Informing a cost-effectiveness threshold for Saudi Arabia

Abstract

Background

Saudi Arabia’s Vision 2030 aims to reform health care across the Kingdom, with health technology assessment being adopted as one tool promising to improve the efficiency with which resources are used. An understanding of the opportunity costs of reimbursement decisions is key to fulfilling this promise and can be used to inform a cost-effectiveness threshold. This paper is the first to provide a range of estimates of this using existing evidence extrapolated to the context of Saudi Arabia.

Methods and materials

We use four approaches to estimate the marginal cost per unit of health produced by the healthcare system; drawing from existing evidence provided by a cross-country analysis, two alternative estimates from the UK context, and based on extrapolating a UK estimate using evidence on the income elasticity of the value of health. Consequences of estimation error are explored.

Results

Based on the four approaches, we find a range of SAR 42,046 per QALY gained (48% of GDP per capita) to SAR 215,120 per QALY gained (246% of GDP per capita). Calculated potential central estimates from the average of estimated health gains based on each source gives a range of SAR 50,000–75,000. The results are in line with estimates from the emerging literature from across the world.

Conclusion

A cost-effectiveness threshold reflecting health opportunity costs can aid decision-making. Applying a cost-effectiveness threshold based on the range SAR 50,000 to 75,000 per QALY gained would ensure that resource allocation decisions in healthcare can in be informed in a way that accounts for health opportunity costs.

Limitations

A limitation is that it is not based on a within-country study for Saudi Arabia, which represents a promising line of future work.

Plain language summary

Healthcare in Saudi Arabia is undergoing wide-ranging reform through Saudi Arabia’s Vision 2030. One aim of these reforms is to ensure that money spent on healthcare generates the most improvement in population health possible. To do this requires understanding the trade-offs that exist: funding one pharmaceutical drug means that same money is not available to fund another pharmaceutical drug. This is relevant whether the new drug would be funded from within the existing budget for healthcare or from an expansion of it. If the drugs apply to the same patient population and have the same price, the question is simply, “which one generates more health?” In reality, we need to compare pharmaceutical drugs for different diseases, patient populations, and at a range of potential prices to understand whether the drug in question would generate more health per riyal spent than what is currently funded by the healthcare system. This paper provides the first estimates of the amount of health, measured in terms of quality adjusted life years (QALYs), generated by the Saudi Arabian healthcare system. We find that the healthcare system generates health at a rate of one QALY produced for every 50,000–75,000 riyals spent (58–86% of GDP per capita). Using the range we estimate to inform cost-effectiveness threshold can aid decision-making.

Keywords:

• Cost-effectiveness threshold
• cost-effectiveness analysis
• health technology assessment
• decision making

JEL CLASSIFICATION CODES:

• I18
• I1
• I
• H51
• H5
• I10

View correction statement:

Correction

Introduction

Healthcare in Saudi Arabia

The Kingdom of Saudi Arabia, under Article 31 of the national constitution, provides free healthcare to all its citizens¹. Citizens constitute around 60% of the population, with the other 40% (non-citizens) being required to have private health insurance through their employers since 1999^2,3. The Ministry of Health (MOH) in Saudi Arabia is responsible for providing the majority of healthcare services with the remaining proportion being provided by other government agencies (such as the National Guard and the Ministry of Defense) and private institutions⁴. Each ministry has a budget for its hospitals and is able to enlist in formularies.

As part of Saudi Vision 2030, a comprehensive long-term development strategy announced by the government in 2016, the healthcare system in Saudi Arabia is undergoing significant transformation. It aims to reform health care across the Kingdom, with health technology assessment (HTA) adopted as one tool promising to improve the efficiency with which resources are used⁵. Pharmaceutical spending represented 19.4% of total health expenditure in 2018, which is above the average in OECD countries (18%)⁶. Pharmaceutical spending increased by 12% from 2014 to 2015, climbing from SAR 28.4 billion (USD 7.6 billion) to SAR 31.8 billion. Some of these increases can be traced to a rapidly growing population, change in the demographic characteristics, the rising incidence and burden of respiratory diseases, diabetes, hypertension, and cancer⁷. The Saudi Arabia pharmaceutical market was valued at SAR 38.2 billion in 2020 and is expected to have a constant annual growth rate of more than 6% from 2020 to 2026⁸. In response to these conditions, the government of Saudi Arabia requires “considered pricing” of pharmaceutical products to constrain public and private pharmaceutical costs and participates in the Gulf Cooperation Council (GCC) drug pricing plan⁷.

The National Transformational Program (NTP) 2020 sets out strategic objectives to restructure the healthcare sector and to enable a shift toward a value-based healthcare system. A major theme of the NTP is to transform healthcare to ease access to healthcare, improve quality and efficiency of healthcare services, and promote prevention against health risks⁹. Numerous initiatives are underway to achieve these objectives, such as corporatization of healthcare facilities, establishment of a national HTA center, and implementation of e-health facilities, along with many others. Momentum within policymaking, as exists now in Saudi Arabia, is critical to the success of HTA in improving the efficiency with which resources for healthcare are used.^10,11

Saudi Arabia has been increasingly using pharmaco-economic evaluation methods to inform decisions around the pricing and reimbursement decisions of new health technologies (drugs, devices, procedures or systems)¹². Pharmaco-economic analysis, such as cost-effectiveness analysis (CEA) or cost-utility analysis (CUA), where the estimated costs and benefits of a health technology are judged against a “threshold” representing value for money, typically forms a key component of HTA. However, there is no single definition for what is considered good value for money (and is therefore reflected in the threshold used for decision making), which leads to difficulties in interpreting studies of cost-effectiveness, e.g. Naber et al. (2021)¹³.

Defining value for money

Decision-making thresholds set forth for use by HTA bodies to define value for money differ between healthcare systems with explicit values having been set forth in few countries (a threshold range of GBP 20,000 to 30,000 per QALY is used by the National Institute for Health and Care Excellence (NICE) for the United Kingdom, a benchmark of EUR 45,000 per QALY is used by the National Centre for Pharmacoeconomics (NCPE) for Ireland, USD 100,000 to 150,000 per QALY by the Institute for Clinical and Economic Review (ICER) for the United States, and CAD 150,000 to 200,000 per QALY by the Patented Medicines Pricing Review Board for Canada)¹⁴. An explicit criterion to judge value for money lends transparency and consistency to the HTA process¹⁵, and, if that criterion reflects the opportunity cost of health expenditure, enables the HTA body to make recommendations that improve the efficiency of healthcare expenditure. Failure to set an explicit decision-making threshold reduces the range of analytical tools available to inform decision-making, especially where competing criteria are involved, as well as reducing the accountability of the process, and risks gaming by pharmaceutical companies to target pharmacoeconomic value to just below the implicit threshold¹⁶. Identifying a decision-making threshold for Saudi Arabia is key to the successful implementation of HTA to achieve the health objectives of the NTP⁷.

Existing decision-making thresholds used elsewhere (both explicit and implicit) vary in terms of their conceptual underpinnings, with many reflecting heuristics or based on past decisions. Examples of thresholds reflecting heuristics include the ICER decision-making threshold and the 1x and 3x GDP thresholds initially set forth by the World Health Organization (WHO). The NICE decision-making threshold reflects past decisions. Other decision-making thresholds are based on either a demand- or supply-side perspective to informing decisions. The “demand-side” perspective reflects the value of health gains to society (which may be estimated from willingness-to-pay studies or inferred from an estimated value of a statistical life)¹⁵. Such values can be useful for informing whether the current budget for healthcare reflects society’s preferences, but do not necessarily reflect the opportunity cost of healthcare expenditure.

The “supply-side” perspective represents the health opportunity cost of reimbursement decisions. The opportunity cost of reimbursing a health technology is the health benefit that could have been gained had the money required to fund it been spent elsewhere in the healthcare system¹⁷. This perspective therefore aligns with an objective of health maximization, and reflects the amount of health that is currently generated by the healthcare system (given existing infrastructure, provision, etc.). This perspective is relevant to both publicly and primarily privately provided and funded healthcare systems, such as the UK (where healthcare is largely publicly provided) and the US (where healthcare is largely privately provided)^18–20.

Accounting for opportunity cost in healthcare

An estimate of the marginal cost per unit of health produced by the healthcare system is necessary for calculating opportunity cost in healthcare. This enables the calculation of the expected net health impact of providing a health technology (i.e. the expected health gains from an intervention net of its expected health opportunity cost), and is a crucial calculation to determine whether the value added by a health technology is greater than the health that could have been gained by spending the money required to provide it elsewhere.

Value can also be expressed in terms of the additional funding that would be required to achieve similar net health impacts: the net monetary value is the amount of funding that would be required to deliver the same amount of net health gained. This calculation is useful for determining the maximum price at which a new technology would generate a net health gain to the population (i.e. its value-based price)^21,22.

Key to fulfilling the promise that Saudi Arabia’s Vision 2030 holds for improving the efficiency of health spending is that decision-makers are able to make decisions to prioritize new health technologies informed by a quantification of the expected health opportunity costs of those choices. When used to underpin the decision-making threshold, a value that reflects health opportunity cost in the healthcare system ensures that at a given price the health technology improves population health (net of the health that could have been gained had the money been spent elsewhere in the healthcare system). While some studies have assessed various factors related the productivity of the Saudi Arabian healthcare system, none has assessed the marginal cost per unit of health produced by the healthcare system^23,24.

This study quantifies, for the first time, the marginal cost per unit of health produced by the healthcare system for Saudi Arabia. Bespoke within-country analyses to estimate the marginal cost per unit of health produced by the healthcare system are time consuming and data intensive^25,26. As a result, approaches have been proposed in order to provide an alternative source of evidence²⁶. In order to provide timely input into discussions around potential values to inform a decision-making threshold for use in HTA in Saudi Arabia, we follow the approach taken by Ochalek and Lomas (2020) bringing together the best currently available evidence applicable to the Saudi Arabian context to identify a range of values that reflect health opportunity costs in Saudi Arabia based on existing evidence of the marginal cost per unit of health produced by the healthcare system. This represents the totality of approaches available which are relevant to Saudi Arabia, and has precedent, having formed the basis of recommendations for cost-effectiveness thresholds in policy proposals by Norway and Canada while bespoke evidence is produced^27,28. We begin by reviewing existing estimates of the marginal cost per unit of health produced by healthcare systems. We then describe what has been done to inform policy. Finally, we present an approach for Saudi Arabia, which we implement with local data. The implications of the results for policy are discussed.

Methods

Existing estimates of the marginal cost per unit of health produced by the healthcare system

An empirical estimate of the marginal cost per unit of health produced by the healthcare system was first estimated by Claxton et al. (2015) for the United Kingdom²⁹. More recent work has followed a broadly similar approach using within country data to first estimate the effect of expenditure on health outcomes, and then translating this effect into a cost per unit of health that accounts for survival and health-related quality of life (HRQoL). This is typically measured in terms of quality-adjusted life years (QALYs) gained or disability-adjusted life years (DALYs) averted. Estimates using within country data are now available for Spain, Australia, the Netherlands, Sweden, South Africa, China, Colombia and Indonesia^30–38. Edney et al. (2021) provides a review of the methods used to estimate the marginal cost per unit of health produced by the healthcare system²⁵. The data requirements for this type of analysis are stringent and the time required to undertake the analysis is non-negligible – often measured in years rather than months²⁶.

Alternative approaches for quantifying these values have been taken by Woods et al. (2016)³⁹, Ochalek et al. (2018)⁴⁰ and Ochalek and Lomas (2020)⁴¹. Woods et al. (2016) extrapolate from the UK estimate to provide estimates for a range of countries by applying data on the income elasticity of the value of health to generate ranges of cost per QALY gained estimates. Ochalek et al. (2018) draw from the body of existing work to estimate the effect of different levels of healthcare expenditure on mortality outcomes using cross-country data and expand upon a particular methodology and dataset from Bokhari et al. (2007)⁴² to estimate the effect of changes in health expenditure on health outcomes for a range of low- and middle-income countries. Following the methods applied in within country work, they translate the estimated effect of expenditure on health outcomes for each country into a cost per “unit of health” that accounts for survival HRQoL using country-specific data on health expenditure, epidemiology, and demography from an international dataset. Ochalek and Lomas (2020) provide a range of estimates for 33 high-income and BRIICS (Brazil, Russia, India, Indonesia, China, South Africa) countries using the same method as used in Ochalek et al. (2018), and additionally applying other existing elasticities drawn from selected UK within-country studies^43,44.

Estimates informing policy

Few countries have access to a bespoke within-country estimate of the marginal cost per unit of health produced by the healthcare system. As a result, a number of approaches have been proposed in order to provide an alternative source of evidence²⁶. The approaches adopted in Ochalek and Lomas (2020) form the basis of recommendations for cost-effectiveness thresholds in policy proposals by Norway and Canada while bespoke evidence is produced^27,28.

Common to both Norwegian and Canadian proposals is the use of estimates from Woods et al. (2016). This study extrapolates from the Claxton et al. (2015) estimate of the marginal cost per unit of health produced by the UK healthcare system using the income elasticity of the value of health to adjust for differences in GDP per capita from the UK. Assuming an income elasticity of the value of health equal to 1⁴⁵ implies that a cost-effectiveness threshold equal to 48% of GDP per capita is appropriate in order to reflect health opportunity costsⁱ. The assumptions of this method are noted by Woods et al. (2016) and further discussed in Ochalek and Lomas (2020)^39,41.

An approach for Saudi Arabia

We generate four estimates for Saudi Arabia. The first three are based on the application of existing relevant estimates of the marginal productivity of healthcare expenditure. We employ two estimates from the UK: Andrews et al. (2017) and Lomas et al. (2019)^43,44; and one estimate from across high-income countries: from Ochalek and Lomas 2020, to the Saudi Arabia context, and undertake sensitivity analysis around these. The fourth estimate is from Woods et al. (2016) and is based on extrapolating the UK estimate of the cost per unit of health produced by the healthcare system to Saudi Arabia using data on the income elasticity of the value of health to determine cost per QALY gained³⁹. The remainder of this section describes the steps requires to obtain the first three estimates from available estimates of the marginal cost per unit of health produced by the healthcare system.

Estimating the marginal cost per unit of health produced by the healthcare system can be understood as requiring two steps²⁵. First, an estimate the effect of expenditure on health is needed. This can be estimated from within-country or cross-country data. Estimation based on within-country data is not constrained by international comparability, which means that more variables may be available, and it may be easier to obtain plausible instrumental variables or natural experiments to inform an identification strategy. On the other hand, using cross-country data enables estimation for a wide range of countries where within-country data of sufficient quality is unavailable. Both within-country and cross-country based estimates are suitable for informing policy decisions. The second step is to apply estimated effect to determine cost per QALY gained. The steps are summarized below, and more detail is given in Supplementary Appendix 1.

Step 1: Identifying the elasticity of health outcomes with respect to expenditure

Given there is no existing estimate of the marginal productivity of the Saudi Arabian healthcare system, the next best alternative requires identifying existing estimates from elsewhere that could be used to inform an estimate of cost per unit of health produced by the Saudi Arabian healthcare system. For the first step, existing published elasticities of the effect of expenditure on health outcomes are taken from the totality of the relevant literature to the Saudi Arabian context. These come from within-country analysis from the UK Andrews et al. (2017) and Lomas et al. (2019)^43,44 and cross-country analysis (Ochalek and Lomas 2020, based on Bokhari et al. 2007)^41,42. These UK studies represent alternative approaches to identification of the effect of healthcare expenditure on mortality⁴⁶. As there is no estimate available from cross-country analysis for Saudi Arabia we apply the average elasticity from high-income countries. Table 1 provides estimates of the elasticities of the effect of expenditure on mortality.

Table 1. Elasticities of the effect of expenditure on health outcomes.

Elasticity source	Source type	Mortality elasticity(ies)
Andrews et al. (2017)	Within-country estimation for UK NHS	All population −0.705
Lomas et al. (2019)	Within-country estimation for UK NHS	All population −1.0278
Ochalek and Lomas (2020)	Cross-country estimation for high-income countries	Under-5 − 0.3549 Adult male −0.2 Adult female −0.1944

Step 2: Converting elasticities to absolute health effects

In the second step, these elasticities are applied to data on health expenditure, demographic characteristics, and burden of ill health for Saudi Arabia using 2018 data. Each of these studies provides estimates of the effect of expenditure on mortality, therefore requiring calculating deaths averted, survival effects and morbidity effects in order to obtain a cost per unit of health.

Deaths averted

We calculate deaths averted from the mortality elasticities from Andrews et al. (2017) and Lomas et al. (2019)^43,44 by applying these elasticities to estimated deaths among Saudi citizens in Saudi Arabia based on mortality data from the Household Health Survey 2018, General Authority for Statistics.ⁱⁱ This provides estimates of deaths averted among the total population.

Estimated deaths based on the average elasticities among high-income countries in Ochalek and Lomas (2020)⁴¹ are calculated by applying each elasticity to mortality data for the relevant population. The elasticity on under-5 mortality is applied to deaths among children under-5 to determine deaths averted in this group by a change in expenditure. The elasticity on adult male deaths to deaths among males and the elasticity on adult female deaths to deaths among females to determine deaths averted among these groups by a change in expenditure. Summing these gives deaths averted among the Saudi Arabians aged 0–4 and 15–60.

Survival effects

In order to determine the survival effects of a change in expenditure, we apply conditional life expectancy to deaths by 5-year age category. The General Authority for Statistics provides life expectancy at birth². The Institute for Health Metrics and Evaluation (IHME) Global Burden of Disease (GBD) database provides estimates of conditional life expectancy⁴⁷. Given the life expectancy at birth from this source matches that from the General Authority for Statistics we opt to use this more granular data for better population coverage. This results in estimates of survival effects for the total population based on the Andrews et al. (2017) and Lomas et al. (2019) elasticities and for the population age 0–4 and 15–60 from the average elasticities from Ochalek and Lomas (2020)^41,43,44.

In order to determine the survival effects of a change in expenditure across the remaining (age 5–14 and 61+) population the average elasticities from Ochalek and Lomas (2020) requires additional steps. Following methods outlined by Ochalek et al. (2018) we first calculate the survival burden of disease among the population age 0–4 and 15–60, which is mortality multiplied by conditional life expectancy for each 5-year age group within the 0–4 and 15–60 ranges summed. We then calculate the survival burden of disease among the total population. Finally, we calculate the proportion of the survival burden of disease among the population age 0–4 and 15–60 estimated to be alleviated, and apply this proportion to the survival burden of disease among the total population. This assumes that the effect of a change in expenditure on survival does not differ markedly between the population in the age ranges 0–4 and 15–60 and the population in the age ranges 5–14 and 61+.

Morbidity effects

We expect that an increase in expenditure will both reduce and increase morbidity. Increases in morbidity are expected through improvements in survival, while decreases in morbidity are expected as a direct result of health improvement through increased expenditure. We use national morbidity estimates for Saudi Arabia from the IHME GBD database, which bases its estimates for Saudi Arabia on, for example, national survey series’ such as the Gallup Poll that is conducted among 1,000 randomly selected Saudi Arabians ages 15+ once per year, disease-specific studies, sources shared by in-country collaborators, and surveys identified in major multinational survey data catalogues, as well as through country Ministry of Health and Central Statistical Office websites⁴⁸. We assume that the same proportion of the morbidity burden of disease among Saudi Arabian citizens is averted as survival burden of disease alleviated by a change in expenditure, and this accounts for the direct effect. To account for the indirect effect, we apply the per capita morbidity burden of disease to the calculated survival effects.

Health effects

The overall health effects of a change in expenditure are calculated by summing the survival and morbidity effects. Cost per unit of health is then calculated based on a 1% change in expenditure. Government expenditure on health includes expenditures falling under the umbrella of the Saudi Health Council, and all relevant sectors, including, but not limited to the Ministry of Health, medical services at the Ministry of Defence, the National Guard, and the Ministry of Interiors, university hospitals (Ministry of Education), King Faisal Specialist Hospital & Research Center (Riyadh & Jeddah), etc^49,50.

Table 2 lists the data sources used to calculate cost per unit of health from the estimated elasticities. More details are available in Supplementary Appendix 2.

Table 2. Population, demography, epidemiology and expenditure input data for Saudi Arabia.

Data	Source
Population	General Authority for Statistics^2
Mortality rate	Household Health Survey 2018, General Authority for Statistics^2
Absolute deaths
Conditional life expectancy	IHME Global Burden of Disease database^47
Survival	Calculated
Morbidity	IHME Global Burden of Disease database^47
Government health expenditure	Saudi Health Council^49^,^50

Results

Table 3 provides the mortality burden of disease, morbidity burden of disease and quality-adjusted life years (QALYs) estimated to be gained by a 1% change in government health expenditure using the elasticities for the UK from Andrews et al. (2017) and Lomas et al. (2019), the average elasticities in high-income countries from Ochalek and Lomas (2020), and the extrapolation from Woods et al. (2016)^39,41,43,44. As anticipated given the greater magnitude of the elasticity from Lomas et al. (2019), this source results in the greatest estimated QALYs gained when comparing the estimates based on elasticities. Conversely, given the relatively smaller elasticities from the cross-country analysis, the estimated QALYs gained from this source are the lowest. The implied QALYs gained from Woods et al. (2016) is the highest among all of the approaches followed and is based on an income elasticity of the value of health equal to 1.

Table 3. Estimates of the survival burden of disease, the morbidity burden of disease, and overall ill health averted for a 1% change in government health expenditure.

Source	Basis of estimate	Survival burden of disease	Morbidity burden of disease	Overall ill health burden of disease
Andrews et al. (2017)	UK within-country elasticity, year 2005	8,861	7,912	16,774
Lomas et al. (2019)	UK within-country elasticity, year 2012	12,918	11,535	24,454
Ochalek and Lomas (2020)	Cross-country elasticity, year 2000	2,869	2,562	5,430
Woods et al. (2016)	Extrapolation of UK estimate, year 2008, using data on the income elasticity of the value of health			27,997

Table 4 provides the estimated cost per QALY gained based on each source. These are calculated using a 1% change in government health expenditure, which is equal to 1% of SAR 118 billion SAR^49,50. Dividing the 1% change in government health expenditure by the estimated QALYs gained for each of the methods given in Table 3 gives the cost per QALY gained. The resulting range is SAR 42,046 (48% of GDP per capita) to SAR 216,770 (247% of GDP per capita). The average elasticity among high-income countries from Ochalek and Lomas (2020) results in the highest estimate while the estimate from Woods et al. (2016) is the lowest. Using elasticities from the UK healthcare system results in estimates of SAR 70,019 and 48,138 (80 and 55% of GDP per capita).

Table 4. Estimates of cost per DALY averted.

Source	Basis of estimate	Cost per QALY gained (SAR 2018)	Percent of GDP per capita (%)
Andrews et al. (2017)	UK within-country elasticity, year 2005	70,179	80
Lomas et al. (2019)	UK within-country elasticity, year 2012	48,138	55
Ochalek and Lomas (2020)	Cross-country elasticity, year 2000	216,770	247
Woods et al. (2016)	Extrapolation of UK estimate, year 2008, using data on the income elasticity of the value of health	42,046	48

Table 5 provides potential central estimates of cost per QALYs gained, which are calculated from the average of estimated health gains based on each source, a form of model averaging as an approach to deal with model uncertainty⁵¹. Calculating a central estimate based on all four sources results in a value of SAR 63,072 or 72% of GDP per capita, which is based on the average across estimates of QALYs gained for a 1% change of government health expenditure (18,664 QALYs gained). Excluding the cross-country source results in an estimate of 51,015 SAR or 58% of GDP per capita. Excluding the extrapolation source results in an estimate of SAR 75,688 or 86% of GDP per capita. The resulting range of potential central estimates is SAR 50,000–75,000. The Saudi Arabia estimates are broadly in line with estimates for other countries from the literature on the marginal cost per unit of health produced by the healthcare system.

Table 5. Potential central estimates of cost per QALY gained.

Andrews et al. (2017)	Lomas et al. (2019)	Ochalek and Lomas (2020)	Woods et al. (2016)	Cost per QALY gained (SAR 2018)	Percent of GDP per capita (%)
✓	✓	✓	✓	63,072	72
✓	✓		✓	51,015	58
✓	✓	✓		75,688	86

Discussion

In line with the Vision 2030, a national HTA entity is currently being established in Saudi Arabia aiming to provide evidence-based recommendations for a given intervention (e.g. a medicine or technology) and inform policy decision making. The foreseeable impacts of implementing HTA which mainly focuses on evidence and value-based approaches in Saudi Arabia include: (1) facilitating improved and equitable quality of care with reduced disparities across populations; (2) providing optimal patient outcomes and enhanced quality of life by making informed clinical choices; (3) ensuring sustainable healthcare owing to rational utilization of healthcare resources and (4) promoting innovation in health through appropriate use of state-of-the-art technologies^5,52. Though the implementation of HTA is in its infancy in Saudi Arabia, roadmaps and activities involved for successful implementation are rapidly progressing. In the Saudi context of establishing HTA, the estimated cost-effectiveness threshold will be used to judge if an intervention represents sufficient value for money to merit adoption in the Saudi healthcare system thus assisting in informed resource allocation decisions and improving overall population health.

This paper brings together evidence to inform the marginal cost per unit of health produced by the healthcare system for Saudi Arabia, which can underpin the cost-effectiveness threshold. Depending upon the specific method chosen, there is a range of estimates of SAR 42,046–216,770 per QALY gained (48–247% of GDP per capita). This is in line with other estimates of this kind from around the world. Three central estimates are proposed based on different inclusion strategies, which give a range of SAR 50,000–75,000.

An estimate of the marginal cost per unit of health produced by the healthcare system enables the calculation of the health that would be expected to be gained across the population, net of any health forgone. By informing health opportunity costs, it can be used as a crucial element to inform funding decisions in a way that is explicit and transparent. The concept of health opportunity costs underpins the guidance of a number of HTA agencies, but existing cost-effectiveness thresholds do not all reflect existing empirical evidence, including NICE in the UK^iii,53. Agencies in Norway and Canada both found themselves in a similar situation to Saudi Arabia and used existing evidence to inform a cost-effectiveness threshold while developing plans for the estimation of a bespoke within-country estimate. The Norheim Commission recommended the adoption of the estimate for Norway from Woods et al. (2016)²⁸, which extrapolates the UK within-country estimate using income elasticity of the value health, while Canada initially proposed an estimate based on Ochalek et al. (2020b), which uses estimates of the effect of healthcare expenditure on health outcomes from cross-country and UK analyses applied to Canadian data on health expenditure, epidemiology and demography, in addition to Woods et al. (2016)^iv,27.

According to the current pharmaceutical pricing policy in Saudi Arabia, prices should take into account the therapeutic value of the new technology and evidence from pharmacoeconomic evaluations, and should be benchmarked according to the price of other similar technologies that are already registered in Saudi Arabia, ex-factory, wholesale and retail prices in the country of origin, prices from official pricing certificate, and the price recommended by the pharmaceutical company⁶. Such benchmarking does not account for the opportunity cost, in terms of health forgone, to the Saudi Arabian population of funding the provision of a new technology whether through government funding or funds raised via premiums on insurance. Used in conjunction with an estimate of the marginal productivity of the healthcare system, pharmacoeconomic evaluations providing evidence on the costs and benefits of a new health technology, can inform calculations of the opportunity cost of funding it, which may be used to inform decisions around whether or not to fund the new technology such that they improve population health outcomes overall. Indeed, this approach is not limited to pharmaceuticals and can be applied to healthcare interventions more generally to ensure that new technologies adopted improve population health.

Among countries that use explicit cost-effectiveness thresholds in HTA, some apply a single value to judge cost-effectiveness while others cite a range for decision-making^15,16. The estimates provided in this paper are suitable to inform calculations of the expected health opportunity cost of reimbursing a new health technology in Saudi Arabia and provide a sound base for proposing a cost-effectiveness threshold which may be applied to the results of pharmacoeconomic studies using QALYs or DALYs as an outcome measure (as done in e.g. Japan, Philippines, etc.)^54,55. Recent evidence supports the interchangeability of cost per QALY and cost per DALY thresholds for use in informing resource allocation decisions^56,57. The data that was available to inform morbidity in Saudi Arabia was based on weights from the Institute for Health Metrics and Evaluation (IHME) international Global Burden of Disease Study. Ongoing collection of HRQoL data in Saudi Arabia may facilitate future estimation of a marginal cost per unit of health produced by the healthcare system informed by national utility weights. The choice of which central estimate to adopt depends upon judgements about the strength and transferability of the alternative sources of evidence. The use of such a cost-effectiveness threshold will help to ensure that HTA processes properly account for health opportunity costs, which can improve the efficiency of health care resources. Of course, the impact on population health is just one consideration, but we would argue that it is an important one¹⁸. Nevertheless, we do not advocate that a cost-effectiveness threshold is used to create a binding decision rule.

There are other reasons why a cost-effectiveness threshold would inform decision-making, but not serve as a binding decision rule. There may be other objectives of interest, beyond population health, and health benefits may be valued differently depending upon the context under consideration. For example, health benefits for patients suffering rare or severe diseases may be given greater weight than health benefits for the average member of the general population. In principle, the estimation of health opportunity costs allows for this to be done in a coherent manner with different weightings being given to health benefits and health opportunity costs depending upon these characteristics.

In policy terms, however, one approach to such equity concerns has been to develop different thresholds for different disease areas. While there may be justifications for such an approach, it cannot be justified on the grounds of opportunity costs when resources are obtained from the same budget with the same marginal cost per unit of health produced by the health care system. This approach is sometimes used to promote equity objectives, but may be flawed on the grounds that the characteristics of individuals that bear the greatest share of the health opportunity cost are generally not known, with the proportion of these individuals that have a rare or severe disease also being unknown, and therefore these health effects are typically unweighted⁵⁸. In principle, these challenges can be overcome by considering weighting of outcomes once opportunity costs have been accounted for. In practice, cost-effectiveness thresholds conflate these issues, which is a limitation that needs to be acknowledged and deserves careful consideration. Of course, where resources are obtained from altogether different budgets for health care the health opportunity costs could be expected to differ, but it is not always practically possible to provide separate estimates and the use of different cost-effectiveness thresholds could itself be problematic.

Existing estimates of the marginal cost per unit of health produced by the healthcare system come primarily from healthcare systems that are largely publicly funded²⁵. However, the use of cost-effectiveness thresholds is not limited to public sector decision-making. For example, ICER in the US is increasingly cited by private healthcare decision makers in the US. The conceptual foundation of its cost-effectiveness threshold is that of health opportunity cost²⁰. Their range of USD 100,000 to 150,000 per QALY is supported by the study by Vanness et al. (2021), which analyses the effect of insurance premiums on short-term mortality and morbidity in the US, a primarily privately funded healthcare system, and finds a central estimate of approximately USD 100,000 per QALY (160% of GDP per capita)⁵⁹. Estimates for publicly funded healthcare systems tend to be lower, with the UK, Australia and Canada estimates being around 50% of GDP per capita while the Spain estimate is close to 1× GDP per capita. Our estimated range, 58–86% of GDP per capita, falls between these.

Like any study, ours has limitations. The biggest limitation is that we have relied upon extrapolating existing evidence on the marginal productivity of healthcare expenditure from outside of Saudi Arabia. Each of the approaches is based on cross-country evidence, taking an average of the elasticities estimated for 33 other high-income countries (HICs), or based on within-country evidence from the UK. Ideally, a within-country study would be conducted using data from the Kingdom in the future to address these concerns, but this will be a resource intensive process in itself. Our study relies on the transferability of the elasticities used to the Saudi Arabian context. In the absence of evidence to inform a bespoke elasticity of the effect of expenditure on health outcomes in a healthcare system such as Saudi Arabia’s, we rely on cross-country evidence and evidence from the UK. The reliability of the cross-country evidence rests on the econometric specification developed in Bokhari et al. (2007) that seeks to identify variation in health care expenditure that is independent of health need using an instrumental variable approach. No estimate is produced for Saudi Arabia in this study, which is why the average elasticity estimate across HICs was used. Given the limited variation in the elasticities estimated for HICs, it seems reasonable to suggest that an estimate for Saudi Arabia would have been similar. Using evidence from the UK relies on the validity of the econometric models in addition to the generalizability of such evidence to the Saudi Arabian context.^v In particular, there is uncertainty in how reliably the UK evidence, generated in the context of a health care system that is almost wholly public, can be extrapolated to the mixed public and private health care system of Saudi Arabia.

Given the uncertainty surrounding our estimates, it is worth reflecting upon the health consequences of setting a cost-effectiveness threshold that is either too high or too low to accurately reflect health opportunity costs. A simple framework for considering this question in the context of drug development is outlined in Pandey et al. (2018)⁶⁰. They consider two stakeholders: producers (the pharmaceutical industry) and consumers (population health). Assuming that producers always price such that their product is at the cost-effectiveness threshold, producers will always prefer a higher cost-effectiveness threshold since it means that they can charge higher prices and it also means that a greater range of products will be profitable for them to develop. The population health perspective is more complicated as there is a trade-off when lowering the cost-effectiveness threshold between achieving net health benefits from products and reducing the number of products that are profitable to develop. In summary, over-estimation of the marginal cost per unit of health produced by the healthcare system will benefit producers and cost population health, while the effect of under-estimation will cost producers and benefit population health (although the relationship between the size of the benefit and the scale of the under-estimation is non-monotonic: it increases and then it falls). Consideration of this framework also poses the question of whether setting the cost-effectiveness threshold equal to the marginal cost per unit of health produced is optimal in terms of incentives for pharmaceutical innovation. This question is beyond the scope of this paper and is an active area of research. It is argued that the marginal cost per unit of health produced is at least a key piece of evidence required for creating conditions for dynamic efficiency, while others claim it appropriate in itself^21,61,62.

We do not expect an estimate of the marginal cost per unit of health produced by the healthcare system to be static⁶³. It is likely that it will evolve over time in a way that is complicated and depends upon a number of factors. Paulden et al. (2017) outline three key determinants of change: (1) The overall quantity of resources available for health care, (2) The efficiency of health care in producing health outcomes, and (3) Changes in the patterns of demand for different types of health care⁶⁴. As part of ongoing reforms in Saudi Arabia, it might be expected that more resources will be made available for health care over time, which would be expected to increase the marginal cost per unit of health produced by the healthcare system in combination with general price inflation within healthcare. However, this would hopefully be offset by improvements in efficiency, which would have the opposite effect. The effect of changes to demand are ambiguous. Thus, the evolution of marginal cost per unit of health produced by the healthcare system is an empirical question and thought needs to be given to the regularity with which estimates ought to be updated.

Where an estimate of the marginal cost per unit of health produced by the healthcare system is significantly lower than current benchmarks, there may be a concern from policymakers that implementing a decision-making threshold based on an estimate of the marginal cost per unit of health produced by the healthcare system could result in major shocks to the reimbursement system, affecting patient access to valuable technologies. To alleviate this concern, policymakers may implement a transition period acknowledging that there are costs (in terms of population health) to not implementing it in full immediately. Implementing a decision-making threshold based on the marginal cost per unit of health produced by the healthcare system ultimately ensures the transparency and predictability of decision-making.

There may be a temptation to apply a threshold representing a demand-side approach, as these are typically higher than estimated based on a supply-side concept¹⁵. In comparison to a recent demand-side study estimating the consumption value of health⁶⁵ roughly equal to 1–1.5 times the GDP per capita of Saudi Arabia, the estimates of the (supply-side) marginal cost per unit of health produced by the healthcare system for Saudi Arabia from this paper are slightly lower. This is a familiar result from the empirical literature that has looked into comparisons of these two separate but related quantities^30,66. However, the expected range of estimates based on a demand-side concept is wide. In a review of demand-side estimates representing the societal monetary value of health gain, Gloria et al. (2021) find median estimates from studies using a willingness to pay approach to estimation to be roughly half of GDP per capita while median estimates using a value of a statistical life approach to be nearly 8x GDP per capita⁶⁷. Consideration of the consumption value of health is also appealing in that different values can be ascribed to different types of health. For example, assumptions about risk aversion within the Generalized Risk-Adjusted Cost-Effectiveness framework provide a justification for greater consumption value for health gains among people with more severe diseases⁶⁸. The “demand-side” perspective does not account for resource constraints, and so applying a demand-side estimate as a decision-making threshold risks decisions that result in a reduction in overall population health.

Conclusion

Existing evidence of the effect of expenditure on health outcomes to Saudi Arabia reveals that a cost-effectiveness threshold representing health opportunity costs would lie in the region of SAR 50,000–75,000 per QALY gained. Our estimates provide a placeholder, based on the best currently available evidence, to guide decision-making in Saudi Arabia while further evidence is developed.

Transparency

Peer reviewer disclosures

A reviewer on this manuscript has disclosed that they are an employee of PRECISIONheor a consulting firm that provides health economics and outcomes research services to life sciences companies. They also hold equity in the parent company Precision Medicine Group. The other peer reviewers on this manuscript have no other relevant financial relationships or otherwise to disclose.

Acknowledgements

None stated.

Disclosure statement of financial/other relationships

JO, IQVIA, individual consultant. This work was carried out as private consultancy by JO with input from JL. Although JO and JL are otherwise employed by the University of York, the work and recommendations produced in this paper do not imply an endorsement or commitment from the University of York.

RFB has received honoraria for speaking in the last year for Novartis and Medtronic.

The other authors have no potential conflicts of interest.

Correction Statement

This article was originally published with errors, which have now been corrected in the online version. Please see Correction (http://dx.doi.org/10.1080/13696998.2024.2334567)

Additional information

Funding

The article received funding from the Ministry of Health – Kingdom of Saudi Arabia.

Notes

i The GDP per capita of Saudi Arabia is less than that of the UK and so a higher income elasticity of the value of health would result in a marginal cost per unit of health that is a smaller proportion of GDP per capita following this approach.

ii Not publicly available

iii UK empirical estimates are, however, used by the UK government in assessing policies⁶⁹, were proposed as an appropriate cost-effectiveness threshold for new vaccines that was rejected following opposition (primarily) from the pharmaceutical industry⁷⁰, and are thought to have indirectly influenced the recently negotiated cost-effectiveness threshold for NICE⁷¹.

iv Opposition (primarily from pharmaceutical companies) to the proposed threshold prompted PMPRB to revise upward their proposed cost-effectiveness threshold basing it instead on other existing thresholds as opposed to empirical evidence⁷².

v The econometric estimates from the UK are themselves subject to uncertainty. Taking the upper (magnitude) confidence interval bound from the Lomas estimate -1.37 and the lower (magnitude) confidence interval from the Andrews estimate -0.225 would result in a range of cost per QALY gained of SAR 36,114 – 220,089.