Application of new methods of data mining and data analysis in bariatric surgery

Postprandial hypoglycemia is increasingly recognized as a complication of bariatric surgery. The clinical characteristics of hypoglycemia in patients underwent bariatric surgery typically emerge gradually over time and are often relatively nonspecific. To date, the literature established that the pathogenesis of hypoglycemia is poorly understood and is probably based on multifactorial causes. In order to prevent post-prandial hypoglycemia, pre-operative individual risk assessment would be helpful. Lacking factors able to prevent hypoglycemia episodes, assessing dietary patterns among patients underwent bariatric surgery was useful to understand which kind of foods are more likely to be associated with hypoglycemia episodes. Both statistical analysis used in our study, considering single principal meals and overall food groups consumed, confirmed the positive association with sweet products and negative with complete meal. Further studies on a more large sample are needed to confirm these results. Although in this pilot study (only 54 patients analyzed), new statistical tools, as Structural Equation Modeling, not frequently used in epidemiology as in the social sciences, gave us the opportunity to delineate causes and effects from observational data. In particular, we found the direct effects of single group foods on the susceptibility to have hypoglycemia episodes, but especially, the indirect effects mediated by other foods or eating pattern. In addition, we found also direct and indirect effects for taste modifications. Such patients require complex nutritional and medical management strategies to reduce postprandial insulin secretion in order to stabilize glucose excursions as well as frequent monitoring of glucose values. At least, the implementation of a large integrated health record system point out three principal benefits: 1) to monitor quality and safety outcomes; 2) to assist the clinical personnel responsible for the care of these patients; and 3) collecting all the information available about single patients using a bariatric registry can facilitate future studies, as well as develop new hypothesis