Analysis of cancer diagnoses from 2015-2019 within Machakos County, Kenya, support establishment of Cancer Centre in 2019 likely changing referral patterns [version 1; peer review: awaiting peer review]

Background: In Kenya, cancer is an increasing public health challenge, with an estimated 48,000 new cancer cases and over 33,000 deaths recorded each year. Machakos County, Kenya, recently opened a cancer centre with an aim of bringing cancer services such as surgical oncology, chemotherapy, radiotherapy, and pathological services closer to the residents of Machakos County. The county is also one of four counties selected for the universal health coverage pilot programme making it uniquely poised to inform cancer control programs at the local, national, and international levels. Methods: This study aimed to build a cancer database to enable future population-based cancer studies by reviewing cancer diagnosis records for selected major public hospitals in Machakos County from 2015-2019. Medical records data were retrieved from Machakos Level 5 Hospital and Kangundo, Matuu, Kathiani and Mwala Level 4 Hospitals. Results: A total of 522 cancer cases were recorded across the study period with more than a third (N=172, 33%) diagnosed AugustDecember of 2019 when the Machakos Cancer Centre opened. Among the cancers diagnosed, the majority were in women (59.2%), with cervix uteri (n=106, 34.3%) followed by breast (n=62, 20.1%) as the most common cancers. For males, oesophagus (n=52, 24.4%) followed by prostate (n=43, 20.2%) were the most common cancer types. The highest crude rates per 100,000 persons were for Kangundo 67.4 and Matungulu 53.2 subcounties. Conclusion: It is clear that access to cancer care treatment will change referral patterns for residents in Machakos County and with Open Peer Review Reviewer Status AWAITING PEER REVIEW Any reports and responses or comments on the article can be found at the end of the article. Page 1 of 10 Wellcome Open Research 2020, 5:290 Last updated: 17 DEC 2020


Introduction
In Kenya, cancer is the third-leading cause of death after infectious and cardiovascular diseases. An estimated 48,000 incident cancer cases and more than 33,000 deaths are recorded every year in Kenya (Globocan, 2018). Age-standardised mortality rates are 1.4 fold higher in Kenya with 129 deaths/100,000 persons compared to North America (Globocan, 2018;Siegel et al., 2020). Kenya, like many countries in Africa are undergoing a demographic transition and there is consensus that the cancer burden is increasing due to multiple factors including an aging population and increased prevalence of risk factors associated with cancer (Brinton et al., 2014;Brinton et al., 2017;Parkin et al., 2014). Compounding the burden of increasing cancer incidence is also higher frequency of poor prognosis tumours, delayed presentation, late diagnosis and access to preventative care. For example 60% of breast cancer cases in sub-Saharan Africa are diagnosed at late stage and 18% at metastatic stage (Joko-Fru et al., 2020;Makau-Barasa et al., 2018;Sayed et al., 2013). Yet, the true burden is likely underestimated given a deficiency in cancer registry and surveillance networks.
In Kenya, cancer incidence and mortality estimates have been derived from three population-based cancer registries in Nairobi, Eldoret and Kisumu (AFCRN, 2020; Korir et al., 2016;Kruk et al., 2018). Additional cancer registries that have been proposed and are at different levels of development include Meru, Embu, Mombasa, Garissa, Nakuru, Kitui, Kakamega, Machakos and Kiambu (KEMRI, 2019). Data from 2018 based on these registries show crude cancer incidence rates estimated at 93.5 per 100,000 persons and a mortality rate of 63.7 cases per 100,000 persons. In total, cancer accounted for 7% of all deaths (Globocan, 2018). The majority of those affected by cancer in Kenya have an earlier age of onset compared to European countries, with approximately 60% diagnosed below the age of 70 years. Indeed, studies have shown that African ancestry is associated with earlier onset of some types of cancers such as breast cancer (Churpek et al., 2015). The Ministry of Health further estimates the risk of getting cancer at 14%, while the risk of dying from cancer is estimated at 12% in a lifetime (Ministry of Health, 2018a).
Rising cancer incidence and increased public concerns in Kenya led to the passing of the Cancer Control and Prevention Bill by Parliament in 2011 (Ministry of Health, 2018a), followed by publication of the national Cancer Control Strategy and increased investment in cancer diagnostic equipment. Yet, investments in cancer surveillance and registry activities to allow objective assessment of programs to monitor and improve cancer control programs are limited.
Machakos County borders the eastern side of the Nairobi metropolitan area and has a total population of 1,421,898 people according to the Kenya National Bureau of Statistics (KNBS). The county covers an area of 6,208 square kilometres, giving a population density of about 229 persons per square kilometre (KNBS, 2019). The county is dominated by the Akamba people with a major emphasis on farming and agriculture (Machakos Assembly, 2019). In December 2018, the Government of Kenya launched the Universal Health Coverage (UHC) Pilot Program, which aims to enable Kenyans to have access to affordable healthcare by reducing the cost of health services. The pilot phase targeted four counties in the country, which are Kisumu, Nyeri, Isiolo and Machakos (Ministry of Health, 2018b). In August 2019, the Machakos County government established a new cancer centre to tackle the increasing burden of cancer care. At the centre, residents of Machakos County who are registered in the UHC Programme have access to free cancer screening, diagnosis, counselling and treatment (Mutavi, 2019). Before these developments, most of the residents of Machakos County preferred to seek health care in Nairobi where advanced laboratory services and medical services are available. Given Machakos County's UHC pilot programme and newly opened cancer centre here we aimed to establish the cancer burden and determine if establishment of access to cancer services change referral patterns across the five county public hospitals in order to inform future national efforts for population-based reporting of cancer incidence and mortality for improved monitoring of cancer prevention and control programs.

Ethical statement
The study protocol was approved by the Institutional Ethics Research Committee of Mount Kenya University (MKU/ ERC/1515) and licensed by the National Commission for Science, Technology, and Innovation (NACOSTI/P/19/3043). Cancer registrars were trained on how to abstract data manually using a standardized data extraction tool and on maintaining confidentiality while handling patient files. The filled abstraction forms were kept in a secure room under lock and key within the Directorate of Research and Innovation, Mount Kenya University with limited access to only authorised and trained personal on data ethics. Authorization to conduct the study was obtained from the county department of health and administrators at each participating health facility. Patient consent was not sought since the study was only retrospective and no active follow up was done. The ethics committee waived consent of individual patients due to the nature of the work being minimal to no risk and the nature of it being secondary data analysis.

Study population
Machakos County borders Nairobi and Kiambu counties to the West, Embu County to the North, Kitui County to the East, Makueni County to the South, Kajiado County to the South East and Murang'a and Kirinyaga counties to the South West ( Figure 1). The county is divided into nine sub-counties, which are Masinga, Yatta, Kangundo, Matungulu, Kathiani, Mavoko, Machakos, Kalama and Mwala sub-counties. The classification of health facilities in Kenya is into six levels of hierarchy; Level 1 are community health providers; Level 2 are dispensaries and clinics; Level 3 are health centres and nursing homes; Level 4 are sub-county hospitals; Level 5 are county referral hospitals and large private hospitals; and Level 6 are national referral and large private teaching hospitals (Murkomen, 2012). In this study, all major Level 4 and 5 hospitals in Machakos County were selected for data abstraction since these are the facilities where cancer patients receive medical and palliative care.

Data collection
Data from 2015-2019 calendar years for five hospitals were obtained, specifically: Machakos Level 5 Hospital and Kangundo, Matuu, Kathiani and Mwala Level 4 Hospitals. In all the Level 4 health facilities, records are centrally located at the records departments (Figure 2). At Machakos Level 5 hospital, hospital files for patients with chronic conditions such as diabetes, and cancer are kept in the specialty clinics. In some cases, the files are centrally located at the health records department and are retrieved and taken to different clinics or wards whenever the patients have a scheduled clinic or are admitted, respectively. The Health Records Officer (HRO) in-charge at each hospital organized retrieval of patient files a day prior to the visits to the facilities. Retrieved files were reviewed and information abstracted to a predesigned abstraction form. Information abstracted from the files included patient's name, age, sex, county of residence, cancer's primary site, histology, facility name date of incidence, and any comorbidities. Initially, the study aimed to analyse cancer cases reported from 2013 to 2019. However, most of the patient files for 2013 and 2014 were missing in the central records department since patient files are archived after five years since the last visit. A total of 587 cancer cases were recorded from the five facilities visited. Incomplete entries and any duplicates belonging to the same person based on name, gender, age, primary site, and residency, as well as those cases diagnosed in 2013 and 2014, resulted in removal of 65 entries. The cases were given unique patient identifiers and then imported into the Statistical Package for Social Sciences programme (IBM-SPSS) version 26.

Statistical analysis
Crude rates for the period under study were calculated by dividing the number of cancer cases reported at each sub-county by the number of people in the sub-county. The 2019 national census (KNBS, 2019) was used as the source of population denominators for the whole of Machakos County and each sub-county and to describe the population age structure (population pyramid). The cancers were listed according to the International Classification of Diseases for Oncology (ICD-O) third edition. With an aim of establishing a cancer registry at the hospital, the abstracted data was entered into the CanReg5 software from International Agency for Research on Cancer (IACR, 2020), which the hospital is expected to continuously update.

Results
Machakos County population demographics (Figure 3) show that it is a relatively young population with more than 85% under the age of 50 (Kamita et al., 2020). During the fiveyear period (2015-2019), a total of 522 cases were registered in Machakos County public health facilities. The majority of patients (80.6%) were residents of Machakos County, 56 (10.1%) were from Makueni County and 23 (4.2%) were from Kitui, while the rest were from other counties. Table 1 summarizes the frequency distributions and age at diagnoses for cancer cases. Women accounted for 59.2% (309 cases).
For males, oesophagus (n=52, 24.4%) followed by prostate (n=43, 20.2%) were the most common cancer types. Among females, the most common types were cervix uteri (n=106, 34.3%) followed by breast (n=62, 20.1%). A little less than half of cancer patients (42.1%) in this population were diagnosed between 50-70 years of age, and the median age at diagnosis was 58 years (Table 1).
The number of cancer cases per facility varied, with Mwala Level 4 hospital documenting the fewest cancer cases (n=9, 2%) and Machakos Level 5 hospital, where the new cancer centre was established, seeing to the highest number of cases (n=353, 66%). Kangundo Level 4 hospital was the other facility that saw a high number of cancer patients (n=123, 24%). Figure 5a shows that, in the two facilities that saw the highest number of patients (Machakos and Kangundo), the number of cancer patients seen per year did not vary much from 2015 to 2018. However, in 2019, there was a significant increase in the number of cancer cases reported at Machakos Level 5 with not much increase at Kangundo Level 4 hospital. In terms of the four most common cancers, Kangundo Level 4 hospital more frequently saw oesophageal cancer as the most common cancer, while cervix uteri cancer was the leading cancer at Machakos Level 5 hospital (Figure 5b).
Using the 2019 census, crude rates per 100,000 people for each sub-county were calculated for the cancer cases of the patients who indicated Machakos County as their county of residency. Matungulu (86 cases) had the highest number of cases, while Mavoko sub-county had the least number of cases (five cases). Machakos sub-county, where the Level 5 hospital is domiciled was second in terms of the number of cancer cases (N=76). Crude rates were highest in Kangundo (67.4/100,000) followed by Matungulu (53.2), while Masinga (15.5) and Mavoko (1.6) had the lowest rates. Overall, the crude rate of cancer in Machakos was 30.9, which is lower than the rates in most of the sub-counties (Figure 4). In our data abstraction form, we aimed to capture additional information such as the patient's HIV status, comorbidities and their smoking or drinking status. However, such information was missing in most of the files. For instance, only 28% (n=144) and 26% (n=134) of the patient files had indicated the smoking and drinking statuses of the patients, respectively, while only 41% (n=213) and 24% (n=123) of the files indicated the occupation of the patient or whether the patient had a family history of a familiar disease, respectively.
Machakos County opened its first cancer centre in August 2019 and within just five months of its opening we observed over an eight-fold increase in the number of cancer cases diagnosed at the Level 5 hospital with little to no change in the peripheral facilities.

Discussion
The current study aimed to establish cancer referral patterns and the most common types of cancers diagnosed at Level 4 and 5 Machakos County hospitals based on hospital records from 2015 -2019 to facilitate future cancer data science, cancer registration and epidemiologic studies that could improve cancer control and services programmes.
Our data show that women have the higher burden of cancers, with cervix uteri and breast the most common cancer types. Our data are consistent with those from Moi Teaching and Referral Hospital and Kenyatta National Hospital (Macharia et al., 2019). In Eldoret cancer registry, data collected between 2008 and 2012 showed the leading cancer in women is cervix uteri (20.7%) followed by breast cancer (19.7%). In 2018, Globocan reported that the leading cancer in women is breast cancer (12.5%) followed by cervix uteri (11.1%). In men, prostate cancer (14.9%) was the leading cancer type followed by oesophagus cancer (12.4%) (Globocan, 2018). The data are also comparable with reports from the Kenya National cancer control strategy and Tenwek Hospital in Western Kenya (Parker et al., 2010), where prostate and oesophageal were reported to be the major cancer types affecting men.
Although oesophageal cancer is ranked as the eighth most common cancer worldwide, more than 80% of cases and deaths from oesophageal cancer are reported in developing countries, including Kenya (Van Loon et al., 2018). Higher incidence of oesophageal cancer has been well documented, and our data are consistent with other reports showing it to be a major contributor of cancer incidence and deaths affecting the eastern corridor of Africa, which spans from Ethiopia to South   (Parker et al., 2010). Indeed, the number of oesophageal cancer cases has been increasing in Kenya and at a higher rate compared with other populations (Patel et al., 2013) with an earlier age at onset (Cheng et al., 2015). A study conducted by Cheng et al. (2015) reported that up to 11% of all cases of oesophageal cancer reported at Tenwek Hospital occurred in patients who were aged 30 years or less. In Kenya, more than 70% of oesophageal cancer cases are diagnosed when the disease is at its late stages, primarily due to lack awareness amongst both the patients and healthcare workers, as well as lack of sufficient diagnostic capabilities. These factors have resulted in the use of stents as the only viable palliative treatment for oesophagus cancer, with a lower chance of using curative treatments (Odera et al., 2017). In contrast, Linxian, China, where oesophageal cancer incidence has also been noted to be high, cases in those under 30 years of age are rare and account for less than 1% of all cases (Zhang et al., 2012). In the USA, studies have shown that the proportion of oesophageal cancers in patients aged 30 years or less is under 0.2% (SEER, 2020). Hence, further studies are needed to determine risk factors and target those that re modifiable that could lower the incidence of oesophageal cancer in Machakos County. The current data show that 80% of the cancer patients receiving care from the health facilities in Machakos County are residents and that the number of cancer cases diagnosed increased significantly in 2019, coincident with the recently established Machakos Cancer Care and Research Center. The opening of this centre, which is now providing diagnostic and chemotherapy services along with the pilot UHC services, has likely impacted a change in referral patterns for this population, allowing more patients to receive care at Machakos Level 5 Hospital as opposed to other referral centres in Nairobi or Kiambu counties. Out of the five health facilities that were included in the study, about 90% of the cases were from Kangundo Level 4 hospital and Machakos Level 5 hospitals. This illustrates the important of these two facilities in their role in cancer patient management either as a screening and treatment centre or a palliative centre. Limitations of our analysis are that we cannot compute population-based rates due to incomplete cancer case assessment since given its specialised nature, cancer care was predominantly obtained outside of Machakos County prior to 2019. The data presented are also limited to the patients seeking cancer care in public hospitals in Machakos County. The data, therefore, miss those patients who may have sought medical care from private hospitals in Machakos County where screening of some cancer types such as cervical cancer is also done. While we aimed to obtain additional data on patients, there was incomplete ascertainment and documentation of histology information, mortality status as well as risk factor data such as smoking and drinking status. Another limitation is that accuracy of the place of residency cannot be ascertained since some patients take residency to mean county of birth, and would provide the county where they were born as the place of residency even when they reside in a different county.

Conclusions
The study established a cancer database for public hospitals in Machakos County from patient files who attended the hospitals between January 2015 and December 2019. As reported in other parts of the country and from the data at some of the already established cancer registries in the country, cervix, breast, and oesophagus cancers are the major cancer types reported at the health facilities in Machakos County. Although the current data do not include cancer cases of all the county residents that have been handled in private hospitals in the county or other counties where cancer care is common (e.g. Nairobi, Eldoret and Kiambu), it documents the changing referral patterns of cancer care and baseline data for establishing a cancer registry. Future efforts to obtain a national cancer registry would allow a more comprehensive picture of cancer burden and referral patterns. This study also allows the possibility for future population-based epidemiology studies to assess risk factors to understand the etiologic factors associated with increasing cancer incidence in the county.

Data availability
Kamita