In this study, we found that overall tobacco use, including both cigarettes and smokeless tobacco products, was extremely low with no women reporting current use and less than 1 % reporting having ever used cigarettes or smokeless tobacco. In addition, the intention to use tobacco was remarkably low with nearly all participants reporting that they had no intention to try either cigarettes or smokeless tobacco in the next year. This is the first study to explore SHS exposure, at the community level, in the Southern region of Ethiopia. While the majority reported that smoking was never allowed in the home (i.e. high level of home smoking bans), one in seven women also reported potential exposure to SHS in their homes on a daily basis. Sociodemographic factors and household-level behaviors associated with reports of SHS exposure in the home were identified and significant differences were observed by place of residence.
Although the sample was homogenous in many ways, significant demographic differences were found based on place of residence or degree of urbanicity in relation to ethnicity, religious affiliation, number of persons per household, level of crowding, years of education, type of house lived in, household decision-making, and in nearly all of the MPI indicators (Table
1). The percentage of the sample classified as “multidimensionally poor” (13.6 %) is considerably lower than that of the national average (87.3 %) and the subnational SNNP regional level (89.7 %), as was reported in the 2011 EDHS [
40]. Information on nutritional status was not included in this analysis, however, while the MPI allows for re-weighting of the indicators when these data are not available, the exclusion of nutritional status may have limited the ability to identify additional sources of deprivation. Yet, when compared to the national data, the study sample had lower levels of deprivation in all of the remaining nine indicators, except for use of solid (biomass) fuel for cooking, with a higher level of solid fuel use in the study sample than in the national sample (99.7 % vs. 87 %) [
40]. Notably, less than 21 % of the participants were deprived of access to clean drinking water, as compared to more than 66 % of the national sample. The study sample also had higher levels of education, with only 11.3 % deprived in this indicator, as compared to more than 45 % in the national sample [
40]. Therefore, based on these indicators, the households in this study sample appear to have a higher standard of living than the national sample. However, when interpreting these findings it should be noted that Lakew & Haile, using national data from the 2011 EDHS, found that adults in the poorest wealth quintile were more likely to use tobacco than those in the richest quintile [
41].
Tobacco use
The tobacco use prevalence rate among women found in this study was lower than the available national data, which reported less than 2 % use of tobacco of any kind [
12], and was lower than other regional cross-sectional studies; 0.7 % in Gilgel Gibe, southwest Ethiopia [
23], and 0.2 % in eastern Ethiopia [
15]. When compared to the findings from the study conducted by Bloch et al., among pregnant women of reproductive age in nine LMICs, from which the current survey was adapted, the ever-tried rate in this sample was also less than that found in the two sub-Saharan African countries included in the original study, namely the Democratic Republic of Congo (14.1 %) and Zambia (6.6 %) [
28]. However, it is noted that participants in these two sites in the Bloch et al. study were primarily drawn from large urban cities.
Barriers to the social acceptability of smoking among women are still intact in sub-Saharan African region, and therefore, as has been reported in other LMICs, low social acceptability for female tobacco use and tobacco-related stigma may have led to underreporting of personal tobacco use [
1,
28,
42]. However, the overall low smoking rates and low prevalence of smoking among women of reproductive age in this area of Ethiopia highlight the opportunity to focus on primary prevention of tobacco-related diseases among women and their unborn children. Understanding the current social norms associated with smoking, the perceived benefits of all forms of tobacco use, and identification of contextual factors influencing tobacco use, while the prevalence is low, will help to inform the development of tailored primary prevention interventions that take into consideration the unique gender-specific motivations associated with tobacco use, or decision not to use. Given the size and regional variability in Ethioipa and the documented expansion of the tobacco industry, these varied findings underscore the urgent need for comprehensive tobacco use monitoring at the national level as recommended by the WHO FCTC [
43].
Khat use
Khat use was explored in this study as a covariate, and the ever use prevalence rate among women was the same as the overall prevalence reported nationally in the 2011 EDHS (11 %); however, only 3.4 % of the respondents in this current study who had ever used, reported having chewed
khat in the last 30 days versus 43 % of among the national sample (Table
2) [
12]. There were significant differences between rural versus urban participants, with more urban participants having ever used
khat as compared to rural participants (
p < .001). Additionally,
khat use was associated with reports of daily occurrence of smoking/SHS in the home. These findings support the findings of previous studies that underscore the importance of continuing to monitor the role that
khat use in this setting plays in relation to tobacco use and subsequently SHS exposure [
17‐
19].
Secondhand smoke exposure
Less than 8 % of women reported having a member of her household who was currently using tobacco products; however, over 14 % of the total sample, and 22 % of the urban participants, reported that smoking occurred in the home daily. While this item was not a measure of direct exposure to SHS, it does provide an estimate of the potential for daily exposure to SHS in the home. At the same time, only 5 % reported that their young children were “sometimes/frequently” or “always” in close proximity to someone who was smoking in the home. On the surface, there appears to be possible discrepancies between reported number of smokers in the home and reported frequency in which smoking/SHS occurs in the home, with approximately twice as many reports of smoking occurring daily in the home than number of households with a current tobacco user. These findings warrant further exploration, however, possible explanations include guests (non-family members) smoking while visiting, differences in definitions of “family member” and/or intentional efforts to limit young children’s exposure. Additionally, these descrepancies may speak to the presence of tobacco-related stigma and hence a hesitancy on the part of participants to report on family members’ smoking behaviors.
Less than 15 % of the sample reported that smoking was allowed in the their home. This is much less than reported by Reda et al. in the study conducted among a rural population in Eastern Ethiopia, where more than 52 % reported allowing smoking indoors [
15]. It is unknown whether the reported low rate of SHS exposure among children (5.1 %), and the high percentage of households in which smoking/SHS is “never allowed” (85.5 %) found in this current study are the result of the overall low smoking prevalence, differences in cultural norms, and/or intentional efforts to communicate and enforce smoking rules and maintain a smoke-free environment in the home. However, it is also possible that the prohibition of smoking in the home represents a dominant cultural norm that could be strengthened through public awareness interventions. Further study would be needed to identify the influencing factors.
In this sample, the strongest predictor of smoking/SHS occurring daily in the home was having a member of the household who used tobacco products. While this may seem intuitive, in a setting where the research on smoking behaviors is nascent, it would be important to be able to document that smokers are smoking indoors in their homes versus in courtyards or outside. Absence of a smoking ban in the home (i.e. “allowing” smoking), and exposure to point-of-sale advertising within the last 30 days, were also predictive of daily occurrence of SHS in the home. Both of these factors have significant policy and programmatic implications, as adoption of indoor smoking bans and restriction of tobacco advertising are among the key evidence-based policy recommendations that have been demonstrated to be the most successful tobacco control policies in the reduction of risk exposure and smoking prevalence across a range of settings, and are required by the WHO FCTC [
36,
43].
Interestingly, in this study, while the HDM scores were higher among urban participants and those with higher education [data not presented], the measure did not perform as expected in relation to the whether or not smoking was permitted in the house. The mean composite HDM score for those who reported never allowing smoking in the home was actually lower than the mean HDM score of those that did allow smoking in the home (Table
3). These findings appears to differ from qualitative studies that have described women’s engagement in household decision-making as a factor contributing to adoption of smoke-free homes [
44‐
46]. However, it is noted that the related item only asked whether smoking was “allowed” in the home, and not about the participants’ involvement in establishing or enforcing home smoking rules. There is also evidence in the literature that suggests that the concept of “avoidance self-efficacy” may be more predictive of behaviors associated with SHS exposure [
47]. While these findings may also speak to the need for more education in regards to the importance of maintaining a smoke-free home environment, further research is needed to understand the role that awareness, decision-making, self-efficacy, empowerment, and social status play in a woman’s ability to limit her own exposure to SHS and that of her children, particularly in in low-income settings.
Urbanicity
The stratified analysis of participants by place of residence resulted in a range of observed differences in both outcome variables and covariates. In previous studies increased urbanicity, or differing degrees of interaction and identification with urban settings, has been found to be predictive of more favorable attitudes toward smoking [
27,
42,
48]. Urban settings have also been associated with an increased prevalence of smoking among women both in high and low-income countries [
49]. If the EDHS criteria is applied, the entire sample in this study would be considered rural; yet, even this relatively small degree of difference in place of residence (i.e., residents from a small rural town compared to those from the immediate outlying rural districts) resulted in an increased likelihood of smoking/SHS occurring daily in the home, even after controlling for other factors (Table
3) [
12].
A number of reasons for the observed urban-rural differentials have been considered. Overall, the urban participants had less sources of deprivation, and higher levels of education than the rural residents, which may provide greater levels of expendable income that can be used for tobacco products. This observed difference may also be attributed to differences in social network tobacco-use norms in more urbanized settings. In a study conducted in 2008, among women in South Africa, Williams et al. found urbanicity to have an independent effect on smoking-related attitudes [
42]. In addition, urbanicity moderated the effect of network smoking norms on smoking related attitudes; but, it did not moderate cigarette advertising exposure. In a more recent study, also conducted in South Africa, smoking prevalence among women was associated with having spent more than half of their lives in urban settings (
p < .001), coping poorly with stress, and an increase in adverse life events; however, these factors were not significant among men [
50]. On the other hand, being poor was significantly associated with a higher smoking prevalence among both men (
p = .024) and women (
p = .002), while education level, employment status, and housing quality were not found to be significantly associated with smoking prevalence for men or women [
50].
While no studies conducted in sub-Saharan Africa were found which reported rural versus urban differences in SHS exposure, a number of studies from India have reported significantly higher levels of SHS exposure, in both household and workplace environments, among households living in rural versus urban settings [
51,
52]. Predictors of smoking have also varied based on place of residence; Singh and Sahoo found place of residence to be the strongest predictor among participants living in rural areas, while education was the most significant among participants in urban settings [
51]. These divergent findings indicate that further exploration is needed to elucidate factors in both rural and urban settings that are contributing to differences in smoking prevalence and indoor smoking-related behaviors.
In this study, significant differences were noted between rural and urban reports of involvement in household decision-making, with the urban participants reporting more involvement in decision-making in all areas (
p < .001). However, it is noted that the reported level of urban women’s involvement in decisions related to major household purchases (54 %) and visitation of family and friends (71.3 %) are still lower than that reported by women at the national level (66.2 and 78 % respectively) [
12]. The 2011 EDHS used a 4-item version of the household-decision making scale, which included an additional item on decisions related to accessing healthcare. Use of a different version of the tool may account for this difference; however, comparisons have yet to be made between these two versions of the tool. Another explanation for this difference may be the degree of urbanicity; if household decision-making increases with urbanicity, the lower level of household decision-making among this sample may be related to overall lower levels of urbanicity. However, these findings underscore the need for further research to understand the relationship between urbanicity, empowerment, decision-making, and SHS exposure.
Participants from the urban
kebeles differed significantly from rural participants in relation to ethnicity, with greater representation from various ethnic groups, level of adoption of smoke-free homes (i.e. smoking “never allowed”), exposure to point-of-sale advertising, and prevalence of ever
khat use. In bivariate analysis, each of these factors were also significantly associated with daily occurrence of smoking/SHS in the home. These factors were not highly correlated (
r < .50) with each other and thereby may begin to help characterize factors in the more urbanized environments that are influencing tobacco use and risk of SHS exposure. For example, the greater percentage of non-Sidama ethnicities found within the urban
kebeles may be indicative of migration from other areas of the country that have different social norms concerning tobacco use and SHS exposure [
42].
A number of recent studies from China and India have demonstrated a strong association between rural-to-urban migration with an increase in smoking prevalence [
53‐
56]. Migration has been associated with changes in social networks and influences, changes in self-definition, as well as increased vulnerability to a range of other health risks which may also begin to give insight into the observed differences in rates of smoking and indoor SHS exposure [
42,
57]. An emerging body of knowledge outlines the unique risk factors that can be attributed to urban settings [
57‐
60]. These risk factors can vary based on differences in social, cultural and physical environmental contexts [
57]. Additionally, factors such as gender and ethnicity can mediate these risk factors [
58]. Therefore, it is critical that tobacco control policies and interventions be informed by local data. Furthermore, the findings from this study provide support for prioritization of tobacco use prevention efforts in both rural towns and
emerging urban areas. Further research is needed to understand the dynamics that contribute to the increased rates of smoking by household members and SHS exposure in these settings. Additionally, there is a need to standardize methods used to characterize neighborhoods and communities, and to incorporate more complex evaluation of urban-rural differences versus simple binary urban-rural measures, in order to generate comparable data across studies [
60,
61].
Strengths and limitations
This study adds to an emerging body of evidence on women’s tobacco use and SHS exposure in Ethiopia. The study was strengthened by the use of items from standardized validated instruments, and use of translation and adaptation guidelines that have been used previously in LMICs, [
28,
32,
35,
62]. The generalizability was enhanced by use of a systematic sampling technique to select households and the reliability of the self-report of SHS exposure was strengthened by use of an interviewer-administered survey [
63].
At the same time, several limitations of this study should be considered. First, the outcome variables relied on self-report, which can be susceptible to a number of non-sampling errors. However, self-report methodology has been used extensively in tobacco-related research and has been demonstrated to yield accurate estimates of prevalence in validation studies using biomarkers [
36,
63]. Furthermore, in this study a number of recommended steps were taken to mitigate these types of error, including careful consideration of question wording assurance of privacy and confidentiality, and use of an interviewer-administered questionnaire to improve accuracy [
36,
63,
64]. An interviewer-administered survey also increases the inclusion of women with low literacy levels, yet, it is noted that this may also have introduced social acceptability bias, particularly in settings, such as this one, with a potential for high levels of tobacco-related stigma and defined gender roles that limit women’s autonomy, thereby potentially resulting in underreporting of personal and/or family member tobacco use. Second, during the data collection exercises, approximately 20 % of the households approached (
n = 137) were not at home. This introduces a potential for selection bias, as it is unknown if these residents may have differed in some way from participants that were home during the day. Additionally, while differences were found between rural and urban
kebeles, rural
kebeles still had relatively close proximity to the town (i.e., less than one hour walking distance) and it is unknown whether the other
kebeles in the Aleta Wondo district, which are further away from town, may have differed in the variables measured. Finally, the small sample size prevented the testing of 2-way interactions in the logistic regression model.