Article Text

Increasing prevalence of smoke-free homes and decreasing rates of sudden infant death syndrome in the United States: an ecological association study
  1. Ilan Behm1,
  2. Zubair Kabir2,
  3. Gregory N Connolly1,
  4. Hillel R Alpert1
  1. 1Center for Global Tobacco Control, Department of Society, Human Development, and Health, Harvard School of Public Health, Boston, MA USA
  2. 2Research Institute for a Tobacco Free Society, Dublin, Republic of Ireland
  1. Correspondence to Ilan Behm, Harvard School of Public Health, Center for Global Tobacco Control, 677 Huntington Avenue, The Landmark Center, 3rd Floor East, Boston, MA 02115, USA; ibehm{at}


Purpose This study utilises an ecological design to analyse the relation between concurrent temporal trends in sudden infant death syndrome (SIDS) rates and prevalence of smoke-free households with infants in the USA, controlling for an important risk factor, infant supine sleep position.

Methods Annual state-specific SIDS cases were computed using period linked birth/infant death files; the prevalence of 100% smoke-free homes with infants using Tobacco Use Supplement to the Current Population Survey data, and percentage of infants in supine sleep position from National Infant Sleep Position data, for years 1995–2006. Incidence rate ratios relating trends in SIDS cases and risk factors were determined using time-series negative binomial regression. Population-level health effects were assessed with secondhand smoke (SHS) exposure population attributable fractions and excess attributable SIDS deaths.

Results For every 1% absolute increase in the prevalence of smoke-free homes with infants, SIDS rates decreased 0.4% from 1995 to 2006, controlling for supine sleep position. Nationally, it is possible that 20% of the 1326 total SIDS cases were attributable to childhood SHS exposure at home in 2006 with potentially 534 fewer infant deaths attributable to SHS exposure in 2006 than in 1995, owing to an increasing prevalence of 100% smoke-free homes with infants. Cumulatively, 4402 (lower 95% CI) to 6406 (upper 95% CI) excess SIDS cases may have been attributable to SHS exposure in the home over the 12-year study period.

Conclusions The uptake of voluntary restrictions on smoking inside the home may present a public health benefit for infants in their first year of life. In light of inherent ecological study design limitations, these results warrant further individual level research linking postnatal SHS exposure and SIDS.

  • Epidemiology
  • sudden infant death syndrome
  • ecological study
  • statistics
  • infant sleep position
  • environmental tobacco smoke
  • prevalence
  • smoking caused disease
View Full Text

Statistics from


Sudden infant death syndrome (SIDS) is a diagnosis of exclusion defined as the death of an infant less than 1 year of age that cannot be explained following autopsy, examination at death scene and review of clinical history.1–4 Between 1995 and 2005, SIDS rates in the USA fell from 87.2 deaths to 54.0 deaths per 100 000 live births.5 6 Despite the declining trend, SIDS remains the leading cause of death among infants in the post-neonatal period (1–12 months).2 3 6–9

The underlying cause for SIDS has yet to be identified1–3; however, several prenatal and postnatal modifiable factors have been associated with SIDS, including postnatal exposure to secondhand smoke (SHS).2 3 8 In 2006 the US Surgeon General concluded that there exists a causal relation between exposure to SHS and SIDS.2 A more recent meta-analysis by the Royal College of Physicians in the UK (RCP–UK) has further strengthened this causal relation.10

Infant sleep environment, including infant bed-sharing, over-heating, sleeping on soft surfaces and prone or side sleep position, has also been established as a modifiable postnatal factor contributing to increased likelihood of SIDS.2 3 7 9 11 Additionally, other independent SIDS risk factors include maternal use of drugs during pregnancy12 as well as lack of pacifier (dummy) use.13 14 The American Academy of Pediatrics particularly stressed the importance of supine sleep position in its 1992 guidelines followed by its ‘Back to Sleep Campaign’, a public awareness campaign launched in 1994 aimed at increasing supine sleep position among infants.7 9 Published data show that the percentage of infants sleeping in the supine position has increased from 12.4% in 1992 to 75.1% in 2005.11 15 Research suggests that declines in SIDS rates have in part been attributed to improved infant sleep position practices.2 3 7 9

Concurrent with the improving trend in supine sleep position, improvements have also been made since the early 1990s in reducing childhood exposure to SHS.2 Childhood exposure to SHS is involuntary, ‘an unavoidable consequence of breathing in a smoke-filled environment’ and especially for children in their first year of life, the home is the primary source of exposure.2 16 The national prevalence of all households with smoke-free rules increased from 43.2% in 199216 to 79.1% in 2007,17 with increases seen in every state over that time period.16 Temporal trends in state-specific prevalence of smoke-free homes with infants have yet to be reported.

In this study, an ecological study design was utilised with multiple data sources to assess the population-level effects of postnatal SHS exposure in the home on SIDS rates in the USA using state-specific measures of smoke-free homes with infants, infant supine sleep position and SIDS cases for years 1995–2006.


Sudden infant death syndrome and live births

Live births and cases of SIDS were computed for years 1995 through 2006 using Period Linked Birth/Infant Death Files from the National Vital Statistics System (NVSS),18 which links birth and death certificates in all 50 states and allows for the analysis of infant deaths in relation to risk factors present at birth.19 Over 90% of infant deaths annually are linked to birth certificates data. For example, all but five states in 2005 had linkage rates ≥97.5%, with the majority having 100% and no state having less than 94%.6 19

The present study included birth-linked SIDS cases with the following birth selection criteria: mother's age between 19 and 35 years, a gestational period of ≥37 weeks and a birth weight ≥2500 g. These criteria were selected based on research showing values outside of these ranges to be possible independent risk factors for increased infant mortality.1 6 Annual state-specific live births were computed as the total number of births meeting the selection criteria.

State-specific SIDS cases were computed as any death under the age of 1 year with an International Classification of Diseases Clinical Modification (ICD-CM) code of 798.0 (ICD-9-CM) during years 1995 through 1998 and R-95 (ICD-10-CM) during years 1999 through 2006.20 21 ICD-9-CM and ICD-10-CM SIDS codes have a comparability ratio of 1.0362.22 SIDS rates were computed per 1000 live births.

Percentages of smoke-free homes with infants

State-specific percentages of smoke-free homes with a child under 1 year of age were computed from the Tobacco use Supplement to the Current Population Survey (TUS-CPS), a National Cancer Institute (NCI) sponsored monthly survey of civilian, non-institutionalised households in the USA.23–27 Data were analysed from all survey waves in years 1995–1996, 1998–1999, 2001–2002, 2003 and 2006–2007.23–27 Percentages of households with at least one infant and a rule disallowing smoking anywhere in the home were calculated for each wave by state and nationally. A smoke-free home with an infant was defined as the response, ‘No one is allowed to smoke anywhere inside your home’ to the question, ‘Which best describes the rules about smoking in your home?’ among homes with a child under 1 year of age. Households with discrepant (eg, one respondent indicated that smoking was not allowed anywhere in the home while another respondent indicated that smoking was allowed in all or in some parts of the home (mean 6%) or all non-responses (mean 3%) to this question were excluded. Estimates of smoke-free homes with infants were obtained accounting for the multistage stratified sample design of TUS-CPS and adjusting final weights with replicate weights provided by NCI with a balanced repeated replication and Fay's adjustment factor of 0.5.28 Smoke-free home percentages were linearly interpolated for years in which the TUS-CPS was not conducted (1997, 2000, 2004 and 2005). Given the increasing linear trend in smoke-free home prevalence, the interpolation is not hypothesised to alter the result findings.

Infant sleep position

The percentage of infants in supine sleep position was calculated from National Infant Sleep Position (NISP) data, an annual telephone survey of night-time caregivers of infants less than 8 months of age in the 48 contiguous states developed by the National Institute of Child Health and Human Development (NICHD).11 NISP data are intended to assess infant care practices with respect to the American Academy of Pediatrics' recommendation regarding infant sleep position9 and were analysed at the state and national level. State stratification of the percentage of infants in supine sleep position with cell sizes of ≥20 counts were taken to be reliable state estimates (n=25 states). The percentage of infants in supine sleep position was calculated with and without imputation for states that did not meet this reliability criterion, as follows: (A) including all states by imputation of national supine sleep position percentages for all states (n=50 states); (B) including only selected states with reliable state estimates of supine sleep position (n=25 states); (C) including all states using reliable state estimates (n=25 states) where appropriate and imputation of national percentages for states lacking reliable estimates (n=25 states).

Statistical analysis—time series negative binomial regression

SIDS rates were regressed on smoke-free home with infant prevalence and supine sleep position using cross-sectional time-series negative binomial regression analysis using Stata SE 10.0.29 Observations were processed as vectors of the form xit30 with i=state and t=year controlling for the vector's number of live births. The vector format allowed for two types of comparisons simultaneously: change over time within groups and differences among groups.31 Negative binomial regression exponentiated coefficients (eb) have the interpretation of incidence-rate ratios (IRRs).30 A negative binomial distribution rather than Poisson was used because of over-dispersion of the count data.

To test the significance of the independent temporal trends in SIDS rates, smoke-free homes with infants and supine sleep position, join point regression analyses were performed using NCI's Joinpoint 3.4.3 software.32 Annual average percentage changes (AAPC) were computed on all observations to summarise the temporal trends.33

The final model selected to best estimate the relation between outcome and predictors utilised the (C) calculated supine sleep position variable, which accounted for state variation in infant sleep position and allowed for inclusion of all states. A sensitivity analysis was undertaken on supine sleep position (C), re-running the regression models twice, substituting in the alternative two computed supine sleep position variables (A) and (B), respectively, in each iteration. Results of this analysis can be found in appendix A (online only).

Population effects of SHS exposure in the home

Estimated population attributable fractions (PAFs) with upper and lower 95% CI estimates were calculated for each xit vector using a pooled OR and CI equal to 3.15 (95% CI 2.58 to 3.85) for risk of SIDS from postnatal exposure to SHS, based on a recent meta-analysis reported by the Royal College of Physicians in the UK.10 The formula for PAF, where Pe is the exposure defined as the fraction of smoke-free homes with infants subtracted from 1, and OR is the pooled OR, is:PAF=Pe(OR1)/(1+Pe[OR1])3436

Upper and lower 95% CI estimates of PAFs were calculated by substituting the pooled OR with the lower and upper limits of the OR's 95% CI (95% CI 2.58 to 3.85). respectively. Excess attributable deaths were summed across states to estimate the cumulative national SIDS mortality as well as annual rates per 100 000 live births, all with lower and upper 95% CI estimates.


National rates of SIDS exhibited an average annual decline of 3.6% and a relative decline of 33% from 0.66 to 0.44 cases per 1000 live births between 1995 and 2006 (figure 1). Median state-specific SIDS rates decreased from 0.79 cases per 1000 live births in 1995 to 0.50 cases per 1000 live births in 2006, a relative decline of 37% over the time period.

Figure 1

Rates of sudden infant death syndrome (SIDS)—United States (1995–2006).

The national prevalence of homes with an infant and no-smoking allowed increased by an average annual 3.1%, from 64.1% in 1995 to 88.3% in 2006, a relative increase of 38% over the time period (figure 2). Median state-specific percentages of smoke-free homes with infants increased from 62.2% in 1995 to 91.4% in 2006, a relative increase of 47% over the same period.

Figure 2

Temporal trends in key risk factors of sudden infant death syndrome—United States (1995–2006).

Regression analyses yielded IRRs for the relation between temporal changes in risk factors and trends in SIDS rates. Unadjusted for other risk factors, for every 1% increase in homes with an infant covered by 100% smoke-free policies, SIDS rates declined by 1.6% (IRR 0.984, 95% CI 0.981 to 0.986) from 1995 through 2006. Adjusting for changes in percentage of infants sleeping in the supine sleep position (IRR 0.993, 95% CI 0.991 to 0.995), for every 1% increase in homes with an infant covered by 100% smoke-free home policies, SIDS rates declined by 0.4% (IRR 0.996, 95% CI 0.993 to 0.999) over the same period.

Assuming an OR=3.15, in 1995, 43.5% of the 1842 total SIDS cases in the USA may have been attributable to SHS exposure in the home (figure 3), resulting in 801 excess deaths with a lower 95% CI estimate of 36.2% (667) and upper 95% CI estimate of 50.6% (932) of SIDS cases attributable to SHS exposure in the home. In 2006, the estimated national PAF attributable to infantile smoke exposure in the home decreased to 20.1% of the total 1326 SIDS cases (figure 3), resulting in a possible 267 excess deaths, 534 fewer SIDS cases than in 1995. In 2006 the lower and upper PAF 95% CI estimates were 15.3% (203) and 25.0% (332), respectively.

Figure 3

Temporal trend in sudden infant death syndrome population attributable fraction from secondhand smoke exposure in the home—United States (1995–2006).


To our knowledge, this is the first study to examine the association between changing infantile SHS exposure levels at home with the occurrence of SIDS in the USA while controlling for changing patterns in infant sleep position. The results indicate that for every 1% absolute increase in homes with an infant covered by complete smoke-free policies, SIDS rates may have declined by 0.4% (IRR 0.996; p<0.05).

Furthermore, an estimated 20% of SIDS cases may have been attributable to childhood SHS exposure at home in 2006, a finding that is consistent with recent studies.34–36 State-adjusted PAF estimates show that a cumulative 5403 excess SIDS cases in the USA were attributable to childhood SHS exposure at home over the 12-year period resulting in a possible 534 fewer SHS-attributable SIDS cases in 2006 than in 1995.

The steady decrease observed in PAF for SHS exposure in the home from 1995 through 2006 (figure 3) represents a public health benefit associated with the uptake of voluntary no-smoking policies in homes with infants. Similarly, the increasing trend in supine sleep position practices (figure 2) is also hypothesised to result in a decreasing contribution of infant sleeping position to SIDS cases over the same period. The improvement in these two important risk factors suggests that by 2006, other SIDS risk factors accounted for a larger proportion of SIDS cases than either infant sleep position or smoke in homes with infants accounted for. SIDS prevention efforts tend to focus on infant sleep position and the infant sleep environment in general. While continuation of these efforts should be encouraged, these results suggest that public health messaging urging smoke-free homes among homes with an infant may serve as an additional important prevention measure.

The estimated PAFs and excess attributable deaths quantify the potential public health burden of SIDS associated with infantile exposure to SHS in the household. The sizeable lower 95% CI estimate underscores the significance of SHS exposure as an environmental risk factor for infant health. The assumed pooled OR used associating SHS in the home with SIDS (OR=3.15) is based on postnatal maternal smoking and has been validated as a proxy for all infant exposure to SHS.34 The RCP-UK meta-analysis, and resulting pooled OR, was inclusive of studies reviewed by the US Surgeon General in 2006 as well as 75 additional studies including studies from the USA; as such, it was appropriate to apply this OR to US SIDS data. Additionally, the RCP-UK reported a pooled OR of 1.45 (95% CI 1.07 to 1.96) associating paternal and others' household smoking with SIDS.10 The true OR associating infants' risk of SIDS from SHS exposure might be higher (ie, because of possible combined effect of maternal with paternal and other household smoking) or possibly lower (ie, absence of maternal smoking in home in presence of paternal and others' smoking) than 3.15, dependent on the number of sources and amount of exposure in the home.

Study strengths and limitations

The ecological study design alone cannot demonstrate causality, but it is suitable for examining the association of temporal trends in multiple risk factors at the population level concurrent with trends in infant mortality.31 It may also be prone to ecological fallacy where the ecological level associations may not accurately reflect individual level associations.37

Additionally, at the ecological level, consideration must also be given to group confounding. Within group confounding may have been partially offset in the present study by examination of a subset of infants (ie, term normal birthweight babies born to mothers ages 19 and 35); nonetheless, the inherent limitations of this study design must be considered when interpreting results and warrant further research relating the two studied important risk factors with SIDS at the individual level.

A strength of the study was the calculation of national estimates adjusting for state-level variation within both outcome and explanatory variables. Controlling for differences among states adds validity to the ecological effect estimates in this study31 and serves to control any state-level confounding in risk factors and outcomes. The explanatory variables of smoke-free homes with infants and supine sleep position are both derived from self-reported data and are open to self-report bias. However, at least for SHS exposure, self-reported data may provide a more sensitive measure over time when compared to biomarkers of exposure.38

Nonetheless, the inclusion of infant sleep position is essential given the importance of this risk factor to SIDS and major efforts to improve infant sleep environment.6 9 15 16 18 36 39 The present finding showing an association between infant sleep position and SIDS (IRR=0.993) is consistent with previous research.2 7 9 The method of imputation for the infant sleep position variable was chosen to allow for inclusion of all states into the regression model. To ensure that this method was valid, a sensitivity analysis was undertaken based on alternative computation of the infant sleep position variable. Results of the sensitivity analysis can be found online in appendix A. These analyses of alternative samples based on completeness of the infant sleep position data further strengthen the stability of the present findings and justify the method of imputation used. Although the NISP data only cover infants through 8 months of age, research suggests that infants are at most risk for SIDS at 2–4 months of age with the majority of SIDS cases occurring in the first 6 months of life.39

A possible limitation of this study is outcome misclassification which could arise from the nature of SIDS as a diagnosis of exclusion.2–4 40–42 The quality of the postmortem investigation may vary by state and over time,43 44 thereby increasing state variation in a non-systematic manner. In addition, various analogous terminologies, such as cot death, crib death and sudden unexplained infant death (SUID) have been shown to be recorded on death certificates of SIDS deaths45 possibly introducing further misclassification; however a recent study has shown the majority of ICD-coded SIDS cases to be classified correctly.45 Nonetheless, the unknown magnitude and frequency of any possible outcome misclassification could serve to bias results towards the null hypothesis.

Another possible limitation of this study is the inability to control for maternal smoking during pregnancy. Existing reliable sources for estimates of maternal tobacco use during pregnancy (NVSS birth certificate data and Centers for Disease Control and Prevention Pregnancy Risk Assessment Monitoring System data) were found to lack sufficient annual state-level data over the study period. Additionally, an interaction may exist between maternal smoking during pregnancy and the prevalence of homes with smoke-free policies as pregnant women smokers residing in homes with smoke-free policies may be more likely to abstain from smoking with a new infant in the home or possibly be more inclined to quit smoking.40–42 Regardless of smoking status, the exposure of an expectant mother to SHS in the home by other household members may also endanger the health and wellbeing of the fetus, possibly resulting in SHS effect modification.2 Future studies should incorporate measures of prenatal smoke exposure, as well as maternal exposure during pregnancy to SHS in all settings.

It is also important to note other established independent risk factors for SIDS. These include but are not limited to maternal use of drugs during pregnancy as well as lack of infant pacifier use (OR=2.61)14 and infant sleeping on soft bedding (OR=5.1).46 While incorporation of these risk factors was beyond the scope of this study, their exclusion may result in residual confounding in the present results. The unknown magnitude and impact of such confounding is a limitation of this study.

Future research might explore the possibility of confounding by gender-specific and/or race-specific correlates as well as further sleep environment factors such as bed-sharing and over-heating, or other SIDS risk factors.

In summary, the study's findings of a relation between potential SHS exposure of infants and SIDS are generalisable to the US population of infants born to young mothers (19 and 35 years of age), with term normal birthweight babies. The reported population-level IRRs suggest the need for further individual-level research associating postnatal SHS exposure with SIDS rates.


The findings of this ecological study indicate a statistical association between the increasing prevalence of smoke-free home rules among homes with infants and decreasing cases of SIDS in the USA from 1995 through 2006, while controlling for the key risk factor of infant sleep position. Adjusted PAF estimates based on pooled OR from meta-analysis relating postnatal SHS exposure and SIDS potentially suggest that 534 fewer SIDS cases than expected were observed in 2006 relative to 1995. Under those assumptions, an overall 5403 SIDS cases in the USA may have been prevented had 100% smoke-free home policies among homes with an infant been in place from 1995 onwards. The results suggest that increased voluntary smoke-free home policies among homes with at least one infant may have an important role in reducing the rates of SIDS, the largest cause of infant mortality in the post-neonatal period. Inherent ecological study design limitations and the present results warrant further individual level research linking postnatal SHS exposure and SIDS.

What this paper adds

  • In 2006, the US Surgeon General concluded that postnatal exposure to secondhand smoke (SHS) is causally related to sudden infant death syndrome (SIDS); however, infant sleep position is also an important known risk-factor. Concurrent improvements in both childhood SHS exposure and infant sleep position have been observed since the mid-1990s. It is unclear how these two important risk factors as well as state variation interplay and relate to the improvement observed in national SIDS rates.

  • This ecological study suggests that improvements in smoke-free homes may have resulted in a lessening of the proportion of SHS attributable SIDS cases from 1995 to 2006. The uptake of smoke-free homes among homes with an infant represents a public health benefit. While much SIDS prevention efforts have rightly focused on the infant sleep environment, these results suggest that a smoke-free environment for infants should also be promoted in an effort to continue reducing SIDS rates in America.


We acknowledge the support of Dr Michael Corwin in obtaining the NISP data.


View Abstract

Supplementary materials

  • Web Only Data

    Files in this Data Supplement:


  • Funding Research support for this study is provided by the Flight Attendant Medical Research Institute, Clinical Innovator Award 072085.

  • Competing interests None.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

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.