Professional Caregiver Insurance Risk, Medical Outliers, and the Duty to Treat

Thomas Cox, Seton Hall University - College of Nursing

10 June 2005

To fully understand and respond to the ethical issues raised by high cost clients (Medical outliers) we must consider the relatively incomplete specification of the impact of prospective payment systems, DRGs, and capitation contracts. All these mechanisms involve an element of "risk/profit" sharing for providers and this point is well acknowledged by the authors. However, the risk/profit sharing has another name as well. When one entity assumes financial risks from another entity in return for a payment that is, on average, approximately equal to the expected costs of those financial risks, the transaction is usually called “Insurance.” I use the term “Professional Caregiver Insurance Risk” for the insurance risks assumed by physicians, hospitals, and other health care providers when they engage in prospective payment agreements. While on first glance this may appear to be a relatively benign transaction, encouraging efficient health services production, and controlling costs at the point where cost control is most feasible, when a sincere provider distinguishes between useful and not useful treatments, it is far from benign in practice. As the authors note there are ethical concerns when some clients' costs far exceed the average costs anticipated in prospective payment plans. Furthermore, the authors highlight the fact that health care providers rarely comprehend their insurance risk assumption and management roles.

The central problem of Professional Caregiver Insurance Risk is that it places health care providers in the role of ‘mini-insurance companies.’ There are two extremely important consequences of this that bear on the arguments advanced by the authors. First, providers are inefficient insurers, and second, the duties to patients change dramatically when health care providers assume insurance risks.

If we assume, because of the relatively large numbers of patients involved at the insurance company and provider levels, that the normal distribution can be used to model average claim costs, we can expose and detail the problem of provider inefficiency qua insurer. Assume that a health care provider accepts responsibility, through some form of prospective payment, for a cohort of 1,000 patients. The insurer, which could include a governmental entity such as Medicare or the National Health Service, or a private insurer or managed care company, has far more patients, say 1,000,000. The insurer very accurately predicts its average costs for insuring all 1,000,000 patients as 85% of its premium income. The reason it can do this is due to two statistical or probabilistic 'Laws' or 'Theorems': The Law of Large Numbers and the Central Limit Theorem. As the number of policyholders rises, the accuracy of the insurer's estimate of the average cost per policyholder rises as well. It is not a completely efficient process, the rise in accuracy is related to the square root of the increases in sample size.

These characteristics bode well for the insurer. However, the problem of insurance 'efficiency' is that this process of increasing accuracy of average cost estimates, occurs in reverse when small portfolios of patients are transferred to health care providers. As small cohorts of patients are transferred to health care providers, the efficiency and the benefit of the insurance mechanism is lost. Using the normal distribution we can estimate the impact of the greater variability in average costs at the level of the providers, that is solely due to the decrease in the number of patients in the portfolio. From the provider perspective, we might think in terms of their ability to make a profit on their entire portfolio of 1,000 patients. To do that, the provider needs to incur total business and treatment costs that are less than 100% of their guaranteed contract payments.

Clearly, the insurer cannot pay out, on average, more than their best estimate of the losses they expect to incur, the balance of their premium income will, of course, be devoted to operating expenses and/or profits. Assume that the insurer expects an 85% loss ratio and that the standard error in their estimate of that loss ratio, based on 1,000,000 patients, drawn at random, from the pool of all possible patients, is 5%. The probability that the insurer's losses exceed 90% is about 0.1587. The probability that the insurer's losses exceed 95% is about 0.0228 and the probability that the insurer's losses exceed 100% is about 0.0014. That is, of course, exactly what one would expect. Large insurers serve a public good by aggregating exposure to risk and managing it far more efficiently than any individual, or smaller insurer, can manage risk.

Turning our attention to the provider however, the picture is not so cheerful. Due solely to the smaller portfolio size at the provider level (1,000 v 1,000,000 patients), the standard error for the sample of 1,000 patients drawn, at random, from the pool of all possible patients, is 31.7 times the standard error for the insurer (31.7 = square root of 1,000). The 85% of the insurer's premiums for those 1,000 patients, is the maximum income for the provider since it is the expected loss for the insurer. An insurer that pays all its providers more than 85% of the premium income generated for their clients will soon be either bankrupt or uncompetitive when compared with insurers who do not pay such excessive amounts to their providers. What is the probability that the provider experiences a loss ratio that exceeds the 85% maximum available from the insurer? Remember though, the 85% is the maximum available to the provider, so it is 100% of the provider's income from this insurance risk transferring contract. We will, however, continue to discuss the provider's losses in terms of the average loss ratio for the insurer. The reader is cautioned to remember that a loss ratio of 90% in insurer terms, really means a loss ratio of 105.9% (105.9 = 90/85*100%) for the provider.

While the chance of a 90% loss ratio for the insurer is only about 15%, the provider, due to the much larger standard error of its portfolio, has a much greater probability that it will experience a loss ratio of 90% or more, about 49% (See Table below). But first, notice that as is true for the insurer, the provider has a probability of a loss greater than the expected amount of 0.50. This is the only place that the provider and insurer have the same probability of an average loss (in terms of insurer loss ratios). If we compare the probability of an average loss of size LR for both the insurer and the provider, this is what we have, presented in vectors of size '3' form because of the formatting problems here):

Each vector has a: (loss ratio; the insurer's probability of a loss ratio that high or larger; and the provider's probability of a loss ratio that high or larger):

(LR; Insurer P; Provider P) =

( 85; 0.5000; 0.5000)

( 90; 0.1587; 0.4874

( 95; 0.0228; 0.4748

(100; 0.0014; 0.4623

(105; 0.0000; 0.4498

(110; 0.0000; 0.4373)

Clearly, this is an extreme example to make a point. But the general pattern does not change. Providers are terribly inefficient insurers when compared with real insurers. At every loss ratio higher than 85% the provider has a much greater probability of a loss that high or higher when compared with the original insurer. While the insurer has a 16% chance of a loss ratio higher than 90%, the provider has a 49% chance of a loss ratio that high or higher. The probability that the insurer experiences a loss ratio of 100% of premium is a mere 0.0014, while the probability of a provider loss ratio to the insurer's premiums of 100% is 0.46. Providers, due to portfolio size alone, have a far more difficult time dealing with variation in costs than do large insurers. While the portfolio size differentials may not be this extreme, they could be even more extreme. The standard error of the provider's portfolio will always be greater than that of the insurer as long as the number of patients the insurer insures exceeds the number of patients the provider insures. In particular, when the provider is a small community hospital or local practitioner, all other things held equal, the size differential of the standard errors may greatly exceed the example above, further exacerbating the ability of smaller, more local providers to manage these insurance risks. It is inescapable that in an efficient health care system, when all operating processes have been made as perfect as they can, where patients require specific treatments and the variation in costs are due to unique characteristics of patients and providers that cannot be further refined: such as experience; complicating conditions; local resources and available technologies, the government entities operating results will be far more accurately predicted than the operating results of the providers. Providers simply cannot manage the highly predictable variability in operating costs as well as the larger government entities or insurers can. That, to be clear, is why we encourage insurance companies to operate by furnishing legal and financial incentives to them to manage risk.

All of this, of course, assumes that the patients are selected at random from the insurer's patient portfolio. In practice this is highly unlikely, leading to additional risk management problems when statistical independence of patient selections cannot be assumed to apply. If a local provider maintains their office close to a small factory, their risks that a conflagration: illness, toxic release, or an industrial accident affecting those employees, present much greater problems than for providers who maintain offices and draw patients from many different employers. Likewise, providers that are located close to areas of poverty, or areas with unusually high numbers of physically and mentally ill residents, also may be harmed by non-random selection patterns that significantly shift the average loss characteristics. While geographic corrections may be made for the average, it is impossible to compensate providers for the risks due to sample size differences.

The consequence of this is that health care providers are far too inefficient insurance risk managers to be placed in the position of mini-insurers. Of course, there is another side to this. Based purely on portfolio size, the provider also has a much greater probability of low loss ratios than is true for the insurer. The problem is that once a provider has a high loss ratio, they cannot offset it by the possibility of a lower loss ratio at some future time. Providers have to pay their bills to remain in practice. Some providers will benefit greatly when their patients make few demands. Other providers will suffer greatly because their patients make excessive service demands. On average, all providers will be expected to have a combined loss ratio equal to 85% of the insurer's total premium income. Some providers, those who are financially weakest, have a lower portion of high-paying and low cost patients, or who have fewer other resources to sustain operations, may have to declare bankruptcy or accept affiliation with stronger providers, before they ever have another contract. Of greater concern from the original authors' standpoint, such financially compromised providers may violate their obvious ethical duties to their patients.

This brings us to the second problem of insurance risk assumption by health care providers. Having entered the insurance business, how does the ethical duty of a provider to the 'medical outlier' patient change? This is no longer a question of beneficence, good will, or duty to steward a patient through a difficult period. The provider is now the patient's 'Insurer of record.' Insurers should anticipate that they will occasionally have large claims and they ought to be prepared to deal with them when it happens. To meet their legal and ethical responsibilities, insurers maintain liquid reserves and operating capital that are sufficient to meet high, though infrequent operating losses. Insurers are regulated by the government and they are obligated to maintain funds in secure assets so that they can meet their obligations even through several years of poor operating results. Health care providers rarely, if ever, meet these standards, despite the fact that their exposure to extreme losses is far greater due to the loss of risk aggregation benefits that true insurers enjoy. Above all else, insurers have a clear legal as well as an ethical duty to honor their insurance contracts with their patients. When health care providers engage in prospective payment agreements they become 'Insurers' and their obligations to 'medical outliers' are no longer solely ethical or medical, they are the same obligations that insurers have to absorb the financial risks of rare, though high cost, claimants with equanimity. Providers that rationalize failing to meet the needs of 'medical outliers' on the basis that their needs are too great and out of the range of that expected, simply do not understand the agreements they have made to meet the needs of all claimants, not just the ones whose costs of care are relatively small.

Clearly providers ought not be in the insurance business in the first place, cannot perform as efficiently as insurers, and every health insurance dollar is inefficiently consumed when health care providers serve dual roles as providers and as insurers. Eliminating providers’ point of service insurance risk assumption and management roles would eliminate a costly and unnecessary inefficiency in financing health care services. More to the point in the current case, eliminating insurance risk assumption and management by health care providers would eliminate an ethical and legal conflict that exists not at all in the absence of these contracts.

In any event, unless, in the aggregate, the provider maintains a portfolio size equal to that of the insurer, the efficiency of the insurance mechanism as well as the ethical considerations of the situation are irreparably compromised when health care providers accept insurance risks.

## Professional Caregiver Insurance Risk, Medical Outliers, and the Duty to Treat

Thomas Cox, Seton Hall University - College of Nursing

10 June 2005

To fully understand and respond to the ethical issues raised by high cost clients (Medical outliers) we must consider the relatively incomplete specification of the impact of prospective payment systems, DRGs, and capitation contracts. All these mechanisms involve an element of "risk/profit" sharing for providers and this point is well acknowledged by the authors. However, the risk/profit sharing has another name as well. When one entity assumes financial risks from another entity in return for a payment that is, on average, approximately equal to the expected costs of those financial risks, the transaction is usually called “Insurance.” I use the term “Professional Caregiver Insurance Risk” for the insurance risks assumed by physicians, hospitals, and other health care providers when they engage in prospective payment agreements. While on first glance this may appear to be a relatively benign transaction, encouraging efficient health services production, and controlling costs at the point where cost control is most feasible, when a sincere provider distinguishes between useful and not useful treatments, it is far from benign in practice. As the authors note there are ethical concerns when some clients' costs far exceed the average costs anticipated in prospective payment plans. Furthermore, the authors highlight the fact that health care providers rarely comprehend their insurance risk assumption and management roles.

The central problem of Professional Caregiver Insurance Risk is that it places health care providers in the role of ‘mini-insurance companies.’ There are two extremely important consequences of this that bear on the arguments advanced by the authors. First, providers are inefficient insurers, and second, the duties to patients change dramatically when health care providers assume insurance risks.

If we assume, because of the relatively large numbers of patients involved at the insurance company and provider levels, that the normal distribution can be used to model average claim costs, we can expose and detail the problem of provider inefficiency

insurer. Assume that a health care provider accepts responsibility, through some form of prospective payment, for a cohort of 1,000 patients. The insurer, which could include a governmental entity such as Medicare or the National Health Service, or a private insurer or managed care company, has far more patients, say 1,000,000. The insurer very accurately predicts its average costs for insuring all 1,000,000 patients as 85% of its premium income. The reason it can do this is due to two statistical or probabilistic 'Laws' or 'Theorems': The Law of Large Numbers and the Central Limit Theorem. As the number of policyholders rises, the accuracy of the insurer's estimate of the average cost per policyholder rises as well. It is not a completely efficient process, the rise in accuracy is related to the square root of the increases in sample size.quaThese characteristics bode well for the insurer. However, the problem of insurance 'efficiency' is that this process of increasing accuracy of average cost estimates, occurs in reverse when small portfolios of patients are transferred to health care providers. As small cohorts of patients are transferred to health care providers, the efficiency and the benefit of the insurance mechanism is lost. Using the normal distribution we can estimate the impact of the greater variability in average costs at the level of the providers, that is solely due to the decrease in the number of patients in the portfolio. From the provider perspective, we might think in terms of their ability to make a profit on their entire portfolio of 1,000 patients. To do that, the provider needs to incur total business and treatment costs that are less than 100% of their guaranteed contract payments.

Clearly, the insurer cannot pay out, on average, more than their best estimate of the losses they expect to incur, the balance of their premium income will, of course, be devoted to operating expenses and/or profits. Assume that the insurer expects an 85% loss ratio and that the standard error in their estimate of that loss ratio, based on 1,000,000 patients, drawn at random, from the pool of all possible patients, is 5%. The probability that the insurer's losses exceed 90% is about 0.1587. The probability that the insurer's losses exceed 95% is about 0.0228 and the probability that the insurer's losses exceed 100% is about 0.0014. That is, of course, exactly what one would expect. Large insurers serve a public good by aggregating exposure to risk and managing it far more efficiently than any individual, or smaller insurer, can manage risk.

Turning our attention to the provider however, the picture is not so cheerful. Due solely to the smaller portfolio size at the provider level (1,000 v 1,000,000 patients), the standard error for the sample of 1,000 patients drawn, at random, from the pool of all possible patients, is 31.7 times the standard error for the insurer (31.7 = square root of 1,000). The 85% of the insurer's premiums for those 1,000 patients, is the maximum income for the provider since it is the expected loss for the insurer. An insurer that pays all its providers more than 85% of the premium income generated for their clients will soon be either bankrupt or uncompetitive when compared with insurers who do not pay such excessive amounts to their providers. What is the probability that the provider experiences a loss ratio that exceeds the 85% maximum available from the insurer? Remember though, the 85% is the maximum available to the provider, so it is 100% of the provider's income from this insurance risk transferring contract. We will, however, continue to discuss the provider's losses in terms of the average loss ratio for the insurer. The reader is cautioned to remember that a loss ratio of 90% in insurer terms, really means a loss ratio of 105.9% (105.9 = 90/85*100%) for the provider.

While the chance of a 90% loss ratio for the insurer is only about 15%, the provider, due to the much larger standard error of its portfolio, has a much greater probability that it will experience a loss ratio of 90% or more, about 49% (See Table below). But first, notice that as is true for the insurer, the provider has a probability of a loss greater than the expected amount of 0.50. This is the only place that the provider and insurer have the same probability of an average loss (in terms of insurer loss ratios). If we compare the probability of an average loss of size LR for both the insurer and the provider, this is what we have, presented in vectors of size '3' form because of the formatting problems here):

Each vector has a: (loss ratio; the insurer's probability of a loss ratio that high or larger; and the provider's probability of a loss ratio that high or larger):

(LR; Insurer P; Provider P) =

( 85; 0.5000; 0.5000)

( 90; 0.1587; 0.4874

( 95; 0.0228; 0.4748

(100; 0.0014; 0.4623

(105; 0.0000; 0.4498

(110; 0.0000; 0.4373)

Clearly, this is an extreme example to make a point. But the general pattern does not change. Providers are terribly inefficient insurers when compared with real insurers. At every loss ratio higher than 85% the provider has a much greater probability of a loss that high or higher when compared with the original insurer. While the insurer has a 16% chance of a loss ratio higher than 90%, the provider has a 49% chance of a loss ratio that high or higher. The probability that the insurer experiences a loss ratio of 100% of premium is a mere 0.0014, while the probability of a provider loss ratio to the insurer's premiums of 100% is 0.46. Providers, due to portfolio size alone, have a far more difficult time dealing with variation in costs than do large insurers. While the portfolio size differentials may not be this extreme, they could be even more extreme. The standard error of the provider's portfolio will always be greater than that of the insurer as long as the number of patients the insurer insures exceeds the number of patients the provider insures. In particular, when the provider is a small community hospital or local practitioner, all other things held equal, the size differential of the standard errors may greatly exceed the example above, further exacerbating the ability of smaller, more local providers to manage these insurance risks. It is inescapable that in an efficient health care system, when all operating processes have been made as perfect as they can, where patients require specific treatments and the variation in costs are due to unique characteristics of patients and providers that cannot be further refined: such as experience; complicating conditions; local resources and available technologies, the government entities operating results will be far more accurately predicted than the operating results of the providers. Providers simply cannot manage the highly predictable variability in operating costs as well as the larger government entities or insurers can. That, to be clear, is why we encourage insurance companies to operate by furnishing legal and financial incentives to them to manage risk.

All of this, of course, assumes that the patients are selected at random from the insurer's patient portfolio. In practice this is highly unlikely, leading to additional risk management problems when statistical independence of patient selections cannot be assumed to apply. If a local provider maintains their office close to a small factory, their risks that a conflagration: illness, toxic release, or an industrial accident affecting those employees, present much greater problems than for providers who maintain offices and draw patients from many different employers. Likewise, providers that are located close to areas of poverty, or areas with unusually high numbers of physically and mentally ill residents, also may be harmed by non-random selection patterns that significantly shift the average loss characteristics. While geographic corrections may be made for the average, it is impossible to compensate providers for the risks due to sample size differences.

The consequence of this is that health care providers are far too inefficient insurance risk managers to be placed in the position of mini-insurers. Of course, there is another side to this. Based purely on portfolio size, the provider also has a much greater probability of low loss ratios than is true for the insurer. The problem is that once a provider has a high loss ratio, they cannot offset it by the possibility of a lower loss ratio at some future time. Providers have to pay their bills to remain in practice. Some providers will benefit greatly when their patients make few demands. Other providers will suffer greatly because their patients make excessive service demands. On average, all providers will be expected to have a combined loss ratio equal to 85% of the insurer's total premium income. Some providers, those who are financially weakest, have a lower portion of high-paying and low cost patients, or who have fewer other resources to sustain operations, may have to declare bankruptcy or accept affiliation with stronger providers, before they ever have another contract. Of greater concern from the original authors' standpoint, such financially compromised providers may violate their obvious ethical duties to their patients.

This brings us to the second problem of insurance risk assumption by health care providers. Having entered the insurance business, how does the ethical duty of a provider to the 'medical outlier' patient change? This is no longer a question of beneficence, good will, or duty to steward a patient through a difficult period. The provider is now the patient's 'Insurer of record.' Insurers should anticipate that they will occasionally have large claims and they ought to be prepared to deal with them when it happens. To meet their legal and ethical responsibilities, insurers maintain liquid reserves and operating capital that are sufficient to meet high, though infrequent operating losses. Insurers are regulated by the government and they are obligated to maintain funds in secure assets so that they can meet their obligations even through several years of poor operating results. Health care providers rarely, if ever, meet these standards, despite the fact that their exposure to extreme losses is far greater due to the loss of risk aggregation benefits that true insurers enjoy. Above all else, insurers have a clear legal as well as an ethical duty to honor their insurance contracts with their patients. When health care providers engage in prospective payment agreements they become 'Insurers' and their obligations to 'medical outliers' are no longer solely ethical or medical, they are the same obligations that insurers have to absorb the financial risks of rare, though high cost, claimants with equanimity. Providers that rationalize failing to meet the needs of 'medical outliers' on the basis that their needs are too great and out of the range of that expected, simply do not understand the agreements they have made to meet the needs of all claimants, not just the ones whose costs of care are relatively small.

Clearly providers ought not be in the insurance business in the first place, cannot perform as efficiently as insurers, and every health insurance dollar is inefficiently consumed when health care providers serve dual roles as providers and as insurers. Eliminating providers’ point of service insurance risk assumption and management roles would eliminate a costly and unnecessary inefficiency in financing health care services. More to the point in the current case, eliminating insurance risk assumption and management by health care providers would eliminate an ethical and legal conflict that exists not at all in the absence of these contracts.

In any event, unless, in the aggregate, the provider maintains a portfolio size equal to that of the insurer, the efficiency of the insurance mechanism as well as the ethical considerations of the situation are irreparably compromised when health care providers accept insurance risks.

## Competing interests

None