For example, one additional medical claim (physician visit) is associated with a.0031 (about 8%) increase in the probability of death. Although utilization of medical services presumably reduces mortality, given initial (pre-treatment) health status, people in the worst initial health utilize the most medical services.

Including the three utilization variables (as well as the diagnosis category and demographic variables) is therefore likely to control for initial health status, which is unobserved and might be correlated with drug vintage (although the sign of the potential correlation is not clear).
The coefficients on the diagnosis category variables indicate differences in mortality rates associated with different diagnoses. They are plotted in Figure 2. By a large margin, people diagnosed with neoplasms had the highest mortality rate. Diseases of the blood and blood-forming organs and diseases of the skin and subcutaneous tissue are two other high-mortality conditions. The lowest-mortality conditions include mental disorders, diseases of the nervous system and sense organs, and diseases of the musculoskeletal system and connective tissue.

The coefficient on N_DISEASE is negative and highly significant, which is somewhat surprising. This indicates that, conditional on the distribution of diagnoses a person had (and other covariates), the greater the effective number of diseases the person had, the lower the probability of death.
The coefficients on the region dummy variables indicate differences in mortality rates associated with different regions. They are plotted in Figure 3. The mortality rate in Suroeste is almost 1.5 percentage points higher than the mortality rate in the second-highest region, Este. The regions with the lowest mortality rates are Montana, Guayama, and Arecibo.

Estimates by condition. In addition to estimating the model using data on the entire population of ASES beneficiaries with pharmaceutical claims, I also estimated the model separately for people with three specific diagnoses: diseases of the circulatory system; endocrine, nutritional and metabolic diseases, and immunity disorders (primarily diabetes); and neoplasms. All three of these conditions exhibit relatively high mortality rates (above 6%), and the first two are highly prevalent (affecting at least one-sixth of the population). There were more than 2600 deaths in each group of people.

Rather than reporting the complete set of estimates (as in Table 1) for each group, we report just the coefficients of the three drug-vintage variables, as well as means of key variables for each group, in Table 2.

Category: EFFECT OF DRUG VINTAGE / Tags: Clinical studies, myocardial infarction, survival rates