OBJECTIVE: To evaluate Type 2 diabetes hospitalization costs and their determinants by applying a proper methodological approach, taking into account the presence of several observations with zero costs. METHODS: A cohort study using per-patient hospital discharge abstracts in a period of 4.5 years of follow-up (from 1/1/1996 to 30/6/2000). Potential cost predictors were age, sex, body max index, hypertension, diabetes duration, hemoglobin A1c levels, insulin treatment, retinopathy, coronary artery disease, peripheral artery disease, nephropathy, death and presence of comorbidity (cancer, chronic liver disease, chronic obstructive pulmonary disease, and psychiatric disease). Among risk factors, total cholesterol, HDL cholesterol and smoking were considered. A two-part model has been adopted in order to take into account the presence of patients with zero hospital costs: the probability of any hospitalization has been modeled via a standard logit generalized linear model (GLM); the actual level of total costs has been modeled via a GLM, with a gamma cost distribution and a LOG link function. RESULTS: In 4.5 years the median total cost per hospitalized person was $4404 (mean $8180). In line with existing evidence, diabetes complications showed a high impact on average costs. In particular, peripheral and coronary artery diseases determined more than $1000 increase in the median costs. Chronic comorbidity were responsible for the highest incremental hospitalization costs ($1771). CONCLUSIONS: Hospitalization costs were significantly increased by the presence of diabetes complications and chronic conditions. The adoption of a two-part model has allowed to obtain estimates not neglecting the effect of covariates on the probability of having no hospital care.
OBJECTIVE: To evaluate Type 2 diabetes hospitalization costs and their determinants by applying a proper methodological approach, taking into account the presence of several observations with zero costs. METHODS: A cohort study using per-patient hospital discharge abstracts in a period of 4.5 years of follow-up (from 1/1/1996 to 30/6/2000). Potential cost predictors were age, sex, body max index, hypertension, diabetes duration, hemoglobin A1c levels, insulin treatment, retinopathy, coronary artery disease, peripheral artery disease, nephropathy, death and presence of comorbidity (cancer, chronic liver disease, chronic obstructive pulmonary disease, and psychiatric disease). Among risk factors, total cholesterol, HDL cholesterol and smoking were considered. A two-part model has been adopted in order to take into account the presence of patients with zero hospital costs: the probability of any hospitalization has been modeled via a standard logit generalized linear model (GLM); the actual level of total costs has been modeled via a GLM, with a gamma cost distribution and a LOG link function. RESULTS: In 4.5 years the median total cost per hospitalized person was $4404 (mean $8180). In line with existing evidence, diabetes complications showed a high impact on average costs. In particular, peripheral and coronary artery diseases determined more than $1000 increase in the median costs. Chronic comorbidity were responsible for the highest incremental hospitalization costs ($1771). CONCLUSIONS: Hospitalization costs were significantly increased by the presence of diabetes complications and chronic conditions. The adoption of a two-part model has allowed to obtain estimates not neglecting the effect of covariates on the probability of having no hospital care.