The estimates in Tables 5.1 - 5.9 were created by formula, using country-by-country estimates of diabetes prevalence by age and sex, population size by age and sex, total healthcare expenditures by age and sex, and the ratio of expenditures per person with diabetes to expenditures per person without diabetes, matched for age and sex. None of these inputs are known with certainty in industrialized countries and, in the rest of the world, no direct measurements have ever been published. One parameter, the diabetes cost ratio, is known by 10-year categories of age and sex only for a US population; assumptions about the relative magnitudes of expenditures for persons without diabetes were also based on US data. Although IDF and WHO are sponsoring studies that will obtain estimates from more settings, the estimates presented here rely on limited information.

Definition of health expenditures
The expenditures displayed in the Tables are the estimated total health expenditures caused by diabetes. Initial data on per capita total health expenditures by country were obtained from Annex Table 6 to the WHO World Health Report for 2004   1 . WHO defines ‘total health expenditure’ to include all expenditures for medical care regardless of who paid for them. The WHO definition also includes expenditures for public health programmes, water supply and hygiene activities, nutritional support activities, education, training, and research — but only when these activities intentionally and primarily address a health problem. The WHO definition excludes the unpaid care-giving of relatives and others, and the opportunity costs of this care-giving, including loss of paid employment. It also excludes other opportunity costs, such as loss of educational opportunities for children who must stay home to care for disabled parents.

A significant portion of healthcare spending in the poorest countries comes from governmental programmes and from external donors, who focus on communicable and parasitic diseases rather than on diabetes and cardiovascular disease (despite diabetes causing as many deaths as HIV/AIDS). The IDF estimates of expenditures for diabetes in poor countries may therefore be exaggerated.

Estimation of general per capita expenditures for persons without diabetes
In each country, expenditures for persons without diabetes were estimated for 42 subpopulations based on sex and five-year strata of age, ranging from age zero to ages greater than 100. Because data on total health expenditures by age and sex are very rare, even for developed countries, estimates of population-wide total health expenditures per capita published by WHO for the year 2002   1  were used.

Per capita WHO expenditure estimates were divided into two components: a portion that was assumed to vary with age- and sex-specific mortality (about 80% of total expenditures) and a portion (about 20%) that was assumed to be constant within each age- and sex-subgroup. Total expenditures per subgroup were calculated as the sum of (a) constant expenditures multiplied by subgroup size and (b) the product of mortality-related expenditures and the predicted number of annual deaths in the subgroup.

Reliable mortality statistics by country are not universally available so each country’s mortality rates were assumed to match the rates published for its WHO Demographic Group (N=14)   2 . Because mortality rates vary more widely with age than medical care expenditures do, rates by sex and five-year age group were transformed using a log function, ln(3.00 + mortality rate). Subgroup mortality rates were further modified to account for the fact that only about half the children who die in countries with high and very high childhood mortality receive medical care, and for the generally lower average expenditures for conditions that cause death in childhood. The 20% of annual expenditures that were assumed not to vary with mortality were adjusted by age and sex to account for natural differences medical care utilization, such as the higher use of medical care services by women of child-bearing age.

The resulting estimates were then fine-tuned by approximately equalizing, for each age and sex subgroup, the ratio of total per capita medical care expenditures predicted via these methods for the US population to the per capita medical care expenditures observed in a US sample of persons who did not have diabetes. (The sample without diabetes were members of the Kaiser Permanente medical care programme in the United States, selected and analyzed by one of the authors [GN].) Per capita expenditures for men were further adjusted to maintain the male/female ratios to expenditures found in the Kaiser Permanente data. The resulting relative distributions of per capita health expenditures by age and sex, when multiplied by the population in each subgroup, yielded for each country an aggregate total health expenditure.

To ensure that average total per capita expenditures in each country still matched the estimates published by WHO, aggregate expenditures were compared to the WHO estimates for 2002, country by country. Expenditures were adjusted up or down so that the estimated countrywide expenditure equalled the country’s WHO-published expenditure for 2002. This result was then increased to account for population growth since 2002, the year of the WHO estimates, by dividing the UN medium-variant projected population for each country in either 2007 or 2025, as appropriate, by the population assumed in WHO estimates for 2002, and multiplying by the resulting ratio.

The diabetes expenditure ratio, R
The Tables give alternative estimates for values of a parameter called R, which is the ratio of all medical care expenditures for persons with diabetes to all medical care expenditures for age- and sex-matched persons who do not have diabetes. By comparing the total expenditures of matched persons with and without diabetes, the expenditures that diabetes causes can be isolated. Because R varies from country to country and over time, the Tables show results for likely lower and upper bounds of R, R=2 and R=3.

The present analysis attempts to improve on the ground-breaking estimates calculated for the second edition of the Diabetes Atlas, by explicitly accounting for demographic variation in R. R is quite sensitive to age and sex, and countries differ markedly in the age structures of their populations. In industrialized countries, R is higher in younger age groups because younger persons without diabetes do not usually incur large medical expenditures. Conversely, R is lower at older ages because old persons without diabetes use substantial medical care. Younger men without diabetes also use less medical care than younger women in industrialized countries.

When the single global R’s used in the second edition of the Atlas were replaced with age- and sex-specific R’s, most country estimates of expenditures for diabetes increased, often quite substantially. Global expenditures nearly doubled. This is an encouraging result because IDF’s earlier results appeared to underestimate true diabetes care expenditures, when compared with published national studies.

To obtain an empirical basis for age- and sex-specific values of R, authors [GN and JBB] affiliated with Kaiser Permanente Northwest Region (KPNW), a large not-for-profit pre-paid medical care system in the United States, calculated ratios from this organization’s large diabetes registry. The mean R for all KPNW registrants aged 20-79 for 2004 was 2.066 for women and 2.088 for men. To create age by sex distributions of R’s for standard R’s with a population-weighted mean of 2.0 and 3.0, the KPNW distributions of ratios were adjusted up and down. Table 1 displays these observed and adjusted ratios and the numbers of subjects that contributed data. Because of low sample sizes in the age groups between 20 and 50, R’s for these ages were estimated en bloc.

Table 1 Diabetes expenditure ratios (R) by age and sex*

Computational details
For the third edition of the Diabetes Atlas, data and calculations for each country were broken down into 10-year age-sex subgroups, starting with age 20-29 and ending with age 70-79. (Persons aged less than 20 or more than 79 years were omitted because data on the prevalence of diabetes in these age groups are lacking for most countries.) Expenditures were calculated for each subgroup, one at a time, for men and for women, using a different value of R for each subgroup. The subgroup expenditures were then combined, first within sex, and then combining the sexes, weighting each subgroup’s contribution to the total by the proportion of the country’s diabetic population that fell into each age-sex subgroup.

Specifically, countrywide and per capita expenditures of medical care in 2007 were estimated by combining data describing: 

• the estimated current prevalences of diabetes in 2007 (P_as, as estimated in the report on epidemiologic studies);
  
• estimated 2007 populations (N_as, based on United Nations projections, median fertility variant  9  or, for non-UN members, the CIA World Factbook  10 );
  
• total current healthcare budgets in 2002 (C_as, obtained from WHO estimates and projections  11 ); and
  
• ratios (R_as) of medical care expenditures for persons with diabetes compared to persons without diabetes.

All these data were divided into age deciles (a=1–6), by sex (s=1,2). The formula used to calculate the expenditures of medical care for diabetes in each country was:

, where

D = the total expenditure of care for diabetes in a country,
C = the estimated annual budget for all healthcare in the country in 2002,
N_as = the total population of persons, in each age and sex subgroup, projected for a country in 2007
N = the total population of the country of all ages
P_as = the prevalence of diabetes in the country, by age and sex,
R_as = the ratios of expenditures for persons with diabetes to persons without diabetes, by age and sex, and where
a is an indicator for age decile (20-29, 30-39, …70-79), and
b is an indicator for sex (men, women).

This formula corrects for the fact that per capita health expenditures per person with diabetes include expenditures caused by many conditions, not just diabetes. The formula yields the costs caused by diabetes.

Projection to 2025
Estimates of expenditures in 2025 differ from estimates for 2007 only as a result of projected changes in population structure (total size, sex, age, and percent urban). Expected growth in diabetes incidence is not included. Also ignored are increases in medical care expenditures due to economic growth and/or relative inflation in prices for medical care. For these reasons, these projections underestimate future diabetes expenditures.

US and international dollars
Expenditures are shown both in US dollars (USD) and international dollars (ID), valued as of the year 2002, the most recent year for which national healthcare expenditure data for all countries are currently available. (Projected expenditures in 2025 are also shown in 2002 dollars.) Expenditures in USD estimate the amount of internationally traded currency that is spent for diabetes care. These expenditures can be used to compare how much individuals and institutions paid or will pay for diabetes care.

A unit of internationally traded currency can buy many more goods and services in some countries than in others. Converting USD to ID corrects for such differences, which economists call differences in purchasing power. Expenditure estimates in ID can be used to compare the amounts of diabetes care that countries actually produce.

The market-basket studies from which ID multipliers are calculated involve a wide range of products and services. These multipliers might not be accurate for the medical care sectors of some countries. For example, healthcare workers in many poor countries are said to be underpaid relative to workers in other occupations in the same country. If so, the true difference between USD and ID estimates might be greater than is reported here. On the other hand, medicines and medical supplies are often imported and, in many low-income countries, medicines are taxed upon entry. Some manufacturers of diabetes medicines lower their wholesale prices to poor countries, but shortages and black-market distribution can erode these efforts. Consequently, estimates in ID could overestimate the amount of medicine that can be purchased in poorer countries.

Summary of limitations
Expenditure estimates derived by the methods described above have many limitations. First, they depend on estimates of population size, diabetes prevalence, aggregate health expenditures and rates of mortality that are imperfect. Second, they depend on assumptions whose accuracy has not been confirmed in most of the world. For example, the data used to calculate R and to adjust mean per capita expenditures for age and sex came from a single country, the USA, and from a single medical care system within that country. Almost nothing is known about R or about general medical expenditure patterns by age and sex in poor and middle-income countries. It was also assumed that estimates of R derived from persons with diagnosed diabetes apply to persons with undiagnosed diabetes, which could be wrong. And it was assumed that estimates of R derived from data on medical care can be generalized to apply to all money expended for health purposes, the only definition of expenditure for which there are estimates for every country. This may be especially inaccurate in low-income countries, which receive large portions of their health budgets from external donors, who generally want to focus their giving on public health initiatives and infectious disease. Finally, purchasing power parities estimated for general baskets of goods and services may not describe purchasing power medicines and medical care, increasing the uncertainty of the estimates in international dollars.

These limitations mean that the estimates here for most countries will be very imprecise. Nevertheless, some conclusions shine through clearly. Diabetes causes huge amounts of spending and loss. Wealthy countries do almost all the spending. Low- and middle-income countries bear most of the loss. Better diabetes treatment would be cost-effective everywhere.