Costing Interventions for Developing an Essential Package of Health Services: Application of a Rapid Method and Results From Pakistan

Background: The Federal Ministry of National Health Services, Regulations and Coordination (MNHSR&C) in Pakistan has committed to progress towards Universal Health Coverage (UHC) by 2030 by providing an essential package for health services (EPHS). Starting in 2019, the Disease Control Priorities 3 (DCP3) evidence framework was used to guide the development of Pakistan’s EPHS. In this paper, we describe the methods and results of a rapid costing approach used to inform the EPHS design process. Methods: A total of 167 unit costs were calculated through a context-specific, normative, ingredients-based, bottom-up economic costing approach. Costs were constructed by determining resource use from descriptions provided by MNHSR&C and validated by technical experts. Price data from publicly available sources were used. Deterministic univariate sensitivity analyses were carried out. Results: Unit costs ranged from 2019 US$ 0.27 to 2019 US$ 1,478. Interventions in the cancer package of services had the highest average cost (2019 US$ 837) while interventions in the environmental package of services had the lowest (2019 US$ 0.68). Cost drivers varied by platform; the two largest drivers were drug regimens and surgery-related costs. Sensitivity analyses suggest our results are not sensitive to changes in staff salary but are sensitive to changes in medicine pricing. Conclusion: We estimated a large number of context-specific unit costs, over a six-month period, demonstrating a rapid costing method suitable for EPHS design.

• This paper demonstrates a rapid method to calculate context-specific and comparable unit costs which was successfully used in the Essential Package of Health Services design process in Pakistan • We report the results for the first comprehensive dataset of unit costs for 169 health interventions based on localised evidence in Pakistan.
• Our method is transparent and unit cost estimates are disaggregated by input type, highlighting the cost drivers for each intervention.Our method can be replicated for future priority setting exercises in Pakistan as well as used in other contexts.

Background
Countries around the world have strengthened their commitment to universal health coverage (UHC) in recent years.UHC has been enshrined in Sustainable Development Goal target 3.8 and calls for access to quality essential healthcare services for all [1].There are many types of health services that countries could potentially deliver, but budgetary constraints force policymakers to limit the number and coverage of interventions financed through public expenditure.In order to set health sector priorities, many countries have embarked on designing or revising essential packages of health services (EPHS).This approach allows for explicit system-wide priority setting within a given budget envelope [2].
Pakistan's commitment to providing a UHC EPHS was stated in its 12 th Five-Year Plan  and National Action Plan (2019-23) for the health sector [3].The Federal Ministry of National Health Services, Regulations and Coordination (MNHSR&C) decided to use the Disease Control Priorities 3 rd edition (DCP3) as a starting point and framework of reference for the process of priority setting of health services provided by the public sector at the district level [4].DCP3 is a multi-year project that sought to synthesise global evidence of cost and cost-effectiveness evidence across disease areas, and provides the evidence base for an essential universal health coverage (EUHC) package composed of 218 interventions as a guide for low-and middle-income countries (LMICs) [5].
In 2019, the MNHSR&C, jointly with the provincial departments of health and key stakeholders, compared the current scope of essential health services offered in Pakistan This paper has two objectives: (i) to demonstrate a rapid costing methodology that can be used in the process of estimating unit costs for HBP design in LMICs, and (ii) to present the first comprehensive dataset of unit costs for health interventions based on localised evidence in Pakistan.These unit costs are the building blocks needed to estimate the package's total cost, the relative budget impact of individual interventions and the affordability of the HBP given available fiscal space, which are key analytical components in the HBP design process.

Methods
The general costing approach presented here was designed during a meeting with national stakeholders convened by the Health Planning, System Strengthening & Information Analysis Unit (HPSIU) of the MNHSR&C in Islamabad in July 2019.
Ideally, cost estimation for a UHC EPHS design would rely on local primary data collection.
However, we used secondary data sources for several reasons.Firstly, costing current service provision would likely reflect service delivery of low quality; as the EPHS aims to deliver high quality services, collecting primary data could lead to cost underestimation.Further, it was estimated that 135 (62%) of the 218 EUHC package interventions were not routinely available in Pakistan's public sector [6].Lastly, primary data collection is a resource-intensive exercise which was not feasible in the project timeframe.
Obtaining unit costs from the published global literature was also considered.However, a review of cost-effectiveness databases found a scarcity of high-quality costing estimates appropriate for Pakistan [7,8].Adapting and transferring cost data across settings can be misleading as resource use (such as lengths of patient consultations, health worker salaries and prices of medications and equipment) varies greatly between countries [9][10][11].Further, variation in the context-specific service configuration can have affect costs and efficiency [12].
Lastly, published data is often not sufficiently disaggregated and costing studies often employ different methodologies leading to evidence of varying levels of quality, a challenge when attempting comparability across many interventions.
Consequently, in consultation with stakeholders, we instead opted to develop a contextspecific, rapid normative method to estimate the cost of DCP3 interventions.The costs were estimated by a joint team from the Aga Khan University and the London School of Hygiene & Tropical Medicine, using data provided by the HPSIU and reviewed and validated by local technical experts.This rapid costing was conducted over a six-month period.

General approach
We carried out an ingredients-based costing and took an economic costing approach, meaning that we accounted for the value of all resources used, regardless of whether financial expenditure was expected.A bottom-up approach to costing using both secondary sources and expert opinion was applied.We assumed a provider's perspective and used a one-year time horizon.Our approached followed the principles set in the Global Health Costing Consortium reference case [13], a gold standard for the costing of health interventions in LMICs.
We costed interventions across all five DCP3 EUHC delivery platforms: community-level, health centre, first-level hospitals, referral hospitals, and population-level.However, the priority setting exercise focused on a district package of services; population-level interventions were therefore excluded from the main analysis as they are operated and implemented at the national level.See Supplementary File 1 for population-level unit cost estimates and further information.
The DCP3 EUHC contains 218 interventions [5].Following a preliminary review carried out by MNHSR&C, as well as consultations with provincial-level stakeholders and within HPSIU, 47 The costing approach comprised several steps as specified in Figure 1 above.: 1) development of a costing template, 2) development of intervention description sheets for each intervention, identification of intervention-specific inputs and validation by technical working groups (TWGs), 3) identification and assessment of price sources and price data extraction, 4) combination of prince and resource use data, and 5) sensitivity analysis.We calculated unit costs by estimating the resources needed and relevant prices per beneficiary per year (e.g., cost per person treated for hypertension over a year).Costs were estimated in Pakistani rupees and converted to 2019 US dollars at an exchange rate of 155 PKR:USD [14].

Costing template
The stakeholders felt that a cost estimation tool for EPHS design needed to show transparency, flexibility and ease of use, and should work across multiple platforms in the health system.As a result, the team designed a semi-automated user-friendly costing Step 0

•Review of methodological approaches with stakehodlers and development of research plan
Step 1 Step 2 Step 3 • Identification and assessment of price sources and price data extraction Step 4 •Multiplication of price data by quantity data to estimate intervention unit costs Step 5 It also granted the flexibility of entering multiple price lists per input.

Determining resource use
To capture all resources used for each individual intervention, the HPSIU prepared intervention-specific description sheets detailing inputs necessary for service delivery: staff requirements (staff type and time), drug regimens, laboratory-based diagnostics, radiology, other supplies and equipment per patient/year.These descriptions were developed based on the latest existing national guidelines (or, in their absence, international guidelines) for each intervention and expert opinion from senior Pakistani clinicians.For hospital-based interventions, the average number of inpatient bed-days, and whether the intervention involved surgery, were also specified by senior clinicians.The clinicians were identified by the primary academic partner Aga Khan University and HPSIU.Feedback from the clinicians was collected using two rounds of 3-day workshops. .The different inputs needed for inpatient bed-days and surgeries were obtained from the published sources [15,16].
The first draft of the intervention description sheets was compiled by the HPSIU and revised and amended by experts during several rounds of disease-specific TWG consultations.Final intervention descriptions were used for mapping individual interventions and compiling a list of resources used and quantities [17].Further details on resource use quantification can be found in Supplementary File 3.
We only accounted for direct resource use in service provision.We did not include any indirect costs, above-service delivery costs or other overheads.We also do not include health system costs such as the cost of governance at the district level.Indirect costs were accounted for and included in a later stage of the design process.See Torres-Rueda et al. for further details [18].

Determining prices
A variety of price sources were available for each input.An assessment of strengths and weaknesses of different price sources was conducted.The quality of sources was assessed using three main criteria shown in Table 1, namely how recently the source had been

Criteria Rationale
Is the price recent?Account for changes in inflation and use real prices Is the source of price a public source?
This costing exercise uses a provider perspective for the public sector.The prices for inputs purchased by the public sector such as medicines and equipment can often differ substantially from private sector prices.We therefore attempted to use the purchase prices for public sector providers Is the source nationally representative?At this stage, the UHC benefit package and its unit costs are a national exercise therefore the price sources used were ranked higher if they were nationally representative Details on price sources assessed and used can be found in Supplementary File 4. In summary, federal-level healthcare worker pay scales were used to determine average staff time pricing per health worker cadre [19].The primary source for medication price data used was the Sindh Health Department Procurement Price list of 2018-19 [20], as it was both recent and listed public sector prices.The first choice for supplies and equipment was a list of procurement prices from the Medical Emergency Resilience Fund 2019-2020 [21].Building prices were obtained from Federal budgets (spaces) [22] and a costing study carried out in Khyber Pakhtunkhwa Province (utilities) [23].A generic cost of furniture was added (10% of the cost of space) [24].We were unable to construct diagnostic and radiology costs through an ingredients-based approach due to time constraints and the complexity of supplies and equipment used.We used the 'Costing and Pricing of Services in Private Hospitals of Lahore: Summary Report' as our primary source for diagnostic prices as it also used an ingredientsbased approach consistent with our methodology [25].Generic prices for surgery and inpatient bed-day were obtained from the same sources as the resource use data [15,16].

Published unit costs and prices
In nine instances, unit costs or the price of the main input of an intervention were obtained from the peer-reviewed literature.These alternative options were used when the unit cost or main price estimate was grossly out of line with available global evidence and no Pakistan-specific reason for the discrepancy could be identified after consultation with health economists knowledgeable of the specific area.See Supplementary File 5.

Sensitivity analysis
We carried out deterministic univariate sensitivity analyses for two key parameters.Staff salaries vary by province in Pakistan but, in our analysis, we used federal pay scale salaries for our base case unit costs (which are used in Islamabad Capital Territory and Baluchistan province).We carried out a sensitivity analysis using pay-scales for Sindh and Khyber Pakhtunkhwa (KP) provinces, estimating percentage changes in average intervention costs per platform.We also examined the sensitivity of unit costs to different medicine prices for the ten most costly interventions.Using the different medicine price sources reviewed, we recalculated unit costs applying the lowest and highest medicine costs available and present those ranges in relation to the base case unit costs.

Ethical Issues/Statement
Approval for this research was obtained from the Research Ethics Committees at the London School of Hygiene & Tropical Medicine and at the Aga Khan University in Karachi.The authors do not have any conflicts of interest.

Results
There is a high variation in the unit costs for interventions, ranging from 2019 US$ 0.27 to 2019 US$ 1,477.76 as shown in Table 2. Interventions with the top five highest unit costs were all delivered in the referral hospital platform: treatment of early-stage childhood cancers    The overall largest cost drivers of unit costs overall were drug regimens and surgery-related costs, but the cost drivers also varied by service delivery platform.Equipment costs were low in all platforms (<1%).See Figure 4.

Discussion
We have presented the first comprehensive set of unit costs of health interventions in Pakistan estimated for the purposes of EPHS design.These unit costs were subsequently used during the priority setting process to calculate total costs, intervention-specific budget impact and to assess package affordability [18].We also demonstrate a rapid method which could be costing method is appropriate for priority setting and is consistent with frameworks for fair decision-making, such as Accountability for Reasonableness (A4R) [26].The dataset produced combines a large number of interventions, considers a comprehensive set of inputs per intervention and presents resource use and price data in a highly disaggregated manner and in a format that is highly accessible.Consequently, the data produced are transparent, relevant to the context and the decision problem faced by decisionmakers and would be accessible during any revision or reversal processes.Further, a lack of sufficient and appropriate evidence is commonly cited as a barrier to decision-making around health investment and disinvestment; we address this gap by providing a broad set of countryspecific cost estimates.
During the planning stages, a number of publicly-available priority setting tools were reviewed, including, the Cost Revenue Analysis Tool Plus [27], the Health Interventions Prioritization (HIP) Tool and the One Health Tool [28].While these tools have important added value for other parts of the priority setting process stakeholders, Pakistan opted for developing semi-automated user-friendly costing templates in Microsoft® Excel for a number of reasons.Firstly, using a commonly available software that did not require extensive training was preferred in order to engage a broader set of stakeholders, both at national and provincial level, leaving the door open for regional adaptations or future package revisions.Secondly, a spreadsheet-based tool provided transparency in inputs and calculations which facilitate external validation of the data.Thirdly, it provided much-needed flexibility that allowed unit costs to be estimated for interventions with multiple service delivery configurations, different delivery platforms, as well as accounting for all intervention-specific fixed resources used.
Excel-based tools have been used for health benefit package design in other LMIC settings [29].
HPSIU developed the service descriptions that served to determine resource use with a highquality service in mind.This allowed for reflection on what constitutes a high-quality service and what aims the health system should strive towards.Although the process of cost calculation was carried out in six months, it was labour intensive.It required considerable input from several clinically trained members of staff at HPSIU (working full-time on this specific task), and, importantly, the aid of a wide range of TWG experts whose input enhanced the accuracy of the service descriptions.For a similar process to be successful elsewhere a number of factors will be required: a high level of technical expertise within the Ministry of Health, the ability to convene a wide range of actors and a high degree of political commitment.This process could also be expedited by having an available inventory of interventions, guidelines and resource needs to form the basis for Excel-based templates that could be adjusted locally.Such an inventory is currently under development at the World Health Organization with the UHC Compendium [30].
Several price sources for different inputs, with different strengths and weaknesses, were identified.This prompted reflection on what constitutes desirable attributes of a price source when estimating costs for HBP design.With the input of all stakeholders, we settled on three main criteria: (i) using recent prices, important in settings like Pakistan with high price fluctuations, (ii) using prices of purchase from the public sector, as the exercise was framed around public provision of services, and (iii) using prices reflective of the entire country, a difficult task in a highly heterogeneous setting as Pakistan.Although we found it helpful to keep these criteria in mind, we did not come to a resolution on how these three criteria should be traded off when one source did not possess all three attributes.More work needs to be done to develop a validated rapid process for selecting local prices to cost health benefit package interventions, as well as potentially finding adjustment factors between price sources.Further, estimates such as ours should be updated periodically to account for changes in prices due to medications going off patent, or reflecting a decrease in pricing due to government bulk purchasing or manufacturer rebates [31].
It is key to draw a distinction between priority setting and budgeting; our costs should not be directly translated into budgets without further adjustments.Our estimates will underestimate future financial expenditures in four ways.Firstly, our costs do not include indirect costs, above-service delivery costs or other overheads.Secondly, the approach does not capture health system inefficiencies or wastage.Thirdly, these estimates do not capture the costs incurred in carrying out initial consultations or diagnostics with patients whose conditions are not covered by the EPHS.Lastly, we calculated economic costs instead of financial costs.As a result, increased expenditure needed to purchase fixed resources at scale (e.g., purchasing a dental chair for every health centre) will not be reflected.This highlights the importance, within the priority setting process, of thinking through issues of implementation, such as the bundling of interventions, which would enable more efficient use of common inputs.These under-estimations could be corrected by applying a mark-up, either a generic health system-wide one [32], or a context-specific one based on empirical evidence from primary costings, and by presenting economic and financial costs separately.Once the composition of the package has been agreed, further work will be needed to better calculate incremental financial expenditure.Such exercises have been carried out elsewhere in the South Asian subcontinent and may be helpful in Pakistan [33].
On the other hand, our unit costs may overestimate future provider expenditure.Firstly, certain goods presently procured as donations from international donors are accounted for even if no financial costs are incurred by the public sector.Secondly, the service delivery descriptions capture a delivery of high quality.Therefore, compared to current provision, some of our input costs may be higher than those of current interventions in order to compensate for both coverage and quality gaps.However, higher upfront costs for higher quality may paradoxically lead to lower overall costs; for instance, the United States National Academy of Medicine's Crossing the Global Quality Chasm estimates an annual cost of $1.4-1.6 trillion in lost productivity due to poor-quality care [34].Nonetheless, correcting these overestimations may require reviewing costs as eligibility for donor funding schemes changes and as service delivery practices and guidelines are updated.
Our unit costs were not highly sensitive to variations in staff salaries and are therefore, in that respect, likely to be generalisable across provinces.Some of the more costly interventions were highly sensitive to changes in medicine costs, particularly, early-stage cancer treatment and treatment for drug-resistant tuberculosis.Further work to understand province-specific medication procurement sources is important to produce accurate estimates.
Our cost estimates were used to feed into the priority setting process, namely to understand budget impact of interventions, as well as to assess affordability within a budget constraint.
Stakeholders reviewed these data, along with data on other decision criteria, such as costeffectiveness and avoidable burden of disease, to make decisions on intervention inclusion and exclusion.This process, including the values that may have been prioritised and the trade-offs made, has been described in detail elsewhere [18].

Limitations
Our study is subject to a number of limitations.Firstly, while we attempted to develop a method that could be used across all HBP cost calculations, we were not able to apply it consistently across inputs and intervention types.Our methods allowed for a rapid ingredients-based estimation of resource use.However, we found it difficult to apply this method when the inputs costed involved large numbers of components (generic inpatient bedday and surgery costs) or complex diagnostic pathways and therefore had to rely on external estimates and prices from a study conducted in a large hospital in an urban setting, which may not be representative of resource use requirements in other areas of the country.Further, we used certain ballpark assumptions (e.g., assuming the cost of furniture was 10% of building costs) which need to be further refined or tested.Secondly, we used expert opinion to fill gaps in certain inputs (e.g., determining the length of time of an average consultation for a specific intervention).While we have explained why this method was preferable, and more feasible, than primary data collection, we acknowledge that eliciting information in such a way may introduce bias to our estimates.Thirdly, our analysis shows that average cost drivers vary by platform.However, the averages here presented are unweighted (in other words, they don't reflect the frequency of each intervention).Therefore, the average proportion of inputs per platform presented in this analysis is not reflective of the actual proportion of inputs that that will be needed for implementation per platform.Such calculations require combining unit costs with detailed geospatial data on effective coverage cascades detailing service need, use and quality, which was outside the scope of this analysis.Lastly, we were not able to adjust for within-country variation of resource use and prices (beyond our sensitivity analysis on staff salaries).Future exercises may want to use econometric or other methods to present a range of sub-national unit costs that better capture heterogeneity [35].

Conclusions
We estimated 167 unit costs for the EPHS design process in Pakistan.We developed methods that allowed us to rapidly produce a large dataset of unit costs which were both contextspecific and comparable.Further refinements to our methods are needed to better estimate costs of diagnostics, inpatient bed-days and generic surgeries for future application in Pakistan and for similar projects in other settings.A global inventory of interventions and their resource needs would greatly enhance efforts by countries wishing to develop their own costing tools in Excel, rather than use programmed global ones.We have demonstrated that it is possible to estimate local costs, with expert validation and local acceptance, in rapid timeframes, rather than rely on extrapolated estimates from the limited global available of costs from other settings.
against the services covered by the EUHC.They recommended that a subset of EUHC interventions should be assessed for inclusion in Pakistan's UHC-EPHS [6].This paper reports INTERNATIONAL JOURNAL OF HEALTH POLICY AND MANAGEMENT (IJHPM) ONLINE ISSN: 2322-5939 JOURNAL HOMEPAGE: HTTPS://WWW.IJHPM.COM 5 on the costing of this subset of DCP3 interventions as part of a broader process of UHC-EPHS design in Pakistan.

7 Figure 1 .
Figure 1.Steps for costing EPHS in Pakistan

(
2019 US$ 1,477.76),treatment of early-stage breast cancer (2019 US$ 1,304.04),management of refractory illness (2019 US$ 673.43), repair of cleft lip and cleft palate (2019 US$ 515.11), and specialized tuberculosis services, including management of drug resistant tuberculosis (2019 US$ 493.21).Interventions with the five lowest unit costs were all delivered through the community-based platform: screening of hypertensive disorders in pregnancy (2019 US$ 0.27), providing guidance on early symptoms during emerging infectious outbreaks (2019 US$ 0.28), adolescent-friendly health service provision (2019 US$ 0.33), antenatal and post-partum education (2019 US$ 0.34) and larviciding and water management in high malaria transmission settings (2019 US$ 0.41).

Figure 4 .
Figure 4. Cost drivers by platform

Figure 5 .Figure 6 .
Figure 5. Tornado diagram of provincial staff salaries by platform •Development of intervention description sheets for each intervention •Identification of intervention-specific inputs • Validation of intervention descriptions by technical working groups (TWGs) •Conduct sensitivity analysis template in Microsoft® Excel.The template separated resource use data and prices and divided costs by input category.It allowed calculations of multiple unit costs per intervention (e.g., interventions carried out in multiple platforms or by multiple types of health workers).

Table 2 .
Unit costs of DCP3 interventions in Pakistan by platform and by cluster (2019 US$).Abbreviations: Reproductive, maternal, new-born, child adolescent health/age related (RMNCH); NCIP (NCIP) health C19 Vision pre-screening by teachers; vision tests and provision of ready-made glasses on-site by eye specialists $1.84 schistosomiasis, soil-transmitted helminthiases and trachoma, and food borne trematode infections (Also included in NTDs package of services) $1.76 Community RMNCH Adolescent health C23 Adolescent-friendly health services including provision of condoms to prevent STIs; provision of reversible contraception; treatment of injury in general and abuse in particular; and screening and treatment for STIs $0.33 implemented in other countries where context-specific and comparable unit costs are required for EPHS design.In this discussion, we reflect on the process and outcome.