Download pdf - How should resources be reallocated between physicians and nurses in Africa and Latin America?

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

LATIN AMERICA?

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 population’s 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 country’s preferences.

Key words-health economics, health production mortality reduction

INTRODUCHON

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 developing 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-effectiveness of health providers and health expenditures in relation to improvements in the population’s health status is an important task for today’s policy makers.

In Africa and Latin America 48% of the population is under 15 years of age. The health status of such a young population depends primarily on expenditure and manpower allocation in primary health services [2]. 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 [3]; 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

SSM 33,6-o

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 [4]. 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 physician’s salary, the question arises as to whether the physician’s 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 rehydration therapy [5]. 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 resources 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?

METHODS

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 comprised 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 [9]. 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.

THE MODEL

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 observations; we limited our sample to 45 or 135 observations 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 population has increased by 35.3%, while the number of nurses per 10,000 population decreased by 5.6%.

RESULTS

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-income 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 mother’s level of education.

The regression coefficients from column 1 for all African and Latin American countries show differences in the level of effectiveness for preventing infant mortality between physicians, nurses and expenditures. The differences between middle and low- income countries are significant at a 1% level using Fischer’s 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 live births. This difference is due to differences in the distribution of the main causes of death and therefore, in the differences of most of mortality prevention within countries. An infant

Table 2. Regression analysis. Dependent variable: infant mortality per 1000 live births in Africa and Latin America-1973. 1978 and 1983

Countries

Low and mid-income Low-income coefficient coefficient (r-ratios) (I -ratios)

Mid-income coefficient (r-ratios)

Intercept

Expenditure per capita

Physicians per 10,000

Nurses per 10,000

Physicians times Nurses per 10,ooo

Number of countries Number of observations F value R=

‘151.66 ‘134.03 (24.37) (12.54)

-0.26 (-1.55)

-7.47 (- 5.04)

-3.41 (- 3.99)

0.32 (2.90) 45

135 262.4

0.95

160.73 (12.53) -1.74

(-1.66) - 19.47 (-0.93)

-3.18 (-1.60)

1.61 (0.74) 27 81 6.84 0.40

131.34 (9.74)

-0.18 (-1.32)

- 6.98 (-4.21)

-3.57 (-2.94)

0.31 (2.38) 18 54 10.12 0.61

The selected Low-income countries are: Ethiopia, Mauritania, Chad, Egypt, El Salvador, Ghana, Guinea, Honduras, Kenya, Lesotho, Liberia, Malawi, Mali, Morocco, Nicaragua, Niger, Nigeria, Rwanda, Sierra Leone, Senegal, Sudan, Tanzania, Togo, Uganda, Zaire, Zambia, Zimbabwe. Middle-income: Argentina, Botswana, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, Guatemala, Jamaica, Mauritius, Mexico, Panama, Paraguay, Peru, Tunisia, Uruguay, Venezuela.

726 VERONICA VAROAS-LAWS

mortality of 67 per 1000 live births in one country and of 131 in another not only signals the existence of different levels of access to basic services, but also suggests differences in the distribution of the main causes of death. Thus in Honduras, a low-income country with an infant mortality rate of 81 per 1000 in 1983, the list of causes of infant mortality is headed by diarrhea1 diseases, followed by disorders relating to short gestation and unspecified low birth weight (mother’s malnutrition). By contrast, in Paraguay, a middle-income country where infant mortality was 45 per 1000 live births in 1983, the list of causes includes certain conditions originating in the perina- tal period [17]. Infectious causes ranked foremost in low-income countries with high mortality rates, where the impact of infections can be reduced by basic and inexpensive measures such as vaccination, oral rehydration therapy. Conversely, non-infectious causes such as congenital abnormalities and heart disease, were characteristic of middle-income countries with low infant mortality; these conditions require more sophisticated and expensive health equipment.

At the primary health care level, where infant mortality is a key indicator, there are significant differences in the impact of physicians between middle and low-income countries. According to Table 2 in low-income countries one physician is able to save about 19.4 infant lives per 1000 and in middle-income countries one physician is able to save 6.9 infant lives per 1000 live births. The r-values are statistically significant especially in middle-income countries. However, we have to be cautious with the results for low-income countries because the statistics are not very significant. This can probably be ex- plained by the low utilization of services, so that infants in those countries die of diarrhoea, usually without ever getting near a physician or nurse.

With respect to nurses our regression shows that they are equally effective in both low-income countries and in middle-income countries. In low- income countries one nurse may reduce infant mortality by 3.18 per 1000 live births, while in middle- income countries this figure would be 3.57 per 1000 live births. The computed r-values for both coefficients are statistically significant.

The parameters from Table 2 permit us to estimate physicians-and-nurses combined impact on saving infant lives. If we express equation (1) in a reduced- form, as the marginal effect physicians and nurses have respectively on preventing infant mortality then:

and

d(IMR)/d(phy) = B2 + 84 x (nurse) (2)

d(IMR)/d(nurse) = 83 + 84 x (phy) (3)

substituting the parameters from Table 2 in equation (2) we see that a physician-and-nurse could save 7.7 lives and one physician may save 7.4 lives. Substitut- ing parameters from Table 2 in equation (3) we see that a nurse-and-physician will save 3.7 lives per 1000 live births, and a nurse could save 3.4 lives. The effect of either kind of professional on the infant mortality rate seems to go toward zero as the other kind of professional increases.

With the results from Table 2 it is possible to calculate the optimal mix of the two types of professionals in economic terms. If w2 and w3 are physicians and nurses’ wages, respectively, and the total wage bill is constant, then the constrained minimization of infant mortality rate from equations (2) and (3) gives:

(82 + B4 x Nurse)/(B3 + 84 x Phy) = w2/w3,

or

Nurse = (B3 x w2 - 82 x w3

+ B4 x w2 x Phy)/B4 x w3, (4)

which allows solving for the optimal staffing ratio as a function of the parameters and the wages. For illustrating those results let’s consider the regression results from Table 2 and assume that: (a) 60% of the health budget is spent on physicians’ and nurses’ salaries, which is the budget constraint [3], (b) physicians salaries are double nurses salaries, (c) in middle-income countries manpower salary is double that in low-income countries and (d) policy-makers have decided to secure a nurse-based system because the cost of training them is cheaper than training physicians and because nurses are less likely to emi- grate than physicians [18, 191. Based on those assumptions and considering from Table 1, the actual number of physicians by different income level, where the number of physicians in low-income countries varies from < 1 to 2 per 10,000 population, and > 2 to 6 in middle-income countries and then substituting in equation (4) gives us the optimal proportion of nurses per the actual number of physicians.

Figure 1 shows the optimal mix of physicians and nurses, according to a health budget constraint, and how some selected countries differ from this optimal. In middle-income countries such as Brazil, where there are about the same number of physicians as nurses (1: 1 .l) [IS] and the number of nurses has decreased in the 1973-83 decade (Table I), it would be more cost-effective to increase the number of nurses and substitute them for physicians.

DISCUSSION

At the manpower level, our estimators suggest that the care given by physicians and nurses is inter- changeable at the primary health care level. If this is the case, and it also happens that the health manpower salary is a proportional to their ability to save infant lives, as is the case in middle-income countries,

A

I

I I I I

0 1 2 3 4 5 6 Physiclanr per 10,000

+ Chile a6 EthloDiO 0 Tonzonlo x Rwanda

0 El Salvador 0 Brazil 1 Colombia

Fig. 1

Resource allocation between physicians and nurses 121

then policy-makers can make clear choices about whether to have a nurse-based primary health care or a physician-based system.

The coefficients in low-income countries are not very significant, it is imaginable because the delivery system is very fragile, so that neither the health expenditures nor the health professionals are reach- ing the people who demand them the most. However, the coefficients from Table 2 do suggest that physicians are three times more effective than nurses in low-income countries. This could be because their expertise is broader than nurses, and therefore more useful in the context of a great variety of problems connected to a deficient health delivery system. Ad- ditional, if their wages are only double the wages of the nurses, and physicians save three times as many lives as nurses do, therefore it seems reasonable to increase the number of physicians. Since we have found that monetary expenditure in low-income countries is more effective than in mid-income countries (Table 2), there is evidence to make a case for more public spending on health in low-income countries. This is especially true for those countries that utilize their manpower close to the most optimal mix and could not improve their health status by a reallocation of resources, but only from additional and new resources, i.e. Tanzania (Fig. 1).

Acknowledgements-I thank Professor Randy Ellis and Professor Tom McGuire from the Economics Department of Boston University, Dr Philip Musgrove from Pan- American Health Organization, Professor Donald Shepard from Harvard Institute for International Development and an anonymous referee for their essential suggestions. The views expressed here are solely my responsibility.

REFERENCES

1. The World Bank. Financing health services in develop ing countries; an agenda for reform. A World Bank Policy Study, p. 15, Washington, 1987.

2.

3.

4.

5.

10.

11.

12.

13.

14.

15.

16.

17.

18.

19.

United Nations Demographic Yearbook 1985, Vol. 39, pp. 394-400. New York, 1987. Henry M. and Chen L. Child survival: research and policy. In Child Survival. Strategies for Research, p. 12. Cambridge University Press, Cambridge, 1984. Mejia A. World trends in health manpower develop ment: a review. Wld Hlth Stat. Q. 33, 116117, 1980. United Nations Development Program. Human resource development for primary health care. Evaluation Study 9, 9, 1983. United Nations Demographic Yearbook 1980 Vol. 32, pp. 34&341. New York, 1987. International Monetary Fund. Government Finance Statistics 1978, Vol. 2, 24-309. New York, 1978. International monetary Fund. Government Finance Stat- istics 1987, Vol. 11, pp. 100-1013. Washington, 1987. International Monetary Fund. International Financial Statistics Supplement of Exchange Rate, pp. 16-17, Washington 1987. U.S. Bureau of The Census. Statistical Abstract of the United States, 108th edn, p. 450. Bureau of Census. Washington D.C., 1987. United Nations Demographic Yearbook 198.7, Vol. 34, pp. 151-152. New York, 1984. United Nations Demographic Yearbook 1986. Vol. 38. pp. 150-151. New York, 1988. World Health Organization World Health Statistics Annual 1977, Vol. 3, pp. 72-14. Geneva, 1971. World Health Organization World Health Statistics Annual 1980, Vol. 6, pp. 2-4, pp. 9-l 1. Geneva, 1980. World Health Organization World Health Statistics Annual 1983, Vol. 8, p. 397. Geneva, 1983. World Health Organization World Health Statistics Annual 1988, Vol. 9, pp. 44, 52. Geneva, 1988. Pan-American Health Organization. Health Conditions in the Americas 1981-1984, p.42. Pan American Sanitary Bureau, Regional OfBce of the World Health Organization, Washington 1986. Abel-Smith B. The price of unbalance health manpower. In Health Manpower Out of Balance. Conlpicts and Prospects, p. 30. CIOMS Conference Mexico, 1986. Mejia A., Pizurki H. and Royston E. Physician and Nurse Migration Analysis and Policy Implications, pp. 95, 125. World Health Organization, Geneva 1979.