This Web site is a component of the SAMHSA Health Information Network

This Site Mental Health at HHS All Mental Health

About the Program Mental Health Parity Resources Journal Articles Featured Publications In the News Related Links Organization & Financing Homepage  CMHS Programs:   Go to... Children's Campaign Community Support Emergency Services Financing HIV/AIDS Homelessness Managing Anxiety Mental Health Statistics Protection & Advocacy Special Programs State Planning Traumatic Events Suicide Prevention Programs Homepage  CMHS Activities:   Go to... 15+ School Violence Stigma Suicide Prevention Surgeon General

printer friendly page e-mail this page bookmark this page shopping cart  current or new account

This Web site is a component of the SAMHSA Health Information Network.

Organization & Financing

SAMHSA Managed Care Initiative Training Sessions

May 1997

Date: May 20, 1997
Morning Session: Mental Health Liaison Group, American Psychological Association, Washington, D.C.
Afternoon Session: SAMHSA Group, Parklawn Conference Center, Rockville, Maryland

Title: Managed Care: New Tools for New Challenges

Speaker: Anthony Broskowski, Ph.D., President, Managed Care Solutions, West Orange, New Jersey

Overview

Definition of Risk

Dr. Broskowski began with an explanation of risk. The word derives from the Italian risicare, meaning choice. Risk is composed of three elements-(1) uncertainty of the outcome (this implies a probability feature), (2) opportunity for gain or loss (one that can be accepted or rejected), and (3) choice. Dr. Broskowski acknowledged that most people shy away from risk, thinking it concerns precision, accounting, measurement, and historical data. He stated that precision does not improve the odds and offered a quotation from John Maynard Keynes: "It is better to be approximately correct than precisely wrong."

In Dr. Broskowski's opinion many providers are risk aversive as they do not understand the concepts of risk. Most providers would rather have fee-for-service arrangements and gripe about the level of payment, the rules and regulations, the responsibilities, the clientele, and the regulators. At least they know their misery. He summed up this attitude with words from cartoonist Walt Kelly's Pogo character: "The certainty of misery is better than the misery of uncertainty."

Evolution of Risk in Health Care

Dr. Broskowski explained that there has always been risk in health care, but how this risk is shared has changed over time. Employer insurance was not prevalent before World War II. The person at risk was the patient. When a person got sick, health care was paid for out of pocket. After World War II President Truman and Congress wanted to promote capital investment in new jobs for returning veterans. Employers were not allowed to increase wages, but could offer fringe benefits that would be tax deductible. This was an incentive to offer health insurance as a benefit, and it stimulated the growth of hospitals and health care professions. Tax policy was used to drive social policy.

Starting with the United Auto Workers and heavy industries, employer-sponsored insurance became prevalent in the 1950s. Most of the new generation of employees assumes that health care insurance comes with employment. Insurance companies took risks and charged premiums that covered their risks. If health care costs increased, then next year's premiums increased. This pattern continued as a successful strategy because employers did not switch insurance companies. There was stability and predictability in the employer-insurer relationship.

Now the distribution of risks has shifted. With increasing costs due to inflation, employers began to take some risks. They became self-insured and paid administrative fees to the insurance companies. Now the risk was being shared. A short time later the employees shared in the risk through payroll deductions and co-payments. The co-payment meant that the patient had partial risk; the payroll deduction meant that the eligible person had risk, whether or not they used the care. By the end of the 1970s the risk was shared among insurance companies, employers, patients (or users), and eligibles. Eventually, providers also joined the risk pool.

Probability and Distribution Terms

Dr. Broskowski reviewed a few basic terms and probability distributions. If cost is an outcome, then there is a probability distribution of outcomes; 100 percent of all possible events fall inside a distribution. To understand risk, one has to understand distributions. A normal distribution is symmetrical, such as a symmetrical bell curve. In this situation the mean, the mode, and the median are equal. The mean is the average, or the sum of all values divided by the number of values; the mode is the most common value; and the median divides the distribution into two equal parts, of 50 percent of values above and below the median.

But normal distributions frequently do not reflect health care use or cost; instead, an asymmetrical or right skewed distribution is more common. In such a skewed distribution the average is greater than the mode. In a distribution representing health care costs, many patients would cost just a few hundred dollars, the vast majority would cost between $2000 and $10,000, and very few would cost more than $75,000.

Managed Care Concerns

Variables in Cost

Dr. Broskowski stated that by the 1980s three variables primarily determined total health care costs-(1) the number of people using health care (referred to as access, penetration, or users per 1,000 eligibles); (2) how much care is used (i.e., the utilization or units of care); and (3) the price or what is paid for the care. Total cost can be calculated using the following formula:

Total cost = (users/1000) x (units/user) x (cost/unit)

The cost per member per month (PMPM) is calculated by using the equation:

PMPM = (total costs)/(total members)/12

Dr. Broskowski used these equations to explain that the risk for providers is very different from that for intermediary insurers. For the intermediaries, costs are variable; money is not spent until a service is authorized and used. If utilization goes down, costs go down. The provider, however, has more fixed costs. For example, the staff in a clinic is fixed whether there are 10 patients or 100 patients. If utilization decreases, staff costs do not decrease. Fixed costs remain high with high-capacity elements such as institutes, group homes, hospitals, and other facilities. To decrease risk, fixed costs should be converted to variable costs.

More recently another factor has been introduced, the quality of care. An underlying principle in business can be applied to managed care-if quality goes up, costs come down. In business if quality improves, less time, money, and manpower are spent on recalls or redoing the project. As a result customer satisfaction increases and brings repeat business. Some providers think this theory also applies to health care.

Health status will be added as a fourth factor, Dr. Broskowski suggested, for the next generation of care. Instead of providers enrolling only healthy members and excluding the sick, they will have incentives to make the entire population or community healthier. The emphasis will be on prevention. The shift into the next generation of health care comes with the realization that sick people do not disappear. They show up in emergency rooms, and we all end up paying for the care. Taking care of the entire community in a planned way is smarter. Dr. Broskowski stressed that we need the right incentives to expedite the transition to the next level of heath care. One of those incentives is long-term contracts. In the long run the commitment to improve community health is paid off only by reduced demand for care.

Types of Risk

Dr. Broskowski reminded the audience that risk does not equal cost; risk is uncertainty. He outlined the three types of uncertainty, or risk, in health care costs:

1. The risk of upward variations in the rate of users per 1,000 members in the population of eligible lives, that is, persons becoming patients

   For example, if the planned rate is for 5/1000 over a year, but the actual rate is 7/1000.

2. The risk for upward variations in the volume and types of units of services used by each patient

   For example, the predicted use is 10 days of inpatient care and 12 outpatient visits per user, but actual use is 14 days of inpatient care and 14 outpatient visits.

3. The upward variation in the cost of producing the various units of service

   For example, the cost of therapy is calculated at $65/hour, but with inflation the actual cost is $67/hour.

Dr. Broskowski noted that the key element is variation. He stressed that risk does not necessarily mean high cost; under planned treatment the risky patient is not the high-cost patient. Instead it is unexpected variation that is risky, such as the alcohol and drug abusers who show up in the emergency room and are admitted with costs running more than $2,000 per incident. As a group the high-cost patients do not consume as many health care dollars as the large volume of moderate-cost patients. The successful plan must assume variation and deviation from the average.

Risk Components

Dr. Broskowski explained that risk also varies with the type of risk sharing arrangement. He alined the three risk components-price, units/user, and users/1000-with fee-for-service, case-rate, and capitation arrangements. In fee-for-service arrangements only the price component is at risk. There is risk that the cost of production, such as an hour of therapy, will exceed the negotiated reimbursement. In case-rate arrangements there is risk in the price and units/user components. Thus, there is potential for loss if the price for care is more than expected or if the average patient uses more care than planned, such as an excessive number of emergency room visits. All three components are at risk under the capitation arrangement. In addition to price and units/user, if the number of users increases beyond expectations, there is financial loss. Conversely, with the right management, each risk of loss can be turned into a gain. Gains are realized if costs are reduced through alternative treatments, the pattern of use is more controlled, and the number of users is lowered due to fewer return visits as the patient population becomes healthier.

Evolution of Providers' Risk

Dr. Broskowski noted that the providers' risk evolves from fee-for-service-the lowest risk-to the moderate risk of case rate (with its focus on patients), to the high risk of capitation (with its focus on the membership or community). The highest risk is seen with the percentage of premium reimbursements that encourages doctors to work collectively and places everything at risk-cost/unit, units/user, number of users, and all operating expenses.

In addition, Dr. Broskowski discussed the cycle of carve-outs for specialty diseases and their reintegration into a health maintenance organization (HMO). Specialty diseases, such as cancer, are often carved out, and a company formed to manage the disease. The company finds providers who are efficient, improve access, narrow the cost per patient, and establish good guidelines for treatment. When the payer sees a pattern of smaller variability, with the errors and extremes worked out, the care is likely to be reintegrated into the HMO.

Important Factors to Consider for Capitation Rate Calculations

Dr. Broskowski reviewed the following factors that one should consider in calculating capitation rates:

1. Characteristics of the population related to its use of service

   Characterizing the population as a whole, not by individuals, is important. The three components one should consider are (a) rate of use among members (number of users per 1000 eligibles), (b) the amount of use among users, and (c) the patterns of use. The pattern can be predicted through research information on comparable groups, or an intelligent estimate can be made based on management and clinical expertise. Using historical data is often misleading.

2. Factors that could affect enrollment, producing adverse selection

   In examining enrollment, the following questions should be answered: Is enrollment required across a broad range of populations? Is enrollment voluntary? Is it competitive or an exclusive provider franchise? Are different benefits being offered to attract a selective (disproportionately healthier or sicker) set of population members?

3. Benefit Design: factors that affect revenue and expenses

   Benefit issues that one should consider include (a) the range of services required (these affect expenses), (b) provider stop-loss provisions that may limit expenses, and (c) copayment requirements that can affect utilization and increase revenues.

   The current structure, which is dominated by fee-for-service, interferes with social incentives to take care of the community. Providers are often burdened with regulations; advocates want guarantees for small slivers of the population; and benefits are viewed as simply something against which to bill. The structure and incentives need to change to improve the health of the entire community, for example, by offering good prenatal care to lower the risk of premature births.

4. Range of financial and contractual adjustments and protections

   There are several important items to ask for in financial and contractual agreements. The first is a multiyear contract with provisions for adjustments in capitation tied to the consumer price index (CPI) and the managed cost improvement (MCI). Provisions for provider aggregate stop-loss or catastrophic-case cost-sharing and retroactive adjustments for unanticipated referrals or utilization should also be included. In addition, adjustments for "out-of-woodwork" phenomena in the first 1 to 2 years should be in place. The reason for this is that often people who have been denied care come "out of the woodwork" in the early years as members are encouraged to use benefits. Consequently, the last provision protects against an increase in early utilization that a model may not predict.

5. The ability of the delivery system to influence utilization patterns

The following activities improve management and lower risk: (a) reducing the rate of use, particularly of the most expensive services, (b) reducing the amount of use per member, (c) reducing repeated use by members, and (d) promoting positive alternatives and a healthier population.

Statistical Considerations

Dr. Broskowski explained that good data on as few as 5,000 lives can generate reliable statistical measures for routine predictions, such as the number of inpatient days per 1,000 members. For rare disorders like autism, a larger sample of perhaps 50,000 lives is required. If the health plan covers diseases and disorders that are rare but costly, a large population base is needed to spread the risk. Looking at the distribution of patient costs and the underlying variation is also important.

Statistical Data

Network Design

Dr. Broskowski used statistical data from the actuarial company Milliman and Robertson to illustrate that high costs may be a function of network design, not population variation or sickness. The data were based on a population with no health care complications or comorbidity. Dr. Broskowski compared a loose network design, such as an independent physicians association (IPA), to a tight network design, such as an HMO staff model. For mental health, the following statistics were found:

Dr. Broskowski again warned that historical rates on utilization or users can be misleading. One has to ask if the data reflect the population, social characteristics, benefit design, or delivery system.

Treatment Prevalence

Dr. Broskowski stated that most providers overestimate the prevalence of disorders and the treatment population. They assume that everyone who needs treatment recognizes the need, wants treatment, and seeks it out. He used data from several sources to show that prevalence is different from treated prevalence. There may be several reasons for this. It could be a problem with access or awareness or the wrong incentives. Recognizing this difference, however, is important; the provider has to gamble on the number of patients who will actually seek treatment.

Dr. Broskowski then reviewed national data that revealed that 28 percent of the adult population has a mental health or addictive disorder. However, only 14.7 percent of the adults with these disorders receive treatment. Of that 14.7 percent, almost half have subthreshold disorders and do not need to be treated. Many people need treatment and are not receiving it, and many people who do not need treatment are being overtreated. Only 3 to 4 percent of the adults with diagnosable disorders are treated in specialty centers, such as mental health centers or hospitals, or by private practitioners. The majority are treated in the general medical sector or voluntary care sector.

The next data sample-from Kessler and colleagues and the Institute of Social Research-showed that 24 percent of the adult population has a mental health disorder. Within that 24 percent, 5.7 percent have a serious mental illness (SMI). Within the 5.7 percent with SMI, 2.7 percent have a severe and persistent mental illness (SPMI). The vast majority have less than serious-and less costly-mental health problems.

Callahan and colleagues found 222.6 unduplicated users per thousand Massachusetts Medicaid enrollees in 1993. This large number suggests an incentive to get people in for treatment. However, when the utilization data were examined, most of the patients had come in for one visit, a few for two visits, and very few for three or more visits. The high volume of single visits may signify bad clinical practice and a way for fee-for-service providers to make money.

Christianson and colleagues examined Utah's prepaid mental health plan, comparing capitated sites with noncapitated sites in 1988 and 1992. They found that while admission rates stayed the same, the costs (dollars PMPM) increased at the noncapitated sites but dropped at the capitated sites. The difference might be explained by the units of care or the price per unit.

Variability

To illustrate the importance of variability, Dr. Broskowski used Medicaid costs for 1 year in populations in New York counties receiving supplemental security income (SSI) and aid for families with dependent-children (AFDC) funding. The information revealed wide variation between the counties. A sample of the data sets lists the mean, standard deviation (the measure of dispersion or variability), and coefficient of variance (the ratio of the standard deviation to the mean) for SSI and AFDC populations.

These samples show that predicting reasonable costs is very difficult; there is too much variability or not enough consistency. The SSI adolescent population shows the greatest variability and risk; the AFDC adult population is least risky due to lower costs and less variability. Variability may be due to inefficient or inappropriate delivery systems, or it could reflect the pathology of the population. Dr. Broskowski stressed that it is not enough to look at statistical averages as they do not tell the whole story. Two distributions may have the same average, yet very different variabilities. Risk decreases in parallel with variation: the smaller the variation, the smaller the risk.

Models to Estimate Cost

Monte Carlo Simulation

Dr. Broskowski explained that it is possible to model health care needs prospectively by using a Monte Carlo simulation that samples from a range of possibilities. Distributions, which more accurately reflect variation, are substituted for averages.

The model that Dr. Broskowski discussed predicted costs for inpatient and outpatient care using users/1000, units/user, and price/unit as variables. Each of the six variables possible for both types of care can change. If limited to 10 possible values for each variable, there are 610 combinations. However, not all combinations are equally likely to occur, and many combinations would have comparable cost effects.

A different distribution is substituted for an average of each variable. The Poisson distribution, which is used for the users/1000 variable, is asymmetrical with a slight tail and is used to describe the probability of rare events. The gamma distribution, which represents the units/user variable, describes the deviation from a likely value; it also is skewed. The greater the deviation, the further the average is from the mode. The normal distribution represents the price/unit variable.

Dr. Broskowski demonstrated how quickly the computer can simulate 1,000 different combinations of the six variables. Within a couple minutes the computer calculated the PMPM dollars, average cost per inpatient episode, average cost per outpatient episode, total average, and average cost per user. Dr. Broskowski pointed out that this example was a simple model with three variables for two types of care. More typically the models have 250 or more variables to reflect the real world.

Cumulative Probability Curve

Using the computer, Dr. Broskowski converted information from the Monte Carlo simulation to a cumulative probability curve showing the distribution of PMPM dollars. The x-axis represented the variation in PMPM costs (e.g., from $1.39 to $68.09). The y-axis represented the probability of PMPM being equal to or greater than an x-axis value. In the example used, if the Monte Carlo simulation reveals that PMPM costs are $13.69, the cumulative probability curve shows that there is a 75 percent chance that PMPM costs will be $13.69 or less. The cumulative probability curve is useful in determining how much to bid for costs. In the same example, a $21 PMPM bid assures that 100 percent of the time the cost would be equal to or less than $21. However, this bid is not competitive. A bid of $9 PMPM is certainly competitive but only has a 20 percent chance of success. Dr. Broskowski recommended a bid set at an 80 percent chance of success. This allows enough reserve to sustain a 10 percent loss for 3 months in a row. The cumulative probability curve allows providers to find a point where the risk is reasonable, and the price is affordable or competitive.

Distributional Effects on Costs and Revenues

Dr. Broskowski continued with a data table depicting a population of 100,000 and a range of costs per member per year from $0.00 to $47,251. The total costs for the population were $7,287,000; the cost per member per year was $72.87; and the capitation rate was $75.00-leaving a profit margin of 2.9 percent. Dr. Broskowski pointed out that with a $250 deductible, 96 percent of the members are treated with no cost to the company. A very small percentage, about 0.2 percent, costs the company more than $10,000 per year.

Dr. Broskowski stated that if people are asked how they would double the profit margin using this population, invariably the reply is to cut the costs of patients consuming more than $40,000 in health care. He argued that a greater profit margin is realized if costs are cut in the moderate-cost group of members because there are more such members. He illustrated his argument using a computer simulation to shift the number of patients into different cost brackets. The following is an example:

Per Year Cost Range	Original number of members	Shift in the number of high-cost members	Shift in the number of moderate-cost patients
1,001 - 2,500	400	400	475
2,501 - 5,000	300	300	275
5,001 - 10,000	200	200	225
10,001 - 15,000	100	125	75
15,001 - 20,000	50	25	50
20,001 - 25,000	20	27	20
25,001 - 30,000	14	11	14
30,001 - 35,000	8	7	8
35,001 - 40,000	5	4	5
40,001 - 45,000	3	1	3
	profit margin = 2.9%	profit margin = 5.9%	profit margin =9.9%

Dr. Broskowski's simulation showed that shifting the number of moderate-cost patients into lower expense groups was more profitable than cutting the care of high-cost patients. He illustrated that it is possible to reduce costs without threatening the care of high-need patients.

Dr. Broskowski also showed data from a Tennessee agency that treated only seriously mentally ill, high-cost patients. These data, along with the previous sample, illustrate the Pareto distribution or 80-20 rule. In these examples 80 percent of the patients accounted for 20 percent of the costs, or, conversely, 20 percent of the patients accounted for 80 percent of the costs. This is a maldistribution of resources. Again, Dr. Broskowski stressed that to minimize the deviation, the focus should be on the large volume of moderate-cost patients and not on cutting the care of the seriously ill.

Question and Answer Session
(Mental Health Liaison Group)

1.
Q: One thing that is showing up a lot in public health is performance measures and outcome measures, and it reminds me of what you are talking about with health care. We are getting a lot of signals from the Federal Government that it wants to have proof that things are really working.

A: Yes, GTE, for example, is developing a plan to buy outcomes. The company expects to use statistical methods of sampling to sample the covered population to see how healthy employees are; then different HMOs will serve the population. In 2 to 3 years another sample of eligibles would be taken. As the group gets healthier, bonuses would be given to the HMOs that enrolled those lives. The concept here is that they want to pay for a healthy workforce. So the HMO changes its strategy and it pays, not only to treat the sick but, to keep the healthy ones healthy. Well, government is trying to do that too. But this approach can work only if there is a good relationship between the payer, the risk-taker, and the provider. Someone has to have a long-term incentive to make people healthy, but right now those incentives do not exist.

Q: The other problem is that the National Academy of Sciences mentions many things that are done that do not have a direct correlation to an outcome.

A: It is very hard to know what works. No research shows that a particular form of psychotherapy works better than others. It is very difficult research to try to match people to drug and alcohol treatment. Many studies show that many factors make a difference in outcomes. However, a whole lot of people who have anxiety and depression get better without treatment. There is a myriad of factors. So what happens is that Government wants to buy process measures but without knowing what produces the outcomes? People don't know how to measure outcomes or for which outcomes they can be held accountable, so they tend to measure the process. The advocates are concerned that managed care gives an incentive to providers to give less care.

2.
Q: Currently, within managed care plans, as we talk about individual providers or provider groups, how does the distribution of those different types of risk-sharing arrangements sort out?

A: Most, risk-sharing arrangements in mental health, are discounted fee-for-service. Colorado and a few other states have jumped to capitation, and some places like Oregon are trying case rates.

Q: I was under the impression that the majority are discounted fee-for-service.

A: You're right.

3.
Q: What is the discounted global fee?

A: Many types of individual specific services may be commonly delivered as a "bundle" of services, as when a day in the hospital always included a therapy session and usually some group therapy. When these individual components are combined under a single fee, it is called a global fee or bundled fee. It can also be offered and paid for at a discounted rate.

Question and Answer Session II
(SAMHSA Group)

1.
Q: Could you speak about Medicaid's disabled persons-a smaller proportion of them use a larger proportion of the services?

A: I am not suggesting we deny services to the seriously mentally ill person. In fact, I am saying do not try to save money in the system on the backs of high-cost users. These people are seriously ill; they need group housing, expensive drugs, and vocational care therapy. Do not try to save money in the distribution of cost/patient by removing cost from the deep end; it hurts the patient, and it does not work. Respect the tail of the distribution unless there is some evidence the high cost is inappropriate. Usually high costs for the seriously ill are justified. It is better to focus cost efficiency efforts on other strata such as moderate-care patients.

Q: I am still not convinced that if you concentrate on the majority of people who are not high-cost users, you are going to save more money than if you concentrate on the people that use a greater percentage of your dollars. In other words, look at your costs instead of your number of people.

A: You must look at cost per person. [Dr. Broskowski then ran a computer simulation changing the distribution of high-cost and moderate-cost patients. It showed a greater saving when the distribution of moderate-cost patients was shifted.]

2.
Q: When you do an analysis like this, do you ask HMOs for their philosophy on things like waiting time for initial appointments and the number of weeks between appointments? I will argue if the HMO is trying to curtail use, it might have a waiting time of 4 to 6 weeks and then some months between subsequent visits. It may be that people are being lost and seen only once because there is not enough access.

A: If I am an HMO leader in a competitive situation, and word gets around that you call me and do not get an appointment, but call the competitor and you do get an appointment, I will lose membership and my revenue will drop. This won't work under a monopoly franchise, but it will if you get a competition going and let people make choices based on their experiences. You are right that it is important to measure a lot of other things-the quality of administrative services, customer service, answering the phone, and giving good readable material. I can do many things as an HMO to build up membership, reduce costs, and improve outcomes without denying care.

3.
Q: This reminds me of the wave in the early 1980s of just-in-time inventory-let us try to reduce the inventory maintained by a producer or car manufacturer and ask for it only when needed. My distributor, someone with whom I have a long-term relationship, will provide a bumper and provide it just as the car comes down the line. What it really does is transfer the fixed costs onto the providers. This works terrifically as long as you are a managed care company-but if I am a psychologist in independent practice, all of a sudden-

A: The risk is shifted to you. If you accept that kind of risk as the provider, what is your goal? Your goal is the right size warehouse with the right amount of inventory. But you are going to have a very special relationship with your buyer and have a very good idea at any one time of the likelihood of a certain part being needed off the shelf. You charge him for storing that inventory because you specialize in that, and you can charge him less than it costs if he does it for himself.

Q: That works only when you have a stable, long-term relationship with your providers, and right now what we have is instability. In fact, what we have is a number of our big managed care plans that do not want to be a major income stream for any of their providers.

A: I am not saying they are doing it right; the goal is to develop a long-term relationship with providers, such as 3-year contracts. The States ought to be interested in giving providers 5-year contracts and letting them figure out how to do this themselves before you bring in the MCOs.