An official website of the United States government
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.
Christine Dehlendorf, MD, MAS, Eric Vittinghoff, PhD, Judith Fitzpatrick, Edith Fox, MPH, Kelsey Holt, ScD, Reiley Reed, MPH, Maria Paula Campora, MD, Abby Sokoloff, MPH, and Miriam Kuppermann, PhD, MPH.
Author Information and AffiliationsThe quality of contraceptive counseling has important implications for women's experience of family planning care and reproductive health outcomes. Studies suggest substantial room for improvement in the patient-centeredness of contraceptive counseling. Decision aids can help facilitate shared decision-making (SDM), a patient-centered counseling approach that is appropriate for preference-sensitive decisions such as those about contraception. We developed a decision support tool, My Birth Control, to support SDM in contraceptive counseling.
We examined (1) whether My Birth Control increased contraceptive continuation and improved patient experience and knowledge compared with usual care, (2) how implementation of My Birth Control affected counseling delivered by providers, and (3) how implementation of My Birth Control impacted providers and health care settings.
We conducted a cluster randomized controlled trial of contraceptive counseling with My Birth Control compared with usual care at 4 clinics serving low-income, racially/ethnically diverse populations in the San Francisco Bay Area. Participating providers (nurse practitioners, nurse midwives, physician assistants, and health educators) were randomized to the tool or to usual care. Patients seeing these providers who reported to research staff that they wished to discuss changing or starting a contraceptive method during their visit were recruited for participation. Consenting participants then either did or did not interact with the tool, according to the arm of their provider. Interacting with My Birth Control consisted of viewing the interactive web-based tool, which provided education about contraceptive methods and elicited preferences related to contraceptive use. Following completion of the tool, a printout was generated with information about these preferences, and was given to the provider for use during the visit. Patient participants were then seen by their provider for care, including contraceptive counseling. Our primary outcome was contraceptive continuation at 7 months. Secondary outcomes included patient experience of counseling, quality of decision-making, and knowledge; qualitative evaluation of counseling quality; and provider experience with counseling patients who had used the tool. To assess patient outcomes, we used baseline, 4-month, and 7-month surveys. To assess provider outcomes, we used baseline and follow-up surveys, clinic visit audio recordings, and provider interviews. We assessed clinic staff outcomes using staff focus groups. We used mixed effects regression models for quantitative outcomes and explored potential modification of the effect of treatment assignment for outcome variables with evidence of a statistically significant overall intervention effect by 5 prespecified factors: patient age, race and ethnicity (combined), language, parental education, and clinic. We used thematic coding of interviews, focus groups, and audio recordings for qualitative outcomes.
Twenty-eight providers and 758 patients enrolled in the trial. In addition, we audio-recorded 64 patient visits before the start of the trial and 31 patient visits after the trial. Patient intervention participants (56.6%) and control participants (59.6%) did not differ in contraceptive continuation at 7 months (odds ratio [OR], 0.89; 95% CI, 0.65-1.22). More intervention participants (66.0%) reported the highest level of interpersonal quality of counseling compared with control participants (57.4%) (OR, 1.45; 95% CI, 1.03-2.05). More intervention participants reported complete satisfaction with information given about side effects (83.5%) compared with control participants (75.7%) (OR, 1.61; 95% CI, 1.11-2.33). More intervention participants (50.5%) gave the highest rating on the informed decision subscale of the Decisional Conflict Scale (DCS) compared with control participants (43.2%) (OR, 1.34; 95% CI, 1.00-1.80). We found no modification of these effects by patient characteristics, including whether they used the tool in English or Spanish. Providers reported a positive experience with My Birth Control and low impact on their work setting. Audio recordings of visits showed providers engaged more with patient preferences during visits in which patients used the tool, including preferences for methods and method characteristics.
Our results indicate that patients who use My Birth Control experience more patient-centered counseling and higher decision quality than those who receive usual care. The tool did not impact contraceptive continuation, but because it improved patient experience and was acceptable to providers, this study supports use of My Birth Control to enhance patient-centeredness in counseling.
Limitations of our trial include the potential for contamination across groups and lack of blinding of patients and providers. This trial examined routine counseling, so we need further research to understand use of the tool by women with more complex counseling needs.
In the United States, use of contraception is almost universal, with 99% of women who have had sexual intercourse with a man reporting they have used a contraceptive method at some point.1 Interaction with providers of family planning care is a common experience for women of reproductive age, because many methods of birth control—pills, sterilization, intrauterine devices (IUDs), implants, patches, rings, implants, and injections—require consultation with a health care provider, either to receive a prescription or to have a method placed. A nationally representative survey of women of reproductive age showed that over 50% of sexually active women reported receiving birth control-related care in the past 12 months.2
Because family planning care is so common, the quality of this care has implications for women's experience of health care and for their reproductive health outcomes. The quality of contraceptive counseling is 1 particularly important aspect of quality, given the uniquely personal nature of considerations of reproduction and sexuality that are integral to discussions about pregnancy prevention. Providing patient-centered counseling that focuses on women's needs and preferences can enable women to choose a method they feel fits within the context of their lives. Further, patient-centered counseling is associated with contraceptive continuation.3 Whether women receive patient-centered contraceptive counseling also has implications for women's trust in reproductive health providers,4 in that women who do not feel their preferences for contraception are valued and prioritized are less likely to access reproductive health care in both the short and the long term.5
Studies in the United States indicate substantial room to improve the patient-centeredness of contraceptive counseling. In qualitative research, women report they highly value comprehensive information and decision-making support.4 However, studies looking at their actual experiences found that women felt they were often unable to discuss their concerns and did not receive sufficient information.4,6-9 Quantitative studies have also found that many women express dissatisfaction about the patient-centeredness and adequacy of counseling.2,8-10 Analysis of audio recordings of contraceptive counseling has confirmed the need for improvement, with clinicians inconsistently eliciting or engaging with patient experiences and preferences during counseling.11,12
Shared decision-making (SDM) is increasingly recognized as a patient-centered counseling approach that is particularly appropriate in preference-sensitive decisions,13 which are decisions for which the best option is highly dependent on the patients' values and preferences. In this model of interaction between patients and providers, providers contribute their medical knowledge, while patients provide expertise on their own values and preferences.13
The choice of a contraceptive method is an archetypal preference-sensitive decision. Multiple contraceptive methods are medically appropriate for most women, and women have strong preferences for the characteristics of contraceptive methods.14,15 As such, facilitating SDM in contraceptive counseling holds promise to improve patient-centered contraceptive care.16 Qualitative research also supports the use of SDM in contraceptive counseling. For example, studies show that many women in the United States desire interactive engagement with health care providers,2 and that satisfaction with a chosen contraceptive method and with counseling experiences is associated with provider engagement in the decision-making process.16
Decision support tools are 1 means to enable SDM. These tools have received attention in the broader medical literature, where studies have shown they have a positive impact on knowledge, values-congruent choices, and risk perception.17 Decision support tools for contraceptive decision-making may help women identify their preferences and values and compare contraceptive methods accordingly. Such tools also encourage providers to structure their decision support around women's expressed preferences.
Decision support tools have the added advantage of addressing several challenges in contraceptive counseling. One prominent challenge is the need to provide comprehensive counseling about a range of options within the limited time of a clinic visit.18-20 An interactive decision support tool can provide individualized information to a patient about her options, ensure her provider is aware of her preferences, and support more efficient interaction between patient and provider. Another common challenge faced by family planning providers is the prevalence of patient misconceptions about contraceptive methods.21-23 A decision support tool provides structured and evidence-based education, addressing this challenge in an objective and efficient manner.
To facilitate contraceptive counseling using SDM, we developed My Birth Control, an interactive web-based tool delivered on tablet computers. We designed the tool to be used immediately before a health care visit about contraception. By sharing information with women about their options and helping them consider which method characteristics are most important to them, My Birth Control informs and empowers women in their interactions with providers (Figure 1).24 In addition, by giving providers information about women's preferences for contraceptive methods as a printout, providers are prompted to engage with women about these preferences. The ultimate goal of My Birth Control is to improve women's experience of contraceptive counseling and to help them select contraceptive methods that are consistent with their values and preferences.
Hypothesized Pathway Between Use of My Birth Control and Patient-Centered Care.
To design My Birth Control, we used best practices for the development of decision support tools25 and patient and provider stakeholder input. We incorporated research on women's preferences for and experiences with birth control counseling, including the documented value of written content to supplement verbal communication,3,26-28 use of visuals,26-28 and decision-making support.3,26,29 A description of the process of developing and pilot testing My Birth Control has been previously published.24 My Birth Control modules contain information about the 5 areas most relevant to the choice of a contraceptive method: effectiveness, side effects, return to fertility, mode of administration, and frequency of administration.14 We designed each section to provide an overall understanding of the range of options, as well as individual methods' characteristics, allowing users to reflect on the method characteristics that align with their preferences. To elicit their preferences for method characteristics, users complete an interactive values clarification exercise. They also answer health history questions related to their medical eligibility for different contraceptive methods. My Birth Control then shows a summary page depicting how users' preferences relate to the characteristics of specific methods. This page helps users identify methods they wish to discuss with their provider and questions they could ask. Providers receive a printed copy of this summary page and women's identified preferences for method characteristics.
In this study, our objective was to evaluate the impact of My Birth Control from both patient and provider perspectives. Our specific aims were to determine the following:
Our primary outcome was continuation of the contraceptive method chosen at 7 months after the visit, with a secondary outcome of continuation of contraception at 4 months. We measured contraceptive continuation at 4 and 7 months to account for continuation of the contraceptive injection (Depo Provera), which is effective for 15 weeks. We selected contraceptive continuation as our primary outcome for 2 reasons. First, we conceptualized it as an indicator of whether counseling enabled women to select a method that met their needs for an extended period of time. Second, it has public health implications regarding prevention of unintended pregnancy. As described in the Discussion section, our conceptualization of the patient-centeredness of this primary outcome evolved during our project.
As described in the Methods section, we assessed many secondary outcomes, including patient experience of counseling and knowledge of contraception. We performed qualitative and quantitative analyses to investigate patient, provider, and clinical staff outcomes.
Before receiving PCORI funding, we engaged with stakeholder groups of patients and providers throughout the process of developing My Birth Control and designing the trial. We met on a quarterly basis with a 5-member patient stakeholder group. We formed this group from a patient advisory council at a local family planning clinic that serves a diverse low-income population. The patient stakeholder group provided input to ensure the tool's tone and content was appropriate for and met the needs of our target population—women aged 15-45 seeking contraceptive services. The group also assisted with planning the evaluation of the tool. All members of the patient stakeholder group were women who had sought contraceptive services at the clinic, and thus were members of our target population.
We also met quarterly with a 5-member provider stakeholder group, which included nurse practitioners, nurses, and counselors who provided contraceptive services in several San Francisco Bay Area clinics serving diverse low-income clients. None of these clinics were recruitment sites for our trial. One provider in the group worked at the clinic that hosted the patient stakeholder group. The provider stakeholder group assisted with structuring the intervention to optimize feasibility for clinics. Our team knew members of the provider stakeholder group due to their participation in our previous studies.
During our PCORI funding, we continued to engage with our existing patient and provider stakeholder groups. We collaborated with these groups to finalize the tool, conduct research, and plan for dissemination. To avoid potential contamination resulting from this collaboration, neither the clinic that hosted the patient stakeholder group nor any clinics employing provider stakeholders participated as study sites.
To balance stakeholder input from patients and providers, we met with both groups with equal frequency. We incorporated diverse patient and provider perspectives into tool design, evaluation, and real-world use. We prioritized patient perspectives by hiring patient stakeholder representatives who consulted on project activities more regularly and intensively than either stakeholder group. We describe the contributions of the stakeholder groups, patient stakeholder representatives, and organizational stakeholders below.
The 5-member patient stakeholder group voiced patient perspectives on diverse aspects of the project, including postdevelopment finalization of the tool and all research activities. The group recommended ways to streamline text-heavy sections of the tool and develop a more user-friendly navigation system. During the development of our trial protocol, the group helped identify outcomes of interest. For example, group members recommended we include outcomes focused specifically on the patient experience of counseling.
In preparation for study recruitment, the patient stakeholder group advised on sensitive issues related to participant privacy and comfort. For example, the group discussed with research staff how to appropriately approach patients about the study in waiting rooms and to ask patients a sensitive eligibility question about literacy level without stigmatizing low literacy levels. This group's advice consistently helped the research team build and maintain positive relationships with patient participants throughout the trial and improved the rigor of the trial by helping the research team more accurately assess patient eligibility criteria.
In fall 2015, the patient advisory council was dissolved by the clinic for reasons unrelated to this project. The patient stakeholder representatives, whose work is described below, consulted on this project over the entire project duration.
Our 3 patient stakeholder representatives were members of the original patient stakeholder group. They had an in-depth consulting role, working an average of a half day per week on the project. They attended weekly research team meetings. In preparation for the trial, they helped review and edit patient-facing documents. During the trial, they often observed and assisted research staff with in-clinic recruitment and provided feedback on patient interactions. During analysis, the representatives were involved in interpretation of results. They contributed to dissemination efforts by helping brainstorm possible implementation sites and strategies for introducing the tool to new providers and patients.
The provider stakeholder group participated in quarterly meetings during the project. Before recruitment began, provider input led to changes in tool content. For example, providers shared their experiences of patients' concerns about the safety of hormonal contraceptive methods that cause amenorrhea. In response, we added information to the tool to assure users about the safety of amenorrhea caused by contraception. The provider stakeholder group improved trial processes by reviewing and editing surveys and interview guides for providers and focus group guides for clinic staff. Their input helped us gather information about provider and clinic staff interaction with and experience of the tool. This group contributed to dissemination by recommending a toolkit on the decision support tool for providers. They recommended the toolkit include an instructional video and job aid to assist with implementation. In sum, provider stakeholder contributions strengthened the tool, its evaluation, and our plans for dissemination, and enhanced the relevance of this project to family planning providers and patients.
Organizational stakeholders on this project included The National Campaign to Prevent Teen and Unplanned Pregnancy (The National Campaign), with whom we collaborated to design and improve the tool; the California Family Health Council (now Essential Access Health), who gave input during the project on how the tool could best reach women receiving publicly funded contraceptive services; and local Planned Parenthood affiliates, who provided input on designing a tool that could scale up to large clinical organizations. In addition to communicating with these organizations throughout the trial, we are now engaged in conversations regarding optimal dissemination strategies. We are also engaged with city public health departments in Baltimore and San Francisco regarding dissemination opportunities.
We aimed to evaluate the impact of the decision support tool My Birth Control on patients' contraceptive use and experience of counseling. We also aimed to evaluate the impact of My Birth Control from the perspective of providers and clinic staff implementing it. To achieve these aims, we conducted a cluster randomized controlled trial comparing patients seen by providers randomized to implement My Birth Control and patients seen by providers randomized to continue their standard counseling approach. Using surveys, in-depth interviews, focus groups, and audio recordings of clinic visits, we evaluated the impact of the tool on patient and provider outcomes.
Our cluster randomized controlled trial compared contraceptive counseling with My Birth Control with usual counseling. Our experimental design allowed us to isolate the effect of the tool on our outcomes of interest. We randomized health care providers, stratified by clinical site. We decided to randomize by provider rather than by patient to reduce the potential for contamination between study arms. My Birth Control is designed to increase provider counseling engagement with patients, and we hypothesized that the tool's effect on counseling could extend to future patients even when not in use. We conducted qualitative assessment of the tool using audio recordings of clinic visits with patients who had interacted with the tool and interviews and focus groups with providers and clinic staff.
To ensure inclusion of low-income women and racially and ethnically diverse populations, we recruited providers and women at 4 clinics serving these populations in the Bay Area. Sites included a variety of types of health care organizations: a freestanding family planning clinic that belongs to a large nonprofit family planning care organization, a Department of Public Health primary care clinic, a college student health center, and a hospital outpatient clinic. This diversity of sites serving women of reproductive age supported maximum diversity of patient participants and clinic implementation contexts.
Research staff approached providers—including nurse practitioners, nurse midwives, physician assistants, counselors, and health educators— individually or in staff meetings to invite them to participate in the study. Provider inclusion criteria were providing family planning counseling and planning to remain in the job for 6 months. We randomized providers to intervention or control arms with a 1:1 allocation ratio using a random number generator tool in Microsoft Excel. Because we stratified randomization by clinic, by chance we had 2 more providers in the intervention arm than in the control arm. We invited intervention providers to participate in detailed interviews at the end of the trial.
We invited clinic staff—including front-desk staff and other individuals in nonmanagerial administrative roles—to participate in focus group discussions upon completion of the trial. In the discussions, we assessed the extent to which tool implementation affected clinic flow and other clinic processes. Inclusion criteria for clinic staff were having patient contact with additional responsibilities other than family planning counseling.
Our target population was low-income patients receiving contraceptive counseling in the Bay Area. We targeted the intended audience for My Birth Control through the following inclusion criteria: 15 to 45 years of age; desire to discuss starting or switching a contraceptive method at one of the participating clinics; history of sexual activity with men; and ability to speak, read, and understand English or Spanish. For the audio-recording component of the trial, we excluded those who communicated with their provider in Spanish to facilitate qualitative analysis.
Exclusion criteria included previous enrollment in the trial or previous use of the tool; being at the visit for an IUD or implant insertion, as we presumed these individuals had already received counseling and selected their method; and being infertile, currently pregnant, or desiring pregnancy sometime within the next 7 months. After enrollment, we excluded women if they found out they were pregnant during the enrollment visit. Target sample size was 758 (see Results section for details).
To recruit patients, research staff reviewed the clinic's daily schedule. Research staff communicated with clinic staff to best identify patients to approach and screen for study participation. How potential participants were identified varied depending on the specific processes and policies of the clinic, including the extent to which the purpose of the visit was specified as part of the scheduling process, and whether contraceptive counseling was typically offered as part of visits in which other care was the primary objective. Research staff approached potentially eligible patients in clinics. In one of the clinics, front-desk staff also gave flyers to women arriving for family planning care to inform them of the study before research staff approached. Reasons for nonparticipation were not systematically collected.
Before patient participant recruitment, providers randomized to the intervention arm received a brief orientation to My Birth Control. The orientation described the motivation for developing the tool, explained the modules and how to explore them, and provided logistical information about receiving the tool printout. We encouraged intervention providers to incorporate the tool printout into their counseling however they felt appropriate.
Whether a patient interacted with the tool was based on the study arm assignment of the provider she was scheduled to see during her visit. When an enrolled patient participant was scheduled to see an intervention arm provider, research staff gave her the tool on a tablet computer immediately before her appointment. After the patient finished using the tool, research staff gave her tool printout to the provider.
Providers randomized to the control arm provided usual care, which consisted of contraceptive counseling as typically provided. Usual care was ascertained through a prior study that analyzed audio recordings of counseling visits in 3 of the 4 trial sites.12 In this study, contraceptive counseling was characterized by a lack of engagement related to patient preferences for method characteristics.
Our primary outcome was a yes/no indicator of contraceptive continuation at 7 months follow-up. We also assessed contraceptive continuation at 4 months. At both time points, continuation met these 2 conditions: (1) participants started to use the method they selected at baseline by their 4-month survey, and (2) participants continued to use that method without a gap of more than 4 consecutive weeks. For 4-month continuation, we considered participants to have continuous use even if they were not currently using the method at the time of the 4-month survey when they reported in the 7-month survey that the gap lasted 4 weeks or less.
We assessed multiple secondary patient outcomes: patient experience, satisfaction, and decision-making; knowledge of contraception; use of and interest in highly effective contraception; and unintended pregnancy. All patient outcomes were patient reported, and we summarize them in Table 1.
Patient Study Outcomes.
To understand any effect of the tool on provider experience, we measured change in provider burnout before and after the intervention in provider surveys using the Maslach Burnout Inventory (MBI).35 The MBI is a 22-item measure of workplace burnout among human services workers with subscales in 3 domains: emotional exhaustion (subscore range = 0-54), depersonalization (subscore range = 0-30), and personal accomplishment (subscore range = 0-48). Additionally, we used semistructured interviews to explore intervention providers' experience using the My Birth Control tool.
We audio-recorded a subset of participant counseling sessions to qualitatively assess differences in counseling interactions with and without tool use. Patient participants for audio recordings were identified and recruited in the same manner as described for participants in the trial. We examined both provider and patient behaviors during counseling.
We sought to understand the tool's impact on clinic resources, an important consideration for dissemination and future use of the tool. To measure the total minutes of the participant's visit to the clinic and the duration of consultation with the provider, we recorded the times of patient arrival to clinic, the start of provider–patient consultation, and the end of the provider–patient consultation. We obtained these times through direct observation of patient and/or provider entry into and exit from the examination room, or through referral to patient electronic medical records, depending on the clinic site. We also recorded the number of minutes participants spent interacting with the tool using website analytics. In focus groups with clinic staff after patient recruitment, we explored the impact of the tool on clinic flow and processes.
Providers and patients provided data at baseline and patients completed follow-up activities at 4 and 7 months. We chose the follow-up time frame to account for the 15-week period of effectiveness of a contraceptive injection. By initiating follow-up of participants at 4 months (approximately 17 weeks) and 7 months (approximately 30 weeks), we increased our ability to detect discontinuation of the contraceptive injection either before the first follow-up survey or between the first and second follow-up surveys. Providers and clinic staff also provided data through participation in interviews and focus groups. Figure 2 shows the time frame for study activities.
Data Collection Time Points.
Patient participants completed surveys before and after their baseline visit. At baseline, if participants left the clinic before completing a postvisit survey, we attempted to reach them by phone to complete the survey within 48 hours of enrollment. If 48 hours passed and they had not completed the survey, we attempted to reach them for up to 4 weeks to complete shortened surveys without time-sensitive questions about patient experience and decision-making. At 4 and 7 months postvisit, we made contact attempts for up to 4 weeks using each participant's preferred mode of communication (phone, short message service, or email) to complete follow-up surveys. Research staff were blinded to study arm. We considered a participant lost to follow-up when the 7-month follow-up window ended or if the participant withdrew. Reasons for withdrawal included desiring not to be contacted any longer.
We collected 64 audio recordings of visits before randomization and 31 visits following completion of the trial. We collected pre-randomization recordings for all providers to prevent a cointervention effect, as providers being audio-recorded could elicit self-reflection and improvement in contraceptive counseling. We collected posttrial audio recordings only from intervention providers. Providers completed baseline surveys when they enrolled and follow-up surveys when they completed their participation in the study. We conducted semistructured interviews with providers randomized to use the tool and focus groups with clinic staff after completion of participant recruitment.
For our quantitative sample size, we based our calculation on our primary outcome, contraceptive continuation. Using data from our previous studies in this population,3 we estimated that 50% of women in our control group would continue with their contraceptive method at 7 months and that the intracluster correlations of this outcome by provider and by clinic would be ~0 and 0.021, respectively. Accounting for clustering and 20% loss to follow-up, as well as higher levels of continuation at clinics that provide IUD insertion, we estimated a sample of 758 patients would provide 80% power in 2-sided tests with a type I error rate of 5% to detect a 12-percentage-point increase in continuation. This sample size would also provide 80% power to detect small-to-moderate but clinically significant differences in secondary outcomes.
We conducted an intention-to-treat analysis, including all participants in their assigned groups. We excluded participants who became ineligible when they found out they were pregnant during the enrollment visit. We included participants with multiple enrollments only for their first enrollment visit.
We examined missing data for all independent and dependent variables. We imputed all missing values using 20-fold multiple imputation (MI), based on iterative chained equations. We calculated summary effect estimates, averaged over the 20 imputed data sets, as well as CI and P values, using standard methods that account for imputation error. To examine potential sensitivity to violations of the untestable MI assumption that data are missing at random, given observed outcomes and covariates, we assessed evidence for modification of the treatment effect on 7-month outcomes by baseline predictors of loss to follow-up. We also determined whether adjustment for baseline correlates of loss to follow-up affected the interpretation of marginal treatment effects.
We assessed imbalance in provider characteristics by treatment assignment using Fisher exact tests. We assessed imbalance in patient participant characteristics by treatment assignment using regression models controlling for site and clustering by provider.
We ran mixed effects logistic and linear models with random effects for provider assignment and fixed effects for site and arm assignment for prespecified binary and continuous patient outcome variables. We modeled ordinal and nominal outcomes using generalized estimating equations because of limitations in the software we used for multiply imputed outcomes. We conducted sensitivity analyses using complete case analysis for all outcomes to compare results with those using imputed data sets. We dichotomized outcomes measured using Likert scales, as well as the Decisional Conflict Scale (DCS) and subscales, as the highest possible score compared with all other scores. For the regression analysis of provider burnout subscale scores, we controlled for site and used bootstrapping for inference because of small sample sizes. We considered P values statistically significant at an ɑ level of .05.
Using a Bonferroni-corrected P = .01, we explored potential modification of the effect of treatment assignment for outcome variables with evidence of a statistically significant overall intervention effect, by 5 prespecified factors: patient age, race and ethnicity (combined), language, parental education, and clinic. Because of software limitations on the ability to test interactions with multilevel effect modifiers in imputed data, we ran analyses with nonimputed data. We conducted all analyses using Stata, version 14.2 (StataCorp; 2016). Because we randomized only 1 type of provider (licensed clinician or health educator) at each clinic, we were unable to explore effect modification by this variable.
The goal of qualitative analysis was to assess differences in counseling provided before and after implementation of the tool among providers randomized to the interventions and to assess provider and staff experience of the intervention. For the audio recordings, we estimated that 2 to 3 patient encounters at each time point (before and after implementation of the intervention) for each of the providers randomized to the intervention would allow us to achieve saturation of themes and meaningfully compare the preintervention and postintervention period. We conducted 4 focus groups—1 at each of the 4 clinics involved in the trial—and interviewed each of the providers randomized to the intervention.
We analyzed interviews and focus groups using a modified grounded theory approach, using elaborative coding that allowed themes to emerge from the data.36 To explore provider and staff experiences with My Birth Control and its effects on counseling, we used interview prompts including the impact, acceptability, and feasibility of using the tool in contraceptive counseling as well as concepts of SDM13 and patient-centered care.37 First, we iteratively developed an initial codebook with intentionally broad codes to allow for parsimonious coding. Once the codebook was finalized, we coded all transcripts. Next, we identified 4 broad codes that were particularly relevant to the research question for subcoding. In subcoding, we iteratively developed subcodes unique to each broad code and applied them to all relevant excerpts. For some broad codes, 2 coders subcoded jointly; for others, 1 coder took the lead and the other reviewed all coding. We resolved all disagreements through discussion. We considered coding complete when no new codes emerged.
For audio recordings, we created memos by provider with a brief summary of each of their counseling sessions. We applied codes denoting inductively generated emergent themes related to counseling patterns with and without tool use to transcripts. We used grids with rows for provider and columns for each theme to summarize patterns in counseling before and after tool use. The grids included information from memos and coding output.
The only substantive change in the protocol following submission of our funding proposal was a change in follow-up period. Before initiation of recruitment, we adjusted the proposed follow-up period of 3 and 6 months postenrollment to 4 and 7 months. As described previously under Study Objectives, the reason for this change was that the contraceptive injection is effective for 15 weeks, making it difficult to measure continuation at 3 months.
This study received approval from the University of California, San Francisco IRB. We did not make changes to the original protocol to obtain IRB approval. We did not observe any adverse events, which we measured nonsystematically throughout the study. We registered the trial with ClinicalTrials.gov under “Patient-Centered Support for Contraceptive Decision-making” (https://clinicaltrials.gov/ct2/show/NCT02078713).
In total, 28 providers enrolled in the trial (see Figure 3). One screened provider declined because of lack of time. We enrolled providers between November 2014 and November 2015. We interviewed providers randomized to the intervention arm about their experience with My Birth Control between March 2015 and June 2016. Eleven clinic staff members participated in 4 focus groups between October 2015 and May 2016.
CONSORT Diagram for Providers.
Providers in the study represented different age groups. A small majority (50.0%) were younger than the age of 36, and 60.7% were White (Table 2). More than half (57.1%) were nurse practitioners, certified nurse midwives, or physician assistants, and the remainder were other counselors or health educators. We found no differences in provider characteristics by study arm using Fisher exact test (Table 2).
Provider Participant Characteristics, by Trial Group.
A total of 758 patients enrolled in the trial (Figure 4). We enrolled patients between December 2014 and February 2016. Patients with a diverse set of sociodemographic characteristics participated in the trial (Table 3). Nearly half (46.1%) were 24 years old or younger. Most participants were non-White, and 18.3% chose to use study materials written in Spanish. Regarding parental education (a marker for socioeconomic status appropriate for young adults who may not have completed schooling), 25.2% of participants' parents had not completed high school, while 35.4% had parents with at least a 4-year college degree. A small majority (52.4%) were living at less than 100% of the federal poverty line. Characteristics did not significantly differ by study arm using chi-square tests controlling for site and accounting for clustering by provider.
CONSORT Diagram for Patients.
Patient Participant Characteristics, by Trial Group.
Between November 2014 and June 2016, we audio-recorded an additional 105 participants' counseling visits (Figure 5). We recorded 32 visits with providers randomized to usual care before randomization (1-3 per provider). Because these data were collected primarily to avoid a cointervention effect, as mentioned previously, they were not analyzed. We analyzed recordings from 32 visits with intervention providers before they began to have their patients use My Birth Control (1-3 per provider), and 41 recordings of visits with intervention providers after starting use of the tool (2-3 per provider), to examine differences in these providers' behaviors before and after they began using the tool.
Audio Recordings of Patient Visits With Providers.
At baseline, we had minimal missing values because of nonresponse (Table 4). One exception was values for income in relation to the federal poverty line, because many participants reported that they were unsure of their 2013 family income (n = 124 missing). Responses to time-sensitive questions about patient experience and decision-making were missing for 19 participants who left the clinic before completing the postvisit survey and whom we could not reach within 48 hours but did reach at a later point. Parity was missing for 11 participants because of an error in survey administration.
Participants Missing Surveys at Trial Time Points.
Of the 85 participants missing at 4 months, we reached 23 for follow-up at 7 months. We assessed relevant 4-month outcomes, with the exception of time-specific outcomes related to method use and satisfaction, when they completed their 7-month survey. At 7 months, we did not reach the remaining 62 participants who were missing at 4 months, and all variables are missing at both time points. We imputed all missing data due to loss to follow-up.
Complete case and multiply imputed estimates of treatment effects were virtually identical. We did find evidence that women with prior births were significantly less likely to complete a 7-month survey compared with those with no prior birth. However, parity did not modify the effect of the treatment on any of the variables collected at 7 months (data not shown). Likewise, adjustment for parity did not change the interpretation of the treatment effect on these outcomes.
The intervention and control groups did not differ significantly in our primary outcome of contraceptive continuation. In the intervention group, 56.6% of patients reported continuation of their chosen method at 7 months, compared with 59.6% in the control group (Table 5). We also found no significant difference at 4 months. The intervention and control groups did not differ significantly in terms of continuous use of any moderately or highly effective method of contraception at 4 or 7 months.
PCOs on Continuation, Experience, Satisfaction, and Decision-Making, by Group.
For secondary outcomes about patient experience of care, several differences between groups emerged. A significantly higher percentage of participants in the intervention group reported the highest level of interpersonal quality of counseling compared with the control group (66.0% vs 57.4%) (Table 5). More intervention participants reported complete satisfaction with information given about side effects compared with control participants (83.5% vs 75.7%). Groups did not differ in percentage reporting complete satisfaction with the visit overall (85.8% intervention vs 83.9% control), or percentage rating their visit as much better than the last (49.7% intervention vs 42.6% control).
Intervention participants were more likely than control participants to have optimal scores on the DCS informed decision (50.5% vs 43.2%) and uncertainty (41.6% vs 33.3%) subscales (Table 5). For the other DCS subscales, the groups did not differ significantly. Approximately a quarter reported the lowest level of overall decisional conflict in both intervention (25.4%) and control (22.6%) groups.
Similar percentages of participants in intervention and control groups reported high satisfaction with the method selected at baseline (63.2% vs 59.2%), 4 months (49.0% vs 51.5%), and 7 months (50.4% vs 55.2%) (Table 5). At 7 months, participants in intervention and control groups reported similarly high satisfaction with their current main method of birth control (58.7% vs 61.0%).
The intervention group had greater contraceptive knowledge after their counseling visit compared with the control group on multiple knowledge indicators (Table 6). On questions related to efficacy, higher percentages in the intervention compared with the control group knew that IUDs are more effective than pills (70.8% vs 48.2%, respectively), and that the contraceptive injection is more effective than condoms (60.4% vs 46.4%, respectively). However, there was no significant difference in the proportion who knew that pills were more effective than condoms comparing the intervention and control groups (55.8% vs 50.6%).
Contraceptive Knowledge, by Group.
Intervention participants answered more IUD-related items correctly compared with control participants. More intervention group participants compared with control group participants correctly stated that IUDs are an option for young nulliparous women (79.9% vs 67.9%), and that methods that cause periods to stop are safe (77.4% vs 65.5%). Similar percentages in each group correctly responded that long-acting reversible contraception (LARC) can be removed early (66.2% intervention vs 62.8% control). For the composite IUD knowledge variable, 36.1% of participants in the intervention group answered all questions correctly compared with 19.1% in the control group. When asked about emergency contraception (EC), significantly more participants in the intervention group than control group knew that copper IUDs can act as EC (22.3% vs 9.9%) (Table 6). Intervention and control groups did not differ in the percentage who knew that there is something can be done to prevent pregnancy after sex (94.2% vs 95.2%) or that EC can prevent pregnancy after sex (91.5% vs 93.4%).
With the exception of the pill and the injection, the intervention group had higher levels of knowledge related to methods' effect on fertility compared with the control group (Table 6). We observed a difference of approximately 10 percentage points in this knowledge for the patch, ring, IUD, and implant.
Among participants who had not previously heard about IUDs and implants, similar percentages learned about them during their contraceptive counseling visit in the intervention and control groups (Table 6). For the hormonal IUD, 79.0% heard about the method in the intervention group vs 71.9% in the control group. For the nonhormonal IUD, 81.9% heard about the method in the intervention group compared with 73.0% in the control group. For the implant, 72.7% heard about the method in the intervention group compared with 69.0% in the control group.
When asked about the decision-making process, a large majority in both groups felt their providers were involved the right amount in decision-making in both intervention and control groups (93.7% vs 90.5%), and the majority in both groups felt their providers appropriately expressed preferences (90.6% vs 88.4%) (Table 7). There was no significant difference in the percentages in the 2 groups who expressed complete satisfaction with how providers helped support their contraceptive choice (76.9% intervention vs 71.4% control). Groups similarly reported feeling they made the decision about contraception by themselves (72.6% in both groups) vs sharing the decision with their provider (22.3% in the intervention vs 24.7% in the control) or feeling like providers made the decision (5.1% intervention vs 2.7% control) (Table 7). Participants also reported providers had a preference for which method they chose in similar percentages in the intervention and control groups (49.2% vs 42.7%).
Patient Experience of Provider Role, by Group.
The intervention and control groups did not differ significantly on any outcomes related to use of and interest in highly effective contraception or unintended pregnancy (Table 8). At baseline, similar percentages chose a highly effective method in intervention and control groups (38.1% intervention vs 35.2% control). Participants in intervention and control groups gave similar average ratings of LARC methods, including the hormonal IUD (6.60 vs 6.25), the nonhormonal IUD, (6.29 vs 6.30), and the implant (5.78 vs 6.06).
Use of and Interest in Highly Effective Contraception, by Group.
Similar percentages in the intervention and control groups were using a highly effective method at 7 months (29.7% vs 28.2%) and 4 months (27.7% vs 28.4%), and the same was true for moderately or highly effective methods at 7 months (72.3% vs 69.1%) and 4 months (73.7% vs 75.2%) (Table 8).
Rates of unintended pregnancy did not differ significantly between intervention and control groups (Table 9). At 7 months, 6.7% in the intervention group had experienced an unintended pregnancy compared with 3.8% in the control group. At 4 months, 3.4% in the intervention group had experienced an unintended pregnancy compared with 2.8% in the control group.
Unintended Pregnancy, by Group.
For primary and secondary quantitative outcomes with statistically significant effects of intervention assignment, we found no evidence of effect modification by prespecified variables: patient race, age, parental education, language of study materials, or site (data not shown).
All but 1 provider reported My Birth Control made their counseling more efficient, and all but 2 described the tool as improving the way they allocated time during counseling, enabling them to home in on areas of interest to patients. Providers also perceived patients who used the tool to be more informed about method options and features and allowed them to take a more active role in method selection. One provider said, “It [My Birth Control] allowed me to give the client the floor first to talk about what they had learned and what was interesting to them and what stuck out to them as something that they might be interested in, versus me starting off giving my spiel about birth control methods. I feel like it gave them a little bit more agency in the process.” All providers in the intervention group found implementation of this intervention feasible and indicated that they would like to incorporate it into their everyday practice.
There were no significant changes before and after implementation of My Birth Control in MBI subscales for emotional exhaustion (β = −3.97; 95% CI, −10.62-2.68; P = .24), depersonalization (β = −1.52; 95% CI, −4.76-1.72; P = .36), or personal accomplishment (β = −1.64; 95% CI, −4.61-1.34; P = .28).
For the 15 intervention arm providers, we compared audio recordings of their contraceptive counseling before and after they began using the tool and found differences in counseling approaches. The most notable differences were related to use of the tool printout by providers. For many providers who had previously tended to exhibit a “foreclosed” counseling style (whereby they ask what method a woman would like to use and then do not explore other options) or a “directive” counseling style (whereby they promote a particular method they believe is best for the woman),24 the method list ensured that women's informed preferences for methods were at the center of counseling initiation. This contrasted with foreclosed counseling before tool use whereby women stated their method choice without necessarily being exposed to a range of options or information about those options. This also contrasted with directive counseling before tool use whereby some providers promoted the most highly effective methods of contraception without engaging with women about their preferences.
In some cases, use of the tool printout also instigated exploration of preferences for method characteristics. For example, one provider saw a woman who had been using pills, made an appointment for a refill, and indicated she would like to discuss the possibility of switching methods. The provider referenced the tool printout and said, “Okay, so you do want something that prevents pregnancy but at the same time you might want to be pregnant some time not too distant in the future.” The provider then displayed shared decision-making skills by talking with the woman about different short-acting methods and how they might work for her. Though the woman ultimately decided the pill was still her best option, she had the opportunity to explore in more depth other options and make an informed decision.
The tool also appeared to have the intended effects of equipping patients with knowledge of contraceptive methods and empowering them to ask questions of providers. For example, a patient who had previously used the injection indicated that she was interested in restarting this method, and that she had previously had counseling that emphasized IUDs and implants but was not interested in them. After using the tool, which introduced comparative effectiveness of methods, she discussed with the provider whether the injection would be as effective as LARC methods. In 3 other sessions, patients referenced the tool feature that allows them to annotate questions for the provider before counseling.
Average overall visit time was 11.81 minutes longer in the intervention arm compared with the control arm (95% CI, 8.54-18.66; P < .001). However, time spent with the provider did not differ between arms (β = 1.05; 95% CI, −3.19-5.29; P = .63). The average time a patient spent engaging with the tool was 12 minutes (range = 1-33 minutes; SD = 5 minutes). Analysis of 4 focus groups with 11 clinic staff showed they found tool implementation acceptable and feasible. For example, one registration staff member said, “I think the flow went well. It didn't cause too much traffic … sometimes it got a little hectic but not too much. Overall it went smoothly.” Another described her perception that patients felt positively about the tool: “I think the patients have enjoyed using the iPad [My Birth Control]. They'll come in looking for it, sometimes.” While clinic staff generally indicated positive reception of the intervention, they had limited interaction with the tool because research staff facilitated implementation. This may affect staff's ability to judge how tool integration would influence their workflow.
The results of our cluster randomized trial indicate that My Birth Control had a positive impact on a range of patient-centered outcomes but did not affect contraceptive continuation or contraceptive use. Interpreting these findings requires consideration of the relative importance of these different types of outcomes and how they relate to one another. Below, we discuss the relevance of our findings in the context of contraceptive counseling and the current literature in the field. We discuss the generalizability of our findings, how this evidence can support implementation, outline our plans for dissemination, and consider limitations of this study.
Our trial found that among a predominantly low-income, racially and ethnically diverse sample of women, My Birth Control had a positive impact on important patient-centered outcomes, including patient experience of the counseling interaction, decisional conflict, and knowledge of contraceptive options and features.
Given the documented gaps in counseling quality perceived by patients,6-9 our findings support use of the tool to increase the patient-centeredness of contraceptive counseling. Further, the tool improved the quality of contraceptive decision-making, as patients who used it felt more informed, felt more certain in their decision-making, and were equipped with more contraceptive information during the decision-making process. Qualitative interviews with providers and audio recordings of counseling interactions reinforced these results; both sources of data suggested patients were more empowered and knowledgeable after using the tool.
We must consider these findings on patient-centered outcomes in light of our findings of no impact on our primary outcome and other outcomes related to contraceptive use. We initially chose contraceptive continuation as our primary outcome because we wanted to focus on a patient-centered outcome that also had public health implications. We hypothesized that continuation was a patient-centered outcome, in that it indicated whether women chose a method that was a good fit for their lives for an extended period of time. This outcome also had relevance to the public health priority to reduce unintended pregnancy, as contraceptive discontinuation can lead to gaps in use and risk of pregnancy.38
Recent research has complicated the conceptualization of contraceptive continuation as a patient-centered outcome, however, as considering continuation as an inherently positive result does not necessarily reflect women's experiences of contraceptive use. Specifically, women's preferences and needs regarding methods can change over time, related to issues such as changes in the need for contraception or experiencing unanticipated side effects.39,40 Switching methods can therefore be a positive outcome that reflects women's knowledge of and ability to access alternative methods to meet their evolving needs. It may also be a positive outcome for a woman to return to a clinic to switch methods, reflecting a positive relationship with the system of care, as negative clinical encounters can deter women from seeking contraceptive care again.5 Using contraceptive continuation as an outcome may be particularly problematic from a patient-centered perspective in the case of LARC methods. Recent research has documented that some providers express reluctance to remove LARC methods when patients request removal.41,42 In circumstances like these, contraceptive continuation clearly does not reflect a positive patient experience or patient-centered outcome.
Outcome variables related to method choice—particularly choice of LARC methods—are similarly problematic from a patient-centered perspective. Method choice outcomes place value on the relative effectiveness of methods, rather than on method characteristics individual women may prioritize, such as side effects or mechanism of use. One could argue that placing a priority on effectiveness is motivated by a desire to help women avoid the undesirable outcome of an unplanned pregnancy. However, recent research shows that avoiding an unplanned pregnancy is not equally important to all women.43 Combined with a lack of evidence supporting the traditional view that a planned pregnancy is associated with better outcomes for women and their children,44 we cannot justify focusing on effectiveness over other method features that women consider important. Therefore, these outcomes, which prioritize the effectiveness of the methods chosen, are not inherently patient centered.
In contrast with continuation and method-based measures, outcomes related to patient experience, itself a core component of quality, are inherently patient centered, and both the DCS and the Interpersonal Quality of Family Planning scale have documented reliability and validity.30,31 In addition, direct assessment of the receipt of patient-centered care has implications for women's reproductive health in the longer term, as it reflects the provision of care that supports women's reproductive autonomy and decision-making. This type of care is likely to facilitate ongoing engagement with reproductive health providers and quality decision-making that reflects women's own values and preferences.45 Therefore, My Birth Control's potential impact on patient-centered outcomes is significant in both the short and long term.
While patient experience measures are important in all areas of health care, these outcomes are of particular consequence in reproductive health care. This is due to the history of coercion of women of color and other disadvantaged populations within family planning settings, including nonconsensual sterilization.46 As a result, it is critical to improve women's experience of family planning care and their ability to make informed, autonomous contraceptive decisions. In contrast, using outcomes that position highly effective methods as the “better choice” over other methods or outcomes that prioritize contraceptive continuation over patient preferences does not support patient autonomy.5,46 Further, research has documented that women of color may receive lower-quality contraceptive counseling.9 Use of My Birth Control among racially and ethnically diverse populations therefore has the potential to help reduce disparities in patient experiences of reproductive health care. This, in turn, creates potential for longer-term impacts on engagement with the reproductive health care system more broadly.
Our finding that My Birth Control did not have a significant impact on method satisfaction deserves examination, as we hypothesized that the tool would help women identify a method aligned with their preferences. Further, one might expect that our findings on reduced decisional conflict and increased knowledge would result in improved method satisfaction. The lack of this expected result may be due to the use of a single-item measure of satisfaction. Using a single item to capture complex phenomena such as satisfaction produces less valid and reliable results.47 Another possible explanation is that there was an overall discordance between available methods and women's preferences for method characteristics. Research shows that no methods have all the features considered extremely important for the vast majority of women.14 Improved decision quality may not overcome the lack of desirable options to improve method satisfaction. If this is the case, this further underlines the importance of focusing on patient experience of care. As there is no 1 clear option for an individual, women will often have ongoing engagement with family planning providers as they adjust their method to their needs and experiences. Positive experiences with care can facilitate this engagement and improve reproductive health across the life course.
Despite the documented benefits of decision support tools on patient-centered care and decision quality,17 there has been limited attention to their use in the context of contraceptive decision-making. A 2017 Cochrane review identified 1 decision aid for contraception.48 Several recent trials have examined interventions similar to contraceptive decision support tools. While researchers designed these interventions to elicit patient preferences, they implicitly or explicitly gave higher priority to method effectiveness than other method characteristics when making suggestions for methods.49-52
In contrast to these previously evaluated tools, My Birth Control not only elicits patient preferences, but centers these preferences in the counseling process. Patient preferences guide the tool's method recommendations, resulting in enhanced patient-centeredness during the clinical encounter when a provider uses the tool printout. By including contraceptive attributes that are important to women, My Birth Control also distinguishes itself from previous counseling aids for contraceptive methods. A previous review of decision aids, structured interviews, and questionnaires (including those not formally studied) developed between 1985 and 2013 documented that existing resources inconsistently addressed the contraceptive attributes that are most important to women, such as side effects.52
This trial is therefore the first cluster randomized controlled trial of a patient-centered contraceptive decision support tool specifically designed to support women to consider the range of method attributes relevant to them. Our results provide the strongest evidence to date that such an intervention has a positive impact on the patient experience of contraceptive care. Our findings point to the potential for large-scale improvement in contraceptive counseling with the dissemination of My Birth Control.
As described in the Results section, our trial participants included a large range of ages and diverse racial and ethnic groups. This diversity makes the findings generalizable to women of different ages, races, and ethnicities. Parental education, a marker of socioeconomic status, varied among participants, making the findings relevant for women from different socioeconomic backgrounds. Trial participants also included both parous and nulliparous women, making the findings relevant for women with different pregnancy histories.
The large majority of participants spoke English, and approximately a fifth spoke Spanish. We found no evidence of effect modification on the intervention by language, making the results applicable for English- and Spanish-speaking populations. Whether we can generalize our findings to women who do not speak these languages is unknown.
The clinics represented in this trial include a hospital outpatient clinic, a freestanding family planning clinic that belongs to a large nonprofit family planning care organization, a public health department primary care clinic, and a college student health center. All of these clinics provide routine contraceptive counseling. Our findings are generalizable to similar settings that provide routine contraceptive care.
In addition to evaluating the effect of My Birth Control on patient outcomes, we paid specific attention in our study to factors relevant to dissemination. These factors included the impact of the tool on providers and clinic settings. Our patient, provider, and organizational stakeholders helped us consider potential barriers to implementation (eg, patient access to the tool, complexity, cost, impact on clinic flow) and worked with us to minimize these barriers from tool development through completion of the trial. Findings from provider interviews, clinic staff focus groups, and audio recordings of counseling visits suggest acceptability among providers and staff. These findings also suggest that implementation at a larger scale is feasible.
Use of the tool did not add time to the provider–patient interaction, which is noteworthy considering its substantial effect on various patient-centered outcomes. Tool use added <15 minutes to the total visit time (ie, wait time), and clinic staff reported in focus groups that the tool did not obstruct clinic flow. Given its perceived low impact on the clinical setting, My Birth Control may be implemented in a wide range of practice settings, including family planning clinics, hospitals, and primary care settings.
Our dissemination efforts will be anchored by Greenhalgh's model of diffusion, including attending to the processes of readiness, adoption, and implementation.53 To disseminate the tool to new clinical settings, we will share our trial results with clinic-level decision makers. We will provide guidance for identifying local champions of the tool who can garner buy-in from other providers and serve as a resource for implementation challenges. We will also provide guidance on strategies to promote integration into clinical workflow based on our prior experiences and recommendations from provider stakeholders. In settings where the extra time required for tool use cannot be absorbed by existing wait times, we will include recommendations for how clinical sites may minimize negative effects of this aspect of the tool. We also plan to provide a mobile version of the tool so that patients may use it before arriving at the clinic.
To communicate about the tool with potential provider users, we will develop a toolkit containing orientation resources. The toolkit will describe the tool's intended purposes of advancing patient-centeredness and enhancing the patient experience of care and include information about the advantages of a decision support tool and the tool's alignment with national recommendations for contraceptive counseling. The toolkit will provide guidance on how to use the tool and case studies to enhance provider and staff acceptance of the tool. The intensiveness and scope of our dissemination efforts will be determined by the availability of funding to support these activities.
We found no evidence of effect modification by patient age, race, language, parental education, or clinic site. This indicates that there are not differential effects of the intervention on groups defined by these characteristics. However, because our trial targeted a general family planning population, our results mainly apply to routine contraceptive counseling. Before applying the tool to populations of contraception clients with more complex needs—for example, postpartum women or women with chronic diseases—we need to investigate the utility of the tool for these subgroups.
Limitations of this trial included the potential for contamination. Contamination could have occurred if providers in the intervention arm shared information about the tool with their colleagues in the control arm. Contamination could have occurred if control providers were otherwise exposed to the tool in their workplace. The research team took precautions to minimize such exposure, for example by shredding printouts from the tool immediately after they had been used by intervention providers.
Given the nature of the intervention, we could not blind patients or providers to their group assignment. To ameliorate this limitation, our procedures included blinding of research assistants to patient study group during follow-up data collection. In addition, we had limited power to observe differences in the subscales of the MBI for providers, due to our small provider sample (n = 28). This small sample was due to the fact that providers served as cluster units, and it was not feasible for us to recruit additional cluster units and their patients.
Because we randomized provider-level clusters, there may be systematic, unobserved differences in patient characteristics by study group. This possibility is limited due to the fact that randomization was stratified by clinic, and demographics of individual providers' patients within one clinic are likely similar. Provider counseling practices may also have differed at baseline. We minimized selection bias by ensuring patient and provider participants were not aware to which arm they would be assigned before participation, avoiding preferential enrollment by arm.
Regarding the impact of missing data, we had almost complete data at baseline and only 14% loss to follow-up at 7 months. Our use of MI and examination of the sensitivity of our findings to violations of the untestable MI assumption that data are missing at random support a lack of impact of missing data on our results.
Our tests of multiple outcomes and the fact that there was no difference in some patient-centered outcomes, and that some were of borderline significance, indicate we may need further research on and refinement of My Birth Control. Because we designed this intervention specifically to improve patient-centered care and informed decision-making, our findings related to interpersonal quality of care and decisional conflict are consistent with our theory for how My Birth Control impacts care, and we consider them robust.
Future research should focus on the effectiveness of My Birth Control in real-world settings. The tool is promising for use in the United States and—with adaptation—globally. Future research may be conducted in a wide range of geographic areas and in additional languages beyond English and Spanish.
As My Birth Control may be adapted to diverse care settings, future research may examine the effect of the tool in certain subpopulations of patients with specific needs. For example, postpartum and breastfeeding women have a different range of methods available to them compared with other women. Postpartum women also have the option to receive counseling before giving birth, immediately postpartum, or in well-baby visits, which are distinct settings compared with routine family planning visits. Research for this subpopulation should examine the feasibility, acceptability, and effectiveness of My Birth Control in these different clinical settings.
In addition to research on subpopulations, we may conduct dissemination and implementation research to determine effective strategies for increasing uptake and optimizing implementation in diverse settings. Such research may also examine the uptake and effectiveness of the proposed My Birth Control toolkit.
In this cluster randomized controlled trial, we did not find an effect on our primary outcome of contraceptive continuation, or on other outcomes related to contraceptive use and unintended pregnancy. We found that My Birth Control had a positive effect on a range of secondary outcomes related to the patient-centeredness of contraceptive counseling, including patient experience, informed decision-making, and contraceptive knowledge. These findings are noteworthy given the importance of patient-centeredness in the contraceptive decision-making context. We also found that My Birth Control is acceptable and desirable to providers, and we documented low impact on clinic flow.
The results of this study may inform system- and clinic-level decision makers as they consider strategies for improving their patients' experiences without compromising clinic flow efficiencies. Individual providers may also find these results useful if given the option to incorporate the tool into their work.
The research team worked to minimize threats to internal validity during the trial. Potential threats included contamination between provider groups, lack of blinding, and cluster randomization resulting in possible confounding. Potential threats to external validity included the fact that the tool was used in routine counseling for this trial, rather than in more complex visits. Future research may inform adaptation and implementation of the tool in more complex care scenarios. Decision makers may wish to consider threats to validity and the unique circumstances of their systems and clinics when making decisions regarding tool adoption. Overall, the results of this trial are promising in their indication of My Birth Control's appropriateness and impact.
We wish to thank our patient stakeholders and provider stakeholders for their invaluable support on this project. The success of this project would not have been possible without their willingness to share their experiences and insights. We wish to thank Michaela Gonzalez, Kitty Torres, Sindura Reddy, Liz Steinfield, Alissa Perucci, Shivaun Nestor, Elizabeth Johns, Sarah Siebold, and Dafna Wu, as well as the Women's Community Clinic and members of their Client Advisory Council. We thank Alexis Hoffman, Anna Gabriela Ycaza, Cara Hall, and Nora Anderson for their contributions to the study, and Whitney Wilson for her contributions to this report. We also wish to thank our organizational stakeholders: The National Campaign, Essential Access Health, Planned Parenthood of Northern California, and Planned Parenthood Mar Monte.
Research reported in this report was [partially] funded through a Patient-Centered Outcomes Research Institute® (PCORI®) Award (#CE-1304-6874) Further information available at: https://www.pcori.org/research-results/2013/decision-aid-help-women-choose-and-use-method-birth-control
Dehlendorf C, Vittinghoff E, Fitzpatrick J, et al. (2019). A Decision Aid to Help Women Choose and Use a Method of Birth Control. Patient-Centered Outcomes Research Institute (PCORI). https://doi.org/10.25302/10.2019.CE.13046874
The [views, statements, opinions] presented in this report are solely the responsibility of the author(s) and do not necessarily represent the views of the Patient-Centered Outcomes Research Institute® (PCORI®), its Board of Governors or Methodology Committee.
This book is distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License which permits noncommercial use and distribution provided the original author(s) and source are credited. (See https://creativecommons.org/licenses/by-nc-nd/4.0/
Your browsing activity is empty.
Activity recording is turned off.
See more...
