1445. Optimal Policies for Cancer Screening Under Budget Constraints
Invited abstract in session TB-21: Healthcare Analytics, cluster Healthcare Management.
Tuesday, 10:30-12:00Room: FENH201
Authors (first author is the speaker)
| 1. | Susana Mondschein
|
| Universidad de Chile | |
| 2. | Felipe Subiabre
|
| Universidad de Chile | |
| 3. | NATALIA YANKOVIC
|
| ESE Business School, Universidad de los Andes, Las Condes, Chile |
Abstract
We develop a general framework for the formulation and evaluation of public policies for cancer screening, allowing a national healthcare program to compare the costs and benefits of a set of policies, and select one, optimal for a given measure of effectiveness. This framework also extends to other non-contagious, preventable diseases.
For this purpose, we consider a base model for the natural history of a specific cancer, calibrated to accurately represent the appearance and evolution of a tumor for various risk groups. We use it to build a public policy model for the optimization of an expected quantity across the whole population, e.g. maximizing the total life years or minimizing the total probability of death from this cancer, considering a global budget constraint representing a maximum expected expense per person.
The optimization problem decides over a set of policies for each risk group and includes in its cost-benefit assessment the crucial trade-off between more frequent and costly screening tests versus diagnosing later-stage tumors, with higher treatment costs and risk of death. The model also includes tests with associated risks and false positive and negative results, and its analysis is based on studying how the stationary distributions of semi-Markov chains change due to each policy.
We evaluate our model on two different sets of policies, one corresponding to a fixed testing frequency for each risk group and another to increasing testing ages, showing under general conditions the optimal global policy assigns more aggressive and expensive testing to groups at higher risk, in accordance with intuition.
By applying this framework to some of the main cancers affecting modern society, after selecting adequate base models from the CISNET database of models, we evaluate the current total costs incurred by healthcare programs for these cancers and find better policies according to our analysis, without a change of expected total expenditure.
This evaluation on realistic models and data should also lead to a classification of cancers into three main groups: (i) preventable according to the models, (ii) non-preventable according to the models, and (iii) unclear due to model uncertainty or discrepancy. Also importantly, we can evaluate how the realistic development of cheaper or better exams and treatments could effectively and positively affect this classification.
Keywords
- Healthcare Service Delivery
- Service Science
- OR in the Public Sector
Status: accepted
Back to the list of papers