Use of long‐acting reversible contraception in a cluster‐random sample of female sex workers in Kenya

To assess correlates of long‐acting reversible contraceptive (LARC) use, and explore patterns of LARC use among female sex workers (FSWs) in Kenya.

male partners. 1 LARCs are considered safe for nulliparous women and women with multiple sexual partners, provided that those with a very high risk of sexually transmitted infections (STIs) receive STI screening or treatment at the time of IUD insertion. [3][4][5] The use of LARCs among FSWs is low in many countries, 6 particularly in sub-Saharan Africa, where prevalence is reported as less than 5% in most studies. 1,7 Use of IUDs among FSWs varies considerably between regions, with higher levels in parts of Latin America and Asia. 8,9 Access to family planning services has increased in parts of Africa in recent decades. 10 In Kenya, LARCs are offered at low or no cost in many public health facilities, and are used by approximately 22% of married women. 11 Uptake by FSWs, however, is unknown. Peer-based HIV-prevention programs for FSWs are common in HIV-endemic countries, but they seldom offer family planning or other sexual or reproductive health services. 12,13 Misconceptions and limited knowledge about contraception have further limited access. 1 The WHISPER or SHOUT study evaluated the impact of a mobile phone intervention on knowledge and attitudes to contraception (with a focus on LARCs) and on unintended pregnancy rates in a population of FSWs. 14 Using baseline data from that trial, the aim of the present study was to assess the prevalence and correlates of LARC use, and explore patterns of use among FSWs in Mombasa, Kenya.

| MATERIAL AND METHODS
The present study analyzed data collected in the WHISPER or SHOUT trial 14  The WHISPER or SHOUT study recruited women aged 16-34 years who self-reported sex work in the past 6 months, had a negative urine pregnancy test, and were not planning a pregnancy for the next 12 months. Peer educators recruited the women from sex work venues such as bars and hotels by using two-stage clusterrandom sampling. First, 102 sex work venues (clusters) were randomly selected from a sampling frame of mapped venues. 15 The probability of a venue being selected was proportionate to the estimated number of FSWs at that venue. Next, peer educators consecutively recruited FSWs from the selected venues, aiming for 10 women from each.
Additional venues were approached until at least 860 women were recruited (the target sample size).
After providing written informed consent, participants completed a clinical assessment, point-of-care testing for HIV, and a structured interviewer-administered questionnaire. Data were collected on electronic tablets using REDCap electronic data capture tools hosted at the Burnet Institute (Vanderbilt University, Nashville, TN, USA). 16 The outcome of interest, LARC use, was defined as self-reported current use of either contraceptive implants or IUDs. Highly effective contraception methods were defined as implant, IUD, injection, oral contraceptive pill, and permanent contraception methods (those with at least 90% typical use efficacy 17 ). Full details of the study measures and variable categories are provided in Supplementary File S1.
Knowledge about family planning was classified as high if participants answered at least five of six true-or-false statements correctly. They were considered to have a positive attitude to family planning if they agreed with at least three of four attitude statements. Self-efficacy and stigma were both measured on a 10-item scale, 18,19 each rated between one and four, with four representing greater self-efficacy or stigma. Two additional items measured contraception-specific selfefficacy, defined as high if participants agreed with both statements.

| RESULTS
Among 1728 women invited to participate in the study, 1432 (82.8%) expressed an interest in participating; of these, 120 (8.4%) did not attend screening and 430 (30.0%) were deemed ineligible. The main reasons for ineligibility were age (n=119, 27.7%) and not owning a mobile phone (n=105, 24.4%) (Fig. 1). In total, 882 eligible women were enrolled from 93 venues. Three women were subsequently excluded from the analysis because they did not answer the questions on contraceptive use, resulting in a sample size of 879 women for the analysis.

| DISCUSSION
The present study recruited a large representative sample of FSWs In the present study, gravidity was the strongest independent correlate of LARC use, reflecting similar results in non-sex-worker populations. 23 This may be because women decide to use longer-acting methods after completing their family or experiencing unintended pregnancy. However, it may also reflect an enduring assumption that LARCs are inappropriate for nulliparous women. 21 Unexpectedly, younger age and lower education were independently associated with LARC use. Younger, less educated women may experience greater difficulty in returning to a clinic for shortacting methods, making LARCs more convenient. 24 An association between use of any contraceptive by FSWs and older age has been T A B L E 2 Contraceptive use characteristics of the sample population. Knowledge and positive attitude to family planning were correlates of LARC use, consistent with findings in other populations. 26 Social norms and contraceptive self-efficacy may lie on the same causal pathway as positive attitude, or may measure the same underlying construct. Education about LARCs has been found to improve attitude and uptake. 21 In the present sample, these individual factors had greater influence on LARC use than structural factors such as the presence of a boyfriend or husband, sex-work-related stigma, and violence. This is surprising given the known influence of structural determinants on sexual health risks. 2,25,27 Structural determinants may have a greater influence on use of condoms and other user-dependent methods than on LARC use.
Adverse effects were experienced by most women for all highly effective contraceptive methods and seemed to be more common than reported elsewhere. 28,29 The rate of LARC discontinuation was high, but the duration of use was not known, preventing a comparison with other studies. There was a lower rate of cessation of implants as compared with IUDs, pills, or injections. 30  population. 14,31 Future studies should investigate whether hormonal IUDs, with their tendency to suppress bleeding, would have a higher uptake than copper IUDs. 32 It would be particularly interesting to determine whether negative perceptions of the copper IUD are transferred to the hormonal one. Targeting additional resources at raising IUD uptake might help to overcome these barriers. It is possible, however, that such efforts might not raise uptake and that the method has low acceptability in this setting. If that is the case, then it may be better to target programmatic resources to other family planning priorities.
Pelvic pain was a frequently reported adverse effect and is also a symptom of cervicitis and pelvic inflammatory disease. The long-held misconception that IUDs cause pelvic inflammatory disease 21 may lead to their unnecessary removal, when in fact it is safe to leave them in situ while concurrent STIs are treated. 4 Quality education and counseling on the benefits and adverse effects of LARCs can improve uptake and continuation rates 21,23 by managing expectations, countering common myths, and providing reassurance on the safety of bleeding disturbances. 32 However, counseling is likely to be insufficient or incomplete in many settings. 30 While Kenya has clear guidelines on contraceptive counseling, 5 one study noted that only 60% of women were counseled on adverse effects when they obtained contraception. 11 Further work is required to determine how guidelines are applied in practice, particularly for sex workers who are subject to discrimination by health workers. 33 Research in South Africa has indicated that, to improve uptake, LARCs need to be available from a wider range of trained service providers, including mobile outreach clinics for harder-to-reach populations such as FSWs, and counseling should be reoriented to emphasize LARCs as a "first-line" contraceptive method. 34 Only 20% of private facilities in Kenya supply LARCs, whereas more than 65% supply other methods 20 -an observation reflected in the present data. Sex worker drop-incenters supplied very few contraceptives (other than male condoms).
This highlights a missed opportunity for these acceptable and widely used centers 33 to improve access to all methods including LARCs.
The study has some important limitations. The data were collected by self-report, increasing the risk of recall bias and social desirability bias; however, it would not be possible or practicable to obtain such personal data by other means. Age was an inclusion criterion, so the results cannot be extrapolated to all ages. There are also limitations around the measurement of pregnancy intention, which may affect the reliability of these data. Some participants may not have intended to get pregnant, but nonetheless desired pregnancy for different reasons. FSWs often have mixed pregnancy intentions depending on their partners, so they must rely on shortacting methods with all partners except the desired father. 35 They may also prefer not to disclose a true intention owing to the stigma surrounding sex work and motherhood.
Interpretation of the analysis is also limited by the cross-sectional design. Correlates such as knowledge may follow rather than precede LARC use. Other variables that might be associated with the outcome were not included; for example, stigma from health workers may be an important structural determinant. 33 Because there were very few current users of IUDs, the results of the regression were dominated by implant users. The low number of IUD users also precluded further examination of their adverse effects, patterns of use, and removal. The analysis of adverse effects had some limitations: there were no data on duration of bleeding, which is a predictor of cessation, 30 and prolonged bleeding may have been instead reported as heavy or irregular.
In conclusion, despite the multiple sexual risks and difficul-