Computed tomography (CT) was first introduced in the medical field in the 1970s and has become an invaluable aid in the diagnostic workup of multiple medical and surgical conditions.¹ Computed tomography imaging also plays a role in admission patterns, with younger patients being more likely to be discharged after a CT scan and older patients more likely to be admitted.²

The use of CT has consistently increased in all medical departments, including the emergency centre. In the United States (US) and Canada, CT imaging in the emergency centre increased by almost 60% between 2005 and 2013.² A similar trend has been noticed in South Africa where CT use has doubled between 2011 and 2013 in certain areas.³

The utility of CT scans is well documented with its clinical impact in the emergency centre proven to alter management in 43% of cases.⁴ However, limited data are available with regard to what an appropriate positive yield rate is. A Malaysian audit of CT scans of the head demonstrated a 70% yield rate for non-traumatic cases.⁵ More locally, a study in a tertiary academic hospital in Johannesburg found a 61% positive yield in non-traumatic presentations. It is unclear whether higher yields represent an under-utilisation of CT scanning or if it just reflects effective judicious use.⁶

In 2015, the number of CT scanners available in South Africa was only 1.7 per million people in the public sector, compared with >40 per million people in the US.⁷ Computed tomography scanners are typically situated within regional or central hospitals in the South African public sector,⁸ necessitating the need for district-level hospitals to refer patients for diagnostic imaging. This incurs costs, both monetary through patient transfers and non-monetary due to delays in patient management. A Canadian study indicated that 84% of all patients transferred from a rural facility were specifically to receive a CT scan at the receiving facility.⁹ Similar data are not available in a South African context, but increased availability of CT scanners has the potential benefit of reducing transfer burden, decreasing delays in patient management as well as mitigating the risk of loss to follow up of outpatients because of transport and socio-economic constraints.

In August 2018, Khayelitsha Hospital became one of the first district-level hospitals in the Western Cape province to have an on-site CT scanner,¹⁰ which runs emergency and elective scans during working hours. The utility of a CT scanner at a district-level healthcare facility should theoretically be huge with many potential benefits, but there is limited data available to support this. This study describes the type, clinical indications and diagnostic yield of CT scans performed at a district-level hospital in Cape Town over a 1-year period. In addition, the prevalence of significant abnormal findings and the congruency between the initial clinical impression and the radiological diagnosis on CT were determined.

A total of 3294 CT scans were documented by the radiology department at Khayelitsha Hospital over the 12-month study period. Fifty-two scans were excluded (duplicated reports n = 40, no report or no clinical information available n = 12) and 3242 CT scans were thus analysed.

The mean age of patients scanned was 45.6 years (s.d. = 17.6) and 51.4% (n = 1666) were males. Most CT scans involved a single body region (n = 2940, 90.7%), were contrasted (n = 2639, 81.4%) and were performed on an inpatient basis (n = 2608, 80.4%). There were 250 working days in 2019 with a mean of 12.9 scans performed per working day (3236/250; s.d. = 3.5); six scans were performed during weekend hours. The number of scans performed per weekday was equally distributed (Table 1).

The head and neck area were the most frequently scanned body region (n = 1841, 52.3%) and was substantially more than the chest (n = 755, 21.4%) and the abdomen and pelvis (n = 674, 19.1%) (Table 1). The indications for CT scan were mostly for the work up of general medical conditions totalling 2151 (66.4%) of all scan requests. This includes work up of cerebrovascular accidents, delirium, seizure disorders and cardio-respiratory diseases. A total of 538 (16.6%) scans were requested for suspected malignancy and 160 (4.9%) for the evaluation of traumatic injuries (Table 1).

Table 2 summarises the distribution of departmental scan requests per patient status. Most of the scans were requested by the Emergency Medicine department (n = 1382, 42.6%) and the Department of Internal Medicine (n = 695, 21,4%). Externally requested CT scans are those scans that were booked by the local clinics and 92.5% of externally requested scans were performed on an outpatient basis (Table 2). Emergency medicine and internal medicine requested the overwhelming majority of their CT scans as inpatients (95.9% and 93.1%, respectively). All CT scans requested by psychiatry were as inpatients because of the need to exclude a general medical condition as the cause of an acute psychosis.

Most CT scans demonstrated some abnormality (n = 2710, 83.6%); however, only 2115 (65.2%) were deemed to have had a ‘positive yield’ (Table 3). Requests from the orthopaedic department had the highest positive diagnostic yield (96.4%) because most orthopaedic requests were for pre-operative planning. Surgery had an 80.9% positive yield, whereas Internal Medicine and Emergency Medicine had a positive yield rate of 67.5% and 60.2%, respectively (Table 3).

The CT diagnoses according to ICD-10 categories are presented in Table 4. To note is that 67% (71/106) of scans for a potential pulmonary embolism were positive. Aside from the normal and negative yield (non-diagnostic) scans, the greatest proportion of diagnoses were related to neoplasms (n = 467, 14.4%), circulatory (n = 451, 13.9%) and infectious (n = 419, 12.9%) (Table 4). The secondary diagnoses on CT scan are also listed in Table 4.

There was congruency between the clinical (pre-CT) and final CT diagnoses in 1610 (49.7%) cases, 1629 (50.2%) were not congruent and agreement was not applicable in three (0.1%) cases.

This study, exploring the utilisation of CT in South Africa, was unique from its point of view at the district level. However, the specific objectives of describing the CT types, the clinical indications, assessing the overall diagnostic yield and clinical impact improves our understanding of appropriate CT utilisation in general, regardless of the level of care. Furthermore, it may also offer valuable insights for enhancing efficiency and overall patient care across diverse healthcare settings.

The Canadian Institute for Health Information (CIHI) reported a national average of three CT scans per hour per scanner in 2005.¹⁵ This was almost double the average of 1.6 scans per hour in this study; however, the differences in infrastructure are stark. The reason for the lower number of scans per hour could be multifactorial, especially in our setting where the CT suite is a relatively new service. Infrastructure challenges, such as an inadequate electricity supply and unstable internet connectivity coupled with limited maintenance and servicing options lead to intermittent breakdowns and downtime of CT services. In addition, a lack of trained personnel including dedicated porters, nurses and radiographers are just some factors that can hinder the effective operation of CT scanners. An outpatient facility in the United States embarked on a process to maximise CT productivity.¹⁶ The study indicated that a single technologist (radiographer) needed to complete up to 34 tasks per patient, resulting in only scanning 2.2 patients per hour.¹⁶ The number of CT scans performed per hour increased substantially to 5.2 when there were two technologists per patient. It increased further to 7.5 when there were three technologists per patient.¹⁶

Another notable difference was in the indications for CT, which indicated that only 4.9% of indications were related to trauma, whereas in another South African study, 76% of CT scans were trauma related.⁶ Although the latter study was performed in a tertiary hospital, the main reason for the low percentage of trauma-related CT scans at Khayelitsha Hospital was because of the policy that was in place. At the time of the study, all trauma patients presenting to Khayelitsha Hospital with an indication for a CT scan were transferred to the local tertiary hospital.

The diagnostic yield of 65%, as determined in our study, aligns closely with findings from both local and international research. A Johannesburg-based study conducted in an emergency centre of a tertiary hospital previously reported a comparable positive yield of 61%,⁶ indicating consistency in diagnostic outcomes across different healthcare settings. Similarly, a Malaysian study focusing on CT scans of the head exhibited an overall positive yield of 65.8%, with a higher positive yield (71%) for non-trauma cases.⁵

Furthermore, delving into the yield of specific CT scans, recent research at the University of Iowa hospitals examined the positive yield rate of CT pulmonary angiograms in detecting pulmonary embolism. The reported yield of 10.9% was considered comparable to existing literature and deemed an acceptable positivity rate.¹⁷ In contrast, our study revealed a substantially higher positive yield rate of 67%, with 71 scans indicating pulmonary embolism out of a total of 106 requests over the 1-year period. This discrepancy may reflect the scarcity of resources and potentially indicate an underutilisation of CT scans compared to the international norm.

Half of the performed CT scans were congruent between the initial clinical impression and the CT diagnosis. Pescatori et al. found a similar congruency ratio in an Italian study. Their study evaluated the clinical effect of CT in non-traumatic chest and abdominal scans, and the clinical hypotheses were confirmed in 45% of cases while there was a major variation in diagnoses in 37% of cases.⁴ However, a normal CT scan does not imply the test was unnecessary, and similarly, non-congruency between pre- and post-CT diagnoses is not necessarily an indication of an unnecessarily performed test. Computed tomography scans are often ordered particularly to exclude an alternate or underlying diagnosis. An example is in the workup of patients with new onset seizures; a normal CT scan will have a negative yield and be non-congruent, but a very necessary diagnostic test to exclude potentially treatable causes. Similarly, young patients with mental health illnesses undergoing CT to exclude a delirium or patients newly diagnosed with cancer getting a metastatic workup, etc. It is in these cases, where the basis of the CT is to exclude a diagnosis, that the congruency may become skewed. The requesting information lists the alternate diagnosis as the one under investigation and thus in these cases, a normal scan will be non-congruent even though there may have been a low level of true clinical suspicion.

This study’s findings highlight several areas for potential improvement in patient care at both the facility and broader governmental policy levels. As mentioned, the comparatively low number of scans per day and per hour could imply that the efficiency of the system could be improved. Introducing additional personnel or implementing standardised procedures to optimise the process may aid in increasing efficiency.

The majority of scans (over 90%) requested by internal medicine and emergency medicine were for inpatients. This prompts the question of whether all these scans were necessary for inpatient care, or if there’s a need for a more efficient outpatient system for bookings and follow-ups. Keeping well patients in hospital just to get a CT scan should be a contributing factor to the overburdening of inpatient systems and might aggravate bed shortages.

Finally, hospital policy at the time resulted in the unexpectedly low number of trauma scans being performed in a community with a high trauma burden.¹² This is a contentious issue that needs further evaluation. Performing more trauma-related CT scans at the district level could alleviate the strain on emergency medical services to transport patients only for a diagnostic study. Similarly, the receiving facility can also be negatively impacted by the greater caseload of patients requiring further investigation. On the other hand, upscaling the CT service to provide a 24/7 service will be costly to the district hospital and is sure to increase the burden on an already over-burdened system. Certain specialist services, particularly neurosurgery, are also not available at the district level and a delay in definitive treatment might be detrimental to patients. A birds-eye evaluation is needed to create an optimal system that ensures cost-effective care without overburdening one facility at the expense of another.

The utilisation of CT scans at district-level healthcare facilities yields comparable positive rates to those observed in tertiary centres. This suggests that the service is appropriately utilised and potentially alleviating the burden on higher-level referral facilities. Overall, regionalisation of CT utilisation holds great potential to enhance healthcare quality and accessibility in district-level health facilities, facilitating a more equitable distribution of advanced diagnostic services, ensuring that patients in underserved areas have access to crucial imaging technology and ultimately promoting better patient outcomes. However, the relatively low number of CT scans conducted per hour indicates a complex issue influenced by multiple factors. Infrastructure challenges (e.g. inadequate electricity supply, unstable internet connectivity), limited maintenance and servicing options and a lack of trained personnel are just some barriers to consider.

Further research and dialogue are warranted to refine policies, enhance efficiency and improve patient care in the context of diagnostic imaging at the district level in South Africa and similar settings.