University of South Carolina, Arnold School of Public Health, Dept. of Health Services Policy and Management, HADM J712

Notes on the Demand for Health Care and Cost-Sharing


An important distinction is between Need and Demand
Someone's subjective idea 
(may be based on a formula applied objectively, but the choice to use the formula was someone's subjective idea. 
Money is not a factor.
Objectively observable as behavior in the market. 
Money is a key factor.  "Demand" is also called "effective demand," because it's expressed only by spending money.

Example of "need":  Hill-Burton hospital-building-subsidy program used 4.5 bed/1000 pop formula (more in rural areas).  On the other hand, DHEC's "need" formula for hospital beds is based on historical ratios of utilization to population.  This makes it a demand measure.

Health plans that focus on need and ignore demand will face under- or over-utilization of service capacity.  If one believes quantity demanded is too little (e.g. indigents) or too much (e.g. overuse of emergency room) relative to need, then quantity demanded must be manipulated,
 by changing price or other costs to buyer, or
 by changing demand through marketing or de-marketing.

Demand is a relationship between price and quantity.  We can study
   how high the demand is -- the position of the demand curve -- and
   how elastic (responsive to price changes) demand is -- the steepness and shape of the curve.


This concept is used in some of the articles which follow.  There are two interactive lectures about this.  Here are some details on the concept.

Elasticity of demand is a measure of the responsiveness of the demanded quantity to price changes.

The elasticity of demand for something is:

For instance, If price goes up 1% and as a result sales fall 2%, the elasticity is -2%/1%=-2.
Elasticity, as a measure of responsiveness to price change, is an alternative to the slope, which would be

Example of calculation (from Phil Jacobs' book):
Health Dept. charge per shot Shots done per month
$3 1200
$3.25 1150

Slope of demand for tests is
-50 tests / $0.25 = -200 tests per $ of price change.

I used 1175 and $3.125 in the elasticity formula.  These are averages of the before and after quantity and price levels.  1175 = (1150+1200)/2.  $3.125 = ($3.00 + $3.25)/2.  The elasticity calculated this way is called the arc elasticity.

Notice how the slope has units (tests per dollar), but the elasticity does not.  The elasticity has no units because the numerator is tests/tests and the denominator is $/$.  All the units cancel out, leaving a unitless number.  This means you get the same answer regardless of what units you use.  This lets you compare elasticities of totally different products, like medical care, petroleum, and videocassette rentals.

One interpretation of elasticity of demand:
It's the percentage change in quantity demanded that comes from a 1% change in the price.

Here's another way to interpret elasticity of demand:
 Suppose you raise the price of something you're selling.  Will you make more money?  The answer depends on the elasticity of demand.
Elasticity (absolute value, ignoring the minus sign) If price goes up Demand is categorized as:
> 1 spending goes down. Elastic
= 1 spending stays the same. Unitary elasticity
< 1 spending goes up. Inelastic

What are examples of products or services whose demand is elastic?  Inelastic?

Hot tip:  If you want to impress economists, say:  "That depends on the elasticity, doesn't it?"

The small elasticities observed for medical care demand imply that higher med care prices will cause people to cut back some, but they will not much, so the total amount spent on care will go up.

Elasticity can be used for other things besides quantities and prices.   The elasticity of health status with respect to medical care expenditure is the percentage difference in health status divided by the percentage difference in medical care expenditure.  The cause (e.g. expenditure change) is in the denominator.  The effect (health status change) is in the numerator.

The advantage of elasticity over slope is that, as mentined, elasticities are numbers with no units.
A disadvantage of elasticity is the other side of its advantage:  The unitless number obscures the problems with measuring the quantities in the equation, such as health status.

The Position of the Medical Care Demand Curve

Demand for medical care is derived from demand for health.
Patient does not necessarily demand particular services.
Physician "combines inputs to provide treatment" -- chooses particular services.

What determines how high demand is (i.e. how much medical care will be bought at various prices):
Physical condition 
Social factors
affect need for care as perceived by individual and society
affects how much the person can pay

Physical and cultural/demographic factors can vary regionally or even locally.

Regarding income:  The usual finding is that medical spending goes up as income goes up, but less than proportionally. In other words, the income elasticity of med care demand is between 0 and 1.

Insurance and the Demand for Health Care

Insurance plays a major role in shaping health care, unlike most goods and services.  If we define insurance broadly, there are three kinds of insurance:
  1. Voluntary insurance purchase -- many of us buy insurance to assure us access to services that we could not afford to pay for directly.
  2. Rule of rescue -- we don't like to see people suffer or die needlessly.  We expect providers to give service to people in immediate need even if they don't pay.
  3. Government insurance extends the rule of rescue to general health care for the elderly and some of the poor.
(Irony about rule-of-rescue emergency care: Prices charged do affect incentive to seek care among people who can afford to pay. And you can pay to get priority: In a discretionary situation, choosing to go by ambulance gets you to the front of the line.)

Demand, Need, and Copayments

Brook, R.H., et al, "Does Free Care Improve Adults' Health? Results from a Controlled Trial of Cost Sharing in Health Insurance," N Engl J Med, December 8, 1983, 309, pp. 1426-1434.

How big is moral hazard in health insurance? How much does health insurance induce people to buy health care that they don't much need? The RAND health insurance study sought to answer these questions (which are not the same, by the way).

This article is a major one in a series coming out of the $70 million government-funded RAND health insurance study. In this portion of the study, 3958 adults were followed over 8 years. 14 insurance plan incentive schemes tried, from no incentive -- free care -- through various copayments of 25 to 50%, to catastrophic insurance that required 95% payment for all services until the family had spent 15% of its income or $1000, whichever was less. (I imagine they used 95% rather than 100% to give the families an incentive to send in their medical bills.)

The free-care vs. copayment comparison allows you to trace out a demand curve

Adults who had to pay used about 2/3 of the ambulatory visits and hospitalizations of those who didn't. An earlier article (Newhouse et al, NEJM 305(25) Dec. 17, 1981, 1501-7) reported this. There was little or no significant difference among the pay plans. As far as they could tell, 25% copayment had the same effect as 95%. Low income families responded more to copayments than higher income families, but not by much in most years and locations. The authors point out that low income families will illnesses would exceed their deductible sooner than higher income families, because the deductible was proportional to income.

This article goes beyond demand to need, by exploring the impact of differences in utilization on health status.

For the whole population, differences in health status are found only for persons with poor vision or high blood pressure.

Elevated risk (worse 25%) persons -- better outcome for free plan for blood pressure (at 0.07 level), far vision (0.02 level), risk of dying (0.001 level).

(Apparent misprint in table 7: 1.42 should be 2.11 in last row.)

Other measures of health included self-assessment by questionnaire.

The impact of cost sharing is clearest in lowest income group (bottom 20%).

Not clear if cost sharing has impact in the rest of the population.

One conclusion is that specifically targeted programs (vision, hypertension, especially for poor) are more cost-effective than free care for all at improving public health status. (Cost effective means more bang for the buck.)

A note about the cash payment to those with catastrophic insurance:  People who had the 95% copayment plan got a cash payment of about $80 a month.  This meant that they had cash that they could spend on health care, if they chose, but they could chose to spend it on something else instead.  Orthodox economists believe that it's better to give people money than services of the same value.  That way people can spend the money in a way that maximizes their satisfaction.  If they want medical service, they can buy it, but if they want something else more, they can buy that.  Put another way, the investigators suspected that many people would rather forgo seeking medical care for some minor (to them) conditions and spend the money saved on something else.  If care is free, a person may go to the emergency room and spend $100 worth of society's resources on treatment of a headache, say.  If the person has to pay part or all of that $100, he/she may prefer to take an aspirin at home and spend the other $99.95 on something else.  Either way uses up $100 worth of society's resources, but the second way gives the person with the headache more satisfaction.

In the same issue was an editorial comment, Relman, A. S., "The Rand Insurance Study, Is Cost Sharing Dangerous to Your Health?" New Engl J Med, December 8, 1983, 309, 1453.  This is not in the packet.

Relman notes that the population studied was mostly health adults under age 65.  Limited measures of health status were reported.  The study does show the difficulty and expense of experiments to determine how health care relates to health status.

My comment:  The worst (highest copayment) plan in the experiment is better than what the poor usually have.  It's better than Medicaid for hospitalizations, and certainly better being medically indigent and having to rely on whatever care the public hospitals are willing to provide free.

Himmelstein, D.U., Woolhandler, S., "Free Care, Cholestyramine, and Health Policy," N Engl J Med, December 6, 1984, 311, pp. 1511-1514.

Cholestyramine can reduce the risk of heart disease, but daily therapy costs $1861.50 per year for the drug alone.
      $9,300,000 average cost per life saved
      $780,000 per coronary heart disease death or non-fatal MI (heart attack) averted.
Free health care for all men over 50 saves lives at an average cost of
      $654,000 (compared with 95% copayment), or
      $378,000 (compared with 50% copayment).
Free care for everyone saves lives at average cost of
The 4th stool guaiac test detects colon cancers at an average cost of
      $906,088 in 1984 dollars.  ($1 million in 1987 $)

So, if cholestyramine is cost-effective, why isn't free care?

O'Grady, K.F., Manning, W.G., Newhouse, J.P., Brook, R.H., "The Impact of Cost Sharing on Emergency Department Use" N Engl J Med, August 22, 1985, 313, pp. 484-490.

From RAND experiment, a study of the demand for emergency room visits.

The effect of copayment is greater for less serious diagnoses.  Compared with free care, coinsurance plans combined had
 77% as many visits for "more urgent" complaints (lacerations, 2nd degree burns, urinary tract infection, head injury, chest pain -- for these last two, copayment didn't affect visits at all)
 53% as many visits for "less urgent" complaints (abrasion, sprain, upper respiratory infection, gi, headache [only 11% as many visits], 1st degree burn [28%])

Even controlling for type of insurance coverage, persons in the lower 1/3 of the income distribution used the ER 64% more than persons in the upper 1/3.  Maybe the poor were accustomed to ER use.  Maybe there was a lack of private docs in the poor areas.

Shapiro, M.F., Ware, J.F., Sherbourne, C.D., "Effects of Cost Sharing on Seeking Care for Serious and Minor Symptoms," Annals of Internal Medicine, February 1986, 104, pp. 246-251.

Shapiro et al re-analyzed some of the Rand data, looking at different responses among patients with more serious vs. less serious symptoms.  For those with minor symptoms, cost sharing meant 1/3 less visits as above.  For serious symptoms, cost sharing doesn't affect propensity to seek care, except for poor.  Here "poor" means the low 40% of socioeconomic status (Brook used 20%).

Health status measured by presence of various symptoms in annual survey.  Survey asked about health status during previous month.

Among those who were sick when the HIE (health insurance experiment) began, the poor reported more symptoms than the non-poor.
The poor in the free care plan improved to where they were no sicker (serious symptoms) than the non-poor.
The poor in the copayment plans also improved some, but remained sicker (serious symptoms) than the non-poor.

Why the improvement in both groups?
They say:  Regression towards the mean.  Whenever you divide people into a sick group and a well group, some people in the sick group will get better on their own.  Meanwhile, some people in the well group will get sick.
I might add, as mentioned above:  Even the catastrophic plan was better than what many of the poor had before. They now had cash to spend, and they had insurance against big expenses.


This is not in the packet, but has some amusing ideas if you're interested to read it:
Manning, W.G., et al, "Health Insurance and the Demand for Medical Care," American Economic Review, June 1987, 77:251-277.

Lots of economeze.   Read the understandable parts and skip the rest.

Manning's Points:
Demand.  Summarizing the results of the Brook and O'Grady articles, medical care demand elasticity is generally about -0.2.

Manning says insurance causes a welfare loss of $37-$60 billion / year (1984 $).  Here's the idea:  When goods or services are free (or subsidized), we buy more of them than we would if we have to pay.  We might spend $50 worth of resources on a service that's worth only, say $5, to us.  Manning et al would call that a welfare loss of $45, because somebody else would presumably have been willing to pay $50 for those resources (that's why we say they are worth $50).  Add that up over all free or heavily subsidized medical care and you get $37-$60 billion, or about $160 per person per year in the U.S.  (You estimate total welfare loss from the amount spent on medical care and from your estimated elasticity of demand.)

Was the experiment the money spent on it?
Manning gives the experiment credit for reforms in health insurance that have been sweeping the U.S. since its results were announced, particularly insurers imposing new "first dollar charges" -- deductibles and coinsurance.
The experiment cost $136 million in 1984 present value.  That's only 1.5 days worth of welfare loss, and is only 1 week's worth of the $7 billion annual savings Manning et al attribute to the imposition of first dollar charges so far.  So the experiment was well worth it, he says.

Blustein, J., "Medicare Coverage. Supplemental Insurance, and the Use of Mammography by Older Women," N Engl J Med, April 27, 1995, 332(17), pp. 1138-1143.

Moving up the 1990's, this study took advantage of Medicare starting to pay for mammograms for screening for breast cancer. (Medicare already paid for diagnostic mammograms for women whose examination found a lump.) The effect of copayments are documented. Note, though, that this study could not randomly assign patients to insurance groups, as RAND could

Economic factors affect even those on Medicare.  OK, you're not surprised.

Article looks at demand and need for mammograms.  Women in this age group assumed to need a mammogram every two years.  Study was of Medicare bills during the first two years in which Medicare paid for mammography.  This payment, like all Medicare, is subject to the Medicare deductible, then $100 per year.  After deductible, patient could pay up to about $20 in copayment and "balance billing."  Supplemental insurance would take care of all or most of this, to varying degrees.

14% of women without supplemental insurance had a mammogram in the two-year period.
24% of women with Medicaid paying their share had a mammogram.
40% of women with self-paid supplemental insurance had a mammogram
45% of women with employer-paid supplemental insurance had a mammogram

The problem with making a judgement from these statistics alone is confounding factors.  General economic circumstances and attitudes affect this decision, and also affect the likelihood of having supplemental insurance.  For example, people with good jobs (4th category above) have more income.

"Multivariate" analysis attempts to separate the causes and isolate the effect of insurance given the other factors, like age, race, income, education, self-image, ... [More on this next semester in 716.]  Leads to "adjusted odds ratios" in table 4 (p. 1141), which actually aren't that different from the raw numbers.  For instance, the adjusted odds ratio of 3 for employer-paid insurance means women with that insurance are three times as likely as women without supplemental insurance to have had a mammagram.  45% is about 3 times 14%.

A problem with studies of this type is self-selection.  People who expect or intend to use more services are more likely to buy insurance.  The apparent effect of insurance on demand is actually a result of the prior disposition.  This would be most true of women who pay for their own supplemental insurance.  It would be less true of those on Medicaid and those with employer-paid insurance.

As a general point in marketing (or health promotion and education), there was still plenty of work to be done in encouraging demand, even among those with insurance.  Half of them didn't get a mammogram.  Still, price is clearly a factor.

Rasell, M.E., "Cost Sharing in Health Insurance -- A Reexamination," N Engl J Med, April 27, 1995, 332(17), pp. 1164-1168.

Reviewing twenty years of evaluations of copayments, Rasell considers the costs and benefits.

Cost sharing is becoming even more trendy.  In RAND days, HMOs typically had little or no cost- sharing.  That was one of their attractions.  They relied on their management of care and presumed benefits from prevention to keep utilization down.  Today, some are using copayments as well.

Copayments -- good or bad?  Did the RAND study settle this issue?

Rasell's article seems sloppily written, or sloppily edited.  For example, the two "approaches" to cost control in the first part of the paper (a. copayments and b. different premiums for insurance policies with different copayments) are two sides of the same coin.  Copayments make insurance less costly to provide because (1) there are fewer claims and (2) the company doesn't have to pay 100% of the cost of claims that do come in.  If the employee pays at least part of the insurance premium, the employee has an incentive to choose the policy with greater copayments.  Otherwise, if the employer pays the full premium either way, no employee will choose a plan with copayments unless it has some other compensating advantage.

On page 1165 she says that Manning says that "cost-sharing does not affect the intensity of care, defined as the number and type of services provided per year" [my emphasis].  This is a mistake.  Manning (p. 258) actually said that cost-sharing does not affect the number and type of services provided per encounter. In other words, cost sharing affects how likely you are to see a doctor or be admitted to the hospital, but not how much you spend once you are there.  Manning's actual idea fits better with the point Rasell is trying to make, so this looks like an editing error.

International comparison of contact rates.  US has fewer physician visits per year than other advanced countries.  Hospital bed days same as UK, less than other countries.  Outpatient surgery somewhat compensates for this, she says, but has no figures.

Even though cost sharing is the norm in the US, our health care spending is growing faster than other countries'.  Evidently the other countries' methods of controlling their health spending are more effective than ours yet don't discourage demand as much.

Cost sharing and unnecessary care.
Different Siu article cited to show utilization of both appropriate and inappropriate care affected, but findings are well summarized here.
Part of what happened was that initial Brook article made big splash, while the follow-up studies that showed more effect of copayments, particularly on the poor, didn't.

Rasell's reading of follow-up studies shows serious symptoms were more prevalent for the sick poor on cost-sharing than with free care. The sick poor with free care improved to where their symptoms were no more prevalent than the sick among the higher income participants. Children in low income families got less care under cost-sharing than under free care. In better-off families, cost-haring didn't matter.

Cost sharing and health
Among low-income and unhealthy, a statistically significant increase in serious symptoms, as we saw. Rasell says income-related cost-sharing has administrative cost.

Whose behavior needs to change to control health care costs?  Rasell says don't blame consumers.  She turns to physicians.  Cites example in which physicians raised fees and increased intensity of services in apparent response to reduction in demand for contacts that followed start of copayments in a union insurance plan.

Financial incentives to purchase less expensive health insurance raise concern that allowing plans to compete on price/comprehensiveness will open doors to risk selection by insurers. Risk-adjusted premium methodology not well developed.
Insurance comprehensiveness of would be stratified by income.  Already evidence of that.
The views and opinions expressed in this page are strictly those of the page author. The contents of this page have not been reviewed or approved by the University of South Carolina.