Here's an interesting article that's essentially about trying too hard to develop a comprehensive model for healthcare risk adjustment. It's from the Healthcare Economist blog entitled "Risk Adjustment: Overfitting the Model". It describes a situation in which parameters are introduced to a model in an attempt to qualtify the effect of a factor, but that end up being more sensitive to the characteristics of an individual person (because of small sample size) than the factor itself. Here's the link: http://healthcare-economist.com/2012/06/26/risk-adjustment-overfitting-the-model/