Mining a spinal cord injury clinical database for nursing information: A source of nursing knowledge

This is a descriptive exploratory analysis of a spinal cord injury (SCI) database from a mid-western Veterans' Administration (VA) hospital. The computerization of practice data elements including diagnoses, interventions and patient acuity provides the opportunity to explore the domain of SCI nursing practice from an actual practice perspective. Data mining techniques are used to determine if there are trends, patterns, or relationships in nursing diagnoses and nursing interventions that provide useful information that can contribute to knowledge about this specialty nursing practice. The theoretical framework of the study is Systems Theory and a model is presented to illustrate how data is processed and organized to become information and levels of knowledge.; The study population of 525 unique patients represents 1107 hospital admissions and 74,047 days of hospital care. Thirty-four percent of this veteran population is identified as service-connected for a spinal cord injury with most common level of injury at the C4-6 level resulting in quadriplegia. Frequency of admissions and lengths of stay reinforce the position of persons with SCI as outliers in the healthcare system. Eleven years of data reviewed indicate a stable pattern of nursing diagnoses and interventions in this setting. The 4750 diagnostic labels in the database are shown to represent 161 nursing diagnoses that are clustered into 20 diagnostic categories. These categories are used as variables in the development of four models of neural nets used to determine their predictive power as related to LOS. All four neural net models indicate that the diagnostic categories achieve over a 77% accuracy rate in the prediction of LOS. An awareness of the stable patterns of clustered nursing diagnoses and nursing interventions in this database can be useful in planning for resource allocation, competency identification, and orientation to this specialty nursing practice.; Study results identify as issues in nursing database research the need for the inclusion of nursing data in hospital information systems, the need for standardized language in data capture, and the need to include nursing data in data warehouses. Nursing data must be used to generate information that validates the impact of nursing care, supports nursing research, and ensures that nursing has a voice in the development of healthcare policy.