Clinical knowledge base development for preterm-birth risk assessment

Abstract

Listen

Translate article

The director of the National Center for Nursing Research (NCNR) challenged nurse scientists to develop substantive content in nursing knowledge domains and called for NCNR funding priorities to include both prevention of low-birth-weight infants and information system technologies (Hinshaw, 1989). To address this challenge, this study was the first in an ongoing program of informatics research in which artificial intelligence technology, called “machine learning,” was used to develop and describe a knowledge base for preterm-birth risk assessment. This research should improve outcomes for childbearing families through the development of a knowledge base that provides improved decision support for nurses' preterm risk assessment. Accurate assessment of preterm-birth risk will permit intervention with appropriate educational programs, bed rest, and early symptom management to prolong gestation. Review of the literature found no conceptual or theoretical models of preterm birth risk. The lack of conceptual and theoretical models of preterm-birth risk may account for poor reliability and validity of existing preterm-birth risk screening instruments (Nunnally, 1978). Existing preterm risk screening instruments include factors that are not valid predictors of preterm-birth risk and fail to include factors reported in the literature that may be valid predictors of preterm birth (Alexander, Weiss, Hulsey, & Papiernik, 1991; Lockwood et al., 1991; Woolery, 1992). Although existing instruments are not adequately predictive of preterm risk, current preterm birth prevention programs use these invalid, unreliable tools to intervene with pregnant women on a daily basis. The outcomes where a women is incorrectly assessed “low risk” for preterm birth may result in preterm birth and delivery of an infant that dies or that requires expensive high technology to survive. The outcomes where a women is assessed “high risk” but does not actually manifest preterm birth may result in extra prenatal visits and more contact with perinatal nurses and physicians to provide education and assessment. However, the expense involved in extra prenatal visits is minimal compared with the costs of high technology and infant morbidity and mortality, resulting in a trend that increasingly treats all pregnant women as if they were “high risk” for preterm birth. Solutions to this problem may be found using machine-learning technology.