Industrial and Systems Engineering

Clearly, Lm reflects the dimensions of alveoli but is not a direct estimate of alveolar diameter or mean alveolar size. The functional significance of Lm lies, however, in its inverse relationship to the surface-to-volume (S/V) ratio of pulmonary airspaces. The S/V ratio determines the density of elastic recoil forces resulting from surface tension, as well as the surface available for gas exchange per unit volume of ventilated lung (1, 2). These are important correlates of Lm that are particularly relevant to understanding how emphysema affects lung function.

A case-mix classification system for nursing home residents is developed, based on a sample of 7,658 residents in seven states. Data included a broad assessment of resident characteristics, corresponding to items of the Minimum Data Set, and detailed measurement of nursing staff care time over a 24-hour period and therapy staff time over a 1-week period. The Resource Utilization Groups, Version III (RUG-III) system, with 44 distinct groups, achieves 55.5% variance explanation of total (nursing and therapy) per diem cost and meets goals of clinical validity and payment incentives. The mean resource use (case-mix index) of groups spans a nine-fold range. The RUG-III system improves on an earlier version not only by increasing the variance explanation (from 43%), but, more importantly, by identifying residents with "high tech" procedures (e.g., ventilators, respirators, and parenteral feeding) and those with cognitive impairments; by using better multiple activities of daily living; and by providing explicit qualifications for the Medicare nursing home benefit. RUG-III is being implemented for nursing home payment in 11 states (six as part of a federal multistate demonstration) and can be used in management, staffing level determination, and quality assurance.

The role of dementia and other mental disorders in nursing home case-mix classification systems has been an area of controversy. The role of mental dysfunctions was considered in developing a new case-mix measurement system for facility payment in a national demonstration to understand staff time use in nursing homes. Nursing staff (nurses and aides) time and resident assessment data were collected for 6,663 nursing home residents in 6 states. Measures of signs and symptoms of cognitive impairment (dementia), depression, and delirium were created based on items from the new National Minimum Data Set. These measures then were used to determine whether mental dysfunctions were predictive of resource use (nursing staff times and costs) when controlling for other case-mix variables. Cognitive impairment was associated with slightly higher staff time only in less physically-impaired residents without serious medical conditions and not receiving heavy rehabilitation. Similarly, depression and delirium were associated with higher resource use only in selected types of residents. Based on these findings, the new Resource Utilization Groups Version III (RUG-III) contain a major category of residents who are cognitively impaired but not severely dependent in Activities of Daily Living. Depression is used to differentiate subgroups of residents with major medical conditions such as hemiplegia and aphasia. Delirium, when used together with other resident characteristics, was not found useful in explaining resource use. Case-mix groups defined by mental dysfunctions can foster improved care, but careful consideration must be given to appropriate incentives and documentation requirements for providers.

The vast amounts of chemical and biological data available through robotic high-throughput assays and micro-array technologies require computational techniques for visualization, analysis, and predictive -modeling. Predictive cheminformatics and bioinformatics employ statistical methods to mine this data for hidden correlations and to retrieve molecules or genes with desirable biological activity from large databases, for the purpose of drug development. While many statistical methods are commonly employed and widely accessible, their proper use involves due consideration to data representation and preprocessing, model validation and domain of applicability estimation, similarity assessment, the nature of the structure-activity landscape, and model interpretation. This chapter seeks to review these considerations in light of the current state of the art in statistical modeling and to summarize the best practices in predictive cheminformatics.

The vast amounts of chemical and biological data available through robotic high-throughput assays and micro-array technologies require computational techniques for visualization, analysis, and predictive -modeling. Predictive cheminformatics and bioinformatics employ statistical methods to mine this data for hidden correlations and to retrieve molecules or genes with desirable biological activity from large databases, for the purpose of drug development. While many statistical methods are commonly employed and widely accessible, their proper use involves due consideration to data representation and preprocessing, model validation and domain of applicability estimation, similarity assessment, the nature of the structure-activity landscape, and model interpretation. This chapter seeks to review these considerations in light of the current state of the art in statistical modeling and to summarize the best practices in predictive cheminformatics.