Factors affecting health worker density

Factors affecting health worker density: evidence from a quantitative crosscountry analysis
Rashid U Zaman*, Isla Gemmell†, Tomas Lievens*
22 December 2014
Abstract
In 2006 the World Health Organization identified 57 countries with critical shortage of health
workforce. A number of cross-country studies have explored the effect of the health workforce density
on countries’ health outcomes. However, little is known about the factors driving health workforce
density. The objective of this study was to identify the factors affecting the density of health workforce,
which would provide broader understanding of the underlying causes of this crisis and help formulate
appropriate policies in order to mitigate the challenge. This study analysed data from 183 UN member
countries to assess the association between the various demographic, economic and political factors
and the health workforce density. Out of 183 countries, 66 (36%) had a heath workforce density below
the WHO recommended threshold of 2.3 per 1,000 people. The adult literacy rate (p-value<0.01), total
health expenditure (p-value<0.01) and social stability (p-value=0.04) are statistically significant. Total
health expenditure had the greatest (33%) effect on the density of health workforce, followed by
literacy rates (25%) and social stability (11%). This cross-country study provides a snapshot of the
potential factors affecting health workforce density. Two of the three significant factors (adult literacy
rate and social stability) are not directly related to countries’ health system, which indicates that a
holistic and integrated approach is required in order to alleviate the health workforce crisis. Further
studies triangulating various quantitative and qualitative data would extend the understanding of the
topic.
*
Oxford Policy Management, 6 St Aldates Courtyard, 38 St Aldates, Oxford, OX1 1BN, UK
†
University of Manchester, Oxford Road, Manchester, M13 9PT, UK
Corresponding author: [email protected]
1
Electronic copy available at: http://ssrn.com/abstract=2541690
Introduction
The World Health Organization has identified the health workforce as one of the six discrete ‘building
blocks’ of health systems and selected health workers as the theme of its flagship annual publication
in 2006 – the World Health Report.1 The Report claims a global shortage of 2.4 million doctors, nurses
and midwives and identified 57 countries as ‘crisis countries’ for having ‘critical shortage’ of health
workforce below the threshold of 2.3 per 1000 population. It was estimated that countries which fall
below this threshold fail to attain 80% coverage in skill birth attendance and childhood vaccination.
The report created strong momentum in addressing the shortfall of global health workforce,
particularly in the countries with severe deficit. However, despite the global interest and broad
consensus on the need to address the health workforce crisis, there is little research to guide
policymakers on the strategic options for policy intervention.2 As a result, shortage and geographical
imbalance of health workforce remained as a widespread global problem 3
The association between the density of health workforce and health outcomes is well established.
Recent studies show statistically significant associations between the density of health workforce and
maternal, infant and under-5 mortality rates, and vaccination coverage.4-6 Density of health workforce
was also found to be associated with skilled birth attendance and measles immunization, but not with
antenatal care coverage, caesarean section, diagnosis of tuberculosis and acute respiratory
infection.7 Another cross-country study shows that the density of nurse workforce has significant
association with the prevalence of HIV and AIDS.8 A study analysing data from 125 countries found
significant association between health workforce density and Disability Adjusted Life Years (DALY).9
Statistical modelling with global health workforce data predicted that the infant mortality could drop by
15% by increasing 1 physician per 1000 population.10
A handful of studies have explored the factors affecting the density of health workforce at country or
regional level, but no ecological studies were found that explored the factors using global data. A
recent review has identified five categories of factors affecting the density of health workforce,
namely, individual factors (e.g. social background, ethnicity, age, gender, education, values, beliefs),
organizational environment (e.g. management style, incentives and career structures, salary scales,
2
Electronic copy available at: http://ssrn.com/abstract=2541690
recruitment, posting and retention practices), health care and education system determinants (e.g.
education and training processes, health care system, HRH policy formulation process), institutional
environment (e.g. structure, organization and role of national institutions such as the civil service and
ministries) and sociocultural environment (e.g. economic, political, social and historical parameters).3
As an example of urbanization as potential a demographic factor, a study in Nicaragua showed about
half of the total health care workers lived in the capital Managua, serving 20% of the total population
of the country.11 In the United States physicians’ density was strongly correlated with per capita gross
domestic product (GDP) across states which indicates the importance of the economic factor.12 An
ethnographic study in a conflict prone area in northern Pakistan reported impeded clinic access,
reduced number of physicians and subsequent increase in maternal morbidity and mortality – an
example of a social unrest as a political factor.13
While there are sufficient evidences of the effect of health workforce density on health outcomes, very
little is known about the factors affecting the density of the health workforce. The objective of this
study was to identify the factors affecting the density of health workforce, which would provide
broader understanding of the underlying causes of this crisis and help formulate appropriate policies
in order to mitigate the challenge.
Methods
Outcome and predictor variables
The outcome (dependent) variable for this study was the density of health workforce defined as the
number of physicians, nurses and midwives per 1000 population. Although the term health workforce
should include all people who are involved in health care including community health workers and
auxiliary staff, methodological and data limitations do not allow to include all groups of health
workers.1 Therefore we adopted a restrictive definition by only including physicians, nurses and
midwives – an approach widely used in other relevant studies and policy papers.4-9 Data on the
density of health workforce were available individually for ‘physicians’ and ‘nurses and midwives’ and
they were summed up to obtain a measure of the size of the ‘health workforce’ .
3
Table 1: List and description of outcome and predictor variables
Variables
Description
Unit
Source
Health workforce density
Number of registered physicians,
Per 1000
Global Health
nurses and midwives per 1000 people
population
Observatory
Number of people living in per square
Per square km of
World Population
kilometre of land area
land area
Prospects
The percentage of population aged ≥15
Percentage
Global Health
Population density
Literacy rate
years who can both read and write
Urbanization
The percentage of population living in
Observatory
Percentage
areas classified as urban area by the
Global Health
Observatory
concerned country
Gross national income
Per capita gross national income in
PPP Int. $ per
Global Health
purchasing power parity international
capita
Observatory
Per capita total expenditure on health
PPP Int. $ per
Global Health
in purchasing power parity international
capita
Observatory
Percentage
Global Health
dollars
Total health expenditure
dollars
Government health
Government expenditure on health as
expenditure
a percentage of total government
Observatory
expenditure
Social stability
Index measuring the level of threat of
10 point scale
social unrest
Democracy
Index measuring the state of
Intelligence Unit
100 point scale
democracy
Migration
Economist
Economist
Intelligence Unit
Net migration rate per 1000 population
Per 1000
World Population
per year
population/ year
Prospects
We carried out a comprehensive literature review to identify the potential predictor (independent)
variables. Variables were included if they were independent, quantifiable and likely to be available for
the majority of the countries in a credible and open-access data depository. The predictor variables
were grouped into three categories, namely demographic factors, economic factors and political
4
factors. The demographic factors that we explored were population density, literacy rate and
urbanization. Economic variables are Gross National Income, Total Health Expenditure and
Government Health Expenditure. Political variables are social stability, democracy and migration. All
of the predictor variables were continuous; however, the units and scales were different (Table 1).
For all except social stability, the direction was positive; i.e. higher scores indicated betterment and
thus a positive correlation was expected to be observed between the health workforce density and the
predictor variables. The original variable for social stability was measured by level of social unrest in
each country on a 10 point scale, where 0 in indicates no unrest (e.g. Norway=1.2) and 10 indicates
total unrest (e.g. Zimbabwe=8.8). To maintain the consistency in correlation across predictor variables
and to avoid confusion with the meaning of “social stability”, the variable was modified. This was done
by deducting each observation from the total score of 10. In this way the countries that had more
social stability (i.e. less social unrest) had higher score (e.g. Norway=8.8) and countries at the
opposite end had lower score (e.g. Zimbabwe=1.2).
Data management
We collected latest available data from the following online data depositories: the WHO Global Health
Observatory, the World Population Prospects 2010 edition, published by United Nations and the
databank of Economics Intelligence Unit of the Economist.14-17 Data were available in Microsoft Excel
spreadsheet at the WHO Global Health Observatory and UN World Population Prospects. We
obtained individual files from these websites and merged them to create a master dataset. We
compared country names using the “match” command of MS Excel to avoid possible mismatched
merging. Data from the Economic Intelligence Unit were available in the form of a published report
and were manually entered into the dataset. We used the “text-to-speech” command of MS Excel to
verify the data that were entered. The initial dataset contained observations from all 193 United
Nations member countries. We checked the data for blanks, ranges, outliers and internal consistency.
We verified suspicious data by looking at other sources, mostly the government website and the WHO
Global Health Atlas. We verified the outliers of the health workforce density data by manually
calculating density data from the absolute number of health workers and the population statistics. We
5
also did sensitivity analyses by adding and removing the outliers in simple linear regression to assess
the impact of the outliers on the statistical association.
Out of 193 countries 10 were excluded because of insufficient data. Data on the outcome variable
(density of health workforce) were insufficient for three countries (Liechtenstein, Montenegro and
South Sudan). While seven countries (Democratic People's Republic of Korea, Monaco, Nauru,
Palau, San Marino, Somalia and Tuvalu) lacked data on more than 5 predictor variables
Data analysis
The data were analysed using the statistical analysis packages StatsDirect and SPSS. T-tests,
simple linear regression and multiple linear regression methods were used to quantify the factors
affecting density of health workforce. Tests were considered statistically significant when p-value was
smaller than 0.05.
t-test: We grouped the countries into two categories (“case” and “control”) depending upon their
position with the recognized threshold of density of health workforce.1 Countries with a health
workforce density smaller than 2.3 per 1000 people were case countries and countries with a higher
density were control counties. We carried out t-tests to test whether the difference in the mean value
of the predictor variables was significantly different between case and control countries and reported
the corresponding t-statistics. A significantly different mean value for case and control groups
confirms that the predictor variable is driving the outcome variable.
Simple linear regression: To verify the result of the t-tests, in second stage of the analyses, we
performed simple linear regression to see whether the association between the outcome and
individual predictor variable was statistically significant. We reported p-values, regression coefficients
(B) and correlation coefficients (r2) to express the significance and magnitude of association.
Multiple regression: Finally, we carried out multiple regression to see the association between the
outcome variable and nine predictor variables in a single model. To express the result of the multiple
6
regression, we reported p-value, r2 and F statistics. Since the predictor variables had different units of
measurements, we also generated β-coefficients for each predictor variable which provided a
standardized value for comparison.
Results
Among 183 countries, 66 (36%) countries had a health workforce density less than the threshold of
2.3 per 1000 population and were grouped as ‘case’ countries. The remaining 117 (64%) countries
had more than or equal to the threshold density and were grouped as ‘control’ countries. In 2006, the
World Health Report (WHR) identified 57 countries as ‘crisis countries’ using the same threshold.1
There were some mismatches between the WHR 2006 and this study in classifying the countries as
‘crisis’ (case) or ‘non-crisis’ (control) countries. Two countries (Indonesia and Somalia) were included
as crisis countries in WHR 2006, but not classified as ‘case’ countries in this study. Indonesia was
included as a control country since according to recent data its health workforce density was above
the threshold. Somalia was excluded from the study because of lack of data on ≥5 predictor variables.
However, 11 countries (Cape Verde, Colombia, Costa Rica, Samoa, Solomon Islands, Sudan,
Suriname, Thailand, Timor-Leste, Vanuatu and Viet Nam) were not identified as crisis country in the
World Health Report 2006, but were classified as ‘case’ countries in this report since according to the
latest data their health workforce density was below the threshold of 2.3 per 1000 population.
Table 2: Distribution of case and control counties by continents
Case countries
Control countries
n (%)
n (%)
Africa
40 (77)
12 (23)
Asia
14 (30)
32 (70)
0 (0)
39 (100)
North America
5 (22)
18 (78)
South America
3 (25)
9 (75)
Oceania
4 (36)
7 (64)
66 (36)
117 (64)
Continents
Europe
Total
7
Figure 1: Simple linear regression between outcome and predictor variables
There were considerable differences in the distribution of case and control countries across the
geographical regions. Out of 66 countries with less than the threshold density 40 (61%) were from
8
Africa. In contrast, all the countries in Europe had a health workforce density above the threshold
(Table 2).
Demographic factors
Population density: Data on population density were available for 182 (99%) countries and the
calculated mean (SD) population density was 176.4 (583.4) per square kilometre of land area. The
mean (SD) population density in the case countries was 104.53 (153.13) persons per sq. km in
comparison to 217.29 (719.56) persons per sq. km in control countries. The difference was not
statistically significant (p-value=0.11, t=1.62). Simple linear regression between population density
and health workforce density also revealed non-significant (p-value=0.64, r2=0.00, B=0.00)
association (Figure 1).
Literacy rates: Out of 183 countries 178 (97%) countries had data on adult literacy rate. The mean
(SD) literacy rate was 83.73 (19.04) ranging from 19.08% to 99.8%. There is a significant (pvalue<0.01, t=11.18) association between the literacy rate and density of health workforce. Mean
(SD) literacy rate of case countries was 65.67% (19.79), whereas in control countries it was 94.12%
(7.15). The linear regression for the association with literacy also showed strong (p-value<0.01, r2
=0.39, B=0.16) association (Figure 1).
Urbanization: Data on urbanization was available for all 183 countries. The mean (SD) proportion of
urban population was 55% (22.78). The case countries had a mean (SD) urban population of 38.55%
(17.48) compared to 64.47% (19.96) in control countries. The difference was statistically significant
(p-value<0.01, t=8.81). The linear regression also exhibited significant association (p-value<0.01,
r2=0.31, B=0.12) between urbanization and density of health workforce (Figure 1).
Economic factors
Gross national income (GNI): Data on GNI was available for xx (yy%) of countries. Mean (SD) per
capita GNI in case countries was $2696.56 (3269.69) compared to $17850.35 (13901.66) in control
9
countries and the difference was statistically significant (p-value<0.01, t=11.02). The regression
between the GNI and density of health workforce also revealed significant (p-value<0.01, r2=0.49,
B=0.00) association (Figure 1).
Total expenditure on health: All countries had data on per capita total expenditure on health with
the mean (SD) value of $1020.98 (1356.53). There was significant (p-value<0.01, t=9.19) difference
between the case and control countries in regard to total expenditure on health. The mean (SD)
expenditure for case and control were $185.07 (289.12) and $1492.52 (1489.47) respectively.
Significant association (p-value<0.01, r2=0.51, B=0.003) was also revealed in regression analysis
(Figure 1).
Government expenditure on health: Data were available for all 183 countries for the ratio indicator
government expenditure on health as a percentage of total government expenditure. The mean (SD)
ratio was 11.13% (4.5). Mean (SD) expenditure ratio for case countries was 10.18% (5.19) which was
significantly (p-value=0.04, t=2.04) different than the control countries with a ratio of 11.68% (3.89).
The regression between the predictor and outcome variables also had a significant (p-value<0.01,
r2=0.07, B=0.29) association (Figure 1).
Political factors
Social stability: Out of 183 countries 160 (87%) had data on the social stability index. The mean
(SD) social stability was 4.1 (1.4). Case countries had a mean (SD) of 3.3 (1.1) scale-point compared
to 4.6 (1.4) in controls. The difference was statistically significant (p-value<0.01, t=6.34). Regression
was also significant (p-value<0.01, r2=0.36, B=2.13) (Figure 1).
Democracy: Democracy index data were available for 163 (89%) countries with the mean (SD) scalepoint of 5.5 (2.17). Mean (SD) scale-point for case countries was 4.4 (1.71) compared to 6.19 (2.16)
in control countries. The difference was statistically significant (p-value<0.01, t=5.86). Regression was
also significantly different (p-value<0.01, r2=0.26, B=1.18) between two groups (Figure 1).
10
Migration: Data on migration rate was available for 175 (95%) countries. The mean (SD) was 1.74
(15.57) per 1000 population ranging between -17.33 and 132.89. The mean (SD) migration rate was 1.3 (4.59) per 1000 population in the case countries and 3.58 (19.2) per 1000 population in control
countries. The difference was statistically significant (p-value=0.01, t=2.54). Regression analysis also
found significant (p-value=0.02, r2=0.02, B=0.05) association (Figure 1).
Aggregated effect
Multiple linear regression was carried out to see the aggregated effect of the factors on the density of
health workforce (Table 3). The multiple regression model explained 66% of the variability in the data
(p-value<0.01, F=31.05, r2=0.66). Among the nine predictor variables literacy rate and total health
expenditure were highly significant (p-value<0.01) and social stability was significant (p-value=0.04).
Table 3: Multiple regression of the outcome and predictor variables
Factors
Unstandardized
Standardised
coefficient (B)
coefficient (β)
-0.0005
-0.059
0.27
Literacy rate*
0.079
0.316
<0.01
Urbanization
0.023
0.106
0.14
Gross national income
2.284
0.063
0.7
Total health expenditure*
0.001
0.418
<0.01
-0.094
-0.085
0.17
Social stability*
0.507
0.142
0.04
Democracy
0.029
0.013
0.86
-0.031
-0.078
0.24
Population density
Government health expenditure
Migration
P-value
* Significant
We used standardised β-coefficients to measure the contribution of the different factors affecting the
health workforce. Among all nine factors, total health expenditure accounted for the greatest effect 33% - on the density of health workforce. Literacy rate (25%) and social stability (11%) also had
considerable effect on the density of health workforce (Figure 2).
11
Figure 2: Proportion of factors effecting density of health workforce
Discussion
Our analysis has identified factors affecting the density of health workforce for 183 UN member states
through an ecological study approach. The findings give quantitative evidence on the factors which
were previously identified as potential contributors individually or at a country or regional level. Over
the past few years, there has been a growing interest in health workforce density. Policymakers and
health systems researchers are increasingly paying attention to this topic. A handful of studies have
explored the impact of health workforce density over countries’ health outcomes using global data,
however, we have not found any studies that analysed the global data on health workforce to quantify
the potential factors affecting the density. This analysis therefore added important novel findings to
the existing knowledge on the topic.
All the associations between the outcome variable and the predictor variables were statistically
significant in the individual t-test and simple linear regression, except the association with the
population density. While the t-test and simple linear regression analyses provided important
information on the degree and direction of association between the outcome variable and the
predictor variables, these analyses did not take into account the potential confounding between the
predictor variables. We therefore carried out multiple regression as the final model that allowed
adjusting the possible associations between the predictor variables. In the multiple regression model,
12
three predictor variables stand out as most important explanatory factors. They were literacy rate,
total health expenditure and social stability. The association between the literacy rate and the density
of health workforce was so strong that there were no notable examples of countries where the literacy
rate was low but the health workforce density was high or vice-versa. The finding is consistent with
the cross-district analysis of physician density and literacy rate in India and a review article which
investigated the different aspects of relation between health and literacy.18-19 This suggests that
literacy rate is a binding constraint for health worker density. Per capita total health expenditure also
had highly significant association with the outcome variable. This includes both government and
private expenditure on health. Since two other economic factors, the gross national income and the
government health expenditure were not statistically significant in the final multiple regression model,
it can be stated that it is the total expenditure that matters, not the source (public and private) of its
funding within the country. This finding re-illustrates the fact that health sector financing is a complex
topic and there are no straightforward solutions to how and to what extend the governments should
intervene20. The third most significant factor was social stability and the data has indicated that the
countries that are more stable had higher density of health workforce. If supply would follow demand
then the density of health workforce should be higher in unstable societies. A study on the impact of
conflict on children’s health and disability has shown that children in conflict prone areas are more
vulnerable to a wide range of morbidities including injury, physical disability, infectious disease,
malnutrition, sexual abuse, psychological trauma etc.21 However, quite often the health workers are
also targeted victims in these settings which may also influence the decision for the health care
workers to migrate to more stable countries.13
This study has some important limitations. Most importantly, the variables impacting on health worker
density, and their inter-relations, are not fully understood, and not all relevant variables can be
measured (the job-status, for example, which is important for decisions relations to entry into the
sector). The mal-distribution of health workers is associated with a combination of individual,
community and government decisions and is subject to personal, professional, organizational,
economic and cultural factors.3 For example, a study has reported that the incentive to become a
nurse is low in Bangladesh since it is not considered as a dignified profession. 22 Secondly, this study
only applied quantitative analysis and there are certain aspects of the health workforce which are
13
difficult to quantify. Moreover, ‘density’ as a quantitative estimate does not necessarily capture the
whole notion around its impact. For example, researchers have argued that in USA it is more
important to work on the ‘quality of the physicians’ than the ‘number of physicians’.23
Prescribing a straightforward policy solution to address the health workforce density crisis is too
ambitious; because of the complex economic, political, societal, individual dimensions of the crisis. A
number of recent policy papers have highlighted some key policy areas that require attention.24-27 This
study has provided some quantitative evidence to them.
Two important findings stand out. Firstly, this study suggests that countries with a shortage in health
workforce density should increase total health expenditure. Governments can intervene in two ways:
either by increasing government contribution to health expenditure, or by creating an enabling
environment that allows the number of health workers to increase with available funding, from private
or public sources. This implies ensuring enough capacity to train health workers, and putting effective
accreditation, quality control and licensing arrangements in place. Secondly, it is noteworthy two of
the three significant factors (literacy rate and social stability) are not directly related to countries’
health systems. This indicates that a holistic and multi-sectorial approach is required in order to
alleviate the health workforce crisis. Donors and international agencies can play an important role in
coordinating the multiple sectors.
Author statements
Acknowledgements
We are grateful to Professor Aneez Ismail, Dr Katie Reed and Dr Roger Harrison of the University of
Manchester and Mr Simon Hunt of Oxford Policy Management for their continuous support and
encouragement to undertake this research work.
Ethical approval
Not sought as no primary data has been used
14
Funding
The first author received the Equity and Merit Scholarship of the University of Manchester which
funded the post graduate programme under which this study was carried out.
Competing interests
None
References
1
WHO. World Health Report 2006 – Working together for health. Geneva: World Health
Organization, 2006.
2
Ranson MK, Chopra M, Atkins S, Dal Poz MR, Bennett S. Priorities for research into human
resources for health in low- and middle-income countries. Bull World Health Organ 2010;
88:435–43.
3
Dussault G, Franceschini MC. Not enough there, too many here: understanding geographical
imbalances in the distribution of the health workforce. Hum Resour Health 2006; 4:16.
4
Anand S, Bärnighausen T. Human resources and health outcomes: cross-country
econometric study. Lancet 2004; 364:1603-09.
5
Chen L, Evans T, Anand S, Boufford JI, Brown H, Chowdhury M et al. Human resources for
health: overcoming the crisis. Lancet 2004; 364:1984-90.
6
Anand S, Bärnighausen T. Health workers and vaccination coverage in developing countries:
an econometric analysis. Lancet 2007; 369:1277-85.
7
Kruk ME, Prescott MR, de Pinho H, Galea S. Are doctors and nurses associated with
coverage of essential health services in developing countries? A cross-sectional study. Hum
Resour Health 2007; 7:27.
8
Madigan EA, Curet OL, Zrinyi M. Workforce analysis using data mining and linear regression
to understand HIV/AIDS prevalence patterns. Hum Resour Health 2008; 6:2.
9
Castillo-Laborde C. Human resources for health and burden of disease: an econometric
approach. Hum Resour Health 2011; 9:4.
15
10
Farahani M, Subramanian SV, Canning D. The effect of changes in health sector resources
on infant mortality in the short-run and the long-run: a longitudinal econometric analysis. Soc
Sci Med 2009; 68:1918-25.
11
Nigenda G, Machado MH. From state to market: the Nicaraguan labour market for health
personnel, Health Policy Plan 2002; 15:312-8.
12
Freed GL, Nahra TA, Wheeler JR. Relation of per capita income and gross domestic product
to the supply and distribution of pediatricians in the United States. J Pediatr 2004; 44:723-8.
13
Varley, E. Targeted doctors, missing patients: obstetric health services and sectarian conflict
in northern Pakistan. Soc Sci Med 2010; 70:61-70.
14
WHO. Global Health Observatory Data Repository, 2011. Geneva: World Health
Organization. http://apps.who.int/ghodata (accessed December 26, 2011).
15
UN DESA. World Population Prospects, the 2010 Revision. New York: United Nations
Department of Economic and Social Affairs. http://esa.un.org/wpp (accessed December 26,
2011.
16
EIU. Social unrest, ViewsWire: Economist Intelligence Unit, 2009. London: The Economist.
http://viewswire.eiu.com/site_info.asp?info_name=social_unrest_table&page=noads&rf=0
(accessed February 01, 2012).
17
EIU. Democracy index 2011: Democracy under stress, ViewsWire: Economist Intelligence
Unit, London: The Economist.
https://www.eiu.com/public/topical_report.aspx?campaignid=DemocracyIndex2011 (accessed
February 01, 2012.
18
De Costa A, Al-Muniri A, Diwan VK, Eriksson B. Where are healthcare providers? Exploring
relationships between context and human resources for health Madhya Pradesh province,
India. Health Policy 2009; 93:41-7.
19
Rudd RE, Moeykens BA, Colton TC. Health and Literacy: A Review of Medical and Public
Health Literature, Chapter 5, Annual Review of Adult Learning and Literacy, 1999. New York:
Jossey-Bass.
20
Musgrove P. Public and private roles in health: Theory and Financing Patterns, HNP
Discussion Paper, 1996. Washington: The World Bank.
16
21
Tamashiro T. Impact of conflict on children’s health and disability, Commissioned paper, EFA
Global Monitoring Report 2011. Paris: United Nations Educational, Scientific and Cultural
Organization.
22
Hadley MB, Blum LS, Mujaddid S, Parveen S, Nuremowla S, Haque ME et al. Why
Bangladeshi nurses avoid 'nursing': social and structural factors on hospital wards in
Bangladesh. Soc Sci Med 2007; 64:1166-77.
23
Green LA, Phillips RL. The Family Physician Workforce: Quality, Not Quantity, Am Fam
Physician 2005; 71:2248-53.
24
Dussault G, Dubois CA. Human resources for health policies: a critical component in health
policies, Hum Resour Health 2003: 1:1.
25
GHWA. Reviewing Progress, Renewing Commitment, Progress report on the Kampala
Declaration and Agenda for Global Action, 2011. Geneva: Global Health Workforce Alliance,
World Health Organization.
26
JLI. Human resources for health: Overcoming the crisis, 2004. Cambridge, MA: The Joint
Learning Initiative, Harvard University.
27
Van den Broek A, Gedik G, Poz MD, Dieleman D. Policies and practices of countries that are
experiencing a crisis in human resources for health: tracking survey, Human Resources for
Health Observer, 2011. Geneva: World Health Organization.
17