Health Insurance Treatment and Outcomes Discussion

Health Insurance Treatment and Outcomes Discussion ORDER NOW FOR CUSTOMIZED AND ORIGINAL ESSAY PAPERS ON Health Insurance Treatment and Outcomes Discussion Think about the Doyle (2005) paper and answer two questions. Health Insurance Treatment and Outcomes Discussion 1. Methodologically, how did the author use information about auto insurance to tease out the causal effect of health insurance? Why was that important? 2. What was your reaction to the result? Did it surprise you? Responses should be 1 – 2 pages, neatly f ormatted with double spacing, 12 point font, and 1” margins. Do not use outside sources, just write everything in your own language. doyle_2005.pdf HEALTH INSURANCE, TREATMENT AND OUTCOMES: USING AUTO ACCIDENTS AS HEALTH SHOCKS Joseph J. Doyle, Jr.* Abstract—Previous studies find that the uninsured receive less health care than the insured, yet differences in health outcomes have rarely been studied. In addition, selection bias may partly explain the difference in care received. This paper focuses on an unexpected health shock—severe automobile accidents where victims have little choice but to visit a hospital. Another innovation is the use of a comparison group that is similar to the uninsured: those who have private health insurance but do not have automobile insurance. The medically uninsured are found to receive 20% less care and have a substantially higher mortality rate. I. Introduction E FFORTS to reform the $1.3-trillion U.S. healthcare industry are often motivated by a concern that the uninsured are denied access to healthcare (NCHS, 2002). Indeed, the uninsured are less likely to pay for care, and theory suggests that they would receive less treatment. Previous empirical work supports this conclusion and finds that the uninsured receive approximately 40% less healthcare than the insured.Health Insurance Treatment and Outcomes Discussion Though treatment differences have been labeled the “access gap,” they can be difficult to interpret. First, health outcomes are rarely compared. Without knowing the effect on health, treatment differences may imply that the insured receive too much care, rather than the uninsured receiving too little. For example, moral hazard problems may allow physicians to practice flat-of-the-curve medicine, where diminishing returns to medical care imply small health gains with additional treatment (Enthoven, 1980). Though the insured may receive more care, this may not translate into health differences. Further, there may be selection problems. Individuals decide to purchase insurance, and they decide to seek medical care. If those with a low risk of using medical care are also less likely to purchase insurance, then treatment differences may have little to do with access to care. This positive correlation between risk level and insurance coverage is a standard result in contract theory and suggests that information problems in insurance markets may lead to treatment differences. The measurement of these differences is further complicated by the lack of an adequate comparison group to test whether unobserved differences between Received for publication September 2, 2003. Revision accepted for publication June 15, 2004. * MIT Sloan School of Management I am grateful to two referees, Daron Acemoglu, Joshua Angrist, Gary Becker, Jon Gruber, Jonathan Guryan, Willard Manning, David Meltzer, Casey Mulligan, Jim Poterba, Tomas Philipson, and especially Steve Levitt, Mark Duggan, and Michael Greenstone for helpful comments and suggestions. I would also like to thank the Center for Health Systems Research and Analysis for graciously supplying data. Health Insurance Treatment and Outcomes Discussion This research was supported by a grant from the Social Science Research Council’s Program on Applied Economics. All remaining errors are my own. the insured and uninsured drive the difference in care received. To examine health outcomes and consider selection problems, this paper focuses on an unexpected health shock— severe automobile accidents where incapacitated crash victims have little choice but to use professional medical care. Using a unique data set that links police accident reports to hospital discharge records, treatment levels and mortality rates are compared between the insured and uninsured. These data provide a rich description of each accident and offer a new way to investigate whether the uninsured are denied access to life-saving healthcare. One innovation in the paper is the use of a comparison group that is similar to the uninsured in observed risk-taking behavior, income, vehicle, and injury characteristics: drivers who have health insurance but do not have automobile insurance. The focus on severe automobile accidents has four main advantages: First, police record mortality, which allows a comparison of health outcomes. Second, contact with professional healthcare providers is virtually automatic for severe accidents, because ambulances arrive as part of the crash investigation. It is then possible to compare patients who were injured in a similar way and immediately sent to the hospital. Third, severe accidents are largely unexpected at the time of the health insurance purchase decision, as opposed to the chronic conditions that are usually studied (Brown, 1998). Patients live with chronic health problems and gain private information about future healthcare use.Health Insurance Treatment and Outcomes Discussion These are cases where adverse-selection problems are at their most extreme, as those likely to use healthcare may be more likely to purchase insurance. In contrast, if the decision to purchase insurance is not based on the likelihood of being in a severe automobile accident, then the approach taken here will largely avoid this selection problem. Finally, automobile accidents are a particularly important health problem for the uninsured. Most health problems are common in older individuals, who are almost universally insured. In contrast, the uninsured tend to be younger, and automobile accidents are the leading cause of death among those under the age of 35 (CDC, 1998). The results suggest that the uninsured receive 20% less treatment than the privately insured, controlling for personal, crash, vehicle, neighborhood, and hospital characteristics. The uninsured are also found to have a substantially higher mortality rate—an increase of 1.5 percentage points over the mean mortality rate of 3.8%. The structure of the paper is as follows. Section II briefly presents theoretical considerations and a review of previous The Review of Economics and Statistics, May 2005, 87(2): 256–270 © 2005 by the President and Fellows of Harvard College and the Massachusetts Institute of Technology HEALTH INSURANCE, TREATMENT AND OUTCOMES work to guide the interpretation of the results and place them in context. Section III describes the data and presents mean comparisons of treatment, mortality, and crash characteristics between the insured and uninsured. Section IV presents the empirical estimates, where withinhospital variation is used to test the effect of insurance status on treatment and mortality. Section V discusses a number of specification and robustness tests, including an examination of differences in procedures performed. These tests suggest that the ability to pay of patients, and not unobserved health differences, is driving the differences in treatment and mortality.Health Insurance Treatment and Outcomes Discussion Section VI considers the implications of the findings, and section VII concludes. II. Background At the time of an accident, insurance lowers the price of healthcare faced by the insured and raises the effective price that providers collect. Both factors suggest that the insured will receive more treatment. The overall effect of insurance on treatment decisions, and ultimately health outcomes, is a combination of these supply and demand factors.1 This paper considers serious crashes where the decision to visit the hospital is largely out of the control of the patient, as police or witnesses call for medical professionals. Once in the hospital, the uninsured (and family members) may help decide on treatment, but for the most complicated decisions, such as emergency brain surgery or spinal fusion, the decisions will likely be made on the supply side. The effect of reimbursement rules on treatment decisions has received considerable attention, begun by McGuire and Ellis (1986). In these models, physicians choose treatment levels to maximize an objective function that includes both profits and the benefits of treatment to patients. The models suggest that an increased ability to pay, through insurance for example, would result in greater treatment. This is partially due to moral hazard problems, though it does not depend on profits being part of the objective function. If profits were incorporated into a resource constraint, where insured patients subsidize the uninsured, then the prediction of more treatment for the insured would remain. Another factor affecting the supply response is whether insurance status is in the provider’s information set at the time treatment decisions are made. In fact, it appears that one of the first pieces of information discovered is insurance status. Health Insurance Treatment and Outcomes Discussion Medical professionals search patients to find not only insurance information, but medical and emergency contact 1 Adverse selection and moral hazard suggest that the insured will receive more care as well. In addition, agents will differ with respect to tastes and income. Perhaps the most interesting difference for this current study is that individuals with a higher degree of risk aversion may be more likely to purchase insurance and drive safely (de Meza & Webb, 2001; Jullien, Salanie, & Salanie, 1999). Though this is not a general result, it suggests the potential for a negative correlation between insurance coverage and health risk. The potential for heterogeneity in risk preference is considered below. 257 information as well. Insurance companies regularly recommend that consumers carry their insurance card especially for this type of situation. Once known, the expected payer is reported on the patient’s chart. When treatment decisions are made, insurance status is likely known. Finally, treatment differences may be smaller because of the legal ramifications of refusing to treat patients. For example, federal law mandates that hospitals stabilize trauma patients. Though hospitals have an incentive to transfer patients that are a higher financial risk, little evidence of such transfers is found following severe crashes, as discussed below. Regardless, treatment differences are predicted, as long as hospitals have some discretion over treatment levels. A. Previous Literature Many previous studies find that the uninsured receive less care than the insured, but few studies consider the effect on health.Health Insurance Treatment and Outcomes Discussion 2 The most extensive study was the Rand Health Insurance Experiment, where over 2,000 families were randomized to health plans that varied by copayment level (Newhouse, 1993). The study found few effects of free coverage on health, though a slight improvement in blood pressure was found. Currie and Gruber (1997) also consider a case where the use of a hospital is less of a choice: childbirth. Using exogenous variation in Medicaid expansions, lowereducated women were found to receive more treatment with greater access to insurance. Only minor effects on neonatal mortality were found, unless the mother lived near a hospital with a neonatal intensive care unit (NICU). For these mothers, the expansions had a large effect on neonatal mortality. The 24% increase in eligibility from 1987 to 1992 was estimated to result in an 11% decrease in mortality. This interaction with NICUs suggests that insurance does have an effect on the most costly, potentially life-saving care. It appears, then, that the hospitals chosen by the uninsured may also affect treatment differences. In a related paper, Currie and Gruber (1996) find that the 15.1% increase in eligibility for all groups over the period 1984–1992 was associated with an 8% decrease in child mortality. A few studies in the health literature also consider outcome differences and find mixed evidence of insurance effects. For example, Ayanian et al. (1993) study breast cancer patients in New Jersey and find that the uninsured are more likely to suffer from serious conditions, and have a 49% higher mortality rate. Sada et al. (1998) considered 2 See for example, Currie and Thomas (1995), Haas and Goldman (1994), Long, Marquis, and Rodgers (1998), Marquis and Long (1994), Spillman (1992), and Tilford et al. (1999). Brown (1998) reviews the health literature, and Levy and Meltzer (2001) review the economics literature. In a related literature, moral hazard problems in workers compensation have been studied, with reported injuries seemingly affected by changes in insurance generosity (Dionne & St-Michel (1991); Fortin & Lanoie, 1992). 258 THE REVIEW OF ECONOMICS AND STATISTICS heart attack patients and find that the uninsured were half as likely to receive angiography, but find no difference in mortality. Health Insurance Treatment and Outcomes Discussion Haas and Goldman (1994) consider trauma patients in Massachusetts and find that the uninsured are 32% less likely to receive surgery and are twice as likely to die. Some of the problems with these earlier studies can be dealt with by the focus on automobile accidents, and the use of linked police and hospital data. First, this paper uses within-hospital variation to explicitly control for hospital resources and incentives. Second, previous papers naturally relied upon physician diagnoses and procedures to categorize patients by injury severity.3 These categories are likely endogenous, however, as diagnoses and procedures are likely affected by insurance status if the uninsured receive less care. One advantage of the linked police data is that police measures of accident severity are less likely to be influenced by the victim’s health insurance status. Instead, patients who were involved in similar crashes are compared in an effort to control for injury severity. For example, a patient whose crash on a rural highway resulted in severe vehicle damage will have greater injuries than a patient who crashed on a slower-moving urban street. Further, the choice of when to visit the hospital, even for cases such as heart attacks, can be associated with insurance status and affect the interpretation of the results. This paper instead focuses on a case where the decision of when to visit the hospital is not made by the patient, and offers a new way to investigate the effect of insurance status on treatment and health. III. Data Description The Crash Outcome Data Evaluation System (CODES) offers a unique data set to compare patients following an automobile accident. CODES links police accident reports with hospital discharge records, using identifiers such as patient name, birth date, and time of accident. The National Highway Traffic Safety Administration subsidized individual state efforts to link these two sources of data, and, until now, they have only been used to study highway safety (NCSA, 1998). This paper uses data from Wisconsin for the period 1992 through 1997.4 In Wisconsin, all police accident reports are submitted to the Wisconsin Department of Transportation, and all inpatient hospital records are submitted to the Wisconsin Department of Health and Family Services.Health Insurance Treatment and Outcomes Discussion Analysts at the Center for Health Systems Research and Anal3 Haas and Goldman (1994) show that the emergency case mix differs substantially between the insured and uninsured, though they do not control for injury type beyond a control for penetrating injury, such as gunshot wounds. 4 Wisconsin is one of the original CODES states and the only state that regularly supplies a public-use data set. All 23 CODES states were contacted. Most of the other states are still in the process of linking the data or did not have key variables of interest. ysis (CHSRA), located at the University of Wisconsin at Madison, linked the data and calculated that 80% of all crash-related hospitalizations were linked successfully. Reasons for linkage failure include the following: the crash victim died at the scene, the patient was transported out of Wisconsin, the crash record contained insufficient identifying information, or the crash was not reported to the police. For the severe accidents investigated here, police reports are more likely. Though it is not possible to compare linkage rates by insurance status—such status is only known for linked cases—CHSRA data documentation argues that “there is no reason to expect that the cases not linking are different from cases which do link.” For the same reason, mortality rates at the scene of the accident cannot be compared across insurance groups, so all mortality comparisons are analyzed for crash victims who arrived at a hospital. A. Police Measures The police-report data offer a rich description of each accident. Characteristics such as age, sex, seat location, seat belt use, entrapment in the vehicle, and injury severity are available. Wisconsin Police report injury severity according to the KABCO score: killed; A, B, and C injuries consisting of incapacitating injuries, nonincapacitating injuries, and possible injuries; and other or unknown. At the accident scene, a police officer records the injury category. If the crash victim later died due to the accident, then the injury severity was scored as a K—even if the death occurred after hospital discharge.5 Total hospital facility charges increase with police-measured injury severity, suggesting that the evaluations are informative. Health Insurance Treatment and Outcomes Discussion The linked police data also provide a novel comparison group: drivers who have health insurance but do not have automobile insurance. This information is recorded because it is illegal to drive in Wisconsin without having such insurance. Other crash characteristics include vehicle damage, the time of the accident, alcohol involvement, road condition, population size where the accident occurred, and the types of crashes, vehicles, and roads. The police also record the vehicle identification number (VIN) for crash victims in cars and trucks. These VINs were decoded by Primedia Price Digests to categorize vehicles by engine size, manufacturer’s suggested retail price, two-door status, vehicle weight, and model year. These measures are used to control for automobile quality, which can affect 5 The scores of K are the raw data for the often used Fatality Analysis Reporting System (FARS) distributed by the National Highway Traffic Safety Administration. It is generally regarded as “thirty-day mortality,” meaning that deaths up to thirty days from the accident are recorded in the police data, though deaths after thirty days are also recorded. For a small number of cases the patient died in the hospital and the KABCO score does not equal K. These have been coded as fatal injuries in the analysis; the results are not sensitive to this definition. As noted above, those who die at the scene do not enter the inpatient data. HEALTH INSURANCE, TREATMENT AND OUTCOMES injury severity and serve as a proxy for socioeconomic status. The ZIP code of residence is also included in the report.Health Insurance Treatment and Outcomes Discussion This information allows a comparison of neighborhood characteristics such as the fraction of the population in poverty, race categories, educational categories, and median household income, available from the 1990 U.S. Census. Patients from wealthier backgrounds may receive favorable treatment and are more likely to have insurance. Controls for neighborhood characteristics provide a way to isolate the effect of insurance on treatment and health outcomes. B. Hospital Measures The hospital discharge data also provide a rich set of variables. These include age, sex, and ZIP code of residence, in addition to measures of treatment, such as total facility charges, length of stay, up to three procedure codes, and up to five diagnosis codes. The total facility charges represent the standard room and procedure charges as opposed to the amount billed to insurance companies or the government. This is particularly useful for research purposes, as charges are uniform across insurance plans within each hospital. Facility charges also approximate the cost of care. Audits performed by Medicaid and Medicare suggest that the cost of the treatment is roughly 70% of charges, but items that are excluded from facility charges, such as physician fees, roughly offset this adjustment.6 Major diagnostic categories are also used to compare patients with similar types of injuries, such as nervous system or musculoskeletal. … Get a 10 % discount on an order above $ 100 Use the following coupon code : NURSING10

Read more
15% OFF your first order
Use a coupon FIRST15 and enjoy expert help with any task at the most affordable price.
Claim my 15% OFF Order in Chat

Good News ! We now help with PROCTORED EXAM. Chat with a support agent for more information