Title of Manuscript: Health System Performance Towards Universal Health Coverage in Iran: A Comparative Panel Data Analysis

Building upon decades of continuous reforms, since 2014 under the banner of health transformation plan (HTP), Iran has been implementing various initiatives to strengthen its health system. Improving efficiency of the health system is fundamental to achieve better performance and reach universal health coverage (UHC). This article aimed to measure the efficiency and productivity changes in the Iranian health system during 2010-2015 in comparison with 36 selected other upper-middle income countries. We used panel data to measure the variations in technical efficiency (TE) and total factor productivity (TFP) through an extended data envelopment analysis (EDEA) and Malmquist productivity index, respectively. General Government Health Expenditure (GGHE) per capita (International dollar) was selected as input variable. Service coverage of diphtheria, tetanus and pertussis; family planning; antiretroviral therapy; skilled attendants at birth; Tuberculosis treatment success rate; and GGHE as % of Total Health Expenditure (THE) were considered as output variables. The data for each indicator were taken from Global Health Observatory data repository and World Development Indicator database, for a period of six years (2010-2015). Two main concepts exist for Allocative (AE) and Technical Efficiency (TE). AE refers to scrutinize either the choice of outputs or the choice of inputs. It determines whether limited resources are directed towards producing the correct mix of health outputs. AE also examines whether the entity uses an optimal mix of inputs to produce its chosen outputs, given the prices of those inputs 10 . TE indicates the extent to which the health system is maximizing outputs for a given set of inputs or minimizing inputs for a given set of outputs 15 . In this study, we use the term of efficiency as TE. Major policy changes, epidemiological transitions, changes in the government, climate change, etc., might have positive or negative impact on health system performance . Thus, to measure variations in health system performance, we measured growth over time, which might be an opportunity to improve

secure valued health system outputs. Two main concepts exist for efficiency: Allocative Efficiency (AE) and Technical Efficiency (TE). AE refers to scrutinize either the choice of outputs or the choice of inputs. It determines whether limited resources are directed towards producing the correct mix of health outputs. AE also examines whether the entity uses an optimal mix of inputs to produce its chosen outputs, given the prices of those inputs 10 . TE indicates the extent to which the health system is maximizing outputs for a given set of inputs or minimizing inputs for a given set of outputs 15 . In this study, we use the term of efficiency as TE. Major policy changes, epidemiological transitions, changes in the government, climate change, etc., might have positive or negative impact on health system performance 16 . Thus, to measure variations in health system performance, we measured productivity growth over time, which might be an opportunity to improve public welfare 17 .
As in the course of last four decades, The Islamic Republic of Iran has been initiating a number of health systems reforms, i.e. establishment of an extensive primary healthcare (PHC) network, the family physician program, and recently health transformation plan (HTP), to pave the way towards UHC 18,19 . Iran's national health accounts (NHA) indicate a healthcare efficiency at the intuitional and system levels 26 − 33 , limited evidence exists on efficiency and productivity trends over time at the system level 8 . This study aims: 1) to measure the TE changes and 2) to analyze productivity changes in Iran's health system over the 2010-2015. Our findings can, we envisage, provide evidence to identify the areas in need of greater concern towards gearing up current attempts to reach UHC in Iran, and perhaps beyond.

Methods
To overcome some limits of conventional DEA in measuring efficiency, our team created an extended DEA (EDEA). We adopted a method to analyze UHC performance relative to health spending 8 and applied panel data approach. The sample (or Decision-Making Units: DMUs) were selected 56 countries, classified as Upper Middle-Income Countries (UMICs) by the World Bank (with 896-12055 $ GDP per capita in 2015). EDEA necessitated weighted DMUs, therefore we removed 20 countries with less than 1,300,000 populations, assuming that the efficient management of health system in populated countries would be more challenging. The population of removed countries had the most variations with the population of remaining countries. Table 1 presents the 36 (out of 56) countries and their selected specifications that were included in our study. service coverage of tuberculosis (TB) treatment success rate. We excluded the two indicators: improved water and improved sanitation, because public spending on health generally does not usually pay for these interventions, and excluded antenatal care coverage as well due to lack of official data for several countries included in this study.
The last output variable in terms of financial protection was General Government Health Expenditure as percentage of Total Health Expenditure (GGHE/THE), which is the indirect measure of financial protection. We convened a panel of six recognized national experts to discuss the relevancy of selected variables to our analysis and managed to reach consensus on a guideline for each variable to ensure consistency in data definition and gathering. These experts were selected from public health (= 1), health economics (= 1), epidemiology (= 1), health services management (= 1), health policy (= 1), and health information system (= 1) disciplines.
We took data for each indicator from the WHO's Global Health Observatory (GHO) data repository 34 and World Bank (WB)'s World Development Indicator database 35 , for a period of six years (2010-2015). Since some values were missing, we applied an imputation technique to prepare data for analysis. To do this, we selected a standard value for each variable to measure the difference between actual and standard values. We then imposed a penalty on any DMU if the actual value was far from a standard range. The variables were also ranked according to their importance. The standard range and weight of the variables were defined by the aforementioned panel of experts (Table 2). To examine the TE changes (Aim 1), we used extended data envelopment analysis (EDEA).
EDEA determines conventionally how well a DMU, in our case a country, converts a set of inputs into a set of outputs. It assumes any deviations of DMUs from the frontier are due to technical inefficiency. A frontier is representing the efficient level of output (y) that can be produced from a given level of input (x). Therefore, DMUs are constrained to lie between completely efficient (= 1) and inefficient (= 0). Among different approaches, an output-oriented DEA model with variable returns to scale was selected.
To capture TFP changes (Aim 2), we used Malmquist Productivity Index (MPI) that is derived from the comparison of efficiency changes (Catch-up) to Frontier-shift 36 . Since the frontiers can shift over time, when inefficiency is assumed to exist, the relative movement of any given MUI over time will therefore depend on both its position relative to the corresponding frontier (efficiency change) and the position of the frontier itself (technology change). The TFP > 1 means that the productivity is improved.

Results
The TE scores of Iran' A) Descriptive analysis Table 3 provides descriptive summaries of the inputs and outputs of studied countries during 2020-2015. Descriptive statistics for both service coverage and financial protection is shown in Fig. 1, in which the 36 countries are categorized into quintiles based on their level of per capita public spending on health.   Our EDEA model identified four of 'best performing' countries with better performance related to their level of spending compared to other countries, i.e. Bosnia and Herzegovina, Namibia, Algeria and Belarus (Fig. 2). Namibia had the highest level of spending (674 Int'l Dollar per capita), while the remaining three best performing countries spent less than Namibia, ranging from 182 to 327 Int'l Dollar per capita). Regarding the level of spending for health, the reference country for Iran was Kazakhstan (Fig. 3).

Discussion
This study aimed to measure the TE changes and productivity variations in Iran's health system over the 2010-2015 period. Our findings demonstrate slightly positive changes of efficiency scores in 2014 and 2015. We reviewed the changes occurred in relevant input and output indicators as well as the recent HTP. Understandably, the five indicators of DPT3, FP, TB, ART, and SAB did not change significantly during the study period. This was because the data for these indicators are not usually calculated and reported annually.
Only two indicators of GGHE per capita and GGHE/THE changed over this period.
Therefore, it seems that changes made in these two indicators, especially those made in the output indicator, have improved the efficiency score in 2014 and 2015, which are mostly attributable to the HTP implementation.
As the key health sector reform towards UHC in Iran 18,37,38 , HTP had five main goals: sustainable financing, financial protection against catastrophic health costs, increasing access to quality healthcare services, improving the performance of health system and better public health indicators, ultimately. Several interventions have been implementing to provide additional financial resources and increase the existing ones, control the price of drugs and medical equipment, expand health insurance coverage among all society, reduce the share of inpatient and outpatient payments, and prevent informal payments.
Despite their positive impact, further interventions are still required to improve efficiency in the health system of Iran. These include reforms to improve governance structures, financial resources and service provision 19,39 .
With the efficiency score of lower than average, our findings revealed Iran's low rank among the studied countries. The efficiency model, as well as descriptive statistics of output indicators, suggest an equal to or higher than the average coverage of DPT3, TB, and SAB in Iran, compared to the studied countries over the study years. This demonstrates that Iran's low efficiency score might be due to other factors than these indicators. Nevertheless, the status of other three output indicators is different. FP coverage has slightly increased over the study years, which is significantly lower than the average of the studied countries and dramatically lower than the maximum value. A small increase is also observed in the ART coverage indicator in Iran, which is lower than the average of the studied countries. Although Iran's public sector share of total health expenditure was significantly lower than the average of the studied countries 1.
Interventions to expand the ART coverage: These interventions are suggested to be determined and prioritized using evidence-based practice models in order to identify and effectively cover high-risk groups through public health resources and improve effective HIV coverage.

2.
Targeted interventions to improve universal financial coverage against ever-spiraling health costs at the health system and public levels. At the health system level, a significant part of terrible health costs can be attributed to the wide range of services covered by public resources, as well as the use of inefficient service delivery models. In Iran, a broad range of low cost-effective services have been covered through available public resources, without following any priority setting.
Considering current constraints of financial health resources in Iran, covering some of these less-prioritized services through basic health insurance seems to be economically inefficient 46,47 . Worse still, fee-for-service (FFS) based payment method has increased the risk of informal payments and induced demands, both of which have contributed to the increased share of direct out-of-pocket (OOP) payments, and catastrophic health costs, ultimately. Let alone, after 15 years into the implementation of family physician program and the referral system, the national health system is still behind to establish appropriate rationing of services, which adds to the burden of undesirable health costs 18,22,48 .

Conclusion
There is a growing demand for efficiency improvements in the health systems to achieve UHC. While there is no defined set of indicators or precise method to measure heath system efficiency, using some techniques such as EDEA might help to paint the picture of health system efficiency, which will be in turn a starting point to identify the causes of any inefficiencies and design tailored interventions for improving efficiency in any specific context. While many countries have been tackling the insufficiency of financial as well as non-financial resources in their health systems, analyzing the health system efficiency can help improve the efficiency and expand fiscal space towards universal health coverage.   Descriptive statistics (inputs and outputs) by public spending on health quintile Appendix.docx