Objectives: We estimate for young children the annual excess health service use, healthcare expenditures, and disability bed days for respiratory conditions associated with exposure to smoking in the home in the United States.
Methods: Health service use, healthcare expenditures and disability bed days data come from the 1999 and 2001 Medical Expenditure Panel Survey (MEPS). Reported smoking in the home comes from the linked National Health Interview Survey, from which the MEPS sample is drawn. Multivariate statistical analysis controls for potential confounding factors. The sample is 2759 children aged 0–4.
Results: Smoking in the home is associated with an increase in the probability of emergency department visits for respiratory conditions by five percentage points and the probability of inpatient use for these conditions by three percentage points. There is no relation between indoor smoking by adults and either ambulatory visits or prescription drug expenditures. Overall, indoor smoking is associated with $117 in additional healthcare expenditures for respiratory conditions for each exposed child aged 0–4. Indoor smoking is also associated with an eight percentage point increase in the probability of having a bed day because of respiratory illness for children aged 1–4.
Conclusions: Despite the significant progress made in tobacco control, many children are still exposed to secondhand smoke in their home. Reducing exposure to smoking in the home would probably reduce healthcare expenditures for respiratory conditions and improve children’s health.
Statistics from Altmetric.com
There is now definitive evidence that secondhand smoke exposure causes disease and premature death in children and adults who do not smoke. Because their lungs are not fully developed, children are at particular risk from secondhand smoke. Children who regularly breathe in secondhand smoke are at increased risk of sudden infant death syndrome (SIDS), ear infections and respiratory diseases such as asthma, pneumonia and bronchitis. Smoking by parents has been shown to cause respiratory symptoms and slows lung growth in their children.1
Even more disturbingly, children are more exposed to secondhand smoke than adults.2 The predominant sites of exposure for children are their homes, because children, especially young children, spend most of their time at home and generally cannot remove themselves from harmful living environments. In the United States, despite the general decline in adult smoking prevalence, almost one in three children lives in households with at least one adult smoker,3 and at least one in four children were exposed to secondhand smoke at home.4
We estimate the extent to which exposure to smoking in the home affects young children’s health and health service use and expenditures. We focus on healthcare use and expenditures related to childhood respiratory conditions for two reasons. Firstly, respiratory conditions are among the most prevalent and costly childhood illnesses. From the 2003 Medical Expenditure Panel Survey, we calculate that for children aged 0–17, expenditures for respiratory illnesses were $10.7 billion, 13% of total healthcare expenditures on this age group. Respiratory conditions account for roughly one in every five hospitalisations or emergency department visits. For children aged 0–4, expenditures for respiratory illnesses were $4.3 billion, representing an even higher percentage of total healthcare expenditures on this age group (18%). For these younger children, one in every four hospitalisations or emergency department visits was for respiratory conditions. Secondly, the causal relation between secondhand smoking and childhood respiratory illness has been clearly established in the medical literature.1 Young children who are exposed to tobacco smoke are at increased risk of respiratory track infections such as bronchitis and pneumonia. Tobacco smoke can also exacerbate existing asthma. These adverse events may require costly emergency department visits or hospitalisations. We must understand how much exposure to cigarette smoke at home contributes to the high costs of treating these conditions.
Few studies have attempted to measure the burden of secondhand smoke on the health care of the children in the United States. Early studies were largely single-site small-sample studies that produced mixed findings on whether exposure to secondhand smoke increases children’s medical service use.5–7 Stoddard and Gray’s study8 is the only one that attempted to estimate the additional healthcare expenditures associated with secondhand smoking using nationally representative data. They found that maternal smoking increases annual expenditures for respiratory conditions for children aged 0–5. The most recent data used in the above studies were from the late 1980s. Frequently, parental smoking was used to proxy for exposure to secondhand smoke, but the validity of this assumption may have declined over time. Some smokers refrain from smoking in the home, especially when children or other non-smoking adults co-reside. Gilpin et al9 reported that in California, almost half of smokers live in smoke-free homes, and the trend is towards less exposure. Recent studies from the United Kingdom, Canada and Hong Kong have used more accurate measures of exposure, such as self reported smoking in the vicinity of the child, regular smoking at home or salivary cotinine level.10–12 These studies typically found that smoking in the home increased health service use, particularly hospitalisations because of asthma and other respiratory illnesses.
Our study uses more recent US data and employs one of the better measures of exposure to secondhand smoke, self reported smoking inside the home. We take advantage of linkage between the nationally representative Medical Expenditure Panel Survey (MEPS) and National Health Interview Survey (NHIS). The MEPS collects detailed information on health service use, healthcare expenditures, bed days for medical conditions and whether each adult in the household smoked. The NHIS survey, from which the MEPS sample is drawn, asks whether smoking occurred inside the home. Results from our study will help to quantify the significance of secondhand smoking in terms of children’s health and health care.
We use two nationally representative surveys: the MEPS, which has detailed service use and expenditure data and many family characteristics, is augmented with information about smoking from the NHIS. In the NHIS, one adult per family is sampled for detailed questions. In the 2000 cancer module, the NHIS asked sampled adults how many days anyone (resident or non-resident) smoked in the home in the past week, and in the 1998 adult prevention module, the sampled adult was asked how many days in the week anyone usually smoked in the home. We combine answers from both years to increase our sample size. Although NHIS asked the smoking inside the home question again in 2005, linked MEPS expenditure data are not yet available. The NHIS did not ask about smoking inside the home in other years. Validation studies have found that parental reporting of exposure to tobacco smoke in the home is an accurate measure of secondhand smoke at home.13 14
Each year, the sample for the MEPS is drawn from homes that responded to the NHIS in January through October in the previous year. The MEPS data for this study, therefore, come from the fourth and sixth panels (beginning in 1999 and 2001, respectively). Each family was interviewed five times over two and a half years to obtain annual data reflecting a two-year reference period.15 Each interview asks about hospital, physician, prescription drugs and other services received since the last interview. The household reported expenditure data are supplemented with information from interviews with a sample of the medical providers identified in the household interviews. In an annual self administered questionnaire, each adult in the household is asked whether she or he currently smokes.
The analytical sample is all children aged 0–4 in the first interviews of Panel 4 and Panel 6 of the MEPS. We focus on these young children because, typically, they spend more time at home than school-age children, and therefore face greater risks from smoking inside the home. There are 2759 children in the sample. Results are weighted to reflect the average of the populations in 1999 and 2001.
We compare young children with three patterns of exposure to household smoke: (1) those living in households in which anyone smoked inside the home one or more days of the week (“smoking inside the home”), (2) those living with adult smokers who did not report smoking inside the home any day of the week (“smoking outside the home”), and (3) those living with no adult smokers, and no one else smokes inside the home. We dichotomise the indoor smoking measure because the reported frequency of indoor smoking lacks variation: 80% of the sample exposed to smoking inside the home were exposed seven days a week.
Outcomes and multivariate analyses
We study hospital stays, emergency department use, prescription drug expenditures, total expenditures and bed days related to respiratory conditions. The MEPS household respondent reported the health conditions related to each hospital stay and provider visit; the conditions for which drugs were prescribed; and the conditions that caused days spent in bed. The conditions reported by the respondent were recorded by the interviewer as verbatim text, which was then coded by professional coders to ICD-9-CM codes. Conditions with ICD-9 codes 460 through 519 were classified as respiratory conditions. Visits, stays, expenditures and bed days related to these conditions were identified and summarised. Expenditures were inflated to 2004 dollars using the personal healthcare expenditure price index.16
We use multivariate statistical analysis to control for differences in characteristics of children and their families in smoke-free homes and those exposed to smoking in the home. We sought to control for any factors that could be confounded with the effects of smoking, by including the same large set of covariates, sometimes multiple measures within the same construct, in each model. Nearly all these variables are from the MEPS. Children’s characteristics are age, race, sex, census region and urbanicity. Family characteristics are poverty status (income as a proportion of the federal poverty line for the family size, in four categories); single parent, two parent and no parent families; and changes in the adults in the home between the NHIS and MEPS. We control for time demands on the primary caregiver, usually the mother, which may reduce her available time to seek medical care: number of other children, an indicator for whether the primary caregiver works and wage rate if the primary caregiver does work. We control for the primary caregiver’s attitudes towards medical care and risk, collected in the MEPS adult self administered questionnaires, and seat belt use, because these attitudes could cause both smoking and variation in propensity to take children to the doctor. An indicator for whether the child has a usual source of care controls for access to care and any differences in parental attitudes towards health care for the child. The child’s insurance status controls for the differential costs of care and any differences in parental attitudes towards obtaining insurance for children. We control for state policies that could separately affect children’s exposure to smoke: the tax on cigarettes and an index of state policies restricting smoking in public places. Variables from the Area Resource File control for the supply of paediatricians and hospital beds.
The type of multivariate model varies with the outcome variable. For binary outcomes (any emergency department visits, any inpatient use and any bed days), we estimate logistic regressions. For number of visits, we estimate a zero inflated Poisson model. This model accounts for the 65% of young children who did not have any visits for respiratory conditions, because there are separate coefficients for the likelihood of any visits and the number of visits. For expenditures, we estimate a two-part model. The first part is a logistic regression for whether there were any expenditures for respiratory conditions. Fifty-nine per cent of young children had no such expenditures. The second part is a gamma model with a log link estimated on only young children with expenditures greater than zero. This non-linear model accounts for the highly skewed distribution of expenditures. For each outcome, we calculate the average marginal effect of smoking inside the home, which is the mean change in the outcome for exposed children if they lived in a smoke-free home. We use balanced repeated replication to estimate standard errors that account for the complex survey design of the MEPS.17
Exposure to household smoke is missing for 28% of the sample. Smoking status is missing for about 12% of the adults living with children in our sample because of non-response to the MEPS self administered questionnaire. NHIS data on indoor smoking are missing for 22% of the sample, overwhelmingly because the interview with the NHIS sample adult was not completed. Because data are missing primarily because of non-response to components of the surveys, missingness is not likely to be related to reluctance to answer questions about smoking. Children with missing smoking status and those with complete smoking information had similar characteristics (see Tobacco Control website, table A.1, http://tobaccocontrol.bmj.com/supplemental). They were similar in their gender, race, ethnicity, insurance status, poverty status, number of co-residing parents, parental education, family home ownership, and primary caregiver’s employment status and attitudes towards care. Children with missing smoking status, however, were more likely to live in families that have more children, or reside in the West, and have missing attitudinal data because of non-response to the self administered questionnaire that also collected adult smoking status; and were less likely to reside in the Midwest or live in families that own a home. To control for potentially non-random missing data and maintain sample size, we keep these observations by including a missing smoking status indicator in our multivariate models.
We test whether our results reflect the effects of indoor smoking per se, rather than unobserved factors, such as parenting practices, which may be correlated with smoking. We study adherence to the American Academy of Pediatrics’ (AAP) guidelines18 for well child visits. If the smoking variable reflects unobserved factors, then we would find poor adherence to guidelines among smokers. For example, parents who focus more on the present than the future might not invest in their own or their children’s health by smoking in the presence of their children or not taking them to well child visits, but smoking per se would not be the cause of poor adherence to guidelines. Well child visits were visits identified as such by respondents, and we also include general check-ups and visits for vaccinations or immunisations. For each child, we calculate the ratio of the child’s visits during the calendar year to the number of recommended visits, using a methodology similar to a recent article by Selden.19
Two more analyses assess the validity of our measure of exposure to smoke. Firstly, we exclude from the analysis infants who were born between the NHIS interview and the end of the first year of the MEPS, if there were no other children already in the family, because parents may temporarily stop smoking during pregnancy. We use the same control variables and regression models. Secondly, we restrict our sample only to those who reported indoor smoking and those who reported no smoking by any adults in the household. We exclude smokers who reportedly did not smoke indoors, because they might not accurately report indoor smoking.
Exposure to smoking inside the home is associated with more hospital stays and emergency department visits, higher expenditures and bed days. Among young children exposed to smoking inside the home, 4.3% had at least one hospital stay for respiratory conditions per year, compared with 1.1% of children from homes with no adult smokers (p<0.01) (table 1). Among children exposed to smoking inside the home, 8.5% had at least one emergency department visit for a respiratory illness per year compared with 3.6% of those without adult smokers (p<0.01). Reflecting the big differences in emergency department visits and hospital stays, total annual expenditures to care for respiratory conditions were higher for exposed children than for those from homes without adult smokers ($273 versus $150, p<0.05). There were no statistically significant differences in hospital stays, ambulatory visits, prescription medication expenditures, total healthcare expenditures, and bed days between children living without smokers and children living with adults who smoked only outside the home. Children with missing data on adult smoking status and those in homes with no adult smokers had similar patterns of hospital stays, emergency department visits and bed days, but they were less likely to have any ambulatory visits, any prescription medications expenditures, and any expenditures on any type of service for respiratory conditions (all p<0.01). Adherence to AAP guidelines for well child visits was similar for all four groups of children. These adherence rates are somewhat different from those Selden19 reported, mostly because of the different age group, study years, and smoking measures in the two studies.
Using regression analysis to control for other factors reduces only slightly the magnitude of the differences between children exposed to smoking inside the home and those not living with adult smokers. Regression adjustment reduces but does not remove the statistical significance of some differences. Young children exposed to smoke in the home were more likely by 3.1 percentage points (p<0.05) to have at least one hospital stay, and they were more likely by 4.8 percentage points (p<0.05) to have at least one emergency department stay (table 2). The regression adjusted difference in total annual expenditures is $117 (p<0.05). These results are similar to the unadjusted differences, suggesting that at least for respiratory conditions, exposure to second hand smoking, not other factors, explains excess healthcare expenditures on respiratory care for young children. The probability of having a bed day was 8.4 percentage points (p<0.05) higher for young children exposed to smoking inside the home. There are no statistically significant effects of living with adult smokers who smoke outside the home or having missing smoking status. This suggests that the differences in observed characteristics between those with no adult smokers and those with missing smoking data account for the variation in outcomes between these two groups.
Our models share a large set of covariates to control for potential confounding factors. Consequently, although most coefficients are of the expected signs, they are often not statistically significant. Some of the significant findings are children with a usual source of care are more likely to incur prescription expenditures and total healthcare expenditures, and children aged 0–2 have a greater probability of emergency department visits and bed days. The full regression results are summarised in table A.2 (see Tobacco Control website, http://tobaccocontrol.bmj.com/supplemental).
The lack of any statistically significant effects of indoor smoking on well child visits suggests the effects on respiratory conditions are causal rather than a reflection of unobserved factors associated with indoor smoking. Excluding newborn infants and restricting the estimation sample to children exposed to indoor smoking and children living with no adult smokers produced estimates that are similar in magnitude and did not qualitatively change the results.
Exposure to smoking in the home is associated with large increases in the risk for emergency department visits, hospital stays and bed days for respiratory conditions among children aged 0–4. These results indicate the acute exacerbations associated with smoking, including worse health and reductions in children’s activities. The magnitudes of these effects are large. Relative to having a smoke-free home, smoking inside the home is associated with more than doubling the risk of a child having any emergency department visits and more than tripling of the risk of hospitalisation for respiratory conditions. Based on these increases in risks for exposed children, we extrapolate from these estimates to report the approximate magnitude of young children using services for respiratory conditions associated with smoking. Among young children with at least one emergency department visit for respiratory conditions, roughly 18% (95% confidence interval: 5% to 32%) may have had at least one visit related to smoking inside the home. Similarly, among children with at least one hospital stay for respiratory conditions, more than a third, roughly 36% (95% confidence interval: 7% to 64%) may have had at least one stay related to smoking inside the home. Although these extrapolations lack precision, they none the less suggest the potential magnitude of the aggregate impact of smoking in the home.
We estimate smoking inside the home is associated with $117 additional annual respiratory care expenditures per exposed child aged 0–4. The previous estimate, $120 in 1995 dollars, was based on 1987 data.8 That estimate, inflated to 2004 dollars using the personal healthcare expenditure price index, would be $159 per exposed child. Although we do not know why our estimate is lower, changes in healthcare financing between 1987 and 1999 may account for the difference. Firstly, managed care may have shortened children’s hospital stays, thereby reducing expenditures. Secondly, since the implementation of the State Children’s Health Insurance Program, more children are enrolled in public insurance and fewer have private insurance than in previous years. The growing use of public programmes may have reduced the prices paid for health care.
Extrapolating our expenditure estimates to the population, smoking inside the home adds roughly $415 million (95% confidence interval: 70 to 760) to annual healthcare expenditures for young children. This number needs to be interpreted in context. Firstly, we study only young children. Older children’s exposure to secondhand smoke is harder to measure. Secondly, we study only respiratory conditions. Smoking also causes sudden infant death syndrome and other illnesses. Many severe health consequences of secondhand smoke take years, even decades, to develop. Thirdly, we study only healthcare use and costs, and exclude over the counter medications, transportation costs to obtain care, and parents’ lost wages while caring for their children. Thus, our estimates are clearly only a component of the total costs of children’s exposure to tobacco smoke at home. None the less, even the components studied indicate the extent of the potential harm of children’s exposure to tobacco smoke at home. Our findings suggest a rough lower bound on the magnitude of spending on reducing smoking inside the home that may be cost effective.
Our study has several limitations. Firstly, we rely on reported smoking inside the home. If there is systematic under-reporting of smoking inside the home, our analysis might underestimate the effects of exposure to secondhand smoke at home. Fortunately, studies have validated the accuracy of parental reporting of exposure to tobacco smoke in the home,13 14 and we found no statistically significant effects of smoking outside the home, which is consistent with accurate reporting of smoking indoors. Secondly, even though we control for state cigarette excise tax and smoking regulation at the state level, we do not have data on the degree of exposure outside the home for each child. Unfortunately, large studies which collect biological measures of tobacco smoke exposures typically do not collect information on health service use and expenditures. Thirdly, smoking in the home is potentially endogenous. In particular, adults may reduce their indoor smoking because a child has respiratory problems. Endogeneity of this nature could potentially bias our estimates downwards. Moreover, some unobservable factors could affect both smoking inside the home by parents and level of health service use of their children. We attempt to alleviate this concern by including risk preference and attitudes towards healthcare variables in the model. Our finding that indoor smoking is not associated with less adherence to guidelines for well child visits suggests that the risk preference and attitude variables might have adequately controlled for these unobserved confounding factors.
In conclusion, reducing smoking inside the homes of young children would improve the health of young children. Substantial proportions of hospital stays and emergency department visits by children are related to respiratory conditions. Reducing smoking inside the home would probably reduce the use of these expensive services. Even if parents cannot stop smoking altogether, ceasing to smoke inside the home seems to prevent the harmful effects of secondhand smoking exposure on children.
What this paper adds
Few studies have attempted to measure the burden of secondhand smoke on the health care of the children in the United States. The only nationally representative study of the United States on this issue uses data from the 1980s and maternal smoking as a proxy for secondhand smoking.8
Our study uses more recent US data and employs one of the better measures of exposure to secondhand smoke, self reported smoking inside the home. Exposure to smoking in the home is associated with large increases in the risk for emergency department visits, hospital stays and bed days for respiratory conditions among children aged 0–4. These results indicate the acute exacerbations associated with smoking, including worse health and reductions in children’s activities. We estimate smoking inside the home is associated with $117 additional annual respiratory care expenditures per exposed child aged 0–4.
The authors thank Zhengyi Fang, Social and Scientific Systems, Inc, for excellent research support, and Robert Baskin, Agency for Healthcare Research and Quality, for statistical advice.
Funding: This study was funded by the US Agency for Healthcare Research and Quality. The views expressed in this paper are those of the authors, and no official endorsement by the US Agency for Healthcare Research and Quality or the US Department of Health and Human Services is intended or should be inferred.
Competing interests: None.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.