Determinants and impact of joint decision-making on contraceptive use among women in union in Nairobi’s urban informal settlements: a repeated cross-sectional analysis and agent-based microsimulation using the Nairobi Urban Health and Demographic Surveillance System

Abstract

Objectives

To estimate the prevalence and determinants of joint decision-making on contraceptive use among women in union in Nairobi’s informal settlements, and to project the contribution of different decision-making patterns to the modern contraceptive prevalence rate (mCPR) over time.

Design

Analysis of repeated cross-sectional surveillance data combined with an agent-based microsimulation (Thanzi la Onse-Nairobi Urban Health and Demographic Surveillance System (NUHDSS). Determinants of joint decision-making were examined using multivariable logistic regression, and model performance was assessed using the likelihood ratio test, the Hosmer-Lemeshow goodness-of-fit test, McFadden pseudo R² and the area under the receiver operating characteristic curve. The microsimulation projected mCPR from 2012 to 2040 under joint, woman-only and spouse-only decision-making scenarios.

Setting

Two urban informal settlements, Korogocho and Viwandani, in Nairobi, Kenya, covered by the NUHDSS, 2012–2015.

Participants

2085 women of reproductive age (15–49 years) who were in union, defined as married or living with a male partner, with complete data on the decision-making outcome.

Primary and secondary outcome measures

The primary outcome was joint decision-making on contraceptive use (woman and partner deciding together vs the woman alone or the partner alone). The secondary outcome was the projected mCPR among women in union from 2012 to 2040 under joint, woman-only and spouse-only decision-making scenarios.

Results

Joint decision-making prevalence was 63.7% (95% CI 61.6 to 65.8). In the adjusted model, the odds of joint decision-making were higher in male-headed households (AOR 1.53, 95% CI 1.11 to 2.10, p=0.009), among residents of Viwandani (AOR 1.30, 95% CI 1.04 to 1.64, p=0.024), among women who desired more than five children (AOR 2.27, 95% CI 1.02 to 5.12, p=0.045) and among unemployed women (AOR 1.36, 95% CI 1.09 to 1.69, p=0.006), and lower among women who did not want another child (AOR 0.63, 95% CI 0.48 to 0.82, p<0.001), were undecided about childbearing (AOR 0.62, 95% CI 0.43 to 0.87, p=0.006) or had not discussed family size with their partner (AOR 0.39, 95% CI 0.31 to 0.51, p<0.001). In the microsimulation, the contribution of joint decision-making to mCPR among women in union rose from 40% in 2012 to 52% in 2040, exceeding the contributions of woman-only (13% to 18%) and spouse-only (about 1% to 0.5%) decision-making.

Conclusions

Joint decision-making was associated with higher modern contraceptive use among women in union in Nairobi’s informal settlements, and the microsimulation suggests it may contribute substantially to mCPR over time. Because the design is cross-sectional, these associations should not be interpreted as causal. Programmes that strengthen couple communication, particularly in male-headed households and among unemployed women or couples with undiscussed or discordant fertility preferences, may support progress toward national family planning targets.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• The study draws on the Nairobi Urban Health and Demographic Surveillance System, a long-running and well-characterised platform covering two informal settlements, which supports internal validity within this setting.

• Determinants of joint decision-making were examined with multivariable logistic regression, with model performance assessed using the likelihood ratio test, the Hosmer-Lemeshow goodness-of-fit test, McFadden pseudo R² and the area under the receiver operating characteristic curve.

• Robustness was examined through treatment of age as a continuous predictor, and ordinal analyses of the three-category decision-making outcome.

• The repeated cross-sectional design cannot establish the temporal order between determinants and decision-making, so the associations reported are not causal.

Introduction

Fertility is instrumental in shaping the structure, size and composition of populations. High fertility, commonly defined as five or more births per woman over the reproductive lifespan, has been linked to increased health risks for both mothers and children.¹ Research also associates high fertility with challenges such as reduced investment in human capital, slower economic development and heightened environmental pressures.² Globally, the total fertility rate is estimated at 2.5 children per woman. In Kenya, this rate is currently 3.5, alongside a maternal mortality ratio of 355 deaths per 100 000 live births translating to an estimated 5000 maternal deaths each year.³ Projections indicate that Kenya’s population is expected to grow significantly, potentially reaching 57.8 million by 2030 if current fertility and growth trends persist.⁴

development and heightened environmental pressures.² Globally, the total fertility rate is estimated at 2.5 children per woman. In Kenya, this rate is currently 3.5, alongside a maternal mortality ratio of 355 deaths per 100 000 live births translating to an estimated 5000 maternal deaths each year.³ Projections indicate that Kenya’s population is expected to grow significantly, potentially reaching 57.8 million by 2030 if current fertility and growth trends persist.⁴

Family planning is an essential global tool for balancing population and economic growth. It provides various advantages, some of which are particular to maternal and child health. Others include socioeconomic benefits; for instance, delaying or limiting childbearing allows women to further their education and careers, which can lead to greater economic prospects for their families.⁵ Family planning reduces infant and maternal morbidity and mortality by reducing unplanned pregnancies and unsafe abortions. It also enables birth spacing, which lowers child mortality and improves the nutritional health of both mother and child.⁶ Healthcare coverage, including family planning, has improved over time. Between 2014 and 2020, PMA2020 surveys in Kenya reported that contraceptive use among women of reproductive age improved from 45% to 58%.⁵ This improvement has significantly impacted maternal, neonatal and child health. The maternal mortality ratio decreased from 520 in 2008–2009 to 362 per 100 000 live births in 2014. Family planning is crucial not only for improving the health of women and newborns, but also as a foundation for women’s empowerment. Despite the above-mentioned advances in contraceptive coverage, Kenya has a high prevalence of unmet needs. The Kenya Demographic and Health Survey (KDHS) 2014 report indicated that 18% of reproductive-age women face unmet needs, 9% for spacing and 8% for limiting. This marks a significant decline from 2008 to 2009 when 26% of women had an unmet need.⁷ The latest KDHS report (2022) indicated that the use of any contraceptive method among women in union was 63%, with 57% using modern methods and 6% using traditional methods. While this marks notable progress from 39% modern contraceptive prevalence rate (mCPR) in 2008–2009, it still falls short of the national FP2030 target of reaching 64% mCPR among women in union and reducing unmet need of family planning from 14% to 10%. Promoting a culture of joint decision-making between partners is strongly recommended as a strategy to help close this gap and improve family planning outcomes.

Decision-making power is an important aspect in accessing healthcare services, including utilisation of family planning services. The evidence indicates that women’s decision-making power has a major favourable impact on attending the continuum of maternal care. However, limited attention has been given to whether family planning decisions are made jointly or through collaboration. In many households, particularly in developing countries, decisions are often made by the partner or husband rather than mutually.⁸ Due to the economic, political and sociocultural status of women, they often have less decision-making power.⁹ Studies indicate that male partners can influence contraceptive use negatively by discouraging its use and this results in either discontinuation or convert use.¹⁰ This significant influence contributes to unplanned or unintended pregnancies, which often lead to unsafe abortions, which is a major direct contributor to maternal mortality. Women whose partners oppose the use of contraception opt to use secretly. This secrecy is associated with many challenges both at household level and women’s health. Evidence suggests that the family planning side effects can easily be noticed by the partners, which could potentially lead to couple dissolution.¹¹ It can also cause challenges in accessing prompt medical consultations from healthcare providers if contraceptive-induced side effects occur. Women without autonomy will resort to family planning methods which cannot be seen by their partners such as injectables and pills. These methods indirectly affect women’s health since they have higher side effects compared with other methods like implants.⁸ These sexual and reproductive health issues can be resolved by cultivating an environment that supports joint decision-making.

Joint decision-making between a couple is typically an important aspect of family life. If both relationship partners participate in a decision that impacts one of them, the outcome may be better than if either member decides alone, simply because more options are likely to be investigated when each spouse expresses his or her opinion. It may also have the advantage of allowing each spouse to exhibit respect and concern for the other partner while considering how decisions about one may affect the other. Data support the idea that a couple’s joint decision-making is better for maternal health outcomes than those made only by one spouse. Other research has demonstrated that joint decision-making and couple communication are linked to higher contraceptive usage, increased use of HIV-related health-enhancing behaviours and lower risk of interpersonal violence. Compared with when men or women make decisions alone, couples who jointly make decisions may also have better reproductive health outcomes.¹²

The fundamental difficulty is the engagement of partners in contraceptive decision-making. Sub-Saharan African evidence suggests that partners are frequently reluctant to engage in these decisions.¹³ According to clinical experience, when partners make the key decisions about contraceptive use, women may face unexpected pregnancies or be forced to use contraception in secret, and this is associated with several negative repercussions. Joint decision-making on family planning entails both the woman and her partner/husband agreeing on elements such as contraceptive method selection, its use and maintenance of usage while assessing any ongoing side effects.¹⁴ Identifying the barriers to joint decision-making is key in tackling the challenges women confront. Therefore, this study seeks to address the gap between current modern contraceptive uptake among women in union and the national mCPR target, as well as the issue of covert contraceptive use. Specifically, it examines the prevalence and determinants of joint decision-making on contraceptive use among women of reproductive age in union, living in informal settlements of Nairobi. Using secondary data from the Nairobi Urban Health and Demographic Surveillance System (NUHDSS), we also simulate the impact of different decision-makers on mCPR. The findings offer valuable insights for policymakers aiming to improve contraceptive uptake and promote shared decision-making in family planning.

Methods

Data source

The analysis used repeated cross-sectional fertility preference data sourced from the NUHDSS. The dataset was collected among all women of reproductive age (15–49 years) residing in two urban slums (Korogocho and Viwandani) between 2012 and 2015. The NUHDSS was pioneered by the African Population and Health Research Centre in consultation with the Nairobi City Council and residents of Korogocho and Viwandani communities with the aim of producing data for decision-making specific to Nairobi’s slum dwellers. The information from NUHDSS has facilitated the monitoring of progress toward several Millennium Development Goals and Sustainable Development Goals among urban poor populations.¹⁵ The dataset was restricted to include only women who were in a union, which was defined as being married or living with a male partner. This selection criterion was applied since joint decision-making in the use of family planning methods, which was the outcome variable of interest in the study, is only relevant to women who have a partner. Complete-case analysis was used for the primary outcome: women with missing information on the decision-making variable were excluded, which affected 29 of 2114 women in union (1.4%). To assess whether this exclusion was likely to bias the findings, we compared the socio-demographic characteristics of included and excluded women where distributions of most socio-demographic characteristics did not differ significantly between the included and excluded participants. Missing completely at random (MCAR) assumption was tested for model predictors using Little’s test (p<0.001). Little’s MCAR test was statistically significant indicating that missing data were not MCAR. Consequently, missing data on covariates were imputed using the mice package in R. Participants’ characteristics and model results with and without imputation were compared. The statistical significance of the variables remained unchanged, indicating that the complete-case analysis did not introduce substantial bias into the results. The extent, pattern of missingness for the outcome and for each covariate together with comparison of results with and without imputation are reported in online supplemental tables S1–S3. Because the NUHDSS covers two specific informal settlements, Korogocho and Viwandani, rather than all of Nairobi or all urban poor populations in Kenya, the sample is representative of women in union in these surveillance communities but is not intended to be nationally representative, a point we return to in the Discussion. The study is reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidance for cross-sectional studies, and the completed checklist is provided as an online supplemental file 2.

Study variables

Dependent variable

The outcome variable in this study was joint decision-making on contraceptive use. The decision-making on use of contraceptives was reported in three categories (decision being made by women, husband making decision and joint decision). This variable was divided into two categories: yes (if the women decide jointly with their spouses to use family planning) and no (if the women decide independently to use contraceptives or if the partner makes the decision).

Independent variables

Explanatory variables in the study consisted of socio-economic, demographic and reproductive characteristics of the respondents. A total of 12 individual-level variables (slum area of women, age of women, age of household head, women’s education, women’s employment, women’s desire number of children, women’s ethnicity, sex of household head, religion, parity, fertility preference, husband’s desire for children) were included in this study. The number and proportion of missing values for each covariate are summarised in online supplemental table S2; covariate missingness was low, and the multivariable model was fitted on women with complete data on the included variables. A household wealth index and a validated measure of women’s decision-making power were not collected in this module and therefore could not be included as covariates; women’s education and employment status were retained as available proxies for socio-economic position, and this limitation is considered in the Discussion.

Statistical analysis

Descriptive statistics were performed to obtain the prevalence of joint decision-making and its distribution across the explanatory variables. All explanatory variables were categorical. A χ² test was conducted to assess the association between each variable and contraceptive decision-making. Variables with statistically significant associations (p<0.05) were included in the univariate logistic regression analysis. Bivariate logistic regression was then performed for each significant variable to identify potential candidates for inclusion in the multivariable (adjusted) model. Variables with a p<0.15 in the bivariable analysis were included in the multivariable logistic regression model. We then conducted a multivariable logistic regression to examine the association between the selected explanatory variables and the outcome variable. To check for collinearity among selected variables, a multicollinearity test was conducted using a variance inflation factor (VIF). There was no evidence of collinearity between the variables in the test (mean VIF=1.06, max VIF=1.116, min VIF=1.014). The goodness-of-fit of the regression model was assessed using McFadden’s pseudo R², which was 0.156, suggesting moderate explanatory power, which is acceptable for behavioural models in public health research. The results were presented using adjusted OR (AOR) at 95% CI. Variables in the multivariable model with p value<0.05 were declared statistically significant. Effect sizes (ORs with 95% CIs) are reported alongside p values for all associations. In addition to McFadden pseudo R², overall model fit and discrimination were assessed using the likelihood ratio test comparing the fitted model with the null model (Likelihood Ratio (LR) χ²=130.03, df=16, p<0.001), the Hosmer-Lemeshow goodness-of-fit test (χ²=8.5085, df=8, p=0.3854) and the area under the receiver operating characteristic (ROC) curve (AUC=0.742, 95% CI 0.715 to 0.768). Several sensitivity analyses were performed. First, women’s age and age of the household head were modelled as continuous variables (per 1 year increase) in addition to the categorical specification. Second, because the decision-making variable was originally recorded in three categories (woman alone, partner alone and joint), treating the categories as an ordered gradient of shared involvement, we fitted an ordinal logistic regression. The outcome was ordered as husband decision, women decision and joint decision, representing increasing levels of shared decision-making. A proportional odds model was fitted to estimate the association between the explanatory variables and the odds of being in a higher category of contraceptive decision-making. ORs and 95% CIs were reported. The proportional odds assumption was assessed using the Brant test (p=0.7166) with joint decision-making as the reference category. The results of these sensitivity analyses are presented in online supplemental tables S4 and S5. Analyses were conducted in R V.4.3.2, and a two-sided p value below 0.05 was considered statistically significant.

Individual level model

The impact of joint decision-making on use of modern contraceptives was assessed using the Thanzi la Onse (TLO-NUHDSS) model. TLO model is an agent-based simulation framework designed to evaluate the health and economic effects of public health interventions in low- and middle-income countries, with initial development focused on Malawi. It simulates the lives of individual agents, each with attributes such as age, sex, household membership and socioeconomic status, over time, capturing their interactions with health systems and disease environments. The model operates in discrete time steps (typically monthly) and is modular, allowing users to activate only the relevant components for their analysis. In this study, we use the demographic, labour and contraception modules of the TLO model to explore how different patterns of decision-making around contraceptive use influence the mCPR over time. Full documentation, source code and technical details for the TLO model are available on the GitHub repository of the TLO modelling team, which provides open access to the model and its associated resources. TLO-NUHDSS is forked from the above repository and parameterised using the NUHDSS data to reflect the population of our study. We incorporated partnership status since our study focused on women in union and assigned decision-makers on use of contraceptives based on probabilities from the data. At the individual level, each simulated woman’s contraceptive status evolves through a set of monthly transition probabilities: initiation of a method, discontinuation (estimated through a competing-risk specification that captures the reasons for stopping), method switching and the risk of unintended pregnancy through method failure. These transition parameters, together with the probabilities assigning each woman to a decision-making category (woman alone, partner alone or joint), were calibrated to the NUHDSS data and held fixed across scenarios. The model is subject to the following constraints: all decision-making and transition probabilities are bounded between 0 and 1; the probabilities across the three decision-making categories sum to 1 for each woman; the underlying age structure and partnership distribution are taken from the surveillance data; and only the composition of decision-making was varied between the joint, woman-only and spouse-only scenarios, while all other parameters were held constant, so that differences in projected mCPR are attributable to decision-making rather than to other inputs. This individual-level transition model is therefore fully specified before the simulation results are presented. Figure 1 illustrates how the contraception module is designed. This takes the simulated data from the demography module and selects women of reproductive age (15–49) years whereby default, are not in contraception and not pregnant. During the initial phase of the simulation, women are assigned partnership status (in union or not in union) based on probabilities from the data. Women in union make decisions whether to use contraception or not, these decisions can be made independently (woman making her decision), jointly (women and her partner discussing) or spouse making decisions. Both women in union and those not in union are assigned different contraceptive methods, either short-term or long-term, based on their age. At each stage of the simulation, women who are not using any contraceptive method may initiate one, depending on method-specific initiation rates. They may also become pregnant or remain non-users. Women who are already using contraception may discontinue use for various reasons: some may wish to become pregnant, while others may stop due to side effects or other personal or external factors. Method discontinuation is modelled using discontinuation rates specific to each method, estimated through a competing risk approach that captures the various reasons women may stop using contraception. Women using contraceptives may also switch to a different method over time or experience unintended pregnancy as a result of method failure. Women who become pregnant take 1–9 months to give birth depending on the outcome of pregnancy. After birth, postpartum women may initiate contraceptive uptake or not.

Figure 1. Contraception module model design.

Results

Socio-demographic characteristics of the study participants

Table 1 summarises the socio-demographic characteristics of the respondents, stratified by joint decision-making on contraceptive use. A total of 2085 women of reproductive age, who were in union, were included in this study. Of these, 78.9% of the respondents were in the age group of 20–34 years, and more than half (69.0%) resided in Viwandani slum. Over half (57.5%) lived in households where the head was aged 30 years or younger, and most households (88.0%) were male-headed. Most of the respondents (55.2%) had no formal education or primary education level with more than half (67.9%) being unemployed. Additionally, more than half (57.0%) of respondents were identified as Protestants. A number of variables were significantly associated with joint decision-making on contraceptive use. For instance, age of the woman, age of household head, sex of the household head and slum area of residence were all significantly associated with joint decision-making (p<0.05). Women living in Viwandani were more likely to report joint decisions compared with those in Korogocho (71.9 vs 28.1). Similarly, male-headed households showed higher rates of joint decision-making than female-headed ones. Educational level and employment status also showed meaningful variation, with women who had no education or those with primary education, or were unemployed, more likely to participate in joint decision-making. However, there was no statistically significant difference between those making joint decisions and not making joint decisions by religion and ethnicity.

Table 1. Socio-demographic characteristics of respondents.

Decision-making		No	Yes	Total	Cramer’s V	P value
Total N (%)		756 (36.3)	1329 (63.7)	2085
Age of women in years	15–19	70 (9.3)	131 (9.9)	201 (9.6)	0.161	0.022
	20–34	580 (76.7)	1065 (80.1)	1645 (78.9)
	35–49	106 (14.0)	133 (10.0)	239 (11.5)
Age of household head in years	≤30	417 (55.2)	778 (58.8)	1195 (57.5)	0.158	0.030
	31–49	301 (39.8)	506 (38.2)	807 (38.8)
	50+	38 (5.0)	39 (2.9)	77 (3.7)
Sex of household head	Female	125 (16.5)	126 (9.5)	251 (12.0)	0.204	<0.001
	Male	631 (83.5)	1203 (90.5)	1834 (88.0)
Slum area of residence	Korogocho	274 (36.2)	373 (28.1)	647 (31.0)	0.185	<0.001
	Viwandani	482 (63.8)	956 (71.9)	1438 (69.0)
Religion	Catholic	221 (29.6)	356 (27.0)	577 (27.9)	0.136	0.451
	Muslim	35 (4.7)	57 (4.3)	92 (4.5)
	Others	71 (9.5)	147 (11.2)	218 (10.6)
	Protestant	420 (56.2)	758 (57.5)	1178 (57.0)
Women’s level of education	None or primary	431 (58.6)	698 (53.2)	1129 (55.2)	0.151	0.023
	Secondary or higher	305 (41.4)	613 (46.8)	918 (44.8)
Current employment status	No	473 (62.9)	936 (70.7)	1409 (67.9)	0.18	<0.001
	Yes	279 (37.1)	388 (29.3)	667 (32.1)
Woman’s ethnicity	Kamba	208 (27.5)	420 (31.6)	628 (30.1)	0.172	0.053
	Kikuyu	149 (19.7)	245 (18.4)	394 (18.9)
	Kisii	82 (10.8)	137 (10.3)	219 (10.5)
	Luhya	164 (21.7)	272 (20.5)	436 (20.9)
	Luo	95 (12.6)	124 (9.3)	219 (10.5)
	Other	58 (7.7)	131 (9.9)	189 (9.1)

Reproductive characteristics of the study participants

Table 2 presents the reproductive characteristics of women in relation to joint decision-making on contraceptive use. About half of the respondents (42.7%) were multiparous (having more than one child). More than half (68.7%) expressed a fertility preference for having more children. This was significantly associated with decision-making (p<0.001), with a higher proportion of women who desired more children reporting joint decisions (72.9%) compared with those who did not want another child (19.6%) and undecided (7.5%). More than three-quarters (90.2%) of participants desired two to four children, and 64.9% reported that their spouses shared the same preference. Husband’s/partner’s desire for children showed a strong association with joint decision-making (p<0.001), where 70.7% of women who reported that their husbands/partners desired the same number of children reported joint decision-making. Those who had not discussed the ideal number of children with their partner had a lower proportion of joint decision-making (14.9%). In regards to use of contraceptives, the most used method among the respondents was injectables with 36.1% using and 37.2% were not using any family planning method. Although injectables were the most common contraceptive method used in both groups, the association between contraceptive method type and decision-making was not statistically significant (p=0.185). In addition, parity also showed no significant association. The prevalence of joint decision on contraceptive use among in union reproductive age women in Nairobi informal settlements was found to be 63.7%; 95% (61.6% to 65.8%) (table 1 and table 2).

Table 2. Reproductive characteristics of respondents.

Decision-making		No	Yes	Total	Cramers’ V	P value
Total N (%)		756 (36.3)	1329 (63.7)	2085
Parity	Grand multiparous	49 (6.5)	70 (5.3)	119 (5.7)	0.15	0.164
	Multiparous	340 (45.0)	546 (41.4)	886 (42.7)
	Nulliparous	101 (13.4)	205 (15.5)	306 (14.8)
	Primiparous	265 (35.1)	498 (37.8)	763 (36.8)
Fertility preference	Does not want another child	193 (27.3)	246 (19.6)	439 (22.4)	0.22	<0.001
	Undecided	81 (11.4)	94 (7.5)	175 (8.9)
	Wants another child	434 (61.3)	913 (72.9)	1347 (68.7)
Woman desired number of children	0–1	21 (2.8)	20 (1.5)	41 (2.0)	0.196	<0.001
	2–4	659 (88.2)	1210 (91.4)	1869 (90.2)
	5+	51 (6.8)	89 (6.7)	140 (6.8)
	Undecided	16 (2.1)	5 (0.4)	21 (1.0)
Husband desire for children	Fewer children	17 (2.3)	35 (2.7)	52 (2.5)	0.302	<0.001
	More children	84 (11.3)	154 (11.7)	238 (11.6)
	Not discussed	235 (31.7)	195 (14.9)	430 (20.9)
	Same number	405 (54.7)	928 (70.7)	1333 (64.9)
Type of contraceptive method	Implants	93 (12.8)	136 (10.6)	229 (11.4)	0.161	0.185
	Injectables	241 (33.1)	485 (37.8)	726 (36.1)
	None	287 (39.4)	462 (36.0)	749 (37.2)
	Others	29 (4.0)	43 (3.4)	72 (3.6)
	Pill	38 (5.2)	74 (5.8)	112 (5.6)
	Rhythm	41 (5.6)	83 (6.5)	124 (6.2)

Determinants of contraceptive decision-making

The findings in tables 1 and 2 identified nine variables that were significantly associated with decision-making. These variables were further assessed using bivariate logistic regression analysis. All nine remained significant in the individual models and were therefore included in the multivariable logistic regression model. In the final adjusted analysis, six variables remained statistically significant predictors of joint decision-making on contraceptive use: slum area of residence, sex of the household head, women’s employment status, fertility preference, woman’s desired number of children and husband’s desired number of children. Women in male-headed households were 1.53 times more likely to decide jointly (AOR=1.53; 95% CI 1.11 to 2.10) than those in female-headed households. Women living in Viwandani were 1.30 times more likely to decide jointly (AOR=1.30; 95% CI 1.04 to 1.64) than women living in Korogocho. Unemployed women were 1.36 times more likely to decide jointly (AOR=1.36; 95% CI 1.09 to 1.69) than women who were employed. Women who did not want another child were 37% less likely to decide jointly (AOR=0.63; 95% CI 0.48 to 0.82) than women who wanted another child. Women who were undecided were 38% less likely to decide jointly (AOR=0.62; 95% CI 0.43 to 0.87) than women who wanted another child. Women who desired more than five children were 2.27 times more likely to make joint decision compared with those who desired no children or one child (AOR=2.27; 95% CI 1.02 to 5.12). Women who had not discussed the desired number of children with their spouse or partner were 61% less likely to decide jointly (AOR=0.39; 95% CI 0.31 to 0.51) than women who desired the same number as their spouse (table 3).

Table 3. Predictors of joint contraceptive decision-making.

	Univariate				Multivariate
Characteristic	N	OR	95% CI	P value	OR	P value
Age of women in years	2085
15–19		—	—		—
20–34		0.98	0.72 to 1.33	0.904	0.96 (0.67 to 1.36)	0.828
35–49		0.67	0.45 to 0.99	0.043	0.98 (0.59 to 1.62)	0.939
Age of household head in years	2079
≤30		—	—		—
31–49		0.90	0.75 to 1.09	0.272	1.12 (0.89 to 1.41)	0.342
50+		0.55	0.35 to 0.88	0.011	0.79 (0.44 to 1.43)	0.439
Sex of household head	2085
Female		—	—		—
Male		1.89	1.45 to 2.47	<0.001	1.53 (1.11 to 2.10)**	0.009
Slum area of residence	2085
Korogocho		—	—		—
Viwandani		1.46	1.20 to 1.76	<0.001	1.30 (1.04 to 1.64)*	0.024
Women’s level of education	2047
None or primary		—	—		—
Secondary or higher		1.24	1.03 to 1.49	0.020	1.03 (0.83 to 1.26)	0.808
Current employment status	2076
Yes		—	—		—
No		1.42	1.18 to 1.72	<0.001	1.36 (1.09 to 1.69)**	0.006
Fertility preference	1961
Wants a child		—	—		—
Does not want		0.61	0.49 to 0.76	<0.001	0.63 (0.48 to 0.82)***	<0.001
Undecided		0.55	0.40 to 0.76	<0.001	0.62 (0.43 to 0.87)**	0.006
Woman desired number of children	2071
0–1		—	—		—
2–4		1.93	1.03 to 3.60	0.038	1.63 (0.80 to 3.38)	0.177
5+		1.83	0.91 to 3.72	0.091	2.27 (1.02 to 5.12)*	0.045
Undecided		0.33	0.09 to 1.01	0.063	0.52 (0.12 to 1.97)	0.351
Husband desire for children	2053
Same number		—	—		—
Fewer children		0.90	0.50 to 1.66	0.723	0.83 (0.46 to 1.56)	0.548
More children		0.80	0.60 to 1.07	0.132	0.75 (0.55 to 1.02)	0.066
Not discussed		0.36	0.29 to 0.45	<0.001	0.39 (0.31 to 0.51)***	<0.001

^*p<0.05; **p<0.01; ***p<0.001.

CI, Confidence Interval; OR, Odds Ratio.

The multivariable model showed acceptable fit and discrimination. The likelihood ratio test indicated that the model performed significantly better than the null model (LR χ²=130.03, df=16, p<0.001), the Hosmer-Lemeshow test was consistent with adequate fit (χ²=8.5085, df=8, p=0.3854), the area under the ROC curve was 0.742 (95% CI 0.715 to 0.768) and McFadden pseudo R² was 0.156. In sensitivity analyses, modelling age as a continuous variable did not materially change the findings and age remained not statistically significant, consistent with the results obtained when it was modelled as age groups; ordinal analyses of the three-category outcome were consistent with the dichotomous model (online supplemental tables S4 and S5), with the proportional odds assumption supported by the Brant test (p=0.7166). The small number of women who reported that contraceptive decisions were made mainly by their husband/partner (n=76), compared with those reporting joint decision-making (n=1329) or mainly respondent decision-making (n=680), could have resulted in unstable estimates in a multinomial logistic regression model. Consequently, multinomial regression was not undertaken as a sensitivity analysis. Furthermore, the study sample was unevenly distributed across the two slum areas, with more participants from Viwandani (n=1438, 69%) than Korogocho (n=647, 31%). Therefore, slum stratification analysis was not conducted as part of sensitivity analysis since the imbalance in sample sizes could have yielded imprecise and less reliable estimates, particularly for the smaller stratum, Korogocho.

Individual-based model results

Figure 2 represents the simulated trends in mCPR among women in union, women not in union and all women from 2012 to 2040. Each line in the graph is accompanied by a shaded band, which represents CIs or uncertainty bounds, reflecting the variability or simulation range around the estimates. Women in union consistently show the highest mCPR, starting at around 54% in 2012 and rising steadily to about 71% by 2040, while women not in union show lower contraceptive use, starting at 43% in 2012 and increasing only slightly to about 47% by 2040. The average mCPR across all women, lies between the two subgroups with 50% in 2012 and rises to about 54% by 2040. To further explore the dynamics among women in union, we modelled three distinct decision-making scenarios: contraceptive decisions made independently by the woman, by the spouse alone, or jointly by both partners, as shown in figure 3. The aim of the decision-making simulation was to assess how each decision-making category contributes to the overall mCPR among women in union. Therefore, decision-making categories indicate the relative contribution of each decision-making type to the overall mCPR among women in union.

Figure 2. Projected trends in mCPR among all women, women in union and women not in union, 2012–2040. Shaded bands represent 95% simulation uncertainty intervals derived from multiple simulation runs per draw. mCPR, modern contraceptive prevalence rate.

Figure 3. Trends in mCPR among women in union, disaggregated by decision-making category, 2012–2040. Shaded bands represent 95% simulation uncertainty intervals derived from multiple simulation runs per draw. mCPR, modern contraceptive prevalence rate.

The results from figure 3 illustrate the projected trends in mCPR from 2012 to 2040 among women in union, disaggregated by decision-making dynamics: joint decision-making, woman-only decision-making and spouse/partner-only decision-making. Each trend represents the mCPR among women in the group. For example, the joint decision-making mCPR indicates the mCPR among women in union who make their decision to use or not use contraceptives jointly. The simulation indicates that mCPR for women in union making their contraceptive decision jointly with their partners is higher throughout the simulation period compared with mCPR for women making decisions independently or partners making decisions. The prevalence of mCPR under joint decision-making is progressively increasing from 59% in 2012, reaching approximately 77% by 2040, reflecting growing alignment between partners in family planning choices. Women who indicate that their spouse/partner determines all of their contraceptive decisions, on the other hand, have the lowest mCPR, dropping from about 36% in 2012 to around 15% by 2040. This suggests that when men are the sole decision-makers, modern contraceptive use tends to be lower and even deteriorates with time. This pattern suggests a potential impediment to contraception access or acceptance for women with limited autonomy. The woman-only decision-making group has moderate levels of mCPR, beginning around 44% in 2012 and steadily growing to nearly 60% by 2040. This trajectory indicates that when women are empowered to make independent choices, contraceptive uptake improves, though not as strongly as in joint decision-making scenarios.

Figure 4 shows the contributions of various decision-makers to the overall contraceptive use, as well as trends in mCPR among women in union from 2012 to 2040. The total mCPR annually is displayed by the solid black line, and the stacked coloured areas indicate the proportion of that use that may be attributed to decisions made by the woman alone, the spouse alone or both partners. From roughly 54% in 2012 to roughly 71% in 2040, the mCPR increases steadily throughout the simulated period, suggesting improved uptake of modern contraceptives over time. The highest percentage of mCPR is constantly contributed by joint decision-making, which rises over time from 40% in 2012 to almost 52% in 2040, indicating that the most contraceptive use occurs when both partners are involved in the decision-making process. A potential increase in female autonomy over reproductive health choices is shown in the progressive growth in the contribution from women making independent decisions, which goes from 13% in 2012 to roughly 18% in 2040. Conversely, the percentage attributable to spouse-only decision-making remains very low and consistent across the entire period, ranging from about 1% in 2012 to roughly 0.5% by 2040.

Figure 4. The figure presents annual estimates of mCPR (solid black line) among women in union from 2012 to 2040, overlaid on stacked areas representing the absolute contribution of each type of decision-maker: joint decision-making, spouse-only and woman-only. mCPR, modern contraceptive prevalence rate.

Model validation

To validate the simulated results, we compare mCPR for women in union from the TLO-NUHDSS model and local data. Uncertainty quantification was also conducted by configuring the simulation to have multiple runs per each draw and comparing the results. Table 4 compares the results from locally collected data used in this study and results generated from the simulation. In 2012, the mCPR for women in union was 54.8% from the local data and the model predicted 54.3% with narrow CI, 95% CI 52.8 to 55.8, suggesting more precision in the simulation. Across the years (2012–2015), the model closely aligns with local data indicating that the simulation captures the real-world contraceptive use pattern well. The mCPR CIs from the model are consistently narrow, reflecting greater statistical certainty. The results from multiple simulations also closely align with the simulations reported in this study.

Table 4. A comparison of mCPR from local data and model output.

Variable	Local data results			Model results
mCPR for women in union	Year	mCPR rate (%)	95% CI (%)	Year	mCPR rate (%)	95% CI (%)
	2012	54.8	45.3 to 60.4	2012	54.3	52.8 to 55.8
	2013	51.9	44.7 to 59.2	2013	55.2	53.8 to 56.7
	2014	56.1	48.1 to 64.1	2014	56.3	54.9 to 57.7
	2015	56.9	48.7 to 65.2	2015	57.2	55.7 to 58.6

mCRP, modern contraceptive prevalence rate.

Discussion

In this study, the prevalence of joint decision-making in contraceptive use was 63.7%, which is somewhat lower than figures reported in other studies conducted in sub-Saharan Africa. For instance, a multilevel analysis based on the Ethiopian Demographic and Health Survey reported a significantly higher prevalence of 78.8%,⁸ while a similar study in Mozambique found a prevalence of 71.6%.¹⁶ However, the result observed in our study is comparable to that reported in Zambia, where 65.8% of couples were found to jointly decide on contraceptive use.⁹ The relatively lower prevalence in our study may be attributed to the focus on urban slum populations, where socio-economic constraints, limited access to information and gender power dynamics may influence shared decision-making. In contrast, the cited studies relied on nationally representative DHS data, which includes a mix of urban, peri-urban and rural populations and may therefore reflect more favourable averages across diverse contexts. It may also be attributed to different geographical settings.

The analysis from this study indicates that women living in Viwandani slum area are more likely to make joint decisions compared with women living in Korogocho. This may be attributed to differences in socioeconomic status, education and access to health services. Viwandani, which is approximately 7 km from the city and near the industrial zones compared with Korogocho which is about 12 km from the city, may offer more opportunities for exposure to reproductive health programmes, which can foster shared responsibility in contraceptive use.¹⁷ It is also evident that male-headed households are more likely to make joint decisions compared with female-headed households. This could be due to the fact that male headship often reflects conventional gender norms in which reproductive decisions are framed as shared, but still affected by male authority. In contrast, women in female-headed families, even if in union, may exercise greater autonomy or live apart from their spouses, resulting in more independent decision-making. Unemployed women were 1.36 times more likely to engage in joint decision-making on contraceptive use compared with those who were employed. Their financial reliance on their spouses might be a contributing factor to a dynamic where decisions are made together, rather than independently. This is consistent with a multicountry study conducted across 11 East African countries that reported that employed partnered women were 1.24 times more likely to have autonomy in contraceptive decision-making than their unemployed counterparts.¹⁸ Women who did not want another child or were undecided were less likely to make joint decisions on use of family planning with their partners compared with those who want another child. This aligns with findings from Sarnak and Gemmill, who showed that in cases of discordant fertility desires, including when the woman did not want more children or was unsure, these women were often more likely to use contraception covertly, rather than discussing it with their partner.¹⁹ This study found that women who had not discussed their ideal number of children with their spouses were much less likely to make shared contraception decisions compared with women who discussed and desired the same number of children with their partners. This is backed by existing research reporting that not knowing husbands’ fertility preferences is associated with increased odds of using contraceptives covertly, compared with perceiving that husbands wanted the same or fewer children.¹⁹ Additionally, it is supported by research conducted in Nigeria which found lower probability of having joint decision-making among women who did not know their partners’ desires.²⁰ However, religion was not statistically significantly associated with decision-making. Evidence suggests that religion greatly influences the use of contraceptives. A study conducted in western Ethiopia found that Muslim women were less likely to use contraceptives compared with their Orthodox Christian counterparts.²¹ Similarly, a study conducted in northern Ghana reported that switching from traditional religion to Christianity or Islam was associated with higher contraceptive use.²² These findings highlight the role of religious affiliation in shaping contraceptive behaviours. However, while religion appears to influence whether contraceptives are used, it does not show a significant association with joint decision-making on contraceptive use.

The results from the simulation show steadily increasing mCPR between 2012 and 2040 for all women, with consistent differences by union status. The report from KDHS 2014 reported mCPR at 53% among married women and approximately 49% for all women of reproductive age, with lower values among unmarried sexually active women. These closely align with the results from the model which we found that in 2014, the estimated mCPR was at 54.4% for women in union and 52.5% for all women. The 2022 KDHS preliminary report confirms a similar trend, with mCPR at 57% for women in union, which is close to 57.7% from the model, reinforcing the pattern of higher contraceptive use among this group. These small variations, generally within 1–2 percentage points, between simulated and reported data validate the robustness of the model. The convergence of simulation outcomes with empirical trends underscores the utility of the model in exploring policy and programmatic scenarios for family planning in Kenya. The simulation revealed that women who made contraceptive decisions along with their partners had a greater mCPR than those who made decisions independently or with their partners alone. As a result, joint decision-making had a greater impact on overall mCPR among union women than those made entirely by the woman or her spouse. This finding is consistent with studies from Nigeria²³ and Ethiopia,²⁴ which demonstrate that spousal communication is linked to higher usage of modern contraceptives. When both partners participate in decision-making, communication is likely to improve, mutual consent is promoted and shared reproductive intentions are supported, all of which encourage contraceptive use.

This study has various strengths. It uses data from informal settlements to provide useful insights into contraceptive decision-making among women in underprivileged areas. The study disaggregated decision-making into joint, woman-only and spouse-only categories, which provides a more nuanced picture of partnership dynamics and it combined regression with an agent-based microsimulation to explore longer-term implications for policy. Several methodological limitations should, however, be considered when interpreting the findings. First, the design is repeated cross-sectional, so the temporal sequence between the determinants and decision-making cannot be established and the associations reported are not causal; the language used throughout has been framed accordingly. Second, the primary outcome was analysed using complete cases, and although missingness on the outcome was modest and the characteristics of included and excluded women were broadly similar (online supplemental table S2), the possibility of selection bias cannot be fully excluded if data were not missing at random. Third, decision-making was self-reported and may be subject to social desirability and recall bias, and dichotomising an originally three-category outcome discards information; we therefore examined ordinal specifications as sensitivity analyses, which were consistent with the main model. Fourth, important constructs were not available in the dataset, including a household wealth index and a validated measure of women’s decision-making power, as well as gender norms, partner attitudes and other sociocultural factors, so residual confounding is likely; education and employment were retained as partial proxies for socio-economic position, and future studies should incorporate wealth and validated empowerment measures. Fifth, the modest McFadden pseudo R², while typical of behavioural models in public health, indicates that much of the variation in joint decision-making remains unexplained by the measured covariates. Sixth, because the data are drawn from two surveillance settlements (Korogocho and Viwandani) and cover 2012–2015, the findings are representative of women in union in these communities but are not nationally generalisable and may not reflect the most recent national trends; corroboration with newer data would strengthen the evidence. Finally, the simulation results depend on the calibration assumptions and transition parameters of the TLO-NUHDSS model, which were held fixed across scenarios; the scenario-specific contributions should therefore be read as illustrative of the relative importance of each decision-making pattern under the model’s assumptions rather than as precise behavioural forecasts, and parameter and structural uncertainty in the agent-based framework remain. More research should be conducted to determine how gender norms, specific roles and power inequalities act as barriers to collaborative decision-making.

Conclusion

This study describes the association between joint decision-making and modern contraceptive use among women in union residing in Nairobi’s urban informal settlements. With a prevalence of 63.7%, joint decision-making was associated with the highest modern contraceptive use among the decision-making patterns examined, and in the microsimulation, it contributed more to overall mCPR than decisions made solely by women or their partners. Determinants such as male-headed households, unemployment, fertility preferences and partner communication point to the interplay of socio-economic and relational factors in shaping reproductive health choices. Taken together, and recognising that the cross-sectional design does not allow causal inference, the findings are consistent with the possibility that fostering collaborative decision-making between partners could support higher mCPR; the simulation projects a rise in mCPR among women in union from about 54% to 71% by 2040, with the contribution of joint decision-making rising from 40% to 52% under the model’s assumptions. These observations support targeted interventions that promote open dialogue and shared responsibility in family planning, particularly among vulnerable urban populations.

To bridge the gap in contraceptive uptake, policymakers and health practitioners should prioritise programmes that empower couples to engage in joint decision-making, especially in female-headed households and among employed women or those with discordant fertility desires. Community-based initiatives, such as couple counselling and male engagement strategies, could mitigate contraceptive use and improve reproductive health outcomes. Future research should explore the cultural and gender dynamics that influence decision-making autonomy, as well as the long-term impacts of joint decision-making on maternal and child health. By addressing these barriers, Kenya can advance toward its 2030 family planning goals, ensuring equitable access to family planning and fostering sustainable demographic and economic development.

Data availability statement

Data are available upon reasonable request.