Exploring women’s preferences for HPV-based cervical cancer screening in South Africa

Abstract

Objective:

To determine preferences for HPV-based cervical cancer screening among South African women.

Methods:

A discrete choice experiment survey was conducted among 298 women who attended two public-sector clinics in South Africa from February 1 to May 31, 2018. Participants choose between hypothetical screening scenarios: method of swab collection (self or provider); timing of treatment (same day or return visit); type of clinic (static or mobile); cost of services (US $0 or US $4); and time spent at the clinic (30, 60, or 120 minutes). A logistic regression model was generated to evaluate the importance of each attribute. A market simulation analysis was performed to determine potential uptake of the various screening strategies.

Results:

The participants expressed strong preferences for free services (β =0.50; p<0.001) and same-day HPV testing and treatment (β =0.40; p<0.001). The market simulation indicated that 83.8% of women would be willing to undergo screening if services were free; the swab was collected by the provider; and treatment was offered at a return visit. Including same-day testing and treatment in the model increased uptake to 96.4%.

Conclusion:

Offering same-day HPV testing and treatment could substantially improve uptake of cervical cancer screening in the South African public healthcare sector.

1. INTRODUCTION

Poverty and inequality remain important determinants of an individual’s risk of premature death. In 2018, a woman living in Sub-Saharan Africa was three times more likely to develop cervical cancer—and six times more likely to die from this disease—than was a woman living in Europe [1]. Substantial disparity by region, race, and socioeconomic status also exists within individual countries. For example, in South Africa, black women have an approximately two-fold increased lifetime risk of developing cervical cancer when compared with their white peers [2]. Access to high-quality cervical cancer screening in the South African public sector is limited, despite the fact that most women obtain healthcare services in this setting. Only one-third of all South African women undergo Papanicolaou smear testing at the recommended intervals; furthermore, approximately 50% of the smears taken are found to be of poor quality and so the test must often be repeated [3]. Among women who require follow-up visits, waiting times for specialist gynecology clinics can be as long as 6–9 months; consequently, there is a high level of attrition from care [4].

Same-day HPV testing and treatment could improve the coverage and effectiveness of cervical cancer screening in countries such as South Africa, which lack a well-organized screening infrastructure [5]. When compared with the Papanicolaou smear test, primary HPV testing offers a substantially more sensitive screening strategy that is less dependent on provider competency [6, 7]. Recommendations from WHO [8] suggest that primary HPV testing models that use the so-called ‘test-and-treat’ paradigm can increase health-system efficiencies over those achieved with traditional Papanicolaou smear-based screening models [8]. Another advantage of primary HPV testing is that women can collect their own swabs [9], which in turn reduces the burden on healthcare providers and potentially encourages women not currently engaged in healthcare to undergo screening [10, 11].

The discrete choice experiment (DCE) survey is a technique that has been widely used to guide public-health policy [12, 13], including strategies for cervical cancer screening [14]. To our knowledge, a DCE survey has not previously been used to inform decisions about HPV self-testing and same-day testing and treatment. DCEs are based on the principle that users of healthcare services are willing to accept trade-offs between key attributes (e.g. quality, location, and cost), depending on the relative importance of each attribute [15]. Survey respondents are asked to choose between hypothetical scenarios (known as ‘choice sets’) in a way that is similar to the real-world decision-making process. The DCE approach is often used to obtained quantitative data on preferences for services not yet widely available in a specific context (e.g. HPV testing in South Africa).

The aim of the present study was to use a DCE survey to evaluate preferences for primary HPV testing among South African women, focusing on aspects that are most applicable to low-income and middle-income countries (i.e. same-day testing and treatment; self-collected swabs; and screening in mobile clinics).

2. MATERIALS AND METHODS

A DCE survey was conducted among a convenience sample of women who attended two urban clinics in Johannesburg, South Africa, from February 1 to May 31, 2018. Ethical approval was granted by the Human Research Ethics Committee of the University of the Witwatersrand, Johannesburg, South Africa (M171076) and the Institutional Review Board of the University of North Carolina, Chapel Hill, USA (17–2296). All eligible participants provided written informed consent.

One of the study sites was a hospital-based clinic co-located with a large public-sector HIV treatment program, which provides cervical cancer screening and treatment of precancerous lesions to approximately 3000 women infected with HIV annually. The second site was a mobile clinic, which provides screening services to 2000 women in one of Johannesburg’s poorest townships each year [16].

Eligible participants were women aged 18 years or older. Women who were unable to provide written informed consent were excluded from the present study.

The DCE survey was conducted in English. After completing a short series of sociodemographic and knowledge-related questions, the participants’ understanding of the survey technique was assessed using an everyday example comprising two choice sets (i.e. a consistency check). Participants were then asked to state their preference for cervical cancer screening scenarios that included the following attributes of primary HPV testing: method of collecting the swab; timing of treatment; type of clinic; cost of services; and time spent in the clinic. Three attributes were chosen a priori: method of collecting the swab (provider or self), timing of treatment (same day or on return to the clinic), and the type of clinic (static or mobile). Cost of services (US $0 or US $4) and time spent in the clinic (30 minutes, 60 minutes, or 120 minutes) were selected through formative work conducted among women attending the hospital-based clinic (n=30; results not shown).

The DCE survey was designed, administered, and analyzed using Sawtooth version 9.6.0 (Sawtooth Software, Provo, UT, USA). Our study was powered to detect all two-way interactions between attributes using the method described by Louviere et al. [15]. To achieve the desired level of statstical power, a total of 300 women were enrolled. In addition, the number of attributes was restricted to five and the number of levels within each attribute to no more than three. Each participant responded to a random subset of 10 choice sets [12]. As shown in Figure 1, each choice set comprised two HPV-based screening scenarios, followed by an option to choose no screening [15].

Figure 1.

Example choice set from discrete choice experiment survey on cervical cancer screening. Abbreviation: R, South African Rand.

The share of choices for each attribute level was determined and assessed within attribute differences using the χ² test. Associations between sociodemographic variables (i.e. study site, education, method of paying for healthcare, and cervical cancer screening history) and each attribute was also examined. No statistically significant associations were found in univariate analyses (data not shown); therefore, unadjusted findings were reported. To evaluate the importance of each attribute in determining the choice to undergo screening, a multivariate logistic regression model was constructed. The marginal utility coefficient (β) was calculated for each attribute, with the accompanying P value derived from a two-sided t test (P<0.05 was considered statistically significant). Willingness to pay and willingness to trade time were calculated using the method described by Lancsar et al. [12].

Finally, a market simulation model was created to estimate uptake of HPV-based cervical cancer screening under different scenarios. In the simulation, the base case represented a conventional primary HPV screening paradigm: provider-collected swab; return-visit treatment; attendance at a static clinic; service cost of US $0; and 60 minutes spent in the clinic. The individual attributes were varied to understand how each one would affect screening uptake. Complex scenarios with multiple screening options offered simultaneously were then included in the simulation on the basis of their clinical or public-health relevance.

3. RESULTS

A total of 300 women participated in the DCE survey; however, two women (0.7%) did not complete the example choice sets correctly and so were excluded. Of the 298 women included in the analysis, 196 (65.8%) attended the hospital clinic and 102 (34.2%) attended the mobile clinic.

The sociodemographic characteristics of the participants are shown in Table 1. The median age was 38 years and 223 (74.8%) women had completed secondary level education. Although only 141 (47.3%) women reported being in formal employment, 241 (80.9%) received income from either employment or a social grant. Access to basic utilities (i.e. electricity, running water, and toilets) and technology (i.e. television, mobile telephone, and internet) was assessed as a surrogate for socioeconomic status. The participants’ access to these various services was generally high; however, only 48 (16.1%) women had internet access at home.

Table 1.

Sociodemographic characteristics of the participants (n=298).^a

Characteristic	Distribution
Age, y	38 (32–45)
Education level (grade)
Primary (0–7)	19 (6.4)
Secondary (8–12)	223 (74.8)
Tertiary (>12)	56 (18.8)
Language spoken at home
English	41 (13.8)
Zulu, Xhosa, or Ndebele	140 (47.0)
Sotho, Northern Sotho, or Tswana	86 (28.9)
Afrikaans	4 (1.3)
Other	27 (9.0)
Source of income
Unemployed; not receiving a social grant	57 (19.1)
Unemployed; receiving a social grant	100 (33.6)
Employed; not receiving a social grant	89 (29.9)
Employed; receiving a social grant	52 (17.4)
Access to utilities and technology
Electricity	252 (84.6)
Running water	247 (82.9)
Toilet	268 (89.9)
Television	246 (82.6)
Mobile telephone	255 (85.6)
Internet	48 (16.1)
Marital status
Never married	144 (48.3)
Married or cohabiting	105 (35.2)
Divorced or separated	27 (9.1)
Widowed	22 (7.4)

^aValues are given as median (interquartile range) or number (percentage).

Encouragingly, 222 (74.5%) women had heard of cervical cancer before participating in the DCE survey. Table 2 outlines the participants’ knowledge and experience of cervical cancer screening. The participants tended to be knowledgeable about cervical cancer, correctly identifying risk factors, including HIV (n=254, 85.2%) and smoking (n=230, 77.2%), and selecting HPV vaccination as a preventive tool (n=231; 77.5%). By contrast, few women were correctly able to identify common symptoms of cervical cancer (n=60; 20.1%). In all, 215 (72.2%) women reported previously attending cervical cancer screening, with 202 (94.0%) undergoing a Papanicolaou smear test within in the past 5 years.

Table 2.

Knowledge and experience of cervical cancer screening among the participants (n=298).^a

Variable	Distribution
Who makes decisions about your health care?
I do	246 (82.6)
I do together with my partner and/or family	41 (13.8)
My partner and/or family	10 (3.3)
Other	1 (0.3)
How do you usually pay for healthcare?
Free care	203 (68.1)
Private health insurance	4 (1.4)
Government health insurance	28 (9.4)
Cash	43 (14.4)
Borrow money	20 (6.7)
No. of cervical cancer knowledge questions answered correctly
0	1 (0.3)
1	12 (4.0)
2	46 (15.4)
3	128 (43.0)
4	97 (32.6)
5	14 (4.7)
Most recent cervical cancer screening visit, y
Never received screening	83 (27.9)
<1	135 (45.3)
2–5	67 (22.5)
6–10	10 (3.3)
>10	3 (1.0)

^aValues are given as number (percentage).

Figure 2 details the share of choices for each attribute in the DCE survey. The attributes chosen most often were provider-collected swab; same-day testing and treatment; static clinic; service cost of US $0; and 30 minutes spent at the clinic (P<0.05 for all attributes). The share of choices was similar by study site (mobile vs hospital); years of education (<12 years vs ≥12 years); receipt of free care (yes vs no); and previous cervical cancer screening (yes vs no; data not shown).

Figure 2.

Share of choices for each attribute within the choice set. Statistical significance was determined using the within-attribute χ² test (* P<0.05; ** P<0.01).

The regression analysis of women’s stated preferences is shown in Table 3. The findings suggested that offering free services (β=0.50; P<0.001) and same-day testing and treatment (β=0.40; P<0.001) would have the largest impact on the participants’ choice to undergo primary HPV testing. Moreover, participants were willing to spend approximately 4 hours extra in the clinic if screening services were provided free of charge and approximately 3 hours extra if same-day testing and treatment were available. Despite stating a preference for free care, the participants were willing to pay US $3.33 for same-day testing and treatment, which was notable given that the median daily income in South Africa is US $2.25 [17]. Although statistically significant, spending less time in the clinic (β=0.16 for 30 minutes; P<0.001), having a healthcare provider collect the swab (β=0.12; P<0.001), and access to a static clinic (β=0.07; P<0.008) were all deemed to be less important for the decision-making process. Compared with the 3–4 hour time frame that participants were willing to accept for the most important attributes, they were only willing to spend an additional 1 hour (or no more than US $1.00) for the least important attributes.

Table 3.

Estimates of the weight for each attribute when making decisions about HPV-based cervical cancer screening.

Attribute	Marginal utility coefficienta	Standard error	P valueb	Willingness to pay, US $c	Willingness to trade time, duration
Method of collection
Provider-collected swab	0.12	0.02	<0.001	1.00	57 min
Self-collected swab	–0.12
Timing of treatment
Return to clinic	–0.40	0.03	<0.001
Same day	0.40			3.33	3 h 10 min
Type of clinic
Static	0.07	0.03	0.008	0.58	33 min
Mobile	–0.07
Cost of servicesc
US$0 (0 ZAR)	0.50	0.03	<0.001	NA	3 h 57 min
US$4 (50 ZAR)	− 0.50
Time spent in clinic
30 min	0.16	0.04	<0.001	0.88	NA
1 h	0.05			1.13
2 h	–0.22		<0.001

Abbreviations: NA, not applicable; ZAR, South African Rand.

^aMultiple regression modeling includes the option to select no screening. The model was unadjusted after testing for potential confounding and interaction.

^bP-value derived from a two-sided t-test (p<0.05 was considered statistically significant).

^cExchange rate: 1 US $ = 12 ZAR.

The market simulation of primary HPV testing models is shown in Figure 3. In all, 83.8% of participants would have chosen to undergo cervical cancer screening under the base case strategy (Figure 3A). The effect of providing various screening options was also explored, with combination strategies resulting in the greatest increase in uptake of screening services (Figure 3B). Specifically, offering both same-day and return-visit treatment options resulted in a 12.6% increase in uptake relative to the base case. Modest increases in uptake rates were also projected for static and mobile clinics (5.9%); self-collected and provider-collected swabs (5.6%); self-collected swabs and same-day testing and treatment (4.5%); and same-day testing and treatment with 2 hours spent in the clinic (4.2%). Strategies that resulted in additional cost, increased length of time spent in the clinic, and the need for self-collected swabs were projected to decrease the uptake of screening.

Figure 3.

Market simulation of the models for primary HPV testing. (A) The colored bars represent the absolute proportion of women estimated to undergo cervical cancer screening for each scenario. The base case was defined as provider-collected swab, return-visit treatment, static clinic, service cost of US $0, and 60 minutes spent in the clinic. The y-axis labels indicate attributes that differ from the base case. (B) The colored bars represent the incremental effect on cervical cancer screening uptake estimated for each scenario when compared with the base case.

4. DISCUSSION

Our DCE survey of HPV-based cervical cancer screening identified strong preferences for free services and same-day testing and treatment. Furthermore, most of the participants (83.8%) would be willing to undergo screening if services were free, the swab collected by a healthcare provider, and treatment offered at a return visit. Providing women with more options, including same-day testing and treatment, would potentially increase the rate of screening uptake to 96.4%. Although limited to a small sample of women in an urban setting, our findings may be used to inform future implementation of public-sector cervical cancer prevention initiatives in South Africa.

In 2017, the South African National Department of Health cervical cancer prevention policy was updated to include primary HPV testing [18]. This change in policy signaled a positive shift in governmental preparations to address the growing burden of cervical disease in South Africa, which is projected to rise by at least 40% over the next 20 years [19]. However, specific guidance on policy implementation and plans for program rollout at the provincial and district levels remain suboptimal. Indeed, most local policies and performance metrics are unchanged, thereby perpetuating a system in which access to cervical cancer screening and treatment of precancerous lesions is inequitable [3].

As South Africa works to eliminate cervical cancer as a public-health problem, primary HPV testing will form an increasingly central part of the prevention agenda in this country [18]. One way to facilitate scale-up of molecular testing is to use point-of-care technologies. Performed using benchtop analyzers, point-of-care HPV tests would enable same-day testing and treatment [20]. One such testing platform (GeneXpert; Cepheid, Sunnyvale, CA, USA) is already widely available across Sub-Saharan Africa owing to its use in the diagnosis of HIV and tuberculosis.

Among the most notable results of our study was that the participants expressed a strong preference for same-day HPV testing and treatment. Although this strategy is thought to be cost-effective in the context of low-income and middle-income countries, including those with generalized HIV epidemics [21, 22], few programs in South Africa have previously used the single-visit strategy [5]. The local standard of care follows a multivisit paradigm in which abnormal Papanicolaou smear test results are followed by diagnostic biopsy and excisional treatment if biopsy confirms the presence of high-grade precancerous lesions [18]. However, this burdensome arrangement creates bottlenecks in service delivery and is costly in both money and time to women who must return to the clinic three or four times before undergoing treatment.

Addressing hidden patient expenses is especially important given the current finding that the single most important attribute influencing the decision to undergo cervical cancer screening was cost. Specifically, provision of such services at no cost was favored, with women willing to spend approximately 4 hours extra in the clinic to obtain free cervical cancer screening. Single-visit cervical cancer screening programs in other low-resource settings have reported reduced costs [23]. These programs could serve as examples for South African policy makers and public-health practitioners when seeking to implement same-day HPV testing and treatment.

Another way to facilitate primary HPV testing is to incorporate self-collection of swabs into the cervical cancer screening strategy [9]. Previous studies suggest that self-collected swabs would be acceptable to women, including those living in South Africa [10, 24]. However, among women who prefer provider-collected swabs, the most frequently reported reason was uncertainty over their ability to self-collect the specimen correctly [10]. Anecdotally, the present study participants expressed similar concerns and, when presented with a choice, most favored the option of provider-collected swabs. However, our market simulation model supports previous research findings that suggest that offering a combination of provider-collected and self-collected swabs would increase uptake of cervical cancer screening [11].

Our study had several limitations. First, it involved a small convenience sample of women who had accessed cervical cancer screening from public healthcare facilities within the same metropolitan area. Conducting the DCE survey in a larger and more representative sample of South African women, including those not currently engaged in such care, would help to generalize the results. Second, recruitment of participants from clinical settings might have contributed to the high estimates of screening uptake found for all scenarios examined in the market simulation. Third, the sample size limited the number of attributes and levels that could be included in the DCE survey. Including additional attributes such as distance to the clinic or type of provider would improve understanding of women’s screening preferences. Furthermore, the inclusion of additional levels and a broadened range of costs and time frames would enable a more nuanced analysis.

In conclusion, equitable access to cervical cancer prevention services is a global health priority. DCEs are useful tool to inform the design of patient-centered screening programs. To our knowledge, the present study was the first to use a DCE survey to explore women’s preferences for primary HPV testing in South Africa. This approach provided preliminary evidence that urban South African women would likely use HPV-based cervical cancer screening services if offered at no cost and with convenient access to same-day treatment.

Synopsis.

When surveyed about HPV-based cervical cancer screening options, women in South Africa expressed preferences for free services; same-day testing and treatment; and provider-collected swabs.