Family planning plays an important role in reproductive rights and the protection of maternal health, yet is underutilized in many parts of Sub-Saharan Africa. In the region, approximately 20% of married women currently use a modern method of contraception, and an average of one in four married women want to space or limit pregnancy, but are not using a modern method.1 Family planning programs in developing countries have worked to increase the number of service delivery points and expand into remote areas;2 however, to be effective, they must also address issues of service quality. Many family planning experts hypothesize that low-quality family planning services are a barrier to high contraceptive prevalence.3–6

Substantial increases in contraceptive use and corresponding declines in fertility have been consistently observed throughout the developing world, although such changes have been more limited in Sub-Saharan Africa than in other developing regions.7 In Kenya, the prevalence of contraceptive use has increased since the 1970s: At that time, only 7% of married women of reproductive age used a family planning method, but that figure grew to nearly 40% by 1998.8 At the same time, the total fertility rate (TFR) dropped, from more than eight children per woman in the early 1970s to approximately five by the late 1990s. Since then, however, progress has slowed: In 2013, Kenya’s contraceptive prevalence among married women was 46%, and the TFR was 4–5 children.8,9

Motivated by the hypothesis that improvements in service quality may facilitate greater contraceptive use, two prior large-scale, facility-level, quantitative studies have assessed the quality of family planning service delivery in health care facilities in Kenya. Kenya’s first nationwide assessment of family planning quality, conducted in 1989 among 99 randomly selected public facilities, found several deficiencies in service quality, including restricted choice of methods, little information on management of side effects, failure on the part of providers to ascertain the client’s reproductive goals and a dearth of mechanisms in place to ensure follow-up.10 A subsequent study that focused on public facilities in Nairobi was conducted in 1993;11 its results did not differ markedly from those of the national study.

Other studies have described the quality of family planning service delivery in Kenya, but have been unable to assess the relationship between quality of care and current contraceptive use.10,12–17 Such an assessment typically requires both facility- and individual-level data, as well as the ability to link women to the facility where they reported receiving or are assumed to have received services. A few studies have taken this type of multilevel approach to assessing the relationship between family planning service quality and contraceptive prevalence or continuation, with mixed results. Studies conducted in Peru, Egypt and Morocco in the late 1980s and early 1990s found little to no effect of quality on method use or continuation;18–20 however, studies conducted between 1991 and 2003 in Tanzania, Egypt, the Philippines and Nepal found moderate to strong associations between service quality and use.21–24 The conflicting results in these studies may be explained by variations in how quality was defined and measured. A 1988 study in Egypt that found no relationship between quality and continued method use measured quality solely through interviews with staff and defined quality by the number of trained personnel, number of available methods and presence of female doctors;18 in contrast, a 2003 study in Egypt that found an association between the two measured quality with a variety of tools, including provider and client interviews and observations, and created a quality of care index.23

Studies that fail to find a notable link between quality and use may accurately reflect the absence of a strong relationship between the two. It is possible, however, that such null findings are a result of measurement error, as suggested by the wide variation in approaches to measuring and defining family planning service quality. It should also be noted that prior multilevel studies linked women to a facility based on their location. This linking strategy assumes that women seek services at health care facilities within the geographic cluster to which they have been assigned. Some believe this assumption should not be made and suggest, instead, that women participating in demographic surveys should be asked to report the facility where they seek services, so that correct exposure classification will be ensured.25

This study aimed to investigate the relationship between family planning service quality and current contraceptive use among women in urban Kenya; our ability to link women’s contraceptive use to family planning service quality at the health care facility from which they report receiving care addresses an important research gap. This linking strategy is based on the hypothesis that the quality of family planning service delivery at the facility where a woman reports actually receiving services will affect her decision to use contraceptives—that is, that her direct experience at a facility is more important to whether she uses contraceptives than the quality of the facility closest to her or the average level of quality among facilities in her geographic area. Because urban populations in Africa are expected to double between 2000 and 2030,26 a focus on urban women is timely. We hypothesize that women attending facilities with high quality services will be more likely than those attending facilities with poor quality services to be using modern contraceptives. It is also possible that the relationship between high-quality services and use of modern contraceptives will be stronger among some demographic subgroups, such as younger or less educated women, because they have fewer financial resources and less knowledge than other women to compensate for low-quality services.

METHODS

Data

This study utilized data from the Measurement, Learning & Evaluation (MLE) Project, a six-year endeavor implemented by the Carolina Population Center to evaluate the Urban Reproductive Health Initiative. The Urban Reproductive Health Initiative is a five-year project funded in 2009 by the Bill and Melinda Gates Foundation to increase contraceptive prevalence in selected urban areas of Kenya, Nigeria, Senegal and Uttar Pradesh, India; Tupange, the country-level program in Kenya, is led by Jhpiego, an international health organization.

The MLE/Tupange study covered five urban areas in Kenya; it collected baseline data on individual women between September and November 2010, and on service delivery points between August and November 2011. For this multilevel analysis, we measured service quality at the facility level and the contraceptive use outcome at the individual level. Although service quality data were collected up to one year after outcome data, we do not believe that any meaningful changes in family planning service quality occurred between September 2010 and November 2011, as Tupange did not implement facility-level quality improvement activities until after all baseline data were collected.

•Individuals. To obtain a representative sample of women from each city, the MLE/Tupange study used a multistage sampling design in which government census enumeration areas served as primary sampling units. Within each selected unit, a random sample of 30 households was selected for interview, during which a list of usual household residents was obtained. All eligible women aged 15–49 on the list were asked to participate via an informed consent protocol. The response rate for the individual women’s questionnaire was 85%; survey weights were used to account for nonresponse and differentials in selection probability.

Respondents were asked about their demographic characteristics, current contraceptive use and fertility desires. In addition, respondents answered questions about the source of their current contraceptive method, and about maternal and child health, vaccination and HIV services currently used; these data were used to link women in the individual-level survey to a facility where they had recently received health care services.

Of the 8,932 women who completed the questionnaire, we excluded 626 from our analysis because they reported being currently pregnant or unable to become pregnant (for reasons such as menopause or hysterectomy), and thus were not in need of contraception. Another 762 women were excluded because they reported wanting to become pregnant at the time of the interview. Last, we excluded 1,871 women because they reported not having received any type of health care service at a facility, and thus were not relevant to our examination of service quality. Overall, 5,673 women were eligible to be included in our analytic sample.

•Facilities. The MLE/Tupange study attempted to collect data from 286 public and private service delivery points, including hospitals, health centers and clinics that offer family planning or maternal and child health services. The selected facilities included those in which the Tupange initiative planned to implement quality improvement activities, as well as those identified by women in the individual survey as locations where they sought family planning services. Of the 286 selected facilities, two were unable to participate because they lacked available staff and five refused; the overall participation rate was 98%. Nineteen participating facilities were excluded from our analysis because they did not provide family planning services; thus, our final sample size was 260. These facilities represent approximately 44% of all operational health care facilities with family planning provision in the five study cities. According to the Kenya Master Health Facility List, approximately 60% of all operational hospitals with family planning services were included, and more than half of the excluded facilities were smaller, private-sector facilities.27

Three types of facility-level data were collected: facility audits, provider interviews and client exit interviews. Facility audits consisted of an interview with a facility manager, who also provided access to client logs and commodity supplies; one audit was conducted at each of the 260 facilities. Audits collected data on training and experience profiles of staff, services provided, integration of available services, and the provision and availability of each of 12 types of family planning methods. They also checked for adequacy of storage and standard operating procedures, and the presence of such basic items as sterile equipment, electricity, running water, blood pressure cuffs, specula and private exam rooms.

Interviews were conducted with 1–4 service providers at 255 of the 260 participating facilities; for facilities with fewer than four providers, all were interviewed, and for facilities with five or more providers, four were chosen at random. A total of 692 providers were selected; seven did not complete an interview due to lack of available time or refusal, for a participation rate of 99%. Health care providers who gave informed consent were asked questions on such topics as preservice and in-service training, counseling procedures for family planning, integration of family planning with other health care services and quality assurance.

Client exit interviews were conducted with a convenience sample of 4,230 women visiting one of the 152 higher volume facilities for any health service.* Interviews were conducted at each facility for 1–5 days; between one and 44 interviews were conducted at each facility. Women were asked about the service or services they had received during their visit. Our analysis includes data from only the 1,316 clients of 126 facilities who reported their main reason for visiting was to initiate use of a contraceptive method or continue contraceptive use. These women were asked about the number of methods discussed by the provider, wait time, client satisfaction, perceived treatment and information given during the counseling session on topics including side effects, method use and when to return to the facility.

Measures

•Dependent variable. Current modern contraceptive use, our outcome of interest, was an individual-level measure based on women’s baseline reports about which method or methods they or their partner were currently using. For this analysis, we defined modern methods as the condom, the pill, the injectable, the implant, the IUD, sterilization, emergency contraception, spermicides and the lactational amenorrhea method. A small number of participants (5% in the women’s weighted sample) reported using a traditional method (the rhythm method, withdrawal or the standard days method) and were classified as not using a modern method.

•Independent variables. Quality of service exposure classification was based on the Bruce-Jain framework, which consists of six elements: choice of methods, information given to user, provider competence, client-provider relations, continuity or follow-up mechanisms and appropriate constellation of services.28 Choice of methods is defined as the physical availability of a satisfactory selection of methods, as well as by the provider’s willingness to discuss multiple methods and ascertain client preferences.11 Information given to clients refers to women’s receipt of information from their provider to assist with their selection and proper use of a method, and with the management of side effects and warning signs of potential health problems related to the method.29 A competent provider is one who demonstrates adequate technical competence and adherence to medical guidelines and protocols.28 Client-provider relations are the personal or human aspect of service provision, such as respectful treatment and bidirectional communication.4 Continuity or follow-up mechanisms refers to the presence of protocols, such as scheduling of future appointments or home visits, to encourage contraceptive continuity.28 Finally, appropriate constellation of services is the integration of family planning into additional health services, such as child immunizations, postpartum care and HIV-related care, to ensure convenient access to services.29 Each element of quality was measured using facility-level data; which specific questions were used from each survey instrument and information on the coding of these variables are included in the appendix (Appendix Table 1, page 77).

In regard to assessing the relationship between quality and contraceptive use, some researchers have suggested that achieving a high level of service quality may not be realistic in the absence of adequate service infrastructure.30,31 RamaRao notes that program managers have cited deficiencies in the service infrastructure as a key barrier to providing good quality services.32 Therefore, the term “quality” can be expanded to include not only the dynamics of the interaction between provider and client, but also the degree to which facilities are prepared to offer services. For this reason, we included several variables related to facility infrastructure, such as whether the facility has basic items, family planning guidelines and quality assurance measures.

We also included measures of client satisfaction as a proxy for high quality services: overall satisfaction with services, satisfaction with amount of waiting time, satisfaction with the amount of information provided, the client’s intention to use the facility again and client’s statement that she would recommend the facility to others. Data on these measures came from client exit interviews, and thus describe only higher volume facilities.

All variables from the facility audit were coded as binary variables, except for the number of methods provided and not out of stock in the past year, which was coded as a continuous variable (ranges, 0–8). For variables from provider interviews, the proportion of providers at each clinic responding affirmatively to each item was calculated; we then dichotomized clinics by whether they were at or above the sample-wide proportion or below it for each indicator. Relevant quality-related variables from client exit interviews were averaged for each facility and then the average was multiplied by 4, to range from 0 to 4. This was done so that estimated prevalence ratios would reflect the change in contraceptive prevalence associated with a 25 percentage-point increase in that indicator. Client interview variables were entered into the model as continuous variables.

•Covariates. On the basis of our knowledge of their relationship with both quality of care and contraceptive use, we included the following variables as covariates: age, education, religion, marital status and wealth. These covariates were measured at the individual level using data from the women’s questionnaires administered at baseline. In addition, we included a measure of slum residence. The 2009 census sampling frame was used to classify all primary sampling units as predominantly formal (nonslum) or informal (slum); households were classified as being located in a slum if built on land that the government had not allocated for housing and as not being located in a slum if built on land allocated for housing.

Statistical Analysis

After exploring the facility audit instrument and the questionnaires for interviewing family planning providers and clients, we identified a total of 48 variables related to facility-level service quality, infrastructure or client satisfaction. We employed factor analysis and were able to reduce the number of quality-related exposure variables to 35. The following sets of variables were grouped together on the basis of an alpha greater than 0.70 and a Factor 1 Eigenvalue greater than 1.0, which suggests that the observed variables in each group have a similar pattern of response and are appropriately grouped for the purposes of data reduction: method choice, measured by facility audits (number of methods provided, mix of methods provided, number of methods currently available and mix of methods currently available); method choice, measured by client interviews (provider gave information about different family planning methods and provider asked the client about her method of choice); information given, measured by client interviews (provider explained how to use the method, provider talked about possible side effects and provider told client what to do if they have any problems); bidirectional communication, measured by client interviews (provider asked the client if she had any questions, client felt comfortable to ask questions during the visit and provider answered all of the client’s questions); presence of basic items and private exam room, measured by facility audits (are certain items available on a functioning basis and is there a private examination room); and client satisfaction, measured by client interviews (client would use this facility again and would recommend it to others).

We estimated prevalence ratios using binomial regression. The model was stabilized by using the Poisson distribution for the residuals. Each of the 35 exposure variables was entered into a separate model with the same covariates. We accounted for clustering of observations within facilities using robust standard errors. Our presentation of results includes one model of the full sample of women, and one of only those women who linked to a higher volume facility; this was done because client data were collected only at the higher volume facilities.

RESULTS

Descriptive Findings

•Sample of women. A total of 5,673 eligible and consenting women completed the individual women’s questionnaire. Some 3,990 (approximately 70%) could be linked to a facility for which the MLE/Tupange study collected quality-related facility-level data at baseline in 2011; of those, 3,083 were linked to a higher volume facility from which client interview data were collected.

After weighting, the sample included 3,246 women. Fifty-seven percent were aged 20–29, and 55% had at least some secondary education (Table 1, page 72). Most women were Protestant, currently married and had had at least two live births. Seventy percent resided in Nairobi, and 24% resided in a slum. Two-thirds of women (65%) reported currently using a modern contraceptive method, 5% were using a traditional method and 30% reported nonuse. Of women practicing contraception, 45% were using the injectable, 22% were using the pill and 15% reported using a long-acting or permanent method (the IUD, the implant, or female or male sterilization).

To examine whether there were selection effects with respect to the users of facilities included in the baseline survey, we compared background characteristics and method use of women included in our analyses with those excluded because they linked to a facility not included in the MLE baseline facility-level survey. Fifty percent of excluded women were unmarried, and 30% had had no children; those figures were 27% and 10%, respectively, among included women. Among women practicing contraception, 24% of excluded women relied on male condoms, compared with 9% of included women.

•Sample of facilities. One-third of the health care facilities selected for the facility-level baseline survey were public facilities, and the majority (84%) of these public facilities were nonhospital facility types such as health centers and dispensaries (Table 2, page 72). A similar proportion of the private facilities (87%) were facilities such as clinics and maternity homes, which are smaller than hospitals. Overall, facilities employed an average of nine service providers, and an average of 10 clients were interviewed per higher volume facility.

•Quality of care. On average, facilities provided seven contraceptive methods (Table 3, page 73); however, fewer than six were available at the time of the facility audit, and only about four had not been out of stock at some point in the previous year. According to facility supervisors, integration of family planning with child health, postnatal or HIV-related services was fairly widespread (78–90% of facilities). Private exam rooms, running water, electricity and basic items often used in family planning provision were each available at most facilities (79–95%); far fewer facilities could show written copies of national family planning guidelines on demand (52%) or describe any quality assurance measures that were in place (39%).

In interviews, most providers (81%) reported discussing multiple methods with their clients; fewer than half (48%) asked their clients their preferred method. Between 30% and 53% of providers helped clients select a method, explained how to use the method or discussed warning signs of potential method-related problems; however, 81% reported explaining possible side effects of the client’s chosen method. Half of providers had received in-service training on the provision of family planning services—our proxy for technical capacity. Providers reported slightly lower levels of integration of family planning with other services than facility audits showed (70–81%).

In exit interviews, 47% of clients reported receiving information on multiple methods during their visit, and 57% reported being asked about their method of choice. Of clients initiating a contraceptive method or switching methods, 66% reported that their provider had explained proper method use, while fewer said that their provider had helped them select a method (41%). Among all clients, 58% were told about potential side effects and 65% about how to manage problems. Regarding the relationship between providers and clients, 35% of clients reported that their provider had asked about their reproductive goals and 33% said their provider had treated them “very well”; only 21% said that other facility staff had treated them “very well.” Sixty-six percent of clients had been asked by their provider if they had any questions, 91% had felt comfortable asking questions and 79% had had all their questions answered by the provider. The vast majority of clients (93%) had been told when to return to the facility for method resupply. In terms of client satisfaction, more than eight out of 10 clients reported that they had had adequate privacy during their visit, their services would be kept confidential, they had received the right amount of information, and they were satisfied with services overall; fewer (76%), however, were satisfied with the waiting time. Clients nearly universally reported that they would use the same facility again and would recommend it to others.

 

Multivariate Analyses

In our first analytic model, which included facility audit and provider interview measures among the full weighted sample of facilities, only one variable was associated with use of a modern method: Clients of facilities at which providers reported explaining method side effects had an increased likelihood of contraceptive use (prevalence ratio, 1.1; Table 4, page 74). In addition, one aspect of method choice measured by the facility audit—providing a mix of methods that has not been out of stock in the previous year—was marginally significant (1.1).

Our second analytic model was restricted to higher volume facilities and included client interview measures, as well as facility audit and provider interview measures. In this model, facilities’ having a consistently stocked mix of methods was marginally associated with clients’ contraceptive use, and the magnitude of the relationship was slightly larger than in the unrestricted sample (prevalence ratio, 1.2). Method choice was associated with contraceptive use when measured by provider interviews: Women attending facilities at which providers reported asking clients about method preferences were significantly more likely than others to use contraceptives (1.1). Providers’ discussion of method side effects—significant in the full sample—was not significant in the restricted sample.

Finally, three client interview measures were found to be positively associated with contraceptive use in our second analytic model. Women attending facilities at which clients reported receiving a provider’s help with method selection had a 6% greater likelihood of contraceptive use for each 25 percentage-point increase in the indicator (prevalence ratio, 1.1). Women attending facilities at which clients reported their provider treating them “very well” were more likely than others to be using a contraceptive (1.1). And women attending facilities at which clients reported that they would use the facility again or recommend the facility had a greater likelihood of contraceptive use (1.2).

In general, the association between several aspects of service quality and contraceptive use was much stronger for younger women and those who were less educated. The relationship between contraceptive use and feeling well-treated by their provider was strongest among 15–19-year-olds (prevalence ratio, 1.4; Figure 1), but diminished with age so that it was no longer significant for women aged 35 or older. A similar modification was observed for some aspects of quality and education, where the magnitude of effect was strongest among the least educated women. For example, the strength of the relationship between providers’ offering clients a choice of methods and contraceptive use was strongest among women with no education (1.3; not shown), and it diminished with increased educational attainment.

 

DISCUSSION

We found that several indicators of family planning service quality are positively associated with current contraceptive use, including providers inquiring about client’s family planning preferences and assisting them with method selection, and having clients who feel that they were treated “very well.” Surprisingly, provider competence, follow-up mechanisms and integrated services appear not to be associated with current contraceptive use. It is possible that our means of measuring these aspects do not sufficiently capture their true meaning. For example, just because a provider has received in-service training on family planning provision, there is no guarantee that they are more competent in service provision than their peers who have not received such training. In addition, giving clients verbal instructions on when to return for continued contraceptive supplies may not affect the future behavior of clients to the same extent as additional types of reminders such as appointment cards or follow-up phone calls, which may not be standard practice in many parts of Kenya. It may also be the case that facility managers and providers report higher levels of integrated services than actually exist in practice, in an attempt to exaggerate service quality;6,33 such misreporting may attenuate an existing relationship. It is also possible that these aspects of quality have no association with current contraceptive use. Last, facility infrastructure and most aspects of client satisfaction—including privacy issues, the amount of information given, waiting time and overall satisfaction—were unrelated to contraceptive use.

Many of the prevalence ratios from our multivariate analyses were close to the null value (1.00); however, a prevalence ratio of 1.1 or 1.2, although modest as a ratio measure, equates to a 10% or 20% increase in modern contraceptive use, which is substantial. For example, contraceptive prevalence among our sample of urban Kenyan women not trying to become pregnant was 65%; a 20% increase would raise it to 78%.

Prior to this study, the most recent multiregion assessment of family planning service quality in Kenya using the Bruce-Jain framework took place in 1989 among public facilities and identified several areas of quality in need of improvement.10 Because of methodological differences between our study and the earlier study, however, a direct comparison of our findings and those of the earlier study is not possible.

Prior multilevel studies examining the relationship between service quality and contraceptive use have produced mixed results;18–24 some found a positive association between quality and use, as we did in this study.21–24 Comparisons between our findings and prior research are challenging, however, given stark differences in study region, as well as in measurement and definition of quality. The only previous multilevel study conducted in Sub-Saharan Africa (Tanzania) also found an association between the information provided to clients and current contraceptive use;21 however, information was measured by the availability of educational and promotional material, rather than discussion of side effects, method selection or proper method use, making comparisons between the two studies problematic. In addition, the prevalence of current contraceptive use in the sample of women in the Tanzania study was 13%, whereas the prevalence within our sample of women in urban Kenya was 65%; therefore the same relative change in contraceptive prevalence will correspond to very different absolute differences within the two populations.

Our study identifies several aspects of family planning service quality that could be modified to increase contraceptive use within a country with high fertility and high unmet need, demonstrating the large public health importance of these results. In terms of quality improvements, increases in contraceptive prevalence may be most responsive to in-service and preservice training with an increased emphasis on the ability of providers to excel in client treatment, assist with method selection and impart critical information on the potential side effects of selected methods. Our results also suggest the need for more specific measures of providers’ technical competence, as well as more innovative strategies for encouraging contraceptive continuation.

 

Strengths and Limitations

The MLE project is one of the first large-scale surveys to be able to link individual and facility-level data by individual woman rather than by cluster. This allows us to assess the relationship between service quality and contraceptive use without the restrictive assumption that all women in the sample attend the closest facility or the one preferred by the women in their primary sampling unit. To our knowledge, no other population-based study has been able to link individual women to their current health facility. In addition, the MLE project is the first large-scale survey to focus exclusively on urban populations in developing countries, which allows for an in-depth investigation of these rapidly growing populations. Last, this is one of only a handful of studies to consider all six aspects of quality, as well as facility infrastructure, and is the first comprehensive multiregion situation analysis conducted in Kenya since the early 1990s.

This study has some limitations that warrant discussion. Approximately 30% of the eligible women could not be linked to a facility at which the MLE project collected baseline facility-level data and, therefore, had to be excluded from the analysis; these exclusions suggest some bias in the MLE/Tupange study selection of facilities and caution should be used when generalizing results to unmarried and nulliparous women. Aggregated indicators at the facility level may not have represented the experience of an individual client, and provider performance may have varied from client to client. Some women who did not report seeking family planning services were linked to a facility where they reported seeking other health care services, possibly leading to misclassification of their exposure status; however, integration of family planning and other services was widespread among the clinics. The assumption that a woman is affected by quality of care at a clinic she is known to have attended—even if she is not known to have received family planning services there—is stronger than the common assumption in similar studies that a woman is affected by quality of care at proximal facilities that she is not known to have visited. Providers may have given an inaccurate report of their service delivery behaviors in an effort to portray their performance in a positive light; similarly, client responses may have been influenced by a desire to please the interviewer or protect themselves from retribution from facility staff, or by a cultural reluctance to provide negative information. Given the large number of quality variables included, we would expect one or two spurious significant results at an alpha level of five percent. Last, for some women, data on exposure were collected up to a year after outcome data were collected; however, we do not expect that quality changed meaningfully during the period and, therefore, our results should not have been substantially biased.

CONCLUSIONS

The results of this analysis support the concept of promoting facility-level improvements in the delivery of contraceptive services, especially with respect to discussion of clients’ method preferences, counseling on side effects, assistance with clients’ method selection and client treatment. Encouraging positive and informative interactions between providers and clients has the potential to increase contraceptive use in this region of high unmet need.