An Economic Theory of Patient Decision-Making

If the test shows that the patient’s blood glucose level is greater than 125 mg/dl, then the patient is said to suffer from diabetes, and the higher the level of blood glucose, the worse

An Economic Theory of Patient Decision-Making*

Douglas O Stewart, PhD (Correspondent)

Associate ProfessorDepartment of EconomicsCleveland State UniversityEuclid Avenue at East 24th StreetCleveland, Ohio USATelephone: (216) 687-4515Fax: (216) 687-9206Email: d.o.stewart@csuohio.edu

andJoseph P DeMarco, PhD

ProfessorDepartment of PhilosophyCleveland State UniversityEuclid Avenue at East 24th StreetCleveland, Ohio USATelephone: (216) 687-3901Fax: (216) 752-1208Email: j.demarco@csuohio.eduWord count: main text—6806, abstract 221

An Economic Theory of Patient Decision-Making Patient autonomy, as exercised in the informed consent process, is a central concern in bioethics.The typical bioethicist’s analysis of autonomy centers on decisional capacity – finding the line between autonomy and its absence This approach leaves unexplored the structure of reasoning behind patient treatment decisions

To counter that approach, we present a microeconomic theory of patient decision-making

regarding the acceptable level of medical treatment from the patient’s perspective We show that arational patient’s desired treatment level typically departs from the level yielding an absence of

symptoms, the level we call ideal This microeconomic theory demonstrates why patients have

good reason not to pursue treatment to the point of absence of physical symptoms

We defend our view against possible objections that it is unrealistic and that it fails toadequately consider harm a patient may suffer by curtailing treatment

Our analysis is fruitful in a various ways It shows why decisions often consideredunreasonable might be fully reasonable It offers a theoretical account of how physicianmisinformation may adversely affect a patient’s decision It shows how billing costs influencepatient decision-making It indicates that health care professionals’ beliefs about the

“unreasonable” attitudes of patients might often be wrong It provides a better understanding ofpatient rationality that should help to ensure fuller information as well as increased respect forpatient decision-making

An Economic Theory of Patient Decision-Making

Patient autonomy, exercised in the informed consent process, is a central concern in bioethics The typical analysis of decision-making centers on decisional capacity  finding the line between autonomy and its absence This approach leaves unexplored the structure of the reasoning behind

a patient’s autonomous decisions To understand the structure of patient reasoning, we present a microeconomic theory of patient decision-making This theory is fruitful in several ways: It shows that decisions often considered unreasonably noncompliant may be in the patient’s interest; such decisions should be respected It explains how physician misinformation adversely affects a patient’s decision; it places responsibility on the physician to provide reliable and full information about both costs and benefits of treatment Furthermore, the theory provides an explicit account

of how billing costs influence patient decision-making; it offers information about the impact of third party payment

Our approach may be controversial because it concludes that patients may frequently have good reasons against pursuing treatment to the level at which all physical symptoms are absent We assume the main goal of medical intervention is treatment until physical symptoms are absent Because health care professionals might view decisions not to eliminate physical symptoms as unreasonable, these professionals might present information in a way that improperly influences patients to pursue further treatment A better understanding of patient rationality should help to ensure accurate information and should increase respect for patient decision-making

Our approach is developed mainly in terms of the diagnosis and treatment of diabetes although the approach is generally applicable.1 We offer a series of graphs to explain decision-making concerning a patient’s desired treatment level The graphs are of the sort used in microeconomic analysis We show that the patient’s desired treatment level departs from the level yielding

absence of symptoms After explaining what we take to be the importance of this analysis, we defend it against claims that it is unrealistic and that it fails to consider the harm a patient may suffer by curtailing treatment.

IPhysicians often diagnose diseases by administering tests to find whether the results differ from

a predetermined physiological level Hypertension and diabetes are examples of diseases whose diagnoses depend on such tests One test for diabetes involves determining whether a patient's blood glucose level is greater than 125 mg/dl If the test shows that the patient’s blood glucose level is greater than 125 mg/dl, then the patient is said to suffer from diabetes, and the higher the level of blood glucose, the worse the condition.2

Our presentation of an economic theory of patient decision-making about diabetes and its

treatment involves four levels of blood glucose test results: (1) G o, the starting or original blood

glucose level assumed to be substantially above 125 mg/dl, (2) G I, the current blood glucose

diagnosis level of 125 mg/dl, (3) G n, a blood glucose level somewhat above 125 mg/dl which we show based on the economic theory is the minimum level of blood glucose to which diabetes

should be treated, and (4) G *, the blood glucose level the patient would choose as the goal of

treatment based on the economic theory We assume that G I is the level physicians consider ideal for health and at which they diagnose diabetes because at it no symptoms of diabetes are expectedfor the typical patient

We begin our presentation by using economic theory to examine the rationality of G I as the treatment level for diabetes.3 Suppose that a patient makes an initial visit to a physician and that

the physician conducts a blood glucose test Further suppose that the test’s result is G o, a blood glucose level substantially above the diagnosis level, 125 mg/dl, so that the physician diagnoses

that the patient suffers from diabetes Should the diagnosis level, G I, be adopted as the goal of treatment, or should the treatment goal be different? And, if the diagnosis level of blood glucose,

G I, should not be adopted as the goal of treatment, what level of blood glucose should be used?

G I is determined entirely by a lack of symptoms and does not take all patient benefits or costs into account To be fully rational from the patient’s perspective, benefits and costs as evaluated by the patient should be considered because it is not reasonable for the patient to pursue any activity with costs greater than benefits Consideration of benefits and costs should occur when making health care as well as other decisions From the point of view of the rational patient, treatment is acceptable only if there is a positive net benefit (which is defined as benefits minus costs)

Let us pause at this point in our presentation to clarify our use of the terms rational, costs, and benefits We consider each from the perspective of the patient We use the term rational to

indicate that a person optimizes a stable and consistent set of preferences This is standard usage

in economics and also in much of philosophy For example, John Rawls uses rational in this way.4

Typically, we use the term reasonable in an equivalent way.

Costs and benefits are measured in monetary units.5 Their measurement provides the valuation that the patient places on the costs associated with treatment, such as trips to the physician’s office, billed charges, dietary restraints, exercise time, costs of medication, side effects, and the benefits from avoidance of symptoms Benefits are associated with reduction of damage done by the disease as well as other factors such as the avoidance of future medical costs.6

We can measure either the total value of benefits and costs or the marginal value The total benefits are all of the benefits received by pursuing an activity at its present level compared to the benefits received by pursuing an activity at its initial or reference level For example, total

benefits of treating the patient described above to the point that the blood glucose level falls from

G o to G I is the total value of the change in the patient’s health status that results from the blood glucose reduction Marginal benefits are the additional benefits received from a one unit change

in the level of the activity whose benefits are being measured For example, marginal benefits might be measured for a reduction in a patient’s blood glucose level from 183 mg/dl to 182 mg/dl Measurement of total costs and marginal costs is analogous to measurement of the corresponding benefits concept

We return to answering the two questions we have posed First, would a goal of treatment other

than G I be more appropriate? Treatment to lower the patient’s blood glucose level to G I from a

slightly higher level is likely to result in minimal additional benefits to the patient, i.e., little

avoidance of physical damage, and will have significant costs At or very near the ideal blood glucose level, determination of which does not take costs into account, it is expected that virtuallyall intensified treatment would encounter a net loss, i.e., benefits from intensified treatment would

be less than the costs Consequently, positive or non-negative total net benefit occurs at a higher blood glucose level than the ideal blood glucose level A rational, fully informed patient would not accept intensified treatment at or very near the ideal diagnostic level

Perhaps reference to the graph in Figure 1 helps clarify the result In Figure 1, the line

[Insert Figure 1 about here]

labeled MC shows the marginal cost of resources used to reduce the patient’s blood glucose level, measured from the perspective of the patient Previously we listed examples of elements of this

marginal resource cost How is one to interpret the points on MC? Take for example point Q, the point on MC corresponding to the blood glucose level, G I The vertical distance at G I between Q

and the horizontal axis represents the resource cost of reducing the patient’s blood glucose level

by 1 mg/dl, given a blood glucose level of G I The shape of MC indicates that starting at G o, the

resource cost of reducing the blood glucose level by 1 mg/dl increases as the patient’s blood

glucose level decreases from G o to G I

Further in Figure 1, the line labeled MB shows the marginal benefit of a reduction in the

patient’s blood glucose level measured from the perspective of the patient As previously

mentioned, marginal benefit is the valuation of the reduction of damage done by the disease as well as other factors Interpretation of the points on MB is analogous to interpretation of those on

MC Take for example point P, the point on MB corresponding to the blood glucose level, G n The

vertical distance at G n between P and the horizontal axis represents the benefit from reducing the patient’s blood glucose level by 1 mg/dl, given a blood glucose level of G n The shape of MB

indicates that starting at G o, the benefit from reducing the blood glucose level by 1 mg/dl

decreases as the patient’s blood glucose level decreases from G o to G I

In Figure 1 and other figures following, linear functions are used to depict relationships between

MC and blood glucose level and between MB and blood glucose level Complex curves, convex

or concave to the horizontal axis, may be more accurate representations of the relationships But more complex functions would not change the results of our analysis as long as MC is negatively sloped and MB is positively sloped

With this explanation of the graph in Figure 1 in mind, let us use it to illustrate the result that a rational, fully informed patient would not accept intensified treatment at or very near the ideal diagnostic level

Assume that the patient is presently receiving treatment that results in a sustained level of blood

glucose of G n Would the patient receive a positive net benefit from intensified treatment to

further reduce the blood glucose level? According to Figure 1, this patient has MC greater than

MB at G n With this relationship between MC and MB at G n, the patient would actually receive

negative net benefit from the intensified treatment, i.e., the patient is better off at G n than at G I The preceding analysis is represented graphically in Figure 1 in which the blood glucose levels

G I and G o are compared Moving from the original blood glucose level, G o , to G * provides a net

benefit represented by the sum of all net gains (MB-MC at each blood glucose level) along the way The net benefit from this change in blood glucose level is captured by the area of the triangle

JKL Moving from G * to the ideal level, G I , involves a net loss This loss is represented by the

area of the triangle LQR The area of the triangle LQR is greater than the area of the triangle JKL Thus G I is shown to be a blood glucose level where the total benefits of treatment are less than

total costs of treatment for this person with an initial blood glucose level of G 0

The result that a rational, fully informed patient would not accept intensified treatment at, or very near, the ideal diagnostic level occurs when the appropriate treatment level is viewed from the patient’s perspective Viewed from a medical perspective, treating to the ideal level, a level that virtually every rational patient would reject, is also unreasonable After all, treatment of a disease is for the benefit of patients Furthermore, an unreasonable treatment goal represents an irresponsible use of resources

We have concluded that G I is not the proper level of blood glucose to use as the goal in treating diabetes Then, what is the proper level? Finding the answer to this question requires additional analysis

We denote by G n the minimum level of blood glucose to which diabetes should be treated

Economic theory provides a conceptual basis for determining G n Treatment to any specific level

of blood glucose is unreasonable when, from the starting level of G o, total net benefit from

treatment is less than zero; no treatment of diabetes should be made to a lower level of blood glucose When the total net benefit is greater than zero, treatment would result in greater gains

than losses With total net benefit greater than zero, an even lower level of blood glucose could beconsidered as a goal of treatment At any blood glucose level above the one where total net

benefit of treatment equals zero, the same analysis holds: the treatment goal can be reduced with

a positive total net benefit The stopping point occurs where total net benefit just equals zero

This stopping point is the minimum blood glucose level, G n, which should be used as the goal in treating diabetes.7 G n is the point at which resources are not wasted in the sense that treating to

G n from the starting level of G o involves no net loss

The blood glucose level G n might be relatively close to G I, and so physical damage to the patientfrom diabetes is low From the physician’s medical perspective in this case, setting the goal of

treatment at G n should be unobjectionable due to good medical results and the absence of net loss

to the patient In other cases, the treatment level, G n , might exceed the ideal or diagnosis level, G I,

by a significant amount so that there would be poorer medical results although the net loss from treatment is zero to the patient

The blood glucose levels G I and G n are compared graphically in Figure 1 The triangle LOP has

an area equal to that of triangle JKL Thus G n is a blood glucose level where the total benefits of treatment are equal to the total costs of treatment for this patient with an initial blood glucose

level of G 0 Furthermore comparison of G I and G n shows that G n is a higher blood glucose level

than the ideal level, G I , the level of blood glucose without physical symptoms Arguably, G n is thelowest blood glucose level which should be selected as the goal of treatment for this patient

Is G n the blood glucose level that a rational patient would accept as the target level for

treatment?8 The answer is “No.” Accepting G n means that a patient is not maximizing total net

benefit Recall that G n is the point at which there is no net benefit to the patient compared to G o From the patient’s perspective, pursuing treatment until net benefit equals zero may be

unreasonable because a treatment goal at a higher blood glucose level may lower all evaluated costs so that damage from the additional physical symptoms would be acceptable Instead of choosing zero net benefit, the rational patient will seek to maximize total net benefit.How do we identify the blood glucose level at which total net benefit is maximized? The

patient-answer to this question derives from standard economic analysis.9 The benchmark is this:

maximize total net benefit by pursuing an activity until marginal benefit (MB) equals marginal cost (MC) From the patient’s perspective treatment should be pursued to the point that MB from treatment of diabetes just equals the MC of treatment

To appreciate the implication of equality between MB and MC, consider a blood glucose level that departs from the blood glucose level at which MB equals MC Suppose that MB is greater than MC at the current blood glucose level but that a patient has already made health gains due to treatment Pursuing treatment further, decreasing blood glucose level by one additional unit,

means that additional net gains will be made because the marginal (or added) benefit will exceed

the marginal (or added) cost Remember that treatment activity at the margin means that

increased treatment has resulted in an additional unit of decrease in the blood glucose level Whenever MB is greater than MC at the current blood glucose level, it would be reasonable to pursue further treatment because the next unit of reduction provides gains greater than the

additional cost This net gain is added to net gains already made Because this is from the

patient’s valuation, it is rational for the patient to continue treatment to this lower blood glucose

level The same analysis occurs at every blood glucose level where MB is greater than MC, i.e., at

all such levels, treatment to lower the patient’s blood glucose level should be intensified From the patient’s perspective a blood glucose level is too high if MB is greater than MC Therefore it isnot rational, at any point at which MB is greater than MC, for a patient to fail to intensify

treatment

Now suppose that at the current blood glucose level MB is less than MC It is unreasonable to pursue the treatment intensity that led to this blood glucose level because marginal cost to raise the treatment intensity to this level is greater than marginal benefit It is clearly unreasonable, from a patient’s perspective, to continue intensifying treatment to lower blood glucose level at anylevel at which MB is less than MC

The alternative that remains is the blood glucose level at which MB equals MC This is the

blood glucose level that we denote as the goal for treatment and symbolize by G * A patient with

an original blood glucose level greater than G * should continue to intensify treatment until G * is reached, and it would be unreasonable for a patient to accept intensified treatment to achieve a

blood glucose level lower than G *

In terms of a comparison of G * to G n and G I, the result is that the optimal blood glucose level for

the patient, G *, is greater than the minimum level which could be selected as a goal of treatment,

G n , and greater still than the medically ideal level, G I The typical patient would be unreasonable

to accept a treatment goal leading to a blood glucose level of G n or G I A rational and fully

informed patient will reasonably accept significant physical damage from a blood glucose level that is higher than the medically ideal level.10

It is further demonstrated graphically in Figure 1 that G * is the optimal level of blood glucose

for this patient At G * , MB equals MC as indicated by point L At values of G lower than G * , the

marginal cost (MC) of intensifying treatment exceeds its marginal benefit (MB) Any movement

to a blood glucose level lower than G * produces a marginal net loss and is therefore unreasonable

to pursue Accepting a level of blood glucose higher than G * is also unreasonable because

foregoing the intensified treatment which yields G * would produce marginal benefit (MB) that

exceeds the marginal cost (MC) of the intensified treatment.

II

We turn to several complexities involving the blood glucose level corresponding to G * The first

of these complexities is that the level of blood glucose serving as the target for treatment may varyfrom person to person In Figure 2 we represent the relevant MC and MB relationships for two

[Insert Figure 2 about here]

persons: Samuel and Samantha We center on the results that occur due to their differing MC To

do this, stipulate that they have the same MB relationship, but Samantha’s MC is higher than

Samuel’s MC at any blood glucose level other than the initial level, G o This difference in their

MC relationships means that Samantha considers it more costly than Samuel does to achieve what

is required to reduce her blood glucose level For example, she might find exercise a greater burden than does Samuel Recall that the treatment target of blood glucose is the level at which

MC equals MB In Figure 2 Samuel’s treatment target is shown as G *, and Samantha’s treatment

target is G * ′ Although Samuel and Samantha have the same initial blood glucose level and the same MB relationship, Samantha’s greater MC results in her choosing a higher treatment target and thereby less treatment than Samuel chooses If they both achieve their treatment targets, Samantha is left with a higher blood glucose level than Samuel

Secondly, G * depends on a patient’s original glucose level, G o With a lower G o , we expect G * to

be at a lower level as well Consider Figure 3 The line labeled MC′ represents a different

marginal [Insert Figure 3 about here]

cost curve with an original glucose level at G o ′ The curve MC′ is below MC based on G o because

at any blood glucose level reducing physical damage is less costly, and so we expect rational

treatment to continue to a blood glucose level of G * ′ This is a lower blood glucose level than G *,

which is located using the higher G o

A further complication exists: A third party may pay part or all of the billed medical treatmentcosts Such costs are part of the MC curve Examine Figure 4 When the patient assumes less

than [Insert Figure 4 about here]

the entire cost, the MC curve of the patient is P Given this, preventing damage has a reduced marginal cost from the patient’s individual perspective The result is a lower target glucose level,

G * ′ Even if a third party is responsible for all of billed medical treatment costs, significant

marginal cost is likely to remain for the patient in the form of treatment costs such as time and effort, out-of-pocket expenses, side effects, dietary restrictions, and so on

A fourth complexity involves determination of the minimum level of blood glucose at which to diagnose diabetes Based on our economic theory of patient decision-making, there is no way to determine a single level of blood glucose that should be used as the goal of treatment for diabetes

in all patients It might be argued that the minimum blood glucose level at which diabetes should

be diagnosed is G n, or the level of blood glucose at which net benefit of treatment is zero

However, like G * , G n may vary from person to person based at least partly on the possibility that

G o may vary from person to person If G n is accepted as the blood glucose level used to diagnose

diabetes and given possible variation in G n, then this blood glucose level should be determined by

careful use of empirical information If such determination proves impossible, then perhaps G I, the level of blood glucose considered ideal for health, is the best option

Whether G n or G I is chosen as the level at which diabetes is diagnosed, using either as the target

level for treatment is unreasonable The target level for treatment, G *, is reasonably set by

carefully considering costs and benefits from the perspective of individual patients, based on solid

and reliable information provided mainly by physicians, in terms of physical damage and

significant associated costs

III

An additional complication is that the professional calling of physicians is to eliminate

symptoms One hazard of this calling is that physicians may wittingly or unwittingly exaggerate physical damage or understate or hide costs For example, a physician may describe an atypical patient who suffered more damage than would be expected Or the physician may fail to mention side effects that are merely considered nuisances; or else he or she may point to unusual examples

of patients who react more quickly to treatment This is improper From an economic perspective,

the patient is a principal and the physician is an agent Principals rely on the superior knowledge

of agents in order to make better judgments An agent, such as a physician, who does not provide good information deprives the principal of the opportunity to make better judgments and thus fails

in his or her fiduciary responsibilities

The effect of physician misinformation is that from the patient’s perspective MC and MB could both differ from what they would have been with better information Figure 5 represents the effect

[Insert Figure 5 about here]

of misinformation about MB MBA represents the actual marginal benefit from treatment of the patient and MBP is the marginal benefit from the patient’s perspective based on exaggerated

information about damage from the disease The effect is that the rational but misinformed

patient/principal accepts a lower blood glucose level, G * ′, as the target for treatment.

Denying the patient good information about costs involves a different MC curve This situation

is depicted in Figure 6 by plotting MCA (with good information) and MCp (with information

[Insert Figure 6 about here]

understating cost) Again the effect is that the patient accepts a lower blood glucose level, G * ′, as

the target for treatment Of course, if the physician understated MB or overstated MC, the patient

accepts a higher G * as the target for treatment

IVOur presentation offers a theoretical account of the reasons why an informed consent process ought to include accurate information about costs and expected physical damage It also gives a theoretical basis behind respect for patient decision-making Typically, bioethicists claim that informed consent represents respect for a patient’s decision whenever the patient is decisionally competent Decisional capacity is a minimal requirement partly because it explicitly rejects any special deliberating skills From this point of view, it is easy to think of patient decisions as conflicting with a patient’s best interest However, informed consent is required even when such irrationality is involved This puts stress on health care professionals and bioethicists By

ignoring a theory behind autonomous choices, health care professionals may be tempted to

subvert, one way or another, patient decisions because the decisions are thought to be

unreasonable or harmful when they depart from a medical ideal Our theory of rational making shows that this typically is not the case and that what might appear to be a poorly

decision-conceived decision  because physical damage is accepted  is likely to be a fully reasonable decision Since autonomy is not the same as rationality, an account that supports informed

consent from the point of view of reasonableness, as this economic theory does, strengthens respect for patient decision-making

We are mainly exploring the patient’s perspective on desired treatment Our analysis includes the physician’s perspective only insofar as it identifies the diagnosis level as the point at which a

patient is expected to be symptom free We put this at 126 mg/dl of blood glucose for diabetes diagnosis However, this is a simplifying assumption because physicians might set a diagnostic point at a different blood glucose level that takes costs into account For example, in 1997 the

“Expert Committee” explained that the diagnosis level was set without considering costs due to inadequate data:

Determining the optimal diagnostic level of hyperglycemia depends on

a balance between the medical, social, and economic costs of making

a diagnosis in someone who is not truly at substantial risk of the adverse

effects of diabetes and those of failing to diagnose someone who is

Unfortunately, not all of these data are available, so we relied primarily

on medical data.11

Our analysis indicates that taking such non-medical cost considerations into account tends to be

in harmony with the preferences of the typical patient However, we also pointed out that

marginal curves do depend on patient evaluation of total cost, and that this is unknown in

advance While in general setting a diagnostic level partly on the basis of cost is consistent with patient-based allocation of resources, it might not be in individual situations

V

In this section we examine and reply to possible objections that practitioners and bioethicists might have to an economic analysis of patient decision-making These objections tend to focus onand intermingle three central issues: (1) An analysis of this sort is not realistic because patients

do not have well-formed cost and benefit curves, (2) Cost is not a crucially important

consideration when health is at stake, and (3) Such an analysis runs contrary to a physician’s highest moral standard, not to harm

We have relied on economic analysis to explain a patient’s rational decision-making in the informed consent process This perspective shows how a fully reasonable patient would take into

account costs and benefits To do this we use marginal curves commonly employed in

microeconomics, but the non-economist may be skeptical The approach we take presents

seemingly precise information about marginal damage and marginal cost, and purports to show how individuals use such information to make decisions about treatment related to their blood glucose level, or other health-status variables The objection might be that patients do not have such precise information and even if they did, they would not know how to apply it in a way that would equalize marginal benefit and marginal cost

However this may appear, it is a standard technique used by economists and is applied to

decision-making concerning virtually all goods Even though frequently used in economic

analysis, economists do not believe that such calculations are always actually made Instead they

believe that a reasonable person acts at least as if he or she calculated benefits and costs using

marginal curves This is thought to be analogous to the way we might predict the shots of an expert billiard player No one would claim that the billiard player used complex mathematical calculations, but those calculations could be used to predict and explain his or her expert shots The billiard player acts as if he or she did apply such calculations Furthermore, the more a player’s actions depart from the results of such calculations, the more we know that the player is not expert.12 Similarly, such graphs can predict a rational patient’s perspective on treatment; the more a patient departs from this analysis, the more we can suspect that the patient was not

provided good information or that the patient is decisionally incapacitated

Microeconomic analysis is based on near truisms As cost decreases, we expect people to

purchase more of a good or to pursue more of an activity As benefit declines, we expect people topurchase less It makes intuitive sense that as treatment brings a patient closer to ideal conditions,without physical symptoms, one would expect that additional benefits decrease and that additional