Parental Risk Perceptions of Child Exposure to Tobacco Smoke

Laura Rosen; Inessa Kostjukovsky

Disclosures

BMC Public Health. 2015;15(90) 

In This Article

Discussion

To the best of our knowledge, this is the first time that the recognized risk perception dimensions of likelihood, susceptibility, and severity[6] have been used to understand parental risk perceptions due to child tobacco smoke exposure. For the combined risk perception scale as well as for the three independent elements of likelihood, susceptibility, and severity, the relationship between risk perceptions and parental smoking status was clear: parents who were regular smokers assessed risk as significantly lower than did other participants (Univariate analysis: p = .0003, Multivariate analysis: p = .0158). The relationship between risk perceptions and family smoking in the home was less clear: higher assessments of risk were significantly associated with less family home smoking in the univariate model (p = .0350), but not the full multivariate model (p = .3224). Once ethnicity was removed from the full model, the significance decreased to borderline level (p = .0560). Particularly since ethnicity was significantly associated with risk perceptions (Model 1), it is possible that ethnicity acted as an effect modifier on the relationship between home exposure and risk perceptions, while controlling for smoking status.

Our finding of a clear association between PRETS and smoking status has strong support in the literature, with at least five previous studies with similar findings (Bock,[9] Chen,[10] Evans,[27] Lonergan,[12] and Lund[13]), and a single study with opposite findings.[28]

Our finding that the association between PRETS and family smoking in the home was statistically significant in a simple analysis, but not in multivariate analyses, is similar to findings by Winickoff:[16] In his study, the relationship between secondhand smoke beliefs and a strict home smoking ban was significant in a bivariate but not multivariate analysis.

Evans and Gilmore[27] found that knowledge was associated with smoke-free homes, when using univariate analyses, and Lund[13] found an association between child exposure to tobacco smoke and health risk awareness using a univariate analytic approach. Helgason[29] found that health risk awareness was associated with TSE in the home in a multivariate analysis.

Based on our findings and on these studies, it seems that the association between PRETS and parental smoking may be somewhat more well-established than is the association between PRETS and smoking in the family home (and this relationship may be modified by ethnicity or other variables). This is somewhat counter-intuitive: PRETS is specifically about child exposure, not about parental smoking, and so one might assume that the more direct relationship between PRETS and family home smoking would be easier to establish. One possible explanation may be found in directionality of associations. Associations found in studies at a single point in time do not prove causality of effect in one direction or another.[30] In the case of PRETS and parental smoking, it is possible that the smoking behavior of the parent influenced his/her perceptions of risk of secondhand smoke for children, rather than that risk perceptions about child exposure influenced the decision of whether or not to smoke. This is especially true since the decision to start smoking was most likely made before the children were born. Another possible explanation is that while parental smoking status is unique to the parent, as are their perceptions, family smoking behavior in the home is a function of the behavior of that parent as well as others who live in the home. Even a parent convinced of harm due to tobacco smoke exposure may be unable to determine the behavior of others residing in the home. This could lead to a weaker association between PRETS and family smoking behavior. Finally, our findings suggest that there may be a more complex relationship between PRETS and family smoking in the home, which is modified by ethnicity.

Measuring Parental Risk Perceptions

Previous investigators have explored parental attitudes, knowledge, and risk perceptions regarding child tobacco smoke exposure. Questions asked by some previous authors are presented in Table 5, with a note on inferred dimension (likelihood, susceptibility, or severity; we added the possibility of a fourth category, "knowledge" to categorize questions which dealt with harm, without explicit reference to either likelihood, susceptibility, or severity). With the exception of Wagener,[23] investigators generally collected information on a single dimension. For example, Chen,[10] Drehmer,[22] McMillen,[17] and Winickoff[16] all asked about harm related to tobacco smoke exposure, using a question such as: "Inhaled smoke from a parent's cigarette harms health of infants and children".[17] Other investigators, for example, Evans,[27] Helgason,[29] Lonergan,[12] and Lund,[13] asked about increased susceptibility to illness due to tobacco smoke exposure, with question such as: "Do you think that living with someone who smokes does, or does not, increase a child's risk of asthma/ear infection/cot death/chest infections/other infections?".[27] As in our study, susceptibility due to tobacco smoke exposure – not due to personal characteristics – was of interest. Bock[9] and Farber[11] asked about severity:"How much do you think other people's smoking affects your baby's health?" Wagener built constructs for precaution effectiveness, optimistic bias, and perceived vulnerability; his "optimistic bias" was closest to our susceptibility measure, while his perceived vulnerability included some questions about severity.[23]

As can be seen, there are currently no accepted standards for measurement of PRETS. Nor is it clear what the optimal way to measure PRETS should be, or how this should be determined. Two issues are of interest. First, measures with good predictive power would allow us identify populations subgroups which are likely to expose their children to tobacco smoke, and allow targeting of those groups for interventions. Second, the identification of dimensions of risk perception which differ between smoke-free and other homes could help craft messages to convince smokers to keep their homes smoke-free.

Some research has been done on comparing different types of measurement for risk perceptions and other psychosocial variables.[8,21,26,31] Baghal, in an article on measuring risk perceptions relating to smoking,[8] emphasizes that different measures may lead to different conclusions. He addressed verbal and probability scales, specifically addressing the relative benefits of asking about absolute risk, relative difference between smokers and nonsmokers, relative risk between smokers and nonsmokers, and vague quantifier scales, and concluded that "numeric measures are inconsistent with logical semantic understanding." He also found that vague quantifier scales had better predictive power than did numeric scales.

Measures constructed from multiple questions are sometimes used in order to decrease the variability, or "noise" which would be present if a single question is measured. In addition to decreasing noise, there is an additional benefit to studying different dimensions of a perception: the answers regarding the individual dimension, and not just the constructed measure, may provide directions for successful interventions. For example, if those who allow their children to be exposed have relatively lower perceptions of severity of damage, then it might be worthwhile to address severity in the context of interventions, rather than susceptibility. On the other hand, if the key difference between those who expose and don't expose their children is perceived susceptibility, the messages would better be focused on the additional risk of illness due to tobacco smoke exposure.

Our study showed that of the three components of the risk perception scale measured in the present study, the largest absolute difference between regular smokers and others in the mean risk perception value, and the smallest p-value, was for the severity component. Regarding family smoking in the home, the only dimension to reach statistical significant in a univariate analysis (controlled for smoking status) was susceptibility.

A review by Janz of the literature regarding the Health Belief Model showed that preventive health behaviors were associated with both susceptibility and severity. He found that perceived severity was weakly associated with preventive health behaviors, while perceived susceptibility was a relatively stronger contributor to preventive health behaviors.[7]

Some evidence for the importance of severity in parental smoking behavior around children comes from a qualitative study of mothers of children with respiratory illnesses who smoked.[32] The authors found that some mothers disputed the severity of exposure to passive smoking, with one mother quoted as saying: "I don't know if I believe it's that bad for you. Not that bad" (p.108) Another mother was quoted as saying: "illness [was] not severe enough to warrant behavior change"; and another: "If he was really really severely asthmatic, I wouldn't be smoking now.' While far from conclusive evidence, these quotes suggest that underestimating the severity of harm due to TSE may be an important factor in parental smoking behavior around children. If true, this could provide a clue to creating effective interventions.

Strengths and Limitations

The strengths of this study include an innovative approach to measuring parental risk perceptions concerning child exposure to tobacco smoke, through the application of accepted psychological concepts, and the examination of the association between these risk perceptions and smoking status of respondent and home smoking behaviors. Careful pretesting of the survey instrument, and a very high response rate (nearly 98%), represent further strengths.

The cross-sectional nature of this study limited our ability to determine whether risk perceptions influenced behavior, as is postulated by the Health Belief Model,[5] or visa versa. As previously described,[30] risk perception can influence behavior, and behavior can influence risk perception. Directionality can be assessed by longitudinal designs (experimental or observational), but not by a cross-sectional study such as this one.

The study population was not randomly sampled, however, concerns about possible selection bias are somewhat mitigated because people were recruited as they entered the clinic, and the response rate was nearly 98%. Participants were recruited from a single large clinic which represented a heterogeneous population. It is not possible to provide population level estimates for any variables, due to sampling of equal numbers of regular smokers and others. We did not explore risk perceptions due to third hand smoke exposure.[16] Our question on susceptibility related specifically to susceptibility due to tobacco smoke exposure, and not to innate, unique susceptibility of a particular child for reasons other than tobacco smoke exposure. Our findings suggested that ethnicity may be acting as an effect modifier on the relationship between family home smoking and risk perceptions. A larger sample size is necessary to further explore this relationship and to permit inclusion of more variables in the statistical model. Further study is needed to understand differences in perceptions between occasional, former, experimental, and never smokers, and explore these relationships in other populations.

processing....