Prediction in medicine by integrating regression trees into regression analysis with optimal scaling.

OBJECTIVES: A new data-analysis strategy is proposed to solve the problems of selecting interaction terms in linear regression on the one hand, and of statistically testing the significance of regression trees on the other hand. METHODS: The proposed strategy combines two data mining techniques: regression trees and regression analysis with optimal scaling (CATREG). The method traces small regression trees using the bootstrap and integrates the results as interaction variables (called "trunk variables") into CATREG. RESULTS: An application to data from cardiac patients shows a relative increase of 19% variance accounted for (16% cross-validated variance), by the CATREG model including the trunk variables compared to the model excluding these variables. CONCLUSIONS: This study indicates that trunk variables can be useful to model interaction effects in prediction problems.

Health status development in a cohort of preterm children.

OBJECTIVE: To determine the impact of preterm birth on health status (HS) development at the ages of 5 and 10 years in a cohort of children born before term. SAMPLE: Six hundred eighty-eight children, born in 1983 with a gestational age of <32 weeks and a birth weight of <1500 g. DESIGN: Prospectively collected HS variables, obtained from the parents, were analyzed in a longitudinal perspective by using principal component analyses. RESULTS: One third of the sample had minor to severe HS problems at both ages of measurement. One third had problems on one assessment only. The remainder of the sample had no HS problems at either age. The analyses grouped the HS variables into 3 combinations. Problems in basic functioning, such as mobility or speech, decreased with age. Negative moods substantially increased, and concentration problems increased slightly. Specifically at risk were preterm born children with handicaps, boys, and children who were small for gestational age. CONCLUSION: According to the parents, one third of the cohort had no HS problems at either age. The pattern of HS problems of the preterm born children changed between 5 and 10 years of age.

Empirically based subgrouping of eating disorders in adolescents: a longitudinal perspective.

BACKGROUND: Successive DSM versions struggle with the heterogeneity of the eating disorders. Criteria were mainly based on clinical impressions and on descriptive and inferential studies. METHOD: In a study of 55 eating-disordered adolescents, we investigated whether patients could be grouped on an empirical basis, using principal components analysis (PCA) with optimal scoring (scaling), i.e. PCA with no a priori assumptions. Clustering was based on Morgan-Russell subscales, each measured four times over the course of illness. RESULTS: Contrary to DSM-IV criteria, patients did not cluster primarily on the basis of anorectic symptoms; the occurrence of bulimic symptoms was more dominant. Core symptomatology (preoccupation with food, disturbed body perception and inadequate sexual behaviour) did not differ between patients, either at referral or over time. CONCLUSIONS: These results support the spectrum hypothesis of the eating disorders, which considers them as one syndrome with different manifestations.

Seasonal fluctuations in the cervical smear detection rates for (pre)malignant changes and for infections.

The detection of diseases can exhibit seasonal fluctuations. This can be studied in cervical smears. Over a 9-year observation span (January 1983-January 1992) a series of 504,093 cervical smears obtained from a routine cytology laboratory in The Netherlands were examined for infections (monilia, trichomonas, actinomyces, human papilloma virus [HPV], chlamydia, and herpes) as well as for mild, moderate, and severe dysplasias, carcinoma in situ, and squamous carcinoma. Statistical analysis (principal component analysis) demonstrates clear seasonal rhythms in the detection of infections as well as in precursor lesions. These findings suggest that we are dealing with "true" detection rhythms. For the detection of (pre)malignancy and HPV, yearly fluctuations in women being screened might be the explanation for our observations.

Prediction of various grades of cervical neoplasia on plastic-embedded cytobrush samples. Discriminant analysis with qualitative and quantitative predictors.

The purpose of this study was to investigate whether discrimination into five groups of various grades of cervical preneoplasia and neoplasia is possible using discriminant analysis models. Data were analyzed for 242 cases diagnosed as either slight dysplasia (n = 50), moderate dysplasia (n = 50), severe dysplasia (n = 50), carcinoma in situ (n = 50) or invasive carcinoma (n = 42) and consisted of qualitative and quantitative features of cells derived from a repeat sample taken from the ectocervix as well as the endocervix using Cytobrushes. The samples were embedded in plastic, and thin sections were prepared, resulting in a monolayer of cut nuclei. The percentages of expected correct prediction were obtained by using 10,000 double cross-validation samples; the mean percentage of correct prediction into five groups using cross-validation was 65% (in the original analysis, 72%) and into two groups (dysplasia versus carcinoma in situ and invasive carcinoma) was 91% (93%). The results reflect group discrimination potential; we do not claim reliability of prediction for an individual patient. The patients were not a representative sample of the population; to investigate whether groups of patients could be discriminated on the basis of both qualitative and quantitative features, the data analyzed contain an almost equal number of observations in each of the five groups. The results indicate that features do not classify the cases in the same way; the discriminant analyses suggest that quantitative features play an important role in the discrimination of dysplasia from carcinoma cases, while the majority of the qualitative features are important in discrimination within the three dysplasia groups.