STEPS FETP Scale Up Decision Aid

1. Introduction and purpose of the STEPS tool

The STEPS FETP Scale Up Decision Aid is an interactive decision support tool that combines preference evidence, cost data and outbreak impact assumptions to help policy makers design and assess Field Epidemiology Training Programme (FETP) scale up scenarios in India.

The tool is built around three linked components. First, it uses a discrete choice experiment (DCE) and mixed logit model to quantify how key stakeholder groups value different FETP programme designs, including training tier, delivery mode, mentoring intensity and qualification. Second, it applies detailed cost templates derived from World Health Organization (WHO), National Institute of Epidemiology (NIE) and National Centre for Disease Control (NCDC) data to translate an assumed monthly cost per trainee into a full cost envelope for each tier. Third, it maps training capacity into outbreak equivalents averted and monetary benefits, using a literature based calibration for the economic value of major outbreaks and explicit sensitivity ranges.

The STEPS tool is scenario based. Users specify a combination of programme design attributes and cost, along with assumptions about number of cohorts, planning horizon and whether opportunity costs are included. The tool then produces endorsement probabilities, implied willingness to pay (WTP), epidemiological benefits, benefit cost ratios and net benefits. All outputs are framed as support for strategic planning and do not constitute forecasts.

This technical appendix documents how each part of the tool works. It explains the preference model, programme attributes, cost construction, benefit formulas, sensitivity analysis and the Copilot interpretation tab. The aim is that a technically trained policy analyst can understand, reproduce and defend the calculations without needing to inspect the underlying code.

2. Overview of the preference study and mixed logit model

2.1 Discrete choice experiment and design

Preferences for FETP scale up options were measured using a discrete choice experiment (DCE) in which relevant stakeholders evaluated hypothetical FETP programmes that differed along key design attributes. Each choice task asked respondents to choose between two alternative FETP programme designs and an opt out option in which no additional FETP training would be funded.

The experimental design varied the following attributes:

The attribute levels were selected through expert consultation and piloting to span realistic policy relevant FETP designs while avoiding implausible combinations. Cost levels were selected to support robust estimation of marginal WTP for improvements in mentoring, delivery, qualification and response time.

2.2 Mixed logit model specification

Choice data from the DCE were analysed using a mixed logit model that allows for preference heterogeneity across respondents. For use in the STEPS tool, the mean preference estimates from the mixed logit model are used as fixed coefficients. These coefficients define the utility of a programme option and the opt out option as linear functions of the attributes.

For a given programme option, the indirect utility is written as:

The utility of the opt out option, in which no new FETP funding is provided, is specified as a single constant:

2.3 Coefficient values used in the tool

The STEPS tool uses the following mean coefficients from the mixed logit model (cost in thousands of Rupees per trainee per month):

2.4 Choice probabilities and endorsement

For a given programme design, the STEPS tool converts utilities into probabilities of choosing the programme versus the opt out option using the standard logit formulation:

In the STEPS outputs this probability is reported as the endorsement rate for each scenario. It is also used as an adjustment factor when translating training capacity into outbreak equivalents averted, so that designs that stakeholders are unlikely to support yield smaller effective benefits.

3. Program design attributes and fixed response capacity logic

3.1 Core programme attributes in STEPS

Within the STEPS interface, each scenario is defined by a set of programme design attributes. These align with the attributes used in the DCE but are organised into intuitive input controls on the Configuration tab. The main attributes are summarised below.

3.2 Fixed response capacity at 7 days

In the DCE, response time had three levels (30 days, 15 days and 7 days). Stakeholders strongly preferred shorter response times, and the mixed logit results show positive coefficients for 15 days and 7 days relative to 30 days. In the STEPS tool, response capacity is fixed at 7 days for all scenarios for four reasons.

In the mixed logit calculations, each scenario is evaluated at the 7 day level. This means the 7 day coefficient β_resp,7 is always included in the non cost utility for every programme scenario evaluated in the STEPS tool.

4. Cost construction methodology

4.1 Data sources and cost structure

The cost structure in the STEPS tool is based on detailed programme budget data for existing FETP streams, compiled with support from WHO, NIE and NCDC. Costs are grouped into three tiers that correspond to frontline, intermediate and advanced training programmes. Within each tier, costs are further classified into salary and benefits, equipment and supplies, facilities, trainee support, training materials, workshops and seminars, travel, other direct costs and indirect costs inclusive of opportunity costs.

For each tier, total costs are available at programme level, along with the cohort size. This allows the tool to compute cost shares for each cost category and cost per trainee for a given cohort. These cost shares are then used as templates that are scaled up or down according to the monthly cost per trainee chosen by the user.

4.2 From monthly cost per trainee to programme cost

On the Configuration tab, the user selects a monthly cost per trainee in Rupees for the programme. This cost is assumed to apply for every month that the programme runs and for every trainee in the cohort. The tool first converts this monthly figure into a cost per trainee for the full duration of the programme and then into a cost per cohort.

In the current configuration, the default durations are 24 months for Advanced, 12 months for Intermediate and 3 months for Frontline programmes. The default cohort sizes used in the national templates are 40 trainees for Advanced, 80 for Intermediate and 35 for Frontline, consistent with the cost template data.

4.3 Linking the cost slider to the DCE

The monthly cost per trainee slider is constrained to values that match the range of cost levels used in the DCE. This ensures that all computations based on the mixed logit model remain within the empirical support of the data and avoids extrapolating preferences far beyond the range used in the experiment. The slider minimum and maximum for each tier correspond to the lowest and highest cost levels that were shown in the choice tasks, adjusted slightly if needed for user friendliness.

Internally, the same cost value is used for both the preference calculations (utility and WTP) and the cost template scaling. For the economic results, an additional step determines whether indirect costs and opportunity costs are included in the total economic cost. This logic is described in section 7.

5. Cost component shares and scaling logic

5.1 Cost templates by tier

Table 3 summarises the cost templates used for the three FETP tiers. These are drawn from WHO, NIE and NCDC financial data and represent total programme costs across all participating institutions for a single typical cohort at each level.

5.2 Calculation of cost shares

For each tier and cost line, the STEPS tool computes a cost share equal to the line item cost divided by the grand total for that tier. For example, for Advanced programmes the in country travel share is:

Similar calculations show that indirect and opportunity costs account for about 52 percent of Frontline programme costs (9,089,142 out of 17,413,793), about 26 percent of Intermediate costs (93,200,454 out of 361,096,904) and about 23 percent of Advanced costs (86,711,010 out of 375,747,710). These differences reflect variation in delivery models and the intensity of supervision and travel at each level.

5.3 Scaling logic for user selected costs

When the user selects a monthly cost per trainee for a given tier, the STEPS tool first computes the total programme cost per cohort as in Equation 4. It then allocates this total cost across cost lines using the shares derived from the template for that tier. In other words, the template shares define the internal composition of the cost envelope, while the user chosen cost defines the overall magnitude.

For example, suppose the user configures an Advanced programme with a monthly cost per trainee of ₹ 100,000, a duration of 24 months and a cohort size of 40. The total cost per cohort is:

If the Advanced in country travel share is approximately 0.29, then the implied in country travel cost for this configuration is 0.29 multiplied by ₹ 96.0 million, that is around ₹ 27.8 million. The same logic is applied to all other cost categories to generate a fully disaggregated cost breakdown for each scenario.

6. Program cost per trainee and cohort calculations

6.1 Per trainee and per cohort cost

The STEPS tool reports costs at several levels that are derived from the same underlying calculations:

• Monthly cost per trainee (input from the slider).
• Full programme cost per trainee.
• Total cost per cohort.
• Total cost across all cohorts in a scenario.

The core formulas for a single cohort are given in Equation 4. When a scenario includes multiple cohorts, the tool multiplies the cohort level totals by the number of cohorts to obtain the total scenario cost.

6.2 Relationship to economic cost

Total economic cost in the STEPS tool is defined as the sum of direct costs and indirect costs, including opportunity costs, associated with delivering the FETP programme. Whether the indirect component is included in the economic cost used in benefit cost calculations depends on the opportunity cost toggle described in section 7.

When the toggle is set to include opportunity costs, total economic cost for a given tier and scenario is equal to the full cost envelope computed from the user chosen monthly cost, multiplied by the number of cohorts. When the toggle is set to exclude opportunity costs, the same envelope is adjusted to remove the indirect share for that tier before benefit cost ratios and net benefits are computed.

7. Opportunity cost inclusion and exclusion logic

7.1 Interpretation of indirect and opportunity costs

Indirect costs in Table 3 include both overheads and opportunity costs. Overheads cover institutional costs that are necessary for the programme but not easily traceable to individual trainees, such as management time and shared infrastructure. Opportunity costs reflect the value of alternative uses of staff time and resources that are devoted to the FETP programme instead of other activities.

The high share of indirect costs for Frontline programmes (about 52 percent of total cost) reflects the fact that such programmes are often delivered using existing staff and infrastructure that have multiple competing demands. For Advanced and Intermediate programmes, indirect cost shares are smaller but still substantial, reflecting the seniority of staff and the need for institutional support.

7.2 Inclusion and exclusion toggle

The STEPS tool offers a simple switch that allows users to include or exclude indirect and opportunity costs in the total economic cost used for benefit cost ratios and net benefit calculations. The logic operates as follows for each tier.

Importantly, this toggle does not change the preference model. The monthly cost per trainee entering the mixed logit utility function remains the same, because preferences were elicited for a given total cost envelope. The toggle affects only the denominator used in economic metrics such as benefit cost ratios and net benefit.

8. Willingness to pay calculations

8.1 Separating non cost utility

For each programme scenario, the STEPS tool separates the non cost part of utility from the cost term. The non cost utility is the sum of the coefficients for the chosen tier, mentoring level, delivery mode, qualification and response time (fixed at 7 days) plus the programme constant:

8.2 Marginal willingness to pay per trainee per month

The mixed logit model implies a marginal rate of substitution between the non cost attributes and monthly cost per trainee. The STEPS tool uses this to report an implied willingness to pay per trainee per month for moving from the opt out option to the specified programme, conditional on the attribute levels. The formula is:

Total willingness to pay for a cohort for the full programme duration is then:

The STEPS tool reports both WTP per trainee per month and total WTP per cohort as preference based indicators of stakeholder value. These are shown alongside economic benefits derived from outbreak calibration but are not automatically added to outbreak benefits to avoid double counting.

9. Epidemiological benefit construction

9.1 Concept of outbreak equivalents averted

The epidemiological impact calculations in the STEPS tool are expressed in terms of outbreak equivalents averted. An outbreak equivalent is defined as the avoidance or substantial mitigation of one large multi district or multi state infectious disease outbreak of the type that would impose major health system and economic costs if uncontrolled. The tool does not attempt to model actual outbreak frequencies but instead uses a calibrated mapping from training capacity to expected outbreak equivalents avoided over the planning horizon.

9.2 Training capacity index

For each cohort, the tool computes a training capacity index that reflects the number of trainees, the duration of their training and the tier, mentoring and delivery characteristics of the programme. The index is defined as:

The default weights used in the tool are:

• Tier weights: 0.7 for Frontline, 1.0 for Intermediate and 1.3 for Advanced.
• Mentoring weights: 0.7 for Low, 1.0 for Medium and 1.3 for High.
• Delivery weights: 1.0 for Blended and 1.1 for In person delivery.

These values can be interpreted as relative effectiveness factors. For example, keeping other attributes constant, an Advanced cohort is treated as having about 86 percent more capacity than a Frontline cohort (1.3 versus 0.7), and a High mentoring programme is treated as having 30 percent more capacity than a Medium mentoring programme (1.3 versus 1.0).

9.3 Endorsement adjusted outbreak equivalents

Training capacity translates into expected outbreak equivalents averted once scaled by the endorsement probability. Programmes that stakeholders are unlikely to support are less likely to be implemented effectively, so their capacity is down weighted. The expected number of outbreak equivalents averted for a single cohort is:

In the default calibration used in STEPS, the combination of the capacity weights and the scaling factor of one thousand in Equation 11 yields outbreak equivalents on the order of one tenth to one quarter of a large outbreak per cohort for plausible Advanced and Intermediate programmes, and smaller fractions for Frontline programmes. These values are then combined with the monetary value per outbreak to obtain epidemiological benefits in Rupees.

10. Value per outbreak calibration and evidence base

10.1 Currency conversion assumption

The STEPS tool expresses the economic cost of outbreaks and the benefits of averting them in Indian Rupees. To make use of international evidence, published cost estimates in US dollars are converted into Rupees using an approximate exchange rate of ₹ 83 per US dollar. For example, an estimated cost of US$ 1 billion corresponds to about ₹ 83 billion.

10.2 Literature based calibration table

Table 4 summarises key evidence used to calibrate the value per outbreak in STEPS. The estimates span severe multi country epidemics such as Ebola and SARS, national outbreaks such as pandemic influenza and dengue, and local or regional waterborne and vector borne outbreaks. The last row shows how these values are translated into a working calibration for India.

10.3 Rationale for the STEPS outbreak values

The baseline value of ₹ 20.0 billion per large outbreak is chosen as a conservative central value. It is substantially lower than the economic losses associated with global SARS or regional Ebola outbreaks but is similar in order of magnitude to the estimated national annual burden of dengue in India. It therefore represents a large but not extreme national scale event that could plausibly be influenced by improved surveillance and response capacity.

The lower bound of ₹ 8.3 billion (about US$ 0.1 billion) is intended to represent sizeable but less catastrophic outbreaks. This value is above the lower end of waterborne outbreak cost estimates in low and middle income countries but below annual national dengue burden estimates, and it provides a conservative lower scenario for sensitivity analysis.

The upper bound of ₹ 166 billion (about US$ 2.0 billion) reflects rare but very large outbreaks with significant macroeconomic disruption. It is well below the extreme worst case scenarios for Ebola or global SARS but similar in scale to some national macroeconomic estimates for respiratory pandemics. It provides an upper benchmark for sensitivity analysis.

Within the STEPS tool, users can also select smaller outbreak values such as ₹ 40 million or ₹ 80 million. These values are included in the dropdown menu to allow very conservative assessments where the benefits are restricted to the avoidance of localised outbreaks with limited macroeconomic impact.

11. Benefit cost ratios and net present value

11.1 Monetary value of epidemiological benefits

For each scenario, the monetary value of epidemiological benefits per cohort is given by multiplying the expected number of outbreak equivalents averted by the selected value per outbreak from the Settings tab:

11.2 Benefit cost ratio and net benefit

The core economic indicators are the benefit cost ratio (BCR) and the net benefit (in Rupees). These are defined as:

11.3 Planning horizon and discounting

The planning horizon in the STEPS tool represents the period over which epidemiological benefits from a given cohort of trainees are expected to accrue. For example, a planning horizon of 10 years means that the benefits of training are assumed to extend over a decade of outbreak prevention and control. The default calibration of the capacity index and outbreak equivalents corresponds to a baseline horizon (for example 10 years). When the user changes the planning horizon, the tool scales the capacity index proportionally to reflect a longer or shorter period of impact.

When a non zero discount rate is set, the tool applies a standard annuity factor to adjust benefits and, where relevant, costs that are spread over time. Under a simplifying assumption that benefits are distributed evenly across each year of the horizon, the present value factor is:

In practice, the cohort level calculations reported in this appendix can be interpreted as already expressed in present value terms for a baseline planning horizon. When users change the horizon or discount rate, the tool scales benefits according to the chosen parameters, while the fundamental logic of the outbreak equivalent approach remains the same.

12. National simulation scaling

12.1 From cohort to national scale

The National Simulation tab allows users to define how many cohorts of each FETP tier will be implemented over the planning horizon. For each tier, the tool combines the cost and benefit per cohort with the number of cohorts to derive national totals.

The resulting national totals are then combined into national level BCRs and net benefits. In scenarios where cohorts are phased in over time, the tool can apply discounting in the same way as described in section 11, adjusting for the timing of costs and benefits.

13. Sensitivity analysis framework

13.1 Benefit definition options

The STEPS sensitivity analysis focuses on how benefit cost ratios and net benefits change when key assumptions are varied. The tool distinguishes between two broad notions of benefit:

• Preference based value, captured by WTP per trainee per month and total WTP per cohort.
• Epidemiological value, captured by outbreak equivalents averted multiplied by the value per outbreak.

Main BCR and net benefit indicators in the tool are based on epidemiological benefits (B_epi). WTP measures are reported alongside these as complementary information on how much stakeholders value different programme designs, without being added directly to outbreak based benefits.

13.2 Endorsement adjusted WTP

To provide a more conservative measure of preference based value, the tool also reports endorsement adjusted WTP. This is defined as:

In the sensitivity tables, endorsement adjusted WTP can be compared to economic cost to provide an alternative perspective on value that is based solely on preferences rather than outbreak calibration.

13.3 Outbreak value sensitivity

The Sensitivity tab allows users to evaluate each scenario under multiple values of V_outbreak. The default set of values is:

• ₹ 8.3 billion per outbreak (lower bound).
• ₹ 20.0 billion per outbreak (baseline).
• ₹ 166 billion per outbreak (upper bound).
• Selected smaller values such as ₹ 40 million and ₹ 80 million per outbreak.

For each value, the tool recomputes B_epi, BCR and net benefit. This allows users to see how robust their conclusions are to different assumptions about the monetary cost of outbreaks, including very conservative cases where only localised outbreaks with costs in the tens of millions of Rupees are considered.

13.4 Planning horizon and discounting sensitivity

Users can also vary the planning horizon and discount rate in the Settings tab. Longer horizons increase benefits proportionally through the capacity index scaling, whereas higher discount rates reduce present values. The sensitivity tables show how BCRs and net benefits change across these settings, making trade offs between short term and long term perspectives explicit.

13.5 Interpreting sensitivity results

When reviewing sensitivity tables, a practical approach is to identify scenarios that maintain BCRs above one and positive net benefits across a wide range of outbreak values, planning horizons and discount rates. Such scenarios can be considered robustly favourable. Scenarios that are attractive only under the most optimistic assumptions should be interpreted more cautiously and may require further evidence or targeted design adjustments.

14. Detailed worked examples for illustrative scenarios

This section presents fully worked numerical examples for nine illustrative FETP programme scenarios. For each scenario, response capacity is fixed at 7 days, the value per outbreak is set to ₹ 20.0 billion, opportunity costs are included in economic costs and the planning horizon corresponds to the baseline calibration used in the capacity index.

The scenarios are defined as follows (monthly costs per trainee and durations are inputs to the tool):

14.1 Scenario S1: Advanced, high mentoring, in person, university qualification

14.2 Scenario S2: Advanced, high mentoring, blended, university qualification

14.3 Scenario S3: Advanced, medium mentoring, blended, government certificate

14.4 Scenario S4: Intermediate, high mentoring, in person, university qualification

14.5 Scenario S5: Intermediate, medium mentoring, blended, government certificate

14.6 Scenario S6: Intermediate, low mentoring, blended, government certificate

14.7 Scenario S7: Frontline, high mentoring, in person, government certificate

14.8 Scenario S8: Frontline, medium mentoring, blended, government certificate

14.9 Scenario S9: Frontline, low mentoring, blended, government certificate

14.10 Summary table for scenarios S1 to S9

15. Copilot interpretation and JSON export

15.1 Purpose and design of the Copilot tab

The Copilot tab in the STEPS tool provides a soft integration with Microsoft Copilot. It does not send any data automatically to external services. Instead, it prepares a structured text package that users can copy and paste into their own Copilot environment to generate policy briefs based on the configured scenario.

The Copilot tab has three main functions. First, it collects the current scenario configuration and results, including programme attributes, costs, epidemiological benefits, WTP indicators, benefit cost ratios and key assumptions such as value per outbreak, planning horizon and discount rate. Second, it converts this information into a compact JSON object that is easy for Copilot to parse. Third, it appends a carefully written prompt that instructs Copilot how to summarise the scenario and how to produce a structured policy briefing.

15.2 JSON fields included in the export

The JSON structure used by the Copilot tab includes, at a minimum, the following fields (field names are illustrative):

• scenario_name: User defined label for the scenario.
• tier, mentoring, delivery_mode, qualification: Programme design attributes.
• monthly_cost_per_trainee, duration_months, cohort_size: Core cost and size parameters.
• total_cost_cohort, total_cost_national: Economic costs at cohort and national scale.
• value_per_outbreak, planning_horizon_years, discount_rate: Key settings from the Settings tab.
• endorsement_probability, wtp_per_month, total_wtp, wtp_adjusted: Preference based indicators.
• outbreak_equivalents_averted, benefit_epi_cohort, benefit_epi_national: Epidemiological and economic benefits.
• bcr_cohort, bcr_national, net_benefit_cohort, net_benefit_national: Benefit cost ratios and net benefits.
• include_opportunity_costs: Logical flag for opportunity cost inclusion.
• notes: Optional free text notes from the user.

The actual JSON in the tool is formatted with consistent indentation and quoted field names so that it can be pasted directly into Copilot without further editing.

15.3 Example JSON snippet

An example JSON snippet for a single scenario might look as follows:

This example is illustrative. The STEPS tool automatically populates the values based on the current configuration and results.

15.4 Copilot interpretation prompt

Alongside the JSON, the Copilot tab generates a natural language prompt that instructs Copilot how to interpret the data and what type of output to produce. The prompt typically:

• Explains that the JSON represents a FETP scale up scenario for India, including key costs and benefits.
• Asks Copilot to produce a policy brief of a specified length (for example 3 to 5 pages) in clear, non technical language.
• Requests a structured document with sections such as context, description of the programme, expected benefits, cost and affordability, benefit cost analysis, uncertainties and recommendations.
• Asks Copilot to use tables to summarise key numbers, highlighting BCRs, net benefits and sensitivity ranges.
• Emphasises that the output should be suitable for senior policy makers in the Ministry of Health and Family Welfare, WHO, World Bank and other partners.

15.5 Example interpretation flow for users

A typical user flow for making use of the Copilot tab is as follows. First, the user configures one or more scenarios in the STEPS tool and reviews the summary tab, sensitivity analysis and technical appendix to ensure that assumptions are appropriate. Second, on the Copilot tab, the user reviews the automatically generated JSON and prompt to confirm that they accurately reflect the scenario and that no confidential information is included.

Third, the user copies the combined JSON and prompt text from the Copilot tab and pastes it into Microsoft Copilot in their own secure environment, specifying that Copilot should generate a policy brief. Copilot then parses the JSON, uses the quantitative values for tables and figures, and follows the instructions in the prompt to produce a structured narrative. Fourth, the user reviews Copilot output, checks numbers against the STEPS tool, and edits the text as needed to align with institutional messaging.

The Copilot tab is deliberately one way. It does not call Copilot or send data to any external service. This design ensures that all data handling is under the control of the user and their organisation, and that the STEPS tool remains compliant with requirements for data protection and information security.

16. Summary of assumptions and limitations

16.1 Key assumptions

The STEPS FETP Scale Up Decision Aid relies on several key assumptions. Preference estimates are derived from a DCE using specific attribute levels and a particular sample of respondents. The mixed logit coefficients are treated as fixed means that apply to the user population of interest, and WTP calculations assume that the linear in parameters utility specification is appropriate over the cost range permitted by the slider.

Epidemiological benefits are expressed in terms of outbreak equivalents averted using a stylised capacity index. The weights used for tier, mentoring and delivery are judgement based and are calibrated so that a cohort of Advanced or Intermediate trainees yields a fraction of a large outbreak equivalent over the planning horizon. The value per outbreak is calibrated to be consistent with international and Indian economic evidence but remains uncertain. The tool therefore emphasises sensitivity analysis around outbreak values.

16.2 Limitations and appropriate use

The STEPS tool is not an epidemic model. It does not simulate disease transmission or dynamically represent how FETP training affects outbreak detection and control. Instead, it provides a transparent framework for combining preference evidence, cost templates and outbreak value assumptions in a consistent way.

The tool does not capture all potential benefits of FETP scale up, such as improved routine surveillance, capacity for non infectious threats, or broader system strengthening. It also does not model potential crowding out of other programmes when resources are reallocated. The opportunity cost toggle provides a partial way of reflecting trade offs.

Despite these limitations, the STEPS tool offers a structured and transparent approach for comparing alternative FETP scale up strategies. By documenting all assumptions, equations and data sources in this technical appendix, it enables policy analysts and decision makers to understand, scrutinise and adapt the tool for their own contexts, and to communicate the rationale for investment decisions in a clear and evidence informed way.