Original research

Assessing the impact of COVID-19 pandemic on maternal healthcare usage: evidence from routine health data in Kenya and Ethiopia

Abstract

Objectives Lockdowns and fear of COVID-19 may have reduced access to antenatal care (ANC) and skilled birth attendance (SBA) in sub-Saharan Africa, which could undermine progress towards maternal and child survival and the sustainable development goals (SDGs). We analysed COVID-19’s impact on maternal healthcare usage, focusing on subnational levels, to identify healthcare disruption hotspots that require targeted interventions and help policymakers prioritise resources to accelerate progress.

Methods and analysis Using monthly health management information system (HMIS) data, we tracked changes in healthcare access at subnational levels in Ethiopia and Kenya during the pandemic. We compared service usage before and during the pandemic, using interrupted time series and counterfactual analyses to evaluate the pandemic’s impact on healthcare usage trends. We also performed geospatial mapping of the affected regions to identify hotspots.

Results Our results show significant changes at subnational levels. ANC declined in several Kenyan counties during the pandemic, with improvements observed in others. SBA disruptions were observed in two counties. In Ethiopia, ANC declined in the north but remained unchanged in the south, with some improvements observed in the two regions. Southern regions showed resilience in SBA, experiencing gains, while northern regions showed no change.

Conclusion Future disease outbreaks may continue to cause further disruptions to health service delivery, affecting maternal and child health outcomes. Our analysis highlights the low resilience of subnational health systems to shocks, underscoring the need to strengthen healthcare systems and HMIS data capture for better data quality. Evidence-based research is essential in identifying hotspots and supporting targeted interventions to achieve the SDGs and improve maternal and child health outcomes.

What is already known on this topic

  • The COVID-19 pandemic is a health crisis that crippled healthcare systems and revealed vulnerabilities within the health systems globally. It derailed countries’ progress towards achieving global commitments, including sustainable development goals (SDGs), especially in low-income and middle-income countries, in maternal, neonatal and child health (MNCH) outcomes, leading to one of the highest recorded mortality rates.

What this study adds

  • This study provides an in-depth assessment of COVID-19’s impact on maternal and child healthcare access at the subnational level in Kenya and Ethiopia. Subnational estimates are usually masked when the impact is assessed and reported at the national level (ie, many studies assess and report impact at this administrative level), leading to certain areas and regions that are left out on interventions and hence lag in achieving set SDG targets. The study provides a much-needed scope of MNCH healthcare access vulnerabilities due to the COVID-19 impact within countries of interest (ie, hotspot areas that have been heavily impacted). The identified subnational hotspots can thus be targeted with health interventions to improve people’s lives.

How this study might affect research, practice or policy

  • This study may be considered a practical guide to programme implementers, policymakers and researchers on the evidence-based identification of subnational areas where interventions would have the most health impact and hence contribute to a quick recovery from the effects of the pandemic in MNCH healthcare access in hotspot areas. Healthcare access has been affected by the pandemic in numerous ways; therefore, it is crucial to understand the context of the epidemic (ie, at the subnational level) to ensure effective interventions are designed and implemented. It also provides several policy-relevant recommendations for African governments to implement in preparedness for resilience to shocks that may come from future pandemic outbreaks similar to COVID-19.

Introduction

The devastating impacts and lessons learnt from other epidemic contexts show that health systems in low-income and middle-income countries (LMICs) and fragile and humanitarian contexts have a low capacity to implement an emergency response during disease outbreaks such as the COVID-19 pandemic caused by the SARS-CoV-2 virus.1 2 Such unprecedented emergency and humanitarian situations burden health systems, which, even before the disease outbreak, have struggled to provide quality essential services. Governments have implemented public health and social measures to curb the transmission of the virus.3 Restrictions on the movement of people, transport and the temporary closure of schools and non-essential services adopted by many countries have harmed the provision and usage of reproductive and maternal health services.4 5

Maternal, newborn and child health (MNCH) is a significant public health focus on the global health and development agenda. In sub-Saharan Africa (sSA), an estimated 196 000 women died during pregnancy, childbirth or postnatal periods, and 2.4 million babies died within 1 month of birth in 2019.6 Most maternal and child deaths are preventable. Women die as a result of treatable complications during and following pregnancy and childbirth, including severe postpartum bleeding, infections (usually after childbirth), high blood pressure during pregnancy (pre-eclampsia and eclampsia), complications from delivery, unsafe abortion, malaria and chronic conditions like cardiac disease and diabetes, which can worsen pregnancy complications. These treatable complications are exacerbated because of individual-level factors such as very early or late maternal age, short birth intervals and poor nutritional status; institutional factors such as poor healthcare and lack of facilities and structural factors such as health financing, inappropriate choice and implementation of programmes and weak policies.6 It is estimated that 16%–33% of all maternal deaths may be averted via primary or secondary prevention of maternity complications in maternal healthcare.7–10 The indirect impacts of COVID-19 were estimated to result in an 8.3%–38.6% increase in maternal mortality and a 9.8%–44.7% increase in child mortality per month in 2020 across 118 LMICs, which included Kenya and Ethiopia.11 Furthermore, the main factors implicated are antenatal care (ANC) disruptions and skilled birth attendance (SBA), including emergency obstetric care.12

ANC is critical as it presents the first contact of opportunity for comprehensive health screening for pregnant women.13 The WHO advises that there should be a high priority placed on essential obstetric care during the pandemic and that care should not be interrupted or neglected, as this could result in higher morbidity and mortality.14 Very few studies have demonstrated how COVID-19 affects progress towards achieving sustainable development goals (SDGs), especially SDG number 3 in countries including Kenya and Ethiopia. Documented accounts of localised studies in the subregion suggest that since the start of the pandemic in 2019–2020, there has been an increase in the number of home deliveries and assistance by traditional birth attendants or traditional midwives.15 This is due to fears of contracting COVID-19, testing positive for COVID-19, being forced to quarantine away from their families and other factors such as cost and distance to the health facility.16–20 Pregnant women with obstetric complications seek health facility care too late, sometimes ending in undesirable birth outcomes such as stillbirths and neonatal and maternal deaths.20

A recent study documents the impact of the COVID-19 pandemic on MNCH service usage in 12 sSA countries.21 The authors report reductions in inpatient, outpatient, ANC, delivery care and immunisation services, among others, in general. Evidence, however, shows that reductions have not been uniform across countries, periods and types of services.22–24 In this study, we assessed COVID-19’s impact on antenatal and skilled delivery care, focusing on subnational levels to identify healthcare disruption hotspots that may require targeted interventions and help policymakers prioritise resources to help accelerate progress.

Methodology

Data

Routine facility-based health usage data from the Ethiopian and Kenyan health management information systems (HMIS) were analysed. The HMIS is managed on the District Health Information Software 2 (DHIS2), a global open-source web-based platform coordinated by the Health Information Systems Programme at the University of Oslo.25 More than 73 countries worldwide use DHIS2 for collecting and analysing health data. Data were extracted regarding the number of skilled deliveries and ANC visits (fourth visit), among other indicators. The data comprised monthly subnational level data for the 47 counties in Kenya and 12 regions in Ethiopia from 2016 to 2021, except for the Tigray region (no data from 2020); see maps for study counties and regions in figure 1. This allows for comparing trends across the years and corroborates any regularity or irregularity. The period from March 2020 to October 2021 was operationalised as the reference COVID-19 period (20 months) and from July 2018 to February 2020 (20 months) as the pre-COVID-19 period.

Figure 1
Figure 1

Top panel: map showing 47 counties in Kenya. Bottom panel: map showing 12 regions in Ethiopia.

HMIS data are known for quality issues, including completeness, internal or external consistency and outliers.26–28 We used a five-step process guided by WHO methodology for HMIS data cleaning.29 The steps included running quality checks to determine reporting rates, completion rates, outliers and missing data, data aggregation, analysis and reporting (online supplemental material figure 1). Identified anomalies were adjusted as follows: extreme outliers were identified using a modified z-score methodology30 with a cut-off of greater than an absolute 2.5 SD. We checked each data point identified as an extreme outlier to verify the status; once confirmed, we adjusted these using multivariate imputation by chained equations,31 specifically using the weighted predictive mean matching value for each calendar year. Similarly, missing data were corrected using imputation. Clean and adjusted facility-level count data were then aggregated to the subnational level (counties or regions) and used for analysis and reporting. Data consistency checks over time included triangulation with other data sources, including modelled estimates and survey data from International Health Metrics and Evaluation and Demographic Health Surveys.

Analysis

Proportion estimates were computed using aggregated subnational counts and population estimates from WorldPop32 for each location–year combination. Temporal trends (monthly and quarterly moving averages) for the aggregated proportions were plotted. Subnational geospatial heatmaps for annual proportion estimates, changes in access and usage between pre-COVID-19 and during COVID-19 and significance maps were plotted using R statistical software.33 We fitted interrupted time series (ITS) analysis using the Autoregressive Integrated Moving Average (ARIMA) model34–36 to assess whether COVID-19 interrupted health access for ANC and SBA in each of the subnational counties or regions. We also performed a counterfactual time series analysis37 by forecasting usage data post-March 2020, assuming there was no COVID-19, to assess its impact on the trends of health access usage.

We analysed and tested for mean differences (MD) (at a 5% significance level) for ANC fourth visit and skilled birth deliveries between pre-COVID-19 and reference COVID-19 periods to assess if there had been significant changes in service usage; thus comparing recent monthly data in COVID-19 reference years (2020 and 2021) with monthly data from the pre-COVID-19 period (2018 and 2019). Aggregated annual proportions were used to create geospatial heat maps to visualise the subnational ANC and SBA rates in the pre-COVID-19 and during COVID-19 periods separately and to show differences (no change, increase or decrease in usage) in ANC and SBA between the periods. Geospatial heat maps were also generated to highlight if the differences obtained in the difference maps were statistically significant at the set 5% significance level for each of the counties or regions.

The number of facilities included in the analysis for each country is presented in online supplemental material table 1. We show summary tables, graphs and geospatial maps, highlighting the hotspot counties and regions in Kenya and Ethiopia, respectively. More resilient subnational counties or regions in the study are operationalised as subnational areas that have made progress in increasing ANC coverage or SBA proportions, comparing their levels at the end of 2020–2021 relative to those of the counterfactual pre-COVID-19 years. On the contrary, the least resilient at the subnational level are those that experienced declining ANC coverage or SBA rates, comparing their levels at the end of 2020–2021 relative to pre-COVID-19 years (2018–2019).

Counterfactual analysis

We performed an ITS analysis to assess whether COVID-19 impacted healthcare usage. In an ITS study, a time series of a particular outcome of interest establishes an underlying trend that is ‘interrupted’ by an intervention at a known time. Here, we present the use of ARIMA modelling to quantify the impact of COVID-19. We used health usage data from the pre-COVID-19 period to make projections for the next 12 months, assuming the COVID-19 pandemic did not occur. We model the predictions with their 95% CI and compare this trend with actual data from HMIS observations. In areas where usage was projected to increase, that is, where the counterfactual used to be higher than the actual usage observed, this would infer that the COVID-19 pandemic has potentially impacted practical usage.

Role of the funding source

The study’s funder had no role in the study design, data collection, data analysis, data interpretation or writing of the paper.

Ethics

The study used anonymised and aggregated secondary data from health facilities and did not require ethics approval.

Results

ANC coverage over time (2016–2021)

Kenya

At a national level, there has been a slight, non-significant increase in ANC coverage from 49.6% in the pre-COVID-19 period to 51.4% in the COVID-19 period, representing a change of 1.8% (95% CI −1.03 to 4.5) (online supplemental material figure 2). Overall, service usage slumped from around September 2020 to January 2021, and the lowest usage was recorded in December 2020, at the peak of the second wave of the COVID-19 pandemic. Subnationally, changes in coverage have been heterogeneous, with some counties experiencing significant decreases in ANC usage and others experiencing increases (figure 2, online supplemental material figure 3). Specifically, the counties of Kiambu, Kajiado and Nairobi, among others, experienced higher ANC coverage proportions in the pre-COVID-19 period, but these declined during COVID-19 (figure 2, online supplemental figure 4). The changes that occurred in ANC coverage during the COVID-19 reference period were significant for most of these counties, showing an increase or a decrease (figure 2), but for some of the counties (ie, West Pokot, Busia and Migori, among others), the changes in ANC coverage were not statistically different comparing pre-COVID-19 and COVID-19 periods. At a 5% significance level, eight counties consistently showed significant declines in ANC coverage during 2020–2021, inferring that the recommended ANC usage was disrupted during the COVID-19 pandemic period. The disruptions were evident in the following counties: Kiambu (MD −23.9%, 95% CI −29.1 to –18.8), Kajiado (MD −15.4%, 95% CI −20.6 to –10.1), Embu (MD −11.0%, 95% CI −16.2 to –5.7), Turkana (MD −10.7%, 95% CI −16.1 to –5.3), Nairobi (MD −9.6%, 95% CI −14.7 to –4.6), Mombasa (MD −7.3, 95% CI −12.8 to –1.7), Nyeri (MD −5.70, 95% CI −9.96 to –1.5) and Meru (MD −4.4, 95% CI −7.07 to –1.78) (online supplemental material table 2, online supplemental material figure 3).

Figure 2
Figure 2

Top panel: ANC coverage (%) for Kenya in the pre-COVID-19 period 2018–2019 (left) and during the COVID-19 period 2020–2021 (right) (green=high ANC coverage rates or increase in coverage and red=low ANC coverage rates or decrease in coverage). Bottom panel: difference in ANC coverage between pre-COVID-19 and during COVID-19 period (left) (green=increase in ANC coverage rate, red=reduction in ANC coverage rates and white=no change) and its statistical significance (right). Data source: Kenya health management information system. ANC, antenatal care.

We zoom in on four of the eight counties with the most significant decline in population-level ANC usage between pre-COVID-19 and during COVID-19. Kiambu County declined from an estimated usage of 85.24% in the pre-COVID-19 period to 61.4% during COVID-19. Kajiado County declined from 68.08% to 52.72%, Embu from 60.94% to 49.97% and Turkana from 64.38% to 53.68% (online supplemental material figure 4). The p-values for the differences in service usage are all highly significant at the 5% significance level. The disaggregation of ANC coverage by counties highlights the inequalities in service usage at the subnational level, which would otherwise be masked at the national or country level. Quarterly trends further illuminate the trajectory of antenatal service usage and the periods of disruption concerning events that occurred over specific periods (online supplemental material figure 3), including indicating a time lag in the impacts on health service usage.

On the other hand, 19 counties experienced an improvement in ANC service usage rates between the pre-COVID-19 and during COVID-19 periods. These counties are Nyamira (MD 19.6%, 95% CI 8.61 to 30.49), Trans Nzoia (MD 10.20%, 95% CI 4.95 to 15.43), Makueni (MD 10.40%, 95% CI 5.01 to 15.79) and Siaya (MD 8.50%, 95% CI 2.49 to 14.55), among others (see figure 2 and online supplemental material table 2).

Ethiopia

We observe mixed results when comparing mean ANC rates for Ethiopia’s pre-COVID-19 and during COVID-19 periods, with no real change registered nationally (online supplemental material figure 5). ANC usage rates declined from 17.6% to 13.5% in Gambela (MD −4.1%, 95% CI −7.19 to –0.92), Afar from 52.66% to 48.52% (MD −4.1%, 95% CI −6.14 to –2.12) and Southern Nations, Nationalities and People’s (SNNP) region from 85.3% to 79.7% (MD −5.6%, 95% CI −8.01 to –3.29) (figure 3 and online supplemental material figures 6 and 7), with the Gambela region performing the worst even before the COVID-19 pandemic (figure 3). A few areas have experienced increases in ANC usage during this period, namely Sidama (MD 16.4%, 95% CI 9.65 to 23.23) and Somali (MD 9.6%, 95% CI 3.32 to 15.93) (figure 3, online supplemental material table 3). Data for the Tigray region were unavailable from October 2020 due to the ongoing conflict. The capital, Addis Ababa, hardly experienced any changes before and during COVID-19; however, this needs to be interpreted cautiously, as it could be an artefact of the population from outside the region (ie, Oromia region) that comes to seek care in facilities within the Addis Ababa region. This could inflate observed counts relative to the population in the Addis Ababa region. Overall, ANC usage averaged 60.3% before the COVID-19 pandemic, which increased to 61.8% in the 2020–2021 COVID-19 period (online supplemental material figure 5 and online supplemental material table 3).

Figure 3
Figure 3

Top panel: ANC coverage (%) for Ethiopia in the pre-COVID-19 period 2018–2019 (left) and during the COVID-19 period 2020–2021 (right) (green=high ANC coverage rates or increase in coverage and red=low ANC coverage rates or decrease in coverage). Bottom panel: difference in ANC coverage between pre-COVID-19 and during COVID-19 period (left) (green=increase in ANC coverage rate, red=reduction in ANC coverage rates and white=no change) and its statistical significance (right). Data source: Ethiopia health management information system. ANC, antenatal care.

Skilled deliveries coverage over time (2016–2021)

Kenya

Since 2016, SBA rates had been increasing before postelection violence from around August 2017,33 which led to disruptions in 2017 (see online supplemental material figure 8). Nationally, SBA usage increased from 64.8% in the 2018–2019 period to 76.0% in the 2020–2021 period (MD 11.20%, 95% CI 8.69 to 13.75) (online supplemental material figure 8 and online supplemental material table 4); this is, nonetheless, below the 90% coverage recommended by WHO. The top 10 counties that saw increasing SBA rates include Kericho, Trans Nzoia, Siaya, Nandi, Mandera, Makueni, Kirinyaga, Nakuru, Tharaka Nithi, Nyamira, Bomet and Uasin Gishu, among others (figure 4, online supplemental material figures 9,10 and online supplemental material table 4). It is evident that almost all counties experienced a sharp decline around December 2020, when the second wave of SARS-CoV-2 reached its peak; however, health campaigns encouraging the population to continue using health services during the pandemic led to increased SBA service usage despite the ongoing pandemic, leading to immediate improvements during the pandemic. Despite the reported national-level increase in SBA usage, subnational disparities exist (online supplemental material figure 10). Some counties experienced declines in usage rates during the same period, that is, Turkana (MD −14.9%, 95% CI −19.18 to –10.54), Marsabit (MD −3.3%, 95% CI −7.55 to 0.86) and Embu (MD −2.80%, 95% CI−7.64 to 2.08) counties (online supplemental material figue 10 and online supplemental material table 4). The boxplot in online supplemental material figure 11) zooms in on the decrease in SBA usage in Turkana from 72.52% to 57.66% and Marsabit County from 69.69% to 66.34% during COVID-19. These are some of the poorest counties in Kenya, and the pandemic exacerbated the inequalities in SBA service usage.

Figure 4
Figure 4

Subnational trends in Kenya’s SBA service usage in Kericho, Tharaka-Nithi, Makueni and Siaya counties. Trends from 2016 to 2021. The dark blue trend line denotes the monthly service usage. The blue trend line represents a smoothed estimation of service usage over the same period. The red vertical line indicates a time point (month) when the first cases of COVID-19 were identified and subsequent lockdown restrictions were enforced. The green vertical line denotes a time point (month) when COVID-19 vaccines were rolled out. SBA, skilled birth attendance.

As was done for ANC, the disaggregation of SBA rates by counties highlights the inequalities in skilled or health facility delivery usage at the subnational level, which would otherwise be masked at the national or country level. The quarterly SBA trends shown per county further illuminate the periods of disruption concerning events that occurred over specific periods.

Ethiopia

Nationally, there has been minimal change in SBA usage at the national level, increasing from 41.8% in the 2018–2019 period to 45.4% in the 2020–2021 period (online supplemental material figure 12 and online supplemental material table 5), which falls short of the WHO-recommended 90% coverage. However, at a subnational level, we saw declining SBA rates in the northern regions of Afar (MD −2.4%, 95% CI −4.16 to –0.54), Addis Ababa (MD −3.9%, 95% CI −8.87 to 1.14) and Amhara (MD −0.5%, 95% CI −2.34 to 1.42) (figure 5). Gambela and Benishangul-Gumuz regions in the West traditionally have poor SBA usage, and the changes observed in these regions were not statistically significant (figure 5). It is worth noting that during this period, a good number in the southern and eastern areas of Oromia, SNNP, Sidama, Somali and Harari (figure 5, online supplemental material figure 13) experienced improvements in SBA service usage following health campaigns implemented by the federal ministry of health in Ethiopia.

Figure 5
Figure 5

Top panel: SBA coverage (%) for Ethiopia in the pre-COVID-19 period 2018–2019 (left) and during the COVID-19 period 2020–2021 (right) (green=high SBA coverage rates or increase in coverage and red=low SBA coverage rates or decrease in coverage). Bottom panel: difference in SBA coverage between pre-COVID-19 and during COVID-19 period (left) (green=increase in SBA coverage rate, red=reduction in SBA coverage rates and white=no change) and its statistical significance (right). Data source: Ethiopia HMIS.

Counterfactual analysis to assess the impact of COVID-19 on health usage

ANC usage in Kenya experienced considerable disruptions compared with Ethiopia. Counterfactual ANC usage for Embu County, which was projected to increase to about 75% (95% CI 55% to 97%), is shown in figure 6 (left); however, the observed usage slumped to below 50% in the pandemic period. Ethiopia, however, demonstrated minimal disruptions to health usage; for example, a counterfactual analysis of use trends in the Dire-Dawa region showed that observed and counterfactual usage were in agreement, confirming that the pandemic has had minimal disruptions to healthcare usage (figure 6 (right)).

Figure 6
Figure 6

Counterfactual ANC usage in Embu County, Kenya (left) and Dire-Dawa region, Ethiopia (right). The blue trend denotes the observed ANC usage from 2016 to 2021. The green line indicates projected usage given the data in the pre-pandemic period. The red vertical line represents the time point (month) when the first cases of COVID-19 were observed in Kenya and Ethiopia. ANC, antenatal care.

Discussion

Our study assessed changes in maternal health service usage, especially ANC and SBA, at the subnational level by comparing pre-COVID-19 and COVID-19 outbreak periods. The results highlight variations in progress in maternal health services usage, specifically in skilled births and ANC services, as recorded in the routinely collected Kenya and Ethiopia HMIS databases. Results from the study show that subnational disparities are mostly masked in national-level analytical studies. This highlights the specific needs of the various counties and regions in Kenya and Ethiopia, respectively, as different areas are moving towards meeting national and global targets, such as the SDGs, at very different paces. Due to the variation of subnational contexts in geography, population density, urbanisation, socioeconomics, climate, policy environments and cultures, we should not expect uniformity in progress across counties, states and regions. Besides, the pandemic has not affected all counties and provinces similarly in terms of cases and deaths due to COVID-19. These subnational areas also vary in their responses and capacities to mitigate the adverse impacts of the pandemic. Thus, this study emphasises the importance of conducting analyses and depicting changes at the subnational level, even when there are no significant changes at the national level.

Unlike deliveries, where women may not have control over the timing of birth, ANC services are primarily preventive. They may be subject to lower usage if women perceive the service as less essential.11 38 39 Our study’s results are consistent with other studies conducted in LMICs. For instance, in India, a decline in the usage of antenatal services was observed, with about 33.5% of pregnant women having fewer ANC visits as compared with the pre-COVID-19 period.4 Women avoided routine antenatal check-ups due to the strict lockdown and delays in reaching a health facility.4 One qualitative study in Kenya reported that limited access to health facilities for antenatal services was due to the fear of pregnant women contracting COVID-19 and the insufficient number of health workers providing such services.18

Furthermore, most women started visiting antenatal clinics very late (sometimes in their sixth or seventh month of pregnancy) and missed necessary vaccinations such as tetanus toxoid; antenatal services among adolescents, including revisiting clients, were also reduced in Kenya.3 Again, other indirect impacts of COVID-19 potentially disrupt skilled deliveries as many women opt for home deliveries and deliveries at nearby non-health facilities due to inaccessibility resulting from lockdown and mobility restrictions, lack of transport and fear of contracting COVID-19 at the health facilities.39 Such practices, however, were not uniformly observed across the continent though. For example, Temesgen et al 40 found no change in Ethiopia, as about 8 out of 10 women had appropriate care during delivery at the health facilities, while more than 81.3% mentioned that COVID-19 had not strengthened the desire for home deliveries.

Besides the fear of contracting COVID-19, which has been suggested as a possibility for people avoiding hospitals and other health facilities, thereby under using health services, several reports have also been given on other fears such as police brutality and demonstrations spurred on by the lockdowns.41 Kenya is no stranger to violence sparked by protests and clashes between the people and law enforcement. In Nyeri, dozens of local vendors clashed in a riot with the police on 6 April 2020, over discontent with government-imposed mandatory COVID-19 business closures. The clashes reportedly occurred in the town centre following Nyeri County administration officials’ refusal to allow the vendors to reopen.42 Also, Kiambu, one of the top 10 counties hit by COVID-19, continued to be under stricter surveillance and restrictions due to COVID-19, along with the likes of Nairobi, Kajiado, Machakos and Nakuru. Indeed, after Nairobi and Mombasa, Kiambu was the third hardest-hit county in Kenya.42

The WHO, United States Agency for International Development and the Kenya Ministry of Health earmarked several counties as high-priority areas for focused maternal health services due to high maternal deaths,43 even before the coronavirus outbreak. Turkana and Marsabit, counties with widening disparities in both ANC and SBA, are among these priority counties that were already underperforming43 before the pandemic hit Kenya. Incidentally, their continued declines in maternal health service usage from 2018 to 2020 could be attributed to or exacerbated by disruptions from the pandemic, while acknowledging existing challenges pre-COVID-19. Turkana, Marsabit and Embu counties consistently have recorded declining rates for ANC and SBA compared with all other counties. Incidentally, these counties are reported to have one of the highest poverty and socioeconomic disparities in Kenya, where more than 60% of people live in poverty, according to the 2015–2016 Kenya Integrated Household Budget Survey. Again, for the mean per-person expenditure (a measure reflecting the cost of living determined by policy and market forces), Turkana ranks among the lowest of the 47 counties in Kenya compared with Nairobi and Mombasa.44 Turkana County presents a unique case when looking at the two indicators of interest. The county is lagging behind all others in making progress towards improving SBA as well as ANC rates for women. This could plausibly trickle down to increasing county maternal and under-five deaths.44 Recently, northern Kenya has been experiencing perennial droughts, potentially further exacerbating this area’s inequalities.

In Ethiopia, the data suggest that COVID-19 has not significantly disrupted the usage of ANC and SBA services. However, when focusing on ANC, the data shows that almost all regions have not progressed, far below SDG targets. Afar, Amhara, Benishangul-Gumuz, SNNP and Gambela’s rates deteriorated during COVID-19. In the remaining regions, rates have remained relatively stable, although below recommended WHO coverage. Over the assessment period from 2016 to 2021, we see that for most regions, sporadic or persisting low levels of ANC usage started in the pre-COVID-19 period and continued during the COVID-19 periods. Surprisingly, despite initial low levels of ANC in Harari, Sidama and Dire-Dawa, there has been a post-COVID-19 increase in usage. The lack of improved ANC attendance can be explained by the fact that ANC services are preventive, and women perceive them as unessential, which may lead to lower usage of the services.11 38 39 In addition, there are many barriers that women in Ethiopia have been reported to face when attempting to access ANC services. For example, difficulty navigating local health systems, the shame of unwanted pregnancy, a lack of partner approval, a lack of transport and high transport costs.45 COVID-19 may have compounded these barriers, but the overall downward trend in ANC usage began before 2020.

Assessing SBA in Ethiopia, we find that when comparing pre-COVID-19 and during COVID-19 rates, five southern regions have experienced significant increases in usage (Somali, Oromia, Sidama, SNNP and Harari), although below recommended WHO coverage. However, Amhara, Benishagul-Gumuz and Dire-Dawa city regions experienced no change between the two periods. Only the Afar region experienced a significant decrease. Looking at trend data, we note that COVID-19 has not deterred women from SBA over time, as rates in most regions are increasing during the COVID-19 period, although below SDG targets and WHO-recommended coverage. The southern regions of Ethiopia are performing better than the northern regions. Our results illustrate the importance of focusing on subnational performance. According to the World Bank’s Ethiopia socioeconomic dashboard,46 the poverty rate is highest in the northern regions of Tigray, Afar, Amhara and Benishangul-Gumuz. These regions of Ethiopia, in particular Tigray, Amhara and Afar, have endured many years of political and social unrest. It is, therefore, reasonable to expect service usage to be affected by this climate. It is somewhat surprising that in Amhara, despite these challenges, the region is experiencing improvement in SBA rates even during COVID-19.

It is reassuring to see that the pandemic has not resulted in rapidly deteriorating ANC and SBA rates and that some regions are either maintaining or improving the usage of these essential services. The counterfactual analysis results corroborate previous studies’ findings; for example, Temesgen et al 40 Although Ethiopia is improving and maternal and neonatal mortality rates are reducing, the changing pace is likely insufficient to attain the SDGs.

In Kenya and Ethiopia, urbanisation has been shown to be a key driver of COVID-19 attacks and death rates. People living in the urban settings of Nairobi, Mombasa, Addis Ababa and Harari, among other cities, experienced higher incidence rates of COVID-19.

Urbanisation in LMICs contributes significantly to overcrowding and poor sanitation; these conditions provide a perfect recipe for spreading infectious diseases such as COVID-19.47 In 2018, the urban population of Ethiopia was estimated to account for 21.2% of its 112 million people, and its urbanisation rate stood at 4.9%. At the onset of the COVID-19 pandemic, governments globally implemented non-pharmaceutical interventions to reduce the spread of COVID-19, including partial or total lockdowns of entire regions or countries. These measures meant that people spent more time at home, many of whom, especially in Africa, were in overcrowded conditions.48 Living in crowded housing conditions makes it harder to self-isolate and shield from COVID-19.

Our study reiterates that the impacts of the pandemic at the individual, societal and government levels have possibly affected the health-seeking behaviours of women in need of maternal health services through fewer services, delays in seeking care, delays in reaching healthcare services and delays in receiving adequate care at a health facility, all of which would subsequently increase and worsen the plight of women in using essential maternal health services.

Putting countries back on track to achieving the SDGs: What can governments do to get back on track to attain maternal health targets? How can lessons learnt from the study inform policies?

Ideally, with efforts being made in the policy environment being strengthened to accelerate the pace towards achieving maternal and child health targets, we expect to observe increasing trends in service usage. Nonetheless, with the unprecedented spread of COVID-19 in Kenya, in addition to already existing barriers to service usage, we can allude that in certain parts of the country, the pandemic has taken its toll and more so in Kiambu, Kajiado, Embu, Turkana, Nairobi, Mombasa, Nyeri and Meru counties. In Kenya and Ethiopia, the declines in maternity service usage have implications for women’s reproductive health and outcomes and infants’ safe delivery and survival.

This study makes the following recommendations based on the findings of this analysis, given that the COVID-19 outbreaks in Kenya and Ethiopia have potentially influenced localised interruptions of maternal health services, derailing progress towards SDG targets.

  1. The countries’ governments must act in resolution to safeguard pregnant women, unborn babies, mothers and newborns and ensure they can get the prenatal and postnatal routine care they need as governments alter their healthcare systems to mitigate the adverse effects of COVID-19 as well as any future public health threats.

  2. The government can also ensure that the continuity of maternal care services is adequate by appropriating budgeted funds equitably without diverting away from intended purposes, especially in the counties or states most affected by declines in maternal care services.

  3. Potentially, reported declines infer increased non-facility deliveries during COVID-19; thus, requisite sensitisation (community health talks and advocacy activities, among others) should encourage women to have skilled birth deliveries. Additionally, incentivise women for skilled birth deliveries by creating an enabling environment at the health facilities (through adequate training of birth attendants and midwives and providing essential equipment).

  4. We also recommend strengthening holistic reporting of health facility data in national data repositories, such as the HMIS, to capture public and private health facilities and ensure that the database is updated with new facilities offering healthcare. Additionally, governments, development partners and other stakeholders are encouraged to work together to improve HMIS data quality. This is a prerequisite to ensuring accuracy in data capture and inferences from any analysis employing such data sources to make sound policy recommendations to governments.

Strengths and limitations

This research possesses several notable strengths. First and foremost, it leverages routine health facility data, thereby offering the potential to supply real-time updates regarding the effects of COVID-19 on healthcare service usage. These datasets are generally characterised by increased frequency of data collection over an extended time frame, enhanced cost-effectiveness and expedited availability. Second, the study provides national and subnational estimates, facilitating decision-making for intervention planning at various administrative levels. Of particular significance is the capacity to unveil previously concealed changes in the data when analysed at the subnational level, which would otherwise remain obscured when examined solely at the national level. Furthermore, this investigation employs a comprehensive array of robust statistical methods for data analysis. These methods encompass geospatial techniques and counterfactual analyses, among others, enabling the presentation of a holistic view of the impact of COVID-19 on healthcare usage. Importantly, these analytical tools can be readily adapted for application in the context of other diseases and research domains. Lastly, the study conducts a comparative analysis of the COVID-19 impact in two distinct countries, offering valuable insights and lessons for other LMICs sharing similar demographic and healthcare profiles.

This study is not without limitations. The challenges in data consistent with routine health information systems could be expected to suffer lapses during external shocks to the healthcare system. Data accuracy is another challenge we encountered during analysis; HMIS databases are notorious for data quality issues, and using these data requires considerable cleaning and adjustment efforts. Additionally, the inherent nature and constraints of DHIS2 data did not allow the inclusion of confounding variables (such as urban and rural disaggregation) in the modelling. A critical potential limitation to our study is that it is predicated on the assumption that no external factors (beyond our control in measuring) would have caused interruptions to the healthcare system, such as inclement weather, power outages and individual and social factors. These could not be controlled for different ecological levels using aggregated health facility data. Therefore, our analysis should be interpreted within the context of health facility-based estimations of the disruption due to COVID-19.