Assessing equity and the determinants of socioeconomic impacts of COVID-19: results from a cross-sectional survey in three counties in Kenya

Background: COVID-19 mitigation measures have major ramifications on all aspects of people’s livelihoods. Based on data collected in February 2021, we present an analysis of the socio-economic impacts of COVID-19 mitigation measures in three counties in Kenya. Methods: We conducted a cross-sectional phone-based survey in three counties in Kenya to assess the level of disruption across seven domains: income, food insecurity, schooling, domestic tension/violence, communal violence, mental health, and decision-making. An overall disruption index was computed from the seven domains using principal component analysis. We used a linear regression model to examine the determinants of vulnerability to disruptions as measured by the index. We used concentration curves and indices to assess inequality in the disruption domains and the overall disruption index. Results: The level of disruption in income was the highest (74%), while the level of disruption for domestic tension/violence was the lowest (30%). Factors associated with increased vulnerability to the overall disruption index included: older age, being married, belonging in the lowest socioeconomic tertile and receiving COVID-19 related assistance. The concentration curves showed that all the seven domains of disruption were disproportionately concentrated among households in the lowest socio-economic tertile, a finding that was supported by the concentration index of the overall disruption index (CI = - 0.022; p = 0.074). Conclusion: The COVID-19 mitigation measures resulted in unintended socio-economic effects that unfairly affected certain vulnerable groups, including those in the lowest socio-economic group and the elderly. Measures to protect households against the adverse socio-economic effects of the pandemic should be scaled up and targeted to the most vulnerable, with attention to the constantly evolving nature of the pandemic. the the associated long-term effects, it is imperative for the government to continuously monitor the pandemic’s indirect effects and implement appropriate mitigation measures during and after the pandemic. This paper presents an analysis of the burden, inequalities and the determinants of vulnerability to disruption across seven domains of livelihood (i.e. income, food insecurity, schooling, domestic tension/violence, communal violence, mental health and decision-making) in three Kenyan counties. Our findings reveal that more than a year into the COVID-19 pandemic, households still experience a high level of disruption in critical aspects of their livelihoods, with poor households being disproportionately affected. Given that the study was conducted before the third wave of the pandemic outbreak in Kenya, there are chances that the domains assessed in the study may have worsened. Therefore, it is critical to scale-up and ensure equity in the coverage of support programmes and initiatives to vulnerable households and population groups. Abbreviations: Author Contributions: Conceptualization, E.B., methodology, published manuscript.


Introduction
The World Health Organization declared the 2019 novel coronavirus disease (COVID-19) a public health emergency of international concern and a pandemic on 11th March 2020 (1). Given the uncertainties and negative health implications of the pandemic, countries have adopted various policy measures to flatten the epidemiological curve and preserve health systems capacity to meet the healthcare needs of the general population and COVID-19 patients (2,3). Such mitigation measures include, among others, 1) instituting country lockdowns, 2) imposing travel restrictions, 3) closing ports of entry, 4) closing schools and 5) banning large gatherings (2). Whereas these mitigation measures are critical in slowing down the spread of the virus, they are likely to lead to unintended health, economic and social impacts on populations, especially the poor and vulnerable, as the economic shocks associated with these mitigation measures are not borne equally within and across countries (3)(4)(5). In low-and-middle-income countries (LMICs), for example, evidence suggests a consistent picture of disruptions in livelihoods and economic shocks in the early stages of the pandemic (2,(6)(7)(8)(9)(10).
In Kenya, following the detection of the first COVID-19 case on 13th March 2020, the number of confirmed cases has accumulated to 217,276 and 4,273 deaths as of 13th August 2021 (11). The government has instituted a range of mitigation measures over time (Table 1). Nevertheless, depending on the intensity of the virus transmission, some of these restrictions have been lifted progressively since June 2020. At the outset of COVID-19 outbreak in Kenya, the government implemented a raft of measures to cushion individuals from potential socio-economic impacts of the pandemic mitigation measures. These included 100% tax waivers for individuals earning below KES 24,000 (USD 225), reduction of personal income tax from 25% to 30%, reduction of value-added tax from 16% to 14% and cash transfers to orphans, the elderly and vulnerable members of society (12,13). Also, to restore disrupted economic activity in informal settlements, youths in 23 informal settlements across seven counties were employed under the Kazi Mtaani program to clean streets, drainages and collect garbage (14). These measures notwithstanding, an increasing body of evidence points to significant socio-economic impacts of the COVID-19 mitigation measures. For instance, implementation of the mitigation measures has been associated with an increase in the number of domestic or gender-based violence cases, food insecurity, loss of income or employment, learning disruption, healthcare access disruption, increased levels of stress, among other forms of disruption in people's livelihoods (6,(15)(16)(17)(18)(19). Specifically, the implementation of various mitigation policies in Kenya has been shown to have disproportionate negative impacts on women dwelling in informal settlements in Nairobi as they were more at risk of facing income loss, food insecurity and foregoing needed healthcare (19).
Given the low levels of COVID-19 vaccine coverage, the evolving nature of the pandemic, and the associated long-term effects, it is imperative for the government to continuously monitor the pandemic's indirect effects and implement appropriate mitigation measures during and after the pandemic. This paper presents an analysis of the burden, inequalities and the determinants of vulnerability to disruption across seven domains of livelihood (i.e. income, food insecurity, schooling, domestic tension/violence, communal violence, mental health and decision-making) in three Kenyan counties.  Nationwide curfewtiming of curfew has been varied over time. The curfew in the 13 "hotzone" counties (Busia, Vihiga, Kisii, Nyamira, Kakamega, Trans Nzoia, Bungoma, Kericho, Bomet, Siaya, Kisumu, Homa Bay and Migori ) has been revised to 10 pm to 4 am.

March 2020 Ongoing
Economics State interventions to cushion Kenyans from economic shocks (tax refunds, rebates, waivers and cash transfers) and implementation of food aid programs including suspension of taxes on food items

March 2020 Accomplished
Launch of a National Hygiene Programme that would create jobs for the youth working in 23 informal settlements in across 7 counties April 2020 Accomplished Government asks Nairobi City County and Kenya Power not to disconnect water and electricity over unpaid bills.

Study Design
A cross-sectional study design was employed through a phone-based platform to administer a knowledge, attitudes, practices, and experiences survey in February 2021 as part of a series of studies conducted by the Population Council.

Study Population and Setting
The data used in this analysis was collected in a survey administered to participants sampled from households in three existing Population Council prospective cohort studies across three counties; Nairobi, Kisumu, and Kilifi (20). In Nairobi county, participants were drawn from two cohort studies in five urban informal settlements. The first cohort consisted of 2,565 households enrolled in the Adolescent Girls Initiative-Kenya (AGI-K) study in Huruma and Kibera (20,21), while the second cohort consisted of 4,519 households enrolled in the NISITU study in Mathare, Dandora and Kariobangi (19,20). The cohort study methods have been explained in detail elsewhere (22). The target population in Kisumu county, located in Western Kenya, were enrolled in the PEPFAR DREAMS study (20) and consisted of households in Kolwa (a peri-urban area) and Nyalenda (a large informal settlement). Participants in the DREAMS study were sampled from a list of compiled households by the Population Council. The target population in Kilifi county, located in the coastal region of Kenya, was drawn from households in three sub-counties (Magarini, Ganze and Kaloleni) who were enrolled in the Nia project (20). Participants enrolled in the study came from low-income households and informal settlements. Table 2 outlines the distribution of participants in the three counties. -NIA project n = 1,331 (508 males)

Sample size and sampling procedure
A ratio of 1:3 for male and female interviews was used to randomly sample households with phone numbers in the existing cohorts in Kilifi and Kisumu. In contrast, respondents from Nairobi were sampled by location. Because of inclusion criteria for the three existing cohorts, the randomly sampled participants were from households with at least one adolescent. As such, households that solely had adults or young children were not eligible for inclusion and are therefore not represented in the study.

Data collection
The tool used for data collection gathered information on 1) knowledge, attitudes and practices about COVID-19 reported by households, 2) the economic, education, food, social and mental health effects of COVID-19 mitigation measures, and 3) socio-demographic status. Two questionnaires were administered; one for adults and another one for adolescents. The tools were translated to Kiswahili and Dholuo, piloted and administered by trained interviewers over the phone. Interviews lasted for an average of 30 minutes. Data were collected using Open Data Kit. These included age, sex, level of education, marital status, socio-economic status, household size, location of residence (urban or rural), and receiving COVID-19 related assistance (Table 3).

Data Analysis
Household characteristics and asset ownership (i.e. electricity, piped water source, reliable water source, livestock, and mobile phone ownership) variables were used to generate the socioeconomic status (SES) index. PCA was used to compute the SES index, which was categorized into three tertiles.
Lastly, we conducted an equity analysis using concentration curves (CC) and indices (27) to assess whether the distribution across the seven domains and the overall disruption index disproportionately affected individuals in higher or lower SES groups. The CC plots the cumulative share of disruption for a given domain (y-axis) against the cumulative share of households, ranked from lowest SES rank to highest SES rank (x-axis). So, if everyone experiences disruption for a given domain irrespective of their SES rank, the CC will consistently lie on the equality (45-degree) line.
If, by contrast, the disruption domain is concentrated among individuals in the lower (higher) SES rank, the CC will lie above (below) the line of equality, with the distance between the CC and  issued the research permit (P20/5010) for the study. Verbal informed consent was obtained from respondents before initiating interviews.

Descriptive analysis
A total of 3,907 respondents took part in the study and their socio-demographic characteristics are outlined in Table 4. A higher proportion of the participants were female (67%), were in the 36-57 years age group (54%), were married (63%) and had primary level of education (46%). The average household size was 6.9 people (SD = 4.1) in the overall sample. Specifically, the mean household size was higher (mean of 8.8 people) in the rural county (Kilifi) compared to the urban counties (Nairobi and Kisumu) (mean of 5.9 people). A low proportion (4%) of the respondents reported having received any form of COVID-19 related assistance seven days before the interview.  Figure 1 illustrates the level of disruption across the seven domains. Overall, income disruption was the highest (74%) of the seven domains of disruption, whereas domestic tension or violence was the least (30%). Of interest, with few exceptions (i.e. income and food disruption), the level of disruption across domains was somewhat lower in Kilifi county compared to Kisumu and Nairobi counties ( Figure 1). Also, it is worthy to note that the level of school disruption was higher in urban counties (i.e. Kisumu (48%) and Nairobi (50%)) compared to the rural county (i.e. Kilifi (27%)).

Inequalities in disruption across domains
The concentration curves in Figure 2 suggest that disruption across the seven domains of interest in this study disproportionately affected respondents in the lowest socio-economic tertile (i.e. all concentration curves lie above the line of equality). Except for school disruption, these findings are confirmed with the concentration indices for disruption in Table 6. For instance, although not statistically significant, the overall disruption index was disproportionately concentrated among respondents in the lowest socio-economic tertile (CID = -0.022; p=0.074). Of all the seven domains of disruption, only two (i.e. food disruption and domestic tension/violence) were significantly concentrated among the lowest tertile (Table 6). Supplementary file 1 provides inequality findings at the county level.

Discussion
To inform policy formulation and target resources to mitigate the adverse socio-economic impacts consequences of COVID-19 mitigation policies at the early stages of the pandemic (18,19,29). Other African countries like Sierra Leone (24), Ethiopia, Malawi, Uganda, Nigeria (7) and South Africa (30) have increasingly reported food insecurity and income loss as the most prevalent and unintended consequences of implementing COVID-19 mitigation measures.
It is apparent that the various economic measures adopted by the government of Kenya to cushion Kenyans against the socio-economic impacts of COVID-19 have not had sufficient effects, possibly due to documented implementation challenges and low coverage (14). The finding that respondents who received COVID-19 assistance were still vulnerable to experiencing disruption underscores the fact that their needs are still not being met and hence there is an urgent need to scale-up coverage with these mitigation measures. Moreover, a previous study conducted in informal settlements in Nairobi showed that 86% and 48% of recipients of COVID-19 related assistance reported that their food and cash needs, respectively, were still not being met (19).

The reported increase in domestic tension/violence corroborates media reports and a previous
Kenyan study that reported an increase in the number of sexual violence cases per out-patient visit during the onset of the pandemic (15). Similar findings are reflected in an Ethiopian study that estimated 25% of women reported intimate partner violence (31). The high level of communal violence and mental health disruption (especially in urban counties) that were observed in this study can be attributed to, among other things, loss of income and food insecurity as a result of disruption in food supply chains that have been documented to disproportionately affect urban poor households (32). In addition, loss of employment or income during the pandemic, particularly for those working in the informal sector, has been increasingly documented to result in poor mental health (26,33,34). Our study also established that 42% of adolescents experienced school disruption due to COVID-19 mitigation measures, with urban counties reporting a higher level of school disruption. While the country implemented digital learning initiatives (35), it has been shown that this had little effect in ensuring learning continuity, given that access to digital devices and the internet is low and inequitably distributed (36). Similar trends were observed in a longitudinal study in Ethiopia, Malawi, Nigeria and Uganda that estimated that student-teacher contact dropped from 96% (pre-covid) to 17% a week before the survey (7). As schools in Kenya reopen, the government of Kenya, through the ministry of education, should implement measures to ensure all school-age children return to school.
Second, our findings revealed that being in an older age group and being married significantly increased an individual's vulnerability to disruption. This finding can be explained in several ways.
One, compared to middle-aged adults who are more likely to be economically active, elderly individuals are less likely to be involved in income-generating activities to be able to cushion themselves against the indirect effects of COVID-19 mitigation measures but rather depend on others. This position is corroborated by findings of a study from Uganda that showed that older people (above 60 years), among other things, lacked access to enough food, lost the little income generated from selling farm produce and could not access healthcare or interact with family and friends due to the COVID-19 regulations (25). Two, married women or women, in general, are more likely to be vulnerable to the indirect COVID-19 mitigation measures because compared to men, they are less likely to be employed, take on unpaid care burden for children and other household chores, depend on their male partners and are more at risk of experiencing gender-based violence, especially when employment or income is lost (37)(38)(39). Such gendered-disparities during the COVID-19 pandemic have been reported in previous studies in Kenya (19) and Zambia (33).
Third, the study finds that individuals in lower socio-economic ranks bear a disproportionate burden of disruption across the domains explored. This can be explained in two ways. Firstly, households in the lowest socio-economic tertile are already vulnerable to disruption given their low-income position, the likelihood that their employment is more informal and that they have little or no savings, coupled with limited access to social/financial capital to overcome the disruptions. This vulnerability was exacerbated by COVID-19 mitigation measures, as the concentration curves and indices revealed that households in the lowest tertile were disproportionately affected. Secondly, the finding that lower socio-economic groups are the most affected across the disruption domains suggests that either those who are most in need are not reached by the interventions targeted at them, or there is a need to scale up such interventions, with a focus on urban informal settlements and marginalized rural areas. This finding is particularly interesting since our study sample was drawn from predominantly poor locations, and hence the socio-economic ranking reflects a ranking among the poor. It is apparent that even among the poor, the intensity of poverty is associated with increased vulnerability to socio-economic disruptions from the pandemic. These findings compare well with evidence from nine LMICs, including South Africa, where missed meals, reduced income, delayed health access, and poor health was unfairly concentrated among households in the lowest socio-economic groups (5,6).
When interpreting the findings of this study, several limitations should be considered. First, given the rapidly evolving nature of the pandemic, the socio-economic impacts of COVID-19 mitigation measures quickly become dated, given the time lag between data collection and when findings are published. Second, the findings of this study are not generalizable to the entire population in the three counties since the participants were drawn from low socio-economic status households that have adolescents participating in ongoing cohort studies. As such, households without an adolescent were not selected. Third, given that data was collected from respondents who could such as peer-learning groups or deployment of community health volunteers to mobilize parents and caregivers to support the learning of girls, especially in urban informal settlements where school disruption was high.

Conclusion
Our findings reveal that more than a year into the COVID-19 pandemic, households still experience a high level of disruption in critical aspects of their livelihoods, with poor households being disproportionately affected. Given that the study was conducted before the third wave of the pandemic outbreak in Kenya, there are chances that the domains assessed in the study may have worsened. Therefore, it is critical to scale-up and ensure equity in the coverage of support programmes and initiatives to vulnerable households and population groups.