mHealth Interventions for Contraceptive Behavior Change in the United States: A Systematic Review

Abstract

Ensuring people have access to their preferred method of contraception can be key for meeting their reproductive goals. A growing number of mHealth interventions show promise for improving access to contraception, but no literature review has identified the effects of mHealth interventions among both adolescents and adults in the United States. The purpose of this systematic review was to describe the format, theoretical basis, and impact of mHealth interventions for contraceptive behavior change (contraceptive initiation and continuation) among people of all ages in the US. A systematic review of the literature was conducted using six electronic databases guided by Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) guidelines. Data on study design, frequency, duration, mHealth modality, contraceptive method, behavior change theory, and behavioral outcome were extracted to facilitate comparison. Eighteen studies met eligibility criteria. The majority (11; 61%) used SMS (short message service). Twelve studies focused on contraceptive initiation, most (n=8) of which also measured continued use over time. The remaining six interventions focused on continuation alone, generally through appointment reminders. Very little contraceptive behavior change was identified across studies. Current mHealth interventions may hold promise for some health areas but there is little evidence that they change contraceptive behavior. Future mHealth interventions should focus on assessing person-centered outcomes, including satisfaction, side effects, and reasons for discontinuation to best support people to use their preferred contraceptive method.

Introduction

In the United States (US), an estimated 23% of women who do not want to become pregnant had sex with a man in the last three months without using contraception, based on the 2015–2017 National Survey of Family Growth (NSFG), a nationally representative population survey (Daniels, 2018). Even among those reporting contraceptive use, an estimated 15% are using a contraceptive method sporadically (based on a previous NSFG) (Pazol et al., 2015). Non-use and inconsistent use can drive the persistently high rate of unintended pregnancies in the US: 45% of pregnancies annually can be classified as either mistimed (becoming pregnant earlier than intended) or unwanted (did not want to become pregnant then or anytime in the future) in the most recent estimates from the NSFG (Finer & Zolna, 2016). Simultaneously over the past decade, more options for contraceptive methods have become routinely available and covered by health insurance. Methods available include permanent methods (including sterilization); highly effective long-acting reversible contraception (LARC, such as contraceptive implants and intrauterine devices (IUDs)) more user-dependent, short-acting hormonal methods (including combined hormonal contraceptive pills, patches or vaginal rings,progestin-only pills, and progestin injectable contraceptives), and the less effective behavioral methods (fertility awareness based methods and withdrawal) and barrier methods (including condoms and diaphragms) (Britton et al., 2020). Many contraceptive methods are hormonal, and with the exception of vasectomy and male/external condoms, generally controlled by the female partner or partner with a uterus. While each method may be contraindicated for certain medical conditions, among healthy users the choice is generally a matter of personal preference (Hatcher, 2018).

Non-use and inconsistent use may arise from dissatisfaction with contraceptive methods. Recent research of a population-based sample in Ohio found that at least 25% of women of reproductive age were not completely satisfied with their current contraceptive method or not using their preferred method (Chakraborty et al., 2021), while another study conducted in Delaware and Maryland showed that higher satisfaction or using a preferred method were associated with less intention to switch and greater intention to continue method use (Steinberg et al., 2021). Considering the continued high rates of unintended pregnancy, as well as some women reporting misalignment between their contraceptive method use and preferences, there is a need for additional innovation in interventions to support uptake of preferred contraceptive methods in the US.

There is great interest in harnessing the power of technology to help people access information to help align their contraceptive behaviors with their reproductive goals. Access to mobile phones appears to have reached a saturation point, with 91% of US adults ages 18–49 reporting they own or use a smartphone (Hitlin, 2018) and 58% of young adults 18–29 reporting they almost exclusively use the internet with their smartphone, rather than a home internet connection (Pew Research Center, 2019). As of 2017, 75% of US adults reported that they used the internet first in their most recent health information search (Finney Rutten et al., 2019). While information on contraception can be readily accessed by youth and adults via online webpages (Aubrey et al., 2020; Marques et al., 2015) and may lead to improved self-efficacy to use contraception (Scull et al., 2021), there are barriers to behavior change. In addition, the ways in which different groups seek out contraceptive information on the internet may vary by sociodemographic characteristics (Laz & Berenson, 2013; Zimmerman, 2018). With wide availability of mobile phones, especially smartphones, a more individualized approach to expand health access, knowledge, and support is mHealth, defined by the World Health Organization as “medical and public health practice supported by mobile devices, such as mobile phones, patient monitoring devices, personal digital assistants (PDAs), and other wireless devices” (World Health Organization, 2011). In other domains of health, mHealth interventions for education and behavior change have been widely acceptable and periodically effective (Rowland et al., 2020). To harness this modality to support the alignment of contraceptive behavior and reproductive goals among people of all ages in the US, it is important to fill the gap in the literature by documenting mHealth interventions that may impact contraceptive behavior change for this population.

This important gap remains despite the many scoping and systematic reviews to date on a variety of interventions for women’s health, sexual and reproductive health, or non-behavioral family planning-specific outcomes in many geographic locations. These reviews have examined the impact of various types of digital health interventions on women’s health or sexual and reproductive health broadly in addition to family planning outcomes (Burns et al., 2016; Goldstein et al., 2018; Sanz-Lorente et al., 2018; Swanton et al., 2015), been limited to adolescents/young people (Guse et al., 2012; L’Engle et al., 2016; Wadham et al., 2019; Widman et al., 2018) or to low and middle income countries (Colaci et al., 2016; Ilozumba et al., 2018; Sondaal et al., 2016), or examined the impact of mobile applications, social media, or text messaging separately but have often not looked across mHealth approaches in a single review (Gabarron & Wynn, 2016; Poorman et al., 2015; Rousseau et al., 2019). Other reviews specifically focused on family planning outcomes looked at the efficacy of computer-based electronic interventions, general strategies to improve adherence/continuation of short-term hormonal contraception, and reminder systems for contraceptive users (Dewart et al., 2019; Mack et al., 2019; Zapata et al., 2018). These may have included, but not specifically been focused on, studies of mHealth interventions. Therefore, we lack a current systematic review that specifically synthesizes how mHealth affects contraceptive behavior in the US among both adolescents and adults. This gap is of particular importance considering that the US is unique among high-income countries in that it does not provide universal healthcare coverage, including access to contraception. This results in barriers to contraceptive access related to health insurance coverage and cost that may require specific mHealth interventions.

Understanding the specific mechanisms of action between mHealth interventions and behavioral outcomes is critical to developing successful programs and products. We note that two other reviews had similar goals to our review. One recent review examined the use of telemedicine on contraception and abortion and found that text messaging reminders improved continuation for oral contraceptive and injectable contraceptive users (Thompson et al., 2020). However, this review was focused on telemedicine broadly and included interventions where technology was used to provide a clinical service; in addition, it did not examine the extent to which theories of health behavior might inform targeted interventions to change behaviors. A Cochrane review examined the impact of mobile phone interventions, including voice messages, on contraceptive use in the US and worldwide and found mixed results of text message and automated interactive voice messages to improve contraceptive use (C. Smith et al., 2015). However, due to Cochrane review requirements, it was restricted to randomized controlled trials and only five studies met inclusion criteria. That review is also now more than six years old in a space that continues to expand and develop rapidly.

We aimed to systematically summarize the literature on how mHealth has been used to impact contraception behaviors, including initiating and continuing a contraceptive method among people of reproductive age in the US. The aims of this study are to: 1) describe the type, format, and theoretical bases of these interventions, including the contraceptive methods of focus; 2) summarize the impacts on family planning use behaviors as a result of these interventions; and 3) provide preliminary insights into characteristics of mHealth interventions that are impactful to changes in contraceptive behavior.

Methods

Search strategy

The search strategy, study selection, and data extraction was conducted in line with Preferred Reporting Items for Systematic Review and Meta-Analyses (PRISMA) reporting guidelines for systematic reviews (Page et al., 2021) and the protocol was registered with the Open Science Framework and is publicly available. Six electronic databases that index articles from medicine, public health, gender, and social science were searched: SCOPUS, PubMed, Embase, Web of Science, GenderWatch, and Women’s Studies International. The search terms and search strategy, provided in Table 1, were developed in consultation with a health sciences librarian at the University of North Carolina at Chapel Hill. The initial search was conducted on December 18, 2018. We limited the search start date to January 2010 to provide the scope of mHealth interventions for contraceptive behavior change over the prior decade. A subsequent search was conducted to capture additional eligible articles, and included articles published between January 1, 2010 to May 28, 2020.

Table 1:

Keyword terms used for searches

1. (Women OR woman OR female)  AND 2. (eHealth OR mHealth OR “digital health” OR mobile OR SMS OR “short message” OR “text message” OR texting OR tablet OR app OR “social media” OR facebook OR twitter OR smartphone OR “mobile applications” OR “mobile application”)  AND 3. (“family planning” OR contracept* OR condom OR condoms OR IUD OR “intrauterine device” OR LARC OR “contraceptive implant” OR “etonogestrel implant” OR “injectable contraception” OR “injectable contraceptive” OR DMPA OR Depo OR “oral contraceptive pill” OR NuvaRing OR OrthoEvra OR Paragard OR Mirena OR Skyla OR Liletta OR Implanon OR Nexplanon)

4. In addition, the following MeSH terms were used for the PubMed search:  (Telemedicine OR Cell Phone OR Text messaging OR Reminder Systems OR Social Media OR Smartphone OR Mobile Applications)  AND  (Family Planning Services OR Contraception OR Contraceptive Agents OR Contraceptive Devices).

Note: While the same search terms were used across databases, there was some variability in the search fields. For PubMed and Embase, the search terms were limited to the title or abstract; for SCOPUS, the search included title, abstract, and keyword; for Web of Science, the topic search was used; for GenderWatch, the search included anywhere but the full text (i.e. title or abstract); and the search in Women’s Studies International searched the full text for the search terms.

Selection criteria

Eligibility criteria for articles were: 1) data-based, peer-reviewed manuscripts; 2) within the United States (multi-country studies including the US were also eligible if US-specific results were extractable); 3) participants were any people involved with a pregnancy (e.g., includes non-pregnant women who have sex with men, men who have sex with women) of reproductive age (generally defined as ages 15–49 for women); 4) utilized mHealth interventions delivered via mobile phone or tablet devices specifically to individual patients or participants to support contraceptive use (these interventions could include text messaging/short-messaging service (SMS) and applications (apps), as well as social media, websites, or videos if they were clearly documented as being delivered via mobile phone or tablet); and 5) quantitatively measured contraceptive behavioral outcomes including initiating or starting a method or continuing/discontinuing a method. Condom use behaviors were considered eligible, regardless of whether the study was originally motivated by family planning or sexually transmitted infection prevention, as long as condoms were utilized in pregnancy-possible sex. Excluded were abstracts, literature reviews, or commentaries; research conducted exclusively outside of the US; non-human studies or biomedical or pharmacological studies of efficacy of specific hormonal contraceptive products; usability testing of a product or data collection using mobile devices; mHealth interventions in conjunction with a medical device; computer-based interventions or use of mobile phones exclusively for making phone calls; and studies whose measured outcomes were limited to knowledge of contraception or intention to use.

Screening and data extraction

Search results were imported into Covidence (Covidence Systematic Review Software, 2020), an online tool for systematic review management, which performed the initial screening for and exclusion of duplicate records. The titles and abstracts of the remaining results were reviewed independently by the first and third author for initial inclusion. Discrepancies in decisions for inclusion for full text review were discussed and consensus reached. The full text of the remaining records were further screened for eligibility by the first and third author, resolving discrepancies for final inclusion through discussion. A data abstraction form (publicly available on Open Science Framework) was developed by the first and third authors, who both subsequently independently extracted information on relevant variables for all selected articles in line with the research aims. These included publication year, study aim(s), study design, sample size, recruitment setting (geographic and clinical or other), participant demographic characteristics, intervention/control mHealth modality, frequency, and duration, contraceptive methods included, theoretical basis, and contraceptive behavior outcome. Any data discrepancies during extraction were reviewed and decided by the second author. Extracted data were then analyzed by contraceptive behavior outcome (initiation and/or continuation). Due to the heterogeneity of the study designs and sample sizes, we did not conduct meta-analysis or pool data across studies to examine effect sizes. We reported frequency or mean measures and odds ratios as measures of treatment effects, along with 95% confidence intervals and/or p-values.

Quality assessment

A quality analysis of all the included articles was conducted using a quality assessment tool for quantitative studies developed by the Effective Public Health Practice Project (Effective Public Health Practice Project, 2010; Thomas et al., 2004). Using the tool, each paper was scored “weak”, “moderate”, or “strong” across six domains (selection bias, study design, confounding, blinding, data collection methods, and withdrawal/dropout), after which an overall global rating was calculated. The assessment was conducted jointly by the first and second author and disagreements were resolved through discussion and consensus.

Results

A total of 4457 search results were identified. After exclusion of duplicate references, the titles and abstracts of the remaining 2099 results were reviewed, of which 1994 did not meet inclusion criteria. After reviewing the full text of the remaining 105 articles, an additional 87 records were excluded, resulting in 18 articles included in the final review (Figure 1).

Figure 1:

PRISMA flow diagram for systematic reviews

Participant and study characteristics

The 18 included studies are summarized in Tables 2 and 3 and were published between 2010 and 2020. Two-thirds of the studies (12; 67%) focused on adolescent or young adult participants (age range: 13–25) (Buchanan et al., 2018; S. Bull et al., 2016; Castaño et al., 2012; Chernick et al., 2017; Cornelius et al., 2013; Juzang et al., 2011; Manlove et al., 2020; Rinehart et al., 2020; Shrier et al., 2020; Suffoletto et al., 2013; Trent et al., 2015; Yao et al., 2018). Of the studies that included individuals older than 25, only three studies included participants over age 31 (Dehlendorf et al., 2019; Sridhar et al., 2015; Thiel de Bocanegra et al., 2017). Over three-quarters of the studies (14; 78%) described interventions specifically targeted to women. Six studies (33%) had racially diverse populations, with sizeable proportions (>15%) of at least three different racial or ethnic groups included (S. Bull et al., 2016; Castaño et al., 2012; Dehlendorf et al., 2019; Manlove et al., 2020; Shrier et al., 2020; Thiel de Bocanegra et al., 2017) and 10 studies (56%) recruited predominantly people of color (African American/Black, Hispanic/Latinx, or American Indian/Alaska Native) (Buchanan et al., 2018; Chernick et al., 2017; Cornelius et al., 2013; Gilliam et al., 2014; Hebert et al., 2018; Juzang et al., 2011; Rinehart et al., 2020; Suffoletto et al., 2013; Yao et al., 2018).

Table 2:

Summary of study designs, key results, and quality scores of mHealth studies on contraceptive behavior change

Author, Year	Study Design	Sample size		Participant characteristics	Duration	Contraceptive method(s)	Key results	Quality score
		Intervention	Control
Initiation or Initiation/Continuation
Manlove et al., 2020	RCT	797	791	Racially diverse women ages 18–20 who were not pregnant or trying to become pregnant	6 weeks	Any method (including condoms), with focus on hormonal contraception (DMPA, OCP, patch, ring) and LARC	Intervention group significantly less likely to report sex without hormonal/LARC method (22.1% vs. 29.7%; p=0.001), but this difference disappeared when limited to those sexually active at baseline (48.9% vs. 49.1%, p=0.945).	Moderate
Rinehart et al., 2020	RCT	122	122	Predominantly Latinx women ages 13–18, not trying to become pregnant in next year	3 months	LARC; Short-acting reversible contraception (DMPA, OCP, patch, ring); Condoms	Intervention group had significantly higher odds of prescription birth control use at 6 months (OR: 3.58, 95% CI: 1.18–10.90), but this difference disappeared when limited to those sexually active at baseline (OR not reported).	Strong
Shrier et al., 2020	Cohort	17	--	Racially diverse women 15–24 years	5 weeks	Condoms; Effective contraception (methods not specified)	Median proportion of condom-unprotected sex events declined from 100% at baseline to 0% at 3 month follow-up (p=0.01); little change in hormonal contraceptive use (proportion not reported).	Weak
Hebert et al., 2018	RCT	110	111	Predominantly Black and Latina women ages 15–29 not currently using LARC	In clinic before appointment	LARC	Intervention participants were not significantly more likely to use a LARC method than controls, neither immediately after the visit (3.8% vs. 1.0%, p=0.37) nor at 3 months post-enrollment (8.0% vs. 1.0%, p=0.34).	Moderate
Yao et al., 2018	Cohort	408	--	American Indian/Alaska Native youth ages 15–24 years	3 months	Condoms	The odds of always using a condom (vs. sometimes or never) were significantly higher from baseline at 1 week post-intervention (OR: 2.43, 95% CI: 1.15–5.13) and 9 weeks post-intervention (OR: 2.60, 95% CI: 1.18–5.73)	Weak
Chernick et al., 2017	RCT	50	50	Predominantly Hispanic young women ages 14–19 not currently using hormonal contraception	3 months	Contraception (methods not specified)	No significance test of contraceptive initiation between the intervention (12.0%) and control (22.4%) groups reported.	Weak
Bull et al., 2016	RCT	436	416	Racially diverse young people ages 14–18	25 weeks	Condoms; Contraception (methods not specified)	No significant differences found between intervention and control groups on average percentage of sex acts protected by condoms (94.1% vs. 92.7%; Cohen’s d: 0.06 (95% CI: 0.04, 0.08)) or contraception (97.5% vs. 95.9%; Cohen’s d: 0.09 (95% CI: 0.07, 0.11) in past 3 months. Percentage of pregnancies among Hispanics was lower in intervention (1.79%) than control group (6.72%; p=0.02).	Weak
Sridhar et al., 2015	RCT	60	60	Women ages 18–45 not currently using a contraceptive method or willing to switch methods	In clinic before appointment	Very effective methods (LARC); Effective methods (injectable contraception, OCP, patch, ring); Less effective methods (condom, diaphragm)	No significant difference between the intervention and control groups in choice of a very effective method of contraception (52% vs. 57%, p=0.753).	Strong
Gilliam et al., 2014	RCT	31	31	Predominantly Black women ages 16–30 not current LARC users	In clinic before appointment	LARC	No significant differences in initiation of a LARC method between intervention and control groups (per-protcol: 25.0% vs. 20.8%, p=0.72; intent-to-treat: 22.6% vs. 25.8%, p=0.77).	Moderate
Cornelius et al., 2013	Cohort	40	--	Black adolescents 13–18 years	3 months	Condoms	No significant change in mean number of unprotected and condom-protected vaginal sex between baseline and follow-ups (Time 1: 0.29, Time 2: 0.22, Time 3: 0.20).	Weak
Suffoletto et al., 2013	RCT	23	29	Predominantly Black women ages 18–25 with hazardous drinking and risky sexual behavior	3 months	Condoms	No significant differences between intervention and control groups in proportions reporting condom use at last vaginal sex (53% vs. 38%) or always use of condoms over last 28 days (33% vs. 24%). No p-values <0.05.	Moderate
Juzang et al., 2011	Prospective case-control	30	30	Black men 16–20 years	3 months	Condoms	No significant changes in the proportion of sex acts protected by condoms in past 90 days (point estimates not reported).	Moderate
Continuation
Dehlendorf et al., 2019	RCT	407	351	Racially diverse women ages 15–45 wishing to discuss starting or switching contraceptive method	In clinic before appointment	Chosen contraceptive method (methods not specified), with additional focus on “moderately or highly effective methods”	No significant differences between intervention and control groups for continuation of chosen contraceptive method at 4 (OR: 0.79, 95% CI: 0.57–1.10) or 7 months follow-up (OR: 0.89, 95% CI: 0.65–1.22).	Strong
Buchanan et al., 2018	Retrospective case-control	42	45	Participants who completed all 3 DMPA visits from Trent et al., study	20 months post-intervention	More efficacious method (DMPA or LARC) vs. Less efficacious method (OCP, patch, ring)	Intervention participants were significantly more likely than control to be using a more efficacious method at 20 months post-intervention (OR 3.65, 95% CI: 1.26–10.08)	Moderate
Thiel de Bocanegra et al., 2017	Prospective case-control	365	365	Racially diverse women ages 13–30+ years currently using eligible contraception	115 days	OCP, patch, ring, contraceptive injection	43% of women in intervention group returned on time for their refill appointment (34% of OCP, patch, ring users and 67% of injection users). There were no significant differences between the intervention and comparison groups (overall p=0.27).	Weak
Trent et al., 2015	RCT	50	50	Predominantly Black women ages 13–21 currently using DMPA	9 months	DMPA	Intervention participants returned significantly closer to appointment date than control for 1st appointment (β=−0.75 days, 95% CI: −1.4–0.06; p=0.03), but not 2nd or 3rd appointments (β point estimates not reported).	Weak
Castano et al., 2012	RCT	480	482	Racially diverse women ages 13–25	6 months	OCP	Intervention participants were more likely than control to continue OCP (taken a pill in last 7 days) at 6 months (OR: 1.44, 95% CI: 1.03–2.00).	Moderate
Hou et al., 2010	RCT	41	41	Predominantly White women ages 18–31, new OCP users	3 months	OCP	Mean number of missed pills per cycle did not differ significantly between the intervention and control groups (4.9±2.6 vs. 3.5±3.4, p=0.60).	Moderate

Note: CI = Confidence interval; DMPA = Depot medroxyprogesterone; LARC = Long acting and reversible contraception; OCP = Oral contraceptive pills; OR = Odds ratio; RCT = Randomized controlled trial; SMS = Short message service

Table 3:

Characteristics of mHealth interventions for contraceptive behavior change

Author, Year	Target population	mHealth modality/App name	Study objective	Theory of health behavior change	Intervention	Control group treatment
Initiation or Initiation/Continuation
Manlove et al., 2020	Young women	Web-based “app” and SMS	Assess effectiveness of app-based teen pregnancy prevention program	Theory of Planned Behavior; Social Learning Theory	Web-based “app” accessed through smartphone, plus 16 unidirectional sexual and reproductive health SMS messages for 6 weeks	Web-based “app” on general health topics, plus 16 unidirectional general health SMS messages for 6 weeks
Rinehart et al., 2020	Young women	SMS	Test the feasibility, acceptability, and initial efficacy of a text intervention on supporting the sexual health of young women in primary care	Health Belief Model	Mix of unidirectional and bidirectional SMS messages, including links to websites and graphics; Periodically over 3 months	Standard clinic care
Shrier et al., 2020	Young women	Mobile app	Develop and pilot test counseling plus mHealth intervention to reduce risk for pregnancy and STIs in adolescent and young adult women with depressive symptoms and high-risk sexual behavior	Behavior-Determinant-Intervention Logic Model	Mobile app downloaded during initial counseling session, used for 1 week, additional counseling session to refine app messages, used for 1 month, then booster counseling session	N/A
Hebert et al., 2018	Young and adult women	Mobile app (miPlan)	Evaluate the effect of a waiting-room contraceptive counseling app on interest in and LARC uptake	Transtheoretical Model; Theory of Planned Behavior	Viewed mobile app on tablet computer with information and short (<1 min) videos provided before clinic appointment for approximately 10 minutes, plus routine counseling	Routine clinic counseling with reproductive health counselor
Yao et al., 2018	Young people	SMS	Test whether a text messaging intervention could improve sexual health knowledge, attitudes, self-efficacy, intention and behaviors	Health Belief Model; Social Cognitive Theory; Theory of Planned Behavior	Unidirectional educational SMS messages; 2 per week for 3 months Bi-directional SMS messages to assess survey knowledge, intentions, behavior; Baseline, 1 week after intervention, 2 months after intervention	N/A
Chernick et al., 2017	Young women	SMS	Evaluate the feasibility and acceptability of a text messaging intervention among adolescent ED patients	Health Belief Model	Unidirectional SMS; Ranged from daily to every 5 days over 3 months	Standard counseling on reproductive care from ED physicians and referral card to FP clinic
Bull et al., 2016	Young people	SMS	Assess the impact of a text message intervention on sex acts protected by condoms or contraception	Integrated theory of mHealth	SMS messages (40% bi-directional); 5–7 messages per week for 25 weeks, in addition to in-person youth development program with sex education	In-person youth development program with sex education
Sridhar et al., 2015	Adult women	Mobile app (Plan A Birth Control)	Design a contraception counseling application and evaluate the effects of the app on contraceptive method selection and knowledge	None specified	Viewed mobile app on tablet computer provided before clinic appointment for approximately 15 minutes	Contraceptive counseling from health educator with same content as the app
Gilliam et al., 2014	Young and adult women	Mobile app	Increase LARC awareness and interest prior to the clinic visit	Theory of Planned Behavior	Viewed mobile app on tablet computer provided before clinic appointment for up to 15 minutes	Standard contraceptive counseling by clinic counselor
Cornelius et al., 2013	Young people	SMS	Examine the feasibility and acceptability of a multimedia cell phone approach as an adjunct to in-person curriculum on HIV and condom knowledge and behaviors	None specified	Bi-directional multimedia SMS (text, pictures, videos) responded to by a trained facilitator (following 7 weeks of face-to-face curriculum); Daily for 3 months	N/A
Suffoletto et al., 2013	Young women	SMS	Pilot test a text message sex risk reduction program among young adult female patients	Health Belief Model; Information Motivation Behavior model	Bi-directional SMS messages that assessed risky behaviors, provided personalized feedback, and prompted collaborative goal setting; Once per week for 3 months	Weekly SMS with reminder of follow-up survey at end of 3 months
Juzang et al., 2011	Young men	SMS	Test the feasibility of recruiting and retaining men into an SMS program and to assess the program’s effect on HIV-related risk behaviors	None specified	Mostly unidirectional SMS messages; 3 times per week for 3 months	Unidirectional SMS messages about nutrition, sent over 3 months
Continuation
Dehlendorf et al., 2019	Young and adult women	Mobile app (My Birth Control)	Evaluate the impact of decision support tool on contraceptive continuation, patient experience, and quality of decision making	None specified	Viewed mobile app on tablet computer provided before clinic appointment with survey results at end to share in provider consult	Routine clinical care
Buchanan et al., 2018	Young women	SMS	Evaluate the longitudinal impact of a 9-month text intervention on adherence to highly effective contraceptive methods	None specified	See Trent et al. 2015	See Trent et al. 2015
Thiel de Bocanegra et al., 2017	Young and adult women	SMS	Determine if text message and email reminders increase contraceptive continuation and appointment rates	None specified	Unidirectional SMS messages; Sent at “method-specific” intervals, median enrollment 115 days	N/A
Trent et al., 2015	Young women	SMS	Evaluate the feasibility, acceptability, and preliminary effectiveness of a text message reminder system on DMPA continuation	Geser’s sociological framework	Bi-directional SMS; Daily starting 72 hours before scheduled visit over 9 months	Automated reminders by phone call
Castano et al., 2012	Young women	SMS	Estimate whether daily educational text messages affect OCP continuation at 6 months	None specified	Mostly unidirectional SMS messages; Daily for 6 months	Routine clinic care, including contraceptive counseling by staff and educational handout
Hou et al., 2010	Adult women	SMS	Estimate whether women receiving daily text message reminders have increased OCP adherence	None specified	Unidirectional SMS; Daily for 3 months	None

Note: DMPA = Depot medroxyprogesterone; ED = Emergency department; FP = Family planning; HIV = Human immunodeficiency virus; LARC = Long acting and reversible contraception; OCP = Oral contraceptive pills; SMS = Short message service; STI = Sexually Transmitted Infection

The intervention periods ranged from 15 minutes during pre-clinical visits in four studies (Dehlendorf et al., 2019; Gilliam et al., 2014; Hebert et al., 2018; Sridhar et al., 2015), 5–6 weeks in two studies (Manlove et al., 2020; Shrier et al., 2020), 3 months for 7 studies (Chernick et al., 2017; Cornelius et al., 2013; Hou et al., 2010; Juzang et al., 2011; Rinehart et al., 2020; Suffoletto et al., 2013; Yao et al., 2018), and between 4–9 months for five studies (Buchanan et al., 2018; S. Bull et al., 2016; Castaño et al., 2012; Thiel de Bocanegra et al., 2017; Trent et al., 2015). Two-thirds of the studies (12; 67%) were randomized control designs (S. Bull et al., 2016; Castaño et al., 2012; Chernick et al., 2017; Dehlendorf et al., 2019; Gilliam et al., 2014; Hebert et al., 2018; Hou et al., 2010; Manlove et al., 2020; Rinehart et al., 2020; Sridhar et al., 2015; Suffoletto et al., 2013; Trent et al., 2015) and one-third (6; 33.3%) were quasi-experimental study designs, including cohort (Cornelius et al., 2013; Shrier et al., 2020; Yao et al., 2018) and case-control (Buchanan et al., 2018; Juzang et al., 2011; Thiel de Bocanegra et al., 2017).

Intervention characteristics

Eleven of the 18 identified interventions (61%) utilized SMS messaging, of which six were exclusively or mostly unidirectional (Castaño et al., 2012; Chernick et al., 2017; Hou et al., 2010; Juzang et al., 2011; Manlove et al., 2020; Thiel de Bocanegra et al., 2017), two were bidirectional (Cornelius et al., 2013; Trent et al., 2015), and three were a combination of both (S. Bull et al., 2016; Rinehart et al., 2020; Yao et al., 2018). Three interventions exclusively utilized mobile apps, which were installed on tablets and used by patients ahead of clinical visits (Gilliam et al., 2014; Hebert et al., 2018; Sridhar et al., 2015). One intervention involved the use of a web-based app accessed through an internet browser on a smartphone, supplemented by SMS messages (Manlove et al., 2020). Four interventions utilized mHealth messaging to supplement or reinforce in-person messages: two used SMS messages (one post-program (Cornelius et al., 2013) and one concurrent to programming (S. Bull et al., 2016)), while two utilized native apps installed on a tablet before a clinical visit (Dehlendorf et al., 2019) and on a personal smartphone after counseling (Shrier et al., 2020).

The majority of the studies (12; 67%) measured outcomes associated with initiating contraceptive methods, four of which provided educational messages regarding contraceptive use and for three of the four, focused on specific methods (LARC) and measured initiation of methods after a clinical encounter (Chernick et al., 2017; Gilliam et al., 2014; Hebert et al., 2018; Sridhar et al., 2015). Eight of the 12 looked at both initiation and continuation over time. Six provided both educational messages and messages that they theorized would change norms and prompt behavior change (S. Bull et al., 2016; Cornelius et al., 2013; Juzang et al., 2011; Manlove et al., 2020; Rinehart et al., 2020; Yao et al., 2018), while two focused more in-depth on supporting participants to make a plan for behavior change (Shrier et al., 2020; Suffoletto et al., 2013). The remaining six studies (33%) looked specifically at contraceptive continuation outcomes (Buchanan et al., 2018; Castaño et al., 2012; Dehlendorf et al., 2019; Hou et al., 2010; Thiel de Bocanegra et al., 2017; Trent et al., 2015). Five of these involved SMS interventions which served as reminders to either take an oral contraceptive pill (Castaño et al., 2012; Hou et al., 2010) or return to a clinic for a method-specific appointment (Buchanan et al., 2018; Thiel de Bocanegra et al., 2017; Trent et al., 2015) and one was an app that provided both contraceptive information and facilitated decision-making between clients and providers at the clinical visit (Dehlendorf et al., 2019). Eight studies (44%) evaluated use of all/any contraceptive method (Buchanan et al., 2018; S. Bull et al., 2016; Chernick et al., 2017; Dehlendorf et al., 2019; Manlove et al., 2020; Rinehart et al., 2020; Shrier et al., 2020; Sridhar et al., 2015), while four specifically focused on condoms (Cornelius et al., 2013; Juzang et al., 2011; Suffoletto et al., 2013; Yao et al., 2018), four on short-acting methods (Castaño et al., 2012; Hou et al., 2010; Thiel de Bocanegra et al., 2017; Trent et al., 2015), and two on LARC (Gilliam et al., 2014; Hebert et al., 2018).

Ten of the 18 studies (56%) mentioned theoretical frameworks that informed the design of their interventions or message development. The most commonly used theories were the theory of planned behavior (Gilliam et al., 2014; Hebert et al., 2018; Manlove et al., 2020; Yao et al., 2018) and the health belief model (Chernick et al., 2017; Rinehart et al., 2020; Suffoletto et al., 2013; Yao et al., 2018). Less commonly mentioned theories or frameworks were the transtheoretical model (Hebert et al., 2018), the behavior-determinants-intervention logic model (Shrier et al., 2020), social learning theory (Manlove et al., 2020), the integrated theory of mHealth (S. Bull et al., 2016), and Geser’s sociological framework (Trent et al., 2015).

Contraceptive behavior change

None of the educational interventions found any significant differences in contraceptive initiation or the initiation of a LARC/very effective method between the intervention and control groups (Chernick et al., 2017; Gilliam et al., 2014; Hebert et al., 2018; Sridhar et al., 2015). The remaining eight interventions that measured initiation and continued or consistent use over time had generally null results, with most showing no impact on either outcome. The exceptions were the intervention with SMS messages specifically tailored to American Indian/Alaska Native youth which saw significantly higher odds of self-reported condom use at follow-up (Yao et al., 2018) and a small pilot sample of app use combined with in-person counseling sessions, which saw a decline in the median number of self-reported condom-unprotected sex events, but no change in hormonal contraception use (Shrier et al., 2020). Two additional studies (from the combined web-based app and SMS intervention (Manlove et al., 2020) and an SMS intervention with embedded graphics and website links (Rinehart et al., 2020)) saw significantly less reported sex without hormonal contraception and higher odds of prescription contraceptive use, respectively, at follow-up among the intervention groups; however these differences disappeared when the samples were limited to those who were sexually active at baseline.

For contraceptive continuation, the results of impact on behavior were also limited. While one SMS reminder intervention found no significant differences in missed pills (as measured by an electronic pill pack) between the intervention and control groups (Hou et al., 2010), the other similar reminder intervention found that intervention participants were significantly more likely to self-report having taken a pill the past 7 days (Castaño et al., 2012). Interestingly, while the intervention which sent SMS reminders for injectable contraception appointments found that intervention participants returned significantly closer to the appointment date for their first, but not subsequent, appointments (Trent et al., 2015), a post-intervention chart review of these study participants found that intervention participants were significantly more likely to be using a more efficacious method than controls 20 months later (Buchanan et al., 2018). The final SMS intervention found no significant differences in on-time appointments for method refill/reinjection between intervention and control matched pairs (Thiel de Bocanegra et al., 2017). Finally, the one app-based intervention focused on continuation found no significant differences in continuation of chosen contraceptive method between the intervention and control groups (Dehlendorf et al., 2019).

Quality assessment

Of the 18 articles, 7 received overall scores of “weak” (S. Bull et al., 2016; Chernick et al., 2017; Cornelius et al., 2013; Shrier et al., 2020; Thiel de Bocanegra et al., 2017; Trent et al., 2015; Yao et al., 2018), 8 were rated “moderate” (Buchanan et al., 2018; Castaño et al., 2012; Gilliam et al., 2014; Hebert et al., 2018; Hou et al., 2010; Juzang et al., 2011; Manlove et al., 2020; Suffoletto et al., 2013), and 3 were rated “strong” (Dehlendorf et al., 2019; Rinehart et al., 2020; Sridhar et al., 2015). The most common reasons for overall “weak” ratings were issues of possible selection bias in samples with high rates of participation refusal, cohort study designs, or missing information on potential differences in characteristics between intervention and control groups at baseline.

Discussion

In this systematic review, we identified 18 studies which evaluated the impact of mHealth interventions on contraceptive behaviors in the US. We found that mHealth interventions were primarily utilized in three ways: as part of targeted messaging to people deemed to be at risk of unintended pregnancy; as a corollary to in-person clinical counseling; and as reminders to take oral contraceptives or return to health facilities for method-specific appointments. Our results indicate limited evidence that existing mHealth interventions influence contraceptive initiation or continuation over time.

Twelve of the 18 identified mHealth interventions (67%) over the past decade included educational messages to increase knowledge related to sexual health contraceptive methods, though 10 of these 12 interventions also provided decision support or motivational messaging to encourage uptake and/or continuation of contraception. The research literature has shown that provision of informational and educational mHealth messages has a significant impact on increasing sexual and reproductive health knowledge (Hall et al., 2013; Lim et al., 2008; Zapata et al., 2018). However, a key question that motivated this review was whether the messages that increase knowledge also result in behavior change. Through the process of this review, we confirmed that even when paired with decision support or motivational messages, educational mHealth messages minimally impact contraceptive initiation and continuation, even while simultaneously improving knowledge and attitudes, which is consistent with evidence from contraceptive counseling in clinical settings (Moos et al., 2003). Three of the four studies that reported differences in intervention groups saw any initial changes disappear within a few weeks or months or when the sample was limited to those who were sexually-active at baseline. An additional challenge with determining the impact of these interventions on behavior change is that many had self-reported outcomes, particularly related to condom use, that may be subject to response or recall bias.

The disconnect between increases in knowledge among participants not translating to subsequent changes in behavior is important in considering the theoretical bases from which just over half of the studies grounded the structure of their interventions. Most mentioned the Theory of Planned Behavior or the Health Belief Model as guiding how they hypothesized their interventions would influence contraceptive behavior. These theories include consideration of participants’ increased positive attitudes toward a behavior, the expectation that engaging in the behavior will have a positive influence on their lives, reduced perceived barriers encouraging the behavior, and increased intentions to perform the behavior to stimulate engagement in the behavior (Ajzen, 1991; Champion & Skinner, 2008). For the mHealth interventions taking place within health facilities in this review, participants have seemingly overcome many of the barriers to contraceptive use already; however, for interventions delivered to a disadvantaged, or even general population, existing mHealth interventions may not sufficiently change attitudes or support participants in converting their intentions to actions.

Moreover, as noted in a systematic review of the use of theory in mobile apps, while studies may cite the use of theory in the design of mHealth intervention, few studies actually test the intervening mechanisms, and whether the interventions actually change those intervening outcomes remains unclear (Chib & Lin, 2018). While text-messaging interventions have been shown to be effective for behavior change in other areas, including physical activity and smoking cessation (Scott-Sheldon et al., 2016; D. M. Smith et al., 2020), there appear to be different barriers to contraceptive behavior, especially in the US. For example, without access to universal health care, 28.9 million Americans still lacked health insurance in 2019 (Tolbert et al., 2020), limiting their ability to readily access the full range of health care services, including reproductive health care and contraception.

This systematic review focuses on contraceptive behavior with the awareness that broader attention is being paid to patient-centered frameworks for contraceptive care that consider contextual and historical factors, policies and health systems, and community and family contexts in providing high-quality, respectful and accessible care. Current practice recognizes the need for non-coercive counseling, support to facilitate method switching (which may involve discontinuation), and information on side effects management (Gavin et al., 2017; Holt et al., 2020). None of the three app interventions in clinical settings measured why people did not choose any particular method and the field lacks a validated scale for measuring how well mHealth interventions honor reproductive self-determination. Scales such as the Interpersonal Quality of Family Planning, which measures how well patient preferences are honored in the contraceptive appointment (Dehlendorf et al., 2018), could potentially be adapted for mHealth research (Holt et al., 2019; Upadhyay et al., 2014). While it was not the goal of this review to determine if existing mHealth interventions measure elements of reproductive autonomy, we did identify two studies which measured satisfaction with contraceptive counseling in addition to behavior change. One study determined that intervention participants who received contraceptive information from the app intervention were less satisfied with their contraceptive counseling (Sridhar et al., 2015), while another determined that while the utilization of a tablet-based app to facilitate shared decision-making with the health care provider did not lead to higher rates of contraceptive continuation or contraceptive satisfaction, it did result in higher ratings of interpersonal quality of counseling and more satisfaction with contraceptive counseling about side effects (Dehlendorf et al., 2019). Additional research is needed to determine if and in what formats mHealth can best be deployed to improve contraceptive satisfaction, as both stand-alone interventions and as corollaries to clinical care, and to assure that interventions honor reproductive self-determination.

In addition, it is surprising that none of the identified interventions collected measures of side effects or reasons for discontinuation. Concerns about side effects, particularly menstrual bleeding changes, are the most common reasons for method dissatisfaction and discontinuation (Daniels, 2018; Moreau et al., 2007; Polis et al., 2018), rather than missed appointments, a common target of these interventions. These concerns may be mitigated by information and other support provided through mHealth and would be an important target for future research. Additional statistical analyses that examine the mediating or moderating impact of side effects, demographic characteristics, or previous contraceptive use experiences on the relationship between these interventions and continuation would be another useful area of investigation. Finally, none of the studies measured or were able to control for measures of pregnancy “ambivalence”, which may have been associated with inconsistent contraceptive use or discontinuation (Higgins et al., 2012; Samari et al., 2020).

Interventions outside of clinical settings are essential for providing information, and yet people also need to be linked to services, particularly to initiate hormonal contraceptive use. Only two identified interventions addressed resource or access barriers people may face: one could connect participants with their tribal clinics for sexual health appointments (Yao et al., 2018) and the another provided links to external resources and a clinic locator page (Manlove et al., 2020). More mHealth interventions should consider integrating clinic finder features or online resources for hormonal contraception (such as Nurx (Nurx, 2020)), particularly for young people or people without health insurance.

Before 2015, mHealth interventions were largely reliant on SMS reminders to influence contraceptive use behavior. More recently, apps are increasingly utilized and are even being marketed, both with and without medical devices, as contraceptive tools themselves (J. Bull et al., 2019; Jennings et al., 2019). A tremendous number of family planning apps are available for download in the Apple and Android app stores but their impact on behavior is largely untested (Chen & Mangone, 2016; Lunde et al., 2017; Mangone et al., 2016). As these apps continue to proliferate, it is increasingly important to evaluate the accuracy of their content, as well as their effectiveness and efficacy in changing behavioral outcomes, preventing pregnancy, and whether they will be most effective integrated into other apps or as stand-alone products.

This systematic review has several limitations. Our findings indicate that the science of mHealth and contraceptive behavior change is still in its infancy. We are limited in our capacity to determine the effect of mHealth characteristics because five of the 18 included interventions were pilot studies not sufficiently powered for efficacy, but rather were testing acceptability and feasibility. Additionally, though the majority of included studies were randomized controlled designs, sample sizes were often small. Long-term follow-up was also limited, with most included interventions collecting data for around 3 months post-intervention. Thus, there is currently an incomplete picture of whether or not behavioral change and health outcomes were sustained in all but one of these studies (Buchanan et al., 2018). Larger, statistically-powered trials are critical for understanding if and how mHealth can effectively impact contraceptive behavior change. Additionally, we caution against drawing conclusions about mHealth as a monolithic approach: our criteria were broad and inclusive, including a variety of mHealth strategies and interventions exclusively focused on condom use and hormonal contraception. Therefore, the set of studies were heterogenous in nature, representing the current status of this field.

Questions remain regarding the ability of mHealth interventions to influence contraceptive use behavior. Key takeaways include the importance of pairing current mHealth interventions in clinical settings with in-person counseling, rather than having them serve as a replacement, and that interventions targeted and tailored to specific groups may be more impactful at motivating behavior change. Researchers and practitioners should also consider measuring other outcomes that can contribute to contraceptive initiation and continuation, such as satisfaction with both counseling and chosen contraceptive method over time. While the early research in this area was heavily focused on SMS appointment reminders and educational messages, as more people gain access to smartphones, additional mobile app interventions and other emerging mHealth technology should be investigated as these interventions can be more interactive and personalized. Finally, the majority of interventions to date tracked people for 3 months or less. Longer-term follow-up is needed to investigate the continued impact of mHealth interventions over time.

Conclusion

This review summarizes the literature on mHealth interventions and contraceptive behavior change over the past decade in the US. Because more people than ever have access to smartphones and tablet computers, and with mHealth interventions showing promise in other health areas, there is tremendous potential to impact contraceptive use with mHealth. However, interventions implemented to date show very limited impact on contraceptive initation and continuation. Any new mHealth interventions in this area must reflect on what has not worked and consider designing interventions that focus more specifically on reasons for non-use or discontinuation, including side effects management.