Statistical clues to postoperative blood loss: Moving averages applied to medical data

With the advent of computerized databases, medical data has become easy to accumulate; however, effective use of this data continues to pose significant problems. In other circumstances, smoothing algorithms have been used to uncover non-obvious correlations, trends and relationships in noisy data. We have applied four such algorithms to a large dataset of postoperative blood replacement in cardiopulmonary bypass patients. When applied to this dataset, one of the algorithms proved surprisingly effective. It confirmed several previously observed correlations, and also provided an additional series of counterintuitive and apparently unrelated associations. These associations have been explored in an accompanying paper.     

Nonabsorbable marker and single, random stool samples used for measuring intestinal absorption of macronutrients in infants and children

To measure intestinal absorption by using a single, random stool sample, polyethylene glycol (PEG), 1 g/d, and a constant diet were given to healthy infants, with a constant PEG-to-macronutrient ratio. After 10 d equilibration, apparent intestinal absorption of macronutrients was estimated from a standard 3-d metabolic balance and compared with that estimated by using the ratio of PEG to macronutrients in a single random sample of feces. Correlation coefficients for this comparison were 0.649, 0.715, and 0.924 for nitrogen, carbohydrate, and fat, respectively. Additionally, apparent intestinal absorptions estimated from two separate consecutive 3-d metabolic-balance studies were compared, showing correlation coefficients of 0.106, 0.653, and 0.463 for nitrogen, carbohydrate, and fat, respectively. The random sample-marker technique appears to be acceptable for measuring apparent absorption of macronutrients and is at least as accurate as a standard 3-d metabolic-balance study.