Budget Impact Modelling

Budget Impact Modelling Step by Step

Budget impact modelling is the hands-on work of turning a payer's affordability question into a calculation that the payer can open, check and rerun with its own numbers. The modeller fixes whose budget is at stake and for how long, projects the eligible population, forecasts uptake and displacement, attaches prices, runs both scenarios year by year, then tests and documents the result. The order matters because each step feeds the next, so an error in the population funnel or the displacement rule carries through to every annual figure. Modelling is the working stage of a budget impact analysis. It produces a budget impact model, and that model outputs the budget impact figure.

Scoping the budget holder, horizon and reporting periods

The first decisions fix the boundaries of the model, and the payer sets them. The 2007 ISPOR task force report on budget impact analysis recommends the perspective of the budget holder and calls the drawing of budget boundaries a highly local exercise, since one pharmacy budget may cover only drugs while elsewhere drugs sit inside a hospital budget. A model built for one budget holder therefore needs switches for the cost categories another would include.

The time horizon follows the budgeting process, which the 2007 report says is usually annual, with results given period by period rather than as one net present value. The June 2020 guidelines of Canada's Patented Medicine Prices Review Board (PMPRB) count 12-month periods from the proposed listing date, reportable by contract or fiscal year. The company budget impact template that NICE publishes for technology appraisal asks for the eligible population and market shares for the next 5 years.

Building and projecting the eligible population

The eligible population is usually built as a funnel. The 2007 ISPOR report applies successively more restrictive inclusion criteria to the decision maker's population, and the PMPRB guidelines narrow the covered population to those with the condition, then those diagnosed, then those diagnosed and treated.

$$N_t = P_t \times f_t \times d_t \times r_t$$

where $N_t$ is the eligible population in year $t$, $P_t$ is the covered population, $f_t$ is the disease frequency, prevalence or incidence, $d_t$ is the proportion diagnosed and $r_t$ is the proportion meeting the licence and any access restriction, such as failure of first-line treatment. Each factor must be a distinct stage with its own denominator, or the product double-counts.

The PMPRB guidelines add practical rules. The covered population for a 12-month period is taken halfway through it, which assumes linear growth. Incidence suits shorter treatments and prevalence longer ones. In the absence of any relevant data on diagnosis and treatment, all eligible people with the disease are assumed to be diagnosed and treated. Off-label use stays out of the base case and goes into a scenario. Claims data suit drugs entering mature markets, and estimates should be compared with historical data.

The population is normally open. The 2007 ISPOR report notes that people enter and leave as they develop the indication or stop meeting the criteria, so the model needs yearly inflows and outflows, not just a starting count. When inflows and outflows change with the new technology, the 2014 ISPOR report allows a cohort model in place of a cost calculator.

Forecasting uptake, market share and displacement

Uptake and market share differ. Uptake is the proportion of eligible people treated with any listed option, and the new technology's market share is its proportion of those treated.

$$n_t = N_t \times u_t \times m_t$$

where $n_t$ is the number of people on the new technology in year $t$, $u_t$ is uptake among the eligible and $m_t$ is the new technology's share of treated people.

The 2007 ISPOR report notes that uptake of a new technology, $u_t \times m_t$ here, changes as physicians and patients become familiar with it, and points to producers' market share estimates, explicit switching rules (for example only after failure on current therapy) and diffusion elsewhere. The PMPRB guidelines prefer evidence from markets where the drug is already reimbursed and allow the share to be built as a product of percentages, such as the shares of physicians aware of the drug and of patients likely to switch. Each year's capture builds on the year before, and with no data, market growth caused by the new drug is set at 0% and tested, for example with a 5% increase. The pattern over time is often drawn as an adoption curve.

The modeller must also decide what the new technology replaces. Without better data, the PMPRB guidelines assume that each comparator loses share in proportion to its current share.

$$m_{A,j,t} = m_{R,j} \times (1 - m_t)$$

where $m_{A,j,t}$ is the share of comparator $j$ in the adoption scenario in year $t$, $m_{R,j}$ is its share in the reference scenario and $m_t$ is the new technology's share.

Costing each treatment and pathway

Costing follows the 2007 ISPOR sequence of identifying the resource use that may change, estimating the amount of change and valuing it. Valuation uses transaction prices, including rebates, because those are what the budget holder pays. Freed capacity, such as staff time, may have no monetary effect within the horizon if staff are not redeployed.

The NICE manual (PMG36, section 4.11.6) asks for net NHS costs broken down by organisation and budget category, such as drugs, staffing, consumables and capital, a statement of how far any savings can be realised, whether VAT is included, and prices that reflect what the NHS pays, including confidential price reductions when appropriate. The NICE company template asks for one results table at list price and another with any commercial arrangement. For drug plans, the PMPRB guidelines add markups, dispensing fees and co-payments, wastage from units dispensed but not consumed, and the cost effect of compliance and persistence, two aspects of adherence.

Running the reference and adoption scenarios

Each scenario forecasts total spending in each period. The PMPRB guidelines compare a Reference Scenario projected from the existing market with a New Drug Scenario in which the drug is listed.

$$E_{s,t} = \sum_{j} n_{s,j,t} \times c_{j,t}$$

where $E_{s,t}$ is total expenditure in year $t$ under scenario $s$ ($A$ for adoption, $R$ for reference), $n_{s,j,t}$ is the number of people on treatment $j$ and $c_{j,t}$ is its annual cost per person.

$$BI_t = E_{A,t} - E_{R,t}$$

where $BI_t$ is the periodic budget impact in year $t$.

$$CBI_T = \sum_{t=1}^{T} BI_t$$

where $CBI_T$ is the cumulative budget impact over $T$ years. The NICE company template asks that the annual figures take account of people treated for more than 1 year. Under PMG36 (sections 5.10.4 and 5.10.5), a net budget impact expected above GBP 40 million a year in any of the first 3 financial years leads NHS England to offer commercial discussions, so a model for England reports each year separately.

Worked example of a three-year model with scenarios

The numbers are illustrative and describe no real technology. A regional payer covers 2,000,000 people. Prevalence is 1.5%, 60% of cases are diagnosed and 25% of those meet a restriction to use after first-line failure, so 2,000,000 × 0.015 × 0.60 × 0.25 = 4,500 people are eligible each year. With uptake of 80%, 3,600 are treated. Comparator A holds 75% of them at GBP 3,000 a year and comparator B 25% at GBP 5,000. The new drug costs GBP 6,000 a year, with shares of 10%, 20% and 30% in years 1 to 3, full-year treatment and proportional displacement.

1. Reference spending. The average treated person costs 0.75 × 3,000 + 0.25 × 5,000 = GBP 3,500, so $E_{R,t}$ is 3,600 × 3,500 = GBP 12,600,000 in each year.
2. Year 1 patients. The new drug treats 3,600 × 0.10 = 360 people, comparator A 3,600 × 0.75 × 0.90 = 2,430 and comparator B 3,600 × 0.25 × 0.90 = 810, a total of 3,600.
3. Year 1 spending. The three groups cost GBP 2,160,000, GBP 7,290,000 and GBP 4,050,000, so $E_{A,1}$ is GBP 13,500,000 and $BI_1$ is GBP 900,000.
4. Years 2 and 3. With 720 and 1,080 people on the new drug, $E_{A,2}$ is GBP 14,400,000 and $E_{A,3}$ is GBP 15,300,000, so $BI_2$ is GBP 1,800,000 and $BI_3$ is GBP 2,700,000.
5. Cumulative result. $CBI_3$ is 900,000 + 1,800,000 + 2,700,000 = GBP 5,400,000.
6. Reconciliation. Reference spending plus budget impact returns adoption spending, since 12,600,000 + 900,000 = 13,500,000.

Each new-drug patient replaces an average GBP 3,500, a net GBP 2,500, so year 3 is also 1,080 × 2,500 = 2,700,000. Each scenario changes one assumption.

A patient taken from B saves GBP 5,000, a net GBP 1,000. B has only 900 patients, so in year 3 the model takes 900 from B and 180 from A at a net GBP 3,000, giving 900 × 1,000 + 180 × 3,000 = 1,440,000. A formula taking all 1,080 from B would leave minus 180 people on B, which a check for negative counts catches. Growth adds 180 × 6,000 = GBP 1,080,000 a year, and the lower price gives a net GBP 1,300 per patient.

Around a base of GBP 5,400,000, the displacement rule alone moves the three-year total by GBP 2,880,000, more than half, with no change in price or clinical evidence. None of these figures says whether the drug is good value, which is a question for cost-effectiveness analysis.

Sensitivity and scenario analysis for budget impact

Each scenario analysis is a named set of input switches that moves linked inputs together. The worked example's table already varies market size, market shares and the new drug's price, which with comparator prices make up the PMPRB minimum, so a deterministic sensitivity analysis of comparator prices would complete it.

The 2014 ISPOR task force report by Sullivan and colleagues recommends alternative scenarios chosen from the decision maker's perspective. The 2007 report suggests most optimistic and most pessimistic scenarios agreed with decision makers, and a tornado diagram of the main drivers.

Checking and documenting the model

A budget impact model can give plausible totals from broken patient flows, so checking runs alongside building. Formal model validation starts with face validity and verification, and verification rests on checks built into the workbook. Simple checks catch most errors, such as shares summing to 100% in each scenario and year, no negative patient counts, and reference spending plus budget impact equalling adoption spending.

Documentation lets the payer replace the modeller's numbers with its own. The 2007 ISPOR report recommends one input worksheet, sources beside each parameter, colour-coded input cells and a way to restore defaults, with full references on a reference worksheet.

Practice errors that distort budget impact

Most errors arise in the population and market share steps. The arithmetic there is simple, but the denominators are easy to confuse.

1. Overlapping funnel percentages. Multiplying prevalence by a proportion that already includes diagnosis counts one restriction twice.
2. Share of eligible read as share of treated. A 30% share of treated people is fewer than 30% of the eligible whenever uptake is below 100%.
3. Displacing more people than a comparator has. A displacement rule must stop at zero, as the year 3 scenario shows.
4. Counting freed capacity as cash. Released capacity is not a saving unless the budget changes, as the 2007 ISPOR report and PMG36 section 4.11.6 make clear.
5. Hard-wiring one price. The NICE template asks for results at list price and with any commercial arrangement, and the PMPRB guidelines ask for a model in which reviewers can test price rebates.

Sources

• Mauskopf JA, Sullivan SD, Annemans L, Caro J, Mullins CD, Nuijten M, Orlewska E, Watkins J, Trueman P. Principles of good practice for budget impact analysis: report of the ISPOR Task Force on good research practices, budget impact analysis. Value in Health. 2007;10(5):336-347. https://doi.org/10.1111/j.1524-4733.2007.00187.x
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 31 January 2022, last updated 13 October 2026. Sections 4.11.6, 5.10.4 and 5.10.5. https://www.nice.org.uk/process/pmg36
• National Institute for Health and Care Excellence. Company budget impact analysis submission template. Technology appraisal submission templates page, accessed 26 October 2026. https://www.nice.org.uk/what-nice-does/our-guidance/about-technology-appraisal-guidance/technology-appraisal-submission-templates-and-supporting-documents
• Patented Medicine Prices Review Board. Budget Impact Analysis Guidelines: Guidelines for Conducting Pharmaceutical Budget Impact Analyses for Submission to Public Drug Plans in Canada. 2nd ed. Ottawa: PMPRB; June 2020. https://publications.gc.ca/site/eng/9.885893/publication.html
• Sullivan SD, Mauskopf JA, Augustovski F, Jaime Caro J, Lee KM, Minchin M, Orlewska E, Penna P, Rodriguez Barrios JM, Shau WY. Budget impact analysis: principles of good practice. Report of the ISPOR 2012 Budget Impact Analysis Good Practice II Task Force. Value in Health. 2014;17(1):5-14. https://doi.org/10.1016/j.jval.2013.08.2291