Justification of the model

The efficacy of the TPB in predicting individual health-related behaviours has been demonstrated in several systematic reviews. Systematic review focused on the relationship between intention and behaviour have predicted more than 20% variance in prospective measures of the actual behaviour of individuals. This variance in behaviour as explained by intention is similar in magnitude to that found in the extant literature [11]. Since the TPB has been useful in predicting health-related behaviour, it may also be useful in evaluating behaviour among postnatal mothers’ fertility intentions and choices [3,13].

Application of the TPB in any new context requires a tool to measure the variables related to the behaviour of interest and its correlations and, like any other measurement tool, it should demonstrate evidence of psychometric properties, such as validity and reliability. There is adequate evidence that supports the successful use of TPB to predict the PPFP use. Therefore, there are measurement instruments whose reliability and validity is known with high internal consistency and test-retest reliability to predict fertility intentions and PPFP choice [1,11,14,15].

Lastly, it is true that the TPB is focused on the controlled facets of human judgment, intention and decision-making. According to the theory, subjective information in the form of behavioural, normative, and control beliefs provides the foundation for attitudes, intention and ultimately behaviour. This being a cluster randomized control trial, TPB was best placed to estimate fertility intention and eventually PPFP choices [1]. Thus, the study sought to assess the application of adapted Theory of Planned Behaviour (TPB) in predicting intention for early PPFP in postpartum women in Western Kenya. The study comprised three phases: pre-intervention, intervention, and follow-up as shown in Fig 1.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. CONSORT diagram showing participant flow.

https://doi.org/10.1371/journal.pone.0315029.g001

Conceptualization of theoretical framework

In the current study, TPB was conceptualized into a framework that was used to explain the interaction of variables in determining intention and choice for PPFP. The framework offers means to study the interactions different aspects that determine fertility intentions, choice and usage of PPFP methods.

As much as TPB is a widely validated and used framework for understanding and predicting human behaviour, researchers have recognized that there are other factors that may influence behaviour, and the TPB can be modified by adding these constructs to the model [16]. Other exogenous constructs such as knowledge of behaviour, perceived risk of contrary behaviour, and satisfaction with behaviour-enhancing process can be adopted to TPB to optimize it. These constructs have previously been found to be important predictors of behaviour in various studies and can provide additional insight into the factors that influence intention and behaviour [10,17].

Research has shown that knowledge of behaviour can increase the accuracy of predictions made by TPB, as individuals with more knowledge about a behaviour may be more likely to engage in it. Knowledge of behaviour is an important predictor of behaviour because it can influence an individual’s attitudes and beliefs about the behaviour, as well as their ability to perform the behaviour. Research has shown that increasing knowledge about a particular behaviour can lead to more positive attitudes towards the behaviour and greater perceived behavioural control, which in turn, can increase the likelihood of engaging in the behaviour [6,18]. A previous study found that individuals who had more knowledge about the benefits of regular physical activity were more likely to engage in regular physical activity. The study found that increasing knowledge about the benefits of physical activity increased positive attitudes towards physical activity, which led to greater intentions to engage in physical activity [4]. Another study found that individuals who had more knowledge about the benefits of recycling were more likely to recycle. The study found that increasing knowledge about recycling led to more positive attitudes towards recycling, which in turn, led to greater intentions to recycle. Adding a construct of knowledge of behaviour to the TPB can increase the accuracy of predictions made by the theory because knowledge has a role in shaping attitudes and perceived behavioural control [19].

Perceived risk of not engaging in a desired behaviour, also known as "perceived risk of contrary behaviour," can be added as an adopted construct in the TPB model to understand why individuals may engage in a behaviour after considering the potential risk of not engaging in the said behaviour. For example, an individual may exercise regularly because they perceive risks of not exercising, such as potential negative health outcomes or social disapproval. They may believe that if they don’t exercise regularly, they will gain weight, have high blood pressure or become less attractive to others. In this case, the perceived risk of not exercising would be high and would likely encourage the individual to engage in regular physical activity [20,21]. Another example would be an individual who is considering quitting smoking. They may consider quitting because they may perceive the risks of not quitting (contrary behaviour), such as potential weight gain, stress, or social isolation. Research has shown that the perceived risk of not engaging in a behaviour can be a significant barrier to behaviour change. A study found that perceived risk of not engaging in a behaviour was a strong predictor of intentions [5].

Satisfaction with the process towards behaviour refers to an individual’s level of satisfaction with the steps or actions they take to engage in a behaviour. It can be an important predictor of behaviour because it can influence an individual’s motivation to continue engaging in a behaviour over time. An individual may have positive attitudes towards regular contraceptive use, perceive social pressure to be active, and have the ability to be active, but if they are not satisfied with the process of finding a facility with their preferred method, they may not continue to engage in FP. In this case, their satisfaction with the process towards the behaviour of FP would be low and would likely discourage the individual from continuing to engage in the behaviour [17,22]. Including satisfaction with the process towards behaviour in TPB can help to understand why individuals continue to engage in a behaviour or not, even if they have positive attitudes towards it, perceived social pressure to do so and the ability to perform the behaviour.

Including these constructs in the TPB can help to increase the explanatory power of the theory and provide a more comprehensive understanding of how individuals make decisions about behaviour. A previous researcher found that when perceived behavioural control, subjective norms, and attitudes were combined with satisfaction with the process, the explanation of intentions and behaviour improved significantly [7].

Therefore, an individual’s intention to use contraceptives may be influenced by their attitudes towards contraceptive use (e.g. believing it is important for family planning), their perceptions of social pressure to use contraceptives (e.g. believing that their partner and friends use contraceptives), and their beliefs about the ease or difficulty of obtaining and using contraceptives (e.g. access to contraceptives and knowledge about different methods).

Nonetheless, researchers have recognized that there are other factors that may influence contraceptive use, and the TPB can be modified by adding these constructs to the model. Knowledge of behaviour is an important predictor of contraceptive use. Individuals who have more knowledge about different methods of contraceptives and their effectiveness may be more likely to use contraceptives because they understand the outcomes associated with them and may also have more skills to obtain and use them [23].

Perceived risk of contrary behaviour can be added as an additional construct in TPB model. Perceived risk of not using contraceptives refers to an individual’s assessment of the likelihood and potential consequences of not using contraceptives, such as unintended pregnancy or sexually transmitted infections [24].

Satisfaction with the process towards behaviour is also an important predictor of contraceptive use. Individuals who are satisfied with the process towards obtaining and using contraceptives, such as access to information and services, may be more likely to use contraceptives and continue to use them over time. Studies have found that TPB can be used to predict contraceptive use by including satisfaction with the process to the core TPB constructs [15,25]. Fig 2 represents the conceptualized TPB applied to the current study with Table 1 providing the conceptual and operational definition of the constructs in the adapted TPB.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Conceptual framework constructs in the adapted TPB.

https://doi.org/10.1371/journal.pone.0315029.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Definition of concepts in the conceptual framework.

https://doi.org/10.1371/journal.pone.0315029.t001

Study design

This was a non-blinded pragmatic randomized controlled trial with three arms (nurses’ arm community arm, and control arm) and three phases (pre-intervention, intervention, and post-intervention) aimed at assessing the application of adapted TPB in predicting intention for early PPFP in postpartum women in Western Kenya.

Study setting and procedure

The study focused on pregnant mothers in their second or third trimester attending ANC clinics in Kisumu County, Kenya. A total of 6 facilities were included in the study, with 2 facilities randomly assigned to each of the 3 arms (nurses’ arm, community arm, and control arm), including 1 rural and 1 urban facility for each arm.

The sample size was estimated based on a proposed sample size determination formula for difference in proportions with consideration of type I and II errors and power [26–30], to be: . Where q1 = 1-p1, q2 = 1-p2, , p1, p2 = proportion (incidence) of groups #1 (27% which is the KDHS estimated current PPFP use) and #2 (53% which is the KDHS estimated Contraceptive Prevalence Rate (CPR) in the general population), Δ = |p2-p1| = absolute difference (desired clinical difference between intervention and control arms) between two proportions (0.26 i.e. 0.53–0.27), n1 = sample size for group #1, n2 = sample size for group #2, α = probability of type I error (is set at 0.05), β = probability of type II error (is set at 0.1 i.e. 90% power), z = critical Z value for a given α or β(1.96) and K = ratio of sample size for group #2 to group #1(1). Thus, for practical equal sample distribution with assumed 10% loss to follow-up, the actual sample size was 246 with each study arm having 82 participants. As such, each facility per arm based on the rural-urban dichotomy had 41 participants [31].

The Health Centres were eligible to participate in the study if they met certain criteria such as offering full ANC, delivery, and PNC services, providing at least three modern contraceptive methods, having no recent contraceptive stock-outs, performing at least 10 deliveries per month, and being willing to participate. Pregnant women in their second or third trimester, attending ANC, intending to attend PNC at the Health Centre, providing informed consent, and living within 20km were eligible to participate. Excluded from the study were; participation in another study, latex sensitivity, lack of a male partner in the next 12 months, inability to complete the consent form, or the only male partner having a vasectomy [32].

The study used a multistage sampling method that combined purposive sampling for the sub-counties and facilities and simple random sampling for the subjects. Participants were randomly assigned to the study through simple random sampling using excel random numbers applied to the sampling frame provided in each facility as per those who met the eligibility criteria. Randomization was done by the principal investigator.

The study comprised three phases: pre-intervention, intervention, and follow-up. During the pre-intervention phase which lasted approximately 6 months from July 2021, site selection and preparation were carried out. Study preparation involved research staff training, tool preparation, tool piloting, finalization, publication of study protocols, and preregistration of protocols with a recognized clinical trials registry (PACTR). The intervention phase encompassed recruitment, random allocation of participants, and administering a single session of antenatal early postpartum family planning (PPFP) information provision in the third trimester. The intervention phase of the study lasted from 26^th February to 30^th August 2022. It involved counseling on early PPFP delivered by trained nurses and community health workers. The study followed clients to assess the quality of the counseling session and determine their intention to use early PPFP. The follow-up phase occurred 14 and 20 weeks post-delivery to assess PPFP uptake in the first three months after delivery, commencing 23^rd May 2022 to 10^th Feb 2023. The during the study, there was continuous monitoring and necessary corrective actions were taken. Clients who lost their sexual partner, underwent postpartum psychosis, or were hospitalized for more than 14 weeks postpartum were discontinued from the study.

The study trained research assistants at each Health Centre and community unit to ensure data management compliance. The primary outcomes were the intention and actual use of early PPFP. The ANC service provider filled out the Case Report Forms at recruitment, and the counselor filled out an appointment card after the client agreed to a follow-up date for PPFP. The client’s exit interview was done after the intervention to assess process quality and measure TPB constructs parameters.

Measurement

Adapted TPB model constructs were knowledge PPFP, satisfaction PPFP counseling, perceived normative beliefs on early PPFP, attitude towards early PPFP, behaviour control of PPFP choice, PPFP intention, perceived risk of early postpartum pregnancy, and uptake of PPFP in the early postpartum period.

The core concepts were measured using adapted validated tool as per the manual for studying TPB by Francis et al., 2004. The primary outcome which was the modeled behaviour was uptake of PPFP in the early postpartum period at 3 months post-delivery. Five tools were used for data collection, namely; client exit interview guide, case report form, appointment card, Site appraisal form and questionnaire. All the tools were used to collect quantitative data except site appraisal form and some questions in the questionnaire that need brief explanation.

The Theory of Planned Behaviour was applied to design the tools. Phone based (Kobo Collect) Case Report Forms (CRFs) was used in recruitment of clients to the study. These CRFs had study’s eligibility criteria, client’s biodata, past and current obstetric information of the client, and past and present medical history of the of client.

Client exit interview guide and Site appraisal form were developed based on the procedures set out in the counseling guide and the overall set up of the counseling session. Appointment card was source of information on client details, proposed date for PPFP initiation and vital PPFP information summary. The other quantitative data collection tool was questionnaire during postpartum follow up. The questionnaire was tailored to the recommendations of Francis et al., 2004 and Ajzen and Klobas 2013 [1,33]. The tools were piloted in advance with an appropriate sample of service providers, CHWs and mock clients from the pilot sites before finalizing the forms.

Attitude towards early PPFP was directly measured using four 7-point Likert scale (Cronbach’s alpha 0.844) ranging from strongly agree (7) to strongly disagree (1). Attitude was then summarized into a continuous variable by calculating the mean score of parameter Likert scores. The mean scores were scaled to classify attitude as follows; very negative attitude (≥1, <2); moderate negative attitude (≥2, <3); weak negative attitude (≥3, <4); neutral attitude (≥4, <5); weak positive attitude (≥5, <6); moderate positive attitude (≥6, <7) and very positive attitude (≤7) [33].

Perceived normative beliefs on early PPFP was directly measured using five (5) 7-point Likert scale paired variables (Cronbach’s alpha 0.807) ranging from strongly agree (7) to strongly disagree (1). Perceived normative beliefs was then summarized into a continuous variable by calculating the mean score of parameter Likert scores. The mean scores were scaled to classify perceived normative beliefs as follows; very positive (7), positive (≥6 <7), moderately positive (≥5 <6), neither positive nor negative (≥4 <5), moderately negative (≥3 <4), negative (≥2 <3), and very negative (≥ <2).

Behavioural control was measured by three (3) 7-point Likert scale questions (Cronbach’s alpha 0.779) ranging from strongly agree (7) to strongly disagree (1). Behavioural control was then summarized into a continuous variable by calculating the mean score of parameter Likert scores. The mean scores were scaled to classify perceived behavioural control as follows; very high (7), high (≥6 <7), moderately high (≥5 <6), neither high nor low (≥4 <5), moderately low (≥3 <4), low (≥2 <3), and very low (≥1 <2).

Intention to use early PPFP was directly determined by three (3) 7-point Likert scale questions (Cronbach’s alpha 0.850) ranging from strongly agree (7) to strongly disagree (1). Intention was then summarized into a continuous variable by calculating the mean score of parameter Likert scores. The mean scores were scaled to classify intention as follows; very high intention (= 7), high intention (≥6, <7), moderately high intention (≥5, <6), neither high nor low high intention (≥4, <5), moderately low intention (≥3, <4), low intention (≥2, <3), and very low intention (≥1, <2).

The constructs adapted into the TPB model were knowledge of PPFP, satisfaction with PPFP counseling, and perceive risk of early postpartum pregnancy. Knowledge of PPFP was directly determined by six (6) 7-point Likert scale questions (Cronbach’s alpha 0.688) ranging from strongly agree (7) to strongly disagree (1). Knowledge was then summarized into a continuous variable by calculating the mean score of parameter Likert scores. The mean scores were scaled to classify knowledge as follows; very high knowledge (= 7), high knowledge (≥6, <7), moderately high knowledge (≥5, <6), neither high nor low high knowledge (≥4, <5), moderately low knowledge (≥3, <4), low knowledge (≥2, <3), and very low knowledge (≥1, <2).

Satisfaction with PPFP counseling was directly determined by six (6) 7-point Likert scale questions (Cronbach’s alpha 0.892) ranging from very satisfied (7) to very dissatisfied (1). Satisfaction was then summarized into a continuous variable by calculating the mean score of parameter Likert scores. The mean scores were scaled to classify satisfaction as follows; very satisfied (7), satisfied (≥6, <7), moderately satisfied (≥5, <6), neither dissatisfied nor satisfied (≥4, <5), moderately dissatisfied (≥3, <4), dissatisfied (≥2, <3), and strongly dissatisfied (≥1, <2).

Perceive risk of early postpartum pregnancy was directly determined by two (2) 7-point Likert scale questions (Cronbach’s alpha 0.823) ranging from very satisfied (7) to very dissatisfied (1). Perceive risk of early postpartum pregnancy was then summarized into a continuous variable by calculating the mean score of parameter Likert scores. The mean scores were scaled to classify Perceive risk of early postpartum pregnancy as follows; very high perceived risk (= 7), high perceived risk (≥6, <7), moderately high perceived risk (≥5, <6), neither high nor low high perceived risk (≥4, <5), moderately low perceived risk (≥3, <4), low perceived risk (≥2, <3), and very low perceived risk (≥1, <2).

Data analysis

In TPB, intention is an immediate determinant of behaviour thus the effect of the other constructs; Knowledge of PPFP, satisfaction with counseling process, perceived risk of getting pregnant in the early postpartum period, individual attitude towards behaviour, perceived normative beliefs about the behaviour, and perceived individual control of the behaviour are mediated by intention. Therefore, Structural Equation Modeling (SEM) was done using SPSS AMOS version 21 to assess and propose modifications to the model for improvement of model fit indices. Factor analysis of the parameters and complete path analysis for the constructs in TPB was performed with adjustment for model fit [34]. The exogenous variables were attitude towards PPFP in early postpartum period, perceived normative beliefs towards PPFP in early postpartum period, behavioural control of choice of PPFP in early postpartum period, knowledge about PPFP in early postpartum period, satisfaction with counseling on PPFP in early postpartum period, and perceived risk of getting pregnant in the early postpartum period. The endogenous variable was intention. Intention was theorized as the direct determinant of behaviour was the last endogenous variable in the model. Five steps were inherent in the process of SEM; conceptualizing the model, constructing a measurement model in SPSS AMOS for Confirmatory Factor Analysis (CFA) and optimization of the model for fit, constructing a structural model using optimized CFA, full structural model path analysis and adopting a final model [35].

Reflective formative assessment of the measurement model was done with CFA. Factor loadings, correlation of constructs and their indicators, model fit indices and model modification suggestions generated by the software guided the process optimizing model fit so that its predictions could be generalizable. The model-fit indices were used to assess the model’s overall goodness of fit. Goodness of fit was assessed based on the prescribed thresholds for the C-Minimum Discrepancy Function by Degrees of Freedom (CMIN/df), Goodness of Fit Indices (GFI); the Tucker and Lewis (1973) Index (TLI); the Confirmatory Fit Index (CFI) [36,37]. Further assessment of fit was based on SPSS AMOS computed value of the Standardized Root Mean Square Residual (SRMR) and the Root Mean Square Error Approximation (RMSEA) [37,38]. The thresholds for the critical model indices were CMIN/df ≤3, GFI >0.9, CFI>0.9, TLI>0.9, SRMR<0.08 and RMSEA<0.08). Parameter’s reliability was assessed using Cronbach’s alpha >0.7 and composite R >0.7 for construct reliability. Convergent validity of scale items was estimated using Average Variance Extracted (AVE) >0.50 [39,40]. All P<0.05 were considered statistically significant.

Moderation effect of sociodemographic factors, process factors and quality of counseling, intimate partner relationship on the optimized model was assessed using ordinal regression analysis. The sociodemographic aspects assessed were age, residence, education level, income, and employment status. Residence, education level and employment status were converted to ranked order based on the trends established in the KDHS 2014 [41]. There was an increase in utilization of FP as age, level of education, income quantiles increased and those who were employed and those who lived in urban areas had higher FP utilization rates as compared to their counterparts. Therefore, marital status was assigned ranks as follows; never married “1”, separated “2” and married “3”. Level of education was assigned ranks as; no education and primary level education “1”, secondary education “2” and tertiary level education “3”. Employment status was assigned as follows; not employed “1”, housewife “2” self-employed “3” and formally employed “4”. Quality of counseling was based on 7-point Likert scale self-score on counseling process fidelity and client FP knowledge. Process factors were; staff score on counseling refresher training posttest, counseling waiting time and counseling turnaround time. Intimate partner relationship rating was assessed by Likert scale scoring for physical, sexual violence, and partner support willingness to be involved in FP decisions-making.

Ethical consideration

The study was approved by Masinde Muliro University of Science and Technology (MMUST) School of Graduate Studies (SGS) (Ref: MMU/COR:509099). Ethical clearance was obtained from the MMUST Institutional Ethics Review Committee (IERC) (MMUST/IERC/013/2021). A research authorization permit was acquired from NACOSTI (Ref. No. 522628). An official data collection permission letter was obtained from the County Director of Medical Services (DMS). The trial was prospectively registered with the Pan African Clinical Trial Registry (PACTR), PACTR202109586388973. Signed written informed consent was obtained from all participants after they were introduced to the purpose of the study and informed about their rights. Refresher training was conducted for the nurses offering antenatal care services to the clients in both the intervention and control arms in order to maximize beneficence.

Participant characteristics

The sample size was 246 pregnant women attending ANC were included in the study with each arm attaining 82 (100%) of the sample size. Participants had an age range of 16–42 years (M = 25.2, SD = 4.9) modal age group being 15–24 years. Most (84.1%) of the participants were married. The highest attained level of education for the participants was tertiary and more than 63% of the participants had attained high school or tertiary education. More than 86% of the participants and almost all the participants earned less than 5000 KES and were Christians, respectively. Mean age was 25.2 years (SD 4.9) was grouped into 3 groups with intra-cluster range of 10 years with a minimum age of 16 years and a maximum of 42 years. The modal age group was 15–24 years.

Early PPFP use.

Early PPFP use was high at 85.8%. Study arms in the rural facilities realized higher PPFP commencement rates (90.2%) compared to the urban setting arm (81.3%). Table 2 shows that intervention arm had higher PPFP commencement rates (90.9%) than the control group (75.6%) (OR:3.2; 95% CI:1.5–6.7; P<0.0001). Nurses’ arm showed higher rates of commencement (95.1%) as compared to control OR:6.3; 95% CI: 2.0–19.4; P<0.0001 arm. Equally, community arm had higher numbers (86.6%) of participants starting PPFP than control arm (75.6%). The most common type of contraception was hormonal methods as follows; Implant (36.0%), injection DMPA (19.0%), and oral pills (16.1%). Barrier methods (9.5%) came in second after the hormonal methods and IUDs was third (8.1%). The rest of the methods accounted for 11.4% of the uptake.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Commencement rates for early PPFP.

https://doi.org/10.1371/journal.pone.0315029.t002

Measurement of TPB constructs

The adapted TPB model constructs were analyzed for basic descriptive statistics that summarized parameter Likert scale scores as shown in Table 3. Knowledge of early PPFP was assessed by 5 parameters based on FP choice, benefits, side effects, risk-benefit analysis and knowledge on one of the methods. Participant sentiments on each of the parameter were rated on 7-point Likert scale. The score was then averaged to get the overall knowledge score. The overall knowledge score was then re-rated to Likert score to express knowledge in an ordinal scale ranging from very low knowledge (1) to very high knowledge (7). The average score for knowledge was high 6.3 (SD = 0.5). Eighty-five percent (85.0%) of the participants’ knowledge rating of at least 6 (High). None of the participants rated below 4.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Summary of TPB constructs parameter attributes distribution per study arm.

https://doi.org/10.1371/journal.pone.0315029.t003

Client’s perceived risk of getting pregnant in early postpartum period was assessed by two parameters; perceived general risk and perceived individual risk of pregnancy in early postpartum period. The overall individual rating for perceived risk of getting pregnant in early postpartum period was arrived at by averaging the scores for responses for each of the questions thus very high perceived risk was average of 7, high perceived risk ≥6 and <7, moderately high perceived risk ≥5 and <6, neither high nor low high perceived risk ≥4 and <5, moderately low perceived risk ≥3 and <4, low perceived risk ≥2 and <3, and very low perceived risk ≥1 and <2. More than 85.0% of the participants had at least high perceived risk and the rest fell between low perceived risk to moderately high perceived risk with an average perceived risk of 6.1 (SD = 1.1).

Attitude towards behaviour was assessed using 4 Likert scale parameters targeting attitude towards uptake, attitude towards recommended minimum of 24 months of interbirth spacing, role of early PPFP in attaining recommended interbirth spacing and side effects of PPFP. Individual attitude towards PPFP was assessed using Likert scale indicators and overall rating was converted into a continuous variable by computing mean score of measurement parameters. These were summarized into frequency tables and measures of central tendency and dispersion. Overall attitude was converted into Likert scale as follows; very negative attitude (≥1, <2); moderate negative attitude (≥2, <3); weak negative attitude (≥3, <4); neutral attitude (≥4, <5); weak positive attitude (≥5, <6); moderate positive attitude (≥6, <7) and very positive attitude (≤7). The average rating of attitude towards early PPFP was high 5.8 (SD = 1.0). Average attitude was skewed towards positive with 96.4% of the attitude being rated in the positive spectrum as follows; moderate positive attitude 59.8%, weak positive attitude 26.4%, and very positive attitude 10.2%. The remaining 3.6% was accounted for by neutral attitude 2.0%, weak negative attitude 0.8% and moderate negative attitude 0.8%.

Perceived social normative beliefs was assessed as a precursor of intention which is a direct determinant of behaviour (early PPFP use). This was assessed by 5 Likert scale parameters on social normative beliefs on use of FP, early PPFP use, social pressure on use, partner approval and partner openness about early PPFP. Overall perceived social normative beliefs on PPFP were converted into a continuous variable by getting mean score of the Likert scale scores for each parameter. The mean of social normative beliefs was then calculated and used to classify perceived normative beliefs as follows; very positive (7), positive (≥6 <7), moderately positive (≥5 <6), neither positive nor negative (≥4 <5), moderately negative (≥3 <4), negative (≥2 <3), and very negative (≥ <2). Perceived normative beliefs about early PPFP were distributed across the whole spectrum from very negative perceived norm to very positive perceived norm. The average perceived norm was high 5.3 (SD = 1.3). The participants had a preponderance towards positive normative beliefs about early PPFP with 75.2% having either moderately positive (31.3%), positive (42.7%), or very positive (1.2%) perceived normative beliefs.

In this study, perceived individual control of PPFP choice was assessed by three (3) parameters geared towards confidence in using early PPFP, ease of PPFP use and ease of decision to use PPFP. Perceived individual control of PPFP choice was assessed using Likert scale questions and summarized into a continuous variable which was later converted into an ordinal scale. Before conversion to scale the findings of perceived individual control of PPFP choice were summarized into descriptive statistics for continuous variable. The mean of parameters of perceived individual control of early PPFP were calculated to estimate overall control beliefs thus classified as; very high (7), high (≥6 <7), moderately high (≥5 <6), neither high nor low (≥4 <5), moderately low (≥3 <4), low (≥2 <3), and very low (≥1 <2). The average perceived individual control of early PPFP was high 6.1 (SD = 0.9). Overall perceived behavioural control ranged from low (2%), moderately low (1.6%), neither high nor low (4.1%), moderately high (9.8%), high (61.4%) and very high (21.1%). The participants had a generally high level of perceived control of PPFP choice with 92.3% ranging in the moderately high, high and very high spectrum of perceived control.

Intention to use early PPFP was assessed using 3 Likert scale questions and summarized into a continuous variable which was later converted into an ordinal scale by averaging the score per participant. After getting the mean scores for intention the following criteria was used to classify it; very high intention (= 7), high intention (≥6, <7), moderately high intention (≥5, <6), neither high nor low intention (≥4, <5), moderately low intention (≥3, <4), low intention (≥2, <3), and very low intention (≥1, <2). The distribution of intention was as follows; neither high nor low intention 3.3%, moderately high intention 13.4%, high intention 45.9%, and very high intention 37.4% (M = 6.2, SD = 0.8).

Structural Equation Modeling (SEM) for the adapted TPB

Theory of Planned Behaviour variable relationship with intention which is the key determinant of behaviour was assessed using structural equation modeling using SPSS AMOS. In addition to the traditional 3 constructs (behavioural control, attitude toward the behaviour and perceived normative beliefs about the behaviour) that predict behavioural intention, three more constructs were added to the model for this research. The added constructs were; client’s knowledge about PPFP, client satisfaction with PPFP counseling and client’s perceived risk of getting pregnant in the early postpartum period.

Measurement model analysis

Confirmatory Factor Analysis (CFA) model was designed (Fig 3), and computed using AMOS to test the measurement model. As part of confirmatory factor analysis, factor loadings were assessed for each item and the model modification indices proposed covariation of error terms (e) e8: e9, e13: e14, e14: e15, e18: e19 and e25: e26. This improved the model fit indices to acceptable thresholds.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Measurement model.

SPSS AMOS measurement modelling showing the relationships between latent variables and their observed indicators. Parameters for measurement of the constructs: KN—Knowledge about early PPFP; SAT—Satisfaction with PPFP Counseling; PN—Perceived Normative beliefs on early PPFP; DRA—Attitude towards PPFP; BC—Behaviour control of PPFP Choice; INT—PPFP Intention; PR—Perceived risk of early postpartum pregnancy; e- error terms.

https://doi.org/10.1371/journal.pone.0315029.g003

Model fit assessment

The model-fit indices were used to assess the model’s overall goodness of fit. Goodness of fit was assessed based on the prescribed thresholds for the C-Minimum Discrepancy Function by Degrees of Freedom (CMIN/df), Goodness of Fit Indices (GFI); the Tucker and Lewis (1973) Index (TLI); the Confirmatory Fit Index (CFI), [36,37,42]. Further assessment of fit was based on SPSS AMOS computed value of the Standardized Root Mean Square Residual (SRMR) and the Root Mean Square Error Approximation (RMSEA) [37,38]. The thresholds for the critical model indices were CMIN/df ≤3, GFI >0.9, CFI>0.9, TLI>0.9, SRMR<0.08 and RMSEA<0.08). The seven-construct modified TPB model yielded a good fit for the data: CMIN/df = 1.754, GFI = 0.857, CFI = 0.935, TLI = 0.925, SRMR = 0.0709, and RMSEA = 0.055. All the values were within their respective common acceptance levels [37].

Reliability and validity.

Reliability was assessed by Cronbach’s Alpha and composite reliability and validity was assessed by Average Variance Extracted (AVE) as shown in Table 4. Construct reliability was assessed using Cronbach’s Alpha and composite reliability for indicators of each construct. Cronbach Alpha for each construct in the study was found to be above the required limited of 0.70 except for Knowledge of PPFP (0.688) [11,43,44]. Composite reliabilities were above threshold of 0.70 benchmark [40,42] except for perceived normative beliefs on early PPFP.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Reliability of adapted TPB Model constructs parameters.

https://doi.org/10.1371/journal.pone.0315029.t004

Convergent validity of scale items was estimated using Average Variance-Extracted (AVE). The Average Variance-Extracted values were above the threshold value of 0.50 [14,38,39]. Therefore, the scales used for each construct in the present study had the required convergent validity.

The measurement model demonstrates good overall reliability and validity. However, attention should be directed toward improving the internal consistency of the "Knowledge about early PPFP" construct.

Structural model assessment

A Structural Equation Model generated (Fig 4), through SPSS AMOS was used to test the relationships in the adapted TPB model. Goodness of fit was assessed based on the prescribed thresholds CMIN/df ≤3, GFI >0.9, CFI>0.9, TLI>0.9, SRMR<0.08 and RMSEA<0.08) [36,37,40].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. TPB structural model.

SPSS AMOS Structural model showing the relationships/ paths between variables. Parameters for measurement of the constructs: KN—Knowledge about early PPFP; SAT—Satisfaction with PPFP Counseling; PN—Perceived Normative beliefs on early PPFP; DRA—Attitude towards PPFP; BC—Behaviour control of PPFP Choice; INT—PPFP Intention; PR—Perceived risk of early postpartum pregnancy; e- error terms.

https://doi.org/10.1371/journal.pone.0315029.g004

Step 0: The model modification indices proposed covariation of error terms e8: e9, e13: e14, e14: e15, e18: e19 and e25: e26 from CFA. These model modifications were maintained in the Structural Equation Modeling (SEM).

Step 1: Full structural model analysis focused on the effect of each construct on the other using path standard regression weights, validity of the indicators of each construct and correlation coefficients. It was then noted that knowledge had a low Cronbach’s alpha (0.688) and AVE (0.359) showing construct and convergence validity incoherence of its assessment criteria and indicators. Likewise, knowledge had the highest correlation with client satisfaction with counseling (0.816), PPFP behavioural control of choice (0.560) and perceived normative beliefs about early PPFP (0.352). The last 2 constructs which knowledge had high correlation with are core components of the TPB model. Considering these facts knowledge was removed from the initial model leading to the final structural model (Fig 5). This optimized the standardized regression weights of the other constructs that were fitted in the model. Thus, no further adjustments were made to the model. Model fit indices for the final model shown in table 96 fell within the acceptable range: CMIN/df = 1.874, GFI = 0.881, TLI = 0.937, CFI = 0.948, SRMR = 0.073, and RMSEA = 0.060.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Optimized structural model.

SPSS AMOS Model optimization/modification by; removing paths, covariances, or adjusting factor loadings. Knowledge for PPFP removed due to Multicollinearity. Parameters for measurement of the constructs: KN—Knowledge about early PPFP; SAT—Satisfaction with PPFP Counseling; PN—Perceived Normative beliefs on early PPFP; DRA—Attitude towards PPFP; BC—Behaviour control of PPFP Choice; INT—PPFP Intention; PR—Perceived risk of early postpartum pregnancy; e- error terms.

https://doi.org/10.1371/journal.pone.0315029.g005

Interaction of constructs in the final SEM model.

Table 5 shows the interaction of constructs in the SEM model. The squared multiple correlation was 0.71 for intention to use PPFP in the first three months of delivery, this shows that 71% of variance in intention to use PPFP in the first three months of delivery was accounted for by the exogenous constructs in the model i.e. satisfaction PPFP counseling, perceived normative beliefs on early PPFP, attitude towards PPFP, behaviour control of PPFP choice, and perceived risk of early postpartum pregnancy.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Interaction of constructs in the final SEM model.

https://doi.org/10.1371/journal.pone.0315029.t005

All the constructs had a positive impact on intention to use PPFP in early postpartum period; satisfaction PPFP counseling (β = 0.197, t = 3.221, P = 0.001), perceived normative beliefs on early PPFP (β = 0.238, t = 3.981, P<0.0001), attitude towards PPFP (β = 0.632, t = 6.128, P<0.0001), behaviour control of PPFP choice (β = 0.157, t = 2.371, P = 0.018), and perceived risk of early postpartum pregnancy (β = .026, t = 0.506, P = 0.614).

Moderation effect

Moderation effect of general aspects such as sociodemographic factors, process factors and quality of counseling, intimate partner relationship on the optimized model was assessed by ordinal logistic regression analysis as premeditated in the conceptualized model. It was not operationally feasible to model for each of these factors individually in the SEM phase. Likewise, now that the study had three distinct arms, the moderating effect of the intervention was assessed using ANOVA. These aspects were explored to enrich the interpretation of the results. Their effect was not directly modeled for in the SEM.

Analysis of moderation of general aspects.

The moderation analysis for the general factors evaluated the influence of various factors on the intention for early postpartum family planning (PPFP). These factors were divided into different categories including sociodemographic characteristics of the clients, such as age, level of education, monthly income, marital status, and employment status. The pregnancy aspects analyzed included the number of antenatal care visits, the gestation period when PPFP counseling was offered, the presence of any comorbidities, the number of children, the health education received during pregnancy, and any complications that arose during pregnancy. The aspects related to labor, delivery, and postpartum were also considered, including health education received after birth, complications during labor and delivery, postpartum complications, and the health status of the woman three months postpartum. The previous family planning (FP) experience, including the rating of the woman’s previous experience with FP and the estimated cost of previous FP services, was taken into account, as well as the intimate partner relationship rated on a 7-point Likert scale ranging from very bad (1) to very good (7). Finally, the process-related aspects of the PPFP counseling, such as the waiting time for counseling, the speed of the counseling process, the mode of counseling, the quality of counseling, the fidelity to the process, and the willingness of the woman to set a postnatal appointment for PPFP, were also analyzed.

Ordinal logistic regression was used to demonstrate the moderating effect of general aspects; sociodemographic, pregnancy, labor and delivery, previous family planning experience, intimate partner relationship, and process-related aspects on the intention to use early postpartum family planning (PPFP) (Table 6). The sociodemographic characteristics fitted in the ordinal regression analysis model were; clients age, level of education, monthly income, marital status, and employment status were fitted in the ordinal regression analysis model. Only client’s age had a significant effect on intention to use early PPFP. An increase in age increased the odds of higher intention to use early PPFP (OR:1.1, 95% CI: 1.0–1.2, P = 0.022).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Client and process aspects and intention to use early PPFP.

https://doi.org/10.1371/journal.pone.0315029.t006

The pregnancy-related aspects fitted in the ordinal regression analysis model were; number of ANC visits, gestation when PPFP counselled, comorbidity, number of children, health education in pregnancy, and complication during pregnancy. Number of children and health education during pregnancy had significant effect on intentions for early PPFP. Having more children increased the odd of having high intention for early PPFP (OR:1.3; 95% CI: 1.1–1.6; P = 0.004). Being health educated was a binary variable with a 1 for “yes” and a 2 for “no” thus the tendency towards not having been educated in pregnancy reduced the odds of intending to use early PPFP (OR:0.3; 95% CI: 0.1–1.0; P = 0.042).

The labour, delivery and postpartum aspects fitted in the ordinal logistic regression model were; health education afterbirth, labour complications, postpartum complication, and health status in 3 months postpartum. None of the labour, delivery and postpartum related aspect had a significant effect on intentions for early PPFP. Rating of previous experience with FP and estimated cost of previous FP services did not elicit any significant effect on intentions for early PPFP.

Intimate partner relationship was rated on a 7-point Likert scale ranging from very bad intimate partner relationship (1) to very good intimate partner relationship (7). There was a significant relationship between intimate partner relationship and intentions for early PPFP. As the rating of intimate partner relationship increased, there was an increase in log odds of intentions for early PPFP (OR: 1.5; 95% CI: 1.2–1.8, P<0.0001).

Ordinal regression analysis to assess effects of process related aspects on intention to use early PPFP was done. FP counseling waiting time, counseling turnaround, mode of counseling, counseling quality, fidelity to process and accepting to set postnatal appointment for PPFP were fitted in the ordinal regression model. FP Counseling turnaround, fidelity to process and accepting to set postnatal appointment for early PPFP had a significant effect on intention to use early PPFP. As both fidelities to process and chance of setting postnatal appointment for early PPFP increased there was a congruent increase in the log odds of intention to use early PPFP (OR:2.62; 95% CI: 1.9–3.6, P<0.0001) and (OR:2.4; 95% CI:1.3–4.6, P = 0.007), respectively. The FP counseling turnaround had an inverse relationship with intention to use early PPFP whereby as counseling turnaround time increased the log odds of intention to use early PPFP reduced (OR:0.9; 95% CI: 0.9–1.0, P<0.0001). FP counseling waiting time had a similar relationship with intention as FP counseling turnaround time albeit not significant. Mode of counseling (individual counseling) and quality of counseling increased the odds of intention to use early PPFP although not significant.

Moderating effect of intervention on intention for early PPFP per study arms. A one-way ANOVA was done to assess the moderating effect of intervention on intention for early PPFP. A summary of descriptive statistics per arm was also done as a preamble to the main ANOVA test. Partial eta2 was used to estimate between arm effect size for ANOVA. The Nurses Arm had a mean intention for early PPFP score of 6.6 (SD = 0.9), the Community Arm had a mean score of 6.1 (SD = 0.9), and the Control Arm had a mean score of 6.0 (SD = 0.7). Across all study arms, the total mean intention score for Early PPFP was 6.2 (SD = 0.9).

A one-way ANOVA for intentions for early PPFP was done with Levene’s test showing that homogeneity of variance was met F(2,243) = 0.7, P = 0.490 thus Tukey’s post-hoc test used to estimate which arms had significant difference in intention, effect size estimation between arms. The ANOVA for intentions for early PPFP and study arms revealed a significant difference in the mean client intention scores between arms F(2,243) = 12.4, P<0.0001 with Tukey’s post-hoc test showing significantly higher intention for early PPFP between nurses’ arm and community arm (P<0.0001) and control arms (P<0.0001) with effect size varying for between arm comparisons. There was no significant difference in intention between the community arm and control arm (P = 0.986) as shown in Table 7.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 7. Analysis of variance (ANOVA) for difference in intentions for early PPFP.

https://doi.org/10.1371/journal.pone.0315029.t007

Study limitations

It is important to note that the findings of this study should be considered within the context of its limitations, such as the study setting and generalizability of the findings to other populations. The study is limited by selection bias, as it focuses solely on individuals attending antenatal care (ANC) clinics, potentially skewing the findings’ generalizability to the broader pregnant population. The study recognizes the need for a more diverse sample beyond ANC attendees to enhance external validity. Additionally, the study’s design limitation includes recruiting participants mostly in the third trimester, preventing the assessment of the intervention’s impact earlier in pregnancy and missing insights into factors influencing family planning decisions in the early stages.