Medicaid1986 {AER}

Medicaid Utilization Data

Description

Cross-section data originating from the 1986 Medicaid Consumer Survey. The data comprise two groups of Medicaid eligibles at two sites in California (Santa Barbara and Ventura counties): a group enrolled in a managed care demonstration program and a fee-for-service comparison group of non-enrollees.

Usage

data("Medicaid1986")

Format

A data frame containing 996 observations on 14 variables.

visits

Number of doctor visits.

exposure

Length of observation period for ambulatory care (days).

children

Total number of children in the household.

age

Age of the respondent.

income

Annual household income (average of income range in million USD).

health1

The first principal component (divided by 1000) of three health-status variables: functional limitations, acute conditions, and chronic conditions.

health2

The second principal component (divided by 1000) of three health-status variables: functional limitations, acute conditions, and chronic conditions.

access

Availability of health services (0 = low access, 1 = high access).

married

Factor. Is the individual married?

gender

Factor indicating gender.

ethnicity

Factor indicating ethnicity ("cauc" or "other").

school

Number of years completed in school.

enroll

Factor. Is the individual enrolled in a demonstration program?

program

Factor indicating the managed care demonstration program: Aid to Families with Dependent Children ("afdc") or non-institutionalized Supplementary Security Income ("ssi").

Source

Journal of Applied Econometrics Data Archive.

http://qed.econ.queensu.ca/jae/1997-v12.3/gurmu/

References

Gurmu, S. (1997). Semi-Parametric Estimation of Hurdle Regression Models with an Application to Medicaid Utilization. Journal of Applied Econometrics, 12, 225–242.

Examples

## data and packages
data("Medicaid1986")
library("MASS")
library("pscl")

## scale regressors
Medicaid1986$age2 <- Medicaid1986$age^2 / 100
Medicaid1986$school <- Medicaid1986$school / 10
Medicaid1986$income <- Medicaid1986$income / 10

## subsets
afdc <- subset(Medicaid1986, program == "afdc")[, c(1, 3:4, 15, 5:9, 11:13)]
ssi <- subset(Medicaid1986, program == "ssi")[, c(1, 3:4, 15, 5:13)]

## Gurmu (1997):
## Table VI., Poisson and negbin models
afdc_pois <- glm(visits ~ ., data = afdc, family = poisson)
summary(afdc_pois)
coeftest(afdc_pois, vcov = sandwich)

afdc_nb <- glm.nb(visits ~ ., data = afdc)
ssi_pois <- glm(visits ~ ., data = ssi, family = poisson)
ssi_nb <- glm.nb(visits ~ ., data = ssi)

## Table VII., Hurdle models (without semi-parametric effects)
afdc_hurdle <- hurdle(visits ~ . | . - access, data = afdc, dist = "negbin")
ssi_hurdle <- hurdle(visits ~ . | . - access, data = ssi, dist = "negbin")

## Table VIII., Observed and expected frequencies
round(cbind(
  Observed = table(afdc$visits)[1:8],
  Poisson = sapply(0:7, function(x) sum(dpois(x, fitted(afdc_pois)))),
  Negbin = sapply(0:7, function(x) sum(dnbinom(x, mu = fitted(afdc_nb), size = afdc_nb$theta))),
  Hurdle = colSums(predict(afdc_hurdle, type = "prob")[,1:8])
  )/nrow(afdc), digits = 3) * 100
round(cbind(
  Observed = table(ssi$visits)[1:8],
  Poisson = sapply(0:7, function(x) sum(dpois(x, fitted(ssi_pois)))),
  Negbin = sapply(0:7, function(x) sum(dnbinom(x, mu = fitted(ssi_nb), size = ssi_nb$theta))),
  Hurdle = colSums(predict(ssi_hurdle, type = "prob")[,1:8])
  )/nrow(ssi), digits = 3) * 100