HYPERTENSION TREATMENT RATES AND HEALTH CARE WORKER DENSITY: AN ANALYSIS OF WORLDWIDE DATA

Abstract

Elevated blood pressure is the leading cause of death worldwide; however, treatment and control rates for hypertension are low. Here, we analyze the relationship between physician and nurse density and hypertension treatment rates worldwide.

Data on hypertension treatment rates were collected from the STEPwise approach to Surveillance country reports, individual studies resulting from a PubMed search for articles published between 1990 and 2010, and manual search of the reference lists of extracted studies. Data on health care worker density were obtained from the Global Atlas of the Health Workforce. We controlled for a variety of variables related to population characteristics and access to health care, data obtained from the World Bank, World Development Indicators, United Nations, and World Health Organization. We used clustering of standard errors at the country level.

Full data were available for 154 hypertension treatment rate values representing 68 countries between 1990–2010. Hypertension treatment rate ranged from 3.4% to 82.5%, with higher treatment rates associated with higher income classification. Physician and nurse/midwife generally increased with income classification. Total health care worker density was significantly associated with hypertension treatment rate in the unadjusted model (p < 0.001); however, only nurse density remained significant in the fully adjusted model (p = 0.050).

These analyses suggest that nurse density, not physician density, explains most of the relationship with hypertension treatment rate, and remains significant even after adjusting for other independent variables. These results have important implications for health policy, health system design, and program implementation.

Introduction

The burden of cardiovascular disease (CVD) continues to rise globally and its increase is not isolated to countries of particular income levels, geographical regions, or social conditions.¹ Elevated blood pressure, a major risk factor for CVD,^2,3 is the leading cause of death worldwide.⁴ Treatment for hypertension is well identified and effective.⁵ The challenge, however, is that treatment and control rates are low worldwide.^6,7 Institutional and economic barriers to getting treatment to the individuals who need it are complex, and while some are shared across countries, others vary with institutional aspects of health care delivery and health insurance status across countries.^8,9

One factor that may affect hypertension treatment rates is the density of the health care workforce—specifically physicians and nurses—within a country. Health care worker density has been shown to be favorably associated with vaccination rates¹⁰ and overall disability-adjusted life-years.¹¹ In addition, the relationship between health care worker density and maternal, infant, and child mortality has been studied, with no clear consensus on the relationship between health care worker density and health outcomes. More recent studies indicate that health care worker density is associated with improved health outcomes,^12–14 while other older studies have demonstrated either neutral^15,16 or negative associations.¹⁷

While there has been more recent interest in investigating the relationship between health care worker density and non-communicable disease (NCD) (including CVD and hypertension) outcomes, there is no consensus regarding the direction or strength of the relationship. An analysis of the relationship between health care worker density (including physicians, nurses, and others) and CVD outcomes reported that a univariate analysis revealed that health care worker density is associated with better CVD outcomes, although a multivariable analysis yielded the opposite relationship.¹⁸ Another study reported no significant relationship between physician density and medical guideline adherence (including for CVD) in Germany.¹⁹

Current estimates of global health care worker shortages have largely not taken into account the human resource requirements for managing NCDs. In particular, the relationship between health care worker density and hypertension treatment rates is unknown. Recognizing this gap in the literature, and recognizing the growing importance of CVD, in this study we analyze the relationship between physician and nurse/midwife density and hypertension treatment rates, across countries of all World Bank income classifications.²⁰

Methods

All data were extracted from publicly available databases and publicly accessible publications. Analytic methods will be made available to researchers upon formal request to the authors.

Choice of Variables

Our dependent variable was hypertension treatment rate, defined as the percentage of hypertensive individuals on treatment. Hypertension was defined as systolic blood pressure (SBP) ≥ 140 mmHg or diastolic blood pressure (DBP) ≥ 90 mmHg or self-reported current use of BP-lowering medication. Hypertension treatment was defined as self-reported current use of antihypertensive medication.

We used three measures of health care worker density. First, an aggregate measure was derived by summing physician and nurse/midwife density per 1000 population. We also disaggregated physician and nurse densities. We limited our analysis to these cadres of health care workers since most hypertension management, treatment decisions, and prescription issuances are completed by physicians or nurses.^21,22

To account for differences in population characteristics and access to health care across countries, we adjusted for the following country-level variables: gross domestic product (GDP) per capita, purchasing power parity (PPP; current international $), private health expenditure (% of GDP), land area, population, hospital beds, rural access to an improved water source (% of rural population with access), and disability-adjusted life year (DALY) estimates for 2004. PPP-adjusted GDP per capita was used as a measure of the general level of resources available in the country. Access to an improved water source (% of rural population with access) was also included as a measure of access to resources and poverty. Private health expenditure as percentage of GDP was included as a measure of private health expense burden. Land area likely impacts the geographic density of health care workers as well as the distance from patients to health care worker; therefore, it was included as it potentially influences treatment rates. Hospital beds per 1000 population was included as an indicator of health system resource availability. DALYs for infectious and parasitic diseases (per 100,000 population, age-standardized, 2004 estimates) were used to control for the potential impact of morbidity from communicable diseases.

Data sources

Data on hypertension treatment rates were collected from the World Health Organization (WHO) STEPwise approach to Surveillance (STEPS) country reports,²³ individual studies resulting from a PubMed search for articles published between 1990 and 2010 using the key words “prevalence AND awareness AND treatment AND control AND (hypertension OR high blood pressure),” and a manual search of the reference lists from the extracted studies (full list of references available in the Appendix). A study was included if it contained information on country-specific hypertension treatment rate. The WHO STEPS data were obtained from surveys that follow a standardized framework for collecting, analyzing, and reporting data on NCD risk factors in WHO member countries.²³

Data on health care worker density (physicians and nurses) were obtained from the WHO Global Atlas of the Health Workforce.²⁴ Data for the other independent variables were obtained from the following sources: the World Development Indicators for gross domestic product (GDP) per person in international dollars at purchasing power parity (PPP$), private health expenditure as a percentage of GDP, land area, hospital beds per 1000 population, and access to an improved water source (% of rural population);²⁵ the United Nations’ World Population Prospects: The 2010 Revision for population estimates;²⁶ and the WHO database for DALYs for infectious and parasitic diseases (per 100,000 population, age-standardized).²⁷

The dataset for our analysis was constructed by matching the country and year of our dependent and independent variables, within a +/− five-year window around the year of the hypertension treatment rate data. The DALYs data were obtained from the 2004 WHO estimates and thus did not conform to the +/− five-year criterion due to data availability limitation. World Bank income classification for each country was aligned with the year of hypertension treatment rate data for that country. In total, our dataset included 154 observations, representing 68 countries over the period 1990–2010.

Statistical Procedures

In order to compare data from the STEPS reports and individual studies collected from our PubMed search, we defined treatment of hypertension as self-reported current use of antihypertensive medication. To create our dependent variable, the rate of hypertension treatment, the number of people reporting hypertension treatment was divided by the total number of study participants with hypertension.

All regressions were estimated using a logistic-logarithmic functional form, similar to the analytic approach previously utilized to assess the relationship between health care worker density and vaccination coverage.¹⁰ The logistic form of the dependent variable reflects the boundedness of hypertension treatment rate between 0% and 100%, which in turn prevents predicted values from falling below 0% or exceeding 100%. The logarithmic form of the independent variables allows for the following interpretation of their estimated coefficients (β): a 1% increase in the independent variable corresponds to a β% change in the sum of the percentage increase in the level of the dependent variable and the percentage reduction in the shortfall from the upper bound of 100%.^10,12 One exception to this is the “access to an improved water source” variable, for which a logit transformation was performed, as the variable is a proportion and is bounded by 0 and 1. All regressions used clustering of standard errors at the country level to control for possible correlation of error terms within countries.

We report the results of three regression analyses with hypertension treatment rate as the dependent variable, while controlling for GDP per capita, PPP (current international $), health expenditure, private (% of GDP), land area, population, hospital beds, access to an improved water source, rural (% of rural population with access), and DALYs estimates for 2004. In the first analysis, hypertension treatment rate was regressed against aggregate combined physician and nurse density. In the second analysis, separate models were used to regress hypertension treatment rate against physician and nurse densities, respectively. In the third analysis, hypertension treatment rate was regressed against disaggregate physician and nurse densities in a single model. We formally tested for interaction by World Bank income classification category. We also conducted sensitivity analyses by performing separate regressions for the studies between 1990–1999 and 2000–2010, as well as separate regressions for studies conducted in each of the World Bank income classification categories. Stata version 11.0 was used to perform the analyses.

Results

Hypertension treatment data were obtained from 28 WHO STEPS reports and 41 individual studies. Health care worker density and treatment data were available from 174 country and treatment data combinations, spanning 84 countries and all World Bank income classification categories. However, full data, including all independent variables, were available for 154 of these, representing 68 countries, over the period 1990–2010 (Table 1).

Table 1.

Country-years and number of observations by country income classification. Numbers in parentheses represent the number of studies from that country in that particular year.

High Income (n = 65)
Country	# of Studies	Years of Treatment Data
Australia	1	1994
Belgium	1	1992
Canada	5	1992, 1995, 2002, 2009(2)
Czech Republic	1	2008
Denmark	3	1992, 1998, 2004
Finland	4	1992(3), 2005
France	2	1998, 2006
Germany	6	1992, 1995(3), 2001(2)
Greece	6	1997, 1998, 2001(2), 2002, 2004
Italy	2	1994(2)
Japan	1	1995
Korea, Rep.	2	2001(2)
Kuwait	1	2008
Netherlands	3	2003, 2004(2)
New Zealand	1	1994
Portugal	2	2003(2)
Saudi Arabia	1	2005
Spain	3	1990, 1996, 2004
Sweden	3	1994, 1996
Switzerland	1	1993
United Kingdom	8	1995, 1998(2), 2003(4), 2006,
United States of America	7	1990, 1994, 2000(2), 2003, 2004, 2005
United Arab Emirates	1	2009
Upper Middle Income (n = 26)
Country	# of Studies	Years of Treatment Data
Argentina	1	2009
Barbados	1	1996
Botswana	1	2007
Brazil	1	2009
Chile	3	2004(2)
China	1	2010
Colombia	1	2009
Czech Republic	3	2001(3)
Gabon	1	2003
Iran, Islamic Rep.	1	2009
Korea, Rep.	1	2000
Lebanon	1	2008
Malaysia	4	1996(2), 2004
Mexico	3	1993, 2000, 2002
Romania	1	2005
St. Lucia	1	1996
Turkey	1	2009
Lower Middle Income (n = 30)
Country	# of Studies	Years of Treatment Data
China	8	1999, 2001(4), 2002(2), 2009
Egypt, Arab Rep.	1	2005
Fiji	1	2002
Iran, Islamic Rep.	1	2005
Iraq	2	2003, 2006
Maldives	1	2004
Mongolia	1	2009
Russian Federation	2	2005(2)
Thailand	2	2003, 2004
Tunisia	1	2001
Turkey	4	1995, 1999, 2003(2)
Venezuela, RB	1	1996
India	3	2007(2), 2009
Pakistan	1	2009
Micronesia	1	2002
Low Income (n = 33)
Country	# of Studies	Years of Treatment Data
Bangladesh	1	2009
Benin	1	2007
Cambodia	1	2010
Cameroon	1	2004
China	3	1993, 1994, 1998
Egypt, Arab Rep.	1	1991
Eritrea	1	2004
Ghana	5	2001, 2002, 2004(3)
India	6	1999, 2000, 2002(2), 2004, 2005
Lao PDR	1	2008
Madagascar	1	2005
Malawi	1	2009
Mauritania	1	2007
Mongolia	1	2006
Mozambique	1	2005
Nepal	2	2003, 2005
Nigeria	1	2003
Sierra Leone	1	2009
Solomon Islands	1	2006
Vietnam	1	2008
Zambia	1	2008

Hypertension treatment rate ranged from 3.4% to 82.5% (Table 2). Mean hypertension treatment rate was 34.5% overall, with higher mean treatment rates observed as income classification increased from low- to high-income. Physician and nurse density varied widely, but generally increased as income classification increased.

Table 2.

Summary statistics of published hypertension treatment rate and health care worker density data (per 1000 population) by country income classification.

High-Income (n = 65)
Variable	Mean	Std. Dev.	Min	Max
Treatment Rate	38.9	18.3	11.1	82.5
Aggregate Physician and Nurse Density	11.46	3.01	3.50	18.41
Physician Density	3.03	0.92	1.37	4.91
Nurse Density	8.43	3.01	1.93	14.41
Upper-Middle-Income (n = 26)
Variable	Mean	Std. Dev.	Min	Max
Treatment Rate	35.2	20.2	10.7	79.5
Aggregate Physician and Nurse Density	4.69	3.29	1.72	12.41
Physician Density	1.86	1.36	0.29	6.42
Nurse Density	2.83	2.59	0.48	8.94
Lower-Middle-Income (n = 30)
Variable	Mean	Std. Dev.	Min	Max
Treatment Rate	30.9	16.2	3.4	69.5
Aggregate Physician and Nurse Density	3.56	2.80	1.16	12.79
Physician Density	1.36	0.98	0.30	4.26
Nurse Density	2.20	1.93	0.38	8.53
Low-Income (n = 33)
Variable	Mean	Std. Dev.	Min	Max
Treatment Rate	28.5	16.6	5.7	74.2
Aggregate Physician and Nurse Density	1.58	1.32	0.19	6.26
Physician Density	0.53	0.67	0.02	2.76
Nurse Density	1.04	0.73	0.17	3.50
Overall (n = 154)
Variable	Mean	Std. Dev.	Min	Max
Treatment Rate	34.5	18.2	3.4	82.5
Aggregate Physician and Nurse Density	6.66	5.03	0.19	18.41
Physician Density	1.97	1.39	0.02	6.42
Nurse Density	4.69	4.04	0.17	14.41

As total health care worker density (physician and nurse) increased, hypertension treatment rates generally tended to increase as well, although there was notable variation in the scatterplot (Supplemental Figure, Online Supplement). Total health care worker density was significantly associated with hypertension treatment rate in the unadjusted model, and nearly significant at the 0.05 level in the fully adjusted model (Table 3). In the separate regressions, both physician and nurse density were significantly associated with hypertension treatment rate in the unadjusted model; however, only nurse density remained significant in the fully adjusted model (Table 3, Figure). In the disaggregated model, only nurse density was significantly associated with hypertension treatment rate, in both the adjusted and unadjusted models. Taken together, these analyses suggest that nurse density, not physician density, explains most of the relationship between health care worker density and hypertension treatment rate. This relationship remains significant even after adjusting for all of the other independent variables listed above. Full multivariable regression results for all four models are presented in the Supplemental Table, Online Supplement.

Table 3.

Logistic-logarithmic regression results.

Model	Unadjusted (n = 154)		Adjusted (n = 154)
	β (CI)	p-value	β (CI)	p-value
Total density (aggregate physician and nurse density)	0.27 (0.13, 0.41)	< 0.001	0.33 (−0.10, 0.76)	0.133
Physician density (disaggregate)	0.03 (−0.15, 0.20)	0.740	−0.07 (−0.41, 0.27)	0.678
Nurse (disaggregate)	0.22 (0.04, 0.42)	0.020	0.30 (0.00, 0.60)	0.050
Physician density (separate regression)	0.18 (0.06, 0.30)	0.004	−0.03 (−0.36, 0.30)	0.849
Nurse Density (separate regression)	0.25 (0.12, 0.38)	< 0.001	0.29 (−0.01, 0.58)	0.055

Figure.

Relationship between nurse density and hypertension treatment rates worldwide, with countries of different income classification indicated. Solid curve is derived from the fully adjusted model.

Sensitivity analyses for year of publication suggested that studies conducted between 2000 and 2010, as compared to those conducted between 1990 and 1999, yielded results consistent with the overall results presented above, although it is difficult to arrive at a definitive conclusion due to small numbers. Similarly, low- and lower-middle-income countries appeared to have nurse density-hypertension treatment rate relationships that were consistent with the overall results, but the small numbers in the stratified analyses restrict any definitive conclusion (Supplemental Table, Online Supplement).

Discussion

In our worldwide econometric analysis of the relationship between hypertension treatment rates and health care worker density, we report that hypertension treatment rates vary widely, health care worker density rates also vary widely, and health care worker density was significantly associated with hypertension treatment. Notably, our approach of aggregating and disaggregating physicians and nurses allowed us to highlight that nurse density, not physician density, appears to explain most of the relationship between health care worker density and hypertension treatment rate. The relationship between nurse density and hypertension treatment rate remained statistically significant after adjusting for several potential confounder variables.

The relationship between health care worker density and NCDs has not been well studied, and examples from the literature are limited.²⁸ In fact, to the best of our knowledge, this is the first report examining the relationship between health care worker density and hypertension treatment rates using worldwide data. Given the health workforce shortage worldwide, this is an important first step in the ultimate goal of aligning disease burden with human resources for health.

One notable finding from our study is that nurse density, not physician density, was more strongly associated with hypertension treatment rate. This is striking because, in many parts of the world, only physicians and clinicians, not nurses, are authorized to prescribe anti-hypertensive medications.²⁹ The nurse’s role in hypertension care, in contrast, has generally been limited to enhancing self-management strategies by educating and counseling the patient about medication adherence and lifestyle modification.^30,31 It is possible, however, that even in countries or regions that do not allow nurses to treat hypertension directly, the availability of nurses to work with physicians may enable physicians to shift other responsibilities, which may allow them to focus more time and effort on hypertension management, thus contributing to higher hypertension treatment rates. It is also possible that the presence of other cadres of health worker, such as pharmacists, could be confounding the observed results. Unfortunately, data regarding pharmacist density were not as routinely available as physician or nurse density. Therefore, this potential hypothesis was not able to be tested.

The relationship between health care worker density and health outcomes has been characterized by conflicting results in previous literature. An inherent challenge is that, while health care worker density may have favorable impacts on health care delivery measures, health outcomes per se may be affected by many other factors beyond care delivery alone. For instance, control of hypertension is also impacted by drug supply,^29,32 medication adherence,³³ lifestyle factors,³⁴ and a variety of socio-economic factors.³⁵ Hence, in this study, we limited our analysis to hypertension treatment rates (care delivery measure) rather than hypertension control rates (health outcome measure).

Given the inclusion of NCDs in the Sustainable Development Goals,³⁶ health care worker requirements need to also take into account human resources for health needed to manage hypertension and other NCDs. Health care workers are a critical component of the health care delivery process that can promote population health, and the supply of health professionals worldwide has been highlighted as a significant problem.³⁷ In addition, other cadres of health worker, such as community health workers, can also positively impact outcomes for hypertension and other NCDs.^38,39 On balance, our results support the strategy of task redistribution to meet the human resource challenge of management of NCDs including hypertension. Task redistribution, in which specific tasks are “reorganized and dispersed” among health care workers with different duration of training and different qualifications, can allow for more efficient use of available human resources for health.^40,41 Non-physicians have been effective in hypertension management in LMICs.^42–46 Given the growing global need for hypertension and NCD management, dissemination and evaluation of task redistribution strategies is urgently required.

Indeed, the WHO Global Strategy for human resources for health indicates that the supply of health care workers needs to match health needs and priorities, including the increasing burden of NCDs.⁴⁷ Given that hypertension and other NCDs are chronic conditions that require longitudinal care, repeated contact with the health system, and interaction between patients and health workers, it is critical that governments, health systems, and health facilities account for NCD-related health workforce requirements in their planning and implementation processes.⁴⁸ The most recent health workforce requirements calculated by the WHO do in fact include NCDs such as hypertension, but the empirical basis for these requirements related to NCDs is weak. Our study therefore helps to fill that empirical gap, and we anticipate that our results will further inform the policy, planning, and health system development tasks of the global health community.

One limitation of our study is that, by utilizing a single regression analysis across the entire pooled dataset, we assumed that the relationship between health care worker density and hypertension treatment rate was the same across countries and over time. The sensitivity analyses we conducted were aimed at evaluating whether there were differences by date of study or income category classification, but the resulting smaller sample sizes did not allow for a definitive conclusion. Thus, while it is possible that this assumption may not hold across all health care worker density-treatment dyads, we feel that our contribution is an important first step to incorporate hypertension and other NCDs into health care worker requirements worldwide. Similarly, it is possible that high- vs. low-quality studies could yield different results. While this was beyond the scope of this paper, future research in this area can consider including quality grading in the selection of study data. Second, we acknowledge that, by focusing only on hypertension treatment rates, we do not take into consideration all of the other tasks that health care workers perform (e.g. vaccinations, antenatal care, care for other disease entities). The WHO has recently advocated for the use of a more integrated “index” of services required to meet the Sustainable Development Goals, and future efforts will benefit from that type of approach. Third, health workforce density is only one component that contributes to overall health care worker performance; other important components include accessibility, acceptability, and quality of care.⁴⁹ In addition, nurse density may reflect better access to health care in general, thus the interpretation of our results requires some caution. Fourth, we were not able to use pharmacist density data, as the data were not as routinely available as physician or nurse density. Given the importance of pharmacists in the dispensing of hypertension medications in many parts of the world, future research in this area should include pharmacist density as the data become more routinely available. Fifth, our analytic model aggregated all variables to the country level; therefore, individual-level variables impacting hypertension treatment rates (e.g. individual socio-economic status, access to care) were not evaluated. Relatedly, we pursued an analytic approach analogous to one previously utilized to assess the relationship between health care worker density and vaccination coverage.¹⁰ We recognize that this is not the only way to analyze this relationship, and alternative models and approaches can be adopted.

Perspectives

Our results support the strategy of task redistribution to meet the human resource challenge of management of NCDs including hypertension. Given the growing global need for hypertension and NCD management, our results have important implications for health policy, health system design, and program implementation. Future research assessing dissemination and evaluation of task redistribution strategies is urgently required. Investigating the relationship between health worker performance and other health outcomes will also be critical.

Conclusions

In this worldwide econometric analysis, we found that health care worker density was significantly associated with hypertension treatment. Notably, we found that that nurse density, not physician density, appeared to explain most of the relationship with hypertension treatment rate, after adjusting for several potential confounder variables. Our study contributes to the literature on cross-country analyses of health care worker density and treatment outcomes. Given the growing burden of CVD, hypertension, and other NCDs, these results have important implications for health policy, health system design, and program implementation. Future research assessing the relationship between health care worker performance and other health outcomes will also be critical.

Novelty and Significance.

What is New?

• Health worker density was significantly associated with hypertension treatment rate

• Our approach of aggregating and disaggregating physicians and nurses allowed us to highlight that nurse density, not physician density, appears to explain more of the relationship with hypertension treatment rate

What is Relevant?

• To the best of our knowledge, this is the first report examining the relationship between health worker density and hypertension treatment rates using worldwide data

Summary

• Our results support the strategy of task redistribution to meet the human resource challenge of management of non-communicable diseases including hypertension.

• Future research assessing the relationship between health worker performance and other health outcomes will also be critical.