A data mining analysis of pathological findings in 11259 placentas

Abstract

To systematically review the pathological findings in placentas, we identified a data set of 11259 cases from 1996 to 2007 in our database. The diagnoses, along with available clinical information, were extracted from the reports by text mining, and parsed into a relational database. The associations among various epidemiologic and pathologic features with clinical outcomes were examined by data mining analysis. The maternal age at delivery ranged from 12 to 49 years (mean 24). Eight diagnostic categories included 26 items, such as no significant pathological change (55%), infarct (16%), villitis (15%), and accelerated maturation (9%). Statistically significant findings (p<0.0001) included: increased maternal age since 2001; increased percentage of cases with pathological findings in women ≥30 years; prematurity associated with age > 26 years, obliterative fetal vasculopathy, inflammation, infarct, thrombosis, marginal umbilical cord insertion, and previous preterm birth; accelerated maturation associated with inflammation, obliterative fetal vasculopathy, marginal umbilical cord insertion, but negatively correlated with calcification. Plotted by maternal age versus gestational age, inflammatory processes and circulatory problems clustered in distinct areas. In conclusion, this study showed that data mining is a valuable approach to discover hidden patterns in large clinical data sets.