Cost of HPV screening at community health campaigns (CHCs) and health clinics in rural Kenya

Invasive cervical cancer is highly preventable through organized screening programs. However, globally over half a million women are diagnosed with the disease each year [1]. Cervical cancer is the most common cancer among women in Sub-Saharan Africa [2]. Nine out of 10 cervical cancer deaths occur in low-resource countries, with an 85% mortality rate in Sub-Saharan Africa [3, 4]. While in developed countries a major decline in cervical cancer incidence and mortality occurred after the introduction of cytologic testing, many developing countries still lack effective screening programs [5].

High cancer mortality rates in developing countries may be due to inadequate health facilities and lack of personnel [6‐8]. Organized screening programs are rare, and screening rates are low in developing countries, where only 19% of women are screened for cervical cancer as compared to 83% in the U.S. [9]. Simplified screening strategies employing testing for high-risk human papillomavirus (HPV), the causative agent in most cervical cancers, have been shown to reduce mortality from cervical cancer when coupled with outpatient treatment of precancerous lesions [5, 10, 11]. Programs implementing HPV testing in community-based settings may address the infrastructure and personnel limitations that prevent screening in some low-resource settings.

In Kenya, cervical cancer is the most frequent cancer among women, with about 4800 annual diagnoses and over 2400 deaths per year [12]. HPV testing is recommended by Kenya’s Ministry of Public Health and Sanitation’s Division of Reproductive Health, but is not reliably incorporated into most community or government programs [13]. Visual inspection with acetic acid-based screening is available at some clinics in Kenya. Screening rates are extremely low, at 13.8% nationwide and below 11% in rural areas in 2014 [14]. Even when screening is available, uptake can remain poor, due to low levels of knowledge of cervical cancer [15], and lack of awareness of HPV screening and the risks of cervical cancer [16]. Very few women recognize that routine testing is the primary way to prevent cervical cancer [17].

The community-based health fair is an important strategy to help people in developing countries access important preventive healthcare messaging and services. Community health campaigns (CHCs) occur over a short duration, delivering preventive health services at high-volume close to residential areas. CHCs have been effective in reducing maternal and neonatal mortality [18‐20], prevention of malaria [21], HIV testing and delivery of antiretrovirals [22, 23], tuberculosis detection [24], and increasing mental health care access [25]. A multi-disease CHC in rural Kenya found increased ART coverage and other health benefits with cost savings [26, 27]. CHCs are part of a community-based healthcare model, which has gained traction in recent years because it can mobilize a large proportion of a community, and may be less resource-intensive than receiving preventive care at clinics.

In this study, we used costing data collected during a cluster-randomized trial in rural Kenya comparing uptake of HPV-based cervical cancer screening in community health campaigns and government clinics to compare the costs of the strategies. This is the first study to estimate the costs of cervical cancer screening with a community-based health campaign strategy, and furthermore, compares the costs of two cervical cancer screening interventions in Kenya. The cost estimations will help inform future implementers on funding and resource needs for cervical cancer screening.

This study was conducted as part of a cluster-randomized trial in 12 rural communities in Migori County in the former Nyanza Province of western Kenya between January and September 2016. Two-thirds of citizens of Nyanza live on less than $1 per day, and the province has the highest prevalence of HIV in the country. The target population was women who were eligible for and would benefit from cervical cancer screening per the Kenya Ministry of Health Guidelines: 25 to 65 years old with an intact uterus and cervix. Site selection was based on census data, health facility information, mapping, and demographic data [28]. Each community was a cluster of villages or sub-locations within a defined administrative boundary. The twelve communities were randomized in a 1:1 ratio using an allocation sequence generated by Stata 11 MP (StataCorp, TX). Six communities were randomized to a CHC intervention and six to HPV screening at government clinics.¹ Despite random assignment, women in CHCs were slightly older and more likely to have been screened for cervical cancer in the past [28]. Non-adjacent communities with a population of 4500 to 9000, with at least one health facility were identified. Both arms included three phases for which costs are presented: outreach and mobilization (“Outreach”), HPV-based screening (“Screening”), and notification of results and standard referral (“Notification”). Treatment services were equivalent between the arms, and therefore not costed. More information on the details of the study design can be found in the main outcomes paper, Huchko et al. 2017 [28].

Figure 1a and b summarize the workflows for each of the three phases for CHC communities and clinic communities, respectively. For CHC communities, each of the three phases (Outreach, Screening, and Notification) lasted two weeks. A subset of the CHC team conducted outreach through stakeholder meetings, information sessions, door-to-door mobilization, announcements using a public-address system, and posters. During screening, the entire CHC team traveled to different areas of the CHC community every morning and set up tents. The team then conducted a sequence of activities dedicated to screening for each woman visiting the tents: registration, group education, informed consent, and self-collection of screening specimens using the careHPV test kit. The lab technician then tested whether the woman was HPV-positive using the careHPV™ test system. After two weeks of screening activities, the CHC team moved onto notification of results and standard referral, which again lasted two weeks. Program assistants worked on using text messages and phone calls to notify women of their results, and Community Health Volunteers (CHVs) employed by the program conducted door-to-door home visits to notify women of their status. HPV-positive participants received referral to a treatment site located at Migori County Hospital.

In contrast to CHCs, each of the three phases were conducted concurrently at the six clinic communities for nine months total. After an initial series of key stakeholder meetings and information sessions prior to launching screening, one to two CHVs at each clinic conducted outreach through door-to-door mobilization and posters. During screening, CHVs, with supervision from the program coordinator and assistance from the data manager and program administrator, conducted all the screening activities at the clinic sites. They conducted group education, acquired informed consent, and collected HPV self-collection test kits. Notification also occurred throughout the nine months of screening at each clinic site. Since CHVs attended clinics fulltime and were in charge of notifying women who decided to visit clinics to receive test results, the CHVs were not involved with home visit notification, in contrast to CHCs. Instead, a team of program assistants would conduct home visits (1-3 days per month for each clinic) to notify women of their test results. The program assistants issued text messages and phone calls. HPV-positive women were referred for treatment at Migori County Hospital. More detailed information on the CHC and clinic community workflows is in the Additional file 1.

A series of sensitivity analyses were conducted to assess whether costs per screening estimated are sensitive to changes in implementation. First, cost estimates for CHCs were conducted under two alternate transportation scenarios to purchasing a program vehicle: (1) issuing travel reimbursements, or (2) renting a vehicle. For two of the six CHCs, travel vouchers were issued because a vehicle was not purchased at the time of the CHC. Travel reimbursement rates for these two CHCs were therefore used to approximate cost per kilometer of travel.² A rented vehicle costs 4000 KES per day, or 42 USD a day.

In the second sensitivity analysis, we explored a scenario where clinic personnel are used more efficiently, and therefore measure the costs of reducing the number of personnel dedicated to clinic screening. The research team, based on observation and experience, believed that each clinic could have hired another CHV who was dedicated to conducting home visits for the notification phase, rather than using program assistants. During the study, clinics each had five program assistants conduct home visits 1-3 times a month, which may have driven up personnel costs for notification at clinics. The team also believed that the program coordinator and data manager could have reduced their time spent on clinic activities, and delegated more responsibility to a program assistant, who would manage operations and data collection. The maximum number of clinics a program assistant could have managed was two clinics. Therefore, in the sensitivity analysis, a program assistant would dedicate full time to monitoring screening and notification at two clinics, where he or she would oversee the progress of the CHV and make appointments with certain patients who need more attention. The program coordinator would commit 5% time to each clinic, and lab technicians would dedicate time to each clinic based on the number of women screened (same as before, average 10% time across clinics). Each clinic would therefore only have four team members: CHV, program assistant, program coordinator, and lab technician. The personnel ratio per clinic would reduce from 2 to 1.65.

Finally, we also implemented cost regressions to estimate the fixed cost for the screening program and the marginal costs of each screening. We estimated total cervical cancer screening costs, one from each community, and regressed total costs on numbers of women screened in each community. Each cost regression has 6 observations, one from each CHC and clinic community.

Figure 2a and b depict cost breakdowns of CHC and clinic community costs, by cost type and phase, respectively. Personnel costs were higher for clinics ($14.27 per screening) compared to CHCs ($11.26 per screening). Capital costs were also higher for clinics ($5.55) compared to CHCs ($2.68), due to the need for clinics to pay for renting space at facilities.⁴ Services were almost equivalent ($2.48 for CHCs and $2.29 for clinics). Recurrent goods were the only input type that was slightly higher for CHCs, at $8.59 per screening compared to $7.45 for clinics; this reflects additional transportation costs associated with executing a CHC (e.g. fuel costs and transport reimbursements).

When comparing CHCs and clinics across activities, per-screening costs were higher for CHCs only in outreach ($3.34 vs $0.17 at clinics), whereas per-screening costs were higher for clinics for screening ($17.59 for CHCs and $21.88 for clinics) and notification ($4.18 for CHCs and $7.51 for clinics). CHCs had higher costs for outreach because they conducted more intensive outreach activities than clinics, including door-to-door mobilization, public address systems, and meetings with chief elders. A larger team of personnel was involved with CHC outreach activities. At clinics, the only outreach conducted was door-to-door mobilization by CHVs. During door-to-door mobilization at clinics, CHVs visited homes one-by-one and provided information on cervical cancer screening conducted at clinics. Costs of mobilization were low for clinics because daily compensation of CHVs is lower than wage rates for personnel staff. In clinic communities, CHVs also put up posters with further information about screening. The outreach activities at CHCs generated higher numbers of women screened at CHCs compared to clinics (483 vs. 340) in a shorter period of time (10 days vs. nine months).

Figure 3 shows further breakdown of personnel costs by cadre. While CHCs had higher costs for data management and other miscellaneous personnel costs (e.g. driver, security, tent setup team), clinics had higher personnel costs associated with hired program assistants (type 1 was involved with monitoring screening at the clinics, and type 2 were involved with home visits during notification).

Variation in costs per woman screened across CHCs can largely be attributed to differences in numbers of women screened at each CHC. For example, personnel team size and compensation rates do not change much across CHCs, but per-screening personnel costs for the screening phase among CHCs range from $4.80 to $9.00. Similarly, for capital goods, the cost of capital goods such as the careHPV test system, tablet, and pipettes, do not change for different CHC communities. Per-screening capital goods costs for the screening phase therefore vary because of differences in numbers of women screened at each community. Variation in recurrent goods ($8.09 to $8.82 per screening during the screening phase) occurred across communities because of difference in fuel costs and amount of supplies required for screening. Variation in services offered for notification (texts, calls, and home visits) occurred because of differences in notification options selected at each CHC.

Variation in cost per screening for clinics was primarily due to differences in costs of facility space. For most clinics, facility rent was not entirely allocated for cervical cancer screening, because other health services such as HIV testing were conducted in the same space. One clinic, however, had considerably higher facility costs because 100% of a clinic room was allocated to CCSP screening. Clinic personnel designated to screening activities shifted across the six clinics, which is why there was slight variation in clinic-level personnel costs. The team of personnel involved with notification did not change across clinics, and therefore, personnel costs for notification did not vary.

For capital goods, all six communities used the central careHPV Test System and shared two motorbikes. Similar to personnel, clinic-level costs of capital goods were estimated based on proportion of women screened at each clinic, and therefore did not vary much across the clinics.

In the sensitivity analysis in which transportation strategies were varied, the average cost per screening for CHCs under both transportation scenarios declined very slightly, at $23.12 per screening with a range of $20.53 to $27.51 for the transport reimbursement scenario, and at $24.98 per screening with a range of $21.68 to $30.42 for the rented vehicle scenario. The CHC cost per screening estimates are therefore not very sensitive to changes in transportation type used for the program.

In the second sensitivity analysis, personnel costs at clinics were estimated with an adjustment to team structure. During the program, a total of 11 individuals comprised clinic personnel: eight working on screening, and nine working on notification, with a personnel ratio of about two per clinic. Additionally, each clinic had one CHV working on both screening (trained to help women self-sample using HPV self-test kits) and notification.

Table 3 shows the results of the sensitivity analysis. Average cost per screening reduced to $25.69 for clinics, with a range of $20.57 to $32.12. The results measured a change in personnel efficiency. The average cost per screening from the sensitivity analysis was similar to per-screening costs of CHCs, differing by only 69 cents.

Finally, we conducted two cost regressions for the six CHCs and six clinic communities to empirically estimate the program fixed cost per community screened and the marginal cost of a screening. For CHCs, the regression results showed an intercept of $5368.24 representing fixed costs, and marginal costs of $13.89, representing variable costs. For clinic communities, the regression results showed an intercept of $982.27 representing fixed costs, and marginal costs of $26.68, representing variable costs. CHCs had higher fixed costs than clinics, and clinics had higher marginal costs than CHCs.