The OHS Prevention Ledger: How to Show the Working on Prevention

A guide to using the Prevention Ledger — what it records, how it calculates, how to read its results, and the mistakes that most often make prevention look better or worse than the evidence allows.

Why a ledger, not a calculator

Most prevention claims arrive as a single number: every pound spent saves ten, or fifteen, or thirty. The flagship article in this issue, The Economics of Prevention, argues that the structural case for prevention is strong but that these numbers are often weak, because the steps between the money and the benefit are not shown. A ratio that assumes an intervention would have removed the whole cost of an epidemic, or that borrows an estimate from a different country and decade without saying so, is advocacy dressed as analysis.

The Prevention Ledger is designed to make that harder to do by accident and easier to spot when someone does it on purpose. It does not ask for the benefit of prevention. It asks for the pieces from which a benefit could be estimated — the capability the money buys, the change in risk that capability produces, the consequences that change could avoid — and it asks where each piece comes from. The ratio is the last thing it produces, and it is never displayed alone.

The intended effect is simple: a spectacular 20:1 ratio built largely from transferred assumptions should look visibly weaker than a modest 2:1 ratio supported by direct evidence.

Before you start

Four decisions shape everything that follows. Make them first and write them down.

What decision is this informing? A Ledger for a funding bid, a programme review and a policy argument will draw on the same evidence but emphasise different parts of it. Be clear which you are doing.

What is the analysis horizon? The Ledger compares costs and consequences over a fixed number of years. Choose a horizon long enough for the intervention’s effect to appear — a vaccination programme that takes three years to reach coverage should not be assessed over two — but not so long that the counterfactual becomes speculation.

What currency and price year? Enter all money in one currency and one price year. Mixing 2019 outbreak costs with 2026 programme costs silently changes the ratio.

What do the three scenarios mean? In the Ledger, Low is always the scenario least favourable to the intervention: the smallest plausible risk reduction, the smallest plausible consequences, and the highest plausible cost. High is the most favourable. Central is your best estimate. Use the same meaning in every section. If the low scenario is built from optimistic assumptions in one place and pessimistic ones in another, the range stops meaning anything.

The eight steps

The workbook and the online version follow the same sequence. Workbook tab numbers are shown in brackets.

1. Intervention and cost (tab 02)

Describe what is being funded, where, against which biological threat, and who pays. Then enter the costs for each scenario: the capital cost, meaning one-off spending at the start in year 0 — equipment, facilities, set-up; the recurring annual cost in each year of the horizon — staff, vaccine, consumables, maintenance; and any other implementation cost falling in year 0, such as training or programme design.

The Ledger converts these to a present value using the discount rate you enter. The default is 3.5%, the rate in HM Treasury’s Green Book; change it if your setting uses another. Remember that the low scenario carries the highest plausible cost.

2. Capability created (tab 03)

Record what the money actually creates. Capability is the link between spending and risk, and it is the step most often skipped. Typical capabilities are vaccination coverage, detection speed, laboratory throughput, trained staff, tracing coverage and reporting relationships with farmers or private vets.

For each one, record the baseline state, the state the intervention is expected to achieve, the mechanism by which it changes, and — importantly — how long it lasts without further investment. Capability decays: staff leave, equipment ages and relationships weaken. A Ledger that assumes a capability built in year one still works in year ten should say why.

3. Counterfactual and risk change (tab 04)

This is the centre of the Ledger. First, state in words what you believe would happen without the intervention and what evidence supports that belief. Then enter, for each scenario, the probability that the event occurs over the horizon without the intervention, and the probability that it occurs with it. The Ledger calculates the absolute and relative risk reduction from these. Those probabilities drive the avoided value in step 5, so the counterfactual is visible and argued, not hidden inside a formula.

Finally, record your counterfactual confidence — High, Moderate or Low. This is a structured judgement about how well the “without” scenario is evidenced, not a statistical confidence interval. A counterfactual based on a well-observed pre-programme period in the same population deserves more confidence than one assembled from another country’s outbreak history.

For an endemic burden that occurs every year — dog-mediated rabies, for example, or a production disease — the event is certain in both cases. Set both probabilities to 100% and record the change in burden in steps 4 and 5.

4. Outcomes potentially avoided (tab 05)

Before money, record the outcomes that plausibly differ if the event occurs, in natural units: human cases, deaths, animals culled, farms restricted, days of trade disruption, courses of post-exposure prophylaxis. Enter baseline and with-intervention values for each scenario; the Ledger shows the difference.

This step exists because many readers care more about the outcomes than the money, and because an economic estimate that cannot be traced back to a change in outcomes is not credible.

5. Economic consequences (tab 06)

For each category of consequence — public response expenditure, animal losses, healthcare costs, business interruption, trade disruption and so on — enter the total consequence over the horizon if the event occurs without the intervention, and the total if it occurs with it.

Enter these undiscounted, in the stated currency and price year. The intervention figure captures effects on severity: earlier detection may not prevent an outbreak but may make it much smaller. If you leave the with-intervention figure blank, the Ledger treats it as equal to the baseline — meaning the intervention changes only the likelihood of the event, not its size.

Mark whether each row is included in the ratio, and record its evidence basis and an evidence ID. The Ledger then weights each consequence by the probabilities from step 3 and discounts it.

6. Who pays? Who benefits? (tab 07)

Allocate the present-value cost and the expected avoided value between the actors who bear them: government, producers, the health system, supply chains, insurers, households, other governments. Record who still bears residual risk.

The Ledger checks that the allocation adds up to the totals, and calculates the payer’s own ratio: the benefit that reaches the primary payer divided by the cost that payer bears. Comparing this with the society-wide Prevention Ratio is how the Ledger demonstrates the prevention paradox examined in The Prevention Paradox: Who Pays When Prevention Works? An intervention can have a strong case for society and a weak one for the budget holder, and that gap is often why it is not funded.

7. Evidence register (tab 08)

Give every important parameter an ID and record its source, the population or setting it comes from, whether it transfers to your setting, and your confidence in it. Classify each piece of evidence as Observed (directly measured in this setting), Empirically estimated (derived from relevant empirical research), Modelled (produced by epidemiological or economic modelling), Transferred (taken from another setting or population) or Assumed (analyst judgement where better evidence is unavailable).

None of these categories is disqualifying. Most real Ledgers will depend on modelled and transferred estimates. The point is to make that visible.

8. Material effects not monetised (tab 11)

Record the effects that matter but are not in the ratio: animal welfare, mortality or morbidity the economic model does not capture, institutional trust, professional capacity, ecological consequences, distributional effects and social disruption. For each, record the direction, the materiality and why it is not monetised. Something does not matter only if it can be priced, and a Ledger that leaves these out invites the reader to assume they are zero.

How the calculation works

The calculations are deliberately simple, so that anyone can check them.

Present value of cost. Capital and other costs fall in year 0. Recurring costs fall in years 1 to n, discounted at rate r:

PV cost = capital + other + recurring × [1 − (1 + r)⁻ⁿ] ÷ r

Expected avoided value. For each included consequence and each scenario:

Expected avoided = (baseline probability × baseline consequence) − (intervention probability × intervention consequence)

This captures both routes by which prevention works — making an event less likely and making it less severe — and ties the result directly to the counterfactual in step 3.

Discounting the benefit. Consequences are assumed to fall evenly across years 1 to n, so each is multiplied by a benefit timing factor equal to the annuity factor above divided by n. Costs and benefits are therefore discounted on the same basis. Dividing discounted costs by undiscounted benefits would flatter any programme longer than a year; the Ledger does not do this.

Prevention Ratio.

Prevention Ratio = PV of expected consequences potentially avoided ÷ PV of prevention cost

It is called a Prevention Ratio rather than a return on investment because most of the benefit usually does not return to the organisation that pays.

Evidence strength. The Ledger reports the share of the avoided value supported by Observed or Empirically estimated evidence, and the share resting on Transferred or Assumed evidence. When more than half of the central avoided value rests on transferred or assumed evidence, the result is flagged.

Reading the results

The summary (tab 10, or the results panel online) places each stage of the chain beside its evidence and the question it raises. Read it in this order.

First, the range. Look at all three ratios. If the low scenario is below 1×, the case depends on the central or high assumptions being right, and you should say which assumptions matter most. A central ratio of 3× with a low of 0.4× is a very different finding from a central of 3× with a low of 2×.

Second, the evidence. How much of the avoided value is observed or empirical, and how much is transferred or assumed? A high ratio resting on transferred assumptions is a hypothesis, not a finding.

Third, the counterfactual confidence. If it is Low in any scenario, treat the ratio as illustrative.

Fourth, the distribution. Compare the payer’s own ratio with the society-wide ratio. If they diverge sharply, the policy question may not be whether the intervention is worth doing, but who else should pay for it.

Last, what is not monetised. A modest ratio alongside large unmonetised effects on animal welfare or trust may justify investment; a high ratio alongside significant harms borne by one group may not.

A worked illustration

The following example is fictional. It is included to show how the Ledger behaves, not to make any claim about a real programme. The first real worked case — mass dog vaccination for rabies control in Ethiopia — is set out in What Does Prevention Actually Buy? in this issue.

A government veterinary service is considering a five-year enhanced surveillance programme for an exotic livestock disease. The central estimate is a capital cost of £1.0 million and running costs of £200,000 a year. The low scenario assumes £1.2 million and £250,000; the high scenario £0.9 million and £180,000. The discount rate is 3.5%.

Without the programme, the probability of an outbreak over five years is judged to be 30% (low 20%, high 40%). With it, 15% (low 15%, high 12%), because earlier detection allows some incursions to be stamped out before they spread. If an outbreak does occur, earlier detection is expected to halve public response costs, from £6 million to £3 million in the central case (modelled). Animal losses of £15 million if an outbreak occurs are taken from a well-documented recent outbreak in the same country (observed), and the programme is assumed to change only their likelihood, not their size.

The Ledger’s reading of this example is instructive.

The central ratio of 1.7× suggests a reasonable case, and 62.5% of the central avoided value rests on observed evidence. But the low scenario falls to 0.4×, so the case depends on the risk reduction being closer to the central estimate than the low one. The most useful next step would be better evidence on how much earlier detection actually reduces the chance of an established outbreak.

The distribution is more striking still. The government pays the whole cost, but the only benefit it receives directly is the avoided response expenditure — £1.22 million in the central case. Its own ratio is therefore 0.64×. Producers receive £2.03 million in avoided losses and pay nothing. Judged by its own budget, the veterinary service would not fund a programme that is worth 1.7 times its cost to society. That is the prevention paradox, stated in numbers rather than asserted.

Common mistakes

Attributing the whole cost of an outbreak to one intervention. Prevention rarely eliminates an event with certainty. The Ledger forces the benefit through the change in probability and severity; resist entering an outbreak’s full cost as the avoided value.

Counting the same loss twice. Lost farm income and lost processor margin may be the same money seen from two ends of a supply chain. Allocate each loss once.

Counting compensation as an economic cost. Compensation paid to farmers is a transfer from taxpayers to producers, not an additional loss to society. It matters enormously for who bears the cost — the subject of Paying for Cooperation in Issue 001 — and belongs in step 6, but it should not be added to animal losses as if it were a separate consequence.

Forgetting the with-intervention consequence. If earlier detection makes outbreaks smaller, enter the smaller figure. If it does not, leave it blank — but make sure that reflects the evidence and not an oversight.

Mixing price years and currencies. Convert everything to one currency and one price year before entering it.

Letting scenarios drift. Low should mean least favourable everywhere, including cost.

Leaving the evidence basis blank. The Ledger flags any included consequence without an evidence basis. A blank is not neutral; it hides the weakest part of the case.

Limitations

The Ledger is a structured record, not a full economic evaluation. In particular, it treats the event as a single probability over the horizon rather than modelling repeated or concurrent outbreaks; it assumes consequences fall evenly over time, which may understate or overstate discounting for front-loaded or back-loaded effects; its three scenarios are not a probabilistic sensitivity analysis and do not produce confidence intervals; it does not apply distributional or equity weights, although step 6 records the distribution explicitly; and it relies on the analyst’s classification of evidence, which should itself be open to challenge.

Where a decision requires a full cost–benefit or cost-effectiveness analysis, the Ledger is best used as the structured evidence record that sits beneath it.

Using the online version

The online Ledger on the One Health Security website follows the same eight steps and uses the same calculations as the workbook. Results update as you type. Entries remain in your browser and are not sent to One Health Security. You can download your assessment as an Excel file or print it to PDF. For larger assessments, or to keep a version history, use the workbook.

Both the online Ledger and the workbook are free to use, with no registration and no email address required.

Use the Ledger

Feedback and versions

This is version 1.0. We expect it to change as people use it. If you apply the Ledger to a real programme, find a case it handles badly, or disagree with a design choice, we would like to hear from you through the contact page.

Cite as: One Health Security (2026) OHS Prevention Ledger v1.0: methodology note. onehealthsecurity.com/resources/

Questions & Answers

References

1. Brown, G.W., von Agris, J., Tacheva, B. and Bell, D. (2026). An investment too good to be true?: Reassessing the World Health Organization and World Bank return-on-investment estimates for pandemic preparedness. Health Economics, Policy and Law, published online 6 August 2026.
2. CIPFA (2025). Evaluating Preventative Investments. London: Chartered Institute of Public Finance and Accountancy.
3. HM Treasury (2022). The Green Book: Central Government Guidance on Appraisal and Evaluation. London: HM Treasury.