Persistence in health expenditures in the short run: prevalence and consequences

Med Care. 2003 Jul;41(7 Suppl):III53-III64. doi: 10.1097/01.MLR.0000076046.46152.EF.


Background: Knowing whether persons in the top percentiles of the health expenditure distribution exhibit persistently high expenditure is fundamental to developing health plan payment policies, containing costs, and understanding the consequences of costly illnesses.

Objectives: To determine the extent of high expenditure persistence over a 2-year period. To identify the correlates and consequences of expenditure persistence.

Subjects: A national sample of the population from a longitudinal panel of the Medical Expenditure Panel Survey (MEPS).

Methods: Changes in a person's position in the expenditure distribution were examined. chi2 tests were used to identify differences in characteristics between high and low spenders. Logistic regression was used to predict the likelihood of expenditure persistence. Changes in income, employment, out-of-pocket expenditure burden, and health insurance were compared for high and low spenders.

Results: Of the top 5% of spenders in 1996, 30% retain this position in 1997 and 45% are in the top decile of 1997 spenders. High expenditures begin to regress to the mean over the study period. Cancer, mental disorders, diabetes, and infectious diseases and being in the top decile of 1996 spenders increase the probability of expenditure persistence (P < 0.05 for all). This probability also has a strong random component. An increased proportion of persons in the top expenditure decile for both years had out-of-pocket health spending greater than 20% of income in 1997 (P < 0.10). Persons with persistently high expenditures were less likely than low spenders to lose employment-based coverage (5.4% vs. 8.8%, P < 0.05) but no changes in income or employment status were detected.

Conclusions: A sizable minority of persons exhibits persistently high expenditures, creating incentives for favorable risk selection. Few consequences of short-run expenditures persistence are observed.

Publication types

  • Research Support, U.S. Gov't, P.H.S.

MeSH terms

  • Adult
  • Aged
  • Demography
  • Employment / statistics & numerical data
  • Ethnicity
  • Female
  • Financing, Personal / statistics & numerical data
  • Health Care Surveys*
  • Health Expenditures / statistics & numerical data*
  • Health Status
  • Humans
  • Income / statistics & numerical data
  • Insurance Coverage / statistics & numerical data
  • Insurance Selection Bias
  • Likelihood Functions
  • Longitudinal Studies
  • Male
  • Middle Aged
  • Mortality
  • Socioeconomic Factors
  • United States / epidemiology