Open Access

Technical efficiency of rural primary health care system for diabetes treatment in Iran: a stochastic frontier analysis

  • Mostafa Qorbani1, 2, 3,
  • Farshad Farzadfar3,
  • Reza Majdzadeh4,
  • Kazem Mohammad4 and
  • Abbas Motevalian1Email author
Journal of Diabetes & Metabolic Disorders201716:33

https://doi.org/10.1186/s40200-017-0312-8

Received: 27 May 2017

Accepted: 24 July 2017

Published: 14 August 2017

Abstract

Background

Our aim was to explore the technical efficiency (TE) of the Iranian rural primary healthcare (PHC) system for diabetes treatment coverage rate using the stochastic frontier analysis (SFA) as well as to examine the strength and significance of the effect of human resources density on diabetes treatment.

Methods

In the SFA model diabetes treatment coverage rate, as a output, is a function of health system inputs (Behvarz worker density, physician density, and rural health center density) and non-health system inputs (urbanization rate, median age of population, and wealth index) as a set of covariates. Data about the rate of self-reported diabetes treatment coverage was obtained from the Non-Communicable Disease Surveillance Survey, data about health system inputs were collected from the health census database and data about non-health system inputs were collected from the census data and household survey.

Results

In 2008, rate of diabetes treatment coverage was 67% (95% CI: 63%–71%) nationally, and at the provincial level it varied from 44% to 81%. The TE score at the national level was 87.84%, with considerable variation across provinces (from 59.65% to 98.28%).Among health system and non-health system inputs, only the Behvarz density (per 1000 population)was significantly associated with diabetes treatment coverage (β (95%CI): 0.50 (0.29–0.70),p < 0.001).

Conclusion

Our findings show that although the rural PHC system can considered efficient in diabetes treatment at the national level, a wide variation exists in TE at the provincial level. Because the only variable that is predictor of TE is the Behvarz density, the PHC system may extend the diabetes treatment coverage by using this group of health care workers.

Keywords

Stochastic frontier analysis Technical efficiency Human resources Diabetes mellitus

Background

Although the number of people living with and dying of diabetes mellitus (DM) is increasing across the world [1, 2] the major increase is occurring in low and middle-income countries [3]. It has been reported that the prevalence rate of type 2 diabetes in Iran among 25 to 64 year-old age group is 11.37% (95%CI: 9.86–12.89) [4]. According to a recently conducted study in Iran, DM treatment coverage was 39.2% in 2005 [5]. In Iran, the National Plan for Diabetes Control and Prevention (NPDCP) was developed and implemented between 1996 and 2002,in order to prevent and control DM and its complications in an early phase through screening ‘individuals at risk’ and pregnant women. The NPDCP is integrated into the primary health system (PHC) of Iran [6]. As a part of its protocol in rural areas, Iranian rural primary health workers (Behvarzes) are trained to identify high-risk groups and refer them to physicians in health houses for DM testing and, if required, treatment. The main duties of Behvarz are taking care and follow up of patients.6

Human resources for health are known to have imbalanced distribution in low and middle income countries [7] and a few studies have investigated the association between this geographical imbalance in the distribution of human resources for healthand health outputs or intervention coverage [816]; the results of these studies were contradictory.

The first study that measured the efficiency of health systems in attaining health outputs applying production function models was the world health report in 2000 [17]. Technical efficiency (TE) was defined as the ability to produce the maximum possible output from a given set of inputs and wasmeasured in terms of the relationship between actual level of health output and the maximum attainable output for the observed inputs [18].Stochastic frontier analysis (SFA) is a parametric technique for estimating efficiency in health care systems, which is has been introduced and used recently.The SFA uses regression analysis to estimate the production frontier and measures the efficiency of a unit using the residuals from the estimated equation. In this method, in addition to a random error, another error term is included which represents systematic inefficiency.

In this study, we explored the TE of Iranian rural PHC system for DM treatment coverage rate using the SFA technique in 2008. We also tested the strength and significance of the effect of health and non-health inputs on levels of DM treatment coverage.

Methods

Study area and data set

In this cross-sectional study, rural areas of all provinces in Iran in 2008 were assessed for estimating TE of the rate of self-reported DM treatment coverage using the SFA model. According to this model DM treatment coverage rate as the output was a function of health system inputs (such as Behvarz worker density, physician density and rural health house density)and non-health system inputs (urbanization rate, median age of population, and wealth index) as a set of covariates. The data about the rate of DM treatment coverage was obtained from the Non-Communicable Disease Surveillance Survey (NCDSS) 2008; the data about health human resources and rural health houses density were collected from the health census database called D-Tarh 2008; median age, and rate of urbanization were collected from the census datas 2006 and 2011; and wealth index were collected from the household survey of 2008.

Definitions

The rate of self-reported DM treatment coverage was defined as the proportion of people with diabetes who reported taking medication for diabetes treatment. Rural health center, Behvarz, and physician density are defined as the ratio of the number of rural health centers, Behvarzes and physicians per 1000 total populations. Median age of population was estimated by interpolating population characteristics attained by census 2006 and 2011 using the Growth Method. Urbanization rate was defined as the ratio of urban to rural population which was estimated by interpolating population characteristics attained by census 2006 and 2011 years using Growth method. According to previous evidence in the “Progress in the International Reading Literacy Study” (PIRLS) [19] and by using principle component analysis (PCA) method, family assets and characteristics of living place were summarized in one main variable to construct wealth index. This main variable explained 69% of total variance.

Conceptual framework of SFA model

In present study we used the SFA model suggested by Battese and Coelli [20]. The model is as follows:
$$ yi={X}_i\beta +{v}_i-{u}_i $$

Where yi is the logarithm of the output in province i, X i is the vector of inputs associated with province i, β is the vector of parameters to be estimated, v i is the random error components and u i is the inefficiency component. The noise component vi, have a normal distribution N(0,σ3 u) and ui which are nonnegative random variables, have truncated normal distribution; N+ (μ,σ3 u) . The μ of the truncated normal distribution is modeled as a linear function of a set of covariates.

According to this SFA model, we modeled DM treatment coverage (C i ) as a function of a vector of inputs x i for the i th province, C i = f(x i ). The xi vector of inputs is composed of Behvarz density B i , physician density P i, rural health center density RHC i ,
$$ {TE}_i={C}_i/f\ {\left({B}_i,{P}_i,{RHC}_i;\beta \right)}^{\ast}\mathit{\exp}.\left({v}_i\right) $$

The stochastic frontier model (f(B i , P i , RHC i ; β)* exp.(v i )) consists of two parts: a deterministic part and a producer-specific part (exp(vi)) which captures the effect of random variables on each producer. Deterministic part [f (B i , P i , RHC i ; β] takes the log-linear Cobb–Douglas form24:

According to above formula, TEi = 1 if observed output (C i ) is equal to maximum attainable output in an environmental characterized by exp.(v i ). TEi < 1 when observed output (C i ) is less than maximum attainable output in an environmental characterized by exp. (v i ) which can be vary across provinces.

Results

In 2008, the rate of DM treatment coverage in rural areas was 67% (95% CI: 63 to 71), nationally. In this year, the density of Behvarz workers was 1.61 per 1000 total population (with range of 0.84 to 2.30) and the density of physicians per 1000 total populations was 0.29 (range: 0.18–0.52) (Table 1).
Table 1

Descriptive statistics of input and output variables in the SFA model for DM treatment coverage in rural areas of Iran, 2008

Variable

Mean

95%CI

Minimum

Maximum

DM treatment coverage (self-report)

0.67

0.63–0.71

0.44

0.81

Physician densitya

0.29

0.26–0.32

0.18

0.52

Behvarz densitya

1.61

1.5–1.7

0.84

2.30

Rural health center densitya

0.13

0.11–0.14

0.08

0.19

Wealth index

0.004

−0.35-0.35

−1.96

1.66

Urbanization rate

0.66

0.6–0.7

0.50

0.95

Median age

24.5

23.8–25.3

19.6

28

aPer 1000 total population

Figure 1 shows the rate of DM treatment coverage at the provincial level in Iran, 2008. This rate varies considerably across provinces from 44% (Ilam province) to 81% (West Azerbaijan province).
Fig. 1

DM treatment coveragein rural areas of Iran at provincial level, 2008

Table 2 presents the results of the SFA model for TE of the PHC system for DM treatment coverage rate. Only the effect of Behvarz worker density (per 1000 total population) on DM treatment coverage rate was statistically significant (β coefficient: 0.50, 95%CI: 0.29–0.70), whereas the effect of physician density on this rate was not statistically significant (β coefficient: 0.05, 95%CI: -0.12-0.22).The results of Wald x 3 statistics showed that the PHC system was not efficient in attaining DM treatment coverage rate (Wald x 3: 30.35, p-value < 0.001). In addition, the estimate for variance parameter of inefficiency components (σ3 u )indicated that this effects was statistically significant (P-value < 0.001).
Table 2

Coefficients of the SFA model for DM treatment coveragein rural areas of Iran, 2008

Independent variable

Coefficients

95%CI

P-value

Production function

Behvarz density (per 1000 total population)

0.50

0.29, 0.70

<0.001

Physician density (per 1000 total population)

0.05

−0.12,0.22

0.60

Rural health center density (per 1000 total population)

−0.07

−0.08,0.21

0.38

Wealth Index

−0.51

−7.2, 6.23

0.88

Urbanization rate

−2.86

−15.58,9.86

0.65

Median age (years)

−17.27

−43.86,9.32

0.20

Constant

−0.47

−0.74,-0.19

0.001

Distribution of u and v

σ3 u

0.15

0.11,0.20

<0.001

σ3 v

0.06

0.03,0.08

0.35

Number of observation: 31

Wald x3

30.35

 

<0.001

Table 3 shows the estimated TE for the PHC system at the provincial level, using the SFA model. The TE score at the national level was 87.84%, with considerable variation across provinces (from 59.65% (Chaharmahal & Bakhtiari) to 98.28% (Markazi province)). Figure 2 demonstrates TE of Iranian PHC system for DM treatment at provincial level.
Table 3

DM treatment coverage and technical efficiency at national and provincial level in rural areas of Iran, 2008

ID

Province

Number of rural total population

DM treatment coverage%

Efficiency %

1

Markazi

408,630

77

98.28

2

Kermanshah

616,599

80

97.41

3

East Azerbaijan

410,522

79

97.04

4

Fars

1,578,334

81

97.03

5

Qazvin

297,505

77

95.66

6

West Azerbaijan

1,072,773

81

95.40

7

Boshehr

280,057

66

95.03

8

Golestan

833,255

70

94.40

9

RazaviKhorasan

1,752,463

65

93.42

10

Ardabil

477,219

74

93.91

11

North Khorasan

382,365

79

92.42

12

Gilan

1,044,515

73

92.16

13

Mazandaran

1,311,380

71

90.79

14

Lorestan

672,328

62

90.13

15

Semnan

120,825

81

90.06

16

South Khorasan

246,589

75

89.14

17

Kohgiluyeh&Boyer-Ahmad

350,286

57

88.16

18

Khozestan

1,229,588

67

87.72

19

Zanjan

395,989

49

87.30

20

Sistan and Baluchestan

942,249

48

86.69

21

Hormozgan

663,401

57

85.96

22

Kurdestan

611,689

63

85.46

23

Hamadan

724,877

50

85.35

24

Alborz

130,883

66

82.74

25

Isfahan

622,441

78

82.28

26

Ilam

209,650

44

80.01

27

Qom

48,870

66

79.04

28

Yazd

184,674

63

78.61

29

Tehran

395,202

66

76.86

30

Kerman

895,607

54

76.05

31

Chaharmahal&Bakhtiari

1,087,876

56

59.65

-----

National

19,998,643

67

87.84

Fig. 2

Technical efficiency of rural primary health care system for DM treatment coverage at provincial level in Iran, 2008

Discussion

According to our results, nationally DM treatment coverage rate in rural areas was 67%(95% CI: 63 to 71). Farzadfar et al. reported that DM treatment coverage rate in urban and rural areas in 2005 according to the NCDSS was 39.2% (95% CI 37·7 to 40·7); this discrepancy can be attributed to the definition of DM treatment coverage; in our study, DM treatment coverage rate was based on self-report which may justify the observed higher rate [5].

We also found a significant association between Behvarz worker density and DM treatment coverage rate. Our results show that each additional Behvarz worker density per 1000 total population increased DM treatment coverage rate by 0.5(95% CI: 29–70), consistent with Farzadfar et al. who show that each additional Behvarz worker density per 1000 total population decreases fasting plasma glucose by 0.09 mmol/L (0.01 to 0.18) on average [5]. Findings from other studies on the association between the density of human resources and health system outputs are contradictory. Some show that the density of health-care workers is associated with improved health system outputs [811]; however other do not support this association [1216]. Most studies have conducted cross-country analyses, but the associations at subnational level are considered by only a few of them [5]. Previous studies showed that in developing countries efficient utilization of human resources for attaining health outputs at the subnational units faces two major problems: uneven distribution of health workers at the subnational level, and difficulty in identifying the threshold of human resource density to achieve the minimum acceptable level of health outputs [722]. To address these problems the WHO in 2006 dedicated world health reports to the topic of human resources for attaining health outputs, and declared that “health workers are crucial for attaining acceptable heath outputs through the PHC system” and that “achieving PHC and Millennium Development Goals are possible only by an appropriately prepared and efficient health workers” [23].

In present study, the effect of physician density (unlike that for Behvarz density) was not statistically significant on DM treatment coverage, in line with Farzadfar et al.’s study [5]. The likely reason for this observation has to do with the defined roles of these groups in the Iran health system. Iran has a National Diabetes Control and Prevention Program, with specific roles for Behvarz workers. Thus, although physicians in the PHC system also have a key role in the definite diagnosis and treatment of referred patients in rural areas, Behvarz workers are in charge of the initial identification of patients according to a determined protocol as well as supervision of their diet and life style modifications, along with follow- ups to check that patients take their medication and visit physicians periodically, if they have symptoms of DM.

We found a wide variation in the levels of TE across Iranian provinces, from 59.65% to 98.28% which suggests that some of provinces could increase their level of DM treatment coverage by using their human resources, particularly Behvarz workers, more efficiently. Overall, the average score of efficiency was 87.84(%). This implies that the provinces, in terms of TE, could produce the same level of DM treatment coverage by using 12.16(%) fewer resources (inputs).

Previous studies have shown that increasing the number of human resources for health, especially Behvarz worker, with well-defined guidelines is a major strategy for attaining acceptable management of NCD [5]. However; our finding suggests that the PHC may extend the coverage of DM treatment by using Behvarz worker more efficiently. In other words, on average, provinces are using more inputs than they need.

To the best of our knowledge, the present study is the first study in Iran that estimates TE of human resources for attaining DM treatment coverage at the sub-national level using the SFA model.

There were some limitations in our study. The main one was the definition of DM treatment coverage which was based on self-report. In addition using survey data for calculating output instead of obtaining data from rural health house may limits our data. We also were not able to include public health expenditure per capita of each province in the SFA model.

Conclusion

Our findings show that although the rural PHC system can be considered efficient in DM treatment coverage at the national level; a wide variation exists in TE at the provincial level. Because the only variable that is predictor of TE is the Behvarz density, the PHC system may extend DM treatment coverage by using this group of health care workers. Our results can help health policy makers to attain higher health outputs by using community health workers; we recommend that the PHC system in Iran rely more on this type of workforce, which perhaps is more cost effective than employing physicians.

Endnote

Any substance to which subjects were sensitive were mentioned it in the questionnaire.

Abbreviations

DM: 

Diabetes mellitus

PHC: 

Primary health care

SFA: 

Stochastic frontier analysis

TE: 

Technical efficiency.

Declarations

Acknowledgments

We would also like to appreciate Dr. Mansourian, Dr. Kasaeian and Mr. Khajavi for their kindly contribution.

Funding

Iran University of Medical Sciences, Tehran, Iran.

Availability of data and materials

Please contact author for data requests.

Authors’ contributions

MQ and FF, design, data analysis, and interpretation, RM, KM, design, AM and MQ design and revision. All authors read and approved the final manuscript.

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

There are no financial or non-financial competing interests (political, personal, religious, ideological, academic, intellectual, commercial, or any other) to declare in relation to this manuscript.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.

Authors’ Affiliations

(1)
Department of Epidemiology, School of Public Health, Iran University of Medical Sciences
(2)
Non-communicable Diseases Research Center, Alborz University of Medical Sciences
(3)
Non-communicable Diseases Research Center, Endocrinology and Metabolism Population Sciences Institute, Tehran University of Medical Sciences
(4)
Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences

References

  1. Worldwide trends in diabetes since 1980: a pooled analysis of 751 population-based studies with 4.4 million participants. NCD Risk Factor Collaboration (NCD-RisC). Lancet. 2016 9;387(10027):1513–30. doi: 10.1016/S0140-6736(16)00618-8. Epub 2016 Apr 6.
  2. Global, regional, and national comparative risk assessment of 79 behavioural, environmental and occupational, and metabolic risks or clusters of risks, 1990–2015: a systematic analysis for the Global Burden of Disease Study 2015.GBD 2015 Risk Factors Collaborators. Lancet. 2016;388(10053):1659–1724. doi: 10.1016/S0140-6736(16)31679-8.
  3. King H, Aubert RE, Herman WH. Global burden of diabetes, 1995–2025: prevalence, numerical estimates, and projections. Diabetes Care. 1998;21:1414–31.View ArticlePubMedGoogle Scholar
  4. Esteghamati A, Etemad K, Koohpayehzadeh J, Abbasi M, Meysamie A, Noshad S, et al. Trends in the prevalence of diabetes and impaired fasting glucose in association with obesity in Iran: 2005-2011. Diabetes Res Clin Pract. 2014;103(2):319–27. doi:10.1016/j.diabres.2013.12.034. Epub 2014 Jan 3View ArticlePubMedGoogle Scholar
  5. Farzadfar F, Murray CJ, Gakidou E, Bossert T, Namdaritabar H, Alikhani S, et al. Effectiveness of diabetes and hypertension management by rural primary health-care workers (Behvarz workers) in Iran: a nationally representative observational study. Lancet. 2012;379:47–54.View ArticlePubMedGoogle Scholar
  6. Mahdavi AR, Etemad K, Haider M, Alavinia SM. The effect of seeing a family physician on the level of glycosylated hemoglobin (HbA1c) in type 2 diabetes mellitus patients. J Diab Metab Disord. 2013;12:2.View ArticleGoogle Scholar
  7. Gupta N, Zurn P, Diallo K, Dal Poz MR. Uses of population census data for monitoring geographical imbalance in the health workforce: snapshots from three developing countries. Int J Equity Health. 2003;2:11.View ArticlePubMedPubMed CentralGoogle Scholar
  8. 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.View ArticleGoogle Scholar
  9. Anand S, Bärnighausen T. Human resources and health outcomes: cross-country econometric study. Lancet. 2004;364:1603–9.View ArticleGoogle Scholar
  10. Mitchell AD, Bossert TJ, Yip W, Mollahaliloglu S. Health worker densities and immunization coverage in Turkey: a panel data analysis. Hum Resour Health. 2008;6:29.View ArticlePubMedPubMed CentralGoogle Scholar
  11. Frankenberg E. The effects of access to health care on infant mortality in Indonesia. Health Transit Rev. 1995:143–63.Google Scholar
  12. Cochrane AL, St Leger A, Moore F. Health service'input'andmortality'output'in developed countries. J Epidemiol Community Health. 1978;32:200–5.View ArticlePubMedPubMed CentralGoogle Scholar
  13. Aakvik A, Holmås TH. Access to primary health care and health outcomes: the relationships between GP characteristics and mortality rates. J Health Econ. 2006;25:1139–53.View ArticlePubMedGoogle Scholar
  14. Hertz E, Hebert JR, Landon J. Social and environmental factors and life expectancy, infant mortality, and maternal mortality rates: results of a cross-national comparison. Soc Sci Med. 1994;39:105–14.View ArticlePubMedGoogle Scholar
  15. Gulliford MC. Availability of primary care doctors and population health in England: is there an association? J Public Health. 2002;24:252–4.View ArticleGoogle Scholar
  16. Kim K, Moody PM. More resources better health? A cross-national perspective. Soc Sci Med. 1992;34:837–42.View ArticlePubMedGoogle Scholar
  17. World Health Organization. Health systems: improving performance. Geneva: The World Health Report; 2000.Google Scholar
  18. Tandon Ajay, Jeremy A. Lauer, David B. Evans, and Christopher JL. Murray. 2003. “Health System Efficiency: Concepts.” Health Systems Performance Assessment: Debates, Methods and Empiricism. Edited by Christopher JL Murray and David B. Evans. Chapter 50, pp. 683-691.Google Scholar
  19. Caro DH, Cortés D. Measuring family socioeconomic status: an illustration using data from PIRLS 2006. IERI Monograph Series Issues and Methodologies in Large-Scale Assessments. 2012;5:9–33.Google Scholar
  20. Battese GE, Coelli TJ. A model for technical inefficiency effects ina stochastic frontier production functionfor panel data. Empir Econ. 1995;20:325–32.View ArticleGoogle Scholar
  21. Battese GE, Coelli TJ. Frontier production functions, technical efficiency and panel data: with application to paddy farmers in India. J Prod Anal. 1992;3:153–69.View ArticleGoogle Scholar
  22. Gupta N, Diallo K, Zurn P, Dal Poz MR. Assessing human resources for health: what can be learned from labor force surveys. Hum Resour Health. 2003;1:1–16.View ArticleGoogle Scholar
  23. Organization WH. World health report 2006: working together for health, WHO, Geneva, 2006. Sines, Erin, Anne Tinker and Julia Ruben,‘The Maternal-Newborn-Child Health Continuum of Care: A collective effort to save lives’, Policy Perspectives on Newborn Health, Save the Children and Population Reference Bureau, Washington, DC 2006:2-3.Google Scholar

Copyright

© The Author(s). 2017

Advertisement