ASES provided me with all medical and pharmacy claims of ASES beneficiaries during the period January 1-June 30, 2000. There were about 12.2 million claims.

Demographic information. The following demographic variables could be determined from the claims records:

• The person’s sex

• The person’s age

• The geographic region in which services were provided

Unfortunately, we lack data on other personal attributes, such as education and income. But unobserved heterogeneity with respect to income is limited by the fact that, to be eligible for Medicaid in Puerto Rico, annual income of a family of four could not exceed $16,440 (in the year 2002).

Person’s utilization of services.

• The person’s number of medical claims (physician encounters)

• The person’s number of hospital admissions

• The person’s number of pharmacy claims

Vintage distribution of pharmacy claims. Each pharmacy claim3 included the National Drug Code (NDC). I determined the active ingredient(s) contained in each NDC from Multum’s Lexicon. I determined the earliest FDA approval date of each active ingredient from standard commercial pharmaceutical databases, i.e. Gold Standard Multimedia’s Clinical Pharmacology 2000 and Mosby’s Drug Consult. Using this information, I calculated, for each pharmaceutical claim, the values (0 or 1) of P0ST70, P0ST80, and P0ST90. I then calculated, for each individual, the average values of P0ST70, P0ST80, and P0ST90, i.e. the fraction of the individual’s Rx’s that were for drugs approved after 1970, 1980, and 1990. read more

Nature of person’s illnesses. The medical claims include ICD9 (diagnosis) codes. I grouped these codes into the following 15 broad disease groups:
I then calculated DISEASE_SHAREij (j = 1, 2,…, 15): the fraction of person i’s diagnoses that were in each disease category. For example, if all of person i’s diagnoses were diabetes, then DISEASE_SHAREij = 1 if j = 3 and DISEASE_SHAREij = 0 if j ф 3. If person i had 3 circulatory diagnoses and one digestive diagnosis, then DISEASE_SHAREij = 0.75 if j = 7, DISEASE_SHAREij =.25 if j = 9, and DISEASE_SHAREij = 0 for all other j.

In addition to measuring the shares of diagnoses in each disease category, I calculated the person’s “effective number” of diseases. Rather than simply counting the number of disease categories in which a person’s diagnoses fell, I computed the following index:


If all of a person’s diagnoses fell in one disease category, then N_DISEASEi = 1. If half of a person’s diagnoses fell in one disease category, and half fell in a second category, then N_DISEASEi = 2. If 90% of a person’s diagnoses fell in one disease category, and 10% fell in a second category, then N_DISEASEi = 1 / (.92 +.12) = 1.22.

Mortality. The Department of Health provided me with a list of (encrypted) social security numbers of all Puerto Rican residents who died during the period 20002002. I merged this list with the January 1-June 30, 2000 ASES claims data; this allowed me to determine whether or not an ASES beneficiary who had utilized services during January 1-June 30, 2000 had died by the end of 2002:

DIEDi = 1 if person i died by the end of 2002 = 0 otherwise

Descriptive statistics. Sample means of the variables are shown in the following table.

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