The purpose of this proposal is to develop the use of text mining and data mining tools to investigate the relationship between physician practice and health outcomes. In the absence of specific guidelines, there remains substantial variability in physician practice for patients with similar diagnoses. It is possible to investigate that variability to find an optimal practice pattern. Since much of the information concerning patient outcomes is written as chart notes, text mining must be used to extract useful intelligence from those notes.
Aim I. To develop an algorithm to extract useful information from clinical records. This includes abstract information from databases, physician notes in patient records, and text data electronically recorded in pharmacy, nursing, and care management databases. This will enable to automatic extraction of meaningful intelligence from patient charts when the current method is primarily by using coders who perform manual extraction.
Aim II. To apply the developed tools to longitudinal data collected during pharmacist-consultant review of patient databases to determine the extent of polypharmacy in patients in long-term and hospital care who were treated for cardiovascular conditions and other co-morbid diseases. Standard statistical techniques cannot generally be used to examine polypharmacy because of the complexity of the data. Text mining can be used to automatically relate similar medications into meaningful clusters so that the relationship between drug interaction and medical outcome can be examined.
Aim III. To explore the relationship between treatment and disease. Data mining tools can be used to examine the relationship between physician treatment and quality. Text mining can be used to categorize descriptive measures of patient quality.
Aim I V. To provide a series of workshops to present the results to medical researchers to demonstrate the benefits of data mining techniques in analyzing clinical outcomes and to build a program of medical internships for students enrolled in the recently approved PhD program in Industrial and Applied Mathematics. The PhD program has a mandatory application requirement and internship.
| Cerrito, Patricia B; Pecoraro, David (2005) Visits to the emergency department as transactional data. J Healthc Manag 50:389-95; discussion 396-7 |
