Sot. Sci. Med. Vol. 33, No. 6, pp. 723-727, 1991 Printed in Great Britain
0277-9536/91 53.00 + 0.00 Pergamon Press plc
HOW SHOULD RESOURCES BE REALLOCATED BETWEEN PHYSICIANS AND NURSES IN AFRICA AND
VERONICA VARGA~-LAGO~* Economics Department, Boston University, Boston, MA 02215, U.S.A.
Abstract-This paper examines ways in which health resources could be reallocated between physicians, nurses and other medical inputs in Africa and Latin America, according to their cost-effectiveness. An underlying question concerns whether countries in Africa and Latin America with decreasing health budgets in the 1980s should reduce the number of highly trained and more expensive workers, i.e. physicians, and redirect resources to less trained and less expensive workers, i.e. nurses. This paper designs a methodology for quantifying the cost-effectiveness of physicians, nurses and government health expenditure in relation to improvements in the populations health status. Direct estimation of the professionals effectiveness is unsuitable in this 4Scountry study. Instead, for measuring the unobserved effectiveness of health providers and health expenditures, infant mortality rate has been chosen as the indicator. Infant mortality is an accepted indicator of the health status in a given population. From another viewpoint, neonatal health is dependent on contact with health care services; this means that inappropriate care may increase the likelihood of infant mortality. Therefore, at the same time infant mortality is an indicator of the effectiveness of services. We used a general linear model as a way of estimating the relationship between infant mortality, health manpower and health expenditures. Forty-five countries were examined over three years and 135 observations were included in the final sample. Three scenarios were estimated: (1) African and Latin American countries, or low and middle-income countries, (2) only middle-income and (3) only low-income countries. From the regression model the marginal productivity for each kind of professional was calculated, and then the optimal mix of manpower physicians and nurses was estimated according to low and middle-income country budget constraints. Some countries were selected for illustrating how they differed from the optimal mix. For example, in countries such as Tanzania, where there is close to an optimal mix of personnel, population health status may not be improved with reallocation of resources, but only with new additional resources. The results suggest differences between low-income and middle-income countries. Physicians in low-income countries, primarily African countries, save about three times the number of lives as do nurses, while their salaries are twice as high as nurses salaries. This finding suggests that investing in physician training can make a significant contribution to primary health care in African countries. However, in middle-income countries, since the impact of nurses and physicians on infant mortality is proportional to their salaries, a health delivery system could be nurse-based or physician-based according to each countrys preferences.
Key words-health economics, health production mortality reduction
In Africa and Latin America, the government is an important provider of health services. It plays a central role by training health workers, assuring professional competence, setting health care stan- dards, and implementing regulations. Continued gains in health care quality depend heavily on the capacity of the government health system to deliver basic services and information to households particularly those in the low socioeconomic strata. Total government spending on health care in devel- oping countries grew in the 1960s and 1970s [l]. Slow economic growth and budget deficits in the 1980s have forced reductions in public spending; thus spending on health has increased more slowly during the 1980s and in some cases has actually decreased. These changes have forced policy makers
*Address correspondence to Veronica Vargas-Lagos, Har- vard Institute for International Development, One Eliot Street, Cambridge, MA 02138, U.S.A.
function, health manpower effectiveness, infant
to focus greater attention on whether health resource allocations can bc made more cost-effective, given the budget constraints. Quantifying the cost-effective- ness of health providers and health expenditures in relation to improvements in the populations health status is an important task for todays policy makers.
In Africa and Latin America 48% of the popu- lation is under 15 years of age. The health status of such a young population depends primarily on expen- diture and manpower allocation in primary health services . The infant mortality rate is an accepted measure of the effectiveness of health expenditures and manpower in relation to the health status of a population. Infant mortality is due mainly to an interrelated set of economic, social and health causes ; however in this context we assume-for the sake of simplicity-that it depends on the number of physicians and nurses and health expenditure. Infant mortality is a good indicator of the effectiveness of health manpower services and health expenditure because the newborn is highly dependent on health
124 VisomcA VAFaGAS-LAGGS
care services, and inadequate services often lead to high mortality. The first year of life is among the most hazardous in terms of survival.
It is estimated that between 60 and 80% of any given health budget is spent on manpower. Recently, Mejia, using an imaginary country not unlike many in Africa and Latin America, found that payments to health personnel could consume nearly 60% of expenditures on health services . Of this proportion nearly half is typically devoted to the payment of physicians and dentists, a fifth goes to nurses, and the remainder to support personnel. The author made assumptions that led him to estimate that the cost of medical education per student in many developing countries is closer to the equivalent of U.S. $60,000 than to the U.S. $10,000 generally estimated. In contrast, they estimated that the cost of training a nurse is closer to U.S. $8000. This suggests that about 7.5 nurses could be trained for the cost of training one physician. When such training costs are added to the costs of paying the physicians salary, the question arises as to whether the phys- icians contribution to improving health justifies the substantially higher training costs and the higher salary.
In most Latin American countries, as everywhere, nurse education has been improved as a way to advance health services. In special supplemental courses (l-3 months) which follow basic training, nurses are taught to make diagnoses and to give appropriate treatment for common illnesses such as infant diarrhoea, which can be treated by oral rehy- dration therapy . In many developing countries nurses are being prepared to take care of the most common causes of infant mortality. In primary health care it may be possible to substitute nurses for physicians, especially in countries where the number of physicians and nurses are very similar.
This paper examines ways in which health re- sources could be reallocated between physicians, nurses and other medical inputs in Africa and Latin America. The question is: should countries in Africa and Latin America reduce the number of physicians being trained and redirect resources toward training more nurses?
In most African and Latin American countries, national statistics for mortality of infants under 1 year of age are incomplete because of underreporting. We used estimates prepared by the Population Division of the United Nations [2,6]. We utilized data for 45 countries in 3 years: 1973, 1978 and 1983.
Expenditure values are based on the International Monetary Fund Government Yearbook. However, some countries such as Mozambique and Cuba do not follow this financing regime, and they were omitted from this analysis. The data were not com- prised prior to 1980 [7,8]. Expenditure values were originally expressed in national currencies, and then converted into U.S. dollars by means of the annual average exchange rate . All expenditures expressed in current dollars were adjusted to U.S. dollars at 1980 purchasing power, as determined by the United States Consumer Price Index [lo].
Population values represent estimates of mid-year population and refer to de facto population on 1 July. Those estimates were prepared by the Population Division of the United Nations Secretariat [l 1, 121. Manpower values were expressed in population per physicians and midwives, as prepared by the Popu- lation Division of the United Nations Secretariat and the World Health Organization [13-161.
We estimated the effects of health manpower and expenditures on infant mortality and examined the extent to which these independent variables ac- counted for decreases in infant mortality. We used a multiple regression analysis of the form:
IMR=BO+Bl x Exp+B2 x Phy
+ B3 x Nurse + B4(Phy x Nurse) + e (1)
Where: IMR = infant mortality rate expressed as the number of deaths of infants under 1 year of age per 1000 live births in a given period. Exp = Government health expenditure per capita in one year expressed in 1980 dollars. Phy = Number of physicians per 10,000 population. Nurse = Number of nurses and midwives per 10,000 population. Phy x Nurse = Number of physicians times number of nurses and midwives per 10,000 population.
CHARACTERISTICS OF THE SAMPLE
Developing countries are characterized by their low income per capita and insufficient access to basic services for most of their population. According to The World Bank in 1987, low-income countries have per capita income ranges between $110 and $1190 1984 dollars and mid-income economies are those in which income per capita ranges between $1200 and $5000 1984 dollars [l]. In order to conform to accepted analysis we are going to use the income difference in our examination. However, because the geographical criteria are comparable to economic distinctions, we will be referring mainly to Africa when we consider low-income countries, and to Latin America when we consider middle-income countries.
The sample comprised originally 60 countries which examined over three years created 180 obser- vations; we limited our sample to 45 or 135 obser- vations because those were the ones for which we had complete information. If we divide the 45 African and Latin American countries to conform to their income level we have 27 countries in the low-income sub- set, with 24 African countries, and El Salvador, Honduras, and Nicaragua. Eighteen countries fall in the middle-income range, all of them Latin American except for Botswana, Mauritius and Tunisia (Table 1).
The average infant mortality rate for the 45 countries was 105.0 per 1000 live births. While in the 27 low-income countries the infant mortality rate varies from 131.2 to 50.8 deaths per 1000 live births, making the average 124.9, in the 18 middle- income countries the infant mortality rate averaged 65.5 and varied from 131.2-50.8. There was a general decline of 17.8% for Africa and Latin America from 1973 to 1983. However, the drop was greater in
Resource allocation between physicians and nurses 125
Table I. Infant mortality and health expenditure by country income in Africa and Latin America over the years 1973, 1978 and 1983
Country Infant Mortality Expenditure Physicians Nur.%X income per 1000 per pmon per 10,000 per 10,wO level Year birth live (dollars) population population
Low income countries 1973 141.91 4.87 I .03 5.04 I978 126.01 7.08 1.31 5.92 I983 124.99 5.70 I .86 6.96
Middle income countries 1973 81.12 21.90 6.49 14.23 1978 64.76 33.09 6.30 10.76 1983 55.27 21.94 8.78 13.44
The s&cted Low-income countries are: Ethiopia, Mauritania, Chad, Egypt, El Salvador, Ghana, Guinea, Honduras, Kenya, hotho, Liberia, Malawi, Mali, Morocco, Nicaragua, Niger, Nigeria, Rwanda, Sierra Leona, Senegal, Sudan, Tanzania, Togo, Uganda, ~&q Zambia, Zimbabwe. Middle-income: Argentina, Botswana, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, &&or, Guatemala, Jamaica, Mauritius, Mexico, Panama, Paraguay, Peru, Tunisia, Uruguay, Venezuela.
middle-income countries: 31.8%, while in low-income countries the decrease was 11.9%.
Government health expenditures per capita for the sample at the end of the 1973-1983 decade were the same as it were at the beginning, with increases in the middle of the decade. However, in 1983 low-income countries showed an increase of 25% over 1973 levels of spending.
In terms of health manpower, the number of physicians per 10,000 population has increased on average 35.3% in all countries during the decade in question while the number of nurses per 10,000 population has decreased on average by 2.5%. How- ever, there are important differences between countries. Low-income countries show an important increase in the supply of physicians and nurses, whereas in middle-income countries the supply of physicians was smaller, and the supply of nurses decreased. In low-income countries, the number of physicians per 10,000 population increased by 80.5%, whereas the number of nurses per 10,000 population rose by 38.1%. On the other hand, in middle-income economies the number of physicians per 10,000 popu- lation has increased by 35.3%, while the number of nurses per 10,000 population decreased by 5.6%.
We used a general linear regression model as a way of estimating the relationship between infant mor-
tality, health manpower, and health expenditure. Regression equations calculated for the three samples are shown with relevant statistical coefficients in Table 2. The first column of Table 2 shows the regression coefficient of the sample that comprises all African and Latin American countries.
In the first column income level has been used as a dummy variable, so the first intercept is for low-in- come countries and the second is for middle-income countries. The intercept coefficients of the three samples captured a variety of factors that cause infant mortality that are not included in the model: poor sanitation, lack of safe water, malnutrition, and other important features such as the mothers level of education.
The regression coefficients from column 1 for all African and Latin American countries show differ- ences in the level of effectiveness for preventing infant mortality between physicians, nurses and expendi- tures. The differences between middle and low- income countries are significant at a 1% level using Fischers statistic (F(4,l) = (17.1)).
The estimators show that in low-income economies one dollar spent on health per capita could save 1.7 infant lives per 1000 live births. However, in middle- income countries the same dollar only saves 0.2 infant lives per 1000 li...