The impact of the COVID-19 pandemic on health service use in sub-Saharan Africa

Abstract

SETTING:

Six hospitals in four sub-Saharan African countries.

OBJECTIVE:

To examine the indirect effects of COVID-19 on health service utilisation and to explore the risk of bias in studies on prediction models.

DESIGN:

Monthly data were analysed using interrupted time-series modelling. We used linear mixed-effect models for the analysis of antenatal care visits, institutional deliveries, vaccinations, outpatient visits and hospital admissions, and generalised linear mixed-effect models for hospital mortality.

RESULTS:

During 2018–2020, the six hospitals recorded a total of 57,075 antenatal care visits, 38,706 institutional deliveries, 312,961 vaccinations, 605,925 out-patient visits and 143,915 hospital admissions. The COVID-19 period was associated with decreases in vacci-nations (− 575 vaccinations, P < 0.0001), outpatient visits (− 700 visits, P < 0.0001) and hospital admission (− 102 admission, P = 0.001); however, no statistically significant effects were found for antenatal care visits (P = 0.71) or institutional deliveries (P = 0.14). Mortality rate increased by 2% per month in the pre-COVID-19 period; however, a decreasing trend (by 2% per month) was observed during the COVID-19 period (P = 0.004). Subgroup and sensitivity analyses broadly confirmed the main findings with only minor inconsistencies. A reduction in outpatient visits was also observed in hospitals from countries with a higher Stringency Index and in urban hospitals.

CONCLUSIONS:

The pandemic resulted in a reduction in health service utilisation. The decreases were less than anticipated from modelling studies.

The COVID-19 pandemic has become one of the main causes of death in many countries, and the real number of deaths are at least double if not triple those confirmed.¹ The magnitude of the COVID-19 pandemic has created unprecedented challenges for health authorities, as well as concerns about disrupted healthcare provision and access.² This may weigh most heavily in settings such as sub-Saharan Africa (sSA).

As reported during the West African Ebola virus outbreak, the response of the health institutions led to restrictions in the coverage of health services, both directly and indirectly.^3–8 Available evidence suggests that disrupting healthcare provision may have a long-term impact on demand, as it erodes trust in providers and patients may continue to avoid providers for fear of infection even when services resume.^9–11

At the beginning of the COVID-19 crisis, it was forecast that sSA might well record large numbers of cases and deaths.¹² To date, however, the region seems relatively less affected (i.e., fewer confirmed cases and deaths) than other continents and regions.^13,14 However, it is unclear to what extent these figures are due to lower testing rates, less severe clinical presentation, or other factors.^15,16 Nonetheless, the pandemic has placed a strain sSA healthcare systems and disrupted healthcare provision and access.^2,17

Only a few observational studies investigated the pandemic’s impact on the use of health services in sSA, finding a decrease in obstetric and paediatric health services,^18–20 as well as barriers to care and fear of COVID-19 infection.¹⁹ Outside sSA, evidence suggests declines in vaccinations, institutional deliveries and healthcare use during the COVID-19 pandemic.^21–23

Modelling studies may provide useful insights into the projected ‘collateral damage’ of the current epidemic.²⁴ However, their accuracy in predicting outcomes depends on the model used, assumptions made, and quality of available data.²⁵ This can be challenging at the beginning of an outbreak or in low-resource settings, where data quality is often poor. A systematic review of the impact of COVID-19 on healthcare utilisation worldwide found no eligible studies from sSA.²¹ Another systematic review of 40 studies assessed the collective evidence on the effects of the pandemic on maternal, foetal and neonatal outcomes. The report found one eligible study from sSA.²⁶

While the epidemic is ongoing, it is difficult to assess the entire impact of COVID-19 on health service access and health outcomes. The present study aimed to evaluate the indirect impact of COVID-19 pandemic on antenatal care visits, institutional deliveries, vaccinations, outpatient visits, hospital admissions and hospital mortality in six hospitals of four sSA countries: Ethiopia, Sierra Leone, Tanzania and Uganda.

METHODS

Setting

This study evaluates the indirect impact of the COVID-19 outbreak in six hospitals in four sSA countries: Ethiopia, Sierra Leone, Tanzania and Uganda. Table 1 shows the progress of the pandemic in these countries. In the four countries considered, the first case of COVID-19 was identified between the second week of March (Ethiopia) and the first week of April (Uganda). Schools were closed in all four countries in the second half of March (16 March in Ethiopia, 31 March in Sierra Leone). Mass gatherings were prohibited, home confinement was mandated and borders were closed, with different modalities and times, between the second week of March and the first week of April.

TABLE 1.

COVID-19 indicators by country

Country	First reported case	First reported death	Schools closed	Mass gatherings events prohibited	Average Stringency Index*
Ethiopia	13 March	5 April	16 March	16 March	67
Sierra Leone	31 March	23 April	20 March	20 March	42
Tanzania	16 March	31 March	19 March	2 April	27
Uganda	21 March	23 July	18 March	18 March	71

^*The stringency index is a composite measure based on nine response indicators, including school closures, workplace closures and travel bans, rescaled to a value from 0 to 100 (100 = strictest).

It was not clear which aspects of the restrictive measures to contain the epidemic would impact on health services. Measures were implemented differently in different settings, and included a broad range of restrictions. Restrictive measures are presented here according to COVID-19 Stringency Index. This index is a composite based on nine response indicators and ranges from 0 to 100 (100 = strictest) (Figure 1 and Table 1).^27,28 Doctors with Africa (DwA) CUAMM (Collegio Universitario Aspiranti Medici Missionari) is an Italian non-governmental organisation (NGO) working in 23 sSA hospitals.²⁹ The low quality of data collection precluded the participation of some hospitals where DwA provides only technical support without a continuous presence of staff (which means that monthly records are not available) and other hospitals where the presence of DwA personnel was intermittent due to the pandemic, and continuous data collection was not possible. This study looked at data from hospitals where DwA has been working for years and where there has been ongoing consolidated data collection without interruption during the pandemic. Data came from Wolisso Hospital (Ethiopia), Pujeuhun Hospital (Sierra Leone), Tosamaganga and Songambele Hospitals (Tanzania), and Aber and Matany Hospitals (Uganda). Table 2 gives general information about these hospitals. The six hospitals analysed did not discontine or reduce healthcare services during the pandemic.

FIGURE 1.

COVID-19 Stringency Index by country.

TABLE 2.

General information on the six participating hospitals, sub-Saharan Africa, 2020

Hospital, country	Served population	Location	Beds	Outpatients	Admissions	Antenatal visits	Births	Vaccinations	Total staff	Qualified staff
Wolisso, Ethiopia	1,198,149	Urban	200	66,522	12,578	5,794	3,950	8,296	353	228
Pujeuhun, Sierra Leone	397,171	Rural	59	5,558	2,067	3,361*	1,081	NA	121	71
Tosamaganga, Tanzania	687,460	Rural	165	38,210	5,812	1,349	2,659	8,859	165	109
Songambele, Tanzania	123,400	Rural	84	12,882	2,600	1,965	903	8,879	79	71
Aber, Uganda	217,141	Rural	178	36,133	10,521	6,837	2,637	35,037	163	110
Matany, Uganda	159,409	Rural	250	33,946	14,761	4,391	1,531	37,108	267	149

^*Data of 2019.

NA = not available.

Data collection

The outcome measures included antenatal care visits, institutional deliveries, vaccinations, outpatient visits, hospital admissions and hospital mortality. To include a pre-COVID-19 period with adequate length for comparison, aggregate data were taken from monthly hospital records from January 2018 to December 2020. Data extraction was carried out by a researcher who was not involved in any clinical activities.

Analytical strategy

This was a retrospective, observational, multi-centre, multi-country study. Monthly data were analysed using interrupted time-series modelling to evaluate the indirect impact of the COVID-19 pandemic by changes in the level and slope of the time series of each outcome measure. As the pre-COVID-19 and COVID-19 periods included different subsets of months, seasonality was expected to contribute to the observed change between the two periods without being associated with the pandemic itself. We ruled out the option of comparing similar months in the two periods due to disadvantages in reducing the amount of data and limiting the estimation capability of the time trends. Since the indirect impact of the pandemic could be measured by the change in the trend component of the time series, we estimated the trend component and the seasonal component of each time series separately, and further analyses focused on the trend component.

We used linear mixed-effect models in the analysis of data on antenatal care visits, institutional deliveries, vaccinations, outpatient visits and hospital admissions, and generalised linear mixed-effect models using the beta distribution family for hospital mortality. The models included the time (January 2018–December 2020), the period (pre-COVID-19 period vs. COVID-19 period) and the interaction term time^*period, with the centre contributing as random effect.

Given the heterogeneity of the participating hospitals in terms of setting and health policies, a subgroup analysis was performed for each country to provide further insights. For the subgroup analysis, linear regression models were used for prenatal care visits, institutional deliveries, vaccinations, outpatient visits, and hospital admissions, whereas beta regression models were used for hospital mortality. The models included the time (January 2018–December 2020), the period (pre-COVID-19 period vs. COVID-19 period), the centre (when more hospitals participated from the same country) and all interaction terms.

In the main and the subgroup analyses, the pre-COVID-19 period ranged from January 2018 to February 2020, while the COVID-19 period ranged from March 2020 to December 2020. However, it was uncertain if data from March 2020 would show the impact of the COVID-19 outbreak. Therefore, a sensitivity analysis was also performed with the pre-COVID-19 period ranging from January 2018 to March 2020, and the COVID-19 period ranging from April 2020 to December 2020.

Further sensitivity analyses explored the possible indirect effects on the study of the restrictive measures to contain the epidemic (measured by the Stringency Index) and the location of the hospital (urban or rural). In these sensitivity analyses, additional interaction terms were included in the models (see Supplementary Data). All tests were two-sided, and P < 0.05 was considered statistically significant. Statistical analysis was performed using R v4.1 (R Computing, Vienna,Austria).³⁰

Ethics statements and patient involvement

Ethical clearance for accessing and using data for scientific purpose was granted by the Ethics Committee of each participating hospital. This study used hospital-level data only, and no personal information was retrieved from hospital charts. As this research was performed without patient involvement, informed consent was not required.

RESULTS

Main analysis

During 2018–2020, the six hospitals recorded a total of 57,075 antenatal care visits, 38,706 institutional deliveries, 312,961 vaccinations, 605,925 outpatient visits and 143,915 hospital admissions, with an average hospital mortality rate of 5.0%. Figure 2 shows overall trends in antenatal care, institutional deliveries and vaccinations during the study period. The COVID-19 period was associated with a reduction in average monthly vaccinations (− 575 vaccinations, P < 0.0001), while no statistically significant effects on the number of monthly antenatal care visits (P = 0.71) or institutional deliveries (P = 0.14) were estimated.

FIGURE 2.

Interrupted time series of antenatal care, institutional deliveries and vaccinations in six sub-Saharan African hospitals in 2018–2020. White background = pre-COVID-19 period (January 2018–March 2020); grey background = COVID-19 period (April–December 2020); line = predicted trend based on the regression model.

Overall trends in outpatient visits, hospital admission and hospital mortality during the study period are shown in Figure 3. The COVID-19 period saw a drop in the average monthly outpatient visits (− 700 visits, P < 0.0001) and hospital admissions (− 102 admissions, P = 0.001). Mortality rate was increasing by 2% per month before the COVID-19 period, while a decreasing trend (by 2% per month) was observed during the COVID-19 period (P = 0.004). Full details of the analysis are reported in Supplementary Table S1.

FIGURE 3.

Interrupted time series of outpatient visits, hospital admission, and hospital mortality in six sub-Saharan African hospitals in 2018–2020. White background = pre-COVID-19 period (January 2018–March 2020); grey background = COVID-19 period (April–December 2020); line = predicted trend based on the regression model.

Subgroup analysis

When data from each country’s hospitals were analysed separately, it was observed that the COVID-19 period showed widely disparate local changes in antenatal care visits and vaccines. The findings of the main analysis of institutional deliveries, outpatient visits, hospital admissions and mortality were locally confirmed by most hospitals, but some differences were observed in a few hospitals. Full details are provided in Supplementary Figures S1–S4 and Supplementary Tables S2–S5.

Sensitivity analyses

When we shifted the beginning of the COVID-19 period from March to April 2020, the sensitivity analysis confirmed the findings of the main analysis and the subgroup analysis, with minor changes in the estimated effects of the COVID-19 period. Full details are reported in Supplementary Figures S5–S10 and Supplementary Tables S6–S10.

In the investigation of the possible indirect effects of the restrictive measures and the location of the hospital, the findings of our analysis suggested a reduction in the average monthly ante-natal care visits (− 31 visits, P = 0.03) and outpatient visits (− 903 visits, P < 0.0001) in hospitals from countries with a higher Stringency Index, i.e., Uganda and Ethiopia. In addition, urban hospitals showed reductions in the average monthly antenatal care visits (− 62 visits, P = 0.0002), institutional deliveries (− 58 deliveries, P < 0.0001), outpatient visits (− 1921 visits, P < 0.0001), hospital admissions (− 306 admissions, P < 0.0001) and hospital mortality rates (− 3.4%, P = 0.03) with respect to rural hospitals (Supplementary Data).

DISCUSSION

In the six hospitals studied in four sSA nations (Ethiopia, Sierra Leone, Tanzania, and Uganda), our findings regarding the indirect effects of COVID-19’s on health service utilisation were inconsistent. The COVID-19 period was associated with a reduction in vaccinations, outpatient visits and hospital admissions. No significant effects were seen on antenatal care visits and institutional deliveries. The mortality rate, which had been increasing in the pre-COVID-19 period, began to decrease in the COVID-19 period. Overall, we did not observe decreases in health service utilisation or the higher mortality rates that had been predicted by modelling studies at the outset of the pandemic.^24,31,32 For example, Abbas et al., have estimated the additional maternal and under-5 child deaths resulting from the potential disruption of health systems due to COVID-19 in LMICs: under different scenarios, these deaths would represent a 9.8–44.7% increase in under-5 child deaths per month and an 8.3–38.6% increase in maternal deaths per month across the 118 countries.³²

Changes in healthcare utilisation might occur concurrently with shifts in health seeking behaviour and trust in the healthcare system. Trust is affected by how patients and the community perceive the infection risk, service availability, and cost of health centres.¹⁹ We can speculate that limited decrease in healthcare utilisation is, above all, related to the perceived level of the threat. During the 2013–2016 Ebola outbreak, the largest decreases in service utilisation were seen in the districts with the highest Ebola incidences.⁷ Unlike Ebola, the number of cases and reported case fatalities has remained low in the current pandemic, and the public probably did not see the disease as a serious life-threatening infection.

Previous studies from sSA reported some reduction in healthcare access during the pandemic. In rural South Africa, a single-centre study found no significant change in total admissions, but did find significant changes between subgroups of admissions, with a large decline in the use of health services among children.¹⁸ Hospital-level data from South Africa and Nigeria showed that the number of antenatal visits had fallen. Evidence was mixed for facility-based deliveries and caesarean sections.¹⁹ Wanyana et al. reported that the utilisation of 13 maternal and child health services across Rwanda had decreased since the COVID-19 outbreak, particularly in the use of services related to health facility deliveries and child vaccinations.²⁰ Chelo et al. assessed the effect of the COVID-19 pandemic on hospitalisations, as well as the mortality of children in a paediatric hospital in Cameroon. A drastic drop in the number of hospitalisations was reported, with a parallel increase in the number of deaths.³³ Caniglia et al. evaluated the association between the COVID-19 period and the adverse birth outcomes in Botswana: they found that the number of deliveries remained constant and no significant difference in adverse birth outcomes was observed.³⁴

Our findings showed a decrease in vaccinations and outpatient visits at the beginning of the pandemic. The decrease in outpatient visits was more marked in urban hospitals and in hospitals located in countries with higher restrictive measures to contain the epidemic. We also observed that vaccinations and outpatient visits did not return to pre-pandemic levels at the end of 2020, suggesting medium-term effects, with the possibility of more long-term impact. A recent risk-benefit study warned that the deaths prevented by sustaining routine childhood immunisation in Africa could outweigh the excess risk of COVID-19 deaths associated with vaccination clinic visits, especially for vaccinated children. Routine childhood immunisation should be sustained in Africa as much as possible, taking into account logistical constraints and reallocation of resources during the COVID-19 pandemic.³²

We can speculate that the informative impact and the economic impact of the pandemic, as well as the limitations on access to care, were greater in urban compared to rural settings.¹⁹ It is hard to disentangle the extent to which restrictive measures and stigma/fear drove the change in care-seeking behaviour. Nonetheless, the fact that the reduction in some health services was greater in hospitals of countries with the highest Stringency Index suggests that reduction in care-seeking for some types of visits was an unintended negative consequence of the restrictive measures to contain the epidemic. This finding is remarkable, especially in light of the fact that these unintended consequences could outlast the duration of the policy.

After an early drop at the beginning of the pandemic, the sensitivity analysis also showed an increase in hospital admissions during the second half of 2020, which might be explained by the relaxing of the restrictive measures and the delayed access to hospital care.¹⁸ However, this increase did not coincide with an increase in hospital mortality that was reported elsewhere.¹⁸ In fact, we observed a decrease in hospital mortality during the pandemic, which may suggest that the most vulnerable people did not access hospital care. The absence of concurrent rising trends in hospital admissions and hospital mortality during the pandemic may imply a higher external mortality (i.e., those who did not seek care and died at home may not have been included in the hospital mortality statistics).³⁵ It is reasonable to assume that several factors, including fear of contracting COVID-19, restrictive measures and changes in resource allocation, might have contributed to limiting access to care.¹⁹ In sSA, approximately 90% of people are employed in the informal sector with low productivity.³⁶ This sector was greatly affected during the pandemic, and faced serious financial difficulties.

Overall, prenatal care visits and institutional deliveries in the hospitals surveyed were unaffected by the pandemic, in contrast to other studies from South Africa and Asia that reported significant declines in the use of these health services.^11,19,20 We believe that pregnant women may have preferred the hospital to the health centres, as they viewed the hospital to be a safer environment.

The subgroup analysis showed some heterogeneity in the effects of the pandemic on the participating hospitals. Such differences might be explained by a complex mix of local factors, such as the role of the hospital in the community, organisational features (urban hospitals showed a larger reduction in monthly ante-natal care visits, institutional deliveries, outpatient visits and hospital admissions), country-level response to the pandemic (hospitals from countries with higher restrictive measures showed a larger reduction in monthly antenatal care visits and outpatient visits) and population factors (fear, affordability of healthcare service, income loss).³⁷ Hence, understanding the local constraints on access to care is crucial in ensuring healthcare coverage during the ongoing epidemic.

Our study has strengths and limitations that should be considered when interpreting the results. First, the multi-centre, multi-country design provided a wide overlook at the indirect impact of COVID-19 pandemic on sSA healthcare and increased the generalisability of the findings. On the other hand, generalisability was limited by possible selection bias when selecting participating hospitals (i.e., sSA hospitals where DwA had worked for years and where a consolidated data gathering methodology had been established over time). Second, the analysis benefited from the inclusion of a wide time series covering adequate pre-COVID-19 and COVID-19 periods; this allowed an adequate estimation of the indirect impact of the COVID-19 pandemic by using interrupted time-series modelling and seasonality adjustments. However, monthly hospital-level data precluded any patient-level analyses (including stratification for patient age), and the quality of the retrospective data collection may have been wanting. Finally, the large sample size and the sensitivity analyses strengthened the interpretation of the findings.

Our study showed indirect effects of the COVID-19 pandemic on health service utilisation in six hospitals in four sSA countries. The decreases in health service utilisation were less than anticipated from modelling studies, but long-term effects should be expected. From a policy perspective, understanding the role of several factors – such as availability and accessibility of services, fear and financial access – can shed light on the underlying causes of the observed changes. Therefore, a regular monitoring of health service utilisation in epidemics plays an important role in guiding and evaluating public health response measures. Integrating health information system data analysis with social sciences evidence can contribute to a better and more comprehensive interpretation of the data.