i JOMO KENYATTA UNIVERSITY OF AGRICULTURE AND TECHNOLOGY COLLEGE OF PURE AND APPLIED SCIENCES SCHOOL OF MATHEMATICAL AND PHYSICAL SCIENCES DEPARTMENT OF STATISTICS AND ACTUARIAL SCIENCE BACHELOR OF SCIENCE IN BIOSTATISTICS Bridget Nyaguthii Waineina - SCM 224-0563/2019 John Segera Oyaro - SCM 224-0588/2019 Glen Washington Ochieng - SCM 224-0589/2019 Collete Chepkemoi Kanja - SCM 224-0631/2019 TOPIC: FACTORS ASSOCIATED WITH TIMES TO CONTRACEPTIVE DISCONTINUATION IN KENYA; A SURVIVAL ANALYSIS STUDY DECLARATION ii We declare that this proposal is our original work and has not been submitted elsewhere for examination, award of a degree or publication. Where other people’s work has been used, this has properly been acknowledged and referenced. We have not sought or used the services of any professional agencies to produce this work. iii DEDICATION We would like to dedicate this research paper to our parents and our siblings. Their unconditional support has enabled us to reach this far. We also like to dedicate this to our close friends. Their support was immeasurable. iv ACKNOWLEDGEMENT We acknowledge the Performance Monitoring Action project for responding to our request and providing us with the dataset we have based our proposed research on. We are grateful to our supervisor Lecturer. Caroline Mugo for her great support and guidance in the whole process of this research project development. To Mr. Dennis Ogot, we also want to acknowledge your efforts in guiding the entire class in what is expected of us in terms of coming up with a project. Our lovely parents for their unconditional love and support. Above all, God for sustaining us. v Table of Contents DECLARATION ................................................................................................................. i DEDICATION ................................................................................................................... iii ACKNOWLEDGEMENT ................................................................................................. iv LIST OF TABLES ............................................................................................................ vii LIST OF FIGURES ......................................................................................................... viii List of Nomenclatures. ....................................................................................................... ix Abstract ............................................................................................................................... x CHAPTER ONE ................................................................................................................. 1 INTRODUCTION .............................................................................................................. 1 1.1 Background of the study ........................................................................................... 1 1.2 Statement of problem ................................................................................................ 3 1.3 Objectives of the study .............................................................................................. 3 1.3.1 Main objective .................................................................................................... 3 1.3.2 Specific objectives .............................................................................................. 3 1.4 Significance of the study ........................................................................................... 3 1.5 Assumptions of the Study ......................................................................................... 4 1.6 Strengths and limitations of the study ....................................................................... 4 1.6.1 Limitations of the study ...................................................................................... 4 1.6.2 Strengths of the study ......................................................................................... 4 CHAPTER TWO ................................................................................................................ 5 CHAPTER THREE ............................................................................................................ 9 3.1 Introduction ............................................................................................................... 9 3.2 Study Design ............................................................................................................. 9 3.3 Data collection methods .......................................................................................... 10 3.4 Data source .............................................................................................................. 10 3.5 Study setting ............................................................................................................ 11 3.6 Inclusion and exclusion criteria............................................................................... 11 3.7 Variables.................................................................................................................. 11 3.7.1 Outcome variable .............................................................................................. 11 vi 3.7.2 Explanatory variables ....................................................................................... 11 3.8 Data Extraction........................................................................................................ 12 3.9 Data Analysis .......................................................................................................... 12 3.9.1 Generation of events (time to discontinuation) ................................................ 12 3.9.2 Prevalence of contraceptive discontinuation. ................................................... 14 3.9.3 Significance in levels of covariates .................................................................. 15 3.9.3.1 Log Rank Test ............................................................................................... 15 3.9.4 Determinants of discontinuation ....................................................................... 16 3.10 Ethical approval..................................................................................................... 19 3.11 Statistical analysis ................................................................................................. 20 CHAPTER FOUR ............................................................................................................. 21 4.1 Results ..................................................................................................................... 21 4.1.1 Generation of times to event (discontinuation). ............................................... 21 4.1.2 Discontinuation rates ........................................................................................ 29 4.1.3 Reasons for discontinuation.............................................................................. 32 4.1.4 Testing difference in levels of covariates to the survival time. ........................ 33 4.1.5 Determinants to time to discontinuation........................................................... 36 4.2 Discussion ............................................................................................................... 40 CHAPTER FIVE .............................................................................................................. 43 5.1 Summary ................................................................................................................. 43 5.3 Recommendations ................................................................................................... 46 References ......................................................................................................................... 47 Appendix ........................................................................................................................... 54 Analytic Code ................................................................................................................... 55 vii LIST OF TABLES Table 1: Discontinuation rates across method types ..................................................................... 30 Table 2 : Distribution of contraceptive discontinuation episodes across covariates at 12, 24 and 36 months .......................................................................................................................... 30 Table 3: Results for log rank test .................................................................................................. 34 Table 4: Factors influencing discontinuation after 36 months duration of contraceptive use ................................................................................................................................................. 36 Table 5: Significant variables associated with discontinuation for the nested Cox proportional model ........................................................................................................................ 39 viii LIST OF FIGURES Figure 1 : Histogram of times to event. ............................................................................ 21 Figure 2 : Stratification times to event by education level ............................................... 22 Figure 3 : Stratification times to event by marital status. ................................................. 23 Figure 4 : Stratification times to event by county. ............................................................ 24 Figure 5 : Stratification times to event by Age group. ...................................................... 25 Figure 6 : Stratification times to event by Wealth. ........................................................... 26 Figure 7 : Stratification times to event by Contraceptive method used. ........................... 27 Figure 8 : Stratification times to event by residence ......................................................... 29 Figure 9 : Percentage distribution of discontinuation by type of discontinuation for 12, 24 and 36 months ................................................................................................................... 33 ix List of Nomenclatures. Parameters Bi The i-th covariate Ei Total number of expected events in group i I ij Cumulative probability of not discontinuing in the i-th month for reason j L 0 Likelihood of the null model L 1 Likelihood of the full model Oi Total numbers of observed events in group i Q 12,j Cumulative probability of discontinuing by 12 months S 0 (t) Baseline survival function S(t|X) Survival function conditional on X Xi Covariate associated with the i-th person d ij Number of episodes in the i-th month e i Total number of episodes that reached the i-th duration h 0 (t) Baseline hazard function h(t) Hazard function at time t q ij, Discontinuation rate for the i-th month since start of episode and j-th reason for discontinuation x Abstract Contraceptive discontinuation is defined as the halt of a contraceptive usage that had priorly been in application within a year from its initial adoption for reasons not associated with pregnancy. This phenomenon is particularly derogative to family planning programs and additionally among the chief causes of unwanted pregnancies and thus this report aims to further analyze the factors associated with the discontinuation of contraceptive methods. Contraceptive discontinuation for reasons other than the desire for pregnancy is associated with a high rate of unintended pregnancies leading to unsafe abortions and adverse effects on maternal, neonatal and child health outcomes. Knowledge of the factors that may lead to contraceptive discontinuation remains incomplete and a better understanding is hampered by the lack of a comprehensive framework that acknowledges the multiple and complex reasons that influence the decision to switch methods or stop using one altogether. To carry out the study, secondary data from Performance Monitoring for Action was utilized. The study participants used for this study were a sample of women who were between the ages of 15 to 49 in Kenya with calendar data and questionnaires were the main data collection tool. To accomplish our main objective, survival analysis methods were applied and a Cox proportional hazard model was fitted. There were 1054 discontinuation episodes in total. Pill users had the highest overall propensity to discontinue at all periods followed by users of injectables. Users of the Intrauterine device were least likely to discontinue use. Women in the age groups 20-24 and 25-29 had a higher discontinuation rate compared to other age groups. Women with a secondary education were observed to be more likely to discontinue across all the three periods and those in rural regions also had higher discontinuation rates. The survival analysis findings implicated that the overall discontinuation rates were 35% (12 months), 47% (24 months) and 51% at (36 months). Wanting a more effecting method accounted for about 25% of discontinuations and side effects accounted for 23% of discontinuations, whilst injection and pill recorded the highest rates. Type of method emerged as a predictor of discontinuation at 36 months with the following hazard ratio at 95% confidence interval [CI]; intrauterine device (HR = 0.168, CI = 0.074-0.382), injection (HR = 0.670, 95% CI = 0.548-0.820), implants (HR = 0.180, 95% CI = 0.139-0.234). Age (30+ years) displayed influence: 30-34 years (HR = 0.721, 95% CI = 0.568-0915), 35-39 years (HR = 0.584, 95% CI = 0.445-0.766), 40-44 years (HR = 0.601, 95% CI = 0.434-0.833) and 45-49 years (HR = 0.610, 95% CI = 0.395-0.944). Contraceptive discontinuation are associated with age categories and an individual’s method of choice. The study recommends that contraceptives should be provided to women of different age groups to ensure effective use of contraception methods. In doing so, the gap created by contraceptive discontinuation will reduce. There is a need for greater education about contraceptive techniques, including their advantages and disadvantages. Therefore, the issue of contraceptive discontinuation should be addressed in family planning programs intended to prevent unintended pregnancy. This could be resolved by increasing access to a variety of contraceptive techniques and providing greater counseling regarding the side effects of those methods. Keywords: Factors, contraceptive discontinuation, contraceptive methods, health outcomes, family planning. 1 CHAPTER ONE INTRODUCTION 1.1 Background of the study Contraceptive discontinuation is a major public health issue that leads to unwanted pregnancies and as a consequence, increased prevalence of unsafe abortions. Understanding contraceptive discontinuation is of fundamental interest because it affects women's reproductive health both positively and negatively over the course of their lives, especially younger women whose contraceptive trajectories are dynamic. Studies on discontinuation are important for family planning program managers and policymakers as they may point to several problems within the programs. Contraceptive discontinuation for reasons other than the desire for pregnancy is associated with mistimed and unwanted pregnancies, unwanted births and unsafe abortions which has increased risks of maternal morbidity and mortality (World Health Organization, 2019). Previous research on contraceptive discontinuation has established that a major obstacle to achieving the Sustainable Development Goals (SDG) for maternal and child health is the interruption of contraceptive use for one month or longer by women who had used a method of contraception in the previous 12 months but discontinued it at least once without switching to another method. (Coulibaly et al ., 2020) A large number of sexually active women who want to prevent pregnancy do not use any form of contraception for a variety of reasons, some of which may include; menstrual disruption, method failure, spouse disapproval, menopause, fear of infertility, and a desire for additional children (Hussain et al ., 2016). Existing research from the World Health Organization (WHO) indicates that globally, in 2018, 65% of sexually active women who had an unmet need for contraception discontinued using it out of fear of its side effects and health risks. (Sundaram et al ., 2019) A study by Sedgh et al ., (2016) showed that about 74 million unplanned pregnancies occur annually in developing countries, mostly among women who do not 2 use any kind of contraception. However, consistent and efficient use of current contraception methods could reduce the number of unplanned pregnancies and the number of fatalities of pregnant women. As of 2019, around 270 million sexually active women worldwide did not use contemporary contraception in an effort to prevent pregnancies and/or delay births. (Kantorova et al ., 2020) It has been shown that certain socio-demographic characteristics such as younger women, higher parity, and unmarried or not with a partner are the most likely determinants of discontinuation. It has also been studied that discontinuation rates are higher among short-term method users as compared to long-acting reversible contraceptive (LARC) users, such as intrauterine devices (IUDs) and implants. (Jain et al ., 2017). Reports by Kenya National Bureau of Statistics(2015) stated high discontinuation at 31% within 12 months of initiating contraception, 18% unmet need for family planning and 35% unplanned births. The information from this survey shows that 32% of unintended births in Kenya were due to discontinuation (Jain et al ., 2017). It is therefore important to identify the factors that influence discontinuation so that they can be addressed where possible and the high rates of discontinuation can be minimized. Analysis of the Performance Monitoring for Action (PMA) calendar data derived from the women’s questionnaire is the major source of information on contraceptive discontinuation; it contains robust historical data on episodes of contraceptive use, recalled by women month by month, three years preceding the survey. Even though there are some concerns about recall bias and the validity of calendar data due to its complexity and information, an analysis of studies that used calendar data and other types of questionnaires found that calendar data performs just as well or better in terms of reliability and validity when capturing information on contraceptive use. 3 1.2 Statement of problem Contraceptive discontinuation for reasons other than the desire for pregnancy is associated with a high rate of unintended pregnancies leading to unsafe abortions and adverse effects on maternal, neonatal and child health outcomes. Knowledge of the factors that may lead to contraceptive discontinuation remains incomplete and a better understanding is hampered by the lack of a comprehensive framework that acknowledges the multiple and complex reasons that influence the decision to switch methods or stop using one altogether. This study therefore aims to offer more insights and better knowledge on this major public health issue. 1.3 Objectives of the study 1.3.1 Main objective To determine the factors associated with times to contraceptive discontinuation in Kenya. 1.3.2 Specific objectives 1.To determine time to discontinuation. 2.To determine the prevalence of contraceptive discontinuation. 3.To determine whether there is significant difference between the levels of the covariates 1.4 Significance of the study It is important to examine contraceptive dynamics such as discontinuation so that family planning programs can provide quality services that meet client needs. Furthermore, such data would allow policymakers and program implementers to monitor progress toward achieving international development goals for family planning. 4 1.5 Assumptions of the Study Women are not declared in fecund. 1.6 Strengths and limitations of the study 1.6.1 Limitations of the study 1. Recall bias- since the respondents are required to report their months of contraceptive use from memory, there may be a slight deviation in inaccuracy of data collected. 2. Some variables that are considered by past studies to affect the outcome are found to be missing from the dataset. 1.6.2 Strengths of the study 1. The data used in this study is reliable and is nationally representative for contraceptive discontinuation in Kenya. 2. The use of calendar year episodes of contraceptive history is considered robust from previous studies, and can be used as a corroboration in policy formulation and for designing family planning programs. 5 CHAPTER TWO LITERATURE REVIEW Globally, modern contraceptive usage is reported to be at 45% among women of reproductive age. Modern contraceptive use has also increased in Sub-Saharan Africa from 13% in 1990 to 29% as of 2019 (Nations, U. 2020). Kenya's adoption of long-acting reversible contraceptives (LARC) rose consistently from 4.1% in 2003 to 13.3% in 2014 (Kenya National Bureau of Statistics, 2015). In Saharan Africa, modern contraceptives have a prevalence of 22%, the three most common methods in the region being injectables (39.4%), followed by implants (26.5%) and condoms (17.5%) (Boadu, 2022). In Kenya’s latest report, the prevalence of modern contraceptives is 59.2% (KNBS, 2023) which is an improvement from 53.2% in 2014 (Kenya National Bureau of Statistics, 2015). Despite the increment, from 2019, the contraceptive prevalence in Kenya has plateaued at around 58% (Statista, 2023). This relative stagnation in increment can be attributed to an increased prevalence of contraceptive discontinuation which continues to plague the nation evidenced by (KNBS, 2023), which has the prevalence at 59.2%, an increment of less than 2%. Contraceptives have proved essential in numerous aspects of reproductive health. It is attributed to potentially preventing 25% of under-five mortalities if birth intervals were at least three years, 35% of maternal deaths and 13% of child mortality could be precluded (Eliason et al. , 2014). Additionally, the embrace of contraceptives has helped families in spacing their pregnancies and preventing unintended pregnancies (Kavanaugh & Anderson, 2013). However, there has been a significant decline in fertility rate from 8.0 per woman in the 1970’s (Shapiro, 2015) to 3.4 per woman by 2022 (KNBS, 2023) Contraceptives and abortions are thought to account for about 70% of this plummet (Robinson & Harbison, 1995). Despite the fore-mentioned benefits of contraception, there are still reports of discontinuation following various reasons. Contraceptive discontinuation, from previous studies, has been categorized into three: the desire to become pregnant or other reasons for ceasing usage; method failure, that is, accidental pregnancy while using a 6 contraceptive method; and finally, method-related reasons: which comprise of health related issues such as side effects and dissatisfaction with a method (comprising of reasons such as; the search for a more effective method, partner’s dislike of the method, cost of the contraceptive of choice and the lack of availability) (M. M. Ali & Cleland, 2010). There is low motivation to discontinue contraceptives for the reason of avoiding pregnancy, however, discontinuing because of quality-related reasons has been proven to be more common rather than by the reduced need for contraception (Curtis et al., 2011). According to Rizvi and Irfan, (2012), method use, side effects, age, parity, quality of care, fertility intentions and changes in marital status are reported to be determinants of contraceptive discontinuation. Contrarily, there is less of a consistent relationship between discontinuation and the number of methods provided, socioeconomic factors, residence, partner's disapproval, cost, and accessibility of the method (M. Ali & Cleland, 1999; Blanc et al. , 1999). Additionally, studies conducted have implied that women who have higher levels of education are more inclined to discontinuing their initial method, while further studies indicated that they are more likely to switch rather than cease usage post discontinuation (Alvergne et al., 2017; Ontiri et al., 2021). This may be because they are more educated on contraceptive methods and can make more informed choices. Over the years, the 12-month contraceptive discontinuation rate has been on a decline. According to Ontiri et al ., (2021), rates for all methods have reduced from 37.5% in 2003, to 36.7% in 2009 to 30.5% in 2014. In recent studies, overall discontinuation rates were estimated at 37% (24-months) and 74% (36 months) (Kungu et al., 2022). Intrauterine devices were reported to have the lowest 12-month discontinuation rates at 6.4% in 2014, then implants at 8.0% in 2014. This decline in discontinuation rates was attributed to an increase to those that reported side effects and those seeking a more effective method of contraception (Ontiri et al ., 2021). In Kenya, side effects associated with hormonal contraception have been linked as the main reason for discontinuation (Kenya National Bureau of Statistics, 2015). A study 7 stated that side effects accounted for 40% of discontinuations (Kungu et al ., 2022). Additionally, studies by Castle and Askew, (2015); Blanc et al ., (2002) have associated contraceptive discontinuation to poor quality of care, specifically the lack of proper counseling on side effects. According to the Performance Monitoring and Accountability 2020 survey, among women with unmet need, 28% of them discontinued their method of choice in the last 12 months. According to reports, about 69% of all women were informed about side effects upon seeking counseling on their contraceptive method while 9 out of 10 women were informed on what actions to take upon experiencing side effects (Performance Monitoring for Action, 2021). Apart from women who desire to be pregnant, there are cases reported of women that discontinue despite the fact that they are still in need. Unmet need of contraceptives can be defined based on three indicators: the woman should be fecund; she should be sexually active and not using a method; or reported to want spacing of children or no children at all. Women reported to be in need of contraception by 2014 were 18% in Kenya (Kenya National Bureau of Statistics, 2015). In Kenya, 54% of pregnant women and 38% of non- users reported to discontinue despite still being in need. About 36% were reported to discontinue whilst in need (Kungu et al ., 2022). Injectables and pills were reported to have the highest rates of discontinuation at about 30% (Kungu, 2022). Therefore, it is crucial to reduce unmet need to discourage discontinuation. Developing countries have accounted for 20% of unmet need from discontinuation due to side effects (Jain et al ., 2013). Contraceptive discontinuation has been affiliated to the risk of unintended pregnancies. According to Ayalew et al ., (2022), the prevalence of unwanted pregnancies in Sub Saharan Africa was reported to be 30% and 35% in Kenya (Kenya National Bureau of Statistics, 2015). Also, previous studies are said to account 32% of recent unintended births to discontinuation in Kenya (Jain & Winfrey, 2017). As per a study by Bellizzi et al ., (2020), in Central Asian and in six African countries, over 80% of women with a current unintended pregnancy had not used any contraceptive 8 method in the previous five years. Use of long-acting modern methods remained consistently low across all countries. Among women who last used a traditional method, 83.8% discontinued due to failure, while among women who last used a long-acting modern method, 40.2% discontinued because of side effects.(Bellizzi et al., 2020). According to Dorman et al ., (2018), the introduction of a reversible vas occlusion would reduce unintended pregnancies by 3.5% to 5.2% in the United States, by 3.2% to 5% in South Africa, and by 30.4% to 38% in Nigeria, assuming that only 10% of interested men would take up a novel male method and that users would comprise both switchers (from existing methods) and brand-new users of contraception. Unwanted pregnancies predispose women to induced abortions and other health related complications which can be curbed by improving women’s knowledge on the significance of family planning services. In an attempt to minimize the high rates of discontinuation, it is vital to identify factors that influence discontinuation of contraceptives. Therefore, understanding contraceptive discontinuation patterns, fertility related behaviors, calculating unmet need and comprehending the implications that follow discontinuation are essential in the field of reproductive health and need to be addressed where possible. The main objective of this study is to determine the prevalence of contraceptive discontinuation and the factors associated with contraceptive discontinuation. 9 CHAPTER THREE METHODOLOGY 3.1 Introduction This section identifies the techniques and procedures adopted in conducting this study. It focuses on the following sectors: study design, data collection techniques, data source, potential variables, procedures carried out and the study model. Our study used secondary data sourced by the Performance Monitoring Action project (PMA). PMA is a study that was implemented in the country to promote the use of data for policy and program impact and enhance family planning in the country. Using their data, we focused our study on women of reproductive age (15-49) who have a history of contraceptive use, and explored the use of calendar data to determine time to event durations. The study focuses on women with varying socio-economic backgrounds, using different contraceptive methods in different geographical locations in the country, and use this data to determine factors that are associated with discontinuation and its prevalence in the country across various determining factors. 3.2 Study Design PMA nationally representative data collected in 2021 was used in this analysis. PMA uses a multistage stratified cluster sampling design, with rural urban stratification. In the first stage, counties are selected based on household density and then enumeration areas (EAs) are selected in the second stage using probability proportional to size (PPS). In the 2021 survey, 308 enumeration areas were selected from eleven counties (West Pokot, Bungoma, Kakamega, Siaya, Nandi, Nyamira, Kericho, Kiambu, Nairobi, Kitui and Kilifi). A sample of 35 households were then randomly selected for the list of all the households in each EA. Household and female questionnaires were administered to the participants once consent was granted. More information regarding the PMA sample 10 design can be found on their website (PMA Kenya Phase 3 SOI Narrative | PMA Data, n.d.). 3.3 Data collection methods PMA used questionnaires to collect data. Data based on family planning was collected using the household questionnaire (HQ) and the female questionnaire (FQ). The FQ was administered to women of reproductive age (15-49) who were registered in the household roster. The female questionnaire gathers specific information on education; fertility and fertility preferences; family planning access, choice, and use; quality of family planning services; and exposure to family planning messaging in the media, migration, empowerment, and the impact of the Covid-19 pandemic on household and family planning access. (PMA Kenya Phase 3 SOI Narrative | PMA Data, n.d.) These questionnaires were based on model questionnaires designed by PMA staff at the Bill & Melinda Gates Institute for Population and Reproductive Health at the Johns Hopkins Bloomberg School of Public Health in Baltimore, Maryland, USA; The International Centre for Reproductive Health Kenya; and Kenya National Bureau of Statistics 3.4 Data source Performance Monitoring for Action (PMA) is a national and county level representative data on knowledge, practice, and coverage of family planning services. (Kenya Phase 3 HQFQ Survey | PMA DataLab, 2021). The data collected for this study focuses on female participants with calendar data for their contraceptive use. Data can be sourced from: (Kenya Phase 3 HQFQ Survey | PMA DataLab, 2021) datalab.pmadata.org. https://doi.org/10.34976/ecre-cf28