Some informative figures are presented in Table 1. Each entry in the ‘other causes’ row is the amount by which a combined alcohol, tobacco and illicit drugs figure falls short of the corresponding all causes figure. Except for person-years of life lost before age 70 years (PYLL), the entries are simply the numbers for deaths, hospital episodes and hospital bed-days that are not attributed to alcohol, tobacco and illicit drugs. Strictly speaking, the ‘other causes’ PYLL entry cannot be interpreted as years lost from other causes because the technical procedure by which PYLL numbers are calculated means that they cannot be added over different risk factors. The ‘other causes’ PYLL entry is simply the difference between the all causes PYLL and the combined total for alcohol, tobacco and illicit drugs. For brevity, we will refer to it as the ‘other causes’ PYLL but it should be interpreted as only a balancing deficit. In terms of percentages, alcohol, tobacco and illicit drugs contribute respectively 2.9, 15.3 and 0.4 percent of deaths; 7.3, 11.6 and 2.4 percent of PYLL; 2.5, 3.4 and 0.2 percent of hospital episodes and 4.4, 4.9 and 0.3 percent of hospital bed-days.

Figures such as those in Table 1 have been seen as supporting the claim that tobacco is the leading addictive substance causing death and premature mortality and morbidity, and is responsible for a large burden on hospital services. But all they tell us is that more deaths, more hospital episodes and bed-days are attributed to tobacco than to alcohol and illicit drugs. It is perhaps a matter of predisposition rather than considered judgement to see the excess 3.4 percent of hospital episodes that are said to originate from the large number of ever-smokers, about 50 percent of the adult population, as a disproportionately large burden on hospital services. Moreover we are not given the perspective of corresponding percentages from other specific causes.

For example in Victoria for the year from July 1990 to June 1991, fractures and accidental falls accounted for about 3.1 percent and 2.5 percent respectively of all hospital episodes in that period. For the same period, reported misadventures during surgical and medical care whilst in hospital accounted for about 1.3 percent of hospital episodes with somewhat larger than average annual patient bed-days⁷. However this counts only iatrogenic illness arising after admission into hospital and does not include non-hospital medical injuries that led to hospitalisation in the first place. Moreover reported iatrogenic injuries induced in hospital do not include unreported cases that might have been so classified by an independent external agency. Perhaps that is why higher rates of iatrogenic injury have been reported from the USA where investigations have been carried out by independent external agencies. The 1977 California Medical Association’s Medical Insurance Feasibility Study estimated medical injuries to occur in about 4.6 percent of cases and the Harvard Medical Practice Study found hospital medical injuries to occur in about 3.7 percent of cases. Both of these rates are comparable to the Australian 1992 percentage of hospital episodes attributed to tobacco. In the same vein it is worth noting that Prescott has argued that higher nursing to patient ratios can decrease patient morbidity and mortality by as much as 5 to 10 percent.

Recognising that causes other than tobacco might impose comparable burdens on hospital services does not exonerate tobacco, but it does help one view the problems posed by it from a wider perspective and hence in a more balanced way. Moreover it raises the possibility that Health Promotion schemes financed by a special tax on tobacco to fund ad hoc health programmes may not be the most advantageous way of using that money. It might for instance yield more immediate and greater benefits to use it to finance higher nursing to patient ratios. Again, while most iatrogenic injuries are not the result of negligence, it may be more sensible, if the American estimates are reliable and applicable here, to improve hospital facilities and provide a less stressful workplace for the medical profession than to fund extensive anti-smoking campaigns and sponsor sporting activities that previously relied on tobacco advertising.

The plain fact is that, by themselves, gross numbers like those in Table 1 serve little purpose other than to provoke shock and horror, and promote unthinking concern about how large they are. To see them in a balanced way we have to see them, not only in the light of other burdens, but also from the perspective of the corresponding numbers of people at risk of contributing to them. For example, while Table 1 states that in 1992 more people died from tobacco than from alcohol, it does not tell us whether there were more or less smokers than drinkers of alcohol; it does not tell us which of alcohol or tobacco is the more likely to kill us; nor indeed does it indicate whether either of them is more or less likely to kill us than are other causes and if so at what sorts of ages. One can give a limited, but nonetheless informative, internal perspective to the figures in Table 1 by dividing each of the PYLL, Hospital Episodes and Hospital Bed-days by the number of deaths in question to give corresponding per death rates. These are presented in Table 2.

Table 2 shows that the greatest years of life-lost per death were due to illicit drugs whereas the least were for tobacco-attributed deaths. Moreover the PYLL per death for tobacco-attributed deaths was almost 1.5 years less than that for all causes and over 14 months less than that for causes other than alcohol, tobacco and illicit drugs. There were almost 4 times more hospital episodes per death due to alcohol than due to tobacco and over 5 times more hospital episodes per death from other causes than from tobacco. Furthermore there were almost 5 times as many bed-days per death spent in hospital due to alcohol than due to tobacco and almost 3.5 times as many hospital bed-days per death from other causes as from tobacco. Thus the figures in Table 2 reveal that the per death loss of life and the per death burden on hospital services that are said to be due to tobacco are in fact smaller than those due to each of alcohol, illicit drugs and other causes. This does not mean that people would live longer or have shorter hospital episodes if they smoked, but it does call into question the extent to which the gross numbers in Table 1 can be said to portray tobacco as a leading cause of mortality and morbidity.

One might be tempted to say that on the contrary Table 2 suggests that if only more people smoked then there would be fewer person-years of life lost before 70 years of age and a much smaller burden on hospital services. But amongst the deaths from other causes are those of children and infants unlikely to die from smoking and these young deaths could contribute substantially to the all cause and ‘other causes’ PYLL, thereby inflating the corresponding per death rates. The effect of such age differences is examined in the next section.

We present ages-at-death patterns for males in Figure 1 and for females inFigure2 by plotting the percentage of deaths in question by age and cause of death. To interpret these figures observe that figure 1 tells us that about 21 percent of male deaths from alcohol occur in the 25-44 years age group whereas that age group accounts for almost 70 percent of the male deaths from illicit drugs, about 2 percent of tobacco-related male deaths and about 6 percent of male deaths from other causes. Inspection of figures 1 and 2 shows that, for both males and females, the age-distribution of deaths from tobacco is similar in shape to that from causes other than tobacco, alcohol and illicit drugs, both of them suggesting steady increases with age, whereas those for alcohol and illicit drugs are markedly different in shape, though for alcohol the disparity is less extreme in females.

Figures 1 and 2 suggest that the deaths attributed to alcohol and to illicit drugs form two separate subgroups of all deaths, both of which are distinguished by an unusual pattern of ages-at-death. In both cases it is the abnormality of that pattern which indicates that something unusual might be going on and suggests one search for underlying reasons, perhaps alcohol abuse in the one case and the abuse of illicit drugs in the other but, at a deeper level, perhaps social factors such as unemployment, a childhood history of sexual or physical abuse and so on.

The disparity between the ages-at-death patterns for tobacco and all causes is not so immediately eye-catching and, on closer analysis, seems to have a number of conflicting interpretations. For although those patterns do have roughly similar shapes, there is an indication that the proportion of tobacco deaths occurring after 74 years of age is smaller than the corresponding proportion of all causes deaths, in fact about 39 percent and 43 percent for males, 44 percent and 63 percent for females. This could be seen as suggesting that tobacco kills an undue proportion of smokers before they can reach 75 years of age, especially female smokers. On the other hand the same sort of logic could be seen as suggesting, on the contrary, that tobacco use provides some protection against other causes of death, at least for males, because about 90 percent of the male tobacco deaths occur after 54 years of age whereas only about 83 percent of both all causes and other causes do so. But for females the corresponding three proportions are all much the same, each of them is about 90 percent, and it could be argued that the alleged protective effect for males arises because other causes account for a much higher proportion of male deaths before 45 years of age than they do for corresponding female deaths, viz. about 11 percent against 3.1 percent for males and only 5.9 percent against 4.4 percent for females. Nevertheless it would be difficult to claim, on those grounds alone, that the alleged effect for males is harmful, even if one did not see it as protective. In the same vein one could note that for males the proportion of deaths at age 65 years or more is, at about 70 percent, much the same for both tobacco and all causes, whereas for females all causes now accounts for the higher proportion, viz. 82 percent as against only 73 percent.

The ambiguous message in these conflicting results calls into question the extent to which the ages-at-death pattern for tobacco can be said to be substantially different from that for all causes. By itself this does not mean that smoking is relatively harmless because it is conceivable that an avoidable cause of death, while causing death at about the same ages as do other harmful but unavoidable causes, does so more frequently. But it does make it problematical that the label ‘tobacco-use’ does in fact identify a definite subgroup of the general population that is characterised by markedly unusual ages-at-death, as might be claimed for the labels ‘alcohol use’ and ‘illicit drug use’. In other words it suggests that there is perhaps something amiss with the common perception that smoking kills at younger ages than is normal and that, as a consequence, the age at death of a smoker is, in general, younger than that of a non-smoker.

To investigate that perception more closely, Table 3 gives the Australian ages-at-death patterns in the 25 to 74 years old age-bracket for tobacco and causes other than tobacco and illicit drugs by sex for the year 1992. We restrict ourselves to those ages to avoid the biasing effect of the higher proportion of deaths at less than 25 years of age that are attributed to other causes, viz. for males 4.9 percent as against only 0.9 percent for those attributed to tobacco and, for females, 2.8 percent as against 1.9 percent. It should be noted, however, that 38.6 percent of the male and 43.5 percent of the female tobacco deaths, and 46.5 percent of the male and 65.6 percent of the female other causes deaths occur after 74 years of age. We also give nominal mean ages at death for both causal categories. These were obtained by supposing that deaths are uniformly spread throughout each age group, so that its average age at death is its mid-point. Multiplying each mid-point by the corresponding percentage of deaths and adding over age-groups gives the nominal mean age at death for the category in question. For the age range in question, this Table shows that of those who allegedly died because of tobacco, the males had lived on average a little over 2 years and six months longer and the females on average about 1 year and 4 months longer than those who had died from causes other than tobacco and illicit drugs. While this casts doubt on the correctness of the perception that smokers die at unusually young ages it does not mean than an individual smoker is likely to live a little longer than a comparable non-smoker, because proportionally more smokers than non-smokers may die in a given year. At most it says only that in 1992, dead smokers in the 25 to 74 years old age-bracket died on average at an older age than did comparable dead non-smokers. It could still be the case that in any given year a live smoker is more likely to become a dead smoker than a similarly aged live non-smoker is to become a dead non-smoker. It needs to be noted too that if one considers only the 45 to 74 year old age-bracket, then one does find that the nominal mean age at death for deaths attributed to smoking is smaller than that due to other causes, by nearly 7.5 months for males and 4.5 months for females. This reinforces the overall impression that smoking deaths do not occur in the main at young ages but from middle age onwards and then at about the same sorts of ages as do deaths from other causes.

This could be seen as confirming our earlier suspicions about Table 2, viz. that the more favourable per death person-years of life lost for tobacco may well reflect only differences between ages at death; tobacco accounting for proportionally fewer young deaths than do causes other than alcohol, tobacco and illicit drugs. However it does so in a way that calls into question procedures that inflate the total burdens imposed by tobacco by including such young deaths. It seems a matter of predisposition to include them, and so obtain a larger estimate of the total burdens tobacco imposes, but to ignore the fact that the corresponding per death rates then put tobacco in a more favourable light than other causes of death. Moreover the fact that to turn the balance against tobacco one has to focus on the 45 to 74 years age-bracket, where the estimated differences in nominal mean age at death are, in any event, at most a matter of a few months, does not support the common perception that tobacco kills at considerably younger ages than do other causes. It might, of course, be objected that the smokers in question, though hardy enough to resist other causes of death that long, would have lived even longer had they not smoked. But such an objection, though possibly true, adds little of substance because it is nothing more than a tautology to note that people would have lived longer than they did had they not died, when they did, from the particular causes that did kill them; this is as true of a death due to a motor vehicle accident as it is of one allegedly due to tobacco.

While deaths allegedly caused by tobacco do not occur at ages that are predominantly very much younger than those of deaths from causes other than alcohol, tobacco and illicit drugs, it could still be the case that, at each age, the death rate among smokers is higher than it is among non-smokers. If that is true, then at each age smokers face, in addition to those risks of dying at that age which they share with non-smokers, an excess risk of death at that age which, because the non-smoker avoids it, could be attributed to something that is closely associated with smoking, perhaps indeed simply to smoking itself. That there is such an excess risk and that it is due to smoking is the principal message of the anti-smoking movement. It is not our purpose here to question the correctness of that message but, granting it, to estimate the magnitudes of the risks in question.

Table 4 gives annual death rates by age and sex, first for deaths from tobacco-related conditions split into two separate groups, those due to some cause other than smoking itself and those among past and present smokers that are attributed to smoking, and then for all causes of death, including conditions that have not been related to smoking but excluding smoking itself. These death rates are compared in Table 5, firstly by comparing ever-smokers and never-smokers vis-a-vis their respective death rates from tobacco-related conditions and secondly by comparing the ever-smokers’ chance of dying from causes other than smoking with that of their dying because of their smoking. The first relative risk in Table 5 is the overall death rate for ever-smokers, viz. the sum of the two rates for tobacco-related conditions in Table 4, divided by the first of them, viz. the death rate for never-smokers. It assesses how much more likely an ever-smoker is to die of a tobacco-related condition than is a never-smoker of the same age and sex, and is the figure usually stated in warnings about the harmfulness of smoking. The second relative risk in Table 5 is simply the death rate for ever-smoker deaths caused by something other than smoking divided by the corresponding rate for those that are caused by smoking. For each age group it assesses how much more likely an ever-smoker is to die from causes other than smoking than he or she is to die because of smoking.

Table 4. Annual Death Rates: In tobacco-related conditions both for causes other than smoking, among smokers and non-smokers alike, and those among ever-smokers because of their smoking, together with those for all conditions, other than smoking, among smokers and non-smokers alike, by Age and Sex in Australia 1992

These two Tables need to be studied together. The first thing to notice is that Table 5 shows that for male ever-smokers of all ages the annual relative risk of dying from a tobacco-related condition fluctuates about 2, roughly the overall value of all male ever-smokers 20 years of age or older, sometimes above it, sometimes below it, increasing with age up to 64 years and thereafter decreasing with age. Female ever-smokers exhibit a similar but slightly more variable pattern with a somewhat overall lower relative risk of 1.4. For both sexes the relative risk peaks between 50 and 64 years of age. But while, for both sexes, these relative risks remain roughly constant at all ages before 65 years, Table 4 shows that the magnitude of the underlying actual risk to the ever-smoker varies enormously with age. For both sexes the actual annual risk that smoking will kill an ever-smoker at an early age is quite small, of the order of 1 in 30,000 before 30 years of age, about the same sort of chance as that of throwing 15 heads in a row with a fair coin. This annual risk increases steadily with age. By 60 years of age it has risen to 1 in 142 for males and 1 in 241 for females, about the same respective odds as those of throwing only 7 and 8 heads in a row with a fair coin. As they approach 70 years of age about 1 in every 100 male ever-smokers and 1 in every 200 female ever-smokers die each year because of their smoking and the corresponding figures for those who reach their eighties are 1 in 28 for males and 1 in 36 for females. At all ages ever-smokers of both sexes are more likely to die of causes other than smoking than they are to die because of their smoking and until they reach 40 years of age considerably more likely to do so. Overall, male ever-smokers are each year more than twice as likely, and female ever-smokers over four times as likely, to do so.

While what is deemed to be rare on the one hand and common on the other are largely matters of taste and colouring, it is nevertheless true that we regard young adult male deaths as rare events even though they occur needlessly often. Since deaths among 20 to 24 year old males from causes other than smoking occur among smokers and non-smokers alike at an annual rate of 1 in 850, it is consistent with ordinary linguistic usage to regard deaths that occur at an annual rate that is less than 1 in 1,000 as rare events. The terminology is also consistent with medical practice. For example in the 1970s anticipated mortality rates of one per thousand from diagnostic coronary arteriography were seen as acceptable because iatrogenic death would occur from it only rarely and its benefits would outweigh the small risk involved. With that terminology in mind, Table 4 shows that deaths from tobacco-related conditions are rare before 45 years of age for males and before 50 years of age for females, among both smokers and non-smokers. Moreover smoking rarely kills male ever-smokers before 50 years of age and female ever-smokers before 55 years of age, and does so very rarely at earlier ages. While deaths attributed to smoking do occur much more frequently with increasing age, so too do deaths from other causes and it is not clear how the ever-smoker’s age-increasing annual risk of death due to his or her smoking should be apportioned between smoking on the one hand and simply aging on the other. The anti-smoking movement’s message that smoking kills has to be interpreted from the balanced perspective of not only how likely it is to do so but of how likely it is that other causes will pre-empt that possibility by leading to death before it eventuates. For instance, while it may be a cause of concern to a 65 to 69 year old male ever-smoker and to a 70 to 74 year old female ever-smoker to be told that they have a yearly chance of about 1 in 100 that their smoking will kill them, that particular concern will not, perhaps, seem quite so overwhelming when they learn that, in any event, they also have about a 1 in 50 chance that they will die from other causes. To put the extreme case, an 80 plus year old ever-smoker is unlikely to be overly concerned that he or she has about a 1 in 30 yearly chance that it is their smoking that will kill them, when the yearly chance that other causes will do so is about 1 in 10.

The stark message that "smoking kills" deliberately suppresses salient facts about when and how likely it is to do so in order to evoke a fearful response. For the young and middle-aged ever-smoker the chance of it doing so soon are very small and, in the light of the much bigger chance of dying from other causes, perhaps too small to worry about. For the older ever-smoker the likelihood of imminent death due to smoking, though no longer small, is overshadowed by the concomitant higher chance of impending death from other causes. For ever-smokers of all ages the risk that their smoking will kill them is always less than the risk of dying from other causes, even if it is additional to it. If never-smokers have age-specific chances of imminent death that are small enough for them to go through life without constantly fretting about when death will come, then it seems just as reasonable that the ever-smoker should not be overly concerned about the even smaller additional chance that smoking may be the cause of his or her death.

The preceding comparison of smoker and non-smoker death rates is based on the estimated 1992 age-specific deaths attributable to tobacco that are given by English et al. in their QDM report⁵; those estimates were obtained by a "Condition-Specific Analysis". The report also considers another way of estimating those deaths viz. an "All-Cause Mortality Analysis". That analysis gives larger numbers of deaths at young ages, fewer deaths among the oldest, and overall it gives a smaller total number of deaths that can be attributed to tobacco; for those aged 35 years or more, 12,546 such deaths compared to the 18,775 of them attributed to tobacco by the Condition-Specific Analysis. It should come as no surprise that anti-smoking propaganda uses the higher estimate as the authoritative figure, and does not mention the smaller one, even though English et al⁵. do point out that both methods have limitations and that each has its strengths and weaknesses. But of the two estimates one is 50 percent larger than the other and, in the face of that, the layman might well wonder at the accuracy and usefulness of the two statistical methods that purport to estimate the number of deaths caused by smoking. The attitude of the anti-smoking lobby seems to be that the large disparity between the two methods of estimation is of no importance because even with the smaller estimate it can still be claimed that smoking kills a lot of people. While this may well be true it misses the point that it is difficult to be sure of the accuracy of either method when the discrepancy between them is so large. It is misleading to select one of the estimates to support one’s prior convictions and then disregard the other. Perhaps the ‘true’ number of deaths caused by smoking is even bigger than the higher estimate or even smaller than the lower estimate. One simply doesn’t know. It should be noted that the issue here is not the accuracy of a prediction, in the face of great uncertainty, about what might happen,; it is the accuracy of postdiction about what has already happened. If the Bureau of Meteorology stated that yesterday’s recorded maximum temperature was 45⁰C on one thermometer but was 30⁰C on another, then one might well question the usefulness of one or other thermometer, perhaps even both. It would rightly be considered unsatisfactory to argue that, while they each had strengths and weaknesses, it didn’t matter whether or not the thermometers were entirely accurate because both of them indicate that it was pretty warm yesterday and that is what we want to claim in order to persuade tourists to visit us.

The plain fact is that in medicine it is sometimes difficult to be very confident about the accuracy of one’s postdictory estimates. Different methods of estimation can give different estimates. It serves neither medicine itself nor the general public well to create an illusion of accuracy by selecting one set of estimates as authoritative and use them to bolster one’s prior convictions. To illustrate this and emphasise that the figures in Tables 4 and 5 should not be regarded as anything more than suggestive guidelines, Table 6 gives the results of analysing the same underlying mortality data in another way. The problem addressed in Table 6 is the same as that considered in Table 4, viz. how do the age-specific death rates due to smoking compare with those of non-smokers and those from other causes? The difference is that we now split ever-smokers into those who are ex-smokers and those who are still smoking, and use an "All-Cause Mortality Analysis" to obtain the death rates in question.

To compare Tables 4 and 6 it is helpful to start with Table 6 and, ignoring its figures for current and ex-smokers, to focus only on the remaining columns which purport to estimate the same quantities as their named counterparts in Table 4. For both sexes, the death rates among smokers and non-smokers from causes other than smoking are on the whole a little smaller in Table 4 than they are in Table 6, whether we focus on tobacco-related conditions only or on all causes of death other than smoking. In compensation the death rates among ever-smokers for deaths caused by smoking tend to be higher in Table 4 than they are in Table 6, markedly so at older ages, and for both sexes are over two and a half times bigger when all age groups are combined; 1 in 273 against 1 in 727 for males and 1 in 530 against 1 in 1,437 for females. This illustrates quite clearly how dependent one’s estimates can be on the method adopted for their estimation.

Perhaps the most notable feature of Table 6 is that, for both sexes, ex-smokers in each age group have a very much smaller risk of dying because of their smoking than do comparable current smokers, suggesting that ceasing to smoke leads to an effective reduction in the risk of death that is associated with smoking. But this leads to inconsistencies between some of the estimates in Table 6 and their counterparts in Table 4. For if the ex-smokers of an age group are dying, because of their smoking, at a lower rate that its current smokers, then so too should its ever-smokers since these consist of its ex-smokers together with its current smokers. While this is true within Table 6 itself, the Table 4 estimates of the ever-smoker death rates due to smoking are not always lower than the Table 6 estimates of those for current smokers; for males that is true for those 20 to 54 years of age but it is false for those 55 years of age and older, for females it is everywhere false except for the 75 to 79 year old age group.

Table 7 gives the "All-Cause Mortality Analysis" estimates of the age-specific relative risks that a current smoker dies from a tobacco-related condition, when compared to a peer never-smoker, and the relative risks that assess how much more likely a current smoker is to die from a cause other than smoking than he or she is to die because of their smoking. It is analogous to Table 5 which gives the corresponding estimates for ever-smokers calculated by a "Condition-Specific Analysis". Disparities between the two Tables are obvious. For example, from Table 5 both male and female ever-smokers aged 65 to 69 years are estimated to be about twice as likely to die soon of a tobacco-related condition as is a never-smoker, whereas for the current smokers in that age-group Table 7 gives not higher estimates, as one might expect, but lower estimates. Again from Table 5, a male ever-smoker of the same age-group is about twice as likely to die soon of a cause other than smoking as he is to die soon because of his smoking, whereas Table 7 says that if he is a current smoker he is almost 11 times more likely to do so.

The discrepancy between the two methods of analysis is disturbing because it makes it difficult to set much store by either of them. Nevertheless there is broad agreement that the risk of imminent death due to smoking, though initially very small, increases steadily with age in concert with that from all other causes. Moreover the relative risks in Table 7 suggest that at all ages both male and female current smokers, and not just the ever-smokers as in Table 5, are more likely to die of causes other than smoking than from their smoking itself, though only marginally so for male smokers in their early sixties. For the young and middle-aged current smoker, the chance of being killed soon by his or her smoking is remote and very much smaller than the chance of early death due to other causes. For the older current smoker the likelihood of imminent death due to smoking, though no longer remote, is overshadowed by the concomitant higher chance of impending death from other causes. Perhaps the Government message on tobacco-products that "smoking kills" would be closer to the truth if it were emended to read "smoking may kill, but relatively infrequently". Even so, it might be said, the large-scale aggregation of the morbidity associated with smoking imposes an intolerable and unfair burden on hospital services. In the next section we examine the extent to which that is so.

As we noted at the beginning of this paper, the anti-smoking movement usually presents the total number of hospital bed-days that are attributed to smoking as if its size alone will move us to see that the virtual elimination of smoking is needed to free the hospital system from that burden. But one does not even need to look at the figures in question to see that this could be a seriously misleading argument. If the burdens a person imposes on hospital services increase with aging, then presumably both smokers who quit smoking, and those who do not now take up smoking because of the perceived danger to their health, are expected to live longer than they would otherwise have done. As they age and eventually become victims of the morbidity due to causes other than smoking, they will enter into the hospital system at older ages than they would have done and at that time, they will impose correspondingly heavier burdens on hospital services. That this is indeed the case is shown very clearly when one compares the age-specific per person annual hospital bed-days rates of smokers and non-smokers, instead of just citing the total number of hospital bed-days for which smoking is claimed to be responsible.

Table 8 gives the annual hours in hospital per person at risk by age and sex first for tobacco-related conditions, both for those not caused by smoking and those that are, and second for all causes of morbidity other than smoking. Annual per person hospital bed-hours rates do increase with age and, at all ages and for both sexes, the rates due to causes other than smoking are much higher than those due to smoking. The aim of the anti-smoking movement is to delete eventually both the male and female columns of morbidity caused by smoking. But if that goal were achieved, then the people in question would remain at risk from causes of morbidity other than smoking and, if they remained healthy longer for not smoking, they would contribute to ‘All Causes’ columns in another year at a later age. Thus would-be smokers who had been persuaded never to smoke would swell the ranks of the older age groups where the per person rates at which hospital services are needed to treat morbidity for causes other than smoking are much larger than the would-have-been rates had they smoked and, because of it, become ill earlier. For example, males in the 55 to 59 year old age group who had they smoked would each contribute, on average, about 9 hospital bed-hours to the annual total, would, if they survived as non-smokers to the next age-group, then contribute, on average, almost 36 bed-hours to the annual total, a four-fold increase in the burden in question. If they survived even longer the annual burden would increase further. The only way in which this annual increase could fail to eventuate would be if all the would-have-been smokers became non-smoking victims of smoking-related conditions shortly after they would have done so as smokers, and that would call into question the extent to which smoking could be seen as a cause of those conditions.

Of course this annual increase has to be set against the possibility that the morbidity then commencing at a later age does not last as long as that which, with smoking, would have occurred at an earlier one. But while the preceding example of male smokers 55 to 59 years of age shifting to higher age groups, and further examination of Table 8, suggest that there would, on balance, still be an overall increase, the possibility in question does highlight the fact that annual figures, whether they be the total numbers favoured by the anti-smoking movement or the rates used here, do not give the whole picture.

It is difficult to see morbidity attributed to smoking as a burden on hospital services, let alone an unfair one, when without smoking the long-run overall burden would, in all likelihood, be considerably higher than it is now. Smokers who have allegedly died prematurely because of their smoking might, if they had not smoked, now be aged members of the population consuming many hospital bed-days. But while one would agree that premature death and sickness due to smoking is regrettable, and accept that people should not smoke if they want to be healthy and live longer, the claim that eliminating smoking will reduce the need for hospital services is misleading; on the contrary, in the long run it is likely to increase the need for them.

It is worth noting that annual per person hospital bed-days rates are much lower for tobacco-related conditions than they are for all causes other than smoking. But annual bed-days per episode are generally a little higher for tobacco-related conditions than they are for all causes other than smoking, even though the former account for fewer annual bed-days per person. Episodes due to conditions that are not related to smoking occur at a higher rate than those due to conditions that are so related but, on average, each of them involves slightly fewer hospital bed-days. We omit the details.

Finally it should be kept in mind that while the morbidity rates in Table 8 are informative guidelines they should not be regarded as any more accurate than the corresponding mortality rates in Table 4.

There have been several attempts to estimate the monetary costs that are said to result from the harmful effects of cigarette smoking. For example Collins and Lapley estimated that the economic costs of tobacco added up to 12,736.2 million dollars in 1992, but ACIL estimated that they were 8,600 million dollars for 1992-93 and the National Centre for Health Program Evaluation and the Australian Institute of Health found a much lower cost for 1989-90, viz. 1,300 million dollars. The disparity between these estimates does little to suggest that any of them should be accepted with confidence. It should be noted that when Doran, Sanson-Fisher and Gordon compared the publicly financed health care expenditure attributed to smoking with the tobacco taxes paid by smokers they found that the taxation benefit to the government was more than 3 times the cost of that expenditure. Since they also found that medical costs accounted for 57 percent of the governm