Factors associated with age of death in sudden unexpected infant death.

AIM: This study aimed to systematically analyse the pregnancy, birth and demographic-related factors associated with age of death in sudden unexpected infant death (SUID). METHODS: Data were analysed from the Centers for Disease Control and Prevention's Cohort Linked Birth/Infant Death data set (2011-2013; 11 737 930 live births). SUID was defined as deaths from sudden infant death syndrome, ill-defined causes, or accidental suffocation and strangulation in bed. There were 9668 SUID cases (7-364 days; gestation >28 weeks; 0.82/1000 live births). The odds of death at different ages were compared to determine which variables significantly affect the SUID age of death. RESULTS: Forty-three features indicated a significant change in age of death with two main patterns: (a) younger chronologic age at death was associated with maternal smoking and factors associated with lower socio-economic status, and (b) older age was associated with low birthweight, prematurity and admission to the neonatal intensive care unit. However, when age was corrected for gestation, these factors were associated with younger age. CONCLUSION: Factors that varied with age of death are well-documented risk factors for SUID. The majority of these risk factors were associated with younger age at death after allowing for gestational age at birth.

Identifying smokers with a medical extraction system.

The Clinical Language Understanding group at Nuance Communications has developed a medical information extraction system that combines a rule-based extraction engine with machine learning algorithms to identify and categorize references to patient smoking in clinical reports. The extraction engine identifies smoking references; documents that contain no smoking references are classified as UNKNOWN. For the remaining documents, the extraction engine uses linguistic analysis to associate features such as status and time to smoking mentions. Machine learning is used to classify the documents based on these features. This approach shows overall accuracy in the 90s on all data sets used. Classification using engine-generated and word-based features outperforms classification using only word-based features for all data sets, although the difference gets smaller as the data set size increases. These techniques could be applied to identify other risk factors, such as drug and alcohol use, or a family history of a disease.