Academic Articles

Should employers provide health insurance to their employees? There are many reasons why they should. One is that employees are attracted to firms that offer health insurance, especially since their are tax and cost advantages to group health insurance purchased through an employer. Another reason is that if a worker becomes sick, that reduces productivity. But how much does it cost a firm when either 1) a worker is absent from work (absenteeism) or 2) the worker shows up for work but their productivity is impaired (presenteeism)?

A paper by Pauly et al. (2008) attempts to quantify these costs. Unlike most studies, they use manager estimates of lost productivity. While workers may have better information regarding how much productivity is lost during an illness, workers may have an incentive to answer strategically and further, managers will be the ones ultimately deciding how much health care preventing employee illness is truly worth.

The authors hypothesize that jobs with the following three characteristics will be more affected when a worker becomes ill:

• jobs with high values of team production,
• jobs with high requirements for timely output, and
• jobs with high difficulties of substitution for absent or impaired workers.

A firm of course, can invest in protection measures in the case of illness. For instance, it could cross-train workers, it could accumulate inventory to smooth out periods of down time, or might pay workers to work harder if an employee is missing. Nevertheless, the authors estimate the costs of absenteeism as follows:

The authors assume that the cost of being absent must be at least the employees wage if the labor market is competitive. We see that the cost of an illness is significantly higher than the wage. Also, jobs that were found to involve more teamwork, had times sensitive products, and for which workers were not substitutable had larger absentee multipliers.

For workers who are sick, but come to work, here is the cost to the firm as a percentage of the employees wage.

One problem with this analysis is that it assumes that this is a one-time illness. If these are long term illnesses, it may be more cost effective for the firm to fire the employee because their sickness is either 1) driving up insurance premiums or 2) causing them to miss too much work. Offering a generous health insurance benefit may help to prevent illness, but may also attract sicker people to the firm. Thus, despite the article’s demonstration that the cost of employees missing work is significantly higher than their wage and that there are large costs when workers come to work when they are sick, it still does not mean that employers will want to offer generous health plans.

• MV Pauly, S Nicholson, D Polsky, ML Berger, C Sharda (2008). “Valuing Reductions in On-the-job Illness: ‘Presenteeism’ from Managerial and Economic Perspectives.” Health Economics, vol. 17(4): 469-485.

The Scientific American magazine has an interesting article (”Science 2.0“) about the web, open-access, blogging and research. Should researchers post their results online? Should scientists blog about their methodology?

Pros

It seems like academic research is the perfect forum for social networking and blogging. The sharing of ideas is a key means towards scientific invention/innovation. Posting raw data is a great way for other researchers to verify results, or utilize the same data for different purposes. One cancer researcher noted:

• “To me, opening up my lab notebook means giving people a window into what I’m doing every day,” Hooker says. “That’s an immense leap forward in clarity. In a paper, I can see what you’ve done. But I don’t know how many things you tried that didn’t work. It’s those little details that become clear with an open [online] notebook but are obscured by every other communication mechanism we have. It makes science more efficient.”

The site OpenWetWare let’s laboratories share their daily experiences online. Further, researchers who are traveling can access their lab notebooks from anywhere in the world with OpenWetWare.

Further, social networking can allow easier collaboration between colleagues working in different parts of the country or different parts of the world.

It seems like researchers would be some of the first people to utilize Web 2.0, but…

Cons

• “It’s so antithetical to the way scientists are trained,” Duke University geneticist Huntington F. Willard said at the January 2007 North Carolina Science Blogging Conference, one of the first big gatherings devoted to this topic. The whole point of blogging is getting ideas out there quickly, even at the risk of being wrong or incomplete. “But to a scientist, that’s a tough jump to make,” Willard says. “When we publish things, by and large, we’ve gone through a very long process of drafting a paper and getting it peer-reviewed. Every word is carefully chosen, because it’s going to stay there for all time. No one wants to read, ‘Contrary to the result of Willard and his colleagues….’”

Beside the fact that writing about unfinished results is not the way scientists are usually trained, most individuals worry about having their ideas stolen. Having your idea “stolen” by another individual means you will not get the recognition you deserve for coming up with an idea, and your career path can be adversely affected. Doling out credit for work accomplished is an important component of the “old school” journal system.

Other worries include the fact that when junior faculty post critical comments of the work of senior faculty, they may fear some sort of reprisal. This has lead some individuals to use pseudonyms.

Summing up

There are some serious drawback to Science 2.0, but as Timo Hannay, head of Web publishing at the Nature Publishing Group, states, “Our real mission isn’t to publish journals but to facilitate scientific communication.”

Differences in the health outcomes between white and minority patients has been well documented in the medical and economics literature. Reasons for this difference could be:

• Unequal access to treatment. Minorities are poorer and less likely to be covered by insurance than whites.
• Unequal treatment - Minorities are less likely to have a regular doc, which leads to discontinuities in care.
• Unequal quality of care available to minorities - For instance, doctors who treat blacks are less likely to be board certified.

A recent paper by Emilia Simeonova tries to dig deeper into what is causing the racial mortality gap for chronic heart failure (CHF). CHF is one of the leading causes of death for the elderly and one of the major components of the racial mortality gap.

Methods

Ms. Simeonova uses a six-year panel data set from Veterans Affairs [i.e.: the VHA Medical SAS inpatient and outpatient datasets,the Beneficiary Identification Records Locator Subsystem (BIRLS) death files, the VHA Enrollment files, and the Veterans Service Support Administration (VSSA) clinic performance measures database]. The data allow the author to compare treatment within facilities rather than just between them. This is important because it is possible that blacks go to bad doctors and whites go to good doctors and this may constitute the entire mortality gap. By comparing outcomes within a clinic or within the same doctor, the author can better analyze what is causing the mortality differences.

The author calculates 3 year survival probabilities conditional on surviving two years. This should help to eliminate different CHF severity levels. In her regression, Simeonova uses patient and clinic characteristics, as well as clinic fixed effects, and time and cohort dummies. Simeonova measures doctor quality as the probability the doctor prescribes beta blockers and ACE inhibitors to patients with chronic heart failure (CHF). However, another aspect of the quality of medical care is patient compliance. Patient compliance is calculated as the number of prescriptions filled on time divided by the total number of prescriptions filled.

Results

Simeonova finds that doctor quality accounts for 5% of the CHF mortality gap and socio-economic factors account for 20% of the differences in CHF mortality. However, the vast majority of the mortality differences are due to the fact that blacks are less likely to take their medication than whites.

“I show that doctor quality significantly influences patient outcomes. While minority patients visit slightly less competent doctors, this does not explain the large gap in survival. Individual doctors are found to treat their patients similarly regardless of race. On the patient side, I demonstrate that variation in compliance triggers a racial mortality gap. Differences in patient response to treatment significantly alter survival probabilities. Considerable reductions in medical costs could be achieved by convincing patients of the importance of strictly following the therapy regimen. I estimate that targeting compliance patterns could reduce the black-white mortality gap by at least two-thirds.”

Also interesting is that the paper found that when blacks have a regular doctor, they end up seeing a lower quality doctor. Nevertheless, compliance rates and mortality decrease for blacks when they have a regular doctor despite the fact that this doctor may be of a somewhat lower quality.

• Simeonova, Emilia. 2007. “Doctors, Patients, and the Racial Mortality Gap: What Are the Causes?” Working Paper, Nov 2007.

Economists and health researchers have generally shown that when doctors are paid on a fee-for-service basis, they will advice the patient to undergo more medical procedures than when the doctor is paid on a capitation or salaried basis (see my own paper: “Operating on Commission“). Which payment method maximizes welfare has not been proven and is likely to vary based on the medical procedure in question.

To add to the confusion, Astrid Selder writes in “Physician Reimbursement and Technology Adoption” that physician reimbursement mechanism may also affect the rate at which new technologies are adopted.

Selder lays out the following simple theoretical structure in which the social planner maximizes the following:

• (1-π)U(Y-P)+πU(Y-P-acm-ε +G(m) -nm)
• s.t.: P + πacm-%pi;R=πpm>=0
• s.t.: R + (p-c)m>=0
• s.t.: m<=B

The first constraints represent the zero-profit conditions for the insurer and the physician respectively and the third is a technology constraint. The probability of falling ill is π. Individuals have income Y, an insurance premium of P, and a coinsurance rate of a to may for the marginal cost, c, of a given level of medical treatment, m. Non monetary treatment costs (e.g.: side effects, time costs) are represented by the parameter n, while the benefits of the treatment are represented by the function G(m). The doctor receives a capitation payment of R and a marginal revenue amount given by p-c. B represents the technology constraint.

It seems obvious, but Selder shows that an increase in monetary medical costs (c) or non-medical costs (n) are welfare destroying. Technological innovations, represented as an increase in B are welfare improving if the technological constraint is binding or have no impact on welfare is the technological constraint is not binding. This results is intuitive: higher costs are bad, technological innovation is good.

This result, however, only holds when there is perfect information. With asymmetrical information, technological innovations may be welfare enhancing or welfare destroying. The welfare enhancing argument is similar to above: better technology leads to better medical care. If there is a moral hazard problem, however, using more basic technologies may actually help to counteract the tendency to over-consume medical care in the presence of insurance. Also, in the presence of moral hazard, technological innovations will cause insurance companies to charge higher premiums in order to cover the additional cost of this care.

How does the technological constraint, B, affect welfare in different physician compensation schemes?

• “If B is a binding constraint and increases in B are welfare enhancing then a fee-for-service system should be used. Increases in B and reductions in [monetary medical costs] c are induced all of which are welfare improvements. Incentives with respect to [non-monetary costs] n cannot be improved upon.
• If B is a binding constraint and increases in B are welfare reducing then a system of cost sharing [e.g.: capitation] should be implemented which induces reductions in B and c and gives no clearcut incentives with respect to n. Incentives with respect to B and c are thus optimal, and incentives with respect to n cannot be improved upon.
• If B is not binding the optimal reimbursement system depends on the demand elasticities with respect to costs.”

In words:

“It has turned out that the state should classify diseases or treatments into subgroups which are reimbursed differently in order to achieve the desired welfare effects in each subgroup. Especially for the case of extremely severe illness shocks the introduction of a fee-for-service system may be socially desirable. It is these very severe diseases where the capitation system seems to induce negative welfare effects with respect to the technologically feasible boundary of treatment. Cost sharing/capitation systems are most suitable for treatments where a reduction of the amount of health care consumed is desirable.”

While the solution Selder proposes is logical, it does not take into account the cost to the state to collect information regarding these disease classes. Lobbying will likely influence whether a given treatment falls into the fee-for-service or cost-sharing grouping. Nevertheless, Selder’s general findings seems reasonable and applicable to the medical care finance design in the real world.

• Selder, Astrid. 2005. “Physician Reimbursement and Technology Adoption” Journal of Health Economics Volume 24, Issue 5, Pages 907-930

In February of 2o07, the UK’s Office of Fair Trading (OFT) recommended reform to Britain’s current Pharmaceutical Price Regulation Scheme (PPRS). The PPRS sets maximum and minimum profit levels from the sale of branded drugs to the NHS. The PPRS allows companies freedom to set prices as they please on new substances, but restricts subsequent prince increases. A system of price cuts has also been instituted as well, yet it is likely that firms take these future price cuts into account when making their original pricing decisions.

Reform

Are there any other options? Simeon Thornton (Health Econ 2007) argues that a value-based pricing (VBP) scheme would be superior. In VBP, the NHS would pay pharmaceutical companies based on the value of the pharmaceutical to the patient base. One question is how value is determined. Thornton proposes cost effectiveness studies, which in effect means that the government or academics will determine the price.

Pricing will also be allowed to vary by subgroup since people with certain diseases may benefit more from a disease than others. Also, incremental improvement will be encouraged since marginal improvements in treatment will receive higher payments.

This program does seem to be an improvement. It is dynamically efficient since pharmaceutical companies will be paid more for more cost effective treatments. Further, if it turns out patients do not like the medicine and no one takes it, then NHS will not be paying a lot for failed drugs. Also, after generics are available, the price will adjust downward.

In the static environment, the pharmaceutical company will capture all the consumer surplus since price will equal marginal benefit. However, as time passes and generics enter the market, a large consumer surplus will occur.

The major impediment of this reform is the problem of any centralized system: information. How will the NHS determine cost-effectiveness? Will it be impartial? Will the conclusions be manipulable by interested parties? These questions are easy to answer theoretically, but very difficult to predict empirically.

• Thorton S. (2007) “Drug Price Reform in the UK: Debunking the Myths” Health Economics, v16, pp. 981-992.

I recently read a Health Affairs article analyzing a pay-for-performance (P4P) demonstration. The Local Initiative Rewarding Results (LIRR) demonstration in California involved seven Medicaid-focused health plans in California between 2003 and 2005. Here are some of my most recent thoughts on P4P:

• The article seemed to show that P4P worked best when there was much consensus over how doctors should treat patients. The irony of P4P is the following: P4P programs with high levels of physician consensus work, but if everyone physician is already acting appropriately P4P does not make a major change in behavior; larger behavioral changes can occur when physicians are greatly deviating from the optimal practice, but in this case it is very difficult to implement P4P.
• P4P plans are often implemented so those that do well gain income and those that preform poorly maintain their status quo income. When no one loses any money, P4P is supposed to gain more acceptance among physicians. In equilibrium, however, an economist would predict that those with low P4P scores will receive lower wages in the future (or wage increases below inflation) due to P4P.
• Unsurprisingly, the authors of the Health Affairs paper found that meager financial incentives do not significantly change physician behavior, while larger financial incentives have a greater impact. Also, good communication with physicians helps to increase physician compliance with P4P mandates.
• If P4P evaluates doctors on care levels which depend on patient compliance, health plans can help physicians to increase the quality of care. For instance in the above study, “many providers used the plan lists of members turning fifteen months old as the basis for outreach…”
• When P4P financial incentives vary across plans, it is difficult for physicians to focus on specific quality indicators due to complex reimbursement schemes in each plan.
• When HEDIS scores improve, some of this improvement is due to better quality while much of it may be attributed to better documentation in a patient’s medical records.

Suzanne Felt-Lisk, Gilbert Gimm and Stephanie Peterson (2007) “Making Pay-For-Performance Work In Medicaid” Health Affairs, v26(4) pp. w516-w527.

Recently, there has been much controversy regarding whether or not the RAND Health Insurance Experiment (HIE) results are truly robust. Many blogs have been questioning the results (see here, here and here). One of the major conclusions of the HIE are that higher co-insurance rates lead to lower levels of medical utilization and lower medical cost, but do not have any adverse impact on health outcomes.

A paper by John Nyman (2007) in the Journal of Health Politics, Policy and Law calls these findings into question. He notes that there was differential attrition in the insurance plans with cost-sharing for the patients compared to those with no cost-sharing. If people who became sick in the cost sharing insurance plans elected to drop out of the experiment and seek care from their previous insurance plan, then an attrition bias would occur. This bias would incorrectly deflate the medical utilization of the cost sharing group and thus lead to the erroneous conclusion that more cost sharing causes lower medical utilization patterns.

Nyman’s paper also wisely notes that moral hazard is only a problem with individuals afflicted with a disease. For instance, no matter if one has insurance that would reduce the price of a mastectomy to zero, no one ever elect to have a mastectomy unless they had breast cancer. There are personal costs of the side effects of treatment and time costs which imply that moral hazard will generally only be a problem for those afflicted with a disease. There are exceptions. For cosmetic surgery procedures, individual do not need to be sick to suffer from the problem of moral hazard with insurance.

Nyman’s points are valid, but I believe that the RAND HIE results are robust. First, each RAND HIE participant receive a participation payment each month. Even if the individual had large medical expenses and had to pay a large deductible (capped at $1000, which $4000 in 2007 dollars), the monthly participation fees would add up to more than $1000 so it was in the individual’s best interest to stay in the experiment. Nyman argues that cash flow problems many have affected some participants. If I receive $100 per month for a year, but have $1000 in medical expenses today, I may decide to leave the experiment. Yet because of this participation fee, it seem that this would not be a major problem.

Also, the problem of attrition was recognized in the original RAND HIE. Nyman himself states on his website:”When the RAND data were re-analyzed to account for the differential utilization from attrition, the difference in hospitalizations between free care and all cost-sharing arms was about 19 percent, not 25 percent as has been reported (Manning, Duan, Keeler, 1993, p. 13).” While the magnitude of the cost sharing effect may be somewhat smaller once we take into account attrition, the general finding remains the same: more cost sharing leads to less utilization.

Finally, while RAND is the only large scale RCT that has been conducted, many other studies have shown that increased cost sharing leads to lower medical utilization. Newhouse states that “enormous number of observational studies over many years, in many settings, with many different methodologies, that find utilization of medical services responds to relatively modest cost sharing.” Nyman’s response is that:

“I do not dispute that cost sharing reduces utilization. I think it does and agree that the non-experimental studies that Newhouse et al. (2007) refer to are generally convincing. The issue that I am addressing, however, is whether cost-sharing in the RAND Experiment actually produced a 25 percent reduction in equally effective hospitalizations and whether such a reduction would actually have no effect on health.”

One question you may still have is how is it possible that lower medical utilization leads to the same health outcomes. One possibility is that the moral hazard may have resulted in only the excess use ineffective medicine. It could be the case that much of medicine has little impact in the overall health of the patient and too much treatment can actually harm the patient (see my post on Overtreated). It could be the case that health was measured poorly by the HIE, but Nyman concedes that a “broad spectrum of health status measures” where used in the RAND experiment.

Overall, I do believe Nyman brought up some reasonable points. No experiment is ever perfect…even the RAND HIE. Attrition was a problem and may affect the magnitude of the result. Nevertheless, the conclusions from the RAND HIE remain robust and I believe that the attrition issue was addressed previously.

Can we still conclude that more cost sharing reduces medical utilization? Is moral hazard a problem in health insurance markets? We can still confidently say: YES!

• Nyman, John A. “American Health Policy: Cracks in the Foundation,” Journal of Health Politics, Policy and Law vol. 32, no. 5, October 2007, pp. 759-783.
• Newhouse et al. (2007) “Attrition and Refusal in the RAND Health Insurance Experiment” (from Newhouse and the Insurance Experiment Group, Free for All: Lessons from the RAND Health Insurance Experiment, pages 17-26).
• Nyman, John A. (2007) Attrition in the RAND Health Insurance Experiment: A Response to Newhouse et al. (2007).
• Manning, Willard G., Joseph P. Newhouse, Naihua Duan, Emmett B. Keeler and Arleen Leibowitz. “Health Insurance and the Demand for Medical Care: Evidence from a Randomized Experiment.” American Economic Review, 1987, 77 (3), pp. 251-277.

Traditional instrumental variables (IV) econometric methodologies often fail to take into account response heterogeneity. Response heterogeneity based on characteristics not observed by the researcher can create a heterogeneity in the self-selection process. For instance, one group of people who elect to receive surgery may have knowledge of a family history where surgery is typically successful, whereas another group may elect not to receive surgery due to a different family history. If this information is unobservable to the researcher than an analysis of the average of effect of surgery may be biased. In the medical context, traditional IV assumes that:

1. treatment effects are constant conditional on observed characteristics, or
2. if treatment effects are heterogeneous, patients or physicians cannot anticipate these effects and use this information to select the most beneficial treatment.

In traditional IV, the treatment parameter gives researchers a local average treatment effect (LATE). But can a researcher characterize a heterogeneous response using IV? A solution to this problem is presented by Basu, Heckman, Navarro-Lozano and Urzua in a 2007 Health Economics paper. They use a local IV to estimate marginal treatment effect (MTE) parameters.

Basic Econometrics Review

Let us assume that a person will have two different outcomes based on whether or not they are treated:

• Y₁ = μ₁(X) + U₁
• Y₀ = μ₀(X) + U₀
• Δ = Y₁ - Y₀ = {μ₁(X) -μ₀(X)} + (U₁ - U₀)
• Y=μ₀(X)+D*{μ₁(X)-μ₀(X)} + {D(U₁ - U₀) + U₀}

The variable Y₁ represents the outcome if the person is treated and Y₀ represents the outcome if they are not treated. We only have one observation per person, however, since we cannot observe the counterfactual. If we could observe the counterfactual, Δ would give us the effect of the treatment for each person. Unfortunately we only observe Y. The dummy variable D is equal to unity if the person is in the treatment group and zero otherwise. If there were a randomized trial where people are randomly placed into the treatment and control groups, it would be easy to estimate the treatment effect by comparing the mean outcomes of the treated and control groups. We could examine the mean outcomes for individuals with similar characteristics to determine the treatment parameter by subgroup. However, if individuals can select whether or not to be treated, the error term–which may be composed of unobserved heterogeneity in the effectiveness of the treatment–may be correlated with the regressors that impact the outcome.

The traditional solution to the endogeneity problem is IV. Let X be the set of regressors and Z represent the instruments. “LATE computes the mean gain to those induced to switch from no treatment to treatment by a change in Z from z to z‘.”

• LATE={E(Y|X=x, Z=z‘)-E(Y|X=x, Z=z)} / {P(D=1| X=x, Z=z‘) - P(D=1| X=x, Z=z)}

Marginal Treatment Effect (MTE)

Developed by Björklund and Moffitt (1987) and furthered by Heckman (1997), the MTE measures “the average gain to patients who are indifferent between receiving treatment 1 [the treatment] versus treatment 0 [the control] given X and Z.” The benefit of using MTE is that one can calculate the marginal treatment effect for different subgroups based on the propensity score. This places a high degree of reliance on the accuracy and precision of the propensity score in order to determine these subgroup treatment parameters.

Let V denote a latent variable which measures the difference in benefits from being in the treated and control groups. Treatment choice can be modeled as follows.

• V= μ_v(Z,X) + U_v
• E(U_v)=0
• D=1(V>0)

The authors use a propensity score to determine the probability of selecting treatment.

• P(z,x)=P(D=1|Z=z, X=x) = P(U_v > -μ_v(z,x)) = 1 - F_Uv(-μ_v(z,x))
• F_Uv() is the cdf of U_v.

Now we can define MTE to be:

• MTE(x,z)=E(Δ|X=x, Z=z, V=0)
• =E(Δ|X=x, Z=z, U_v=-μ_v(z,x))
• =μ₁(x) -μ₀(x) + E{U₁ - U₀|U_v=-μ_v(z,x)}
• =μ₁(x) -μ₀(x) + E{U₁ - U₀|U_D= F_Uv(-μ_v(z,x))}

where F_Uv(U_v)=U_D. The last equation after the ‘|’ is a monotonic transformation of the terms after the ‘|’ in the third equation.

Local IV (LIV)

The LIV estimates the derivative of the expected outcome conditional on observed characteristics and the probability of electing to be in the treatment group, E(Y|X=x, P(z,x)), with respect to the probability of treatment, P(z,x). The term E(Y|X=x, P(z,x)) is defined as follows:

• E(Y|X=x, P(z,x))=E{ DY₁ - (1-D)Y₀ |X=x, P(Z,X)=P(z,x)}
• =μ₀(x) + P(z,x){μ₁(x) -μ₀(x)} + E{U₀|P(Z,X)=P(z,x)} + P(z,x){E{U₁-U₀ | P(Z,X)=P(z,x), D=1)
• =μ₀(x) + P(z,x){μ₁(x) -μ₀(x)} + K{P(z,x))

The term K(P(z,x)) is a general function of the propensity score, P(z,x). Often, K() will be a polynomial of the propensity score. The MTE can be computed mathematically as below:

• {∂E(Y|X=x, P(z,x)) / ∂P(z,x)} |_1-P(x,z)=U_D
• = μ₁(x) -μ₀(x) + ∂K(P(z,x))/∂P(z,x)

The equation above “…is implemented by regressing the outcome Y on all covariates [X], the propensity score, the interaction of the propensity score with all covariates and a polynomial on the propensity score.” This procedure is carried out in the paper empirically by applying these methods to data on breast cancer patients and their choice of breast-conserving surgery with radiation compared to mastectomy.

• Basu A, Heckman JJ, Navarro-Lozano S, Urzua S (2007) “Use of instrumental variables in the presence of heterogeneity and self-selection: an application to treatments of breast cancer patients,” Health Economics v(16), pp. 1133-1157.

Many people believe that better information technology (IT) can help improve the quality of medical care in the U.S. and around the world. For instance, if a doctor prescribes a drug which interacts harmfully with a drug the patient is already taking, a computer program could notify the doctor of this problem. If a patient has a certain disease, a computer program could help to narrow down the physician’s treatment options.

The problem with this top-down, one-size-fits-all doctoring is that every patient is unique. Of course, there are medical standards which should generally be followed, but often these standardized treatments have to be altered slightly or completely depending on patient wishes, the prevalence of multiple diseases, or economic constraints. Further, the physician typically has superior information regarding the patient’s condition than a standardized computer algorithm.

An NBER working paper by Javitt, Rebitzer and Reisman investigates a randomized trial of the implementation of a new IT system at a commercial HMO plan. The program was ‘turned on’ for some patients for each physician in the HMO and ‘turned off’ for other patients. The program would give three care consideration (CC) alerts: stop a drug, do a test, or add a drug. Each of these warnings had 3 severity levels.

The authors compare data on the CC alerts which occurred for the ’switched on’ patients with the hypothetical CC alerts which would have occurred if the software had been running. The authors found that the IT program lead to 6% lower charges in the turned on group compared to the control. The authors interpret this to be due to lower hospitalization rates and medication complications in the treatment group.

One key aspect of the program is that the CC alerts could be ignored by doctors.

“Physicians often have better information about their patients than does the error detecting software. Actions that look like a misstep to the computer may in fact be the result of informed physician choice, informed patient choice and/or patient non-compliance. For this reason, the HMO and the software company viewed CCs as recommendations that physicians were free to ignore if they disagreed.”

Physicians seemed to heed these CC alerts more than other studies had previously thought they would. Why?

Our results, however, suggest something more: the messages in the CCs seem to have had a bigger effect on physicians than the conventional medical channels used to promote the HOPE trial findings. It seems plausible that the CCs were influential because they linked a general recommendation (“take ACE inhibitors�?) to a specific patient [my italics] and a specific cite to the medical literature.”

To me, this seems to be the most appropriate way to implement medical IT.

• Javitt JC, Rebitzer JB, Reisman L, (2007) “Information Technology and Medical Missteps: Evidence from a Randomized Trial ” NBER WP #13493.

Many studies have revealed patient non-compliance with medical prescriptions (e.g.: not taking prescribed drugs, not visiting the doctor) at around 50%. Most medical researchers believe that non-adherence is either due to 1) irrationality or 2) misinformation. Yet a Health Economics article by Lamiraud and Geoffard (2007) tests the hypothesis of whether or not this behavior may in fact be rational.

Most medical treatments have costs. There are monetary costs, time costs, inconvenience costs, and physical costs from any side affects which occur. Physicians often focus on the benefits of medication without fully taking into account the discomfort patients experience from medication.

To solve this problem the authors used a data from a randomized controlled trial (RCT) of HIV-1 infected patients. The trial aimed to test the safety and efficacy of two HAART therapies. The authors use the following system of equations to estimate factors associated with compliance.

• θ^*_it=x_1itβ₁+ε_1it
• h^*_it=x_2itβ₂+ θ_itγ +ε_2it
• ε_jit=ν_jt+η_jit

The variable θ^* represents a latent variable as to the level of adherence (θ=1 for full adherence and 0 otherwise) and h* is a latent variable representing the health status of the individual (h=1 if the individual is an a good health status and 0 otherwise). The model is estimated using a panel non-linear simultaneous two-equation system.

The authors find that higher adherence levels are associate with higher patient welfare. Thus, if two drugs are equally effective, the authors recommend that policy-makers, insurance companies, and government should select the pharmaceutical with higher compliance levels. “To the contrary, the clinicians of the trial were tempted to support the treatment in which the adherence level was smaller, based on the rationale that a higher adherence would have pushed the efficacy to upper levels in that group.”

Realizing that adherence is an endogenous behavior made by rational individuals may change the way in which the benefits of prescription drugs are measured in the future.

• Lamiraud K, Geoffard PY (2007) “Therapeutic non-adherence: a rational behavior revealing patient preferences?” Health Economics, vol 16(11), pp. 1185-1204.

Revascularization (bypass surgery or angioplasty) have been frequently used procedures to treat patients who have experienced a myocardial infarction (MI). These procedure are expensive, but are supposed to enhance longevity. Do they?

This is the question analyzed by David Cutler in his NBER working paper titled “The Lifetime Benefits of Medical Technology.” The problem with many MI studies is that they are short term. Cutler uses data on Medicare beneficiaries who had a heart attack between 1986 and 1988. This means that Cutler can utilize 17 years of follow-up mortality data.

Another issue with non-randomized trials is that of patient selection. Very sick MI patients likely will not receive revascularization surgeries because they not be well enough to survive the surgery. Relatively healthy MI patients may not need revascularization. Thus, even the direction of the selection bias is unknown in this case. In order to account for this selection problem, the paper uses an instrumental variables approach.

The instrument is “the distance to the nearest revascularization hospital [defined as a hospital capable of preforming a revascularization] minus the distance to the nearby hospital of any type.” This instrument was used in a paper by McClellan, McNeil and Newhouse (JAMA 1994). In order for the instrument to be valid, “patients who are more likely to benefit from invasive treatments do not select their residential location based on distance to high-tech medical care.” Cutler argues that this likely true since most covariates are balance above and below the median differential distance. It is possible, however, that richer, healthier people live in more affluent areas and revascularization hospitals locate their facilities to attract these types of patients. I do not know whether or not this is the case. Also, “…if hospitals that provide revascularization are also better at providing aspirin at admission, at managing post-acute follow-up, or at treating subsequent illnesses years later, the instrumental variables estimates will overstate the importance of revascularization.”

Cutler does show that his instrument is relevant as “people who live closer to a revascularization hospital are 3 percentage points more likely to receive a revascularization procedure than those who live farther away.”

Results

The affect of revascularization on mortality is not clear.

“The marginal person receiving a revascularization is about 4 percentage points more likely to survive the first day after the MI than if the person did not receive a revascularization (although not statistically significantly). This gap narrows over time and even reverses by 1 year. At that time interval, people who received revascularization are 6 percentage points more likely to have died than people not receiving revascularization.”

After 17 years, people with MI have a mortality benefit of 5% but this result is not statistically significant. Cutler also examines the cost-effectiveness of revascularization procedures:

“…The greater survival for the marginal patients receiving revascularization translates into 1.1 years of additional life expectancy. The cost of this gain is about $38,000. Thus, the cost per year of life is $33,246.”

A year of life is generally valued at around $100,000, which might lead one to conclude that this is a worthwhile procedure. Since the mortality differential is not measured very precisely, however, one should be somewhat skeptical of this conclusions. Further, Cutler wisely notes that he is not sure whether or not the actual revascularization caused the decreased mortality. Being admitted to a revascularization hospital may simply be a proxy for the receipt of other hospital services, or it could be the case that revascularization hospitals provide superior patient management and patient care than non-revascularization hospitals. Also, due to data constraints, this paper only examines mortality effects. The impact of heart surgery on quality of life is not considered.

• Culter, David M. (2007) “The Lifetime Costs and Benefits of Medical Technology” NBER Working Paper # 13478.

A recent paper by Franco et al. (2007) claims that increased poverty may improve health (see also NPR’s Marketplace report). How is this possible? Lower income reduces excess food as well as cigarette consumption.  Further, poverty makes public transportation less affordable and individuals may substitute walking for taking the bus.  The authors study Cuba’s experience between 1989 and 2005.

Cuba has been subjected to an economic embargo by the United States since the 1960s. After the loss of the Soviet Union as a trading partner in 1989, Cuba entered a prolonged economic crisis known as the ‘‘Special Period.’’ The crisis worsened continuously over the next 5 years, with economic output reaching a nadir in 1995 of about half the level in 1990.

The decreased economic activity lead to the following changes:

• Calorie intake: “Average per capita daily energy intake…declined from 2,899 kcal in 1988 to 1,863 kcal in 1993.”
• Physical Activity: “In 1987, only 30 percent of the population living in Havana was characterized as physically active. In national data, approximately 70 percent of Cubans were considered physically active in 1991–1995, and 67 percent were active in 2001″
• Obesity: The prevalences of obesity in Havana were 11.9 percent, 5.4 percent, and 9.3 percent in 1982, 1994, and 1998, respectively. In Cienfuegos, prevalences were 14.3 percent, 7.2 percent, and 12.1 percent in 1990, 1995, and 2001, respectively, reflecting a 49 percent fall during the economic crisis.”
• Cigarette smoking: Cigarette smoking decreased over this period as well.

We see that the prevalence of many of risk factors declined during the “Special Period.”  What were the affects on health?

• “In subsequent years (1997–2002), rates of mortality from type 2 diabetes, coronary heart disease, and all causes dropped 51 percent, 35 percent, and 18 percent, respectively…No significant changes in total cancer mortality were observed, consistent with the current knowledge that obesity is not strongly associated with this condition.”

So poverty is the answer?  It turns out that all the news is not rosy:

• All-cause mortality among persons over the age of 65 years increased 13 percent from 1989 to 1996, primarily because of excess deaths from infections (21). The secular decline in infant mortality was interrupted for 3 years, and the incidence of low birth weight increased from 7.3 percent to 9.0 percent between 1989 and 1993 (21). An epidemic of optical and peripheral neuropathy attributed in part to vitamin and protein deficiencies affected 50,000 people between 1992 and 1993.”

Can these results be extended to other countries? Cuba is a communist country where health is provided publicly.  The government can incur debt in order to provide medical care for its citizens.  In the U.S., increased poverty will likely make medical care less affordable–for those not on Medicaid–and thus health outcomes may suffer.  Further, long run economic decline will make the provision of even government-run high quality medical care unaffordable for a society.  A final critique is that although the data presented in the paper is suggestive, correlation does not imply causation.  One should always maintain a healthy skepticism regarding the conclusion of time-series correlation studies.While no economist would advocate for policy-makers to attempt to increase poverty, the Franco study may guide individuals into believing that “less is more,” at least when it comes to food intake and car usage.

•  Franco M, Orduñez P, Caballero B, Tapia Granados JA, Lazo M, Bernal JL, Guallar E, Cooper RS (2007) “Impact of Energy Intake, Physical Activity, and Population-wide Weight Loss on Cardiovascular Disease and Diabetes Mortality in Cuba, 1980–2005” American Journal of Epidemiology, doi:10.1093/aje/kwm226.

“Medicare adopted its [Medicare as a Secondary Payer] MSP policy in 1982, effective January 1, 1983. This legislation states that for individuals working at firms with 20 or more employees, and otherwise eligible for Medicare benefits, Medicare serves as a secondary payer for health care expenses. The employer’s health insurance is the first payer. Because employer-sponsored health plans tend to be more comprehensive than Medicare, these workers are effectively foregoing their Medicare benefits by working. If these same individuals were not working, they would receive Medicare as their primary health insurance.”

A recent NBER working paper (”A Tax on Work for the Elderly: Medicare as a Secondary Payer“) claims that the MSP policy creates an implicit tax for elderly workers and thus creates disincentives to work. The authors calculate that the implicit tax is 15-20 percent at age 65 and increases to 45-70 percent by age 80. Making the Medicare the primary payer (MPP) will have two fiscal impacts. First, the cost of Medicare will increase. Since Medicare will be the primary payer, it will then be paying for more treatments. Secondly, enacting an MPP policy will decrease the implicit tax, increase the income and hours worked of elderly individuals and thus increase income tax receipts. The authors claim that since the second fiscal impact will dominate the first and thus recommend that an MPP system be implemented.

• Gopi Shah Goda, John B. Shoven, Sita Nataraj Slavov (2007) “A Tax on Work for the Elderly: Medicare as a Secondary Payer” NBER WP #13383.

Local governments provide a variety of services which are highly valued by their residents. From police protection to waste disposal, from snow plowing to utility meter reading, the local government is charged with providing the infrastructure necessary for a smooth functioning economy and a high level quality of life. But should local governments outsource these tasks to the private sector or provide the services themselves?

This is the question analyzed in an NBER working paper by Jonathan Levin and Steven Tadelis (earlier draft for free available here). On the one hand, using government provision is often inefficient due to the low incentives faced by city employees. On the other hand, cost relating to specifying and implementing performance requirements of a private supplier may also create inefficiencies. The authors claim the following:

• “If contracting difficulty m increases, the principal will be more likely to use an employment contract, while the optimal quality may increase or decrease. If the importance of quality s increases, the principal will be more likely to use an employment contract, and optimal quality will increase.

Services where quality or performance is difficult to measure are poor candidates for privatization. According to the authors’ data, the most difficult service to contract are police services, drug/alcohol treatment programs, inspection/code enforcement and fire prevention. The services where quality is easily seen are waste collection, utility meter reading, vehicle towing, and operation of recreation facilities.

Data

The authors use information from 1043 cities from the International City/County Management Association’s (ICMA’s) 1997 and 2002 Service Delivery surveys. Supplementary information is provided by the U.S. Census. To better understand which services are easily contractible, the authors conducted a survey of 23 city managers.

The authors use a multinomial logit regression in order to investigate which factors influence the probability that a city will privatize a given service.

Results

Here is what the authors concluded from their data:

1. Cities are more likely to privatize services for which it is easier to specify, enforce or adjust performance standard.
2. Cities are more likely to privatize services for which sensitivity to quality is low. As city residents are the final consumers of services, and city administrators are ultimately accountable to residents, this suggests that privatization should be less likely for those services where city residents are more likely to react to quality problems.
3. Smaller and rural cities may be less responsive to contracting difficulties. For instance, if the city simply does not have the capacity to provide the service, then they will privatize. On the other hand, if there are no business which will provide the desired service, then they local government may have to do it in-house.
4. The authors claim that city managers who have had experience working with private contractors are more likely to have fewer difficulties working with private firms and thus are more likely to privatize. While the learning argument is somewhat convincing, empirical, this may simply be the result that some cities are more ideologically in favor of privatization and others prefer more socialized services.
5. Cities run by mayors may be less likely to privatize services as compared to cities run by managers. The authors claim that politicians (mayors) may be are more responsive to quality in order to please their constituents but may care less about the cost.
6. Cities with strong government unions will be less likely to privatize.
7. If financial constraints cause administrators to focus more on economic considerations, then debt-constrained cities would be more responsive to contracting difficulties.

Extension

According to this model, it is difficult to tell whether or not medical care should be provided publicly or privately. Quality or performance is extremely difficult to judge and thus contracting based on performance (i.e.: P4P) is destined to fail. If a person dies after a surgery, is that the surgeons fault or was the person just in critical health? This may lead individuals to think that public provision is ideal. However, having a public sector surgeon who is less motivated and cares less about their reputation may lead to inefficiencies.

For vaccinations, quality is easy to judge–people are either vaccinated or they are not. This may lead individuals to think that vaccinations should be privatized to save money. However, it is difficult to contract the number of individuals who are vaccinated. For instance, if firms were paid based on the number of people they vaccinate they may have an incentive to vaccinate individuals even when they should not receive them (if they are sick and should not receive the vaccine until they recover) or the firm could inflate the number of people vaccinated by fraud.

Thus, whether or not medical services should be public or private is not so simple, and a case-by-case analysis may be needed.

Despite much public rhetoric, why is preventative and chronic care so poor in the U.S.? The easy answer is that patients switch plans so frequently that insurance companies who invest in preventative care will incur the cost, but not reap the benefits. The harder question is why patients are switching health plans.

According to a working paper by Cebul, Herschman, Rebitzer, Taylor and Votruba featured on Slate, the answer may be “search frictions.” In the paper, turnover is generated from two sources: 1) from employees leaving the company for new jobs and 2) by having the employer switch to a new health plan. Data from the Community Track Study in 1996/7, 1998/9, 2001 and 2003 show that average annual insurance cancellations are about 21%. More than one third of the turnover is caused by employers switching health plans. Small employers were more likely to switch insurance plans than larger employers. Why don’t they just stay with one plan?

The search friction model is developed from a labor economics paper by Burdett and Mortensen (1998). The authors argue persuasively that extending the model to the case of health insurance makes perfect sense.

“The market for health insurance is a natural place to expect search frictions. Health insurance is a complex, multi-attribute product and this complexity makes it difficult for clients to meaningfully compare more than a handful of proposals. Informal discussions with insurers suggest that they offer customers hundreds if not thousands of different policies. This complexity also makes the marketing of insurance costly so that companies can make only a limited number of appeals to employer groups in a period.”

The authors explain how the price friction mechanism works. The price of the insurance policy is p, the marginal cost of the policy is c, and the firm’s reservation price for buying insurance is p^R.

Suppose all firms made the same price offers p=c and so earned zero profit. Then one maverick firm could clearly increase profits by charging some discretely higher price (less than or equal to the reservation price p^R). This high offer would be rejected more frequently than the going price because any potential client who fielded more than one offer in a period would obviously reject the high offer. But on occasion the contacted client would have no other offers, and a policy would be sold. This would produce positive profit for the firm. Similarly, in a candidate equilibrium in which all firms were charging the same price (a price such that c<p<p^R ), a maverick firm could always increase profit by undercutting slightly the price charged by competitors, thereby increasing the number of clients while reducing by profit per client by only a trivial amount. In short, an equilibrium must entail a distribution of price offers.

Once market friction reach a sufficient level, in equilibrium we will observe a churning of employers going through different insurance policies each year. Introducing the issues that come along with adverse selection is likely to only increase market frictions because insurance companies now will want to screen employees.

Possible Solutions

The authors offer arguments made that may be solutions to the problem.

• Patient-financed health investments. Health care investments (i.e.: preventative care) should be financed by the client. This way, the person reaping the rewards from preventative care will also incur the costs. If the patient switches insurance plans, this will not be a problem since they will be eligible for lower premiums because of their preventative care history. On the other hand, determining what type of care is an expense and what is an investment may be very difficult practically. Further, shifting more risk onto the patient is the antithesis of what insurance is supposed to do. Finally, when switching insurers, it will be very difficult to verify the actual amount of preventative care received without a nation-wide standard for electronic medical records.
• Long term contracts. Another simple solution is just for employers to purchase long term contracts from insurance companies. This way, insurers will be able to reap more of their rewards from earlier years’ health investments. On the other hand, “given constantly-evolving medical technology and treatment protocols, as well as hard to predict changes in governmental regulation and mandates, it is difficult to see how long-term contracts might be implemented.”
• Price Caps. Government set price caps are probably the worse option. As Slate states, “But where should the government set the ceiling? If it’s too low, the government could end up destroying insurance companies’ incentives to stay in business at all.”
• Legislate a basic insurance package. The authors conclude the paper with the following: “It follows from this that much of the distortions resulting from frictions could be mitigated if there were a simple, easily understood and reasonably priced alternative insurance policy that would be available to all market participants. In the context of our search models, we believe we can prove that by making this alternative insurance available on a voluntary basis to all purchasers the inefficiencies resulting from search frictions could be greatly reduced.” Another option would be to offer everyone the choice of a nationalized health plan (a la Medicaid). People who did not want Medicaid, could choose to have vouchers (see Healthcare Vouchers) used to pay towards a private insurance plan of their choice. Many of the basic private insurance plans will likely mimic the nationalized Medicaid, but some plans will offer alternatives which will be more flexible and easier to adapt to new technology advances.

Many papers have attempted to calculate hospital efficiency before and after a policy change. Most research, however, does not take into account quality levels when analyzing these changes in efficiency. For instance, a doctor may increase the number of patients per hour that they visit by 50%, but if this is accomplished simply by lowering the quality of each visit, there may not necessarily be a quality improvement.

Pablo Aroncena and Ariadna García-Prado (2007) look at efficiency measures in Costa Rica after the implementation of management contracts with the local public hospital administrators. Unlike most papers, however, they attempt to take into account how quality plays a role in efficiency metrics.

Costa Rican Health Care System

In Costa Rica, health care is primarily provided by the government. In 2001, 83% of providers worked in the public sector. Approximately 77% of health care expenditures are public compared to only 23% private. The Ministry of Health provides regulatory oversight of the health care system and the Caja Costarricense de Seguro Social (CCSS)–Costa Rican Social Security Institute is in charge of public health care service delivery and financing. Most all physicians in public health care system are salaried and receive civil service status. Physicians and hospital managers receive little or no merit-based pay and thus have little incentive to improve performance.

The 1990s brought much restructuring of the Costa Rican health care system. For example:

• 1994: a population-based model of primary care was created. Basic health care teams–Equipo Básico de Atención Integrada a la Salud (EBAIS) were established to decentralize health care services and offer primary care and preventative services to the entire population.
• 1998: Decentralization Law. This law gave managerial autonomy for public hospitals.
• 2000: Management agreements fully implemented. “These contracts specified targets that managers pledged to achieve in a given time frame and were mainly deemed to get quality improvements and a better utilization of hospital resources.” Spain introduced similar management contracts in the 1990s and while hospital efficiency did increase, the measures were ineffective in primary care centers.

Data and Methods

The authors use Costa Rican public hospital data between 1997 and 2001. The authors define a productive process as follows:

• P(x)={(y,b) : x can produce (x,b)}

In the paper, x are inputs, y are good outputs, and b are bad outputs. The vector x consists of measures of MD and RN labor hours, the number of beds as a proxy for capital, and real monetary expenditures on goods and services. The good outcomes y are the number of discharges and number of outpatient hospital services used, while the bad outcome, b, is given by hospital readmissions.

The authors claim that hospital readmissions is a good proxy for quality. This may not be true. A sicker patient base may have more hospital readmissions. The authors do try to control for health status using DRGs and also their study examines changes in readmission rates so the baseline case mix of a hospital is not as important. Similarly, if a hospital has poorer patients, they may be less likely to comply with the doctors directives (either due to financial issues or misunderstanding the physician). If the case mix remains unchanged, however, this will not bias the results. A final issue noted by Arocena and García-Prado is that managers may have misreported re-admission rates in order to receive favorable reviews on their performance contracts. This problem is mitigated somewhat by the fact that the CCSS conducted audits of each hospital facility.

The distance function employs a nonparametric data envelopment analysis approach. The function used by the authors is:

• D^t(x^t,y^t,b^t,α)=min {λ: (λ^1-αb, λ^-αy) ∈ P^t(x)}

Results

The authors find that hospital efficiency increased significantly after the management contracts only in small hospitals. In large hospitals, there was neither an increase or decrease in efficiency. The paper believes that increased physician compliance with clinical protocols and better record keeping drove the improved performance for small hospitals. Peer pressure mechanisms are less effective in large hospitals, and the finding that absenteeism rates increased in large hospitals after the management agreements may show one reason why efficiency improvements did not increase in large hospitals.

Further, the authors actually find an efficiency decrease when quality is not taken into account. While the management contracts may not have strictly increased throughput, they do seem to have increased quality in small hospitals and the quality gains outweigh the quantity losses, at least in the case of small hospitals.

• Aroncena P, García-Prado A (2007). “Accounting for quality in the measurement of hospital performance: Evidence from Costa Rica,” Health Economics. Volume 16, Issue 7, Pages 667 - 685.

Should hospitals with long waiting times have higher or lower budget transfers? Offering hospitals who have low wait times more money will increase a hospital’s incentives to decrease wait times. On the other hand, thus policy may hurt the busier hospitals and may not alleviate the wait times of those who are waiting the longest. In the case of public school transfers, if the best schools are rewarded, this encourages achievement, but may punish the worst off kids (i.e.: those at poorer schools). Transfers to low-performing school may mute incentives to increase achievement.

The issue of hospital payment structure is analyzed by Luigi Siciliani in his article on optimal contracts B.E. Journal of Economic Analysis & Policy. As with any thesis which claims to give an optimum solution, this optimum is based on some assumptions. This paper uses four major assumptions.

1. Demand for treatment can be controlled by dumping some patients. Doctors can tell patients who wish to have medical treatment that they either a) don’t really need it or b) that they will not provide it
2. The purchaser (i.e.: NHS, an insurance company, Medicare, etc.) can not observe the number of people dumped.
3. Dumping is costly for the specialist. By dumping patients, the specialist receives more complaints about their service level. Thus, either the physician’s reputation is tarnished (a cost) or the physician must spent more time (another cost) convincing the patient that they do not need treatment.
4. Hospitals differ in potential demand for treatment, either due to the catchment area of the hospital or from having a better or worse reputation.

Another key assumption is that no co-payment charges can be issued. This assumption is plausible, because it basically represents the British NHS system. Thus, the optimal solution must be seen not as the ideal optimal, but as the optimal with a centralized payer and no co-payments.

The Model

Hospitals have parameter θ which describes the public hospital type. This parameter θ indicates potential demand in the absence of a rationing system.

For each treatment, patients differ in the value they would receive from treatment. For instance, healthy patients would not benefit from heart surgery, but individuals with coronary artery blockages likely would benefit from surgery. Thus the author assumes that individuals’ value from treatment is uniformly distributed between v₀ and v₁.

Patients have three options:

1. They can be treated at a public sector hospital after a wait of time w, [u^p(v,p)]={∫^T₀ v dt} -p=vT-p]
2. They can be treated in a private sector hospital with no wait, but pay a price of p, [u^NHS(v,w)]={∫^T_w v*g(w) dt} =vg(w)(T-w)]
3. or they can receive no treatment[u^none=0]

There are two costs to going to the public hospital. First, the individual has to wait w weeks longer, so they do not get to enjoy the benefit of the treatment for as extended a period of time. Secondly, since 0<g(w)<1, the treatment becomes less effective or less valued the longer the patient waits.

Thus, from the math above, we can see that a person will choose a public hospital if and only if:

• v<V(w)=p/{T-g(w)*(T-w)

The comparative statics show that longer wait times decrease the probability of using a public hospital, higher prices, p, decrease the probability of using a private hospital, and higher valuations, v, increase the probability of using a private hospital.

Demand for public hospital services is written as:

• D(θ,w,x)=θV(w)-x
• x is the number of patients who are dumped (i.e.: not added to the waiting list)

The number of treatment supplied by hospital θ is y(θ) and since supply must equal demand, we have:

• θV(w)-x=y(θ)

The authors claim that providers receive disutility from dumping patients. Also, hospitals receive more disutility when they dump patients who value the treatment more (i.e.: high v, this is more likely to be the sicker patients). Thus, we are lead to our first major conclusion.

• Conclusion 1: The patients who are dumped are the ones with the lower benefits from treatment. This means that hospitals dump the patients who don’t really need the treatment.

After some more math, the Dr. Siciliani states a second conclusion:

• Conclusion 2: A mix of explicit rationing (through dumping) and implicit rationing (through waiting) is therefore optimal. Siciliani explains that: “Rationing by waiting alone induces excessive disutility for patients. Rationing by dumping alone generates excessive disutility for the specialists.”

The author continues to conclude that a separating or pooling equilibrium may occur.

“Under symmetric information, the optimal contract is for the purchaser simply to over a transfer in exchange for the provision of the desired level of activity and waiting time, without leaving any rent to the provider…Under asymmetric information, we found that a separating equilibrium exists when it is optimal for the purchaser of health services to contract more activity and higher waiting times to hospitals with higher demand. In this case providers with low potential demand have an incentive to mimic hospitals with high potential demand. To induce hospitals to self-select, the purchaser needs to pay a rent to hospitals with lower potential demand. [But] if it is not optimal for the purchaser to contract more activity and higher waiting time to hospitals with higher demand, then a separating equilibrium may not exist.”

Problems

One main problem with the paper is that it assumes that patients with a high value, v, cost the same to treat as low value patients. If v is a proxy for sickness, this is likely not to be the case; sicker patients with a high v are more expensive to treat. If this were the case, then conclusion 1 would not hold. Public hospitals would instead treat patients with the lowest benefit and dump patients with intermediate benefits–the high benefit patients would still go to the private sector hospital.

Also, the paper does not take into account any strategic interaction between hospitals. “If hospitals with higher potential demand are contracted higher waiting times, then patients will switch from the hospitals with high potential demand to hospitals with low potential demand, increasing excessively the amount of dumping and consequent disutility for hospitals with low potential demand.”

• Luigi Siciliani (2007) “Optimal Contracts for Health Services in the Presence of Waiting Times and Asymmetric Information,” The B.E. Journal of Economic Analysis & Policy: Vol. 7 : Iss. 1 (Contributions), Article 40.

This is the question posed by a 2007 NBER working paper by Mary Beth Landrum, Kate A. Stewart, David M. Cutler. The authors use data from the National Long Term Care Survey (NLTCS) between 1994 and 1999. The data is panel in nature and has the benefit of combining Medicare administrative data with survey responses. The research examines individuals over 65 without disabilities in 1994 and see which patients became disabled and how this occurred.

Disability is defined by either:

• 6 specific ADL tasks (eating, getting in and out of bed, getting around inside, dressing, bathing, toileting)
• 8 specific IADL tasks (light housework, laundry, preparing meals, shopping for groceries, getting around outside, managing money, taking medications, using the telephone)
• 9 functional limitations (difficulty climbing a flight of stairs, walking across a room, bending to put on socks, lifting a 10-lb object, reaching above the head, using fingers to grasp and handle small objects, seeing well enough to read newsprint, speaking, and hearing)

Results

“Sixty-six percent of non-disabled respondents in 1994 survived and remained non-disabled to 1999, while 15.1% became disabled over the 5 year period and 18.9% died between survey waves. Six conditions made up about half of new disabilities (48%). These conditions were:

• arthritis,
• infectious disease,
• dementia,
• heart failure,
• diabetes, and
• stroke.

Only 17% of elderly respondents did not have one of these conditions and a majority (54%) had two or more. Of disabled respondents, 96% of newly disabled have at least one of these characteristics. Most disabled respondents had multiple conditions; two-thirds of newly disabled respondents had 3 or more conditions. The authors examine in more depth the interaction effects of having two or more conditions. For instance:

“…the interaction between diabetes and arthritis was positive, suggesting that these two conditions have synergistic effects such that having both conditions was more disabling than would be expected by the effects of each individual condition. In contrast, two interactions with dementia were negative (stroke*dementia and heart failure*dementia), suggesting that in the presence of a highly disabling condition like dementia, other conditions have effects that are dampened relative to what would be expected when the disease occurs in isolation.”

Hopefully, this data can be used to aid health service providers on how to better prevent and treat disabilities which occur in old age.

• Mary Beth Landrum, Kate A. Stewart, David M. Cutler (2007) “Clinical Pathways to Disability,” NBER WP #13304.

Physician scorecards have been a highly touted means to improve healthcare quality. One example is NY state’s coronary artery bypass graft (CABG) Surgery Reporting System. One side effect of scorecards is that surgeons may choose to operate on healthier patients in order to maximize their scorecard grade. In fact, over 60 percent of NY cardiothoracic surgeons reported refusing to operate on high risk patients on at least one occasion (Burack et al. 1999).

A paper by Glance et al. (HSR 2007) investigates whether or not high-quality cardiac surgeons are less likely to operate on high-risk patients. The paper uses a data set with over 57,000 patients treated by 189 surgeons. The authors first estimate a regression as follows:

• log[p_i/(1-p_i)] = α + Σβ_kx_kij + Σλ_jP_j

The equation above estimates the predicted probability of the ith patient treated by the jth surgeon with risk factors x_kij. The mortality of each patient if they are treated by the average surgeon is

• log[p_n/(1-p_n)] = α + Σβ_kx_ki

To determine the surgeon’s quality the authors used the observed to expected mortality rate.

Results

The paper finds that high quality surgeons actually treat higher risk patients. This finding is reassuring for those who favor medical scorecards. The authors due note some issues with the paper. For instance, one must be sure that the risk adjustment calculation is correct, and high-quality surgeons may be miscoding patients more frequently as high-risk. Another explanation could be that high quality surgeons have more experience and are closer to the end of their careers so they care less about scorecards and more about informal reputations. However, as young surgeons age and have been conditioned to believe that scorecards matter, this finding that high quality surgeons treat high risk patients may not hold in the long run. Nevertheless, the study is straightforward, clear and assuages some fear of patient selection by doctors operating under a scorecard system.

• Glance LG, Dick A, Mukame DB, Li Y, Osler TM. (2007) “Are High-Quality Cardiac Surgeons Less Likely to Operate on High-Risk Patients Compared to Low-Quality Surgeons? Evidence from New York State.” Health Services Research (OnlineEarly Articles). doi:10.1111/j.1475-6773.2007.00753.x

The popularity of specialty medical facilities (SMF) has increased over the years. The number of Medicare-certified ambulatory surgery centers (ASCs) has doubled to 3,371 during the past decade. A question remains: are these “Focused Factories” good for society?

In an article by Casalino, Devers and Brewster, (”Focused Factories…“) the authors try to answer this question.

What are SMFs?

At this point there are 2 types of SMFs: specialty hospitals and ambulatory surgical centers (ASCs).

• Specialty hospitals - According to the Community Tracking Study, most specialty hospitals are either “heart hospitals” or hospitals specializing in orthopedic surgery. They are generally joint ventures between physicians and national specialty firms or local hospitals, but a few are wholly owned by either physicians or by local hospitals.
• ASCs - Most ASCs are small and have four or fewer operating rooms. The most common services provided at ASCs are ophthalmology and gastroenterology.

Are they specialized medical facilities (SMFs) good for society?

Physicians who work at or own the SMFs claim that this type of care increases quality and reduces cost. Proponents of SMFs claim productivity increases due to specialization and the fact that there is less down time between procedures. Some hospitals say that these facilities engage in competition that is ‘unfair’, but if lowering cost and increasing quality is ‘unfair’ competition, mark me down as a fan of the unfair. The hospitals also argue that the SMFs are formed around the most profitable services and thus the hospital can not subsidize money-losing departments (e.g.: ERs, burn units, trauma centers). If the hospital is losing money on certain procedures, this does not mean that they should be making excess profits on other procedures, only that health plan compensation schedules should be altered.

The hospitals, however, are not all in the wrong. The SMFs have been accused of ‘cherry picking’ healthy patients. For instance, if physicians are reimbursed $5000 for a surgical procedure, the cost of preforming the surgery may be $2000 for a (relatively) healthier patient and $4000 for a relatively sicker patient due to increased likelihood of complications during surgery for the sicker patient. Thus, the hospital is shouldered with caring for the sicker patients and it may be more difficult for them to turn a profit. This solution to this is of course to make surgery payment in a risk-adjusted manner. In reality, risk adjustment is a delicate process which depends on many unobserved health variables so this solution may not be as easy to implement as it would seem in theory.

Another issue is whether SMFs create incentives for excess medical care. This is related to the problem of integrating the diagnostician of a problem and the treater of a problem (see 10 April 2007 post). If physicians own the SMFs, there may be an even larger incentive for them to recommend that their patients have invasive medical procedures since the physicians themselves often will profit not only from the labor compensation they will receive from preforming the procedure, but will receive additional income as return on capital from their investment in the SMF. Even physicians who do not treat patients and only diagnose them will have an incentive to recommend surgeries if they own a share of the SMF where the surgery would be preformed.

To counteract this problem, some politicians are considering bills which would “prohibit physicians from referring Medicare
and Medicaid patients to specialty hospitals in which the physicians have an investment.” While this is certainly a problem, the authors wise note that the negative aspect of these types of laws “…is that it would cause society to lose any advantages that might come from physician ownership and management of such facilities.”

So are SMFs good for society? At this stage it is difficult to say with certainty whether they are or not. Further investigation on this topic is certainly merited in the future. Any comments on your opinions regarding specialized medical facilities would be greatly appreciated.

• Casalino; Devers; Brewster (2003) “Focused Factories? Physician-Owned Specialty Facilities” Health Affairs, 22, no. 6, pp. 56-67.

The 2005 Deficit Reduction Act aimed to reduce Medicaid entitlements as one means to reduce the national deficit. One portion of the bill changed eligibility rules for the elderly to qualify for Medicaid payment of nursing home (NH) expenses. In the May edition of the Journal of Health Economics, authors David Grabowski and Jonathan Gruber estimate whether or not restricting (or expanding) Medicaid eligibility rules significantly impact NH utilization.

Who pays for nursing homes?

There are generally 4 institutions which pay for nursing home (NH) care.

1. Medicare. Medicare pays 100% of the first 20 days of NH stays after a hospitalization, and a portion of NH costs between days 21-100 after a hospitalization.
2. Long-term care (LTC) insurance. Private LTC insurance is purchased at younger ages and protects against the risk of needing NH care. LTC insurance is still fairly rare; only about 4% of NH expenditures are paid by private LTC insurance plans.
3. Out of pocket payments.
4. Medicaid. Medicaid will generally pay for NH care if the individual meets certain income and asset requirements. Generally, the individual needs to qualify for Supplemental Security Income (SSI) or be in a state which allows the individual to “spend down” income in order to qualify for Medicaid.

Analysis

The authors seek to see how loosening Medicaid restrictions will impact NH usage. Expanding Medicaid income eligibility requirements, of course, will mechanically make more individuals eligible for Medicaid, but it may also make Medicaid more attractive since one does not need to “spend down” such a large portion of their assets or income. Loosening Medicaid restrictions can also affects supply. Higher Medicaid reimbursement rates will likely increase the percentage of patients who are Medicaid patients (vs. private insurance or out-of-pocket patients), but may also increase the total number of nursing home patients.

Complications

The analysis does have some complications. Certificate-of-need (CON) laws restrict the ability of NH supply to adjust to changes in NH demand. Also, the population affected by changes in eligibility are the middle class; poor individuals (always eligible) and rich individuals (never eligible) are not impacted by marginal changes in eligibility requirements. One also has to worry that Medicaid NH patients receive worse care than private insurance NH patients, but Grabowski et al. (NBER 2006) shows that this is not the case. Finally, the availability of care given by family or friends is not measured in this study and would likely impact NH utilization.

Data and Results

The data used is the National Long-Term Care Survey (NLTCS) from 1982, 1984, 1989, 1994 and 1999. The data have information on individual demographics, health, as well as income and asset levels to determine Medicaid eligibility. The econometric specification uses a probit regression with time, marriage, and state dummies included.

The authors generally find that loosening Medicaid eligibility rules has a limited impact on Medicaid NH usage. Further, the “spend down” rule that many states have adopted also does not significantly affect Medicaid NH utilization. The authors conclude: “we find consistent and clear evidence that nursing home utilization is inelastic with respect to state policies. Thus, the large increase in nursing home expenditures over the past few decades is not likely attributable to increased generosity in state Medicaid payment programs.”

• Grabowski; Gruber (2007) “Moral hazard in nursing home use” JHE, vol 26, issue 3, pp. 560-577.

A recent paper in the Health Services Research journal (”Hospice…“) looks at whether hospice care reduces hospitalizations for elderly terminally ill patients in nursing homes. In the introduction, authors Pedro Gozalo and Susan Miller state that there are two main implications which result from end-of-life hospitalizations:

“At the patient level, hospitalizations of frail NH [nursing home] residents have been shown to include hazards that negatively affect the quality of life (Creditor 1993) and, in many cases, are inappropriate (Saliba et al. 2000). At the policy level, hospitalizations represent the main component of total health care costs, particularly during the last few months of life. In an recent study using both Medicare and Medicaid claims for NH decedents in the state of Florida in 1999, Miller et al. (2004) found hospital expenditures to account on average for 78 percent of all expenditures in the last month of life among those patients that did not receive hospice and 33 percent among NH residents who had any hospice in the last 30 days of life.“

The main problem in determining whether or not hospice care reduces hospitalization is the issue of selection. Individuals who enroll in hospice care may be relatively ‘healthier’ than those who are hospitalized without hospice care. Also, it could be the case that hospice patients are ones that prefer less aggressive treatment methods. Local market idiosyncrasies can also lead to erroneous conclusions.

Econometrics

To take these selection issues into account the authors use an inverse-probability-of-treatment weighting (IPTW) [see