The probability of developing cancer at any point in the lifetime of an American is 0.4566 and the probability of dying of cancer is 0.2150. With the current state of affairs about 64,500,000 people in America will die of cancer. Cancer, however, is curable if detected and treated early. Thus, it is our objective to allocate scarce resources efficiently throughout the life-cycle to save more life-years, or alternatively, reduce costs. To this end we use data obtained from SEER (Surveillance, Epidemiology, and End-Results) ...
Read More
The probability of developing cancer at any point in the lifetime of an American is 0.4566 and the probability of dying of cancer is 0.2150. With the current state of affairs about 64,500,000 people in America will die of cancer. Cancer, however, is curable if detected and treated early. Thus, it is our objective to allocate scarce resources efficiently throughout the life-cycle to save more life-years, or alternatively, reduce costs. To this end we use data obtained from SEER (Surveillance, Epidemiology, and End-Results) provided by the National Cancer Institute (NCI). From here we can obtain probabilities of dying of all other causes by age, sex, race; probability of developing cancer by age, sex, race; survival probabilities conditional on cancer stage, and most parameters. The model itself is a sequential Markov model in which the state-space is composed of an always observed state, and underlying state, and an observable state. The action space is to test or not test for cancer every period, which would reduce consumption but would increase the likelihood of being alive and consuming in the future. Agents receive a shock each year in which they can die from all other causes except the cancers in question. If they survive, they receive a cancer shock. If they get cancer, we will follow through the development of the cancer in a Markov process. If the agent has cancer and decides to test and there is no misdiagnosis, the agent is treated and has a probability of survival depending on the cancer stage, and age of the agent. Some agents whose cancers are detected early enough will survive to die of other causes; others will survive a few more periods due to treatment. We then compare a testing policy to what would happen if agents did not test at all and we can obtain the cost per life-year saved of the policy in questions and the average utility achieved. As compared to the existing literature, our model and program allow us to try policies with varying lengths of time between testing. Since the policy set in question is very large and the problem by the nature of the state-space cannot be done recursively, we evaluate a subset of the policy set of at least several hundred thousand policies. This approach complements what has been done already in the medical side of the literature. Clinical-cohort studies with a control group and experimental group would have at most 5-years time span and would provide information on one policy. Other simulation based models that exist are usually Markov trees that usually compare three policies each of which has the same length of interval between tests (i.e. starting at age 40 test every 3 years, 4 years, 5 years). Thus, our approach innovates in the modeling itself using the 3 part state-space from the dynamic programming perspective. The model allows for nonlinear policies, and since it is written in Fortran it allows us to evaluate a greater number of policies with more agents that could be done with an application as has been done previously. We apply the model to colon and rectum cancer to obtain the optimal ages to perform colonoscopies. The currently recommended guideline for colonoscopy is to test at ages 50, 60, 70, 80 and 90. Our Monte Carlo simulation yields a cost/life-year saved of $54,919.62 for the official policy. We ask the following: ff we could reallocate 4 tests between the ages of 51 and 90, which ages for testing would give the lowest cost/life-year saved? Of the 91,309 possible policies 43,026, or 47%, have a better cost per life-year saved than the currently recommended guidelines. The policy with the lowest cost is to test at the ages of 50, 67, 71, 75, 78 with a cost...
Read Less
Add this copy of Optimal Cancer Screening: Life-Cycle Monte Carlo to cart. $98.00, good condition, Sold by Bonita rated 4.0 out of 5 stars, ships from Newport Coast, CA, UNITED STATES, published 2010 by LAP LAMBERT Academic Publishin.
