EC9790 Health Economics May Examinations 2022/23Java

Java Python EC9790

Health Economics

May Examinations 2022/23

Section A: Answer ONE question

1.   Answer both (a) and (b).

(a)  A certain disease has two treatment alternatives. The cost of Treatment A is 1200 pounds while the cost of Treatment B is 2400 pounds. A doctor must decide which treatment to prescribe to a patient.  If the doctor chooses Treatment A, the probability that the treatment will succeed is 0.5, and if they choose Treatment B, the probability that the treatment will succeed is 0.75. If the treatment is successful (under each of the two alternatives), then the patient gains 20 units of QUALY (i.e., Quality Adjusted Life Years) and if the treatment fails, then the patient will gain only 4 units of QUALY.

i)    Assume that the doctor is an employee of an HMO that provides treatment at no cost to all its enrollees.  If the doctor’s objective is to maximize each patient’s expected QUALY, which of the two treatments will the doctor choose? (6 marks)

ii)   Will the doctor’s behavior lead to a problem of moral hazard for the HMO?   (6 marks)

iii)  Assume now that the doctor sees numerous patients and their objective is to maximize the sum of the expected QUALY of the patients being treated.  Further assume that the HMO sets a “fixed budget” of 24,000 pounds for each doctor treating this specific disease. That is, once the total cost for treating the doctor’s patients for this disease reaches 24,000 pounds, the doctor cannot accept any additional patients with the disease. Which of the two treatment alternatives will the doctor choose and is there a moral hazard problem in this case? (6 marks)

iv)  Assume now that some of the patients suffering from this disease do not respond to Treatment A (i.e., if they are treated using this alternative, the treatment will fail) and the doctor can identify these patients in advance. Will the HMO’s policy described in section iii above create a problem of adverse selection? (6 marks)

v)   Assume again that all patients are identical and that, for each patient, if the doctor chooses Treatment A, the probability that the treatment will succeed is 0.5, and if they choose Treatment B, the probability that the treatment will succeed is 0.75. However, assume now that the doctor is not an employee of the  HMO but rather, they work as a subcontractor for the HMO and they (the doctor) has to bear the entire cost of treatment themself.  Suppose that the doctor can choose between two contracts. Under the first, they are paid 1200 pounds for each patient that they treat. Under the second, they are paid 800 pounds if the treatment fails and 1400 pounds if it succeeds. The doctor’s utility from treating each patient is given by:

U(Q, F, C) = 100Q + (1600 + F − C)1⁄2

where Q is the QUALY of the treatment’s outcome, F is the reimbursement the doctor receives from the HMO and C is the treatment’s cost. Which of the two contracts will the doctor choose if their objective is to maximize expected utility?  (8 marks)

(b) Below are three different mechanisms that have been applied, in different countries around the world, in order to improve the efficiency of their health care systems. For each of the above mechanisms explain in not more than five sentences how they were supposed to improve social efficiency and provide an explanation of why they have not been so successful.

i)   Prospective payments. (6 marks)

ii)  Co-payments. (6 marks)

iii)  Quality measures. (6 marks)

2.   Answer both parts (a) and (b).

(a)  A large number of hospitals around the world are compensated according to the DRG (Diagnosis Related Group) model. According to this model, a list of diagnoses is determined, and a price is determined for each one. When a patient arrives at the hospital, the doctors determine the "main” diagnosis of their medical problem, and the hospital is reimbursed according to the price assigned to that diagnosis (regardless of the length of the hospitalization or the procedures performed during it). Please answer the following questions on how to determine the diagnosis groups and their prices:

i)   One possibility for determining the diagnosis groups is to classify all diseases into a relatively small number of "diagnoses” such that each one includes various conditions of the same disease or even several relatively similar diseases. Examples of diagnoses might be “Prostate Cancer”, “heart problems”, etc.  An alternative is to classify all diseases into a large number of diagnoses where at least some of the DRGs will also include a procedure. Examples of diagnoses might be "heart problem with catheterization", "heart problem with drug treatment", etc. Using the ideas discussed in class, compare the advantages and   disadvantages of the second method relative to the first. [Your answer should be no longer than half a page] (10 marks)

ii)   Another option is to classify diseases in a more fine-grained way, that is, not according to the main procedure called for by the diagnosis, but rather according to the degree of severity of the problem (for example, "severe heart problem", "moderate heart problem", etc.).  Compare the advantages and disadvantages of this method relative to the first one described in part a (i.e., the coarse-grained method). [Your answer should be no longer than half a page] (10 marks)

iii)   In some countries where the DRG system is used, the hospital also receives additional compensation for "exceptional" cases. Consider the following (“retrospective”) mechanism: If the cost of the treatment for a certain patient exceeds the compensation the hospital will receive according to the relevant DRG, then it will receive additional compensation at the rate of 50% of the amount by which the cost exceeds the DRG payment. What are the advantages and disadvantages of this method relative to the method with no retrospective payments? [Your answer should be no longer than half a page] (10 marks)

(b)  For each of the statements below, please state whether it is “Right” or “Wrong” and briefly (in no more the 3-4 sentences) explain your answer:

i)   “If demand for health services is not perfectly inelastic, competitive health insurers may not provide full insurance.” (5 marks)

ii)   The more “predictable” or “predictive” a service is, the smaller are the plans’ incentives to provide it. (5 marks)

iii)   “Pay for performance mechanisms are socially efficient as they affect both the demand for high-quality sellers and the incentives to improve quality.” (5 marks)

iv)   “A major benefit of bundled payment is the ability to target incentives to the   group of physicians primarily responsible for the care delivered over a defined episode.” (5 marks)

< EC9790 Health Economics May Examinations 2022/23Java b>Section B: Answer ONE question

3.   During 2013, some US states expanded Medicaid coverage (which provides health care access at very low cost) to some low-income childless adults. Researchers used a survey representative at the state level to assess the effects of this policy.

(a)   The figure above plots the share of survey respondents with a certain characteristic.

i)    Describe the figure (2 marks)

ii)   What does the figure suggest about whether costs might be a barrier to health care for some people?  (2 marks)

iii)  Can you conclude that this relationship is causal? Why yes or why not? (3 marks)

(b)   Table 1, Panel 3 reports the impact of the policy on preventative care measures (in rows) by subgroup of respondents (in columns). Each cell reports about a separate regression.

i)    Describe the table (2 marks)

ii)   Why is it important to compare Column 8 to Column 10? (2 marks)

iii)  Can we interpret these results as causal? Why yes or why not? (3 marks)

iv)  How do these results compare with what you know from the literature about the price-elasticity of healthcare? (4 marks)

(c)    Table 1, Panel 4 reports the impact of the policy on risky behaviours by subgroup of respondents. 

i)   Describe the table (2 marks)

ii)  Why would we expect access to health insurance to affect risky behaviours? Name two mechanisms and relate them to the model of risky behaviours discussed in class (5 marks)

(d)   Based on findings about the price-elasticity of preventative care, an employer decides to randomly offer a gym membership to some employees. The table below show results from the experiment.

i)      Describe the table (2 marks)

ii)     Does the table support the hypothesis that gym costs are a barrier for employees? Why yes, why not? (5 marks)

(e)   The experiment was an encouragement design: individuals in the treatment group were encouraged to sign up. In order to sign up, they had to complete a series of screening and activities. The table below shows the probability that individuals randomly assigned to the treatment group completed these activities based on their observable characteristics. 

i)   Describe the table (2 marks)

ii)  Does lower baseline productivity cause higher probability of completing activities? Discuss (3 marks)

iii) Can the patterns in this table explain the findings under part (d)? (3 marks)

(f)    Based on your answers to the previous questions, is the following statement true or false? Discuss.

“Government and employer subsidies for health insurance and wellness programs are always welfare improving” . (10 marks)

4.   Between February 2004 and September 2006, Facebook was rolled out across US colleges in a staggered fashion.

(a)   Figure 1 above shows the impact of the introduction of Facebook at different colleges on students’ mental health as measured by surveys conducted at the time.

i)  Describe the figure (2 marks)

ii)  Can we interpret these patterns causally? Why yes or why not? (3 marks)

iii)  What could be the concern about the use of survey data in this case? What do we need to assume about survey responses for the results to be valid?  (3 marks)

(b)   Researchers recruited social media users for an experiment.

To participate in the study, people had to install an app, only available for Android phones, that records which app is being used at any time. Does this requirement concern you for the internal validity of the study? And for the external validity? (3 marks)

(c)   The figure below shows the distribution of answers to a baseline survey (prior to any experiment) of participants.

i)  Describe the figure (2 marks)

ii)  Can we conclude that social media are welfare decreasing from this figure? Explain your answer. (3 marks)

(d)   Participants were randomly assigned to a control group, a bonus group, and a limit group. The bonus group received an announcement in period 2 of the study that indicated they would get a bonus if they used social media less than a given number of hours in the next period (period 3). No incentive was available for periods 2, 4, 5.  The limit group was given access to an app that allows to set limits on next-day social media use. Users would then receive notification when their social media use got close to the limit, but were able to snooze away the limit (with some delay) once it hit. Figure 4 below shows social media use in these groups relative to the control group. 

i)  Describe the Figure. (1 marks)

ii)  What do the temporal patterns in social media use for the bonus group show about digital addiction? Relate your answer to the model of addiction seen in class.  (5 marks)

iii)  What can explain the effect of the limit treatment? (3 marks)

(e) The table below shows the results from a regression of wages at age 50 on indicators for depression at ages 27-35 with different controls.

 

Robust standard errors in parentheses. All columns include individual, family and environmental controls.

Individual controls include age, gender dummy, race dummies, dummies for education at age 26, a dummy for marital status in 1992, work experience by 1992, the average percentage of weeks the person’s work history data is missing, dummies for health status during childhood, a dummy for smoking behaviour in 1992 and missing data indicators. Family controls include dummies on parental education and missing data indicators.

Environmental controls include year dummies, local unemployment rate in 1992, 1998, 2004, and the year outcome variables is collected, and missing data indicators.

i)  Describe the table (2 marks)

ii)  In column 2, the authors add controls for depression episodes between 35 and 50 years of age to test whether persistence of depression can explain the productivity result. What can you conclude from this test? Is this a valid test? Why yes or why not? (5 marks)

iii)  In column 3, the authors add controls for work experience after the first depression diagnosis (human capital accumulation). Is this a good control? Can it be affected by early depression? (5 marks)

iv)  Are the patterns in this table causal? Why yes or why not? (3 marks)

(f)   Based on your answers to the above questions and what you have learnt in class, discuss the following statement.

“Banning social media would be welfare improving