Health econ stuff
In a post I published a few weeks ago I mentioned that I had decided against including some comments and observations I had written about health economics in that post because the post was growing unwieldy, but that I might post that stuff later on in a separate post. This post will include those observations, as well as some additional details I added to the post later. This sort of post is the sort of post that usually does not get past the ‘draft’ stage (in wordpress you can save posts you intend to publish later on as drafts), and as is usually the case for posts like these I already regret having written it, for multiple reasons. I should warn you from the start that this post is very long and will probably take you some time to read.
Anyway, the starting point for this post was some comments related to health insurance and health economics which I left on SSC in the past. A lot more people read those comments on SSC than will read this post so the motivation for posting it here was not to ‘increase awareness’ of the ideas and observations included in some kind of general sense; my primary motivation for adding this stuff here is rather that it’s a lot easier for me personally to find stuff I’ve written when it’s located here on this blog rather than elsewhere on the internet, and I figure that some of the things I wrote back then are topics which might well come up again later, and it would be convenient for me in that case to have a link at hand. Relatedly I have added many additional comments and observations in this post not included in the primary exchange, which it is no longer possible for me to do on SSC as my comments are no longer editable on that site.
Although the starting point for the post was as mentioned a comment exchange, I decided early on against just ‘quoting myself’ in this post, and I have thus made some changes in wording and structure in order to increase the precision of the statements included and in order to add a bit of context making the observations below easier to read and understand (and harder to misread). Major topics to which the observations included in this post relate are preventable diseases, the level of complexity that is present in the health care sector, and various topics which relate to health care cost growth. Included in the post are some perhaps not sufficiently well known complications which may arise in the context of the discussion of how different financing schemes may relate to various outcomes, and to cost growth. Much of the stuff included will probably be review to people who’ve read my previous posts on health economics, but that’s to be expected considering the nature of this post.
Although ‘normative stuff’ is not what interests me most – I generally tend to prefer discussions where the aim is to identify what happens if you do X, and I’ll often be happy to leave the discussion of whether outcome X or Y is ‘best’ to others – I do want to start out with stating a policy preference, as this preference was the starting point for the aforementioned debate that lead to the origination of this post. At the outset I should thus make clear that I would in general be in favour of changes to the financial structure of health care systems where people who take avoidable risks which systematically and demonstrably increase their expected health care expenditures at the population level pay a larger proportion of the cost than do people who did not take such avoidable risks.
Most developed societies have health care systems which are designed in a way that implicitly to some extent subsidize unhealthy behaviours. An important note in this context is incidentally that one way of looking at these things is that if you are not explicitly demanding people who behave in risky ways which tend to increase their expected costs to pay more for their health care (/insurance), then you are in fact by virtue of not doing this implicitly subsidizing those unhealthy individuals/behaviours. I mention this because some people might not like the idea of ‘subsidizing healthy behaviours’ (‘health fascism’) – which from a certain point of view is what you do if you charge people who behave in unhealthy ways more. Maybe some people would take issue with words like ‘subsidy’ and ‘implicit’, but regardless of what you call these things the major point that is important to have in mind here is that if one group of people (e.g. ‘unhealthy people’) cost more to treat (/are ill more often, get illnesses related to their behaviours, etc., etc.) than another group of people (‘healthy people’), then if you need to finance this shortfall – which you do, as you face a budget constraint – there are only two basic ways to do this; you can either charge the high-cost group (‘unhealthy people’) more, or you can require the other group (‘healthy people’) to make up the difference. Any scheme which deals with such a case of unequal net contribution rates are equivalent either to one of those schemes or a mix of the two, regardless of what you call things and how it’s done, and regardless of which groups we are talking about (old people also have higher health care expenditures than do young people, and most health care systems implicitly redistribute income from the young to the old). If you’re worried about ‘health fascism’ and the implications of subsidizing healthy behaviours (/’punishing’ unhealthy behaviours) you should at least keep in mind that if the health care costs of people who live healthy lives and people who do not are dissimilar then any system that deals with this issue – which all systems must – can either choose to ‘subsidize’ healthy behaviours or unhealthy behaviours; there’s no feasible way to design a ‘neutral system’ if the costs of the groups are dissimilar.
Having said all this, the very important next point is then that it is much more difficult to make simple schemes that would accomplish an outcome in which people who engage in unhealthy behaviours are required to pay more without at the same time introducing a significant number of new problems than people who are not familiar with this field would probably think it is. And it’s almost certainly much harder to evaluate if the proposed change actually accomplished what you wanted to accomplish than you think it is. Even if we are clear about what we want to accomplish and can all agree that that is what we are aiming for – i.e. we are disregarding the political preferences of large groups of voters and whether the setup in question is at all feasible to accomplish – this stuff is really much harder than it looks, for many reasons.
Let’s start out by assuming that smoking increases the risk of disease X by 50%. Say you can’t say which of the cases of X are caused by smoking, all you know is that smoking increases the risk at the population level. Say you don’t cover disease X at all if someone smokes, that is, smokers are required to pay the full treatment cost out of pocket if they contract disease X. It’s probably not too controversial to state that this approach might by some people be perceived of as not completely ‘fair’ to the many smokers who would have got disease X even if they had not smoked (a majority in this particular case, though of course the proportion will vary with the conditions and the risk factors in question). Now, a lot of the excess health care costs related to smoking are of this kind, and it is actually a pretty standard pattern in general with risk factors – smoking, alcohol, physical inactivity, etc. You know that these behaviours increase risk, but you usually can’t say for certain which of the specific cases you observe in clinical practice are actually (‘perfectly’/’completely’/’partially’?) attributable to the behaviour. And quite often the risk increase associated with a specific behaviour is actually really somewhat modest, compared to the relevant base rates, meaning that many of the people who engage in behaviours which increase risk and who get sick might well have got sick even if they hadn’t engaged in those risky behaviours.
On top of this problem usually it’s also the case that risk factors interact with each other. Smoking increases the risk of cancer of the esophagus, but so does alcohol and obesity, and if a person both smokes and drinks the potential interaction effect may not be linear – so you most likely often can’t just identify individual risk factors in specific studies and then pool them later and add them all together to get a proper risk assessment. A further complication is that behaviours may both increase as well as decrease risk – to stick with the example, diets high in fruits and vegetables both lower the risk of cancer of the esophagus. Exercise probably does as well – we know that exercise has important and highly complex effects on immune system function (see e.g. this post). Usually a large number of potential risk factors is at play at the same time, there may be multiple variables which lower risk and are also important to include if you want a proper risk assessment, and even if you knew in theory which interaction terms were likely to be relevant, you might even so find yourself in a situation unable to estimate the interaction terms of interest – this might take high-powered studies with large numbers of patients, which may not be available or the results of such high-powered studies may not apply to your specific subgroup of patients. Cost-effectiveness is also an issue – it’s expensive to assess risk properly. One take-away is that you’ll still have a lot of unfairness in a modified contribution rate model, and even evaluating fairness aspects of the change may be difficult to impossible because to some extent this question is unknowable. You might find yourself in a situation where you charge the obese guy more because obesity means he’s high risk, but in reality he is actually lower risk than is the non-fat guy who is charged a lower rate, because he also exercises and eats a lot of fruits and vegetables, which the other guy doesn’t.
Of course the above paragraph took it for granted that it was even possible to quantify the excess costs attributable to a specific condition. That may not be easy at all to do, and there may be large uncertainties involved. The estimated excess cost will depend upon a variety of factors which may or may not be of interest to the party performing the analysis, for example it may be very important which time frame you’re looking at and which discounting methodology is applied (see e.g. the last paragraph in this post). The usual average vs marginal cost problem (see the third-last paragraph in the post to which I link in the previous sentence – this post also has more on this topic) also applies here and is related to ‘the fat guy who exercises and is low-risk’-problem; ideally you’d want to charge people with higher health care utilization levels more (again, in a setting where we assume the excess cost is associated with life-style variables which are modifiable – this was our starting point), but if there’s a large amount of variation in costs across individuals in the specific subgroups of interest and you only have access to average costs rather than individual-level costs, then a scheme only taking into account the differences in the averages may be very sub-optimal when you look at it from the viewpoint of the individual. Care needs to be taken to avoid problems like e.g. Simpson’s paradox.
Risk factors are not the only things that cluster; so do diseases. An example:
“An analysis of the Robert Koch-Institute (RKI) from 2012 shows that more than 50 % of German people over 65 years suffer from at least one chronic disease, approximately 50 % suffer from two to four chronic diseases, and over a quarter suffer from five or more diseases .” (link)
78.3 % of the type 2 diabetics also suffered from hypertension in that study. Does this fact make it easier or harder to figure out what is ‘the true cost contribution’ of ‘type 2 diabetes’ and ‘hypertension’ (and, what we’re ultimately interested in in this setting – the ‘true cost contribution’ of the unhealthy behaviours which lead some individuals to develop type 2 diabetics and hypertension who would not otherwise have developed diabetes and/or hypertension (…/as early as they did)? It should be noted that diabetes was estimated to account for 11 % of total global healthcare expenditure on adults in 2013 (link). That already large proportion is expected to rise substantially in the decades to come – if you’re interested in cost growth trajectories, this is a major variable to account for. Attributability is really tricky here, and perhaps even more tricky in the case of hypertension – but for what it’s worth, according to a CDC estimate hypertension cost the US $46 billion per year, or ~$150/per person per year.
Anyway, you look at the data and you make guesses, but the point is that doctor Smith won’t know for certain if Mr. Hanson would have had a stroke even if he hadn’t smoked or not. A proposal of not providing payment for a health care service or medical product in the case of an ‘obviously risky-behaviour-related-health-condition’ may sometimes appear to be an appealing proposition and you sometimes see people make this sort of proposal in discussions of this nature, but it tends to be very difficult when you look at the details to figure out just what those ‘obviously risky-behaviour-related-health-conditions’ are, and even harder to make even remotely actuarially fair adjustments to the premiums and coverage patterns to reflect the risk. Smoking and lung cancer is a common example of a relatively ‘clean’ case, but most cases are ‘less clean’ and even here there are complications; a substantial proportion of lung cancer cases are not caused by tobacco – occupational exposures also cause a substantial proportion of cases, and: “If considered in its own disease category […] lung cancer in never smokers would represent the seventh leading cause of cancer mortality globally, surpassing cancers of the cervix, pancreas, and prostate,5 and among the top 10 causes of death in the United States.” (link) Occupational exposures (e.g. asbestos) are not likely to fully account for all cases, and for example it has also been found that other variables, including previous pneumonia infections and tuberculosis, affect risk (here are a couple of relevant links to some previous coverage I wrote on these topics).
I think many people who have preferences of this nature (‘if it’s their own fault they’re sick, they should pay for it themselves’) underestimate how difficult it may be to make changes which could be known with a reasonable level of certainty to actually have the intended consequences, even assuming everybody agreed on the goal to be achieved. This is in part because there are many other aspects and complications which need to be addressed as well. Withholding payment in the case of costly preventative illness may for example in some contexts increase cost, rather than decrease them. The risk of complications of some diseases – an important cost driver in the context of diabetes – tends to be dependent on post-diagnosis behavioural patterns. The risk of developing diabetes complications will depend upon the level of glycemic control. If you say you won’t cover complications at all in the case of ‘self-inflicted disease X’, then you also to some extent tend to remove the option of designing insurance schemes which might lower cost and complication rates post-diagnosis by rewarding ‘good’ (risk-minimizing) behaviours post-diagnosis and punishing ‘bad’ (risk-increasing) behaviours. This is not desirable in the context of diseases where post-diagnosis behaviour is an important component of the cost function, as it certainly is in the diabetes context. There are multiple potential mechanisms here, some of which are disease specific (e.g. suboptimal diet in a diagnosed type 2 diabetic) and some of which may not be (a more general mechanism could e.g. be lowered compliance/adherence to treatment in the uncovered populations because they can’t afford the drugs which are required to treat their illness; though the cost-compliance link is admittedly not completely clear in the general case, there are certainly multiple diseases where lowered compliance to treatment would be expected to increase cost long-term).
And again, also in the context of complications fairness issues are not as simple to evaluate as people might like them to be; some people may have a much harder time controlling their disease than others, or they may be more susceptible to complications given the same behaviour. Some may already have developed complications by the time of diagnosis. Such issues make it difficult to design simple rules which would achieve what you want them to achieve without having unfortunate side-effects; for example a rule that a microvascular diabetes-related complication is automatically ‘your own fault’ (so we won’t pay for it), which might be motivated by the substantial amount of research linking glycemic control with complication risk, would punish some diabetics who have had the disease for a longer amount of time (many complications are not only strongly linked to Hba1c but also display a substantial degree of duration-dependence; for example in type 1 diabetics one study found that diabetic retinopathy was present in 13% of patients with a duration of disease less than 5 years, whereas the corresponding figure was 90% for individuals with a disease duration of 10–15 years (Sperling et al., p. 393). I also recall reading a study finding that Hba1c itself is increasing with diabetes duration, which may be partly accounted for by the higher risk of hypoglycemia related to hypoglycemia-unawareness-syndromes in individuals with long-standing disease), individuals with diseases which are relatively hard to control (perhaps due to genetics, or maybe again due to the fact that they have had the disease for a longer amount of time; the presence of hypoglycemia unawareness is as alluded to above to a substantial degree duration-dependent, and this problem increases the risk of hospitalizations, which are expensive), diabetics who developed complications before they knew they were sick (a substantial proportion of type 2 diabetics develop some degree of microvascular damage pre-diagnosis), and diabetics with genetic variants which confer an elevated risk of complications (“observations suggest that involvement of genetic factors is increasing the risk of complications” (Sperling et al., p. 226), and for example in the DCCT trial familial clustering of both neuropathy and retinopathy was found; clustering which persisted after controlling for Hba1c – for more on these topics, see e.g. Sperling et al.’s chapter 11).
Other decision rules would similarly lead to potentially problematic incentives and fairness issues; for example requiring individuals to meet a specific Hba1c goal might be more desirable than to just not cover complications, but that one also leads to potential problems; ideally such an Hba1c goal should be individualized, because of the above-mentioned complexities and others I have not mentioned here; to require a newly-diagnosed individual to meet the same goals as someone who has had diabetes for decades does not make sense, and neither does it make sense to require these two groups to meet exactly the same Hba1c goal as the middle-aged female diabetic who desires to become pregnant (diabetes greatly increases the risk of pregnancy complications, and strict glycemic control is extremely important in this patient group). It’s important to note that these issues don’t just relate to whether or not the setup is perceived of as fair, but it also relates to whether or not you would expect the intended goals to actually be met or not when you implement the rule. If you were to require that a long-standing diabetic with severe hypoglycemia unawareness had to meet the same Hba1c goal as the newly diagnosed individual, this might well lead to higher overall cost, because said individual might suffer a large number of hypoglycemia-related hospitalizations which would have been avoidable if a more lax requirement was imposed; when you decrease Hba1c you decrease the risk of long-term complications, but you increase the risk of hypoglycemia. A few numbers might make it easier to make sense of how expensive hospitalizations really are, and why I emphasize them here. In this diabetes-care publication they assign a cost for an inpatient day for a diabetes-related hospitalization at $2,359 and an emergency visit at ~$800. The same publication estimates the total average annual excess expenditures of diabetics below the age of 45 at $4,394. Going to the hospital is really expensive (43% of the total medical costs of diabetes are accounted for by hospital inpatient care in that publication).
A topic which was brought up in the SSC discussion was the question of the extent to which private providers have a greater incentive to ‘get things right’ in terms of assessing risk. I don’t take issue with this notion in general, but there are a lot of complicating factors in the health care context. One factor of interest is that it is costly to get things right. If you’re looking at this from an insurance perspective, larger insurance providers may be better at getting things right because they can afford to hire specialists who provide good cost estimates – getting good cost estimates is really hard, as I’ve noted above. Larger providers translate into fewer firms, which increases firm concentration and may thus increase collusion risk, which may again increase the prices of health care services. Interestingly if your aim is to minimize health care cost growth increased market power of private firms may actually be a desirable state of affairs/goal, because cost growth is a function of both unit prices and utilization levels, and higher premiums are likely to translate into lower utilization rates, which may lower overall costs and -cost growth. I decided to include this observation here also in order to illustrate that what is an optimal outcome depends on what your goal is, and in the setting of the health care sector you sometimes need to be very careful about thinking about what your actual goal is, and which other goals might be relevant.
When private insurance providers become active in a market that also includes a government entity providing a level of guaranteed coverage, total medical outlays may easily increase rather than decrease. The firms may meed an unmet need, but some of that unmet need may be induced demand (here’s a related link). Additionally, the bargaining power of various groups of medical personnel may change in such a setting, leading to changes in compensation schedules which may not be considered desirable/fair. An increase in total outlays may or may not be considered a desirable outcome, but this does illustrate once again the point that you need to be careful about what you are trying to achieve.
There’s a significant literature on how the level of health care integration, both at the vertical and horizontal level, both in terms of financial structure and e.g. in terms of service provision structure, may impact health care costs, and this is an active area of research where we in some contexts do not yet know the answers.
Even when cost minimization mechanisms are employed in the context of private firms and the firm in question is efficient, the firm may not internalize all relevant costs. This may paradoxically lead to higher overall cost, due to coverage decisions taken ‘upstream’ influencing costs ‘downstream’ in an adverse manner; I have talked about this topic on this blog before. A diabetic might be denied coverage of glucose testing materials by his private insurer, and that might mean that the diabetic instead gets hospitalized for a foreseeable and avoidable complication (hypoglycemic coma due to misdosing), but because it might not be the same people paying for the testing material and the subsequent hospitalization it might not matter to the people denying coverage of the testing materials, and/so they won’t take it into account when they’re making their coverage decisions. That sort of thing is quite common in the health care sector – different entities pay for and receive payments for different things, and this is once again a problem to keep in mind if you’re interested in health care evaluation; interventions which seem to lower cost may not do so in reality, because the intervention lead to higher health care utilization elsewhere in the system. If incentives are not well-aligned things may go badly wrong, and they are often not well-aligned in the health care sector. When both the private and public sectors are involved in either the financial arrangements and/or actual health service provision – which is the default health care system setup for developed societies – this usually leads to highly complex systems, where the scope for such problems to appear seems magnified, rather than the opposite. I would assume that in many cases it matters a lot more that incentives are well-aligned than which specific entity is providing insurance or health care in the specific context, in part a conclusion drawn from the coverage included in Simmons, Wenzel & Zgibor‘s book.
In terms of the incentive structures of the people involved in the health care sector, this stuff is of course also adding another layer of complexity. In all sectors of the economy you have people with different interests who interact with each other, and when incentives change outcomes change. Outcomes may be car batteries, or baseball bats, or lectures. Evaluating outcomes is easier in some settings than in others, and I have already touched upon some of the problems that might be present when you’re trying to evaluate outcomes in the health care context. How easy it is to evaluate outcomes will naturally vary across sub-sectors of the health care sector but a general problem which tends to surface here is the existence of various forms of asymmetrical information. There are multiple layers, but a few examples are worth mentioning. To put it bluntly, the patient tends to know his symptoms and behavioural patterns – which may be disease-relevant, and this aspect is certainly important to include when discussing preventative illnesses caused at least in part by behaviours which increase the risk of said illnesses – better than his doctor, and the doctor will in general tend to know much more about the health condition and potential treatment options than will the patient. The patient wants to get better, but he also wants to look good in the eyes of the doctor, which means he might not be completely truthful when interacting with the doctor; he might downplay how much alcohol he drinks, misrepresent how often he exercises, or he may lie about smoking habits or about how much he weighs. These things make risk-assessments more difficult than they otherwise might have been. As for the GPs, usually we here have some level of regulation which restricts their behaviour to some extent, and part of the motivation for such regulation is to reduce the level of induced demand which might otherwise be the result of information asymmetry in the context of stuff like relevant treatment effects. If a patient is not sufficiently competent to evaluate the treatments he receives (‘did the drug the doctor ordered really work, or would I have gotten better without it?’), there’s a risk he might be talked into undergoing needless procedures or take medications for which he has no need, especially if the doctor who advises him has a financial interest in the treatment modality on offer.
General physicians have different incentives from nurses and specialists working in hospitals, and all of these groups may experience conflicts of interests when they’re dealing with insurance providers and with each other. Patients as mentioned have their own set of incentives, which may not align perfectly with those of the health care providers. Different approaches to how to deal with such problems lead to different organizational setups, all of which influence both the quantity and quality of care, subject to various constraints. It’s an active area of research whether decreasing competition between stakeholders/service providers may decrease costs; one thing that is relatively clear from diabetes research with which I have familiarized myself is that when different types of care providers coordinate activities, this tends to lead to better outcomes (and sometimes, but not always, lower costs), because some of the externalized costs become internalized by virtue of the coordination. It seems very likely to me that conclusions to such questions will be different for different subsectors of the health care sector. A general point might be that more complex diseases should be expected to be more likely to generate cost savings from increased coordination than should relatively simple diseases (if you’re fuzzy about what the concept of disease complexity refers to, this post includes some relevant observations). This may be important, because complex diseases also should probably tend to be more expensive to treat in general, because the level of need in patients is higher.
It’s perhaps hardly surprising, considering the problems I’ve already discussed related to how difficult it may be to properly assess costs, that there’s a big discussion to be had about how to even estimate costs (and benefits) in specific contexts, and that people write books about these kinds of things. A lot of things have already been said on this topic and a lot more could be said, but one general point perhaps worth repeating is that it may in the health care sector be very difficult to figure out what things (‘truly’) cost (/’is worth’). If you only have a public sector entity dealing with a specific health problem and patients are not charged for receiving treatment, it may be very difficult to figure out what things ‘should’ cost because relevant prices are simply missing from the picture. You know what the government entity paid the doctors in wages and what it paid for the drugs, but the link between payment and value is sometimes a bit iffy here. There are ways to at least try to address some of these issues, but as already noted people write books about these kinds of things so I’m not going to provide all the highlights here – I refer to the previous posts I’ve written on these topics instead.
Another important related point is that medical expenditures and medical costs are not synonyms. There are many costs associated with illness which are not directly related to e.g. a payment to a doctor. People who are ill may be less productive while they are at work, they may have more sick-days, they may retire earlier, their spouse may cut down on work hours to take care of them instead of going to work, a family caretaker may become ill as a result of the demands imposed by the caretaker role (for example Alzheimer’s disease significantly increases the risk of depression in the spouse). Those costs are relevant, there are literatures on these things, and in some contexts such ‘indirect costs’ (e.g. lower productivity at work and early retirement) may make up a very substantial proportion of the total costs of a health condition. I have seen diabetes cost estimates which indicated that the indirect costs may account for as much as 50 % of the total costs.
If there’s a significant disconnect between total costs and medical expenditures then minimizing expenditures may not be desirable from an economic viewpoint. A reasonable assessment model will/should in the context of models of outlays include both a monetary cost parameter and a quality/quantity (ideally both) parameter; if you neglect to take account of the latter, in some sense you’re only dealing with what you pay out, not what you get for that payment (which is relevant). If you don’t take into account indirect costs you implicitly allow cost switching practices to potentially muddle the picture and make assessments more difficult; for example if you provide fewer long-term care facilities then the number of people involved in ‘informal care’ (e.g. family members having to take care of granny) will go up, and that is going to have secondary effects downstream which should also be assessed (you improve the budget in the context of the long-term care facilities, but you may at the same time increase demands on e.g. psychiatric institutions and marginally lower especially the female labour market participation rate. The net effect may still be positive, but the point is that an evaluation will/should include costs like these in the analysis, at least if you want anything remotely close to the full picture).
Let’s return to those smokers we talked about earlier. A general point not mentioned yet is that if you don’t cover smokers in the public sector because of cost considerations, many of them may also not be covered by private insurance either. This is because a group of individuals that is high risk and expensive to treat will be demanded high premiums (or the insurance providers would go out of business), and for the sake of this discussion we’re now assuming smokers are expensive. If that is so, many of them probably would not be able to afford the premiums demanded. Now, one of the health problems which are very common in smokers is chronic obstructive pulmonary disease (COPD). Admission rates for COPD patients differ as much as 10-fold between European countries, and one of the most important parameters regarding pharmacoeconomics is the hospitalization rate (both observations are from this text). What does this mean? It means that we know that admission rate from COPD is highly responsive to the treatment regime; populations well-treated have much fewer hospitalizations. 4% of all Polish hospitalizations are due to COPD. If you remove the public sector subsidies, the most likely scenario you get seems to me to be a poor-outcomes scenario with lots of hospitalizations. Paying for those is likely to be a lot more expensive than it is to treat the COPD pharmacologically in the community. And if smokers aren’t going to be paying for it, someone else will have to do that. If you both deny them health insurance and refuse them treatment if they cannot pay for it they may just die of course, but in most cost-assessment models that’s a high-cost outcome, not a low-cost outcome (e.g. due to lost work-life productivity etc. Half of people with COPD are of working age, see the text referred to above.). This is one example where the ‘more fair’ option might lead to higher costs, rather than lower costs. Some people might still consider such an outcome desirable, it depends on the maximand of interest, but such outcomes are worth considering when assessing the desirability of different systems.
A broadly similar dynamic, in the context of post-diagnosis behaviour and links to complications and costs, may be present in the context of type 2 diabetes. I know much more about diabetes than I do about respirology, but certainly in the case of diabetes this is a potentially really big problem. Diabetics who are poorly regulated tend to die a lot sooner than other people, they develop horrible complications, they stop being able to work, etc. etc. Some of those costs you can ignore if you’re willing to ‘let them die in the streets’ (as the expression goes), but a lot of those costs are indirect costs due to lower productivity, and those costs aren’t going anywhere, regardless of who may or may not be paying the medical bills of these people. Even if they have become sick due to a high-risk behaviour of their own choosing, their health care costs post-diagnosis will still be highly dependent upon their future medical care and future health insurance coverage. Denying them coverage for all diabetes-related costs post-diagnosis may, paradoxical though it may seem to some, not be the cost-minimizing option.
I already talked about information asymmetries. Another problematic aspect linked to information management also presents itself here in a model of this nature (‘deny all diabetes-related coverage to known diabetics’); people who suspect they might be having type 2 diabetes may choose not to disclose this fact to a health care provider because of the insurance aspect (denial of coverage problems). Insurance providers can of course (and will try to) counter this by things like mandatory screening protocols, but this is expensive, and even assuming they are successful you again not only potentially neglect to try to minimize the costs of the high-cost individuals in the population (the known diabetics, who might be cheaper long-term if they had some coverage), you also price a lot of non-diabetics out of the market (because premiums went up to pay for the screening). And some of those non-diabetics are diabetics to-be, who may get a delayed diagnosis as a result, with an associated higher risk of (expensive) complications. Again, as in the smoking context if the private insurer does not cover the high-cost outcomes someone else will have to do that, and the blind diabetic in a wheel-chair is not likely to be able to pay for his dialysis himself.
More information may in some situations lead to a breakdown in insurance markets. This is particularly relevant in the context of genetics and genetic tests. If you have full information, or close to it, the problem you have to some extent stops being an insurance problem and instead becomes a problem of whether or not to, and to which extent you want to-, explicitly compensate people for having been dealt a bad hand by nature. To put it in very general terms, insurance is a better framework for diseases which can in principle be cured than it is for chronic conditions where future outlays are known with a great level of certainty; the latter type of disease tends to be difficult to handle in an insurance context.
People who have one disease may develop other diseases as time progresses, and having disease X may increase or decrease the risk of disease Y. People study such disease variability patterns, and have done so for years, but there’s still a lot of stuff we don’t know – here’s a recent post on these topics. Such patterns are interesting for multiple reasons. One major motivation for studying these things is that ‘different’ diseases may have common mechanisms, and the identification of these mechanisms may lead to new treatment options. A completely different motivation for studying these things relate rather to the kind of stuff covered in this post, where you instead wonder about economic aspects; for example, if the smoker stops smoking he may gain weight and eventually develop type 2 diabetes instead of developing some smoking-related condition. Is this outcome better or worse than the other? It’s important to keep in mind when evaluating changes in compensation schedules/insurance structures that diseases are not independent, and this is a problem regardless of whether you’re interested in total costs or ‘just’ direct outlays. Say you’re ‘only’ worried about outlays and you are trying to figure out if it is a good idea to deny coverage to smokers, and you know that ex-smokers are likely to gain weight and have an increased risk of type 2 diabetes. Then the relevant change in cost is not the money you save on smoking-related illness, it’s the cost change you arrive at when after you account for those savings also account for the increased cost of treating type 2 diabetes. Disease interdependencies are probably as complex as risk factor interdependencies – the two phenomena are to some extent representing the same basic phenomenon – so this makes true cost evaluation even harder than it already was. Not all relevant costs at the societal level are of course medical costs; if people live longer, and they rely partly on a pension scheme to which they are no longer contributing, that cost is also relevant.
If a group of people who live longer cost more than a group of people who do not live as long, and you need to cover the associated shortfall, then – as we concluded in the beginning – there are really only two ways to handle this: Make them pay more than the people who do not live as long, or make the people who do not live as long pay more to cover the shortfall. Another way to look at this is that in this situation you can either tax people ‘for not living long enough’, or you can tax people for ‘not dying at the appropriate time’. On the other hand (?), if a group of people who die early turns out to be the higher-cost group in the relevant comparison (perhaps because they have shorter working lives and so pay into the system for a shorter amount of time), then you can deal with this problem by… either taxing them for ‘not living long enough’ or by punishing the people who live long lives for ‘not dying at the appropriate time’. No, of course it doesn’t matter which group is high cost, the solution mechanism is the same in both cases – make one of the groups pay more. And every time you tweak things you change the incentives of various people, and implicit effects like these hide somewhere in the background.
No comments yet.