The association between smoking during pregnancy and hospital inpatient costs in childhood

Abstract

Although the health sequelae of smoking during pregnancy are well documented, relatively little is known about its long-term economic implications. The objective of this study was to analyse individual-level data on maternal smoking behaviour and sociodemographic, perinatal and resource utilisation variables in order to estimate the association between smoking during pregnancy and hospital inpatient service utilisation and costs through the first 5 years of the infant's life. Data from the Oxford Record Linkage Study, a collection of birth registrations, death certificates and statistical abstracts of hospital inpatient and day case admissions formed the basis of the investigation. The study population comprised all infants born to women who both lived and delivered in Oxfordshire or West Berkshire during the period 1 January 1980–31 December 1989 ($n=119,028$). The cost of each hospital admission, including the initial birth admission, was estimated by multiplying the length of stay by the per diem cost of the respective specialty (£ 1998–1999 sterling). The effect of maternal smoking behaviour on cumulative 5-year hospital inpatient service utilisation and costs was analysed in a series of multivariate analyses, taking account of confounding clinical and sociodemographic factors. Infants born to women who reported smoking during pregnancy were hospitalised for a significantly greater number of days than infants born to women who had either never smoked or had smoked in the past ($P<0.0001$). Over the first 5 years of life, the adjusted mean cost difference was estimated at £462 (95%CI: £353–£571) when infants born to women who smoked at least 20 cigarettes per day were compared to infants of non-smoking mothers, and £307 (95%CI: £221–£394) when infants born to women who smoked 10–19 cigarettes per day were compared to infants of non-smoking mothers ($P<0.0001$). The results of this study should add an economic dimension to the importance of providing smoking cessation services for pregnant women.

Introduction

Prevalence studies conducted in recent years suggest that between one in seven and one in three pregnant women in developed countries smoke cigarettes at some stage during their pregnancy (Dodds, 1995; Stewart et al., 1995; Cnattingius & Haglund, 1997; Ford, Tappin, Schluter, & Wild, 1997; Tappin, Ford, & Schluter, 1997; Ebrahim, Floyd, Merritt, Decoufle, & Holtzman, 2000; Jaakkola, Jaakkola, Gissler, & Jaakkola, 2001; Schluter, Ford, & Ford, 2002). The risk of smoking during pregnancy is elevated among women characterised by social disadvantage, high parity, being without a partner, low income and experience of depression, stress at work and low levels of practical support (Graham, 1994; Borelli, Bock, King, Pinto, & Marcus, 1996; Dejin-Karlsson et al., 1996; Graham, 1996; Wergeland, Strand, & Bjerkedal, 1996).

Although recent evidence suggests that the prevalence of smoking during pregnancy has been declining in some developed countries (Ebrahim et al., 2000; Cnattingius & Lambe, 2002; Colman & Joyce, 2003), it remains a significant public health concern. It is well established that cigarette smoking by pregnant women is associated with an increased risk of a range of adverse perinatal outcomes. In particular, it has been associated with an increased risk of spontaneous abortion, placental complications, preterm premature rupture of the membrane and ectopic pregnancy (DiFranza & Lew, 1995; Castles, Adams, Melvin, & Kelsch, 1998). Furthermore, smoking during pregnancy is known to restrict intrauterine growth (Kramer, 1987), decrease birthweight of the infant (Meyer, Jonas, & Tonasicia, 1976) and increase the odds of preterm delivery (Shiono, Klebanoff, & Rhoads, 1986). In later life, smoking during pregnancy is significantly associated with respiratory illness (Fergusson & Horwood, 1985), conduct disorder (Wakschlag et al., 1997), attention problems (Olds, 1997) and attention deficit hyperactivity disorder (Milberger, Biederman, Faraone, Chen, & Jones, 1996) among the offspring.

An understanding of the economic consequences of smoking during pregnancy could provide a framework for identifying priorities for research and development; inform future economic evaluations of smoking cessation programmes; and act as an invaluable resource to clinical decision-makers and budgetary and service planners. A recent review of the economic literature in this area highlighted the significant economic consequences of maternal smoking during pregnancy that are mediated through adverse infant outcomes (Adams & Young, 1999). Studies by Oster, Delea, and Colditz (1988) and Li, Windsor, Lowe, and Goldenberg (1992) estimated the United States-wide smoking-attributable costs related to low birthweight at $267 million (1983 prices) and $52–150.5 million (1987 prices), respectively. The difference between these cost estimates can largely be explained by the discrepant data sets analysed by the investigators. Oster et al. (1988) applied estimates of the costs of neonatal intensive care calculated by Phibbs, Williams, and Phibbs (1981) at two San Francisco hospitals to secondary evidence on the probability of admission to a neonatal intensive care unit (Korenbrot, 1984). Li et al. (1992), on the other hand, applied estimates of the cost of the initial hospitalisation, readmissions during the first year of life and discounted long-term costs of disability that result from low birthweight, which had been calculated by the United States Office of Technology Assessment (OTA, 1987), to the same secondary data set (Korenbrot, 1984). A study by Marks, Koplan, Hogue, and Dalmat (1990) applied similar methodology to Li et al. (1992), but generated a rather higher smoking-attributable cost related to low birthweight ($1.03 billion, 1986 prices), largely as a consequence of a higher assumed admission rate to a neonatal intensive care unit. Two more recent studies estimated the smoking-attributable costs related to low birthweight in the United States at $263 million (1995 prices) for the nation as a whole (Lightwood, Phibbs, & Glantz, 1999) and $914 (1996 prices) for each smoking pregnant woman (Miller, Villa, Hogue, & Sivapathasundaram, 2001).

A common methodological feature of these studies is that separate sources of evidence are used to estimate relative risks for adverse infant outcomes and the cost of each adverse outcome. Adams and Young (1999) have noted that this precludes the application of multivariate analysis to account for the complex interaction of sociodemographic factors that can explain health care costs incurred by women who smoke during their pregnancy and their offspring. Furthermore, this attributable risk approach cannot account for potential differences in the duration, complexity and cost of medical conditions for smokers and non-smokers.

Two studies have attempted to overcome some of these methodological limitations by analysing medical data sets with much of the relevant sociodemographic and perinatal information using multivariate statistical techniques. Adams, Solanki, and Miller (1997) applied multivariate analysis to National Medical Expenditure Survey (NMES) data from the United States and estimated national smoking-attributable direct medical costs during pregnancy and through delivery for mothers and infants at $1.4 billion (1995 prices). More recently, Adams et al. (2002) applied multivariate analysis to Pregnancy Risk Assessment Monitoring System (PRAMS) data from 13 states in the United States and estimated that smoking during pregnancy increased national neonatal costs by $367 million (1996 prices). Although these analyses of individual-level data represent an improvement over earlier work, the studies were themselves subject to a number of methodological limitations. Most notably, in both studies, a crucial component of the cost of neonatal care, namely the duration of the infant's stay in the neonatal unit, was drawn from a secondary source rather than the individual-level data set. In addition, the longer-term economic consequences of maternal smoking that are felt throughout the infant's first year of life and beyond were not taken into account.

The study reported in this paper set out to overcome the limitations of earlier studies by conducting a comprehensive assessment of the long-term economic consequences of maternal smoking during pregnancy. In line with the studies by Adams et al. (1997) and Adams et al. (2002), use was made of a large medical data set with individual level sociodemographic and perinatal variables. The unique feature of this study, however, is that the primary data set also incorporated individual level resource utilisation data throughout infancy and early and mid-childhood. The specific hypothesis tested was that maternal smoking during pregnancy is associated with significantly increased hospital inpatient service utilisation and costs through the first 5 years of the infant's life.