The HC-REMA project aims at meeting the needs of health care professionals who want to use decision-analysis for performing both cost-effective choices and time-continuous quality control. The proposed software architecture is based on the following main components.
1. A framework designed to assist the decision-maker in building decision-analytic models, that is to define the cost and effectiveness components for a given problem. The result will be a model in which each component is 'well-documented', i.e. it has a sound ontological definition, and it may be linked with data sources for its own quantification and 'learning': for example, let us suppose that ``population life expectancy'' is node of a belief network; it could have a link to the survival table of a given country, where life expectancy can be found as a function of sex, age and race. As another example, let ``disease prevalence'' be another node of that belief network: it could be initially given a prior by the medical expert, and could be linked to a growing data base that will provide evidence for updating such a prior at the light of new data about the disease.
2. An interface that will translate the models in a format suitable as input to commercial software tools, such as Hugin, which will perform decision-analysis computations on them.
3. An interface to both local and remote databases, which will be support both model construction and model learning and updating.
4. An interface supporting knowledge acquisition and data entry, as well as presentation of results in a 'standard' format.
Each component will use telecommunication networks for acquiring information and delivering results to other tools or to the user.
The starting point of the HC-REMA project is the consideration that it is possible to individuate both general structures of models of economic evaluation on different classes of problems and general structures of models for cost evaluation in different contexts. The precise specification of these general structures will make the development of economic evaluation models more rigorous - e.g. it will become more difficult 'to forget' some important predictive variable - and will encourage the presentation of results according to standard, predefined formats.
For example, the problem of the economic evaluation for alternative therapeutic actions could be represented by the generic model given in Fig.1.
Squares represent alternative choices, circles represent general categories of medical and economic entities which should be taken into account in modeling a specific problem by instantiating domain/institution specific entities and, finally, diamonds represent objective functions to be optimized.
As a matter of fact, every problem can be characterized in terms of both cost and efficacy components, but these vary depending on the domain. For example, instrumentation purchase implies to take into account cost components such as capital investment, management costs, and amortization, while the cost components for a screening program include items such as the cost of the screening tests, and eventual therapies carried out as a consequence of the screening results. The measure of benefits also tends to vary from one domain to another, even though all programs in health care aim to improve life expectancy and/or the quality of life of the population. The reason for these variations is that in general some surrogate outcomes are considered, which are easier to measure and predict, such as the increased number of examinations performed with a new instrument, the number of cancers found by screening, etc. Temporizing costs and benefits is another important point, given that a discount must be computed for future costs and benefits, but 'how to discount' is one of the controversial issues in this field.
The basic components of the HC-REMA project architecture are shown in Fig.2 Upper and lower part of the figure represents the architecture of the economic evaluation and cost evaluation tool, respectively.
The greatest assistance for decision-makers in building such models comes from an ontology server containing a formal description of model components. It will be unique, maintained in a responsibility center, which will be accessible for exploitation through networking facilities. A common ontology will guarantee that the produced models will be easily shared among different users as well as user-tailored in the appropriate contexts, for example by adopting a different quantification of some parameters.
For what concerns economic evaluation tool, according to the problem class a meta-model is formulated. The specific model (i.e. a model addressing a specific problem and obtained by instantiating the corresponding meta-model) may represent knowledge through different formalisms. We will begin by considering the two most widely used formalisms for decision analysis, namely decision trees and influence diagrams. Existing software packages will be exploited in order to implement the 'inference engines' of the models (i.e. to make computations according to decision analysis theory). Thus, it is essential to provide either an interface that will translate the developed models in a format suitable for the package or a direct link among the package itself and the ontology and terminology servers. The choice will depend on whether or not the source code of the decision analysis tools will be available. Luckily, one of the project partners is the producer of the best known package for managing belief networks and influence diagrams, and the software is available for Unix and DOS operating systems.
For what concerns cost evaluation tool, the same type of architecture will allow to represent meta-models for describing organizations, financial and economic processes, etc, in order to build computational models for calculating costs.
Another essential component of the framework will be the graphical user interface. Its design will take into account the fact that the HC-REMA tools will have to be integrated within an existing health care information system (if any). For this reason the HC-REMA tools will be based on one of the available tools for building graphical interfaces which are portable on different hardware platforms and make use of standard semantics.
A mandatory integration will be with those tools that compute effectively sustained costs: for example, to calculate the cost of a hospitalized patient, it is necessary to store, day by day, detailed information about prescribed drugs (dosage, timing, plasma concentration monitoring, etc), tests performed, nurses and physician time devoted to the patient, etc. These computations are then used to compare 'real' costs with the reimbursementd obtained according to the DRG (diagnosis related group) classification. It is clear that results obtained from such a tool will be useful also for individuating situations that could benefit from a cost-effectineness analysis. The following subsection will detail the project of one of such tool.
Finally the HC-REMA tools will interface to both local and remote databases, to support both model construction and model learning and updating. Particularly useful would be to have links to data repositories storing clinical trials' information and meta-analysis results. This kind of repositories are becoming more and more diffused, mainly through Internet, and are being developed by national and international institutions, such as the Cochrane Collaboration. As another simple example, summary statistics about life expectancy by age, sex, and race in a given country, is an information useful for all cost-effectiveness studies in which survival is one of the effectiveness measures.
Cost evaluation
DRGs are a patient classification system relating the type of in-patients treated (case-mix) to his/her resources intensity and therefore to the costs incurred by the hospital fro treating the patient. Each in-patient is reimbursed to the hospital with a fixed amount related to the DRG group. Therefore, it is essential that the hospitals should be able to evaluate the DRGs costs, either to have objective charges definition criteria and to estimate the effects of this kind of financing on the budget. The evaluation of the DRGs costs, makes essential the use of automatic procedures for surveying all the diagnostic procedures and the surgical or therapeutic treatments delivered to the in-patients as well as the quantity of sanitary material consumed during each stay in hospital.
We propose a software application for evaluating the costs incurred by the hospitals for treating the in-patients. The software should represent a framework to monitor the resources intensity and the utilization of services of each hospitalization in a hospital setting. It, also, should support the clinical staff in the daily activity by producing a series of reports (therapeutic plans, exams orders, etc.) and by carrying out some relevant administrative tasks (data coding for patients discharge, DRG classification, discharge reports generation, etc.). To this purpose, it appears important that either a Hospital Information System (HIS) as well as an Administrative Information System (AIS) should be available in order to manage the hospital accounting and to obtain all the financial data that allow to compute the total DRGs costs. The correct health services costings and their auditing on behalf of the hospitals represent the basic element for an efficient hospital management. But a difficult problem is, certainly, the definition of an appropriate costs computation model in order to make the hospitals able to evaluate correctly the costs sustained in treating the in-patients and to verify the performance of the rates imposed by the authorities.
The software should allow the collection of all the health services delivered to each in-patient (from the admission to the discharge), as well as the drugs and the sanitary matter supplied. Moreover, it should manage any information (to say ICD9 diagnoses and complications codes and ICD9-CM procedures codes) needed to assign each discharged patient to a DRG class. The software will produce the hospital total amount invoiced on the basis of the rates imposed, as well as the evaluation of the costs actually sustained.
To the costs calculation task contribute both the information coming from the administrative offices (central level), and from the hospital sections and units (departmental level). The information stored at departmental level (number and type of health services supplied, drugs administered, matter consumed, number of days of hospitalization) joined to the administrative data (health services and drugs or matter costs) are used to calculate the costs incurred by the hospital for treating the in-patients.