Estimating and Projecting Health Expenditures on the Elderly in Low and Middle Income Counties: An Econometric approach IAAHS Colloquium IAAHS Colloquium Dresden, Germany Dresden, Germany April 2004 April 2004 By By A.K. Nandakumar, Ph.D. A.K. Nandakumar, Ph.D. Brandeis University Brandeis University Abt Associates Inc. Jonathan Wilwerding, PH.D. Jonathan Wilwerding, PH.D. In collaboration with: � Development Associates, Inc. Abt Associates, Inc. Abt Associates, Inc. � Emory University Rollins School of Public Health � Philoxenia International Travel, Inc. � Program for Appropriate Technology in Health � Social Sectors Development Strategies, Inc. � Training Resources Group � Tulane University School of Public Health and Tropical Medicine � University Research Co., LLC
“Old age is the most unexpected of all things that happen to a man.” (Tolstoy)
Demographic Trends Demographic Trends
Life Expectancy 90 80 70 60 50 40 30 20 10 0 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 Year World More developed regions Less developed regions Least developed countries
Average age 50.00 45.00 40.00 35.00 30.00 25.00 20.00 15.00 10.00 5.00 0.00 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Year source: UNPOP 98. Post '95 data are projections. World Less developed regions More developed regions Least developed countries
Population share 65+ 0.30 0.25 0.20 0.15 0.10 0.05 0.00 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Year source: UNPOP 98. Post '95 data are projections. World Less developed regions More developed regions Least developed countries
Population share 80+ 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Year source: UNPOP 98. Post '95 data are projections. World Less developed regions More developed regions Least developed countries
Share of 80+ in the old population 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Year source: UNPOP 98. Post '95 data are projections. World Less developed regions More developed regions Least developed countries
Dependency ratio 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Year source: UNPOP 98. Post '95 data are projections. Least developed countries Less developed regions More developed regions World
Elder share of dependent population 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 source: UNPOP 98. Post '95 data are projections Year Least developed countries Less developed regions More developed regions World
Trends in Global Aging L Rapid L Predictable L Pattern -- Oldest Old L Distribution across countries L Lack of preparedness
The Elderly in 2020: In the Next 50 Years L Share of world population 65+ will double L Average age increases from 26.5 years to 36.2 L Share in 80+ age group will quadruple L B y 2025, 75% of world’s elderly will be living in developing countries L Aging populations severely pressure public finances; policy responses must be radical and swift See: See: Bloom, D.E., Nandakumar, A.K., Bhawalkar, M., “The Demography of Bloom, D.E., Nandakumar, A.K., Bhawalkar, M., “The Demography of Aging in Japan Aging in Japan and the US,” and the US,”
Estimating Expenditures on Elderly Estimating Expenditures on Elderly
Current State of Knowledge L Almost all the work done in developed countries L Approaches fall into following categories � Actuarial � Micro-simulation L A number of challenges in low and middle income countries � Lack of longitudinal data � Health systems that are not transparent � Correlations observed in developed countries do not always hold See Research Paper No. 01.23, GBD 2000 in Aging Populations by A See Research Paper No. 01.23, GBD 2000 in Aging Populations by Ajay Mahal and Peter Berman jay Mahal and Peter Berman that reviews estimation methodologies. Report published by WHO that reviews estimation methodologies. Report published by WHO
The Proposed Methodology L Uses a National Health Accounts framework � Classification of sources and uses of health expenditures in developing countries approved by the World Bank, WHO and USAID L Four Step Process � Estimate the base case � Model macro-economic growth � Age Population � Project Expenditures (to 2015 in this case)
Sources of Data Sources of Data � National Health Accounts � Household Health Care Utilization and Expenditure Survey � Costing studies � Annual Reports � GDP Information � Population Data from UN Sources � Government budgets
Estimating the Base Case L Public Expenditures (g) = the sum of (quantity x unit cost) across all functions, j, and public entities, l . ( ) ( ) ∑ ∑ L J = × g Y Q c = = k jkl jkl 1 l j
Estimating the Base Case L Step 1: Estimate total outpatient visits and inpatient admissions at public entities for year k N ∑ = × 1 / Q N w q population jk i ijk k = 1 i
Estimating the Base Case L Step 2: Allocate visits and admissions across public entities. = × Q p Q jkl jkl jk ′ ( ) e x â i m = = | P p m x ijkl i L ∑ ′ + 1 x â e i l = 2 l Multinomial Logit Multinomial Logit
Estimating the Base Case L Estimate private out of pocket expenditures, using household survey data N ∑ = 1 / y N w y k i ik = 1 i = × private Y y populaiton k k k
Estimating the Base Case L Step 4: Estimate expenditures by donors � The best way to get donor assistance information is through a survey of donors � This is because donors give assistance in both cash and kind
Projecting Expenditures L We follow a four step process � Estimate econometric models of private Estimate econometric models of private � expenditures, visits, admissions, and choice of expenditures, visits, admissions, and choice of provider using data for the base year provider using data for the base year � “Age” the base year population data “Age” the base year population data � � Model the macro Model the macro- -economic growth economic growth � � Apply econometric models to aged data to predict Apply econometric models to aged data to predict � private expenditures, utilization and out- private expenditures, utilization and out -of of- -pocket pocket expenditures for the forecast year expenditures for the forecast year
Estimating Utilization in Projection Year Visits, admissions = (the probability that an Visits, admissions = (the probability that an individual had an illness x the expected individual had an illness x the expected number of visits, admissions) number of visits, admissions) ( ) ( ) ( ) = = ⋅ , = 1 | 1 E Q P S E Q S j k k j k k Estimate probability of illness by logistic Estimate probability of illness by logistic regression regression Estimate expected visits, admissions by Estimate expected visits, admissions by econometric model econometric model
Estimating Outpatient Visits in Estimating Outpatient Visits in Projection Year Projection Year Visits, admissions generated by two-stage Poisson Process -- Separate individuals who must have zero counts -- Determine number of visits for others Estimate utilization by zero-inflated negative binomial regression model (ZINB) Other models underpredict zeros; in low, middle income countries utilization drops for older ages
Estimating Out-of-Pocket Expenditures and Choice of Provider in Projection Year L Use OLS to predict private expenditures in projection year, conditional on having > 0 expenditures. L Tobit model and Sample Selection models predict poorly within sample L Use multinomial logit models to predict the choice of provider
Aging the Survey Data to Get the Data Set for Projection Year L Age only data for variables that appear in econometric models used to predict use and expenditures for the forecast year L Add fifteen years to each individual in the survey data (projection year was 2015) L Create observations for age group 0-15 L Use econometric methods to assign predicted values to relevant variables L To account for the effects of mortality, weight the data using 15-year survival rates, by age
Creating the Data Set for Projection Year L Variables whose predicted values were modeled included � Marital Status � Education � Income � Employment � Health Status � Insurance � Cost per visit and cost per inpatient admission
Recommend
More recommend