The role of smoking on mortality compression

Abstract

AbstractBACKGROUNDAdult lifespan variation has been stagnant since the 1960s in most developed countries, despite increases in longevity. However, national averages mask large socioeconomic differences. In Finland lifespan variation among the highest occupational class has continued to decline, while the lower classes have experienced stagnation.OBJECTIVEWe aimed to investigate the role that smoking has played in the occupational social class divergence in lifespan variation since the 1970s.METHODSFinnish register data (1971-2010) by occupational social class, ages 50+, was used. Smoking-attributable mortality was estimated by the Preston, Glei, and Wilmoth (2010) method.RESULTSWe expected smoking-attributable mortality to explain the divergence by occupational class because smoking is socially patterned and particularly important for middle- and young-old-age mortality, ages which contribute greatly to lifespan variation. Instead, we found that among men social class differences would have widened even further without smoking; for women lifespan variation was unaffected by smoking.CONCLUSIONThe maturation of the smoking epidemic will not decrease uncertainty in the timing of death or reduce inequalities in this dimension of mortality by occupational class in Finland.1. Background1.1 IntroductionUncertainty in the timing of death, or lifespan variation, is a fundamental dimension of inequality and has important implications for both individuals and society (Tuljapurkar 2010). A greater uncertainty of surviving to old age reduces the value of public and private investment in training and education. At the individual level, lifespan variation impedes the planning of life's events. Economic models have shown that individuals would be willing to trade half a year in mean lifetime to reduce lifespan variation measured by standard deviation by one year (Edwards 2013). At the societal level, high lifespan variation may signal failing social policies (van Raalte et al. 2011), particularly if caused by high levels of premature mortality (Sen 1998).Traditionally, within-population mortality trends have been monitored by the mean of the age-at-death distribution that is associated with the period life table: that is, the life expectancy. However, given recent uneven age patterns of mortality decline (Himes, Preston, and Condran 1994; Kannisto et al. 1994; Rau et al. 2013; Tuljapurkar, Li, and Boe 2000), the variation of this distribution, what we here refer to as lifespan variation, is becoming a more important dimension to examine. Differences in mortality distributions between high income countries increasingly come from the shapes rather than the locations of the age-at-death distributions. That is, differences in adult lifespan variation, defined as variation in ages-at-death conditional upon surviving childhood, contribute more to the differences between mortality distributions than differences in life expectancies (Edwards and Tuljapurkar 2005). Although a strong inverse relationship used to exist between these two domains, increases in life expectancy since the 1960s have not uniformly been met with compression in the adult lifespan distribution; if anything, stagnation in adult lifespan variation has become the norm in high income countries (Edwards 2011; Edwards and Tuljapurkar 2005; Smits and Monden 2009; Wilmoth and Horiuchi 1999).When mortality across all ages is declining, as has been the typical pattern in most high income countries, adult lifespan variation can decrease only if mortality reductions at younger ages outpace mortality reductions at older ages. Thus it might be expected that the observed stagnation in lifespan variation is owing to harmful personal behavior that has led to high levels of premature mortality. Smoking is the most important behavioral factor influencing mortality at middle- and young-old ages, but its role in contributing to trends in lifespan variation has yet to be investigated. …