Establishing Effective Registration Systems in Resource-Limited Settings: Cancer Registration in Kumasi, Ghana

Abstract

Cancer control programs are needed worldwide to combat the increases in cancer incidence and mortality predicted for sub-Saharan Africa in the next decades. The effective design, implementation, and evaluation of such programs require population-based cancer registries. Ghana’s second largest medical center, the Komfo Anokye Teaching Hospital (KATH) in Kumasi, has made initial progress at developing a cancer registry. This registry, however, is housed in the medical oncology/radiotherapy center at KATH and does not currently include data from other departments that also interact with cancer patients. The aim of this study was to improve KATH cancer registration by compiling cancer data from other major departments that see cancer patients. Using recent population estimates, we calculated crude cancer incidence rates of the “minimally-reported cases” for the Ashanti region. The most common cancers found in this study were breast (12.6 per 100,000), cervix (9.2 per 100,000), and prostate (8.8 per 100,000). These cancers occur at similar crude incidence rates in other West African countries. Females had overall higher incidence rates than males, which is consistent throughout the West African region. This study identified a number of methodological challenges facing cancer registries in Ghana that can be addressed to improve the quality of cancer registries in other resource-limited settings. Such registries should be tailored to the local health system context. A lack of coordination among the sources reporting cancer cases and a lack of understanding of local health-care systems and payment plans may interfere with the quality, completeness, and comparability of data from cancer registries in resource-limited settings. Steps, barriers, and solutions for improving cancer registration in Ghana and countries at similar levels are discussed.

Introduction

More than half of incident cancer cases and two thirds of cancer deaths already occur in developing countries.¹ In sub-Saharan Africa, cancer incidence is expected to increase drastically over the next few decades due to multiple factors, notably increased life expectancy and westernized lifestyles as well as increased quality of diagnostics.^1-4 Despite the known growing burden of cancer in sub-Saharan Africa, little is known about the incidence and distribution of the different tumor types and few resources are allocated to early detection and disease management in this region.^1-3,5 Though cancer registration is indispensable as an accurate and longitudinal measure of the true burden of disease and thus to inform national cancer control programs,^6,7 only 11% of the population in Africa is covered by a cancer registry^1,8 and only 1% of African populations are covered by high-quality population-based cancer registries.⁹ Where registration systems do exist in these resource-limited settings, major challenges, including the lack of trained personnel, insufficient coordination of reporting sources, and the lack of available census data, threaten the quality and overall effectiveness of the registry.^3,10-12 Though Ghana currently lacks a fully functioning population-based cancer registry, there have been earnest attempts to describe the cancer burden in the country. The 2 largest hospitals that diagnose and treat cancer patients in Ghana are the Korle Bu Teaching Hospital (KBTH) in Accra and the Komfo Anokye Teaching Hospital (KATH) in Kumasi. Limited frequency data from these hospitals indicate that the most common cancers are of the cervix, breast, prostate, and liver, with higher proportions of cancer among women than men.13-16 In 2008, GLOBOCAN used a mixture of data from surrounding countries and extrapolations to estimate an age-standardized national cancer rate of 109.5 cases per 100,000 persons per year.17 GLOBOCAN also identified the same most frequent cancers that were reported by the limited patient data available from KBTH and KATH. Importantly, data from each of 32 sentinel hospitals within the 10 regions of Ghana indicate that cancer is among the top 10 causes of death in the country.^13,18

Cancer registration at KATH started with the development of a departmental-based registry in 2004.⁵ This registry was developed with the aim of reporting all cancer cases diagnosed and treated at the medical oncology/ radiotherapy department where most of the patients diagnosed and treated come from the Ashanti region.⁵ Although cancer data from other departments exists at KATH, no previous research has tried to link cancer data from different diagnostic and treatment sources at this hospital to arrive at more accurate Ghana-focused estimates of the true incidence of cancer.

Ghana has made initial progress towards the development of cancer registries, yet still lacks the resources and coordination of reporting sources needed for a fully functioning registration system. A lack of coordination of reporting sources within the hospital is a possible source of underestimation and inaccuracy, as cancer cases are not reported in every department. To this end, this study aimed to coordinate cancer data from the surgery, pathology, and medical oncology/radiotherapy departments to estimate the cancer burden seen at KATH from 2008-2010, obtain estimates of cancer incidence for the Ashanti region, and set the groundwork for the development of an effective cancer registry in Ghana.

Materials and Methods

Study Population and Data Sources

The study population consisted of all confirmed or suspected cases of cancer diagnosed in the surgery, pathology, and medical oncology/radiotherapy departments at KATH from 2008-2010. For each department, available information on cancer cases was abstracted from patient logbooks, medical records, or electronic databases into Excel databases.

Surgery

The electronic database of all 24,889 cases seen for any surgical procedure in the department of surgery during the period of January 2008-December 2010 was obtained. This database contained the following variables: patient name, age, sex, preoperative diagnosis, and postoperative diagnosis. Postoperative diagnosis data was missing for the majority of cases. Based on the available clinical diagnosis reported for each procedure, cases were reviewed by both Ghanaian and US physicians with more than 30 years of combined expertise in the heterogeneous presentations of cancers. The reviewers categorized the cases into the following 4 categories: definite cancer, likely cancer, unlikely cancer, and not cancer. The first 2 groups (cancer and likely cancer) included a total of 4,304 cases. The judgment call of “likely cancer” or “unlikely cancer” was based on experience for predicting with over or under 50% probability of cancer.

Pathology

Logbooks from the pathology department for the same period of 2008-2010 were obtained. Logbook entries included information on patient name, age, sex, site of procedure, and diagnosis. A total of 2,617 cases strongly suggestive of cancer were identified and abstracted.

Medical Oncology/Radiotherapy

Data from the medical oncology/radiotherapy department database was obtained for all cases seen during the same period of 2008-2010. This database contained demographic variables, such as name, age, sex, occupation, and town/region of residence and clinical information such as basis of diagnosis, primary site, histology, grade, stage, and treatment. Individual patient medical records were reviewed for this period to supplement and validate the data contained in the database. This process was completed for the data from the medical oncology/radiotherapy department for 2 main reasons: 1) to supplement the minimal diagnostic data included in the database with the complete diagnoses found in the patients’ medical records; 2) to provide a direct link to cases identified from the pathology department using the histopathology report number found in the medical oncology/radiotherapy medical records. A total of 1,936 cases of cancer were obtained from the medical oncology/radiotherapy department.

Data Linking/Matching

Once data from all departments were reviewed for accuracy, the following variables were used to link cases between departments: histopathology report number, patient name, patient age, date of diagnosis, and cancer site/site of surgical procedure. After linking the data from the 3 departments and removing duplicates, data was stripped of all identifiers and the final dataset for analysis was defined. Approval for this project was obtained from both the University of Michigan Institutional Review Board and Komfo Anokye Teaching Hospital/Kwame Nkrumah University of Science and Technology Committee on Human Research, Publication, and Ethics.

Estimation of Incidence and Statistical Analysis

KATH’s patient population comes largely from the Ashanti region, in which it is located. With a population of 4,725,046, Ashanti is the most populous region in Ghana with one of the fastest growing populations.¹⁹ KATH’s total catchment area covers nearly 50% of the total population in Ghana and also includes the more rural regions north of the Ashanti region, namely the Upper East, Upper West, Northern, and Brong Ahafo regions as well as parts of the Central and Western regions.¹⁹ KATH also sees a majority of the cancer cases in the Ashanti region. As it is estimated that 75% of KATH’s cancer patients are from the Ashanti region, and 25% of cancer cases in the Ashanti region are estimated to receive treatment outside of KATH, 100% of cases from this project were included for the estimation of incidence.

Population data for the Ashanti region was obtained from Ghana’s 2010 Population and Housing Census.¹⁹ Population estimates for 2008 and 2009 were calculated using regression estimate data from Ghana’s 2000 and 2010 census, including data on population growth rates. Age- and sex-specific distributions were obtained from the 2008 Demographic and Health Survey in Ghana²⁰ and combined with the census data to generate the population distribution for the Ashanti region by 5-year age intervals and sex. KATH provided statistics on its patient catchment area, which were used to determine the percentage of KATH cancer patients residing in the Ashanti Region. The combined data on cancer diagnosis obtained from the 3 departments were stratified by age groups (5-year intervals), sex, and cancer site for the primary tumor.

Descriptive statistics and rate analyses were conducted using SAS (Version 9.2; SAS Institute, Cary, NC). Crude annual incidence rates for each year were estimated with the number of cases per year as the numerator and the respective population estimate for that year as the denominator. Age-, sex-, and site-specific incidence rates were calculated for each year. Crude incidence rates for the 3-year period were calculated with the total number of cases identified from 2008-2010 as the numerator, and the person-time followed for the period as the denominator. Age-standardized incidence rates were calculated using the World Health Organization (WHO) world standard population estimates.²¹ Finally, observational comparisons of crude and site-specific incidence rates were made with other West African countries. Côte d’Ivoire, Niger, and Nigeria were chosen for comparison purposes due to availability of incidence data,³ similarities in age structure and distribution,^20,22-24 and geographic proximity to Ghana. While we used the term “incidence,” other terms that may reflect the actual situation in Ghana might be “minimally-reported cases” or “working estimates.”

Results

A total of 4,012 individual cancer cases were identified from the 3 departments. Figure 1 provides details on the numbers of unique cases identified in each department and the overlap of cases between departments. The largest numbers of cases were seen in the pathology department only (35.1%), followed by both pathology and oncology departments (20.1%), and surgery only (18.1%). The majority of cases (64.9%) were females. The most frequent cancers identified were breast (22.4%), cervix (16.4%), prostate (14.7%), and head and neck (10.3%).

Figure 1.

The total crude incidence rate for the Ashanti region over the 2008-2010 period was 29.5 cases per 100,000. Females had an overall higher total incidence rate (37.0 per 100,000) compared with males (21.3 per 100,000). This was consistent over each of the 3 years as well as the total for the 3-year period.

The total crude incidence rates per year increased slightly over time, from 25.5 per 100,000 in 2008, to 30.8 per 100,000 in 2009, and 31.9 per 100,000 in 2010. The incidence rate per year for females varied similarly to the total, from 32.5 per 100,000 in 2008 to 36.9 per 100,000 in 2009, and 41.5 per 100,000 in 2010. Moreover, the total incidence rate for males increased from 17.9 per 100,000 in 2008 to 24.3 per 100,000 in 2009, and then decreased slightly to 21.5 per 100,000 in 2010.

Table 1 displays the crude age-specific incidence rates for the 3-year period by sex. As expected, the incidence rates increase with increasing age. In this population, however, the incidence rate begins to increase noticeably around age 40-44 for both sexes. For females, incidence rates begin to increase substantially at even younger ages, around age 30-34, whereas for males, incidence rates increase notably starting at older ages, around 55-59 years. Interestingly, there appear to be significant increases in crude incidence rates for both sexes at ages older than 70 years compared to age groups younger than age 70.

Table 1.

Crude Age-Specific Cancer Incidence Rates (per 100,000) in the Ashanti Region (2008-2010) by Sex

Age	Male	Female	Total
Groups (males and females)	(n=1408)	(n=2604)	(n=4012)
0-4 (45,39)	4.7	4.4	4.6
5-9 (41,34)	4.0	3.5	3.8
10-14 (33,27)	3.6	3.1	3.3
15-19 (26,37)	3.6	5.4	4.5
20-24 (36,57)	7.1	9.5	8.5
25-29 (54,81)	12.0	13.7	13.0
30-34 (42,139)	11.4	31.5	22.2
35-39 (64,203)	17.3	47.5	33.8
40-44 (69,242)	25.7	73.9	52.3
45-49 (83,319)	30.9	109.3	72.6
50-54 (80,342)	36.1	114.4	82.4
55-59 (98,249)	60.7	129.6	100.3
60-64 (126,203)	81.5	135.8	108.0
65-69 (110,126)	102.3	110.6	106.6
70-74 (221,242)	219.1	200.0	209.1
75-79 (155,125)	288.2	195.1	224.8
80+(125,139)	206.6	162.7	173.4
Total	21.3	37.0	29.5

Table 2 shows the total incidence rates for the 3-year period by cancer site, revealing that the most common cancers in terms of crude incidence rate were breast, cervix, prostate, and head and neck cancers. Breast cancer was the most frequent cancer overall, with an incidence rate of 12.6 per 100,000 for females. The second most common cancer among females was cervical cancer, with an incidence rate of 9.2 per 100,000. For males, prostate cancer was the most common with an incidence rate of 8.8 per 100,000. Head and neck cancers were common among both sexes. The overall incidence rate for head and neck cancers was 2.9 per 100,000 for both sexes combined, 3.5 per 100,000 for males, and 2.5 per 100,000 for females. Other cancers that occurred at low frequencies include cancers of the corpus uteri, ovary, bladder, colorectum, stomach, thyroid, kidney, skin, and lymphomas. Due to unavailability of detailed diagnostic data, many skin cancers and lymphomas were unspecified. Liver cancer is present at low rates (0.20 per 100,000 persons per year for both sexes) and other cancers such as leukemias and multiple myeloma were not found in these departments.

Table 2.

Crude Cancer Incidence Rates (per 100,000) in the Ashanti Region (2008-2010) by Site and Sex^*

Site	Male	Female	CrudeRate/ 100,000
Head/neck
Head/neck total	3.5	2.5	3.0
Nasopharynx	0.0	0.0	0.0
Other phayrnx	0.0	0.0	0.0
Larynx	0.5	0.1	0.3
Esophagus	0.1	0.0	0.0
Lip/oral cavity	1.2	0.8	1.0
Other head/neck	1.4	1.0	1.2
Stomach	0.7	0.7	0.7
Colon/rectum	1.1	1.1	1.1
Anus	0.1	0.1	0.1
Liver	0.2	0.2	0.2
Gallbladder	0.0	0.0	0.0
Pancreas	0.0	0.0	0.0
Lung	0.1	0.0	0.1
Bone	0.2	0.2	0.2
Skin	0.7	1.2	1.0
Breast	2.0	12.6	12.6
Vulva	–	0.4	0.4
Vagina	–	0.4	0.4
Cervix uteri	–	9.2	9.2
Corpus uteri	–	2.3	2.5
Ovary	–	1.3	1.3
Penis	0.2	–	0.2
Prostate	8.8	–	8.8
Testis	0.2	–	0.2
Kidney	0.5	0.6	0.5
Bladder	1.1	1.2	1.1
Eye	0.7	0.5	0.6
Brain	0.1	0.1	0.1
Thyroid	0.2	0.7	0.5
Hodgkin disease	0.2	0.2	0.2
Non-Hodgkin lymphoma	0.1	0.2	0.1
Lymphoma, unspecified	0.4	0.3	0.4
Sarcoma	0.2	0.1	0.1
Abdomen	0.2	0.1	0.1
Small intestine	0.1	0.1	0.1
Teratoma	0.1	0.1	0.1
Other/unknown site	1.9	1.7	1.8
Total	21.3	37.0	29.5

^*n=4012.

Not applicable or not available indicated by dash.

Tables 3 and 4 compare the overall crude cancer incidence rates for males and females and by site for the Ashanti region with other countries in West Africa (Côte d’Ivoire, Niger, and Nigeria). As seen in Table 3, the crude total cancer incidence rates for Ghana as estimated for this study (female: 37.0 per 100,000; male: 21.3 per 100,000) fall within ranges seen for the other West African countries. Table 4 indicates that the higher incidence rates estimated for females by this study are consistent across other West African countries. Also, these countries share similar common cancers; in particular, breast, cervix, and prostate cancers are common to all 4 of these West African countries. In contrast, other common cancers seen in Côte d’Ivoire, Niger, and Nigeria were found at lower frequencies or not at all in the 3 KATH departments. These include liver cancer, lung cancer among males, Kaposi sarcoma, non-Hodgkin lymphoma, leukemia, and multiple myeloma.

Table 3.

Comparison of Crude Total Cancer Incidence Rates (per 100,000) by Sex among West African Countries^*

	Ghana, Ashanti Region	Cote d’Ivoire	Niger	Nigeria
Male	21.3	18.7	32.5	32.6
Female	37	23.3	48.8	42.6

^*Crude incidence data from IARC Scientific Publication No. 153, Cancer in Africa: Epidemiology and Prevention, 2003. Cĉte d’Ivoire: Abidjan Cancer Registry, 1995-1997; Niger: Cancer Registry of Niger, Niamey, 1993-1999; Nigeria: Ibadan Cancer Registry, 1998-1999.

Table 4.

Comparison of Crude Cancer Incidence Rates (per 100,000) for the Ashanti Region by Site Compared with Other West African Countries^*

Cancer Site	Ghana, Ashanti Region		Cĉte d’Ivoire		Niger		Nigeria
	Male	Female	Male	Female	Male	Female	Male	Female
Mouth	0.9	0.7	0.3	0.3	0.5	0.5	0.6	0.1
Salivary gland	0.2	0.2	2.0	–	0.3	0.2	0.3	0.3
Nasopharynx	0.0	0.0	0.2	0.0	0.1	0.0	0.8	0.4
Other pharynx	0.0	0.0	0.2	0.1	0.4	0.1	0.2	0.1
Oesophagus	0.1	0.0	0.2	0.0	0.4	0.4	0.5	0.1
Stomach	0.7	0.7	0.8	0.5	0.7	0.8	0.6	0.7
Colon/rectum	1.1	1.1	0.8	0.6	2.0	1.3	2.0	1.6
Liver	0.2	0.2	2.7	1.0	7.3	3.6	3.8	1.3
Gallbladder	0.0	0.0	–	–	0.0	0.1	0.1	0.1
Pancreas	0.0	0.0	0.2	0.1	0.4	0.4	0.4	0.5
Larynx	0.5	0.1	–	–	0.6	0.2	0.7	0.1
Lung	0.1	0.0	1.0	0.2	1.2	0.1	0.3	0.1
Bone	0.2	0.2	–	–	2.0	1.3	1.3	0.5
Skin, unspecified	0.6	0.8	–	–	1.6	1.3	0.8	0.6
Melanoma of skin	0.1	0.3	0.1	0.1	0.2	0.2	0.3	0.3
Kaposi sarcoma	0.0	0.0	1.4	0.4	0.3	0.2	0.2	0.1
Breast	–	12.6	0.2	5.9	0.4	9.7	0.5	15.0
Vulva	–	0.4	–	–	–	0.4	–	0.5
Vagina	–	0.4	–	–	–	0.3	–	0.3
Cervix uteri	–	9.2	–	5.5	–	9.3	–	10.4
Corpus uteri	–	2.3	–	0.3	–	2.6	–	1.1
Ovary	–	1.3	–	1.1	–	2.8	–	2.0
Penis	0.2	–	0.1	–	0.0	–	0.0	–
Prostate	8.8	–	2.8	–	2.2	–	7.7	–
Testis	0.2	–	–	–	0.1	–	0.2	–
Kidney	0.5	0.6	0.3	0.2	0.8	0.6	0.2	0.6
Bladder	1.1	1.2	0.5	0.2	1.6	1.4	1.1	0.2
Eye	0.7	0.5	0.1	0.3	0.8	0.4	1.0	0.5
Brain, nervous system	0.1	0.1	0.2	0.3	0.3	0.2	0.5	0.7
Thyroid	0.2	0.7	0.1	0.4	0.2	0.7	0.4	0.6
Lymphoma, unspecified	0.4	0.3	–	–	–	–	–	–
Hodgkin lymphoma	0.2	0.2	0.5	0.3	0.5	0.1	0.5	0.1
Non-Hodgkin lymphoma	0.1	0.2	1.9	1.7	2.4	2.1	3.4	1.7
Multiple myeloma	–	–	0.2	0.1	0.2	0.1	0.1	0.3
Lymphoid leukemia	–	–	–	–	1.0	0.9	0.1	0.1
Myeloid leukemia	–	–	–	–	0.5	0.8	0.6	0.2
Leukemia, unspecified	–	–	0.7	0.7	0.0	0.2	0.2	0.1
Other and unspecified†	2.4	2.1	–	–	4.1	5.4	2.6	1.3
All sites‡	21.3	37.0	18.7	23.3	32.5	48.8	32.6	42.6

^*Crude incidence data from IARC Scientific Publication No. 153, Cancer in Africa: Epidemiology and Prevention, 2003. Cote d’Ivoire: Abidjan Cancer Registry, 1995-1997; Niger: Cancer Registry of Niger, Niamey, 1993-1999; Nigeria: Ibadan Cancer Registry, 1998-1999.

^†For Niger, and Nigeria, rate may not reflect total of “other and unspecified“ as seen in this table.

^‡Excluding non-melanoma skin cancer for Niger and Nigeria. Not applicable or not available indicated by a dash.

The total age-standardized incidence rate was 41.9 per 100,000. For females, the age-standardized incidence rate was 50.8 per 100,000. Finally, the age-standardized incidence rate for males was 31.7 per 100,000.

Discussion

This study revealed the following interesting findings. First, this study highlights the importance of developing strategies tailored to the local context in order to enhance the efficiency of cancer registration starting from earnest local efforts across all departments and identifying the major local registrars and health-care professionals involved in the registry. This is clear in our work, which, for the first time, linked patient records across departments and thus identified many cases of cancer when they were not explicitly diagnosed as such, as from surgical cases only, for example. Second, our results show the need to fully understand the health-care system and structure when developing a cancer registry. This is evidenced by the distribution of cancers by both age and cancer site in this study. In our study, some patients tended to report older ages to utilize health insurance benefits available only to older patients. Also, some cancer sites such as leukemia were under-reported because of lack of diagnostic facilities at the study hospital. Finally, comparison of our results to previous estimates for Ghana and other similar countries provide strong insight into the reasonable nature of our estimates and point to potential gaps in our data.

Linking cancer cases across hospital departments is essential to ensure quality and completeness in a registration system. The departmental-based registry at KATH is common in Africa and in many under-resourced regions, and currently only collects data on cancer cases seen within the medical oncology/radiotherapy department and is therefore unable to accurately describe the cancer burden at the hospital. The process currently in use is based on the assumption that all cancer patients who are seen at the department of surgery and/or diagnosed by the department of pathology will subsequently receive chemotherapy and/or radiotherapy. Our study found 53.2% of cancer cases were seen only in the pathology or surgery departments and thus no further linkage to therapy occurred because the patient declined to receive treatment or a referral was not made. Our results emphasize the importance of capturing cases in all relevant departments based on the dynamics of patient referrals between departments in local health-care systems. This process that we conducted for the first time in Ghana is crucial at least until formal and thorough population-based registries are in place.

The need for locally-tailored strategies for cancer registration is also evidenced by our work in identifying cancer cases that were seen, but not explicitly diagnosed, in the surgery department. Due to the incomplete database of surgery patients and lack of separate databases or logbooks for oncologic surgeries, cancer cases were not easily identified. To abstract cancer cases, we developed a classification strategy to rank cases on their likelihood of being cancer and the coauthors reviewed the data systematically. Without this system, our study would not have captured the 725 (18.1%) patients with suspected cancer who were seen only by the surgery department. Thus, although not without its complexities of design, having identified such a high percentage of cases at surgery alone again points out the diverse nature of the patients’ navigation schemes within the KATH system.

Understanding the intricacies of the local health-care system is essential to effective cancer registration, which was made clear through our attempts to age-standardize our estimated incidence rates. The age-standardized incidence rates reported in the results (total: 41.9 per 100,000; female: 50.8 per 100,000; male: 31.7 per 100,000) are notably low compared to GLOBOCAN’s estimates for Ghana (total: 109.5 per 100,000; female: 125.5 per 100,000; male: 93.8 per 100,000) and age-standardized estimates for other sub-Saharan African countries.^3,17 Our lower estimate of the age-standardized rates is due to the large proportion of cases aged 70 and older. It is important to note that the National Health Insurance Scheme in Ghana provides free health care without premiums for patients aged 70 and older.²⁵ Based on one of the coauthors experience in managing cancer patients at KATH (EO-B), patients often report their age as 70 and older to avoid payment of premiums for health care, which is seen in this data. As patients often do not have birth records, it is difficult to validate patients’ age reporting and so the age claimed by the patient is the one reported in the medical records.

Our findings with respect to the distribution and crude cancer incidence rates in the Ashanti region are similar to estimates reported elsewhere for Ghana and other comparable countries. First, our study confirmed previous estimates and observations that cancer incidence rates are higher among females. The International Agency for Research on Cancer (IARC) and other previous studies in Ghana^13-15,17 estimated that females have overall greater morbidity and mortality from cancer than males. Our estimated crude incidence rates for Ghana are similar to the estimates for other West African countries and indicate that females have higher rates of cancers in these countries as well. This is likely a real difference, but the magnitude of the differences between sexes could also arise from the available treatment options for cancer in Ghana. Treatment for breast and cervical cancer at KATH are partially covered by the National Health Insurance Scheme whereas other cancers are not covered.²⁵ This could partly explain the greater representation of women with cancer in treatment at KATH that were captured in this study.

This study found relatively high crude incidence rates for breast, cervix, and prostate cancers. Previous studies^13-15,17 are in agreement these cancers are common in Ghana. Furthermore, comparisons with other West African countries indicate that breast, cervix, and prostate cancers are common in similar countries as well.³

The overall crude rates for males and females as well as the majority of site-specific rates as estimated in this study are quite reasonable when compared with similar West African countries. The crude rates observed in this study fall within ranges reported for other West African countries with similar age distributions to Ghana.³ Previous results from other West African countries support this observation.³

In contrast, such comparisons also show potential gaps in our study. Other estimates for Ghana and other similar sub-Saharan African countries found lung cancer in males and liver cancer among both sexes to be more frequent than indicated in our data. This is likely explained by the fact that this study did not review autopsy records or those in other departments such as internal medicine or laboratory departments. The lower rates of Kaposi sarcoma and non-Hodgkin lymphoma relative to other West African countries could be explained by the lack of availability of detailed diagnostic data for many skin cancers and lymphomas. Also, Ghana’s HIV prevalence is low relative to both Côte d’Ivoire and Nigeria which could partly explain the lower rates of HIV-related malignancies like Kaposi sarcoma and non-Hodgkin lymphoma, although Niger’s reported HIV prevalence is lower than Ghana’s.^26-29 Finally, this study did not identify certain cancers such as leukemia and multiple myeloma that were identified in comparable countries. Again, this is due to incomplete data collection across all relevant departments.

This study has a number of strengths. We built upon existing registration infrastructure at KATH to enhance coordination of case reporting between hospital departments and improve data collection. This study was the first at KATH to provide complete data on all cancer cases seen in 3 departments over a 3-year period. Using the most recent available census data, this study provided the first estimates of incidence based on in-country data.

We also addressed numerous methodological issues in cancer registration that are likely to be highly applicable in other resource-limited settings. For example, KATH lacks a uniform identification system for patients, so there was not one variable common across departments to facilitating case linkages. Instead, we devised a variety of strategies based on the available data to enable these linkages. Also, we identified the most available and efficient data sources available to each department. As these different sources had varying degrees of completeness and quality, we identified other sources in each department for validation purposes. Finally, we devised strategies to identify cases of cancer when they were not explicitly diagnosed in the available data, as evidences by our methods for the surgery department. This study also had a few limitations. There is the potential for underestimation in the incidence rates provided in this study. We did not review cases in all departments and so may have missed cases of cancer only seen in pediatrics, laboratory medicine or found in death registration and/or autopsy records. In addition, KATH does not see all existing cancer cases in the Ashanti region, so those cases that may have only been seen at private health centers, by traditional or alternative practitioners, or that died before diagnosis were missed by this study. We did, however, attempt to compensate for these missed cases using hospital- and census-based statistics to estimate that we missed 25% of cancer cases in the Ashanti region and used this estimate in our calculations.

There is also a potential for errors in our data or cases that were missed in the 3 departments reviewed. The surgery database did not contain explicit diagnoses of cancer and so each surgery case was categorized by its likelihood of being cancer. These cases were, however, reviewed carefully by medical practitioners and matched to both pathology and oncology to ensure accuracy. We also found data entry errors in the oncology database obtained from the departmental-based registry. We attempted to correct this with a systematic review of patient folders, but could not locate 5.4% of folders for validation purposes. In addition, certain cancer types such as leukemia and multiple myeloma are not reported because of lack of collection from the department of laboratory medicine.

In summary, this study provided the first data on cancer cases seen in multiple departments over multiple years at KATH. We estimated crude cancer incidence rates of the “minimally-reported cases” or “working estimates” for the Ashanti region that are comparable to other West African estimates and we confirmed that breast, cervix, and prostate cancers are common in this population. The methodology and results from this study emphasize the need to tailor strategies for cancer registration to the local context with an in-depth understanding of the overall healthcare system and structure. The next steps in enhancing cancer registration at KATH include expanding data collection to all other relevant departments and collaborating with data sources outside of KATH, such as private treatment centers. Future studies should utilize the insights from this study that may be applicable to other resource-limited settings. Future studies should also consider additional methods to identify the proportion of missing data such as capture-recapture analyses³⁰ and verification of completeness of registries.³¹ Future studies should also aim to verify the reliability of the census. Methods for inclusion of non-conforming departments into the registration system must consider steps, barriers, and proposed solutions for ensuring participation in the registration process. Periodic utilization of the registry data by senior and junior physicians and faculty of departments could enhance the interest in supporting cancer registration. Utilization of registry data for periodic reports, theses, and dissertation materials could also help in increasing the support of registration by non-conforming departments.