Public Sector Radiological Resources and Utilization Patterns in the Western Cape Province of South Africa

Background There are marked disparities in radiological resources globally, particularly between metropolitan and rural populations. The World Health Organization (WHO) suggests that 90% of low- and middle-income country (LMIC) imaging needs can be addressed by one X-ray and ultrasound machine for every 50000 people. However, this gure is untested, as limited work on radiological resources and service utilization patterns globally, particularly in LMICs exists. The aim was to analyze provincial radiological service in a middle-income country. Methods An institutional review board-approved retrospective audit of radiological data for the public healthcare sector of the Western Cape Province (WCP) of South Africa (SA) for 2017, utilizing databases of the WCP Department of Health and Stats SA. We conducted population-based analyses of imaging equipment, personnel, and service utilization data for the province, metropolitan and rural areas. Results Metropolitan population density exceeds rural (1682 vs 19 people/km 2 ; 89:1). Rural imaging facilities by population are double the metropolitan (19 vs 11/10 6 people). Provincially, there are 36 X-ray and 18 ultrasound units/10 6 people. Rural X-ray (39.3 vs 33.6/10 6 people), ultrasound (24.7 vs 14.5/10 6 people) and mammography (14 vs 5 units/10 6 women > 40 years) resources exceed metropolitan by 17, 70 and 180 percent, respectively. Metropolitan personnel resources by population (n = 112 vs 53/10 6 people) and equipment unit (1.7 vs 0.7/10 6 people) are double the rural. Provincial imaging studies totalled 1.2778 million, averaging 262 examinations/10 3 people and 1.3 investigations/patient. Radiography ultrasound together studies. imaging

Conclusion Our ndings support the WHO contention that approximately ninety percent of a population's diagnostic imaging needs can be met by plain radiography and ultrasound and underscore the complexity of achieving equitable utilization of services between rural and metropolitan areas.

Background
Diagnostic imaging is currently recognised as a key building block of any healthcare system and is considered essential for effective primary health care. (1)(2)(3)(4)(5)(6)(7)(8)(9)(10) However, radiological services are expensive, labour-intensive, require high levels of technical expertise and are thus amongst the leading drivers of escalating medical costs. (11)(12)(13)(14)(15) The expense of modern diagnostic imaging has the potential to compound existing worldwide inequalities in access to radiological services. At one end of the spectrum are high-income countries with an aging population, a high burden of non-communicable diseases and a relative abundance of radiological resources. (11,16) At the other end are more than half the world's population, living in low-and middle-income countries (LMIC) by World Bank criteria, with diseases predominantly related to poverty, and with only limited access to basic imaging resources. There are marked disparities in radiological equipment resources amongst countries in the same World Bank economic grouping, as well as between geographical regions and healthcare sectors within the same country. The greatest divide is between metropolitan and rural populations. (5,9,17−23) The reduction of inequality and the promotion of universal health coverage are two key United Nations 2030 Sustainable Development Goals (SDGs). (18) The World Health Organization (WHO) suggests that approximately ninety percent of all LMIC imaging needs can be addressed by one plain radiographic unit and a basic ultrasound machine for every 50,000-100,000 people, or 20 units each per million people.
However, this gure is untested, as there has been limited in-depth work on radiological resources and service utilization patterns globally, particularly in LMICs. Although the WHO has published national estimates of high-end medical imaging resources based on questionnaire surveys of member countries, these data do not include basic equipment such as plain radiography and ultrasound units. (24) Furthermore, while Stellenbosch University in Cape Town, South Africa, is coordinating a project to collate detailed data on registered radiological resources in Southern, East and West Africa, these data do not include ultrasound equipment, or utilization data. (17,22,23,25) Despite the European Commission conducting detailed surveys of the use of ionizing radiation for medical purposes, data are not correlated with imaging equipment and personnel resources. (26) Although there has been an analysis of differential geographical utilization of radiological services in Norway, (27) the need exists for such an analysis in LMICs, incorporating all components of the so-called "imaging enterprise". (14) The healthcare infrastructure of the Western Cape Province (WCP) of South Africa (SA) is ideal for such an analysis.
SA is one of ve upper middle-income countries in sub-Saharan Africa. It has adopted the District Health System (DHS) for delivery of comprehensive primary care to its peoples. (28) Health services are devolved to the country's nine provinces. The WCP is SA's southernmost province. It has six administrative districts.
The City of Cape Town Metropolitan District is surrounded by ve sprawling rural Districts ( Fig. 1), (29) in which the main economic activities are agriculture and recreation.
WCP health services, including diagnostic imaging, are managed along geographic lines, and strati ed as metropolitan or rural streams. (30) Mirrored, tiered referral pathways exist in each system. First access to imaging is generally at Community Day Centres or Community Health Centres with sequential referral to District, Regional, and Central Hospitals, the latter being university-a liated tertiary-level teaching institutions.
The WCP has a provincial-wide digital imaging platform, with picture archiving and communication system (PACS)-integration of services at the various levels of care. This facilitates access to all imaging across the platform and eliminates unnecessary duplication of services. Clinicians at lower tiers of service can also get assistance in interpretation of investigations via the digital platform.
The Medical Imaging Services Sub-Directorate within the Directorate of Health Technology in the WCP Department of Health (DoH) is responsible for stringent collation of all data pertaining to provincial imaging, including the number and location of all radiological equipment units and personnel, as well as the utilization of services at each facility. These public sector data provide unique insights into the utilization of imaging services in a high middle-income country.
The aim of this study was to analyse the resources and utilization patterns of the WCP radiological platform, including those features differentiating metropolitan from rural services.

Methods
This was a retrospective audit of diagnostic imaging data for the public healthcare sector of the WCP of SA for 2017. All radiological details were extracted from the databases of the Medical Imaging Services Data on radiological equipment units and examinations performed were captured on a customized spreadsheet and strati ed by imaging modality, healthcare facility and provincial service stream (metropolitan/rural). Plain radiography and uoroscopy units were further subdivided into xed and mobile units. Plain radiographic examinations were categorised as chest radiographs (CXR) and general radiographs (GR). All analyses were for the province as a whole and for the metropolitan and rural areas.
The number of radiology equipment units per million people was calculated by modality. The quantum of radiological examinations for each modality was assessed per thousand people, as well as per thousand patient engagements. Imaging examinations per equipment unit were calculated by modality.
The numbers of registered personnel in the categories of diagnostic radiographer, sonographer, registrar, and radiologist were collated and categorized by healthcare facility and service stream.

Provincial overview
Approximately two-thirds of the WCP population live in the City of Cape Town Metropolitan District which constitutes just two percent of overall provincial land area. Metropolitan population density (1682 people/km 2 ) exceeds rural (19 people/km 2 ) by a factor of almost ninety (89:1).

Imaging facilities
There are sixty-eight (n = 68) provincial imaging facilities, all with plain radiography and fty three (n = 53; 78%) with ultrasound services. The metropolitan and rural areas have 35 and 33 imaging facilities, respectively. As a result, rural resources by population are almost double the metropolitan (19 vs 11/10 6 people).
District Hospitals (n = 35, 51%) and Community Day Centres (n = 14, 21%) together account for almost three-quarters (n = 49, 72%) of all facilities, with Community Day Centres predominant (n = 9/35, 27%) in the metropole and District Hospitals (n = 26/34, 76%) in the rural areas. All Central (n = 3) and Community Health Centres (n = 10) are in the metropole, while the six Regional Hospitals are equally distributed between the two areas.
There are 36 radiography and 18 ultrasound units per million people overall. However, the rural radiography (39.3 vs 33.6/10 6 people) and ultrasound (24.7 vs 14.5/10 6 people) resources exceed those in the metropole by 17 and 70 percent, respectively. Additionally, rural access to mammography is almost triple that in the metropole (14 vs 5 units/10 6 women > 40 years).

Imaging staff
Almost eighty percent (n = 348; 79%) of the 443 provincial imaging personnel are in the metropole and more than three-quarters of all staff (n = 339; 77%) are radiographers.
Metropolitan personnel resources by population (n = 112 vs 53/10 6 people) and equipment unit (1.7 vs 0.7/10 6 people) are more than double the rural. Rural areas have more equipment units than personnel.
The potential average annual workload (excluding ultrasound) by radiologist at provincial, metropolitan and rural levels was 35820, 28910 and 63825 studies, respectively.
Overall population-based utilization of imaging services across the modalities was 30 percent higher in the metropole (279 vs 214 studies/10 3 people), with mammography (24 vs 5 studies/10 3 woman > 40 years; 517%) and CT (21 vs 6/10 3 people; 380%) recording the highest differentials and plain radiography utilization (203 vs 171/10 3 people; 19%) the lowest. However, overall patient-based utilization of services in the rural areas exceeded that in the metropole by 12 percent (1353 vs 1209 studies/10 3 patients) re ecting increased plain radiography and ultrasound usage.

Imaging equipment utilization:
Provincial equipment units performed an average of 3660 studies annually. Average metropolitan equipment outputs by unit exceeded those in the rural area by more than fty percent (4114 vs 2739/unit, 50%). The highest differences were in mammography (n = 6198 vs 458/unit) and uoroscopy (n = 450 vs 140/unit).

Discussion
To our knowledge, this is the most detailed analysis of the usage of radiological services in a low-or middle-income country to date. It broadly reviews a provincial imaging platform, while allowing a better understanding of key differences between metropolitan and rural service provision and utilization. It therefore represents a seminal work in the eld, that contributes signi cantly to discourses on equitable access to basic healthcare, universal health coverage and appropriate utilization of diagnostic imaging. It can serve as a benchmark resource and stimulate further work in this domain.
A key nding was that, in a population with access to the full range of modern imaging modalities, plain radiography and ultrasound examinations constituted 89% of all studies, with minimal (8%) differential between metropolitan and rural areas. This provides compelling support for the WHO contention (4,25,31−34) that approximately ninety percent of all imaging needs can be met by one plain radiography and one ultrasound machine for every 50 000 people, or 20 units each per million people. WCP plain radiography resources (36 units/10 6 people) are fty percent above the WHO benchmark, while the availability of ultrasound equipment (18 units/10 6 people) is aligned with this guideline.
Provincial utilization of imaging services totalled 262 patient encounters/10 3 people, or approximately one engagement for every four people. However, the 30% difference between metropolitan (279/10 3 people) and rural (214/10 3 people) engagements underscores the challenge of equitable service provision to sparse rural populations, given that total rural equipment units/10 6 people exceeded metropolitan by a differential of 15%. Furthermore, despite the rural areas having 17%, 70% and 263% greater equipment resources/10 6 people for radiography, ultrasound and mammography, respectively, corresponding utilization was 72%, 45% and 7% less than metropolitan. This highlights the truism that access to services is not simply a question of the equipment to population ratio.
By contrast, rural plain radiography and ultrasound examinations per patient were 23% and 10% higher than metropolitan, respectively. The increased investigation of rural patients may re ect relative clinical inexperience on the part of doctors working at smaller peripheral facilities, many being in their postinternship year (35) . It may also re ect a mindset that patients who have travelled great distances at considerable expense merit the full ambit of available investigations.
This work provides novel insights into the differential equipment workload across the modalities and regions. CT (n = 5835) and plain radiography (n = 5377) achieved the highest average annual outputs per unit across the province, while uoroscopy recorded the lowest (n = 347), being approximately 17-fold below that of CT. The limited use of uoroscopy re ects declining global trends and suggests that provincial policy on uoroscopic service provision merits review. Of note, less than 4 uoroscopic investigations/10 3 people were performed across the province in the review period.
Rural equipment utilization by unit was less than metropolitan across all modalities. The smallest differential was in plain radiography, where the average rural unit (n = 4365) achieved 72% of metropolitan output (n = 6058), whilst the greatest difference was in mammography, with rural output (n = 458) a mere 7% of metropolitan (n = 6198). The optimal combination of equipment, personnel, and hours of operation for rural facilities remains a conundrum. WCP rural equipment resources currently exceed imaging personnel by 46% (n = 139 vs 95), likely contributing to decreased equipment utilization.
A recent study of Zambian radiological equipment and personnel resources (22) found more than three Zambian diagnostic radiographers per equipment unit, nationally, and at least two radiographers per unit at provincial level, even in the most sparsely populated regions. This study was limited by inability to accurately assess the ultrasound workload of the various categories of personnel, since ultrasound outputs were not strati ed by type of healthcare worker. In the WCP, ultrasounds in the radiological domain are performed by registrars, sonographers, and dual-quali ed radiographers, the latter performing both radiography and ultrasound. Dual-quali ed radiographers and sonographers practice independently, while registrars generally report under consultant supervision. In the future, WCP ultrasound data would be enhanced by allocating outputs by category of healthcare worker. A further limitation is that many obstetric and gynaecological ultrasounds are not performed within the radiological department and are not included in this analysis.
Even allowing for uncertainty in ultrasound outputs by personnel category, this work provides important insights into the reporting workload generated by other modalities. Excluding ultrasounds, 722761 metropolitan and 319123 rural studies required reporting, including 630078 (87%) metropolitan and 305529 (96%) rural plain radiographs. This translates to a potential annual workload of 63 825 and 28 910 studies per radiologist in the rural and metropolitan areas, respectively.
Of note, most metropolitan radiologists have university a liations, and thus both clinical and academic commitments, while rural consultants have exclusively clinical commitments. It is acknowledged that supervision of trainees impacts the clinical output of academic radiologists (36,37) . There is thus wide acceptance that the clinical load of the academic radiologist should be capped (12,37,38) . The Royal Australian and New Zealand College of Radiologists (RANZCR) recommends a threshold of 12 000 examinations per year (12) . It has previously been shown that the increasing clinical workload of the WCP academic radiologist over the past decade has necessitated the prioritization of reporting of special investigations, such as uoroscopy, CT, mammography, MRI and angiography, with resultant decreased capacity for plain radiograph reporting (39) . The average special investigation workload of the metropolitan radiologist in the review period was 3707 cases, representing a manageable annual workload when combined with selective plain radiograph reporting and educational commitments. This is the rst detailed appraisal of the potentially overwhelming workload of the rural radiologist. If all imaging studies performed in the rural areas were to be formally reported by a radiologist (n = 76131) this would be more than 5-fold the average caseload (n = 14900) documented for general radiology consultants working in the United States (40) and more than nine times that of general radiologists (n = 8171) in the United Kingdom (UK). (41) It is clearly not a realistic expectation. In the rural areas of the WCP, the same pragmatic reporting policy applies, with mandatory special-investigation reporting, but selective plain radiograph reporting, making for a more realistic and sustainable consultant radiologist workload. The average special investigation workload of the individual rural radiologist in the review period was 2719 cases.
The challenge of meeting the ever-increasing demand for radiology reporting is not unique to LMICs. (42,43) A recent report showed that 97% of UK radiology departments were unable to meet all reporting requirements, (42) that most delays involved plain radiograph reporting and that only 12% of trusts reported all radiological examinations. There were instances where trusts were not reporting the bulk of in-patient and emergency department plain radiographs.
Our study provides a compelling argument for upscaling undergraduate medical education programs in the interpretation of plain radiographs and the performance of basic ultrasound examinations. Equipped with such skills, medical graduates would be well-placed to address ninety percent of provincial reporting needs. Additionally, competence in basic interpretation of plain radiographs would address the current reporting void in this modality. Our study also shows that educational initiatives in CXR interpretation would be particularly bene cial, since almost half of all plain radiographic studies are CXR.
The increasing population dose from medical exposures to ionising radiation is a global concern. (26) This study provides important utilization data for the WCP. This facilitates international comparisons, notwithstanding the WCP having a younger population and a higher prevalence of tuberculosis and HIV than the well-resourced countries for which comparative data are available. WCP plain radiography utilization was compared with 2014 data from 35 European countries. (26,44) Hungary (n = 581) and Romania (n = 143) have the highest and lowest plain radiographic examinations per 10 3 people, respectively. The WCP (n = 192) utilization is at the lower end of this range. WCP plain radiograph utilization is most closely aligned with Sweden (n = 201/10 3 people) and Denmark (n = 180/10 3 people) To the best of our knowledge, the only comparable analysis of geographic variation in the utilization of radiographic services was conducted in Norway in 2002. (27) The study showed that Oslo (1532.7) and Finnmark (1.6) were the counties with the highest and lowest population densities, respectively. (45) Plain radiographic utilization in Oslo (921 studies/10 3 people) was double that in the more sparsely populated Finnmark (459 studies/10 3 people). By comparison, there was a 1:1.8 differential in plain radiographic utilization between metropolitan (203/10 3 people) and rural (171/10 3 people) areas of the WCP.
Our analysis also allowed comparison of WCP CT utilization with global data in 2014. (26,44) The United States demonstrated the highest CT utilization (n = 271) and Costa Rica the lowest (n = 34) examinations per 10 3 population. CT utilization in the WCP (16/10 3 people) is less than half that of Costa Rica which is classi ed as an upper middle-income country by the World Bank. (46) WCP is below the range of reported international CT utilization gures.

Conclusion
This study has shown that in a population with access to the full range of modern imaging modalities, plain radiography and ultrasound constitute 89% of all radiological studies, with minimal variation between rural and metropolitan areas. Our ndings therefore support the hitherto untested WHO contention that approximately ninety percent of a population's diagnostic imaging needs can be met by these two basic imaging modalities.
The study also highlights WCP attempts to bridge the divide between rural and metropolitan access to imaging, through the provision of an increased number of imaging facilities and basic equipment units per million people in the rural compared to metropolitan areas. However, it also underscores the complexity of achieving equitable utilization of services between rural and metropolitan areas.
The achievement of equity in all aspects of healthcare must be seen as a process, involving incremental improvements and iterative analyses that de ne progress towards the ultimate goal. Studies such as this serve to de ne the baseline and inform future interventions to enhance equity going forward. Authors' contributions RDP conceived the project. BCVZ, MMB and AF contributed to the re ning of the project concept. MMB and AF collection and collation of data. BCVZ conducted the data analysis with assistance from RDP.
BCVZ drafted the initial manuscript. RDP provided editorial input and critical revision of the manuscript for intellectual content. MMB, AF, KB and MM provided critical review and contributed to the nal manuscript. All authors read and approved the nal manuscript.