecol 2006;195:85–91.
21. Hall KS. The Health Belief Model can guide modern contraceptive behavior research and practice. J Midwifery Womens Health 2012;57:74–81.
22. Hillygus DS, Jackson N, Young M. Professional respondents in non-probability online panels. In: Callegaro M, Baker RP, Bethlehem J, Göritz AS, Krosnick JA, Lavrakas PJ, editors. Online panel research: a data quality perspective. West Sussex (UK): John Wiley and Sons; 2014. p.219–37.
24. Kothandapani V. Validation of feeling, belief, and intention to act as three components of attitude and their contribution to prediction of contraceptive behavior. J Pers Soc Psychol 1971;19:321–33.
25. Kim MS, Hunter JE. Relationships among attitudes, behavioral intentions, and behavior: a meta-analysis of past research, part 2. Commun Res 1993;20:331–64.
26. Morrison I, Smith R. Hamster health care. BMJ 2000;321:1541–2.