Direct and Indirect Effects of the Coronavirus Disease 2019 Pandemic on Private Healthcare Utilization in South Africa, March 2020–September 2021

Abstract Background The coronavirus disease 2019 (COVID-19) pandemic caused severe disruptions to healthcare in many areas of the world, but data remain scarce for sub-Saharan Africa. Methods We evaluated trends in hospital admissions and outpatient emergency department (ED) and general practitioner (GP) visits to South Africa’s largest private healthcare system during 2016–2021. We fit time series models to historical data and, for March 2020–September 2021, quantified changes in encounters relative to baseline. Results The nationwide lockdown on 27 March 2020 led to sharp reductions in care-seeking behavior that persisted for 18 months after initial declines. For example, total admissions dropped 59.6% (95% confidence interval [CI], 52.4–66.8) during home confinement and were 33.2% (95% CI, 29–37.4) below baseline in September 2021. We identified 3 waves of all-cause respiratory encounters consistent with COVID-19 activity. Intestinal infections and non–COVID-19 respiratory illnesses experienced the most pronounced declines, with some diagnoses reduced 80%, even as nonpharmaceutical interventions (NPIs) relaxed. Non-respiratory hospitalizations, including injuries and acute illnesses, were 20%–60% below baseline throughout the pandemic and exhibited strong temporal associations with NPIs and mobility. ED attendances exhibited trends similar to those for hospitalizations, while GP visits were less impacted and have returned to pre-pandemic levels. Conclusions We found substantially reduced use of health services during the pandemic for a range of conditions unrelated to COVID-19. Persistent declines in hospitalizations and ED visits indicate that high-risk patients are still delaying seeking care, which could lead to morbidity or mortality increases in the future.


Age-specific patterns in all-cause respiratory admissions
In individuals aged ≥ 20 years, there were 52,764 all-cause respiratory hospitalizations from March 2020 to March 2021, compared with an annual average of 33,423 (95% confidence interval: [28,470,38,376]) during the same period in the previous four years (Figure 3). In individuals aged < 20 years, there were 10,334 all-cause respiratory hospitalizations from March 2020 to March 2021, compared with an annual average of 29,723 [25,973,33,473] during the same period in the previous four years (Figure 3).

National and age-specific trends in outpatient all-cause respiratory encounters
In individuals aged ≥ 20 years, there were 32,026 all-cause respiratory attendances at emergency departments (ED) in private hospitals and 194,075 consultations at private general practitioners (GP)  Outpatient all-cause respiratory consultations spiked nationally and across all age groups during the week of 15 March 2020, consistent with "worried well" health-seeking behavior (national peak increase above baseline: ED, 279% [150,459]; GP, 150% [108, 214]; Figure 2; Figures S9-10). After the onset of Level 5, outpatient respiratory visits declined sharply across all ages, reflecting the impact of lockdown measures. Adult ED and GP respiratory encounters increased after the transition to Level 3 in June 2020, peaked at levels above (ED) or equivalent to (GP) their seasonal baselines in July 2020, and then declined below baseline levels in August 2020 ( Figures S9-10). Similar to trends observed in inpatient admissions, national-level and adult ED respiratory visits during the second wave in December 2020 exceeded levels observed during the first wave (national peak increase above baseline: 159% [70, 242]; Figure 2; Figure S9). GP respiratory consultations also peaked in late December 2020, with consultations in adults aged ≥ 20 years surpassing baseline levels ( Figure 2, Figure S10). The third wave of outpatient visits peaked in early July 2021, with national ED visits peaking at a level equivalent to the projected baseline (12% [-24, 87]) and adult visits exceeding baseline levels ( Figure S9). National and adult GP visits surpassed those observed during the prior two waves (peak increase above baseline: 60% [26, 107]; Figure 2; Figure  S10).
After the initiation of lockdown Level 5 in late March 2020, outpatient respiratory visits in children (< 5 years, 5-19 years) remained substantially below baseline numbers until RSV circulation increased in last quarter of 2020 ( Figures S9-10). ED and GP consultations in young children spiked in February 2021 and then declined below their seasonal baselines. During the third COVID-19 wave, GP visits in individuals aged < 20 years increased to baseline levels, whereas ED visits appeared unaffected.
Like trends observed in inpatient admissions, the weekly percent change in baseline for outpatient all-cause respiratory consultations strongly correlated with weekly COVID-19-coded admissions in adults aged ≥ 20 years but not in children ( Figure S11). During the pre-lockdown period of our study (1 -26 March 2020), weekly attendances for non-COVID-19 respiratory illnesses (pneumonia and influenza, acute lower respiratory infections, chronic lower respiratory illnesses, and asthma) were at levels equivalent to or slightly above their projected baselines ( Figure S13, Table S4). Attendances for intestinal infectious diseases, chronic diseases (chronic obstructive pulmonary disorder (COPD), diabetes, heart disease, heart failure, hypertension), neoplasms, alcohol-related disorders, and injuries were also equivalent to baseline levels. After the initiation of strict lockdown measures (Level 5), attendances for respiratory illnesses, intestinal infectious diseases, heart diseases, cerebrovascular diseases, alcohol-related disorders, and injuries declined to levels 50% below baseline, while attendances for other conditions, such as acute myocardial infarction (AMI), heart failure, and hypertension, exhibited less marked declines ( Figure S13, Table S4). Acute and chronic respiratory illnesses, COPD, intestinal infectious diseases, and injuries experienced the most pronounced declines throughout the pandemic, while attendances for non-communicable diseases returned to numbers closer to baseline levels during periods of more relaxed public health measures. Though attendances for injuries increased during Levels 2 and 1 (August to mid-December 2020), they dropped again to numbers 50% below baseline when stay-at-home restrictions tightened in late December 2020. At the end of the study period in September 2021, attendances for non-COVID respiratory illnesses, intestinal infections, and alcohol-related incidents were approximately 60-70% below projected baselines, visits for heart diseases, heart failure, cerebrovascular diseases, and hypertension were 15-25% below baseline, and injuries were 40-50% below baseline (Figure S13, Table S4). As of September 2021, encounters for diabetes and AMI were at pre-pandemic levels.

Non-COVID-19 consultations at general practitioner providers
In-person consultations at private general practitioner (GP) providers were less impacted by pandemic-related factors than hospitalizations and emergency department visits. From March 2020 to March 2021, there were 1,041,432 total consultations and 789,572 non-respiratory consultations at GP providers, compared to an average of 2,084,919 [1,805,175, 2,364,663] annual total consultations and 1,480,490 [1,308,128, 1,652,852] annual nonrespiratory consultations during the same period in the previous four years.
During the pre-lockdown period in March 2020, consultations for chronic lower respiratory illnesses, and in particular asthma, spiked at numbers 50% above seasonal baselines while consultations for other conditions were, on average, equivalent to baseline levels ( Figure S14, Table S5). Non-COVID-19 pneumonia and influenza, acute lower respiratory infections, chronic bronchitis, and intestinal infectious diseases were the conditions most impacted by shelter-in-place orders, followed by injuries, neoplasms, and heart failure, and then chronic illnesses (e.g., lipidemias, heart diseases, and hypertension). Consultations for asthma, diabetes, and human immunodeficiency virus (HIV) were not, on average, impacted by the nationwide lockdown and remained at pre-pandemic levels throughout the pandemic. Similar to patterns observed for inpatient admissions and emergency department attendances, consultations for intestinal infectious diseases and some respiratory conditions, specifically pneumonia and influenza, acute lower respiratory infections, and chronic bronchitis, were 40-80% below baseline throughout the pandemic ( Figure S14, Table S5). From May to September 2020 (Levels 4 to 2), consultations for diabetes, lipidemias, heart disease, heart failure, hypertension, HIV, and COPD were slightly below or equivalent to baseline levels, whereas those for neoplasms and injuries were approximately 20-25% below baseline. Weekly GP consultations for diabetes, lipidemias, heart disease, heart failure, hypertension, neoplasms, and injuries declined to numbers approximately 50% below baseline from late December 2020 to May 2021 (adjusted Levels 3 and 1) but returned to baseline levels in June 2021, after the transition to adjusted Level 2. At the end of the study in September 2021, GP visits for pneumonia and influenza, acute lower respiratory tract infections, chronic bronchitis, and intestinal infections were approximately 60% below baseline, and those for chronic lower respiratory diseases, asthma, and COPD were 10-40% below baseline (Figure S14, Table S5). As of September 2021, consultations for neoplasms, cerebrovascular diseases, cardiac conditions, and injuries were 20-40% below baseline, while those for HIV, diabetes, and hypertension were at pre-pandemic levels. Figure S1. Numbers of monthly general practitioner consultations. Medicross began its telehealth initiative in March 2020, indicated by the black vertical dashed line. From March 2020 to May 2021, visits are categorized as physical, in person visits (light blue) or as telehealth -telephone or virtual -consults (dark blue). Labels indicate the percentage of monthly general practitioner visits that were telehealth consults.

Supplementary Figures
The Health Professions Council of South Africa (HPSCA) regulates telemedicine in South Africa. Prior to the pandemic, HPSCA prohibited first-time consultations and required all telemedicine appointments to include a faceto-face consultation and physical examination by the 'consulting' practitioner in a clinical setting, with the 'consulting' practitioner communicating information to the 'servicing' practitioner (2014 Guidelines). The COVID-19 pandemic necessitated a relaxation of these policies to reduce healthcare worker contact with patients and to reach patients in remote and rural locations. The HPSCA amended their guidance in April 2020 to allow first-time consultations between clinicians and patients without an established relationship.   (Table 1) and are colored according to the stringency of lockdown measures: red to orange to yellow.

Figure S4. Weekly emergency department consultations by diagnosis and age group.
Vertical dashed lines indicate lockdown alert levels from March 2020 to September 2021 (Table 1) and are colored according to the stringency of lockdown measures: red to orange to yellow.

Figure S5. Weekly general practitioner consultations by diagnosis and age group.
Vertical dashed lines indicate lockdown alert levels from March 2020 to September 2021 (Table 1) and are colored according to the stringency of lockdown measures: red to orange to yellow.

Figure S6. Individual indices for the ban of alcohol sales, containment and closure policies, and health policies in South Africa.
Daily time series for indicators C1 to C8, H1 to H3, and H6 to H7 were extracted from the OxCGRT database [4], which records individual indicators as ordinal or continuous values (     (Table 1). Panels are shaded according to the stringency of lockdown measures: red to orange to yellow. The area between the projected seasonal baseline and observed consultations is shaded green when observed consultations are below baseline ("averted cases") and shaded yellow when observed consultations are above baseline ("excess cases").  (Table 1). Panels are shaded according to the stringency of lockdown measures: red to orange to yellow. The area between the projected seasonal baseline and observed consultations is shaded green when observed consultations are below baseline ("averted cases") and shaded yellow when observed consultations are above baseline ("excess cases"). B. Weekly observed percent difference from seasonal baseline (95% confidence interval) by age group.  (Table 1). Panels are shaded according to the stringency of lockdown measures: red to orange to yellow. The area between the projected seasonal baseline and observed consultations is shaded green when observed consultations are below baseline ("averted cases") and shaded yellow when observed consultations are above baseline ("excess cases"). B. Weekly observed percent difference from seasonal baseline (95% confidence interval) by age group.  (Table 1) and are shaded according to the stringency of lockdown measures: red to orange to yellow.      (Table 1). Points are shaded according to the stringency of lockdown measures: red to orange to yellow. Generalized additive models (GAMs) were used to identify non-linear relationships between the Google Transit Stations metric and inpatient admissions for each diagnosis group. GAM adjusted R 2 values are in the top left of each facet.   (Table 1). Panels are shaded according to the stringency of lockdown measures: red to orange to yellow.

Figure S18. Associations between Google mobility metrics and the weekly percent change from baseline for diagnoses coded as A. superficial injuries and B. trauma or fractures.
Point colors indicate weeks during which alcohol sales were banned (first ban: 27 March -1 June 2020, second ban: 12 July -17 August 2020, third ban: 28 December 2020 -1 February 2021, fourth ban: 28 June -25 July 2021). Generalized additive models (GAMs) were used to identify non-linear relationships between Google Mobility metrics and inpatient admissions. GAM adjusted R 2 values are in the top left of each facet.

Figure S19. Non-linear associations between Google mobility metrics and respiratory admissions in individuals aged ≥ 5 years.
A. Relationships between the weekly percent change from baseline for Google mobility metrics and the weekly percent change from baseline for all-cause respiratory admissions (including COVID-19) among individuals aged ≥ 5 years. B. Relationships between the weekly percent change from baseline for Google mobility metrics and the weekly number of COVID-19-coded admissions among individuals aged ≥ 5 years. Point colors indicate the prelockdown period (light blue: 1 -26 March 2020) and lockdown alert levels from March 2020 to September 2021 (Table 1) (Table 1). Panels are shaded according to the stringency of lockdown measures: red to orange to yellow.   Table S3. Percent change in admissions relative to baseline numbers expected in the absence of COVID-19. The mean percent change from baseline in weekly admissions at a private hospital group during ten phases of the COVID-19 pandemic in South Africa: pre-lockdown (1 -26 March 2020) and lockdown alert levels from March 2020 to September 2021 (Table 1). Brackets include 95% confidence intervals for mean estimates.

Respiratory conditions
All-cause respiratory    Table S4. Percent change in emergency department consultations relative to baseline numbers expected in the absence of COVID-19. The average percent change from baseline in weekly consultations at a private healthcare group during ten phases of the COVID-19 pandemic in South Africa: pre-lockdown (1 -26 March 2020) and lockdown alert levels from March 2020 to September 2021 (Table 1). Brackets include 95% confidence intervals for mean estimates.