Study approach

The study was designed as a retrospective study based on an online/telephonic survey questionnaire. The survey was conducted in four South African provinces, namely Gauteng, Western Cape, KwaZulu-Natal, and Eastern Cape. These provinces were selected for having the highest COVID-19 cases at the time of study preparation [11]. The study population included consenting adults residing in dwellings within the four provinces. Inclusion criteria included individuals aged 18 years or over and who were able to speak English, Zulu, Xhosa and/or Tswana.

Study participants were drawn from a market research company’s panel. Participants were passed through a double opt-in process that confers to the European Society for Opinion and Market Research (ESOMAR) guidelines [12]. This was followed by mobile authentication using a one-time password at registration. For the Computer-Assisted Telephone Interviews (CATI), participants from existing databases where the respondents were interviewed previously and opted to be recontacted in the future for research purposes were contacted to participate in the study. The sample was designed using Statistics South Africa and All Media Products Survey (AMPS) (i.e., re-weighted) population figures and quotas on demographics to match the South African national population demographics. Participants from a range of socio-economic-demographics were invited to participate. Online questionnaires were administered to those participants who could access online platforms and telephone surveys were done for those individuals who did not have access to online platforms (or mobile phone data, for example). Research ethics clearance for the study was granted by the South African Medical Research Ethics Committee (EC029–8/2020).

Survey questionnaire

A short introduction (information sheet) was given at the start of the survey. After written or verbal consent was given for online survey or telephone interview, respectively, the survey/interview was initiated.

The questionnaire was designed on the lines of the questionnaire developed in the UK for national longitudinal cohort studies [13], bearing in mind South Africa’s unique socio-economic circumstances. Answers were captured electronically for the online survey and by the interviewer for the telephonic interview.

The questionnaire was divided into six parts (Table 1) namely: Part A: general participant information; Part B: socio-economic-demographic information; Part C: HAP indoors: the month before Lockdown Levels 5 and 4 (i.e., before March 26, 2020), and during Lockdown Levels 5 and 4 (i.e., from March 26, 2020 to May 30, 2020); Part D: respiratory symptoms/illnesses during Lockdown Levels 5 and 4; Part E: any presence of COVID-19 test/case in the household during Lockdown levels 5 and 4; and Part F: agency/choice and health promotion for informed decision-making and lifestyle changes. Finally, participants were thanked for completing the questionnaire and informed that there are health risks linked to being exposed to HAP, especially when cooking with coal, wood, or paraffin indoors. Moreover, if they had any concerns, or wanted more information, they were advised to contact their healthcare providers.

A Living Standards Measure (LSM) which is a means of segmenting South African people according to their living standards, where 1–4 equate to the lowest, 5–7 intermediate and 8–10 the highest LSM status was known from data in the market research company’s panel and also included in the dataset.

Sample size calculation

To account for the different fuel use patterns across the four provinces (heating with solid fuels versus electricity) [14], it was hypothesized that there would be a change in domestic solid fuel burning post-Lockdown Levels 5 and 4 due to reduced incomes and more time spent at home cooking and heating, leading to higher electricity costs among those using electricity. It was assumed that there would be a 3% increase in solid fuel use, and a 0.1% increased use in electricity given the economic fallout from the pandemic, leading to a minimum total sample size of 1 148 participants, to achieve 80% power.

Data management and statistical analyses

Data were converted from a .txt file to Stata version 16.1 [15]. Demographic characteristics of participants, fuel use patterns and respiratory symptoms the month before and during Lockdown Levels 5 and 4 were summarized using frequencies and percentages for categorical variables and means and standard deviations for continuous variables.

To meet objective 1, changes in fuel use patterns the month before and during Lockdown Levels 5 and 4 were examined using the generalized McNemar’s test, to account for both discordant as well concordant pairs i.e., participants who experienced changes in their fuel use as well as participants who have not changed their fuel use patterns. P-values < 0.05 were considered statistically significant.

To meet objective 2, respiratory symptoms were compared for the month before and during Lockdown Levels 5 and 4 using logistic regression.

Objective 3 (i.e., to consider whether fuel use patterns for heating and cooking and other HAP-related behaviours changed when comparing responses for the month before and during Lockdown Levels 5 and 4) was discussed in light of the findings from objectives 1 and 2 in the discussion. No analyses were done stratified by province since this was an overall analysis.

Sample description

The survey was conducted between October 29, 2020 and December 2, 2020. Table 2 provides the sample descriptive statistics by questionnaire variable. Majority of participants were Black African and female. About 40% of participants were from Gauteng likely since this is the most populous province in South Africa. There is also free Wi-Fi in certain public spaces of major cities, e.g., Tshwane Free Wi-Fi in Pretoria, possibly making it easier to complete the survey, compared for example to the Eastern Cape province where Internet access may have been a challenge. In this instance, CATI was applied to try and include more participants from areas without internet access.

Comparison of fuel use patterns and HAP exposure the month before and during Lockdown Levels 5 and 4

Overall, electricity was the most commonly used energy source for cooking and heating followed by gas (Table 3). Patterns of HAP exposure factors, excluding fuel use types for cooking and heating, were similar based on self-reported responses for questions pertaining to the month before and during Lockdown Levels 5 and 4. Most participants (about two-thirds) reported naturally ventilating their homes and did not have presence of mould, animals, incense burning or environmental tobacco smoke in their dwellings both before and during Lockdown Levels 5 and 4. However, about one in three participants did report presence of one or more of these HAP exposure factors. This being a particular concern for environmental tobacco smoke which is known to contribute to HAP and adverse respiratory health outcomes.

Changes in fuel use patterns for cooking for the month before versus during Lockdown Levels 5 and 4 were not statistically significantly different (Table 4). The main contributions were due to changes from electricity to gas (66 people, 2.6%) and gas to electricity (43 people, 1.7%).

Changes in fuel use patterns for heating was significant (Table 5) though the actual numbers are small. The contributions were due to changes from coal to electricity (5 people, 0.2%), electricity to coal (14 people, 0.6%), electricity to gas (71 people, 2.8%), gas to electricity (44 people, 1.8%), electricity to paraffin (34 people, 1.4%) and paraffin to electricity (16 people, 0.6%).

Self-reported presence of respiratory health outcomes the month before and during Lockdown Levels 5 and 4

Self-reported prevalence of doctor or nurse-diagnosed health outcomes was generally low at less than 10%, except for dry cough. There were statistically significant differences in the responses for dry cough, wet cough and hay fever for the month before versus during Lockdown Levels 5 and 4 (Table 6) when less of the health outcomes were experienced during the latter.

One in three participants reported someone in the household having had a COVID-19 test during Lockdown levels 5 and 4. One in five respondents reported someone who had a positive COVID-19 test in the household during Lockdown levels 5 and 4.

Changes in household air pollution-related behaviours and experiences during Lockdown Levels 5 and 4

Majority of households reported cooking more, cleaning more and spending more time indoors during Lockdown Levels 5 and 4 compared to the month before Lockdown Levels 5 and 4 (Table 7). More than 80% of participants reported finding it difficult to make financial ends meet.

Study limitations

We acknowledge that this was a survey-based questionnaire that relied on self-reporting hence there was a risk of recall (or response) bias. We tried to address this bias by requesting that participants report on doctor or nurse-related health symptoms and diseases and not self-reported symptoms/illnesses that may be more easily forgotten. We also tightened the time frame for occurrence of health symptoms and diseases to the month before and during Lockdown Levels 5 and 4 which were difficult periods in everyone’s lives and more likely recalled.

This study was conducted using non-probability quota sampling which does entail sampling bias. We set quotas for different types of respondents, mainly dependent on their home province, with more people surveyed from provinces affected by COVID-19 (at the time of project set-up) being surveyed. The aim of quota sampling was to control the composition of the final achieved sample at the design stage already. We did conduct the survey using probability sampling using a sample frame, and therefore there were no “non-responses” for which to account. A different sampling approach may be useful to apply in the future with a focus on ensuring good coverage of all LSM groups since we had few participants from the low LSM groups.

Self-reporting happened through the telephonic interviews as well as the online self-complete interviews. This is a limitation because of the current pandemic that requires contactless interviewing methods rather than face-to-face. The number of questions that one can have in an online or CATI survey is also restricted based on previous experience from the company who assisted with the data collection. This limited the questions we were able to ask and therefore prohibited some analyses, for example, comparing an individual who said they had COVID symptoms and whether or not they had a COVID test, followed by whether it was positive or negative. Future studies, preferably face-to-face, would permit for longer surveys where more detailed information would be gathered about all individuals separately living in the household.

Due to budget constraints and the cost of CATIs, the sample was skewed to the online survey which included a higher proportion of people from middle-to-high income groups. This affected the proportion of households using fuels associated with HAP, i.e., not electricity. Future studies should consider spending more funds on CATIs to gather data from lower-income households more likely to be exposed to high levels of HAP from the burning of solid or “dirty” fuels.

Our survey questionnaire did not ask about the urbanicity of the participant’s dwelling, i.e., urban, rural, peri-urban, peri-rural etc. and this should be included in future surveys. We also did not ask about specific type of geographical zone, such as township, suburb, inner city etc. and this information would be useful in future studies.

While several of the HAP respiratory health outcomes are similar to those for COVID-19, the questions about whether anyone in the household had had a COVID-19 test and any positive cases of COVID-19 in the household were included to consider presence or absence of COVID-19 in households. Future studies, with questions about specific HAP-related respiratory outcomes and specific COVID-19 health outcomes, as well as associated risk factors, might be useful if one wanted to assess the impact of HAP exposure on HAP-related respiratory outcomes and also to assess the impact of HAP exposure on COVID-19 specific symptoms/ill health effects.