In his book The Health Gap, Michael Marmot (Reference Marmot2015), who conducted the classic Whitehall study of social determinants of health, asks: “Why treat people and send them back to the conditions that made them sick?” (p. 1). Thus, a fundamental challenge for public health is to understand the socioecological contributors to ill health, perhaps most importantly the factors that lead individuals to make decisions detrimental to their well-being. Building on the insights of evolutionary life history theory (Hill Reference Hill1993; Stearns Reference Stearns1992), Pepper & Nettle (P&N) outline the case for high extrinsic mortality risk (EMR) as a key driver of seemingly “bad behaviour.” The argument they present is both well reasoned and intuitive, and I find little to disagree with. However, in this comment focusing on health behaviour, I make two points: (1) The evidence for a strong association between EMR and detrimental health behaviour is not as substantive as the target article conveys; and (2) there are several areas where observed health behaviour suggests that EMR plays a relatively inconsequential role. It is only by interrogating these areas of apparent weak support that we can refine our understanding and better integrate evolutionary thinking with existing public health research and policy.
Many of the studies cited by P&N to support the idea that EMR predicts poor health behaviour rely on suboptimal methodologies, including crude aggregates of mortality rates and a failure to isolate the effect of individual socioeconomic status (SES) from EMR. A combination of more fine-grained measures of local mortality rates that distinguish between causes of death (extrinsic and intrinsic) and individual-level characteristics is needed to satisfactorily address the question at hand. Studies that fulfil these criteria have found that the effect of the EMR on reproductive timing and health behaviours is small in comparison to the effect of individual SES, and that area crime rates or the adult sex ratio may be just as important (Uggla & Mace Reference Uggla and Mace2015; Reference Uggla and Mace2016). In the same developed population, evidence suggested that EMR positively predicted preventable death among men but not among women, and that EMR effects were greater for men with low SES than men with high SES (Uggla & Mace Reference Uggla and Mace2015). Such heterogeneity does not refute a model with EMR as a key driver of behaviour, but it does suggest that its predictive power may vary considerably within populations.
Another source of evidence P&N use is priming studies. Studies relying on self-rated mortality risk can be informative, but they are prone to response biases, and the ecological validity of such studies is questionable. It is therefore important to test whether individuals' reported risks map onto the ecological conditions to which they are exposed. A recent study that compared perceived and actual area characteristics in Belfast, Northern Ireland, found that whether individuals accurately gauged their neighbourhood varied with the type of characteristic; perceptions of median age at death were more accurate than perceptions of local levels of crime and the local adult sex ratio (Gilbert et al. Reference Gilbert, Uggla and Mace2016). P&N do acknowledge that individuals might respond to other extrinsic factors, such as illnesses, but it is only by testing these factors alongside the EMR and comparing their relative effects that we can achieve a fuller understanding of the root causes of ill health.
The observation that many health interventions achieve behavioural change without altering the individual's extrinsic risks also raises questions about how central extrinsic risks are for health behaviour. Meta-reviews of health interventions targeting, for example, weight loss and physical activity, have presented convincing results even when interventions are minor, such as keeping a food diary or practising mindfulness (Burke et al. Reference Burke, Wang and Sevick2011; Katterman et al. Reference Katterman, Kleinman, Hood, Nackers and Corsica2014). If individuals with low SES are more likely to have poor diets and be overweight because of higher extrinsic risks, why would they change their behaviours in response to interventions that leave their overall conditions unchanged? In many instances, these interventions do not provide any additional information on the health behaviour, which suggests that their success is likely explained by factors other than educational components. Notably, even in developing contexts, where life may be short and uncertain, studies repeatedly show that when women receive cash, they often spend it on causes that benefit their families' well-being in the long run rather than on short-term expenditures (Banerjee & Duflo Reference Banerjee and Duflo2012). These examples are not mutually exclusive to a model invoking extrinsic risks, but they do suggest that further thinking is necessary to explain the malleability of health behaviours.
Finally, a sticky point is why some health behaviours with well-known health risks, such as smoking, show a clear SES gradient, whereas alcohol misuse – which in many ways is a comparable behaviour – does not. Excessive alcohol consumption is a leading cause of premature mortality in developed countries (World Health Organization 2014), but its relationship with SES is a little more complex than P&N depict. Low-SES individuals have higher risk of alcohol-related death, yet evidence suggests that drinking patterns are similar across SES groups with regard to overall consumption and to binge drinking (Mäkelä & Paljärvi Reference Mäkelä and Paljärvi2008). Do high-SES individuals drink excessive amounts of alcohol because they know it is unlikely to end badly due to better social support networks and higher compliance to treatment? Or is it because the health risks of alcohol are not as severe for high-SES individuals due to their otherwise healthy habits (e.g., better diets)? A greater emphasis on compensation effects and the interplay between different health behaviours might provide fruitful insights on this topic. The association between low SES and poor health behaviours is strong, but paying greater attention to behaviours that do not neatly fit the pattern can help refine theory and offer a better understanding of health inequalities and their causes.
In his book The Health Gap, Michael Marmot (Reference Marmot2015), who conducted the classic Whitehall study of social determinants of health, asks: “Why treat people and send them back to the conditions that made them sick?” (p. 1). Thus, a fundamental challenge for public health is to understand the socioecological contributors to ill health, perhaps most importantly the factors that lead individuals to make decisions detrimental to their well-being. Building on the insights of evolutionary life history theory (Hill Reference Hill1993; Stearns Reference Stearns1992), Pepper & Nettle (P&N) outline the case for high extrinsic mortality risk (EMR) as a key driver of seemingly “bad behaviour.” The argument they present is both well reasoned and intuitive, and I find little to disagree with. However, in this comment focusing on health behaviour, I make two points: (1) The evidence for a strong association between EMR and detrimental health behaviour is not as substantive as the target article conveys; and (2) there are several areas where observed health behaviour suggests that EMR plays a relatively inconsequential role. It is only by interrogating these areas of apparent weak support that we can refine our understanding and better integrate evolutionary thinking with existing public health research and policy.
Many of the studies cited by P&N to support the idea that EMR predicts poor health behaviour rely on suboptimal methodologies, including crude aggregates of mortality rates and a failure to isolate the effect of individual socioeconomic status (SES) from EMR. A combination of more fine-grained measures of local mortality rates that distinguish between causes of death (extrinsic and intrinsic) and individual-level characteristics is needed to satisfactorily address the question at hand. Studies that fulfil these criteria have found that the effect of the EMR on reproductive timing and health behaviours is small in comparison to the effect of individual SES, and that area crime rates or the adult sex ratio may be just as important (Uggla & Mace Reference Uggla and Mace2015; Reference Uggla and Mace2016). In the same developed population, evidence suggested that EMR positively predicted preventable death among men but not among women, and that EMR effects were greater for men with low SES than men with high SES (Uggla & Mace Reference Uggla and Mace2015). Such heterogeneity does not refute a model with EMR as a key driver of behaviour, but it does suggest that its predictive power may vary considerably within populations.
Another source of evidence P&N use is priming studies. Studies relying on self-rated mortality risk can be informative, but they are prone to response biases, and the ecological validity of such studies is questionable. It is therefore important to test whether individuals' reported risks map onto the ecological conditions to which they are exposed. A recent study that compared perceived and actual area characteristics in Belfast, Northern Ireland, found that whether individuals accurately gauged their neighbourhood varied with the type of characteristic; perceptions of median age at death were more accurate than perceptions of local levels of crime and the local adult sex ratio (Gilbert et al. Reference Gilbert, Uggla and Mace2016). P&N do acknowledge that individuals might respond to other extrinsic factors, such as illnesses, but it is only by testing these factors alongside the EMR and comparing their relative effects that we can achieve a fuller understanding of the root causes of ill health.
The observation that many health interventions achieve behavioural change without altering the individual's extrinsic risks also raises questions about how central extrinsic risks are for health behaviour. Meta-reviews of health interventions targeting, for example, weight loss and physical activity, have presented convincing results even when interventions are minor, such as keeping a food diary or practising mindfulness (Burke et al. Reference Burke, Wang and Sevick2011; Katterman et al. Reference Katterman, Kleinman, Hood, Nackers and Corsica2014). If individuals with low SES are more likely to have poor diets and be overweight because of higher extrinsic risks, why would they change their behaviours in response to interventions that leave their overall conditions unchanged? In many instances, these interventions do not provide any additional information on the health behaviour, which suggests that their success is likely explained by factors other than educational components. Notably, even in developing contexts, where life may be short and uncertain, studies repeatedly show that when women receive cash, they often spend it on causes that benefit their families' well-being in the long run rather than on short-term expenditures (Banerjee & Duflo Reference Banerjee and Duflo2012). These examples are not mutually exclusive to a model invoking extrinsic risks, but they do suggest that further thinking is necessary to explain the malleability of health behaviours.
Finally, a sticky point is why some health behaviours with well-known health risks, such as smoking, show a clear SES gradient, whereas alcohol misuse – which in many ways is a comparable behaviour – does not. Excessive alcohol consumption is a leading cause of premature mortality in developed countries (World Health Organization 2014), but its relationship with SES is a little more complex than P&N depict. Low-SES individuals have higher risk of alcohol-related death, yet evidence suggests that drinking patterns are similar across SES groups with regard to overall consumption and to binge drinking (Mäkelä & Paljärvi Reference Mäkelä and Paljärvi2008). Do high-SES individuals drink excessive amounts of alcohol because they know it is unlikely to end badly due to better social support networks and higher compliance to treatment? Or is it because the health risks of alcohol are not as severe for high-SES individuals due to their otherwise healthy habits (e.g., better diets)? A greater emphasis on compensation effects and the interplay between different health behaviours might provide fruitful insights on this topic. The association between low SES and poor health behaviours is strong, but paying greater attention to behaviours that do not neatly fit the pattern can help refine theory and offer a better understanding of health inequalities and their causes.