Interventions to increase uptake of cervical screening in sub-Saharan Africa: a scoping review using the integrated behavioral model

A medical librarian (JMC) was consulted to develop a comprehensive search strategy, to identify appropriate databases, and to conduct the literature search. We systematically searched electronic databases PubMed/MEDLINE (1946-May 1, 2019), Clarivate Analytics Web of Science (1900-May 1, 2019), Elsevier EMBASE (1947-May 1, 2019), and EBSCO CINAHL (1982-May 1, 2019). The searches utilized controlled vocabulary searching modified for each database, as well as natural or key terms following a structure of “topic/cervical cancer” AND “outcome/screening uptake” AND “context/population.” No language or publication date restrictions were applied. Additional file 1: Appendix 1 details the complete search strategy. Conference proceedings and abstracts, published as supplements to journals or indexed in the electronic databases, were returned in the search and were screened with the same eligibility criteria as peer-reviewed articles.

The search results were compiled and deduplicated using EndNote (version × 8). Then, the deduplicated entries were uploaded to Covidence, an online software for systematic screening of results. Screening was completed in two rounds: title and abstract screening was performed, then full text review for eligibility. Each search result was screened independently by two reviewers (BEL, MJT, CB). If the two reviewers were not in agreement, the result was screened by a third reviewer (BEL, MJT, CB).

Inclusion criteria were studies with an intervention component, conducted in a sub-Saharan African community, and any type of participant such as women, their male partners or other family members, or health care providers. Studies were excluded if they targeted African participants in a non-African setting as were studies measuring attendance of follow-up care for women who had previously attended cervical cancer screening. All intervention types were considered including one-on-one or group educational interventions or counseling, patient reminders, media and awareness-building campaigns, incentivization schemes, and innovative service delivery approaches, which could be set in health facility- or community-based contexts. The primary outcome of interest was uptake of cervical screening, measured as self-reported or medically verified receipt of any type of cervical cancer screening service, as a proportion of the entire study population. Secondary outcomes of interest were cervical cancer or screening knowledge/awareness, willingness/intent to get screened, and cervical cancer or screening-related attitudes/beliefs. Comparisons varied and included no intervention and standard care. Study designs that could demonstrate a change in screening uptake over time or between two groups were included: randomized control trials, observational cohort studies, and before-and-after studies. Studies that described screening program implementation, without screening rates for a comparison group were excluded.

An excel data charting form, created and tested by our team, was used to elicit information on pre-defined variables from the included sources. The form was modified during an iterative data charting process. Data extracted included: study period, location, sample size, aims of the study, methodology, participants/population, intervention, comparison, findings, and author recommendations. In addition to descriptions of the intervention, information was recorded about the intervention development process and the delivery. Data charting for each included study was completed by two reviewers (BEL, MJT, CB). Reviewers used emergent themes to categorize the studies by intervention type. Content analysis was performed, using the IBM as a framework to describe the scope of interventions included, as described in the following section. Each study was reviewed for evidence that the researchers considered IBM constructs during the development of their intervention, as the mechanism through which the intervention is supposed to act, or as a measure or outcome of the intervention study. Presence or absence of IBM constructs was recorded for each study and the compiled construct frequency data is presented in a table. Narrative synthesis is used to elaborate on ways in which the constructs appeared in the included studies. An optional quality assessment of the included studies was not performed, as it is unnecessary for scoping review methodology and did not help us to meet our study objectives [11].

As the purpose of this review is to describe the nature of interventions that have been conducted to increase uptake of cervical cancer, we chose to use a health behavior model to map the breadth of literature, to understand how the interventions might work to create behavior change, and to identify gaps in intervention types and implementation strategies that have been attempted. The IBM was developed by bringing together constructs from the Theory of Planned Behavior, the Theory of Reasoned Action, and other prominent health behavior theories with overlapping constructs [14]. The main premise of the IBM is that the most important determinant of a behavior is intention to perform that behavior, and that intention is influenced by attitudes, perceived norms, and personal agency [14]. Factors such as knowledge and environmental constraints also act on the behavior directly (see Fig. 1). This review uses the IBM to understand cervical screening as a behavior, “willingness to screen” as the precursive intention to perform the screening behavior, and other theoretical constructs to describe the rationale, action, and evaluation of interventions intended to change screening behavior.

The systematic search returned 3704 unique entries, after duplicates were removed. Screening and application of eligibility criteria produced 19 studies for inclusion in the review [15‐33]. Figure 2 shows the results of the study selection process. The majority of studies were published between 2014 and 2019 (n = 15, 78.9%), and were randomized (n = 9, 47.4%) or quasi-experimental studies (n = 5, 26.3%). Studies were set in seven African countries with almost half set in Nigeria (n = 9, 42.9%), followed by South Africa (n = 5, 23.8%) and Kenya (n = 3, 14.3%). While not all studies specified, both urban and rural settings were represented. Minimum age eligibility criteria were used to define target populations; few studies cited World Health Organization or other screening recommendations when defining age criteria. Pap smear or cytology-based cervical screening was assessed most often (n = 6, 31.6%) [16, 17, 21, 26, 31, 33]. A quarter of studies (n = 5, 26.3%) did not specify which screening test was being performed [20, 22, 24, 29, 32]. Multiple screening methods were used in some studies, such as those that compared uptake of HPV self-sampling to facility-based visual inspection screening methods [19, 25, 27, 28]. Screening was offered at workplaces [28], in women’s homes [19, 27, 28], in community spaces [23], and in health facilities [15‐18, 20‐22, 24‐26, 29‐33]. Study characteristics are presented in Table 1.

We found three emergent themes for intervention types: health education interventions (n = 11, 57.9%), economic incentivization interventions (n = 2, 10.5%), and innovative service delivery interventions (n = 6, 31.6%). Descriptions of the population, intervention, comparison, and outcomes (PICO) of included studies, organized by intervention type, can be found in Table 2. The most prevalent type of intervention was health education which was characterized by use of short-term, lecture-based education, peer health educators, multimedia lessons, and behavior change communication and message framing interventions. Educational interventions operated under the rationale that low knowledge and awareness of cervical cancer and screening services are the primary barriers to screening uptake. These interventions were implemented in-person and remotely, through email and text messaging. There were two economic incentivization interventions, which included a lottery-style game that was played by women to randomize their households to receive subsidized cervical screening and cancer treatment in the unlikely event that cancer was detected and a reward program offered by a health insurance plan that incentivized receipt of preventive health services including cervical screening, where women could earn points and prizes for their behavior [26, 30]. Six of the included studies described innovative approaches to delivering cervical screening including: changes to the type of cervical screening provided, the screening location, or the coupling of cervical screening with other services with the intent to make cervical screening more comfortable, convenient, and accessible by removing or mitigating environmental barriers to screening. Cervical screening was integrated with sexual and reproductive health services in two studies [22, 24], and with HPV vaccination for adolescent daughters in a school-based intervention [19]. Multiple studies also drew upon established social ties; interventions targeted male partners [22] and peer groups of female sex workers [24], utilized trusted community health outreach workers and religious leaders to deliver interventions [18, 25, 28], and leveraged the influence of the King’s council in one community [27].

Uptake of cervical screening was most commonly measured as “ever screened” (n = 8, 42.1%) or screened in a specified amount of time, usually during the study period (n = 7, 36.8%). One study created a screening score that accounted for presence of screening behavior, as well as when screening was last received and whether or not a woman was up-to-date with screening recommendations [19]. Some studies used vague language to describe uptake of cervical screening, which made it impossible to clearly deduce the measure of screening [16, 22, 28, 29, 32]. At baseline, screening coverage ranged from 0 to 53.6%, with most studies reporting coverage of less than 10% (n = 12, 63.2%). Following intervention, coverage ranged from 1.7 to 99.2%, with six studies reporting a significant improvement in screening that achieved ≥60% coverage (31.6%). More than a third of the studies reported non-significant increases in screening uptake associated with their intervention (n = 7, 36.8%); all of them were educational interventions [16, 17, 21, 29, 31‐33].

One educational intervention was particularly effective, with a 64.4 percentage point increase in “ever screened” from 3.2% pre-intervention to 67.6% screened post-intervention (p < 0.001) [18]. This study evaluated a house-to-house education program that provided women with information about cervical and breast cancer and offered free screening services at health facilities within walking distance as well as HPV vaccination for eligible children for a fee. Two other educational interventions resulted in more modest, but significant improvements in coverage. One was an SMS behavior change communication intervention, offered with and without additional eVouchers to cover transportation costs to the screening clinic. Women who received the 15 SMS messages over 21 days had three times greater odds of screening during the study period than women in the control group (12.9% uptake vs. 4.3%, adjusted odds ratio [AOR] = 3.0, 95% CI: 1.5–6.2) [20]. Women randomized to receive the eVoucher for transportation in addition to the behavior change communication had almost five times the odds of uptake compared to the control group (18% uptake vs. 4.3%, AOR = 4.7, 95% CI: 2.93–7.44) [20]. Mbachu et al. designed a health education intervention that trained clergy wives in Anglican dioceses to deliver peer health education on cervical cancer to women they worship with. Women attending at least three group sessions reported a 6.8 percentage point increase in “ever screened” from 10.5% pre-intervention to 17.3% post-intervention (p = 0.02) [25].

As a category, the interventions that used innovative approaches to cervical cancer service delivery provided the most compelling results. A pilot RCT with 500 women in Uganda found that use of outreach workers to distribute and collect HPV self-sampling kits from women in their homes and workplaces was highly effective; 99.2% of self-collection samples were returned for HPV testing [28]. The control group, which received outreach worker visits, appointments for VIA screening at the local health facility, and phone call reminders the day before the appointment, also had good uptake (48.4%) [28]. The use of outreach workers was an effective screening implementation strategy and the use of self-sampling as the intervention further increased the uptake of screening behavior. Another study of home-based HPV self-sampling in Nigeria resulted in similar outcomes with 92.5% uptake in the intervention group post-intervention and 56.5% uptake in the control group post-intervention, compared to 1.5 and 1% uptake respectively in each group pre-intervention [27]. One study in rural Kenya offered community-based HPV screening at pop-up tents during a two-week phase of a six-week community health campaign. Women who screened on-the-spot in the tents had 60% uptake compared to the control community where women were referred to their local government health facility for testing, resulting in 37% uptake [23]. Another approach was to offer cervical screening as an add-on service for patients and their partners who were already seeking care in a facility. In a large RCT in urban Zambia, Hewett et al. offered enhanced client-centered counseling, referral to add-on services, and follow-up by phone if the appointment was missed [22]. Additionally, a second intervention group was provided escort services from the original service to the referral service location. A significant increase in the uptake of cervical screening was observed for both intervention arms compared to standard of care (without escort: 22.2% vs. 9.7%, AOR 2.75, 95% CI: 1.94–3.91; with escort: 23.6% vs. 9.7%, AOR = 2.98, 95%CI: 2.10–4.22) [22]. Another integrated services intervention targeted female sex workers in South Africa, Mozambique, and Kenya [24]. Implementation looked different in each site with shared components of increasing access to general health facilities, targeted peer outreach, targeted clinical services, and female sex worker empowerment. Significant increases in uptake were observed in South Africa (31.8 to 60.0%, p = 0.001) and Mozambique (0 to 25.2%) [24]. Lastly, combining cervical screening with a school-based HPV vaccination program in South Africa increased screening among mothers to 68.9%, though the prevalence of “ever screened” was higher at baseline (53.6%) compared to other included studies [19]. Both self-sampling and facility-based screening were offered with this intervention; a more favorable change was observed for the self-collection group with 45.3% of the screening kits being returned for testing [19].

Almost half of the included studies assessed some outcome other than screening uptake (n = 9, 47.4%): six had some measure of awareness (31.6%) [15, 18, 21, 29, 32, 33], five measured knowledge (26.3%) [15, 16, 19, 29, 32], one included an attitude score [16], and four reported “willingness to screen” (21.1%) [15, 25, 29, 33]. No studies used a validated measure for any of the secondary outcomes. Generally, improvements in secondary outcomes were observed, even when cervical screening did not significantly improve (Table 2). Willingness to screen was high before and after interventions, ranging from 63.4 to 89.7% of participants at baseline to 80.8 to 92.3% at follow-up.

Theoretical frameworks were cited by several of the included studies, including the Health Belief Model [20], Pender’s Health Promotion Model [29], Behavior Change Theory [32], Prospect Theory [17], Diffusion of Innovation Theory [25], and the Intertemporal Choice Problem [26, 30]. We analyzed all included studies by applying the IBM and assessing whether IBM constructs were considered during intervention development, as the intervention mechanism, or as a measure/outcome of the intervention (Table 3). Knowledge and skills to perform the behavior was the construct considered most often (n = 13, 68.4%) followed by environmental constraints (n = 10, 52.6%). These constructs align with the intervention categories of educational interventions and innovative service delivery interventions, presented in Table 2. Economic incentivization interventions acted on the construct of instrumental attitude and operated under a premise that individuals must weigh the perceived costs/consequences and benefits of cervical screening and that behavior change can be prompted by increased immediate benefit or reduced cost of screening. Measures of instrumental attitude included individual risk assessments and perceived severity of cervical cancer. In addition to economic incentives, motivational interviewing was used to act on the construct of instrumental attitude, by helping patients to express and overcome individual barriers to screening and to improve self-efficacy [22]. Interventions employing peer or community health education acted on perceived norms (injunctive norms), in addition to providing knowledge to perform the behavior [18, 22, 24, 25]. Other strategies targeting norms included use of culturally-relevant media to disseminate information or demonstrations that modelled cervical cancer screening behavior [15, 16, 31].

Studies assessed experiential attitudes like fears and misconceptions about cervical screening when developing interventions that would be contextually-appropriate to their populations, and measured experiential attitude following interventions through acceptability questions like “was screening straight-forward/comfortable/convenient?” [23, 27]. Participants reported that screening methods were acceptable, indicating the experiential attitudes toward cervical cancer screening were positive overall. They would also recommend screening to friends, indicating positive descriptive norms (99.4 and 99.0% of self-sampling intervention arms) [23]. Salience of behavior could not be measured for women receiving screening for the first time but agreement with the statement that they would screen again in the future and future screening preferences indicate support for salience [23, 27]. Perceived control, a construct of personal agency, was measured by five studies that asked participants to identify barriers to screening [15, 16, 25, 29, 33]. Lack of awareness of screening services was unanimously the greatest perceived barrier to screening pre-intervention or in the control group among the five studies that measured participants’ perceived control and ranged from 41.4 to 94% [15, 16, 25, 29, 30]. For women participating in interventions, there was a shift following the intervention to citing other reasons for not screening which included lack of access or availability of screening services [15, 29], no symptoms/low perceived risk [25, 33], or not liking the test (Pap smear) [15]. Intention to perform the behavior or “willingness to screen” was measured by four of the studies (21.1%). Intention was high, even prior to intervention, with all four studies reporting that more than 70% of women were hypothetically willing to screen [15, 25, 29, 33].

The IBM operates on the premise that behavior is most influenced by the intent to perform that behavior. In this review, we observed very high willingness to screen, even pre-intervention, but that intent was not always well-translated to uptake of cervical cancer screening. Ndikom et al. described an increase from 75.8 to 91.0% in willingness to screen yet no change in actual screening [29]. Furthermore, one of the included studies saw a significant improvement in uptake without any significant change in willingness to screen [25]. Through application of the health behavior model, we can begin to examine how interventions may work and explore other factors that may be disrupting the translation of behavioral intent to behavior. For example, intention may not translate to desired behavior when knowledge or skills to perform the behavior are low or when environmental constraints prevent a person from following through on the intent. Two educational interventions found that after intervention, participants were less likely to identify knowledge or awareness as a barrier to screening but more likely to identify non-availability of services [15, 29]. The reality is that low awareness and availability issues probably need to be addressed concurrently. We suggest more research into health systems change interventions that address infrastructure and human resource challenges, patient-centered healthcare delivery, and reducing “visit burden” for patients and facilities. While these types of interventions act on the environmental constraints construct of the IBM, another theoretical framework may be more appropriate for systems change interventions where individual behavioral intention is not the major determinant of behavior change.

Habit related to the behavior of cervical screening was not considered by any of the studies, likely because the goal of most interventions was to get women to screen for the first time rather than to instill a new habitual health behavior. Two studies assessed intended salience of behavior by asking participants if they would test again in the future and what their preferences would be [23, 27]. Very few interventions considered the element of self-efficacy, perhaps in part because of the focus on environmental constraints beyond a patient’s control, rather than internal factors. Techniques like motivational interviewing, in combination with the removal of environmental constraints may be one way to increase women’s personal agency and should be explored through future research as a method of improving uptake.

One included study involved males, by offering referral services to patients and their partners seeking care at facilities in the study catchment area [22]. This intervention offered cervical screening as an add-on service for all women seeking care and for female partners of all men seeking care. Male partners of females seeking care were also referred for services like HIV testing and counseling and voluntary medical male circumcision. During the intervention development phase, Risi et al. found that men, particularly father-in-laws and husbands, played an important role in decision-making related to Pap smear uptake in South Africa [31]. Finding ways to meaningfully engage with male partners and relatives may therefore be an effective approach to increasing cervical screening in women. A review of barriers to cervical cancer screening in low-and-middle income countries found that lack of family support was the most commonly reported sociocultural/religious barrier [34]. Indeed, control group participants in one of our included studies, set in Ibadan, Nigeria, reported lack of support from husband (35%) and lack of decision-making ability (37%) as barriers to screening [29]. When executed strategically and sensitively, positive male partner engagement has proven valuable for promotion of other health behaviors such as adoption of HIV prevention behavior following couples-based counseling, testing, and services [39‐41]. Male partner involvement has also shown to effect cervical cancer-related behavior change; in Uganda, male partner involvement reduced loss-to-follow-up among women referred for colposcopy [42].

In this review, we considered uptake of any type of cervical screening as the behavior of interest, but a question remains as to whether screening with different methodologies is one behavior or multiple behaviors. Is the behavior of screening with Pap smear truly the same as accepting visual inspection with acetic acid (VIA) or HPV testing? While some behavioral change constructs, like knowledge of cervical cancer, are shared between screening methodologies, attitudes toward each screening method may differ. For example, Adamu et al. found that 38.4% of their intervention group did “not like Pap smear” post-intervention [16]. Interestingly, this figure was much higher than at baseline when only 6.7% of the intervention group and 1.3% of the control group indicated a dislike for the test. In contrast, HPV self-sampling interventions in Nigeria and Uganda where more than 90% of self-sample kits were returned for testing were very favorably responded to by participants not only in terms of screening behavior but also in self-reported attitudes toward the screening procedure [27, 28]. When asked, 83.2% of women in one study indicated a preference for self-sampling in the future, while 9.2% preferred hospital-sampling, and 7.6% had no preference [27]. Among the women with a preference for self-sampling, they liked the test’s comfort (87%), privacy (6.5%), and found it to be less embarrassing (6.5%) [27]. To women, the act of collecting a cervicovaginal swab in the comfort of their own home may be conceptualized very differently than presenting on an exam table and exposing themselves to a health worker for a cervical exam. Future investigation into women’s conceptualization of each type of screening may inform behavior change interventions targeting each type of screening. Additionally, we recommend transparency of screening methods in reporting; one limitation we encountered was an inability to deduce the type of screening in five of the included studies (26.3%) [16, 22, 28, 29, 32].

In this review, screening uptake was highest among all screening methods for HPV testing of self-collected samples. Screen-and-treat recommendations favor use of HPV testing either alone or in sequence with other screening methods when feasible to identify high-risk, HPV-positive women [10]. Of the four studies in this review that used HPV testing, two studies referred HPV-positive women for VIA as a sequential screening method [28] or to determine treatment eligibility [23], while the remaining studies vaguely described follow-up that included inviting HPV-positive women for further investigation and treatment if necessary [19, 27]. One advantage of using HPV testing is that it frees up the screening program staff to spend more time with high-risk patients by triaging out the HPV-negative women. This screening method may, however, be more prone to loss-to-follow-up (LFU) than visual inspection screening methods. In particular, the out-of-facility location of testing possible with the method, the possible extra step added to a screen-and-treat approach, and test results that are not immediately available to health workers and women create opportunity for LFU. The endpoint or outcome of public health significance may therefore not be just the proportion of women who return a sample, but those who are notified and counseled on their results, and the number of women who go on to receive subsequent complementary cervical screening or treatment. Moses et al. utilized home- and workplace-based HPV self-sampling but found that 53.4% of the HPV-positive women in their sample were unable to be reached by phone after three attempts by the community outreach worker and ultimately 45.2% attended a follow-up VIA appointment at the health center [28]. Huchko et al. observed low treatment acquisition of HPV-positive women identified through their community health campaigns (39.2%) but observed an even lower treatment acquisition rate among women who were tested for HPV in facilities (31.5%), demonstrating that the challenge isn’t necessarily about getting women into the clinic [23]. They describe a long time between screening and treatment, 47 days on average with no difference between the study arms, as one area for LFU improvement. Other recommendations include improved follow-up protocols, use of varied result notification methods like text or in-person home-visits, and having women program the caller’s phone number into their phone at time of sample collection so that they will recognize the caller’s number later at time of notification [28]. The “screen-and-treat” approach common in Africa keeps LFU at a minimum by performing the screening, sharing the results, and treating any possible pre-cancerous lesions during the same health visit. Approaches that use HPV testing should be especially aware of LFU and proactively work to mitigate its effects.

This review highlights greater gains in screening uptake following intervention in rural communities compared with urban study settings. Community awareness about cervical cancer and screening and access to screening may be lower in rural communities at baseline. One study found that only 10% of rural women had heard of cervical screening before and “good knowledge” increased from 2 to 70.5% following intervention [15]. The intervention by Chigbu et al., that increased screening from 3.2 to 67.6%, noted that of the women who were screened following intervention, 94.3% were not aware of cervical screening before the intervention [18]. One study included in this review operated in both urban and rural locations and further demonstrates the urban/rural difference in screening; participants in the SMS behavior change communication intervention with eVoucher for transportation had a more pronounced effect in the rural community than urban community (AOR: 8.78 vs. AOR: 4.67) [20]. Dramatic results in communities previously unexposed to information about screening and with underdeveloped health services may not be indicative of sustained increases in uptake or comparable to what is possible for urban sites. Still, screening even just once in a lifetime is beneficial and the WHO recommends increasing population coverage over increasing the number of times an individual woman is screened [10].

This is the first study, to our knowledge, to review literature on interventions to increase uptake of cervical screening in sub-Saharan Africa, a region with cervical cancer epidemiology distinct from other world regions and from other low-and-middle income countries. We apply a health behavior model to understand how interventions might work and identify opportunities for future research on cervical screening. Rigorous methodology for systematic searching, screening, and data charting and use of PRISMA-ScR reporting maximizes the review’s reproducibility.

One limitation of this review is that 14 of 19 studies come from just two countries, Nigeria and South Africa. While we are attempting to draw conclusions at a regional level for sub-Saharan Africa, the lack of diversity in countries may limit the generalizability of findings. Still, a variety of settings and populations were represented from studies within those two countries and one study compared “diagonal” interventions across countries, juxtaposing results from South Africa to those from Mozambique and Kenya [24]. Careful construction of a search strategy that utilized individual country names for all SSA countries mitigated the possibility that South African studies would be returned more often than other countries for search terms like “Africa”. Rather, we believe that studies from these two countries are just more present in the literature. As a U.S.-based research team, we also acknowledge potential bias in the literature sources searched which may not fully capture the breadth of African literature. Missed citations could exist in Africa-specific databases or journals that we did not hand-search.

Our search may have an over-representation of Pap smear screening because of the study designs and locations of included studies. Pap smear may be seen as a gold standard for screening and may therefore be employed by RCTs and quasi-experimental studies. Pap smears may also be more widely available in certain countries like South Africa. Another review of cervical cancer prevention in sub-Saharan Africa found that VIA was the most commonly used secondary screening method [35]. Both VIA and HPV testing have been declared strategies for closing the cancer divide between LMIC and high-income countries because of their cost-effectiveness and feasibility [43] so the over-representation of Pap smear testing in our study may limit the generalizability and significance of our findings. Lastly, heterogeneity of screening methods, screening uptake measures, and lack of validated measures for secondary outcomes made it difficult to compare results across studies. Work to validate cervical cancer awareness, knowledge, attitude, belief, and personal agency measures should be undertaken in sub-Saharan Africa.