National medical expenditures by smoking status in American adults: an application of Manning’s two-stage model to nationally representative data

OBJECTIVES: To assess the medical expenditures of American adults by their smoking status—Current, Former or Never smokers. We update these expenditures through 2015 controlling for personal characteristics and medical history and assess the impact of years-since-quitting and decade of life. SETTING AND PARTICIPANTS: Weighted sample of American adults, 2011–2015. The linked National Health Interview Survey (NHIS) and Medical Expenditure Panel Survey (MEPS) are annual weighted representations of approximately 250 million adults. Sampling of NHIS is multistage with data collected throughout the year. PRIMARY OUTCOME MEASURES: Using data from NHIS and MEPS, we collected demographic data, self-reported medical history and current smoking status. Smoking status was designated as Never, Current and Former, along with years-since-quitting. Total medical expenditures were collected from MEPS for 2011–2015. We used Manning’s two-part model to estimate average expenditures per individual and marginal costs for individuals at all levels of smoking status. RESULTS: American adults averaged US$4830 in average medical expenditures. Never smokers (US$4360, 95% CI 4154.3 to 4566.3), had lower expenditures than Current (US$5244, 95% CI 4707.9 to 5580.3) and Former (US$5590, 95% CI 5267.4 to 5913.5) smokers. CI for Current and Former smokers overlapped. Results were similarly significant when controlling for disease history. Years-since-quitting did not affect expenditures. In each decade of adult life, Former smokers had the highest annual medical expenditures, followed by Current and then Never smokers. CONCLUSIONS: We updated annual medical expenditures during the Affordable Care Act era by smoking status using the current best practice model. While we identify Former smokers as having higher medical expenditures than Current smokers, we do not examine how care-seeking behaviour varies between levels of each risk factor.