Contingency and risk transfer are the two financial mechanisms a hospital project uses to absorb the uncertainty that the risk register identifies. Contingency funds the uncertainty the owner chooses to retain; insurance and contractual transfer move the rest off the balance sheet — and both fail quietly when sized by reflex rather than by analysis, or when a policy is bought as a commodity instead of read against the acute-care risk profile it actually has to cover.

Contingency is the financial expression of the risk register. It is not a slush fund, a hidden margin, or a number to be defended by withholding. A mature hospital project runs multiple, named, separately-governed contingencies, each sized to a specific class of uncertainty, each with explicit drawdown authority and a transparent ledger — and it confirms that the risks it intends to transfer are genuinely transferred by the policy wording, before a loss rather than after.

The four named contingencies

Hospital capital programs carry distinct classes of uncertainty that mature at different times and are owned by different parties. Pooling them into a single "contingency" line invites the most common failure mode — solving one class of problem by raiding the reserve meant for another. Run four named, separately-governed reserves.

Design contingency is largest early and is consciously drawn down and retired as design completes and pricing firms up. The discipline is to plan its reduction gate by gate; a design contingency still full at GMP signals that estimating uncertainty was never resolved.

Construction contingency covers field conditions, minor coordination gaps, and incidental rework that fall within the contracted scope. It explicitly does not cover owner-directed scope changes, which draw on owner's contingency. Define in the contract the categories it may and may not be used for, to prevent erosion into what should properly be change orders.

Owner's contingency is the reserve that absorbs the highest-consequence healthcare-specific risks: the mid-project FGI or code-edition adoption, the new AHJ interpretation, the late clinical requirement that surfaces after a room is built, and the deductible or uninsured exposure left behind by the insurance program (see below).

Activation / transition contingency is the most frequently omitted and underfunded reserve. Activation carries its own dense risk profile — late equipment substitutions, additional simulation cycles to close a workflow gap found in a mock code, incremental dual-running cost when go-live slips. Fund it as a dedicated reserve so activation problems are not solved by raiding construction reserves at the worst possible moment.

Owner risk appetite sets the confidence level

Before any contingency number can be defended, the owner must make one explicit decision: how much residual risk of overrun is acceptable? This is the owner's risk appetite, and it is the single parameter that sets the confidence level the contingency is funded to. It is a governance decision — made by the owner's board / CFO with the program executive — not a project-controls default, and it should be recorded in the Risk Management Plan.

The mapping from appetite to confidence level is direct:

• P50 funds the project to its expected (50/50) outcome — half the time the contingency is exhausted. Appropriate only for a risk-tolerant owner, or as the target in a portfolio where over- and under-runs wash out across many projects.
• P70–P80 is the normal healthcare capital range — only a 20–30% modeled chance of breaching the budget. Where to sit in that band is the expression of appetite: a system financing a single bond-funded replacement hospital with no balance-sheet cushion sits at P80 (or higher); an owner with portfolio flexibility and tolerance for variance may accept P70.
• Above P85–P90 ties up capital that could be deployed elsewhere and often signals an immature register being padded — the better response is to retire the underlying risks, not to over-reserve.

State the appetite as a sentence the board has signed: "Construction contingency is funded to the P80 confidence level on the integrated cost-risk model; schedule reserve is funded to P80 on the integrated CPM, with regulatory-milestone risks modeled explicitly." That single line makes the reserve auditable and ends the perennial argument over whether the contingency is big enough.

Size contingency quantitatively, not by reflex percentage

A flat-percentage contingency is unanchored — it can be simultaneously too small for the real risk and impossible to defend to a board. Size reserves with cost-risk Monte Carlo analysis driven by the register, and stress the schedule reserve against the regulatory milestones a contractor's float does not protect. The minimal practitioner method has five steps and needs nothing more exotic than a risk-analysis add-in (e.g., @RISK, Safran, Acumen, or Primavera Risk Analysis) or a modest spreadsheet/Python model.

1. Build the cost model from the estimate. Take the line-item or major-WBS cost breakdown as the deterministic base. This base carries no contingency — contingency is the output of the analysis, not an input.
2. Assign a distribution to each uncertain line and to discrete register risks. For estimate uncertainty on continuous quantities (concrete, structural steel, AHU pricing), use a three-point PERT or triangular distribution — Low / Most-Likely / High — captured in calibrated ranging workshops with the estimator and trade leads. For discrete register events (an FGI re-review, a contaminated-soil find), model each as a probability × impact branch: a 30% chance of a $1.2M shielding-redesign-and-re-review. Anchor the three points in real ranges, not arbitrary ±10%.