The global distribution of physicians and nurses

The global distribution of physiciansand nurses

Heather Wharrad BSc PhD

Lecturer, Post-graduate Division of Nursing Studies,

The Medical School, University of Nottingham

and Jane Robinson FRCN MA PhD MIPD RGN ONC RHV HVT

Emeritus Professor, Post-graduate Division of Nursing Studies,

The Medical School, University of Nottingham,

Queen's Medical Centre, Nottingham, England

Accepted for publication 23 September 1998

WHARRADWHARRAD HH. & ROBINSON J. (1999)& ROBINSON J. (1999) Journal of Advanced Nursing 30(1), 109±120

The global distribution of physicians and nurses

Aim: To explore the global distribution of physicians and nurses and the

in¯uence of gross national product per capita on this distribution, using

available United Nations' (UN) sources. Objectives: to compare the international

distribution of physicians and nurses by country; to examine the in¯uence of

gross national product per capita (GNP) on the global distribution of physicians

and nurses; to explore the assumptions underlying the recommendations of The

World Development Report 1993 Investing in Health for health workforce

substitution; and to consider the implications for future studies of global health

labour distribution. Design: A database was compiled from various UN sources

on 147 countries. Using some of the variables from this database, a general linear

regression model for log GNP per capita on each of the two dependent variables

(log nurses and log physicians per 1000 population) was produced. Standard-

ized residuals obtained from these bivariate regressions were calculated and

plotted against each other to determine the relationship between the global

distribution of physicians to population and that of nurses. From this analysis

outlying countries could also be identi®ed. Results: Ratios of physicians to

population by country varied from 0á02 to 4á4 per 1000 population (or from 1 to

227 and 1±50,000 population), and nurses from 0á03 to 16á4 (or from 1 to 61 and

1±33,000 population). There was a positive correlation (r � 0á84, P < 0á001)

between the number of physicians per 1000 population and the number of

nurses per 1000 population. GNP explained 49% of the variation in physicians

and 40% in nurses. Ranking of countries according to their standardized

deviation from mean regression lines for GNP against health personnel in

countries with both the lowest incomes and lowest numbers of health

personnel, resulted in little change from the original rankings of ratios of

physicians and nurses relative to population. For some of the wealthiest

countries, there was a marked fall in global ranking and for some middle income

Correspondence: Jane Robinson, 8 School Road, Coalbrookdale,

Telford TF8 7DY, England. E-mail: [email protected]

Ó 1999 Blackwell Science Ltd 109

Journal of Advanced Nursing, 1999, 30(1), 109±120 Health and nursing policy issues

countries a marked improvement in ranking. Conclusion: 70% of the distribu-

tion of nurses globally can be explained by the distribution of physicians, and

the in¯uence of GNP per capita on the global distribution of physicians and

nurses appears to be substantial. In only a minority of the world's very poorest

countries is there evidence to suggest that higher numbers of nurses substitute

for low numbers of physicians. Standardization of the distributions by GNP

demonstrates that many countries (but not the poorest) regress to within one

standard deviation of the mean expected distribution. This suggests that

countries could set optimum levels of physicians and nurses within the limits of

their GNP. More realistically, the ®ndings suggest that recommendations for

modi®cation of the structure of countries' health labour forces as a component

of health care reform may be more dif®cult to achieve than at ®rst appears. The

potential unreliability of the data sources used, and the implications for the

accuracy of the ®ndings, are discussed.

Keywords: `Giffen Goods', global labour distribution, health economics,

health workforce planning, nurses, physicians, The World Bank, United Nations

INTRODUCTION

Factors in¯uencing the global distribution of health labour

have been the subject of surprisingly little study given the

assumed ability of governments to manipulate the struc-

ture of health labour forces in many of the recommended

policies for contemporary health care reform. Changing

the proportions of skill mixes, substituting less expensive

staff for more costly, and encouraging health staff to work

in areas of greatest health need (such as primary health

care), all represent strategies associated with an assumed

potential to manipulate the health labour force (Journal of

the Royal Society of Medicine 1995, Cooper 1995). Yet

there is a notable absence of any theoretical development

to justify the recommendations and a dearth of evidence

as to whether they produce the intended effects. Human

resource development in health care apparently remains a

largely pragmatic art.

Gray (1993), a health economist, referring to the

Organization for Economic Co-operation and Develop-

ment (OECD, 1991) Health Data suggested that the level of

nursing provision is solely a function of a country's

wealth:

¼ richer countries have more nurses, poorer countries have

fewer. If the ®gure was to be extended to the left, in order to

include countries with very low GNP per head and correspond-

ingly high levels of mortality and morbidity, we would ®nd that

levels of nursing were generally very low indeed.

(Gray 1993 p. 157)

Hertz et al. (1994) observed that although increases

in medical and physician facilities alone were not an

absolute solution to poor health, in the case study countries

identi®ed in their study, health care development had

been adapted to meet their own unique socio-economic

conditions. Curative and preventive services were direct-

ed towards existing patterns of morbidity and malnutri-

tion. Thus the provision and orientation of medical

facilities were seen to be important contributory factors

within an overall context of governmental attitudes

towards social and health development.

Hertz et al.'s conclusions are re¯ected in the approach

of the The World Development Report 1993: Investing in

Health to the interplay between health, health policy and

economic development (The World Bank 1993). The

Report's main recommendations for the improvement of

health focus on poverty reduction measures, in particular

strengthening the health of households through recogni-

tion of the strong positive correlations between: per capita

income; women's education; public health infrastructure;

and improved life expectancy (The World Bank 1993

pp. 37±51). Investing in Health also comments on the need

to address the fundamental problems of the imbalance of

health personnel in nearly all countries. It observes that:

there are too many specialists and not enough primary

care providers; health workers are concentrated in urban

areas; training in public health, health policy and health

management has been relatively neglected; and medical

training is subsidized even though many physicians work

in the private sector and earn high incomes (The World

Bank 1993 p. 139).

Investing in Health recommends a number of health

workforce strategies to rectify this situation including:

forms of labour substitution for improving the balance

between primary care providers and specialists; setting

recommended ratios of physicians to population (0á1 per

1000) and graduate nurses to physicians (between 2 and 4

to 1 physician, or 0á2±0á4 per 1000 population) for

H. Wharrad and J. Robinson

110 Ó 1999 Blackwell Science Ltd, Journal of Advanced Nursing, 30(1), 109±120

minimum essential clinical services to the poor; the

introduction of measures to control physician oversupply;

curtailing specialist training; attracting primary care pro-

viders to under-served areas; reforming the ®nance of

health training; and redirecting the content to public

health, management, policy and planning (The World

Bank 1993 pp. 139±144).

Investing in Health supports strongly the idea of labour

substitution, observing that `¼ the relatively high ratio of

nurses to physicians in Sub-Saharan Africa is a good sign'

and stating that whilst `There is no optimal level of

physicians per capita or optimal nurse-to-physician ratio, a

rule of thumb is that nurses should exceed physicians by at

least two to one' (The World Bank 1993 p. 139). The World

Bank's emphasis on the importance of maintaining an

overview of the ratio of nurses to physicians (`®ve to one in

Africa but well under two to one in China, India, Latin

America, and the Middle Eastern Crescent' (The World

Bank 1993 p. 139) led us in this paper to include graphs

showing the relative numbers of nurses to physicians, in

addition to correlations for the single occupations.

In supporting an inverse relationship between physi-

cians and nursing staff The World Bank's recommenda-

tions discount the evidence that the supply of both

physicians and nurses is associated with a country's

wealth in terms of GNP per capita. Instead, Investing in

Health endorses a view that the supply of nurses behaves

(or ideally should behave) as a `Giffen good' (Whitehead

1970) that is: consumption of a cheaper commodity (i.e.

nurses) is high when material resources are low, and that

replacement with a more expensive commodity (i.e.

physicians) takes place as available resources increase.

In its recommendations for the manipulation of health

labour, The World Bank proposes that by adopting its

recommended measures, governments will be able to

in¯uence positively some of the most persistent health

problems affecting the poor, such as infant, under 5, and

maternal mortality. The World Bank's position on health

labour in the World Development Report 1993 thus

mirrors in several substantial ways that adopted in World

Health Organization reports of the 1980s and 1990s on

nursing and midwifery development (for a summary see

Robinson 1997).

The World Bank's recommended measures therefore

assume that governments have both the will and the

power to in¯uence such events. The study reported here

investigates an alternative premise: that it is the in¯uence

of socio-economic factors, as mediated by GNP per capita

rather than the intentional acts of governments, which

determine the inherent structure of the health labour force

in individual countries. The study tests this assumption

using data available from The World Bank and other

United Nations' sources. The focus is on the global

distribution of health personnel as shown by the relative

numbers of physicians and nurses in relation to GNP.

THE STUDY

Materials and methods

A data base was compiled from the following sources:

· Nurse to doctor ratios (1988±1992) from The World

Bank (1993, Table A.8);

· Ratios of physicians to population (1990±1995) and

GNP per capita (US$ 1995, from The World Bank 1997,

Table A.6);

· Missing data for GNP (US$ 1994) for Cuba and

Afghanistan, from United Nations' Children's Fund

(1997, Table 1);

· Gross domestic product (GDP, US$ 1994) for Malta and

Belize from the United Nations' Human Development

Programme (1997, Table 1).

Complete data for ratios of physicians and nurses to

population (the latter inferred from the nurse to physician

ratios), plus GNP per capita, were established from these

combined sources for 147 countries (Tables 1 and 2).

Issues concerning the reliability of these data are discussed

under `Study limitations' at the conclusion of thispaper.

Copyright permission for the use of the above data was

granted by Oxford University Press and The World Bank.

Using these data, statistical methods were used to

analyse the relationship between GNP and the global

distribution of physicians and nurses. Prior to all regres-

sion analyses, exploratory analysis was carried out to

determine the linearity of the relationship between

variables. Non-linearity between variables (which often

arose for these data) was addressed by log transformation

(Bland & Altman 19961 ). The purpose of the analysis was to

determine how much of the variation in the dependent

variables could be explained by the variation in the

independent variables and also to identify outlying

countries. The sample estimate of R2 tends to be an over-

estimate of the population parameter. The R2 values

presented are the adjusted R2 which compensates for the

optimistic bias of R2. This adjusted ®gure takes account of

the number of variables in the model and the sample size.

Standardized residuals obtained from the bivariate re-

gressions of GNP on physicians/1000 and GNP on nurses/

1000 were calculated. These residuals were plotted to

produce a scatter-gram (physicians/1000 on the abscissa

and nurses/1000 on the ordinate axis) which was divided

into nine quadrants representing standard deviations (SDSD)

from the mean (see results section, Figure 4). The central

quadrant contains countries which lie within 1 SDSD of the

two regression lines (those for physicians and for nurses) Ð

hence the numbers of health personnel were as predicted

by the model. Countries lying in the top right hand

quadrant were those which have high numbers of health

personnel for their level of GNP. The lower left hand

quadrant contains those countries which have fewer health

Health and nursing policy issues Global distribution of physicians and nurses

Ó 1999 Blackwell Science Ltd, Journal of Advanced Nursing, 30(1), 109±120 111

personnel than expected for their level of GNP. This plot

therefore re-distributes the countries according to the

deviations from the expected numbers of health personnel

relative to GNP.

RESULTS

Tables 1 and 2 show that ratios of physicians to popula-

tion vary from 0á02 to 4á4 per 1000 population (or from 1

physician to 227 and 1±50,000 population). Nurses range

from 0á03 to 16á4 (or from 1 to 61 and 1±33,000 popula-

tion). Figure 1 is a scatter-gram showing the number of

physicians per 1000 population against the number of

nurses per 1000 population for 147 countries plotted on a

log scale. The size of the positive correlation coef®cient for

these indices is 0á84, this is highly signi®cant (P < 0á001)

thus indicating that countries having low numbers of

physicians also have low numbers of nurses (Table 3). The

dotted lines superimposed on the graph represent The

World Bank's (1993) minimum recommended ratios of

physicians to population (P1 to P2) and the lower recom-

mended ratio of nurses (N1 to N2) to population.

Countries in the lower left hand segment A (Figure 1)

with lower than recommended ratios of both physicians

and nurses to population are all from The World Bank's

(1997) low-income economy group (between $80 and $730

per capita GNP in 1995). Those in the upper left hand

segment (B) with similarly low ratios of physicians but

recommended ratios of nurses are also low-income group

economies. Countries on line P1 to P2 with the recom-

mended ratio of physicians and a recommended, or

higher, ratio of nurses include further low-income

economies and three from the lower-middle-income group

where the highest GNP per capita income was $1170 in

1995.

No countries fall into the lower right hand segment (D)

of Figure 1 with high ratios of physicians but low ratios of

nurses. However, approximately 10 countries all have

recommended ratios of nurses (that is between 0á2 and 0á4per 1000 population) with much higher than recommend-

ed ratios of physicians. The remaining countries in the

upper right hand segment (C) cluster around the mean

regression line which shows that the ratios of physicians

and nurses to population increase in a positive direction

more or less synonymously (Table 2).

Figures 2 and 3 show the associations between GNP per

capita and rates of health personnel. Forty-nine per cent of

the variation in the ratios of physicians and 40% of the

variation in nurses per 1000 population is explained by

GNP per capita (Table 3).

When standardized residuals from the regression ana-

lyses shown in Figures 2 and 3 were plotted against each

Figure 1 Scatterplot showing the global distribution of physicians per 1000 population and nurses per 1000 population. The dashed

lines (P1±P2 and N1±N2) show The World Bank minimum recommendations for the number of physicians and nurses per 1000

population. Segment A shows countries with below the minimum recommendation for physicians and nurses; segment B shows

countries with below the recommended number of physicians and above the recommended number of nurses; Segment C contains

countries with above the recommended levels for physicians and nurses. No countries lie in segment D. Data are plotted on logarithmic

scales. Countries labelled are either examples used in the text or examples of outliers.



other (Figure 4) some redistribution of countries occurs.

This change in global ranking for physicians can be seen

by comparing the ranks shown in columns 3 and 8 in

Tables 1 and 2; the change in ranking for nurses can be

seen by comparing the ranks listed in columns 5 and 10. In

countries with the least physicians and nurses (Segment

A, Figure 1), ranking according to the standard deviation

from the mean regression lines for GNP against health

Figure 2 Scatterplot of GNP per capita ($) and physicians per 1000 population. The dashed line (P1±P2) shows The World Bank

minimum recommendations for the number of physicians per 1000 population. Data are plotted on logarithmic scales. Countries

labelled are either examples used in the text or examples of outliers.

Figure 3 Scatterplot of GNP per capita ($) and nurses per 1000 population. The dashed line (N1±N2) shows The World Bank minimum

recommendations for the number of nurses per 1000 population. Data are plotted on logarithmic scales. Countries labelled are either

examples used in the text or examples of outliers.



personnel makes little difference to their rank orders.

Malawi, Central African Republic and Rwanda retain their

positions with the lowest ratios in the world of physicians

and nurses to population and are joined by Burkina Faso,

Ghana and Niger from Segment B.

Among countries on the line P1 to P2 (Figure 1) where

initially all physicians to population rank 123rd, several

large shifts are identi®ed but countries in Segment C show

some of the most marked shifts. The relative positions of

some poorer countries improve dramatically for both

physicians and nurses (Albania, Armenia, Krgyzstan,

Libya, Macedonia, Tajikistan, Syria and Surinam) and

the position of China improves for physicians2 . On the

other hand, some of the world's wealthiest countries

experience an apparent worsening of their relative posi-

tions. Australia, Germany, Norway, Sweden, Switzerland,

United Arab Emirates, United Kingdom and the United

States of America all move down the rank order by at least

44 places, and in some cases more than 50 places, when

ranked according to their standardized residuals for

physicians and nurses to population regressed on GNP.

So, although a large proportion of high income countries

have ratios of physicians and nurses to population within

1 SDSD of the mean of that expected by GNP, many less

wealthy countries have more physicians and nurses than

predicted by their GNP. There is an absence of countries

lying in the top left hand segment of Figure 4, which

would have indicated high numbers of nurses and low

numbers of physicians relative to GNP.

DISCUSSION

GNP per capita is an important in¯uencing variable in the

distribution of physicians and nurses. Forty-nine per cent

of the variation in physicians and 40% in nurses can be

explained by this variable. The ®nding that the statisti-

cally signi®cant positive correlation (0á84) shown for the

distribution of nurses against that of physicians suggests

patterns of workforce development more related to health

service expansion than to rational patterns of labour

substitution. This makes intuitive sense, but how these

developments work in practice is more dif®cult to calcu-

late. It may be that with health care reforms, the respon-

sibility for particular health care `tasks' is shifted

deliberately between occupational groups thus giving an

illusion of `substitution' whilst the relative size of an

individual group is unaltered, or even expanding. There is

certainly no indication from these data that countries are

substituting less costly nurses for more expensive

physicians, although the `Giffen goods' hypothesis is very

Figure 4 Scatterplot of the standardized residuals for the regression of physicians/1000 on GNP (per capita $) against the standardized

residuals for the regression of nurses/1000 on GNP (per capita $). Countries labelled are either examples used in the text or examples of

outliers. The ®gure is divided into nine quadrants de®ned by standard deviations from the regression lines. The central quadrant shows

countries within one standard deviation for the ratios of physicians and nurses to 1000 population. The outlying quadrants show

variations of between 1 and 3 standard deviations.



weakly supported in a small number of low income

economies.

Although the data appear therefore weakly to support

The World Bank's observations on the ratios of physicians

and nurses in some Sub-Saharan African countries, if the

supply of nurses globally was inversely related to the

supply of physicians (as in the Giffen goods hypothesis)

then the correlation coef®cient would be negative instead

of strongly positive as shown with an r-value of 0á84

(Table 3). The absence of countries in the top left hand

segment, Figure 4 (low physicans, high nurses), also

demonstrates that no country strongly supports the `Giffen

Table 1 Physicians and nurses per 1000 population with global rankings, GNP per capita ($), standardized residuals from regression of

physicians per 1000 population on GNP and nurses per 1000 population on GNP, with global rankings, for countries with The World

Bank's (1993) recommended ratio of physicians to population (countries on the line P1±P2; Figure 1) or below this recommendation

(countries in segments A and B). Countries in segment A (Figure 1) have below the recommended number of nurses per 1000

population, those in segment B the recommended number or above.

Country

Physicians

per

1000 Rank

Nurses

per

1000 Rank

GNP

$ per

cap

Standardized

residuals for

physicians GNP Rank

Standardized

residuals for

nurses on GNP Rank

1 2 3 4 5 6 7 8 9 10

Line P1-P2 (Figure 1)

Afghanistan 0á10 123á0 0á10 142á5 280 )0á56 103á0 )1á67 137á0Benin 0á10 123á0 0á58 111á0 370 )0á73 114á0 )0á22 87á0Burundi 0á10 123á0 0á43 114á5 160 )0á21 74á0 )0á06 76á0Cambodia 0á10 123á0 1á00 97á5 270 )0á54 101á0 )0á43 45á0Cameroon 0á10 123á0 0á64 106á5 650 )1á08 130á0 )0á42 107á0Comoros 0á10 123á0 0á60 109á5 470 )0á88 119á0 )0á32 96á0Cote dI voire 0á10 123á0 0á40 118á5 660 )1á09 131á0 )0á86 126á0Guinea 0á10 123á0 0á43 114á5 550 )0á98 125á0 )0á70 121á0Haiti 0á10 123á0 0á40 118á5 250 )0á49 96á5 )0á36 100á0Madagascar 0á10 123á0 0á40 118á5 230 )0á44 92á0 )0á32 97á0Mali 0á10 123á0 0á25 132á5 250 )0á49 96á5 )0á78 124á0Mauritania 0á10 123á0 0á50 112á5 460 )0á87 118á0 )0á47 109á0Mynamar 0á10 123á0 0á40 118á5 220 )0á41 91á0 )0á29 93á0Nepal 0á10 123á0 0á20 139á0 200 )0á35 83á0 )0á87 127á0Papua NG 0á10 123á0 1á40 88á5 1160 )1á44 140á0 )0á01 73á0Philippines 0á10 123á0 0á31 128á0 1050 )1á38 138á0 )1á32 134á0Senegal 0á10 123á0 0á26 130á5 600 )1á03 129á0 )1á20 132á0Swaziland 0á10 123á0 0á20 139á0 1170 )1á45 141á0 )1á78 142á0Togo 0á10 123á0 0á40 118á5 310 )0á62 107á0 )0á47 108á0Yemen 0á10 123á0 8á20 22á0 260 )0á51 99á0 )2á35 1á0Zaire 0á10 123á0 0á21 136á0 120 )0á04 61á0 )0á56 114á0Zambia 0á10 123á0 0á90 101á0 400 )0á78 116á0 )0á13 63á0Zimbabwe 0á10 123á0 0á61 108á0 540 )0á97 124á0 )0á37 102á0

Segment B (Figure 1)

Sudan 0á09 135á0 0á24 134á0 480 )0á99 127á0 )1á15 131á0Sierra Leone 0á07 136á0 0á35 125á0 180 )0á61 106á0 )0á31 95á0Ghana 0á04 138á0 0á36 123á0 390 )1á61 145á0 )0á68 119á0Uganda 0á04 138á0 0á34 127á0 240 )1á30 136á0 )0á48 111á0Burkina Faso 0á03 142á0 0á25 132á5 230 )1á54 144á0 )0á74 122á0Niger 0á03 142á0 0á20 139á0 220 )1á51 142á0 )0á92 129á0Tanzania 0á03 142á0 0á22 135á0 120 )1á14 132á0 )0á52 112á0Mozambique 0á02 146á0 0á26 130á5 80 )1á26 135á0 )0á16 82á0

Segment A (Figure 1)

Central Africa 0á04 138á0 0á10 142á5 340 )1á52 143á0 )1á77 141á0Chad 0á03 142á0 0á03 147á0 180 )1á39 139á0 )2á62 145á0Ethiopia 0á03 142á0 0á07 144á0 100 )1á03 128á0 )1á46 135á0Malawi 0á02 146á0 0á06 145á0 170 )1á73 146á0 )1á87 143á0Rwanda 0á02 146á0 0á03 146á0 180 )1á76 147á0 )2á42 144á0



Table 2 Physicians and nurses per 1000 population with global rankings, GNP per capita ($), standardized residuals from regression of

physicians per 1000 population on GNP and nurses per 1000 population on GNP, with global rankings, for countries above the World

Bank's (1993) recommended ratio of physicians to population and the recommended number of nurses per 1000 population (segment C).

Country

Physicians

per

1000 Rank

Nurses

per

1000 Rank

GNP

$ per

cap

Standardized

residuals for

physicians GNP Rank

Standardized

residuals for

nurses on GNP Rank

1 2 3 4 5 6 7 8 9 10

Segment C (Figure 1)

Ukraine 4á40 1á0 12á20 3á0 1630 1á82 11á0 1á77 7á0Belarus 4á10 3á0 11á40 6á0 2070 1á61 15á0 1á58 11á0Georgia 4á10 3á0 8á70 16á5 440 2á56 1á0 2á13 4á0Spain 4á10 3á0 4á30 50á5 13580 0á44 43á0 )0á26 88á0Greece 4á00 5á5 3á60 53á0 8210 0á73 34á0 )0á16 81á0Lithuania 4á00 5á5 9á70 13á0 1900 1á64 14á0 1á48 14á0Azerbaijan 3á90 7á0 10á80 8á0 480 2á46 2á0 2á28 2á0Russian Fed 3á80 8á0 8á70 16á5 2240 1á49 19á0 1á30 19á0Belgium 3á70 9á0 0á37 122á0 24710 )0á03 59á0 )2á78 147á0Cuba 3á60 10á5 6á10 38á5 1170 1á84 10á0 1á31 18á0Kazakhstan 3á60 10á5 9á00 15á0 1330 1á76 12á0 1á60 9á0Bulgaria 3á50 12á5 8á50 19á0 1330 1á73 13á0 1á55 13á0Moldova 3á50 12á5 10á40 10á0 870 2á00 5á0 1á95 5á0Hungary 3á40 14á0 7á00 28á5 4120 1á01 26á0 0á79 30á0Germany 3á30 16á0 5á60 44á0 27510 )0á20 71á0 )0á38 103á0Norway 3á30 16á0 16á40 1á0 31250 )0á28 77á0 0á52 37á0Uzbekistan 3á30 16á0 10á60 9á0 970 1á88 9á0 1á91 6á0Kyrgyzstan 3á20 19á0 6á10 38á5 700 2á05 4á0 1á57 12á0Turkmenistan 3á20 19á0 7á70 23á0 920 1á88 8á0 1á65 8á0Uruguay 3á20 19á0 0á64 106á5 5170 0á81 32á0 )1á49 136á0Armenia 3á10 22á5 6á40 36á0 730 1á99 6á0 1á60 10á0Estonia 3á10 22á5 6á50 33á5 2860 1á15 23á0 0á91 24á0Sweden 3á10 22á5 10á10 11á5 23750 )0á16 69á0 0á22 55á0Switzerland 3á10 22á5 14á00 2á0 40630 )0á50 98á0 0á24 53á0Czech Repub. 3á00 26á0 9á40 14á0 3870 0á93 28á0 1á09 21á0Iceland 3á00 26á0 7á10 27á0 24950 )0á22 75á0 )0á12 77á0Slovak 3á00 26á0 7á50 25á5 8200 0á47 41á0 0á50 38á0Denmark 2á90 29á0 6á70 31á5 29890 )0á37 85á0 )0á26 89á0Latvia 2á90 29á0 6á40 36á0 2270 1á23 22á0 1á02 23á0Portugal 2á90 29á0 3á20 56á0 9740 0á33 47á0 )0á36 99á0France 2á80 31á0 5á70 42á5 24990 )0á29 78á0 )0á32 98á0Argentina 2á70 32á5 2á10 78á0 8030 0á38 45á0 )0á64 118á0Finland 2á70 32á5 12á00 4á0 20580 )0á20 72á0 0á45 43á0Austria 2á60 34á0 8á40 21á0 26890 )0á40 89á0 )0á01 72á0Malta 2á50 36á0 11á80 5á0 6021 0á49 40á0 1á07 22á0Netherland 2á50 36á0 8á50 19á0 24000 )0á37 86á0 0á06 68á0US 2á50 36á0 7á00 28á5 26980 )0á44 93á0 )0á17 83á0Poland 2á30 38á0 5á90 40á5 2790 0á89 30á0 0á84 28á0Australia 2á20 39á5 8á50 19á0 18720 )0á33 81á0 0á19 59á0Canada 2á20 39á5 11á20 7á0 19380 )0á35 82á0 0á42 46á0Macedonia 2á10 42á5 5á70 42á5 860 1á54 17á0 1á41 15á0New Zealand 2á10 42á5 10á10 11á5 14340 )0á21 73á0 0á48 40á0Slovenia 2á10 42á5 5á90 40á5 2950 0á77 33á0 0á81 29á0Tajikistan 2á10 42á5 7á60 24á0 340 2á11 3á0 2á14 3á0Croatia 2á00 45á5 4á40 49á0 3250 0á67 35á0 0á49 39á0Ireland 2á00 45á5 7á50 25á5 14710 )0á27 76á0 0á20 56á0



Table 2 (Continued)

Country

Physicians

per

1000 Rank

Nurses

per

1000 Rank

GNP

$ per

cap

Standardized

residuals for

physicians GNP Rank

Standardized

residuals for

nurses on GNP Rank

1 2 3 4 5 6 7 8 9 10


Egypt 1á80 48á0 2á20 76á5 790 1á45 20á0 0á59 33á0Japan 1á80 48á0 6á80 30á0 39640 )0á98 126á0 )0á40 104á0Rumania 1á80 48á0 4á30 50á5 1480 1á06 24á0 0á88 26á0Italy 1á70 50á0 4á10 52á0 19020 )0á58 104á0 )0á48 110á0China 1á60 52á5 1á10 95á5 620 1á49 18á0 0á09 64á0Jordan 1á60 52á5 3á00 60á5 1510 0á94 27á0 0á54 36á0Panama 1á60 52á5 2á40 71á5 2750 0á57 38á0 0á03 69á0Venezuela 1á60 52á5 2á70 67á0 3020 0á51 39á0 0á09 65á0Ecuador 1á50 55á5 1á60 84á0 1390 0á93 29á0 0á01 70á0UK 1á50 55á5 3á00 60á5 18700 )0á68 112á0 )0á75 123á0Albania 1á40 59á0 4á70 48á0 670 1á32 21á0 1á36 17á0Bahamas 1á40 59á0 6á40 36á0 11940 )0á47 94á0 0á16 62á0Brazil 1á40 59á0 0á40 118á5 3640 0á27 49á0 )1á73 139á0Saudi Arabia 1á40 59á0 0á20 139á0 6180 )0á06 62á0 )2á63 146á0Singapore 1á40 59á0 3á50 54á0 26730 )0á96 123á0 )0á79 125á0Costa Rica 1á30 63á5 2á20 76á5 2610 0á41 44á0 )0á02 74á0Hong Kong 1á30 63á5 5á50 45á0 22990 )0á94 122á0 )0á31 94á0Mexico 1á30 63á5 1á90 80á5 3320 0á26 53á0 )0á28 92á0Taiwan 1á30 63á5 2á60 68á5 12790 )0á58 105á0 )0á68 120á0KoreaRep 1á20 66á0 6á70 31á5 9700 )0á48 95á0 0á31 51á0Barbados 1á10 69á0 5á20 46á0 6560 )0á32 80á0 0á28 52á0Chile 1á10 69á0 2á50 70á0 4160 )0á03 60á0 )0á14 79á0Columbia 1á10 69á0 1á40 88á5 1910 0á45 42á0 )0á27 90á0Domin. Rep. 1á10 69á0 0á80 104á0 1460 0á62 37á0 )0á64 117á0Turkey 1á10 69á0 1á50 86á5 2780 0á22 57á0 )0á40 105á0Libya 1á00 72á5 2á90 62á0 269 1á58 16á0 1á39 16á0Peru 1á00 72á5 2á60 68á5 2310 0á24 55á0 0á19 58á0Antigua 0á90 74á5 4á80 47á0 6770 )0á52 100á0 0á20 57á0St Kits 0á90 74á5 6á50 33á5 5170 )0á35 84á0 0á61 31á0Algeria 0á80 78á5 3á20 56á0 1600 0á27 50á0 0á57 34á0El Salvador 0á80 78á5 0á90 101á0 1610 0á26 52á0 )0á58 115á0Guatamala 0á80 78á5 1á50 86á5 1340 0á38 46á0 )0á02 75á0Surinam 0á80 78á5 3á10 58á5 880 0á64 36á0 0á85 27á0Syrian A 0á80 78á5 1á00 97á5 112 1á91 7á0 0á88 25á0UAE 0á80 78á5 2á30 74á0 17400 )1á21 134á0 )0á95 130á0Nicaragua 0á70 82á5 1á20 92á5 380 1á03 25á0 0á42 47á0Trinidad 0á70 82á5 2á80 64á0 3770 )0á39 88á0 0á01 71á0South Africa 0á61 84á0 2á75 66á0 3160 )0á40 90á0 0á08 66á0Tunisia 0á60 85á0 2á80 64á0 1820 )0á08 64á0 0á38 48á0Belise 0á50 89á0 1á30 90á5 5590 )0á94 121á0 )0á89 128á0Bolivia 0á50 89á0 0á90 101á0 800 0á27 51á0 )0á22 86á0Bosnia 0á50 89á0 1á80 82á5 725 0á33 48á0 0á45 41á0Dominica 0á50 89á0 0á35 125á0 2990 )0á55 102á0 )1á75 140á0Jamaica 0á50 89á0 2á00 79á0 1510 )0á13 66á0 0á17 61á0St Lucia 0á50 89á0 2á80 64á0 3370 )0á63 108á0 0á07 67á0St Vincent 0á50 89á0 3á10 58á5 2280 )0á38 87á0 0á36 49á0Honduras 0á40 94á5 1á20 92á5 600 0á24 56á0 0á19 60á0Malaysia 0á40 94á5 1á56 85á0 3890 )0á92 120á0 )0á54 113á0



good' hypothesis with low levels of physicians relative to

GNP compensated for by high levels of nurses.

On the basis of the ®ndings reported here it appears

that, amongst low income countries, there is a group of

approximately a dozen countries where the absolute

under-supply of physicians and nurses is so great that

even when ranked according to standardized residuals

from the mean regression lines for health personnel

predicted from GNP, they remain bottom of the global

ranking. However, once The World Bank's recommended

ratio of 0á1 physicians and 0á2±0á4 nurses per 1000

population is reached, some countries show marked shifts

in ranking indicating that relative to their GNP per capita

they show high ratios of physicians and nurses to popu-

lation. What variables are contributing to this outcome is

unknown.

Wealthy countries experience the reverse of this situa-

tion. Just as Paul Getty was reported to have observed that

a rich man can only wear one pair of shoes at a time, or eat

so much food, it appears that, in relative terms, wealthier

countries can only absorb so many physicians and nurses

to population. Hence, relative to GNP they are doing worse

than other middle income countries. This begins to

suggest that there may be an optimum ratio of health

personnel to population in relation to income although,

clearly, this would ideally be dependent on a range of

local geographical, demographic and epidemiological

features. This tentative hypothesis takes the argument

back to the notion of health workforce planning and the

evidence presented here suggests that this activity has, to

date, had little demonstrable effect on the global distribu-

tion of health labour. Nevertheless, our ®ndings suggest

that rather than aiming for globally determined ratios of

health personnel to population, countries could instead

set optimum levels within the limits of their own GNP per

capita. Deployment of health personnel should then be

aimed at individual countries' most pressing health

problems.

Table 2 (Continued)

Country

Physicians

per

1000 Rank

Nurses

per

1000 Rank

GNP

$ per

cap

Standardized

residuals for

physicians GNP Rank

Standardized

residuals for

nurses on GNP Rank

1 2 3 4 5 6 7 8 9 10


Morocco 0á40 94á5 1á80 82á5 1110 )0á14 67á0 0á24 54á0Vietnam 0á40 94á5 2á30 74á0 219 0á86 31á0 1á29 20á0Iran 0á32 97á0 0á35 125á0 1033 0á30 79á0 )1á21 133á0Congo 0á30 99á0 0á90 101á0 680 )0á10 65á0 )0á14 78á0EquGuinea 0á30 99á0 0á60 109á5 380 0á26 54á0 )0á21 85á0Paraguay 0á30 99á0 1á30 90á5 1690 )0á67 111á0 )0á27 91á0Bangladesh 0á20 105á0 0á30 129á0 240 0á17 58á0 )0á60 116á0Bhutan 0á20 105á0 0á50 112á5 420 )0á18 70á0 )0á42 106á0Indonesia 0á20 105á0 1á10 95á5 980 )0á70 113á0 )0á14 80á0Iraq 0á20 105á0 0á20 139á0 1036 )0á74 115á0 )1á71 138á0Kiribati 0á20 105á0 2á40 71á5 920 )0á66 110á0 0á59 32á0LaoPDR 0á20 105á0 1á18 94á0 350 )0á06 63á0 0á45 44á0Namibia 0á20 105á0 3á20 56á0 2000 )1á14 133á0 0á45 42á0Solomon 0á20 105á0 2á30 74á0 910 )0á65 109á0 0á56 35á0Thailand 0á20 105á0 1á90 80á5 2740 )1á34 137á0 )0á18 84á0Nigeria 0á15 110á0 0á90 101á0 260 )0á14 68á0 0á35 50á0Sri Lanka 0á14 111á0 0á71 105á0 700 )0á82 117á0 )0á37 101á0

Table 3 Linear regression models for log physicians per 1000 population on log GNP per capita ($), log nurses per 1000 population on

log GNP and log nurses per 1000 population on log physicians per 1000 population

Dependent

variable

Independent

variable

Correlation

coef®cient r

R2

(SESE)

Constant

(SESE)

Slope

coef®cient (SESE) ANOVAANOVA

Log nurses Log physicians 0á840 70% (34) 0á409 (0á029) 0á783 (0á042) F = 348, P < 0.001

Log physicians Log GNP 0á700 49% (47) )2á388 (0á186) 0á676 (0á057) F = 142, P < 0.001

Log nurses Log GNP 0á633 40% (48) )1á578 (0á189) 0á566 (0á058) F = 97, P < 0.001



Whether the apparent improved numbers of health

personnel relative to GNP in some countries translates

into relatively better health outcomes is not addressed in

this paper and requires further investigation and analysis

using the databases described. Hertz et al. (1994), using

United Nations' sources in an ecological study, explored

the relationship between social and environmental factors

and life expectancy at birth, and infant and maternal

mortality in 66 countries representing all The World Bank

economic regions. Univariate statistics for countries

whose data were used in the regression analyses included

available medical personnel and hospital beds per 10 000

population. They identi®ed outlier case study countries

which showed some of the best outcome results.

Twenty years ago, Cochrane et al. (1978) explored

health service ìnput' and mortality òutput' in 18 devel-

oped countries (GNP exceeding $2000 per capita at

current prices). Minimum, median and maximum indices

for health service inputs, including doctors, nurses,

obstetricians and midwives were presented together with

the percentage of GNP spent on health, but correlations

between GNP and health workforce data were not

produced. Despite unexplained anomalies found in the

Cochrane et al. study between nurses' and doctors'

associations with mortality rates, its ®ndings have not

been replicated.

Particular solutions may need to be sought for some of

the world's poorest countries. Several Sub-Saharan and

some other developing world countries appear weakly to

support the `Giffen good' hypothesis of higher ratios of

nurses to population compared with physicians in low

income situations. How health personnel are deployed in

these countries, and with what success in relation to

health outcomes, should be the subject of in-depth case

studies.

Study limitations

Until recently there was a paucity of global data available

on nursing and medical staff. The data used in this study

may therefore be fairly crude and there are undoubted

limitations in the absence of clear de®nitions. First, for

example, we do not know how the health personnel

represented in these data are deployed between the acute

hospital and primary care, between the public and private

sectors, between general community practice and high

technology specialties.

Second, de®nitions of the occupational groups are not

necessarily standard, as United Nations sources rely

heavily on the quality of data provided at country level.

The World Bank de®nes physicians as `graduates of any

faculty or school of medicine who are working in the

country in any medical ®eld (practice, teaching, research).

Data from government statistical year books, World Bank,

OECD and WHO (The World Bank 1997 p. 29). With

regard to nurses, The World Bank (1993 p. 197) states that

`The de®nition of nurse includes only registered nurses

and registered midwives'. However, there is still no global

standard de®nition of a `registered' nurse (International

Labour Organization 1977). Some `nurses' counted at

country level in these data bases may have `quali®ed'

directly from school as young as 14±16 years, whereas

others may have followed 4-year university graduate

programmes after 12 years of formal schooling.

Third, there is no indication from the data de®nitions

whether the ®gures refer to whole time equivalent staff.

Buchan et al. (1997) have discussed the unreliability

for international comparative purposes of the Organiza-

tion for Economic Co-operation and Development (OECD)

data on nursing staff, and it is notable that The World

Bank includes this source, at least in its de®nition of

physicians.

CONCLUSION

The potential inaccuracies of the United Nations' global

data sets currently available raise questions as to how

seriously the conclusions in this paper can be taken. We

would argue that they must be seen as a necessary ®rst

step in trying to understand the forces which determine

the distribution of health labour. More reliable, disaggre-

gated data sets are undoubtedly required in order to

develop these ideas further and we hope that this paper

will be a stimulus to greater discussion of these structural

factors. Current exhortations to countries to manipulate

the structure of their health labour forces as a feature of

health services reform do not appear to take into account

the effects of these larger structural forces. Yet, the

strength of the correlation between the number of physi-

cians and nurses per 1000 population and gross national

product per capita identi®ed in this analysis, shows that

the in¯uence of extraneous variables may make

recommendations for governments to modify the structure

of their health labour forces much more dif®cult to

achieve than might at ®rst appear. The World Bank's

(1993) recommendations for governments to in¯uence the

distribution of physicians and nurses according to

recommended global parameters may not therefore be

entirely feasible. On the other hand, the recommendation

that the activities of available health personnel should be

directed at countries' most pressing health problems,

especially with the poor, is logical and according to the

limited research evidence, cost-effective.

The publication by The World Bank of the largest data

sets yet available on the distribution of physicians and

nurses allows both researchers and individual countries to

begin to question the assumptions that have been made

historically about the distribution of health personnel, and

to test the accuracy of ®ndings such as those presented

here against local evidence and interpretations.



Acknowledgements

Jane Robinson wishes to acknowledge the support of the

United States of America Fulbright Commission in the

award of a Senior Research Scholarship, and of the Health,

Nutrition and Population Sector of The World Bank, in

facilitating this work. She also thanks Dean Rita Carty and

the Faculty of Nursing, George Mason University, Fairfax,

Virginia, for their collegiate welcome during her stay in

the USA.

References

Bland J.M. & Altman D.G. (1996) Transforming data. British

Medical Journal 312, 770.

Buchan J., Seccombe I. & Thomas S. (1997) Overseas mobility of

UK based nurses. International Journal of Nursing Studies

34(1), 54±62.

Cochrane A.L., St Leger A.S. & Moore F. (1978) Health service

ìnput' and mortality òutput' in developed countries. Journal of

Epidemiology and Community Health 32, 200±205.

Cooper R.A. (1995) Perspectives on the physician workforce to the

year 2020. Journal of American Medical Association 274(19),

1534±1543.

Gray A. (1993) De®ning health needs and the need for nursing

care. In Nursing: Art and Science (Kitson A. ed.), Chapman and

Hall, London, pp. 1993.

Hertz E., James R.H. & Landon J. (1994) Social and environmental

factors and life expectancy, infant mortality, and maternal

mortality rates: results of a cross national comparison. Social

Science and Medicine 39(1), 105±114.

International Labour Organisation (1997) International Labour

Conventions and Recommendations 1919±1991, Volume II

(1963±1991). Convention no. 149: Nursing Personnel Con-

vention, Articles 1±8. International Labour Organisation,

Geneva.

Journal of the Royal Society of Medicine (1995) Policy issues and

priorities in health care reform: relevant lessons from the USA,

UK and Canada. Journal of the Royal Society of Medicine

(Suppl. 26) 88.

Organization for Economic Cooperation and Development (1991)

Health Data 1991. OECD, Paris.

Robinson J. (1997) Sustainable Development: Implications for

Nursing and Midwifery. Nursing and Midwifery Discussion

Paper no. 1. WHO/HDP/NUR-MID/97.1 Health Systems Devel-

opment Programme. World Health Organization, Geneva.

The World Bank (1993) World Development Report 1993: Invest-

ing in Health. Oxford University Press, New York.

The World Bank. (1997) Sector Strategy Paper: Health, Nutrition

and Population. Health, Nutrition and Population Family,

Human Development Network Washington, The World Bank,

June 30.

United Nations Children's Fund (1997) The State of the World's

Children 1997. Oxford University Press, New York.

United Nations Development Programme (1997) Human Devel-

opment Report 1997. Oxford University Press, New York.

Whitehead G. (1970) Economics made Simple. W.H. Allen,

London.



Documents

The global distribution of physicians and nurses