In August, the Center for Sustainable Health Spending (CSHS) was awarded a grant from the Robert Wood Johnson Foundation to, among other things, examine the relationship between disease prevention and health care costs.  This project heightened my interest in the wonderfully-researched report from the Congressional Budget Office (CBO) entitled Raising the Excise Tax on Cigarettes:  Effects on Health and the Federal Budget, and its excellent summary in the New England Journal of Medicine (NEJM).

The report was years in the making and is noteworthy for its original research and its thorough and insightful literature review.  As the title suggests, its economic focus is on the federal budget.  In some ways this is a very broad perspective as it brings into play smoking’s impact on employment and earnings (hence tax payments), as well as health care costs and Social Security payments.  But in other ways it is quite narrow, being limited to federal revenues and costs.  Before discussing this CBO report, and the complex economics of disease prevention and longevity it underscores, I’d like to create some context.

Smoking And Spending By Medical Condition

In 2009, I and others published estimates of national health spending from 1996 through 2005 by an all-inclusive list of medical conditions. We noted that “… smoking is a risk factor for all of the conditions showing the lowest annual spending growth (coronary heart disease, cerebrovascular disease, lung cancer, COPD, and pneumonia) … this could reflect, in part, the health benefits associated with reduced rates of smoking.”  We subsequently found that the low growth in spending for these smoking-related conditions was associated with reduced prevalence.   Past successes in reducing smoking have paid off in slower spending growth on associated medical conditions.  Has smoking reduction has also slowed the rate of growth in total health care spending?  Maybe, but maybe not.

Longevity And Federal Outlays

The top chart from CBO’s NEJM article shows the impact of smoking reduction on government outlays (spending).  One component of outlays declines over time due to a healthier population with lower per capita health care costs.  However, another component labeled “Effects of greater longevity on outlays” increases over time as smoking-related deaths are averted and more individuals are alive to collect Social Security and consume federally funded health care.  After about twelve years, the longevity effect begins to outweigh the per capita spending effect, and federal outlays are actually increased by the reduced prevalence of smoking brought about by the excise tax!

Too bad that not smoking couldn’t just keep us healthier without prolonging life!  Without this dreaded longevity effect, we could unambiguously claim to be saving money in addition to producing greater health.  Seriously though, the narrow question of whether reduced prevalence of smoking saves federal dollars hinges on the number of extended life years.  Because some policy-makers will evaluate tobacco control programs on whether they save federal dollars, the delayed-mortality effect of reduced smoking is a negative from this perspective.  This perverse result is no knock on the CBO study; it simply answers the question asked and is careful to note that “consequences for the federal budget are only one factor that lawmakers may consider when developing policies to promote health.”

Longevity And Health Spending

Of course, longevity impacts all health spending, not just the federal component.   Imagine an individual with two hypothetical life paths:  a smoker and a nonsmoker.  The two paths show higher health spending for the smoker, concentrated in smoking-related conditions, up to the age when the smoker suffers a premature death.  Let’s call this phase one.  Phase two follows the nonsmoker from this point until his or her eventual death at a later age from causes other than smoking.  The “longevity effect” refers to this second phase in which the nonsmoker incurs health care costs while the deceased smoker costs nothing.

If the nonsmoker survives the smoker by enough years, lifetime health spending could exceed that of the smoker.  Under such a scenario, the shift from smoker to nonsmoker would result in lower spending on smoking-related diseases but higher total spending due to the longevity effect.  While CBO has estimated that federal outlays, which include Social Security, are increased by smoking reduction, the CBO literature review notes that studies are mixed as to the impact of smoking cessation on lifetime health spending.

This relationship between longevity and health care spending has become so embedded in discussions about primary prevention that when advocating for policies to reduce obesity, for example, it is noted that such measures do not prolong life as much as smoking reduction and, therefore, do not suffer from as much longevity-related extra health spending (not to mention social security, if you are tracking federal outlays).

What’s Wrong With This Picture?

Why should primary prevention be put on the defensive over its successful reduction of premature mortality?  What could be wrong with an intervention that increases health and reduces health care costs (and boosts employment and earnings) during the initial life span and, as a bonus, avoids premature death?  Does it really matter if lifetime costs increase?   As Steven Woolf (among others) has noted:  “The question of whether prevention saves money is incorrectly framed.  The proper question … is how much health the investment purchases.”

Before discussing the question as reframed by Woolf, I note that even those who choose to focus on saving federal dollars might want to promote smoking cessation.  Medicine is continually working to extend the lives of those with smoking-related illnesses and, if past is prologue, new treatments will eventually be both effective and expensive.  In the extreme, smokers and nonsmokers might achieve nearly identical life spans but with vastly greater health care costs for smokers, including in the later years, which are largely federally funded.  In this future, the longevity effect gets smaller and, eventually, smoking reduction actually lowers federal outlays.  (Of course, less-favorable scenarios are possible, such as even more expensive care for nonsmokers in their period of increased longevity.  I would like to hear what the experts have to say about the odds of alternative scenarios for medical advancements.)

Cost-Effectiveness To The Rescue:  Longevity Is Good

Consider reframing the question in terms of the amount of health obtained for the dollars spent.  We’ll use the framework in which health is measured in quality-adjusted life-years (QALYs), a weighted sum of years of life with higher weights given to years of better health, and cost-effectiveness is defined as cost-per-QALY.  How does increased longevity fare in this framework?

Table 3-4 in the CBO report shows that the life expectancy of 30-year-olds is 77 for lifetime smokers and 86 for people who never smoked — a difference of nine years.  Using health care costs as the sole criterion, these extra nine years of longevity would be evaluated as causing an undesirable increase in costs.   But what if cost per QALY is the criterion?

Denote the cost of health care during these nine years by C and the number of QALYs by Q. Then, if you had a switch that could turn the longevity effect off and on, turning it on would increase costs by C and QALYs by Q, for a cost per QALY of C/Q.  Is this cost per QALY competitive with the cost effectiveness of other health spending?

An obvious approach would be to gather data on C and Q, quantify the cost per QALY, and compare it to the alternative.  I may attempt that in a subsequent blog but, for now, I will settle for an approach that does not require estimation: I show that, whatever its value, the cost per QALY from the longevity switch is significantly lower than the cost per QALY obtained via the health care provided during the longevity period.  Assuming that this health care spending is deemed worthwhile (in terms of its cost per QALY), then turning the longevity switch “on” is even more worthwhile.  (See Note 1)

The proof rests on the self-evident “fact” that only a portion of the QALYs achieved during the nine years of increased longevity are the result of health care provided during that period.  To illustrate, imagine that health care became unavailable at the exact instant that the longevity period began.  The lack of health care would shorten the period of increased longevity but would certainly not cut it to zero.  (The lack of health care would have to increase the probability of immediate death to 100 percent in order to eliminate the longevity period altogether, and it is self-evident that this would not happen.)

This shortened longevity period, characterized by the absence of any health care, would contain a number of QALYs denoted as Qa.  The remaining QALYs, Q – Qa, are attributable to health care.  The cost per QALY associated with health care during this period is then C/(Q-Qa).  Since the cost per QALY from longevity is C/Q, and Q is greater than (Q-Qa), the cost per QALY from longevity is less than the cost per QALY of the health care consumed during the period of longevity.  As they say in mathematics, QED!

The Cost-Per-QALY Advantage Of Longevity:  Implications For Investments In Smoking Prevention

In this concluding section, I sketch out some initial implications.   Please forgive (or give thanks) for the lack of methodological precision as, in this constrained space, I can only hint at the answers that a more rigorous study would provide.  (In particular, I will not bother with present valuing.)  I’ll begin by roughly quantifying the cost-per-QALY advantage of longevity and then derive implications for investments in smoking prevention.

In the absence of empirical evidence, I polled my colleagues for their intuition on what share of the nine years of increased longevity would be due to health care consumed during the longevity period.  We agreed that health care likely accounts for less than half of the increased longevity but could possibly account for half of the QALYs.  If health care accounts for half of the QALYs, the longevity cost per QALY would be half that of the health care cost per QALY.  (The “cost per QALY of longevity” is shorthand for the cost per QALY of pushing the longevity “on” button as described earlier.)  Thus, while the longevity effect might keep smoking prevention efforts from being cost saving over the lifetime, it produces QALYs at a large discount relative to the health care delivered during the longevity period.

One reason for the longevity cost-per-QALY advantage is that we have assumed zero investment cost in converting the would-be smoker.  (In this case, it is driven by an excise tax on cigarettes which increases revenue for the government.)  Let us now drop that assumption and ask how large an investment would be justified if it were put on an equal cost-effectiveness footing with health care provided during the longevity period.  At a minimum, we would invest an amount equal to what is spent on health care during the nine year longevity period, as this would equalize costs per QALY during that period.  This would double the cost of pushing the longevity button, thereby doubling its cost per QALY and bringing it into equality with that of health spending.

In addition, there are nonsmoker health care cost savings prior to the period of increased longevity that amount to more than ten percent of health care costs during that period.  (See figure 3-2 of the CBO report.  Note that the increased health of nonsmokers would also lead to some QALY gains during the period preceding the increased longevity period.)  Thus, this equal-footing rule  would justify spending to convert a would-be smoker of an amount equal to more than ten percent of health care spending up to age 77 plus total health care spending between age 78 and expected death at age 86.

A different way to apply the equal-footing rule would be to shift health care dollars (from the increased longevity period) into investments in smoking prevention until the cost per QALYs were equalized.  In either case, the main takeaway here is that the increase in longevity from not smoking tends to justify greater investments in smoking prevention under the cost-effectiveness framework.  The lessons obviously carry over to similar types of prevention such as reducing the prevalence of obesity, although the specifics, such as the timing and duration of extended longevity period, will differ.

This strikes me as a fertile area for future research and I expect our Center to be enthusiastically involved in the months and years ahead.

Note 1.  The point here is that the cost per QALY of longevity is always less than the cost per QALY of the healthcare consumed.  If the healthcare does not have an acceptable cost per QALY, it should be reduced in scope until the cost per QALY becomes acceptable.  This, in turn, reduces the cost per QALY of longevity to level that is even more acceptable.