Mediators of a 12-Month Change in Physical Activity in Ethnically Diverse Sample of Postpartum Women : Translational Journal of the American College of Sports Medicine

Secondary Logo

Journal Logo

Original Investigation

Mediators of a 12-Month Change in Physical Activity in Ethnically Diverse Sample of Postpartum Women

Albright, Cheryl L.1,2; Wilkens, Lynne R.2; Saiki, Kara1; White, Kami K.2; Steffen, Alana D.3

Author Information
Translational Journal of the ACSM 4(19):p 215-224, October 1, 2019. | DOI: 10.1249/TJX.0000000000000106



Every 8 s, a baby is born in the United States resulting in almost 4 million women becoming new mothers each year (1). Thus, millions of women a year face a life-altering transition that is often accompanied by changes in their social, cultural, economic, and personal (family/friends) situations. Such changes can demonstrably alter their priorities in life, including their pursuit of and interest in regular, planned leisure-time moderate to vigorous physical activity (MVPA). Up to a year after delivery, 68% of women are not regularly active, reporting they rarely or never exercise (2), and this often represents a substantial reduction compared with their prepregnancy levels of physical activity (PA) (3,4). Furthermore, epidemiological studies have found that women with children are more inactive than similarly aged women without children (5). Such shifts place this vulnerable population at risk for physical inactivity and weight gain retention that can, over time, increase their risk of developing hypertension, obesity, diabetes, and other chronic diseases (6–8).

Several studies have tested the efficacy of PA interventions, alone or in combination with other health behaviors, to increase postpartum women’s MVPA or achievement of the national recommendation of 150 min of MVPA a week (1,9–11). Many of these interventions were based on a priori theoretical models focused on specific constructs from social cognitive theory (SCT) (12–16) and transtheoretical model (TTM) (17,18). Many involved largely White or African American postpartum women and found a significant MVPA increase in response to short-term interventions (9,13,19). Data from the Hawaii Behavioral Risk Factor Surveillance System has shown that more non-Hispanic White women (22.5%) meet national recommendations for MVPA (i.e., 150 min MVPA and strength building more than 2 times a week), compared with Asian-American, Native Hawaiian, and other Pacific Islander (collectively abbreviated as AANHPI) women, where the range is from 9% to 21% (20). Such race/ethnic disparities led Albright and colleagues (15) to conduct a 12-month MVPA trial that randomly assigned postpartum women, predominantly AANHPI, to a tailored telephone counseling plus website (TTCW) MVPA condition that used SCT/TTM theoretical constructs or a standard/usual care website-only MVPA condition. This trial found positive increases in MVPA minutes per week in both groups, with significantly higher increases in the tailored group. Because of the lack of information regarding long-term change in MVPA in new mothers, determining mediators of PA interventions in postpartum women, particularly from these ethnic backgrounds, who are at risk for physical inactivity and who are underrepresented in behavioral research, is important.

Most PA interventions based on SCT and TTM are designed to facilitate increases in MVPA by operationalizing behavioral constructs, such as self-efficacy to increase PA, overcoming barriers to PA, enlisting social support for PA, emphasizing PA benefits while decreasing PA barriers, engaging self-regulation, and changing outcome expectations for PA, which are hypothesized to act together or independently to increase MVPA (21,22). However, only a few MVPA intervention studies with postpartum women or mothers of young children have conducted analyses that formally tested the mediation effects of these constructs; that is, whether the intervention successfully changed the potential mediators, and whether these changes were in turn associated with an increase in MVPA (12,16,23–25).

Studies have shown that self-regulatory efficacy (i.e., confidence in one’s regulatory skills such as learning how to set a weekly MVPA goal and then planning when/where to walk) is a significant mediator among mothers of infants (23), and that partner support, self-efficacy to overcome barriers, and self-regulatory skills are mediators among mothers of young children (5 yr old or less) (12,26) as well as in working mothers of children up to 15 yr old (24). Other PA intervention studies with postpartum women have found that the theoretically derived constructs from SCT or TTM are not significant mediators of change in MVPA (12,27).

The multicomponent Nā Mikimiki (translated as “the active ones” in Hawaiian) Project’s intervention was hypothesized to increase MVPA by changing intermediate SCT/TTM constructs such as self-efficacy to overcome barriers to PA, social support for PA, perceived benefits, and barriers associated with increasing PA. We have already reported a significant increase in MVPA minutes per week after 12 months of the Nā Mikimiki intervention, compared with a standard PA information website condition (15) (202 vs 110 min per week, P = 0.03). The purpose of this paper is to investigate whether these changes in MVPA were mediated by theoretically derived SCT/TTM constructs. Thus, we a priori hypothesized that counseling participants in the TTCW condition about how to use self-monitoring, self-evaluation, and self-reinforcement through goal setting and problem solving of barriers would develop skills to help them enlist social support and build self-efficacy, which would, in turn, enhance their ability to integrate MVPA into daily routines. The intervention components targeting mediators were delivered during scheduled telephone counseling calls over the 12-month period.


Women were recruited from two sources: community-wide campaigns and a mail campaign for female members of a local health care organization (Kaiser Permanente, Honolulu, HI); thus, the study’s protocols and assessments were approved by both the University of Hawaii’s Office of Human Studies and the Kaiser Permanente Institutional Review Board affiliated with the Center for Health Research–Hawaii, and both boards monitored the project’s accrual rates and consent protocols. Recruitment of Kaiser members included mailed letters and follow-up phone calls to women in the eligible age and postpartum time frames, whereas community-wide campaigns included advertisements in free parenting magazines and other community-wide activities (14). Potential participants attended a scheduled baseline visit at the University of Hawaii during which they read and signed an informed consent. A Data Safety Monitoring Board monitored severe adverse events and other study outcomes over the course of the study.

Study Design

The Nā Mikimiki Project was a parallel randomized controlled trial comparing two 12-month PA interventions: 1) a TTCW condition and 2) a standard website-only (SWO) condition that was conceived as a “usual care” online PA information-only condition (14).

Participant Recruitment and Randomization

Details on the study design and methods of recruitment/randomization (i.e., CONSORT recruitment flow diagram) are reported elsewhere (15). Briefly, to be eligible, women had to be healthy, inactive (reported 30 min or less of leisure-time MVPA over the last week), 18–45 yr of age, 2–12 months postpartum, and not pregnant or planning to become pregnant over the ensuing 12 months and to have health insurance and a body mass index (BMI) between 18.5 and 40 kg per meter squared. Women who were 2–12 months postpartum were selected for eligibility to allow for consistent messaging on MVPA and based on the exercise preferences of women (i.e., walking with their baby in a stroller) determined in our pilot studies (2,28) and in previous exercise interventions with new mothers (29–31). Of the 311 randomized participants, 154 were assigned to the TTCW and 157 to the SWO condition (15). Over the 12 months, 28 women’s participation (7.4%) ended due to a pregnancy (16 from TTCW, 12 from SWO), 31 women (10%) asked to be removed from study, and 3 women ended due to illnesses unrelated to PA. Overall, 62 women (19.9%) did not complete the study, 41 (13.2%) were in TTWC, and 21 (6.8%) were in SWO.

Study Conditions

The conditions’ websites were demonstrably different. The TTCW website included resources tailored specifically to new mothers in Hawaii, whereas the SWO website provided links to standard information on strategies for increasing PA, from credible national online PA resources (i.e., American Heart Association, Centers for Disease Control and Prevention, and National Institutes of Health). The information on these websites often addressed the benefits of PA, goal setting, advice, and resources; however, this information was for adults of any age and did not address issues related to children or infants. Both conditions received incentives for participating in study assessments ($60; $10 gift cards at baseline and 6 months, $20 gift cards at 12 and 18 months) and small gifts/birthday cards over the course of the 12 months. However, SWO participants did not receive any telephone calls over the 12 months other than contacts regarding study-related assessments (that women in both conditions received).

The TTCW intervention used theoretically derived constructs from SCT to initiate and maintain MVPA using tailored resources that were available both on its website and were addressed during personalized counseling delivered via 17 telephone calls. The PA-related resources included information such as the location/names of “stroller friendly” (i.e., had a paved sidewalk/path) parks and tips for giving PA support to a new mother (participant provided this to their partner/other family member/friend), for staying on track even during bad weather, for setting step goals using a pedometer, and for keeping PA goals when her baby was sick/fussy. The website also had a calendar with community classes that were updated monthly, links to local postpartum PA groups (i.e., Stroller Strides [FIT4MOM]/baby boot camp), and programs women could use to map their walking route.

Telephone counselors also used motivational interviewing techniques to problem solve ways new mothers could overcome PA barriers, to highlight the benefits of MVPA during the postpartum period, to enlist social support for PA from family/friends, and to set MVPA goals in reference to type of activity, duration, intensity, and location. Setting of realistic, incremental MVPA goals was used to increase the mothers’ self-efficacy to increase their MVPA gradually and safely. During the telephone calls, counselors discussed how participants could obtain emotional, tangible, facilitative, and companionship social support for PA from her family and friends. For example, counselors asked participants: “Do you have someone who can offer you support this week/month?” “If yes, who is it? How can they support you?” If they could not think of someone, counselors would guide them on how to identify a support person specifically for PA. Counselors also discussed ways to negotiate with their support person (typically their partner) how long of a walk (e.g., 10 min) could she take while her partner watched the baby/other children or how to negotiate so that the walk was a family affair. The counselors also provided emotional support for achievement of MVPA goals. Specific details and examples of the TTCW intervention strategies and materials are provided elsewhere (14).

On the basis of the current national recommendation for MVPA established by the American College of Sports Medicine (32,33), participants were informed that their ultimate goal (by 12 months) was 30 min per day of MVPA, 5 days a week (i.e., 150 min of MVPA per week). The 30 min could be broken into bouts as short as 10 min, although participants were encouraged to progress to longer bouts over time. Although all forms of MVPA were discussed, the counselors focused on planned, purposeful leisure-time MVPA, and for most women, this was achieved by walking while pushing their baby in a stroller in their own neighborhood, which is an activity determined by oxygen consumption testing to be a moderate-intensity activity (34,35). The content and high fidelity of the PA counseling telephone calls is reported elsewhere (15).


Detailed descriptions of all study measures are available elsewhere (14); briefly, outcome data were collected at baseline and at 3, 6, and 12 months postbaseline. The baseline and the 12-month data were collected at in-person visits, whereas data for the other time points were collected via mail or online. Brief descriptions of the measures are provided below.

Primary outcome MVPA

The Active Australia Survey (AAS) assesses frequency, intensity, and duration of activities performed for at least 10 min over the last week: walking for recreation, exercise, or to get from place to place; moderate-intensity PA; and vigorous PA (36). Moderate intensity was defined as “gentle swimming, social tennis, golf,” and vigorous intensity was “activity made which you breathe hard (e.g., jogging, cycling, aerobics, competitive tennis).” The AAS was administered at baseline and at 6 and 12 months. Total minutes of MVPA per week was calculated as the sum of minutes of walking, MVPA, with the latter weighted by two (37). The AAS survey has been shown to have good reliability (38), sensitivity to change (39), and validity when compared with accelerometers using Spearman’s correlation coefficients: rs = 0.28, P = 0.01 (40) and rs = 0.52 for moderate- or higher-intensity PA, P = 0.001 (36). A recent validity study compared ActiGraph MVPA time measurements to AAS MVPA time and found an intraclass correlation coefficient = 0.44, P < 0.01 (41). MVPA in Nā Mikimiki was measured via accelerometer at baseline, and at 3, 6, and 12 months (15). However, MVPA using accelerometry will not be considered as a primary outcome in this mediator analysis. A large proportion of our inactive women in Nā Mikimiki appeared to be physically active based on their baseline accelerometry levels due to reactivity. The intervention effect observed for the AAS was mirrored in women who were classified as sedentary by baseline accelerometry. However, the sample size of this subgroup was small at n = 130, which carries a low power (20%) to detect mediation (with alpha = 0.05, two-sided) based on correlations of magnitude 0.2 as observed in this study. An n of 250 or greater is needed for 80% power in this setting.


Detailed descriptions, including psychometric properties, of the mediators are available elsewhere (14). All measures were validated scales, had high Cronbach alpha values, and were adapted for use with postpartum women. For example, questions assessing self-efficacy to overcome barriers to PA asked women to report their level of confidence to exercise when their infant was sick. The range of scores was 0–10 for self-efficacy, with 10 representing the highest level; 1–5 for perceived barriers, with five representing the strongest barrier; 1–5 for benefits of regular PA with five representing the strongest benefits achieved by regular MVPA; and 1–5 for social support, with five representing the highest level of support from family and friends to be physically active (14). Mediators were collected at baseline and at 6 and 12 months.

Change in MVPA over 12 Months

Participants in the TTCW condition had a significantly larger increase in MVPA over the 12-month study (average increase of 202 min per week) than the SWO condition (average increase of 110 min per week) (P = 0.03), based on self-report. The Cohen’s d effect size for the difference at 12 months was 0.36 (15).

Statistical Analysis

Hypothesized mediators of the intervention on PA were tested individually using longitudinal autoregressive mediation path models as depicted in Fig. 1; each parameter shown was freely estimated (42). The timing of the measurements gave us the ideal setting for longitudinal analysis, whereby the effect of a change in each mediator at an earlier time period, such as from baseline to 6 months, on a change in MVPA in a later period, such 6 to 12 months, is assessed. In general, the model includes both indirect effects of the intervention on MVPA through a potential mediator and direct effects, along with correlations between MVPA and mediator variables. In particular, the term “intervention,” in Fig. 1 and Table 2, represents a dichotomous variable indicating the “test” treatment (TTCW) condition versus the “comparison” (SWO) condition, which was modeled to affect the mediator and outcome (MVPA) at all three time points. The parameters (depicting paths) OT and MT measure the condition differences for the outcome and mediator at time T, respectively. Therefore, the O12 parameter measures the direct treatment/TTCW condition effect on MVPA at month 12. We included paths from mediators to the MVPA outcomes one time point later but did not include direct paths to outcomes at later time points; for example, we did not model the path from a mediator at 1 month to MVPA at 12 months. These directional mediator-to-outcome effects are labeled MOB,6 for the baseline mediator’s effect on 6-month MVPA and MO6,12 for the 6-month mediator’s effect on 12-month MVPA. There are two key paths to establish the mediation of the treatment effect, shown as bold arrows in Fig. 1: the direct effect of intervention on the mediator at 6 months (M6) and the 6-month mediator’s effect on the 12-month outcome (MO6,12). In behavioral prevention, the M6 path is termed the action theory test to determine whether the intervention was successful in affecting the mediator. The MO6,12 path represents a conceptual theory test of the relationship of the mediator to the outcome (43). Other paths included in the model control for expected relationships but are not germane to the mediation results. These include the longitudinal autoregressive paths that account for the relationship of repeated measurements over time, including both the first lag between adjacent measurements (labeled as AR1 paths) as well as lag two effects (AR2), such as the relation between the baseline MVPA and the 12-month MVPA. In addition, we made no assumptions about the causal relation of mediators and outcomes measured concurrently but allowed for their covariance at each time point; these are curved lines with arrows on both ends, labeled C for covariance parameters. All models were tested with these specifications (model test with 4 degrees of freedom) without post hoc adjustment to improve model fit using Mplus version 7 (44). Confidence intervals were created for indirect effects using bootstrapping. Missing data were accounted for by full information maximum likelihood (FIML) estimation, with SE that are robust to nonnormality allowing modeling of untransformed values. FIML produces unbiased parameter estimates under the missing at random mechanism and is superior to listwise deletion when missing at random mechanism does not hold (45,46). Model fit was assessed using the likelihood ratio chi-square and the root mean square error of approximation, and comparative indices such as comparative fit index and Tucker–Lewis fit index (47). One criticism of the autoregressive mediation path model is that change is not explicitly modeled. We supplemented this approach with mixed regression models of MVPA in SAS 9.1, with time and study condition and their interaction (cross-product) terms as fixed effects, and person as a random effect to account for the repeated measures. The following transformations were used to meet assumptions of the mixed model: (x + 1)1/4 for MVPA, log(x + 1) for social support, (x + 1)1/5 for benefits; barriers and self-efficacy were untransformed. Back-transformed means at each time point and the F-tests for condition-specific slopes and for the difference in slopes between conditions (interaction) are provided to inform the mediation model.

Figure 1:
Model of longitudinal mediation of intervention effect on MVPA. * B, 6, 12 = measurement points (at baseline, 6 months, and 12 months). O, direct path from intervention to outcome; M, direct path from intervention to mediator; MO, mediator-to-outcome effects path; AR1 and AR2, longitudinal autoregressive path.
Results for Four Mediator Models of Change in MVPA.



The total sample size was 311 with 154 in TTCW condition and 157 in the SWO condition. There were no significant differences in baseline demographics between the two study conditions. Briefly, participants had a mean age of 32 yr, infants with a mean age of 5.7 months, women had an average of 2.0 children, 37.7% were primipara, 63.9% were working (full or part-time), and the mean BMI was 27.9 kg·m−2. Although all race/ethnicities were eligible to join the study, the majority of the sample (85%) were ethnic minorities, most of whom were Asians (34%) or Native Hawaiians/other Pacific Islanders (32%) (14,15). Previous analyses found that only the number of children (≥2) was a significant moderator of the TTCW intervention on women’s increase in MVPA over 12 months, with women in the TTCW intervention condition who had ≥ two children having a significantly higher increase in MVPA increase than those with one child (15). No other moderators including employment status, race, BMI, education, or age of the infant were significant. The intervention materials provided PA information and resources tailored to Hawaii, including graphics/photos on the website/in materials that depicted women/infants from the most prevalent races/ethnicities in Hawaii (AANHPI).

Table 1 includes the sample sizes and means for the hypothesized mediators (social support, barriers, benefits, and self-efficacy) across all time points (baseline, 6 months, and 12 months) as well as MVPA minutes per week across the 12-month intervention (baseline, 6 months, and 12 months) by study condition. Because an important characteristic of a mediator is that it changes in response to the intervention, this table quantifies how much the four mediators varied over the 12-month intervention period. Most of the mediators’ mean values changed substantially over the 12-month period. Specifically, significant increases over time were observed in social support for both conditions, and the change was greater, albeit not significantly, in the TTCW condition. Barriers decreased significantly and similarly in both conditions over time, and self-efficacy increased in both conditions, significantly in the TTCW condition. The scores for benefits did not change for either condition during the 12-month period.

Mean (95% CI) for MVPA and Mediators by Time Point and Study Condition.

Mediation Results

Mediation model results are shown in Table 2. For each hypothesized mediator, all the path coefficients, SE, and P values are shown; direct and mediated effects of the intervention are italicized where P < 0.10 and emphasized in bold where P < 0.05. Absolute and comparative model fit indices are shown and judged to be adequate for all models (47). Across all models, the baseline measures were not significantly different between conditions for the mediators (MB) and outcome (OB), although there was a suggestive difference of fewer barriers to MVPA, on average, for the TTCW condition(P = 0.065). The intervention (TTCW vs SWO condition) showed a direct effect on social support at 6 months (M6, P = 0.006, effect size = 0.16) controlling for the other variables in the model and their specified relationships, whereas all other hypothesized mediators did not differ by condition at 6 months. No direct intervention effects were detected at 12 months for any mediator (M12). Across most models, direct effects of the intervention on MPVA were evident at 6 months (O6) (although nonsignificant) and at 12 months (O12).

The mediator effects section of Table 2 shows parameters for the effect of mediators measured at a previous time point on the outcome measured at the subsequent time point. For baseline mediators, self-efficacy was shown to have a positive significant association with MVPA at 6 months (MOB,6) (P = 0.023), suggesting that early self-efficacy predicted later success; however, self-efficacy at baseline did not differ by study condition (shown as intervention → mediator baseline (MB), P = 0.941). Similarly, fewer barriers at 6 months were predictive of more MVPA at 12 months (P = 0.02), but again, this did not differ by condition (intervention → mediator 6 months (M6), P = 0.626). Therefore, there was no evidence that the intervention effect was mediated through changing self-efficacy or barriers. For social support (Fig. 2), a mediation effect appeared evident; the key model parameters are shown as bold lines. Social support at 6 months, which was greater for the TTCW condition (P = 0.006), was predictive of MVPA at 12 months (P = 0.04) (the bold line in Fig. 2). However, this indirect effect was not strictly significant when tested with a bootstrapped confidence interval (indirect effect = 0.045 [−0.003 to 0.118]). This is interpreted as a partial mediation because a direct effect of the TTCW condition on MVPA at 12 months exists (O12), although this did not reach statistical significance (P = 0.052). The underlying correlation matrix between the variables in the mediator models are listed in the Supplemental Content 1, Note that correlations between moderators at 6 months with MVPA at 12 months were small, with the largest being 0.24 for barriers.

Figure 2:
Longitudinal mediation of eHealth on MVPA through social support. Bold arrow indicates the pathway for mediation of social support on MVPA at 12 months; only paths significant at P < 0.05 are shown.

Table 2 shows the associations between mediators, and the outcome at the coincident time points (C parameters) were estimated as correlations. Social support was positively related to MVPA at every measurement. At 6 and 12 months, barriers (negative) and self-efficacy (positive) were associated with MVPA, whereas no associations were evident at the initial measurement. The measure used to assess a woman’s perceived benefits of being more physically active was unrelated to MVPA at all concurrent measurements. Generally, the longitudinal autoregressive relationships (AR1 and AR2) were highly significant for MVPA and for mediators, as one would expect for repeated measurements.

Two forms of sensitivity analyses were conducted for the social support model (results not shown). First, we examined if the results were robust to listwise deletion instead of FIML. The statistical conclusions were consistent, but the intervention effect (intervention → mediator 6 months [M6]) was slightly attenuated (P = 0.03) and the mediator effect (mediator 6 months → MVPA 12 months [MO6,12]) was stronger (P = 0.02). Second, we tested the model with adjustment for covariates found to be important in the main effects analyses of this study (15). This adjustment was accomplished by creating adjusted variables for all mediators and outcomes as the residuals of regressions of the mediators and outcomes on mother’s age, race, education, baby’s age, BMI, and postpartum depressive symptoms at baseline. In the mediation model using the residuals, both paths of interest were slightly attenuated, with the intervention–mediator at 6-month association (intervention → mediator 6 months) remaining significant (P = 0.01) but the mediator–MVPA association (mediator 6 months → MVPA 12 months [MO6,12]) no longer statistically significant (P = 0.065). This finding supports the assumption of our model of sequential ignorability (48), which states that randomization will balance characteristics of the two groups, including postrandomization mediators.


Thus, we a priori hypothesized that counseling participants in the TTCW condition about how to use self-monitoring, self-evaluation, and self-reinforcement through goal setting and problem solving of barriers would develop skills to help them enlist social support and build self-efficacy and would, in turn, enhance their ability to integrate MVPA into daily routines. The intervention components targeting mediators were delivered during scheduled telephone counseling calls over the 12-month period.

We hypothesized that the TTCW intervention’s use of theoretical constructs tailored to new mothers (i.e., enlisting social support, increasing self-efficacy via short-term achievable goals, etc.) would be mediators of the increase in MVPA. The results showed that social support, which improved significantly more in the intervention condition than the comparison condition in the first 6 months of the intervention, was a mediator of improved MVPA in the second 6 months. The results also confirmed that increasing self-efficacy and reducing barriers were associated with increases in PA in the first and second 6 months respectively; however, as there were no overall condition differences over time in these constructs, they were not mediators of our intervention.

Mediator results from previous (but shorter, 8–12 wk) postpartum PA interventions have yielded inconsistent findings for mediators (12,23,27). Two studies found none of the potential theoretical mediators were significant (12,27), and one found only self-regulatory efficacy (but not outcome expectancies) partially mediated the change in PA after 8 wk (23). One recent study of weight loss in postpartum Latinas found social support from family and friends increased after a supervised (by trained promotoras/community health workers) walking group intervention that also had increases in aerobic PA over 12 months (25). This study’s PA intervention was a 12-wk group walking program (49) versus our individualized telephone/website intervention that lasted 12 months. The sample consisted only of Latinas most of whom were unemployed/never employed (76%) and fewer had only one child (20%) compared with our sample (37%). Similar to our findings, they found social support for exercise increased from baseline to 6 months; however, it then decreased to baseline levels by 12 months.

A few studies of women with young children (around 5 yr of age) and working women with children up to age 15 yr have shown that social support, self-efficacy, or self-regulation were mediators of the intervention on change in PA (24,26). The duration of our study (i.e., 12 months) was purposefully set to be longer than previous studies with this population because we wanted to determine whether a long-term PA intervention could promote changes in theoretically derived constructs such that changes (from baseline to 6 months) in a construct would increase MVPA from baseline to 12 months. Thus, we did not anticipate our results would be comparable with shorter-term interventions. The finding that social support was a key mediator of our MVPA tailored intervention suggests that the intervention helped new mothers learn how to enlist support from family and friends for their long-term MVPA efforts. There was evidence that reducing barriers and increasing benefits led to a greater increase in MVPA; however, counseling women on these behaviors in the TTCW condition did not lead to a significantly greater change in these mediators than was found for the women who only received standard online PA information. Social support depends on communication with and help from others, and the results may indicate that new mothers benefit from learning strategies to enlist social support for PA during this time when they face the competing demands of a growing infant. Having a counselor as an advocate to remind women about the importance of their own health may also be important.

On the other hand, we found that women in both the TTCW and the SWO conditions were able to increase self-efficacy and reduce barriers over time and increase their PA, but these constructs were not found to be mediators. It is possible that women with new infants are able to overcome barriers and build self-efficacy naturally over time as they adjust to life with their new baby, with modest help such as the generic PA information provided to the SWO comparison condition. However, this cannot be taken to mean that future PA interventions should only focus on increasing social support for MVPA. As our intervention addressed and improved three of the four theoretical constructs, reducing barriers to MVPA and increasing self-efficacy to be more active may have indirectly affected TTCW women’s perceptions or active engagement of social support for PA.

Moreover, our models did not account for any complexity in the way the intervention would affect each potential mediator. Any theorized mediator might fail to show an effect because there was no effect, or because multiple effects cannot be dissected due to the study design, or because the changes in the mediators, while similar in magnitude, differed in type between the study conditions. To address these issues, future research could strive to improve how mediators are measured and tested, using conceptual models that link intervention components with mechanisms, before assessing the efficacy of interventions designed to change barriers, benefits, and self-efficacy, together or in sequence. This could include more complex designs, such as sequential introduction and assessment of potential mediators, such that the sequences of mediation (i.e., removing barriers leads to self-efficacy) or conditional processes, like moderated mediation, could be evaluated.

Our study had several limitations. The mediator analysis used only self-reported MVPA minutes per week. Although the original intervention effect was verified using accelerometry in the subset of inactive women who had less reactivity to the device, the sample size in the subgroup of 130 is too small for a mediator analysis (15). However, the use of self-reported MVPA data allows comparisons with the previous mediator findings from PA interventions among postpartum women, which generally used self-reported measures (interview or survey) as the PA outcome data, with Fjeldsoe et al. (10,12,26) using the same measure (AAS) as was used in our analyses. Thus, our analyses are comparable with other mediator analyses of interventions with postpartum women.

Another limitation is that we did not include every potential mediator (for example, autonomy, self-regulation, or outcome expectations) because we did not want to increase attrition due to the high burden of measurement on the participants. Respecting that mothers of infants are extremely busy, we tried to keep our assessments as concise and as brief as possible. Also, our intervention included women with young babies 2–12 months old at recruitment; therefore, the results may not apply to mothers of toddlers. Lastly, the sequential ignorability assumption of our analyses that supposes that balance achieved by random assignment also applied to our mediator variables may not be valid. However, our sensitivity analysis that adjusted all outcome and mediator values provided similar results and no evidence of confounding. The most important strength of our study was its ability to test mediation prospectively, with changes in each mediator preceding changes in the MVPA outcome. Strengths for our clinical trial include the study duration (12 months), its multiethnic sample of AANHPI women, and the large sample size (14,15). However, the 6-month long intervals between our measurements may have resulted in being underpowered to detect smaller or short-term mediation effects using this stringent longitudinal approach. Also, as women in the comparison SWO condition were provided resources that discussed similar SCT mediators as were used in TTCW (such as setting realistic goals, health benefits, and barrier reductions), the effect of the TTCW condition on the mediators and MVPA may have been underestimated.

In conclusion, although this 12-month multimethod theoretically derived intervention (telephone counseling plus a website) was tailored to meet the needs, barriers, and life situations of new mothers, 84% of whom were ethnic minorities/had mixed race heritage, the most important mediator of their change in MVPA was social support for exercise. During this life stage, in which a mother’s focus is inevitably and understandably on the baby and its needs, learning to enlist help from friends and family appears to be the most important factor for improvement in MVPA.

The project described was supported by NIH Award Numbers CA115614 and CA115614-03S1 from the National Cancer Institute. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institutes of Health.

The authors thank and show sincere appreciation to the following individuals who contributed to the implementation and completion of this study’s intervention and assessments: Paulette Yamada, Ph.D.; Brooke Hedemark; Lillian Cross, M.P.H., R.N.; Katie M. Heinrich, Ph.D.; Trina Orimoto, Ph.D.; Leslie Welsh, M.P.H.; Regina Suyderhoud; Anniken Rose, M.S.; Sonya Niess, M.P.H.; D. Chad Johnson; Dominique Freire, M.P.H.; Elise Davis, M.P.H.; Peter Hinely; Yihe G. Daida, M.S.; Ada Demleitner, M.A.; Aleli Vinoya; Valentyna S. Pishchalenko; Jennifer L. Elia, Dr.P.H.; Ray Browning, Ph.D.; Rebecca E. Lee, Ph.D.; Fedor Lurie, M.D.; Heather Hausenblas, Ph.D.; and all the postpartum women who volunteered for our study.

The authors declare that they have no conflicts of interest. The results in this article do not constitute endorsement by the American College of Sports Medicine.


1. Hamilton BE, Martin JA, Osterman MJ. Births: preliminary data for 2015. Natl Vital Stat Rep. 2016;65(3):1–15.
2. Albright CL, Maddock JE, Nigg CR. Physical activity before pregnancy and following childbirth in a multiethnic sample of healthy women in Hawaii. Women Health. 2005;42(3):95–110.
3. Durham HA, Morey MC, Lovelady CA, Namenek Brouwer RJ, Krause KM, Østbye T. Postpartum physical activity in overweight and obese women. J Phys Act Health. 2011;8(7):988–93.
4. Evenson KR, Herring AH, Wen F. Self-Reported and objectively measured physical activity among a cohort of postpartum women: the PIN Postpartum Study. J Phys Act Health. 2012;9(1):5–20.
5. Engberg E, Alen M, Kukkonen-Harjula K, Peltonen JE, Tikkanen HO, Pekkarinen H. Life events and change in leisure time physical activity: a systematic review. Sports Med. 2012;42(5):433–47.
6. Fowles ER, Cheng HR, Mills S. Postpartum health promotion interventions: a systematic review. Nurs Res. 2012;61(4):269–82.
7. Retnakaran R, Qi Y, Sermer M, Connelly PW, Zinman B, Hanley AJ. Gestational diabetes and postpartum physical activity: evidence of lifestyle change 1 year after delivery. Obesity (Silver Spring). 2010;18(7):1323–9.
8. Aune D, Norat T, Leitzmann M, Tonstad S, Vatten LJ. Physical activity and the risk of type 2 diabetes: a systematic review and dose–response meta-analysis. Eur J Epidemiol. 2015;30(7):529–42.
9. Fjeldsoe BS, Miller YD, Marshall AL. MobileMums: a randomized controlled trial of an SMS-based physical activity intervention. Ann Behav Med. 2010;39(2):101–11.
10. McIntyre HD, Peacock A, Miller YD, Koh D, Marshall AL. Pilot study of an individualised early postpartum intervention to increase physical activity in women with previous gestational diabetes. Int J Endocrinol. 2012;2012(Apr 4):1–5.
11. Phelan S, Hagobian T, Brannen A, et al. Effect of an Internet-based program on weight loss for low-income postpartum women: a randomized clinical trial. JAMA. 2017;317(23):2381–91.
12. Fjeldsoe BS, Miller YD, Marshall AL. Social cognitive mediators of the effect of the MobileMums intervention on physical activity. Health Psychol. 2013;32(7):729–38.
13. Fjeldsoe BS, Miller YD, Graves N, Barnett AG, Marshall AL. Randomized controlled trial of an improved version of MobileMums, an intervention for increasing physical activity in women with young children. Ann Behav Med. 2015;49(4):487–99.
14. Albright CL, Steffen AD, Novotny R, et al. Baseline results from Hawaii's Na Mikimiki Project: a physical activity intervention tailored to multiethnic postpartum women. Women Health. 2012;52(3):265–91.
15. Albright CL, Steffen AD, Wilkens LR, et al. Effectiveness of a 12-month randomized clinical trial to increase physical activity in multiethnic postpartum women: results from Hawaii's Na Mikimiki Project. Prev Med. 2014;69:214–23.
16. Jelsma JGM, van Poppel MNM, Smith BJ, et al. Changing psychosocial determinants of physical activity and diet in women with a history of gestational diabetes mellitus. Diabetes Metab Res Rev. 2018;34(1):1–9.
17. Lewis BA, Schuver K, Dunsiger S, et al. Rationale, design, and baseline data for the Healthy Mom II Trial: a randomized trial examining the efficacy of exercise and wellness interventions for the prevention of postpartum depression. Contemp Clin Trials. 2018;70:15–23.
18. Chasan-Taber L, Marcus BH, Rosal MC, et al. Estudio Parto: postpartum diabetes prevention program for hispanic women with abnormal glucose tolerance in pregnancy: a randomised controlled trial—study protocol. BMC Pregnancy Childbirth. 2014;14:100.
19. Chang MW, Nitzke S, Brown R. Design and outcomes of a Mothers In Motion behavioral intervention pilot study. J Nutr Educ Behav. 2010;42(Suppl 3):S11–21.
20. Hawaii Health Data Warehouse, State of Hawaii, Department of Health, Behavioral Risk Factors Surveillance System (BRFSS). Physical Activity—Meet Recommendations, for the State of Hawaii, for State and Selected Ethnicities, for the Year 2011; 2013. Available from:
21. Bandura A. Health promotion by social cognitive means. Health Educ Behav. 2004;31(2):143–64.
22. Prochaska JO, Marcus BH. The transtheoretical model: applications to exercise. In: Dishman RK, editor. Exercise Adherence II. Champaign (IL): Human Kinetics Books; 1994, pp. 161–80.
23. Cramp AG, Brawley LR. Sustaining self-regulatory efficacy and psychological outcome expectations for postnatal exercise: effects of a group-mediated cognitive behavioural intervention. Br J Health Psychol. 2009;14(Pt 3):595–611.
24. Mailey EL, McAuley E. Impact of a brief intervention on physical activity and social cognitive determinants among working mothers: a randomized trial. J Behav Med. 2014;37(2):343–55.
25. Keller C, Ainsworth B, Records K, et al. A comparison of a social support physical activity intervention in weight management among post-partum Latinas. BMC Public Health. 2014;14(971):1–8.
26. Miller YD, Trost SG, Brown WJ. Mediators of physical activity behavior change among women with young children. Am J Prev Med. 2002;23(Suppl 2):98–103.
27. Fahrenwald NL, Atwood JR, Johnson DR. Mediator analysis of Moms on the move. West J Nurs Res. 2005;27(3):271–91.
28. Albright CL, Maddock JE, Nigg CR. Increasing physical activity in postpartum multiethnic women in Hawaii: results from a pilot study. BMC Womens Health. 2009;9(4):1–7.
29. Østbye T, Krause KM, Lovelady CA, et al. Active mothers postpartum: a randomized controlled weight-loss intervention trial. Am J Prev Med. 2009;37(3):173–80.
30. Dalrymple KV, Flynn AC, Relph SA, O'Keeffe M, Poston L. Lifestyle interventions in overweight and obese pregnant or postpartum women for postpartum weight management: a systematic review of the literature. Nutrients. 2018;10(11):1704–30.
31. Choi J, Fukuoka Y, Lee JH. The effects of physical activity and physical activity plus diet interventions on body weight in overweight or obese women who are pregnant or in postpartum: a systematic review and meta-analysis of randomized controlled trials. Prev Med. 2013;56(6):351–64.
32. Pate RR, Pratt M, Blair SN, et al. Physical activity and public health. a recommendation from the Centers for Disease Control and Prevention and the American College of Sports Medicine. JAMA. 1995;273:402–7.
33. Haskell WL, Lee IM, Pate RR, et al. Physical activity and public health: updated recommendation for adults from the American College of Sports Medicine and the American Heart Association. Med Sci Sports Exerc. 2007;39(8):1423–34.
34. Meckes N, Vezina JW, Herrmann SD, Sawyer BJ, Angadi S, Ainsworth BE. Oxygen cost of performing selected adult and child care activities. Int J Exerc Sci. 2013;6(1):11–9.
35. Gregory DA, Pfeiffer KA, Vickers KE, et al. Physiologic responses to running with a jogging stroller. Int J Sports Med. 2012;33(9):711–5.
36. Brown WJ, Burton NW, Marshall AL, Miller YD. Reliability and validity of a modified self-administered version of the Active Australia physical activity survey in a sample of mid-age women. Aust N Z J Public Health. 2008;32(6):535–41.
37. Brown WJ, Bauman AE. Comparison of estimates of population levels of physical activity using two measures. Aust N Z J Public Health. 2000;24(5):520–5.
38. Brown WJ, Trost SG, Bauman A, Mummery K, Owen N. Test–retest reliability of four physical activity measures used in population surveys. J Sci Med Sport. 2004;7(2):205–15.
39. Reeves MM, Marshall AL, Owen N, Winkler EA, Eakin EG. Measuring physical activity change in broad-reach intervention trials. J Phys Act Health. 2010;7(2):194–202.
40. Fjeldsoe BS, Marshall AL, Miller YD. Measurement properties of the Australian Women's Activity Survey. Med Sci Sports Exerc. 2009;41(5):1020–33.
41. Zafiropoulos B, Alison JA, Heard R. Physical activity levels of allied health professionals working in a large Australian metropolitan health district—an observational study. J Multidiscip Healthc. 2019;12:51–62.
42. MacKinnon DP. Introduction to Statistical Mediation Analysis. New York: Lawrence Erlbaum Associates; 2008.
43. Cerin E, MacKinnon DP. A commentary on current practice in mediating variable analyses in behavioural nutrition and physical activity. Public Health Nutr. 2009;12(8):1182–8.
44. Muthén LK, Muthén BO. Mplus, Statistical Analysis with Latent Variables: User’s Guide. 6th ed. Los Angeles (CA): Muthén & Muthén; 1998–2010.
45. Enders C. Applied Missing Data Analysis. New York (NY): The Guilford Press; 2010.
46. Graham JW. Missing data analysis: making it work in the real world. Annu Rev Psychol. 2009;60:549–76.
47. Byrne BM. Structural equation modeling with Mplus. New York (NY): Routledge; 2012.
48. Lynch KG, Cary M, Gallop R, Ten Have TR. Causal mediation analyses for randomized trials. Health Serv Outcomes Res Methodol. 2008;8(2):57–76.
49. Keller C, Records K, Ainsworth B, et al. Madres para la Salud: design of a theory-based intervention for postpartum Latinas. Contemp Clin Trials. 2011;32(3):418–27.

Supplemental Digital Content

Copyright © 2019 The Author(s). Published by Wolters Kluwer Health, Inc. on behalf of the American College of Sports Medicine.