Determinants of moderate-intensity physical activity during pregnancy based on the COM-B model

Abstract

This study aimed to explore the determinants of moderate-intensity physical activity (PA) among pregnant women using the Capability, Opportunity, Motivation, and Behavior (COM-B) model. A cross-sectional study was conducted with 316 participants recruited from three tertiary hospitals in China. Data were collected through self-reported questionnaires measuring demographic characteristics, PA knowledge, PA social support, PA environment, PA self-efficacy, and moderate-intensity PA. Structural equation modeling was employed for data analysis. Mean moderate-intensity PA was 7.11 ± 6.88 MET-hours/week. Analysis using standardized path coefficients (β) revealed that: PA knowledge had a significant direct effect on moderate-intensity PA (β = 0.265, p < 0.001) and a significant indirect effect mediated by PA self-efficacy (β = 0.035, p = 0.003). PA social support exhibited both a significant direct effect (β = 0.238, p < 0.001) and a significant indirect effect via PA self-efficacy (β = 0.127, p = 0.004) on moderate-intensity PA. PA environment affected moderate-intensity PA directly (β = 0.218, p = 0.003) and also exerted a significant indirect effect through the serial mediators of PA social support and PA self-efficacy (β = 0.205, p < 0.001). Additionally, PA self-efficacy had a significant direct effect on moderate-intensity PA (β = 0.220, p = 0.005). The study revealed a low level of moderate-intensity PA among Chinese pregnant women. In conclusion, PA knowledge, PA social support, PA environment, and PA self-efficacy were significant direct and/or indirect predictors of moderate-intensity PA. These factors represent promising targets for designing effective interventions to promote PA during pregnancy.

Introduction

Physical activity (PA) refers to any bodily movement involving muscle engagement and energy expenditure across various domains, including household/caregiving, occupational, sports/exercise, and transportation activities¹. The World Health Organization (WHO) recommends that all pregnant women without contraindications engage in at least 150 min of moderate-intensity PA per week through pregnancy¹. Regular moderate-intensity PA is associated with substantial health benefits. For the mother, the benefits include low risks of developing excessive gestational weight gain², decreased risk of gestational diabetes mellitus³, less likelihood of gestational hypertension⁴, reduced symptoms of depression and anxiety⁵, and improved sleep quality⁶. For the fetus, regular PA can help newborns have better body composition after birth and improve neurological development^4,7.

Although regular moderate-intensity PA can positively affect maternal health and that of her fetus, most pregnant women do not meet the WHO recommendation. In high-income countries like the United States, studies indicate that 12.70% to 45.00% of pregnant women fail to meet PA guidelines⁸. By contrast, 79.30% of pregnant women in low-income Ethiopia are classified as sedentary⁹. Data from middle-income China reveal a PA compliance rate ranging from 11.10% to 55.30%^10–12. The wide variation in PA engagement worldwide can be attributed to the differences in geographic region, local culture, and assessment methods¹³. Nonetheless, it is evident that most pregnant women remain physically inactive¹⁴. Therefore, strategies to improve participation in moderate-intensity physical activity during pregnancy are urgently needed.

Identifying the determinants of moderate-intensity PA in pregnant women, particularly focusing on modified and intervenable factors, is critical to develop evidence-based interventions that increase their moderate-intensity PA participation. Some factors such as PA knowledge^15,16, PA self-efficacy^17–19, PA social support^20,21, and PA environment^22,23, have been recognized as significant and actionable predictors of pregnant women’s engagement in PA. Although some evidence already links these variables to prenatal PA²⁴, existing studies have often examined these factors in a fragmented and isolated manner. The relationship between the variables in previous studies was still inconsistent and most of them were not grounded in a theoretical framework. Results lacking a theoretical basis are not only challenging to interpret but also hinder the design of effective interventions²⁵. The interaction mechanisms between the various determinants of moderate-intensity PA remains unclear, necessitating further research in this area.

We selected the Capability, Opportunity, Motivation, and Behavior (COM-B) model as the guiding framework because of its explicit linkage of capability, opportunity, and motivation to behavior and its direct mapping to behavior change techniques²⁶. This model recognizes that behavior (e.g., moderate-intensity PA) is part of an interacting system comprising an individual’s capability (both physical and psychological, e.g., knowledge), opportunity (both social and physical, e.g., social influences and environmental resources), and motivation (both reflective and automatic, e.g., beliefs about capabilities). A key strength of the COM-B model lies in its granular taxonomy, which allows clear alignment between its core components and the variables examined in this study: psychological capability maps to PA knowledge, social and physical opportunity maps to PA social support and PA environment, and motivation maps to PA self-efficacy. This precise mapping facilitates a systematic and theory-driven investigation of the determinants of prenatal PA, enabling the identification of modifiable targets for future interventions. This study focused on these components that serve as direct prerequisites for behavior occurrence to ensure model parsimony and statistical power without contradicting the core principles of the model. Within the COM-B framework, capability and opportunity are posited to influence behavior both directly and indirectly, with motivation serving as a key mediator²⁷. Previous research has validated these pathways in other populations, such as medication adherence among patients with rheumatoid arthritis²⁵ and patient safety behaviors among nursing interns²⁸. While this confirms the model’s generalizability, this study draws on the Theoretical Domains Framework (TDF) to inform the selection and conceptual mapping of study variables within the COM-B structure, which comprises 14 domains and 84 component constructs, into the COM-B structure²⁹. This integration enables a systematic identification of modifiable determinants through cross-theoretical mapping. The complete theoretical correspondences are systematically delineated in Table 1. Figure 1 illustrates the hypothesized model guiding this study.

Table 1.

The correspondence between COM-B, TDF and variables.

COM-B		TDF	Variables
Capability	Psychological Capability	Knowledge	PA knowledge
Motivation	Reflective Motivation	Beliefs about Capabilities	PA self-efficacy
Opportunity	Physical Opportunity	Environmental Resources	PA environment
	Social Opportunity	Social Influences	PA social support

Note: COM-B = Capability, Opportunity, Motivation, and Behavior; TDF = Theoretical Domains Framework; PA = Physical Activity.

Fig. 1.

The hypothesized model guiding this study.

Therefore, this study aimed to explore the influencing mechanisms of PA knowledge (Capability), PA social support and environment (Opportunity), and PA self-efficacy (Motivation) on moderate-intensity PA in pregnant women based on the COM-B model and to consider strategies for increasing women’s engagement in moderate-intensity PA. Specifically, we propose the following hypotheses. Hypothesis 1: PA knowledge, PA social support, PA environment, and PA self-efficacy are significantly and positively associated with moderate-intensity PA; Hypothesis 2: The influence of PA knowledge on moderate-intensity PA is mediated by PA self-efficacy; Hypothesis 3: PA self-efficacy serves as the mediator through which PA social support enhances engagement in moderate-intensity PA; Hypothesis 4: PA environment can indirectly affect moderate-intensity PA through the mediator of PA social support and PA self-efficacy.

Methods

Research design and participants

This cross-sectional study was conducted at three teaching hospitals in Yangzhou, China, between December 2023 and January 2024. Participants were included if they met the following eligibility criteria: (1) aged ≥ 20 years; (2) at any trimester with a singleton pregnancy; (3) able to read and write in Chinese and participate voluntarily. Participants were excluded if they had any contraindications to PA, such as unexplained persistent vaginal bleeding, pre-eclampsia, intrauterine growth restriction, incompetent cervix, or musculoskeletal disorders.

All eligible women receiving routine prenatal care during the study period were recruited from the hospital obstetrics departments using a convenient sampling method. Informed consent was obtained from all participants before the investigation. For individuals who declined to participate, trained investigators briefly asked about their reasons for refusal and recorded these immediately after each conversation; this procedure was exploratory and not a formal assessment. Eligible participants were instructed to independently complete the questionnaire in a quiet room. To ensure data quality and completeness, investigators reviewed each questionnaire immediately upon completion. Any missing or ambiguous responses were clarified in person with the participant at that time. This study followed the guidelines of the Declaration of Helsinki and was approved by the School of Nursing Ethics Committee, Yangzhou University (Approval No. YZUHL20230016).

Sample size

The sample size was calculated based on the requirement of the structural equation modeling (SEM) analysis³⁰. A minimum of 20 participants per parameter is necessary to improve the test efficacy. The number of parameters in our original hypothesized model was 8, and the minimum sample size was 160. Moreover, For SEM to obtain stable results, the sample size should be over 200³¹. Combining the above two, the minimum sample size was 200. Considering a 15% invalid questionnaire rate, it was determined that a sample size of at least 230 would be needed.

Measures.

Demographic characteristics

The questionnaire collected individual characteristics, including demographic information (such as age, education level, current employment status, place of residence, and income) and pregnancy-related information (such as trimester, parity, history of abortion, pre-pregnancy body mass index (BMI), preferred delivery mode, pregnancy complications, whether specific guidance was received during pregnancy and pre-pregnancy habitual exercise. This study classified participants according to China criteria for pre-pregnancy BMI (underweight, < 18.5 kg/m²; normal weight, 18.5–23.9 kg/m²; overweight, 24–27.9 kg/m²; and obese, ≥ 28 kg/m²)³². Regular pre-pregnancy exercise was defined as participating in PA or exercise ≥ 5 times/week and lasting ≥ 30 min/time before pregnancy¹.

Moderate-intensity PA

Moderate-intensity PA during pregnancy was measured using the Pregnant Physical Activity Questionnaire (PPAQ) developed by Chasan-Taber et al.³³. This questionnaire includes 32 activities and participants select the best category that estimates the amount of time spent in each activity (e.g., “none”, “less than 1/2 per day”, “1/2 to almost 1 hour per day”, “1 to almost 2 hours per day”, “2 to almost 3 hours per day” or “3 or more hours per day”). The number of hours spent in each self-reported activity was multiplied by its metabolic equivalent of task (MET) level to measure the average weekly energy expenditure (MET-hours/week) attributable to each activity. These activities were rated by intensity: sedentary (1 to < 1.5 METs), light (1.5 to < 3 METs), moderate (3–6 METs) and vigorous (> 6 METs). These activities were also classified by type: household/caregiving, occupational, sports/exercise, and transportation activities. The Cronbach’s alpha coefficient of the PPAQ and each category ranged from 0.780 to 0.930. The PPAQ has been widely used and validated in many countries³⁴, such as Turkey, Denmark, and Vietnam^35–37. Zhang et al. translated this questionnaire to Chinese, and the Chinese version of the scale showed good reliability and validity³⁸. In this study, the Chinese version of the PPAQ, including 31 items, was used to measure pregnancy PA. Consistent with WHO guidelines for PA during pregnancy¹, meeting the guidelines was defined as more than 7.5 MET*hours per week, equal to at least 150 min of moderate-intensity activities in a week.

PA knowledge

PA knowledge was evaluated using the pregnancy physical activity knowledge questionnaire designed based on the WHO guidelines and Lee et al.’s measuring tool^1,39. The 12-item questionnaire included principles of frequency, intensity, time, and type (FITT), benefits, safety precautions, and warning signs. The scoring protocol followed a binary system: each of the 12 items was assigned 1 point for a correct response and 0 points for an incorrect/absent response. The total score ranged from 0 to 12, with higher scores indicating a higher level of PA knowledge. A committee of experts composed of two clinical obstetricians, an obstetric nurse, two associate nursing professors, and a sports professor was invited to evaluate this PA knowledge questionnaire. This questionnaire was modified according to the suggestions of experts. The content validity index (CVI) ranged from 95% to 100%. In this study, Cronbach’s alpha on the questionnaire was 0.772.

PA self-efficacy

PA self-efficacy during pregnancy was measured by the Pregnancy Exercise Self-Efficacy Scale (P-ESES) revised by Bland et al.⁴⁰. The P-ESES includes 10 items, and each item is based on a Likert 5-point scale ranging from ‘strongly agree’ to ‘strongly disagree’. The Cronbach’s alpha of the P-ESES was 0.838, showing good reliability. The Chinese version of P-ESES was modified by Yang⁴¹, with 3 domains (overcoming exercise barriers, emotional barriers, and support barriers). The total score of the P-ESES ranges from 10 to 50 and is categorized as follows: low self-efficacy (10–20), mediate self-efficacy (21–40), and high self-efficacy (41–50). The P-ESES has good reliability and validity, with Cronbach’s alpha coefficient of 0.804. In this study, the Cronbach’s alpha of the scale was 0.930.

PA social support

Social support for PA was assessed by the Physical Activity Social Support Scale (PASSS) revised by Xiang et al.⁴². The PASSS comprises 24 items with four domains (emotional, informational, instrumental, and companion support). Each item is scored on a 5-point scale ranging from 1 (strongly disagree) to 5 (strongly agree). The total score of the PASSS is 24 to 120, and higher scores indicate more social support for PA. The PASSS has good reliability and validity, with Cronbach’s alpha coefficient of 0.950. In this study, the Cronbach’s alpha coefficient of the PASSS was 0.956.

PA environment

PA environment was evaluated by the Chinese Neighborhood Environment Walkability Scale (CNEWS) revised by Zhou et al.⁴³. The CNEWS includes 17 items, with four domains (accessibility to services, neighborhood surroundings/aesthetics, traffic safety, and crime safety). Each item is scored on a 5-point scale ranging from 1= (strongly disagree) to 5= (strongly agree). The total score ranges from 17 to 85, with higher scores indicating a more PA-friendly environment. The CNEWS is reliable and valid, with a reported Cronbach’s alpha coefficient of 0.807. In this study, the Cronbach’s alpha coefficient of the PASSS was 0.899.

Statistical analysis

SPSS version 27.0 was used to analyze the data. As the data approximated a normal distribution, mean ± standard deviation (SD) was used to describe continuous variables. Frequency (percentage) was used to describe categorical variables. The normality of variables was confirmed through skewness and kurtosis. Absolute values of skewness < 3 and kurtosis < 10 are considered indicative of approximate normality⁴⁴. The Maximum Likelihood (ML) estimation method was employed in the SEM analysis, which is robust to minor violations of the normality assumption. Correlations between variables were tested using the Pearson correlation coefficient. Prior to model analysis, multicollinearity among independent variables was assessed by calculating the Variance Inflation Factor (VIF). The results revealed VIF values below 3 for all variables, indicating no significant multicollinearity issues. AMOS version 26.0 was used to perform SEM to verify the hypothetical model of PA knowledge, PA social support, PA environment, PA self-efficacy, and moderate-intensity PA proposed based on the COM-B model. To estimate the goodness of model fit, the following indices were considered: the chi-square to degrees of freedom ratio (χ²/df), goodness of fit index (GFI), comparative fit index (CFI), incremental fit index (IFI), tucker and lewis index (TLI), and root mean square error of approximation (RMSEA). A model with χ²/df < 3, RMSEA < 0.080, and GFI, CFI, IFI, and TLI > 0.900 indicated an acceptable fit. The bootstrapping method was used to test the statistical significance of the model’s direct, indirect, and total effects. All path coefficients reported in the results are standardized estimates. A p-value of < 0.05 was considered statistically significant (two-tailed). For clarity, p-values < 0.001 are reported as < 0.001.

Results

Demographic characteristics

The participation rate was high, with a refusal rate of only 3%. The primary self-reported reason for refusal, obtained through informal verbal inquiries with participants who declined to participate, was lack of time. A total of 324 participants were recruited in this study. Eight questionnaires were excluded due to straight-lining, a pattern of identical responses indicating non-attentive answering. Finally, a sample of 316 were analyzed with a response rate of 97.53%. Most participants (90.8%) were in their third trimester. This distribution aligns with the attendance patterns at the participating prenatal clinics, as women in later stages of pregnancy typically have more frequent scheduled visits. Most of the women (88.60%) were < 35 years. Approximately 80% of participants were employed and living in urban areas. Almost 68% of the participants were primiparous, and less than half of pregnant women (47.20%) wanted to have a natural labor. The demographic characteristics of the participants are shown in Table 2.

Table 2.

Demographic characteristics of pregnant women (N = 316).

Variables		n(%)
Age (years)	< 35	280(88.60)
	≥ 35	36(11.40)
Education level	High school or below	9(2.80)
	Diploma or bachelor’s degree	274(86.70)
	Master’s degree or higher	33(10.40)
Current employment status	Employed	252(79.70)
	Unemployed	64(20.30)
Place of residence	Urban	255(80.70)
	Rural	61(19.30)
Personal monthly income (¥)	≤ 3000	14(4.40)
	3001–5000	66(20.90)
	5001–8000	142(44.90)
	> 8000	94(29.70)
Pre-pregnancy BMI (kg/m2)	Underweight (<18.5)	27(8.50)
	Normal (18.5–23.9)	208(65.80)
	Overweight (24–27.9.9)	63(19.90)
	Obese (≥ 28)	18(5.70)
Trimester (weeks)	First trimester (≤ 13)	8(2.50)
	Second trimester (14–27)	21(6.60)
	Third trimester (≥ 28)	287(90.80)
Parity	Primiparous	216(68.40)
	Multiparous	100(31.60)
Preferred delivery mode	Natural labor	149(47.20)
	Cesarean section	66(20.90)
	Both	101(32.00)
History of abortion	Yes	88(27.80)
	No	228(72.20)
Received specific guidance during pregnancy	Yes	16(5.10)
	No	299(94.60)
Pre-pregnancy habitual exercise	Yes	101(32.00)
	No	215(68.00)

Descriptive statistics and correlation matrix of the study variables

Descriptive statistics and correlation matrix of the study variables are presented in Table 3. The average energy expenditure of moderate-intensity PA was (7.11 ± 6.88) MET-hours/week. According to the criteria recommended by WHO¹, 202 participants (63.90%) did not meet the PA guideline. The average scores of PA knowledge, PA self-efficacy, PA social support, and PA environment were (6.14 ± 1.75), (34.43 ± 7.09), (81.95 ± 15.92), and (58.52 ± 8.46), respectively. Moderate-intensity PA was significantly positively correlated with PA knowledge (r = 0.471, p < 0.01), PA social support (r = 0.421, p < 0.01), PA environment (r = 0.426, p < 0.01), and PA self-efficacy (r = 0.397, p < 0.01). In addition, analyses for normality indicated that all variables were not significantly skewed (p > 0.05).

Table 3.

Descriptive statistics and correlation matrix of the study variables (N = 316).

	Mean ± SD	Skewness (SE)	Kurtosis (SE)	1	2	3	4	5
1. Moderate-intensity PA	7.11 ± 6.88	1.57(0.14)	4.41(0.27)	1.000	-	-	-	-
2. PA knowledge	6.14 ± 1.75	−0.223(0.14)	0.54(0.27)	0.471**	1.000	-	-	-
3. PA self-efficacy	34.43 ± 7.09	0.19(0.14)	−0.52(0.27)	0.397**	0.391**	1.000	-	-
4. PA social support	81.95 ± 15.92	0.62(0.14)	−0.65(0.27)	0.421**	0.413**	0.516**	1.000	-
5. PA environment	58.52 ± 8.46	0.63(0.14)	−0.25(0.27)	0.426**	0.425**	0.473**	0.424**	1.000

Note: ^**p < 0.01 (two-tailed).

Testing of the hypothetical model

According to the hypothetical model, SEM was used to identify the path relationship among PA knowledge, PA social support, PA environment, PA self-efficacy, and moderate-intensity PA (Fig. 1). The model fit indices indicated that the final model (Fig. 2) had a reasonable fit to the data (χ²/df = 2.175 < 3, RMSEA = 0.061 < 0.080, GFI = 0.961 > 0.900, CFI = 0.967 > 0.900, IFI = 0.967 > 0.900, and TLI = 0.951 > 0.900).

Fig. 2.

Final model of moderate-intensity PA in pregnant women. Note: Path coefficients were shown in the standardized regression coefficient. PA = Physical Activity; R² = 0.355; ^***p < 0.001.

In this study, all interaction terms between gestational trimester and independent variables on PA levels were statistically non-significant (p > 0.05). Table 4 revealed the standardized direct, indirect, and total effect of predictors on moderate-intensity PA in the final model. The model explained 35.5% of the variance in moderate-intensity PA. Path analysis indicated the following significant effects: there was a direct relationship between PA knowledge and moderate-intensity PA (β = 0.265, p < 0.001). PA knowledge also had an indirect effect on moderate-intensity PA through the mediator of PA self-efficacy (β = 0.035, p = 0.003). PA social support demonstrated both a direct effect on moderate-intensity PA directly (β = 0.238, p < 0.001)) and an indirect effect mediated by PA self-efficacy (β = 0.127, p = 0.004). PA environment had a direct effect on moderate-intensity PA (β = 0.218, p = 0.003) and an indirect effect mediated by PA social support and PA self-efficacy (β = 0.205, p < 0.001). There was a direct relationship between PA self-efficacy and moderate-intensity PA (β = 0.220, p = 0.005).

Table 4.

Bootstrap test for the standardized direct, indirect, and total effect of the model (N=316).

Predictors	Standardized direct effect				Standardized indirect effect				Standardized total effect
	β	SE	95% CI	p	β	SE	95% CI	p	β	SE	95% CI	p
PA knowledge	0.265	0.137	[0.129, 0.390]	< 0.001	0.035	0.017	[0.010, 0.078]	0.003	0.299	0.067	[0.165, 0.430]	< 0.001
PA self-efficacy	0.220	0.059	[0.068, 0.380]	0.005	-	-	-	-	0.220	0.059	[0.068, 0.380]	0.005
PA social support	0.238	0.025	[0.110, 0.360]	< 0.001	0.127	0.047	[0.041, 0.228]	0.004	0.365	0.051	[0.267, 0.466]	0.001
PA environment	0.218	0.036	[0.079, 0.349]	0.003	0.205	0.036	[0.142, 0.284]	< 0.001	0.424	0.057	[0.296, 0.521]	< 0.001

Note: PA = Physical Activity; β = Standardized Coefficient; SE = Standard Error; CI = Confidence Intervals; R² = 0.355.

Discussion

Evidence on identifying and optimizing the factors associated with moderate-intensity PA is crucial to develop effective interventions aimed at increasing moderate-intensity PA. However, although previous studies have reported associations between individual variables and PA, existing evidence has largely focused on isolated factors without exploring their interaction mechanisms. To our knowledge, this study was the first attempt to examine the determinants of moderate-intensity PA in pregnant women using the COM-B model. The findings support the utility of the COM-B model for understanding PA behavior in this population.

≥ 7.5 MET*hours/week of moderate-intensity PA was the threshold for meeting WHO standards¹. In this study, the mean energy consumption of moderate-intensity PA was (7.11 ± 6.88) MET-hours/week. Pregnant women did not fulfill the PA criterion in a lower percentage of participants (63.9) than in prior studies⁴⁵. One explanation could be that most of pregnant women in this study worked in urban regions, making it difficult for them to be physically active during the pregnancy⁴⁶. Time restrictions often served as an excuse for them not to engage in PA, despite their motivation to do so⁴⁶. The study’s findings suggest that there is still room for improvement in moderate-intensity PA that pregnant women engage with. Therefore, effective interventions are urgently needed to increase moderate-intensity PA in this population. No significant difference was observed in moderate-intensity PA among trimesters, which contrasts with prior studies reporting a decline in PA as pregnancy progresses. This may be due to the heavily skewed trimester distribution (90.8% third trimester), which could have limited power to detect differences⁴⁷. Thus, the impact of trimesters as a potential moderator remains incompletely understood, necessitating further research.

Capability: from knowledge to cognitive transformation

As the cornerstone of psychological capability, PA knowledge operated through a dual-pathway mechanism in this study. Consistent with the hypothesized model, this finding demonstrated a direct and positive relationship between PA knowledge and moderate-intensity PA. This is inconsistent with previous findings on this relationship in other studies^24,45. The conflicting findings on the direct relationship between this study and prior research may be due to the differences in the measure tools and sample characteristics. The theory of knowledge, attitudes, and practices supported the finding that better knowledge can imperceptibly affect beliefs about capabilities to engage in moderate-intensity PA, thereby increasing moderate-intensity PA⁴⁸. On the contrary, a paucity of knowledge contributed to pregnant women experiencing uncertainty about the capability to do PA safely during pregnancy and this uncertainty may make them feel less confident to do PA, thus leading to negative PA participation^49,50. Therefore, it is necessary to improve pregnant women’s level of PA knowledge.

Opportunity: synergy between social and environmental enablers

The synergistic interplay between social support and environmental resources validated the dual-channel mechanism of opportunity within the COM-B framework. PA social support was significantly positively associated with moderate-intensity PA in the model, which is in line with a qualitative study conducted in Singapore showing that women with greater social support for PA are more likely to engage in PA during pregnancy²¹. This finding aligns with this study’s hypothesis and previous studies in other populations, implying that high PA social support may change the incentive to engage in PA and thus improve moderate-intensity PA^51,52. The explanation is that providing emotional, informative, affectionate, and tactile support for pregnant women via the available social network can boost their confidence in overcoming the obstacles and then promote their involvement in moderate-intensity PA^20,42. In this study, the importance of social support indicates that programs designed to increase pregnant women’s engagement in PA should interact with their social support network, including emotional, informational, instrumental, and companion support provided by family members, friends, and healthcare professionals, to promote healthy behaviors. A study of 687 participants indicated that families and friends were the primary sources of providing actionable PA information⁵³. Moreover, instrumental and companion support from families enabled pregnant women to do PA, such as assisting with household chores and walking companionship⁵⁴. Therefore, it is essential to provide more social support for pregnant women.

The result that PA environment is directly and positively associated with moderate-intensity PA in women during pregnancy, was in line with a prospective cohort study conducted in the USA⁵⁵. Leggett et al. reported that neighborhood public parks with good quality features and amenities may help facilitate pregnant women’s PA⁵⁶. They provide places for people to walk or jog, and many have facilities for sports⁵⁶. Greater access to high-quality PA areas can subjectively offer emotional support for pregnant women, increase the motivation of PA, and in turn substantially increase rates of moderate-intensity PA among pregnant women^57,58. Thus, the findings in this study highlight the need for improving PA environment during pregnancy. Some environmental interventions can be utilized to promote moderate-intensity PA. PA environment has several aspects including accessibility to services, neighborhood surroundings/aesthetics, traffic safety, and crime safety. With their growing family and associated responsibilities, pregnant women particularly need places to be physically active that are close to home and accessible⁵⁹. Furthermore, a prior study found that pregnant women were more likely to meet PA guidelines if they lived in walkable and recreational districts⁶⁰. Therefore, good environment plays an indispensable role in promoting pregnant women’s PA.

Motivation: Self-Efficacy as a multilevel catalyst

Self-efficacy, as the nucleus of reflective motivation, demonstrated multilevel moderating effects. This study found a direct relationship between PA self-efficacy and moderate-intensity PA, which is consistent with this study’s hypothesis and previous studies reported that pregnant women with high levels of PA self-efficacy were more likely to meet the PA guideline^45,61. Moreover, PA self-efficacy for overcoming barriers mediates between capability variable (PA knowledge), opportunity variables (PA social support and PA environment), and moderate-intensity PA. This indicates that pregnant women’s beliefs about their capabilities to take part in moderate-intensity PA is an essential underlying mechanism through which knowledge, social influences, and environmental resources are associated with moderate-intensity PA. Social cognitive theory underpins that self-efficacy not only deals with the prominent roles played by cognitive, vicarious, self-reflective, and self-regulatory processes but also plays an important in determining how much effort will be made and sustained in the face of obstacles⁶². Thus, when facing barriers to PA, women with high levels of self-efficacy may regulate their emotions and make more efforts to overcome these barriers. Overall, PA self-efficacy is key to enhancing engagement in moderate-intensity PA⁶³. Given the mediating role of self-efficacy, targeted interventions should address antecedent factors such as PA knowledge, social support, and the environment to effectively promote PA. To help pregnant women maintain regular PA habits, healthcare providers should collaborate with pregnant women to create personalized PA plans, deliver targeted education (e.g., safety guidelines, FITT principles), and track progress via activity diaries. Meanwhile, policy-level efforts to improve access to safe and walkable PA infrastructure (walking trails, prenatal gyms) may support PA engagement among pregnant women.

Limitations and future study

There are several limitations in our study. First, participants in this study were recruited from three teaching hospitals in China, limiting our findings’ generalizability. Future research with multi-center and large samples is warranted to improve the generalizability of this study’s results. Second, the cross-sectional design and uneven distribution of participants across trimesters prevent us from definitively characterizing longitudinal changes or making causal inference. Prospective longitudinal research with balanced sample sizes across all trimesters is needed to enable robust comparisons and a definitive characterization of longitudinal changes. Third, PA was assessed using the self-reported PPAQ. While validated, this method is susceptible to recall bias and social desirability bias. Furthermore, participants might inaccurately classify the intensity of their activities, potentially leading to misclassification. The reliance on self-reported measures for both predictors and the outcome, without objective corroboration (e.g., using accelerometers), is a limitation. Future studies can combine objective measurements with subjective tools to measure study variables. Fourth, although we controlled for several key demographic and pregnancy-related variables, the possibility of residual confounding cannot be ruled out. Factors such as specific comorbidities, and unmeasured social or environmental influences might still account for some of the observed associations. Fifth, the complexity of behavior change means that this study could not address all COM-B-relevant factors affecting moderate-intensity PA. Notably, components of automatic motivation—such as emotional well-being and underlying attitudes toward PA—were not measured and represent a promising avenue for future research to better understand the determinants of prenatal PA. Future researchers should explore the effect of other potential factors. Sixth, PA comprises household/caregiving, occupational, sports/exercise, and transportation activities. Distinct types of activities may yield differing outcomes. Hence, future research could investigate the influence of variables on moderate-intensity PA based on activity types. Finally, future research could benefit from applying brief measures specifically designed for the COM-B model, such as the 6-item measure proposed by Keyworth et al.⁶⁴, to more efficiently and comprehensively investigate the COM-B framework in relation to prenatal PA.

Conclusion

This study demonstrates that low engagement in moderate-intensity PA among Chinese pregnant women is modifiable through targeted interventions addressing COM-B domains: Capability (PA knowledge), Opportunity (PA social support and environment), Motivation (PA self-efficacy). As the first quantitative study to apply the COM-B model to prenatal PA in China, our findings reveal culturally specific barriers (e.g., reliance on family approval for PA) and health system gaps (e.g., limited PA guidance in antenatal care), advancing the model’s contextual adaptation. Healthcare providers should prioritize integrating these evidence-based targets with behavior change techniques (e.g., motivational interviewing for self-efficacy, family-inclusive environmental redesign) to address the issue of insufficient PA among pregnant women.

Author contributions

L.F.Y.: Conceptualization, Funding acquisition, Formal analysis, Investigation, Methodology, Writing – original draft. X.C.S.: Conceptualization, Funding acquisition, Methodology, Writing– review & editing. M.G.: Methodology, Investigation, Writing - original draft. L.P.S.: Investigation, Writing – original draft. H.W.Y.: Investigation, Writing – original draft. Y.Z.: Investigation, Writing – original draft. L.L.: Investigation, Methodology, Writing – review & editing. Y.X.L.: Conceptualization, Formal analysis, Methodology, Writing – review & editing. F.F.W: Funding acquisition, Methodology, Writing – review & editing. All authors read and approved the final manuscript.

Funding

This study was supported by grants from the Key Project of the Nursing Research Special Program of Northern Jiangsu People’s Hospital (SBHL25004), the ‘Huxin Fund’ project of Jiangsu Key Laboratory of Zoonotic Diseases (HX2414), Management Research Project of Northern Jiangsu People’s Hospital (YYGL202406), and the Key R&D Projects of Yangzhou (YZ2021067). The funding body played no role in the design of the study and collection, analysis, and interpretation of data and in writing the manuscript.

Data availability

The datasets used and analyzed during the current study available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.