CRA Administrative Positions on PHSP Plan Documents

A Private Health Services Plan only delivers tax-free benefits if the underlying documentation meets CRA's technical standard for what counts as insurance. That standard comes almost entirely from one bulletin: IT-339R2, archived by CRA years ago but never replaced, and still the document CRA's own payroll guidance points people toward. Archived does not mean irrelevant: the rules are frozen in text from an earlier era, and reading them correctly now requires layering in the technical interpretations CRA has issued since. It means the rules are frozen in text from an earlier era, and reading them correctly now requires layering in the technical interpretations CRA has issued since.

Two bulletins built this framework. IT-85R2 set up the Health and Welfare Trust structure, mainly for incorporated entities. IT-339R2 came next and introduced the PHSP concept for non-incorporated businesses, and it remains the primary reference for what a plan document has to contain. The Income Tax Act itself defines a PHSP as a contract of insurance for hospital and medical expenses, a medical or hospital care insurance plan, or some combination of the two. Simple enough on paper. The compliance work occurs in the details, produced by the gap between the Income Tax Act's simple definition and the specific requirements that must be met in practice, as CRA's technical interpretations show.

The insurance character requirement and its five elements

Paragraph 3 of IT-339R2 lays out five things a plan has to have before CRA will treat it as insurance in substance, not just in name. There has to be an undertaking by one person, to indemnify another person, for an agreed consideration, against a loss or liability tied to some event, and that event's occurrence has to be uncertain.

Arm's length insurance contracts tend to have all five baked in automatically, because that's what an insurance contract is. Self-insured employer plans don't get that for free. An employer setting up its own plan has to build these elements in deliberately, through the plan document's actual language. Paragraph 7 of IT-339R2 confirms self-insured arrangements can still qualify, but only if the insurance elements genuinely exist.

Uncertainty is the element that does the most work here. If the benefit amount is fixed and guaranteed regardless of any actual health event, what's left isn't insurance, it's just a reimbursement account dressed up in insurance language. If an employee's benefit amount is fixed and guaranteed regardless of any actual health event, there's no risk being transferred, no indemnification against an uncertain loss. That's the line between a real PHSP and a disguised form of compensation, and CRA draws it precisely there.

The contractual obligation: what the plan document must say

Paragraph 7 also sets the bar for the employer's side of the bargain: reimbursements only qualify where the employer is obligated, under the employment contract, to make them. Not willing to. Not usually does. Obligated.

Whether that obligation actually exists is a question of fact, decided case by case, and CRA doesn't hand out a template. For unionized workplaces, the bar is lower in practice, because a plan spelled out in a collective agreement and communicated to employees generally satisfies the obligation test on its own.

Outside a union setting, there's no shortcut. The plan document itself, or a written employment contract that references it, has to establish the obligation in writing. An employer who's been reimbursing medical expenses informally for years, with nothing written down, hasn't created a PHSP. It's created a pattern of behavior that CRA can, and will, decline to recognize.

Three-contract structure required in cost-plus plans

Cost-plus arrangements, common with pay-as-you-go HSA administrators, need more than a single agreement. Paragraph 6 of IT-339R2 implies a three-contract structure has to exist for the arrangement to hold up.

First, a contract between the employer and the trusteed plan or insurance company. Second, one between the employee and that same trusteed plan or insurer. Third, a contract between employee and employer, usually embedded in the employment agreement. In practice, that means the employer contracts with an administrator to indemnify employee claims within defined risk categories, promises to reimburse approved claims plus an administration fee, and is bound by the employment contract to cover proper claims the administrator pays out.

Each leg carries its own legal weight, and none of them is optional. Skipping one causes the insurance character argument to fall apart, because the chain of obligations that makes the arrangement look like insurance in the first place has a broken link. A business running an HSA through a third-party administrator shouldn't assume a plan booklet covers this. It needs to confirm all three contractual relationships actually exist on paper.

The "all or substantially all" rule effective January 1, 2015

For years, CRA held every plan to an all-or-nothing standard: every expense covered had to qualify under the Medical Expense Tax Credit, full stop. One ineligible category could jeopardize the whole plan's status.

That changed with a revised position announced in November 2015, applied retroactively to January 1, 2015. Under the new rule, a plan qualifies as long as 90% or more of the premiums paid relate to METC-eligible expenses, and the other conditions in paragraph 3 are still met.

The test isn't identical for every plan type, though. Insured plans get measured on premiums, not benefits paid out. So if an insured plan pays claims where only 88% of the dollar amount going out the door happens to be METC-eligible, that doesn't disqualify anything, because the 90% threshold applies to what's paid in as premiums, not what's paid out in claims. Self-insured plans face a stricter version: at least 90% of benefits actually paid have to be for METC-eligible expenses, full stop, no separate premium calculation to fall back on.

A worked example drawn from industry guidance on the CRA position makes the math concrete. Take a $1,000 monthly premium split across four categories: $700 in prescription drugs, $200 in dental, $80 in medical tests, and $20 in non-prescription vitamins. The first three categories, all METC-eligible, together make up the large majority of the total. The plan qualifies, even though the vitamin category, on its own, would never pass the METC test.

Funds reversion, carry-forward limits, and the forfeiture principle

Money inside a PHSP cannot flow back to the employer. That's a hard rule from IT-339R2, with essentially no wiggle room, because a plan that lets unused funds revert to the person who funded it starts to look less like insurance and more like a savings account with extra steps.

There's one narrow carve-out. A 1998 CRA bulletin allows reversion in the specific case of a notional credit program tied to a flex benefits or cafeteria-style plan. Outside that scenario, reversion kills the plan's status.

Carry-forward provisions get a bit more room to breathe, but only within limits. CRA's stated position allows a plan to carry forward either unused flex credits or unused eligible medical expenses, for up to 12 months, but not both at once. Stacking both types of carry-forward together causes the plan to lose its footing.

What tips a plan over into disqualification is more specific than it sounds. An employee with zero allocation in the prior year shouldn't be able to carry forward expenses from a year they had no coverage. And allocating some token, nominal amount purely to manufacture a carry-forward balance doesn't fix that. CRA's own language on this point is blunt: a plan structured that way carries little or no actual risk, and a plan with no risk has no insurance element to speak of.

Unilateral discretion as a disqualifying factor in employer termination rights

CRA technical interpretation 2012-044701E5 addresses questions about spending caps and employer termination rights. On the cap question, a spending limit is generally understood as a defined boundary on risk, and defined risk is what insurance is supposed to have.

Second, and this one trips up more plan documents than people expect: a clause giving the employer sole discretion to terminate the plan, at any time, for any reason, is a disqualifying factor.

The logic tracks straight back to the contractual obligation requirement covered earlier. An obligation that the obligated party can cancel whenever it feels like it isn't really an obligation. If an employer can walk away from the plan unilaterally and without cause, the binding commitment that made the arrangement look like insurance in the first place evaporates. Plan documents need termination clauses that reflect a genuine, binding commitment rather than unconstrained employer discretion. Unconditional employer exit rights are the opposite of what CRA is looking for.

The sole shareholder-employee problem and the employment-capacity requirement

Small corporations run into a specific wrinkle when the only person covered by the plan also happens to own the company. CRA then has to figure out whether that person is receiving the benefit as an employee, or as a shareholder collecting value from the business they own. The distinction matters enormously for tax treatment.

CRA document 2005-0163771E5 says these plans can work, but only under conditions. The benefit has to flow from employment. CRA's general approach: if the individual is genuinely working as an employee, and the plan's limits and coverage are reasonable, in line with what an arm's length employee doing similar work would get, the benefit gets treated as employment-derived rather than a shareholder perk.

A later interpretation, 2014-0521301E5, revisited the sole shareholder-employee question. Where the sole shareholder is also the sole employee, establishing that benefits flow from employment rather than ownership requires meaningful supporting evidence, and CRA's scrutiny in this scenario is real.

How employee class design affects PHSP status under non-discrimination rules

PHSP coverage must be structured around defined employee classes with a legitimate employment basis, job function, seniority, executive rank, that kind of thing, rather than shareholding status.

What matters is that the classification itself rests on employment criteria rather than who owns equity in the company.

The risk appears when a plan's design, viewed as a whole, suggests that benefits are flowing from ownership rather than from employment, which pulls the arrangement back into the sole shareholder-employee issue described above.

Special constraints for sole proprietors and unincorporated businesses

Unincorporated businesses face a hard limit that corporations don't. A sole proprietorship can't set up a self-insured PHSP simply by making direct payments from the business owner to employees. That structure doesn't qualify, no matter how the paperwork is worded.

An unincorporated business has two real paths available. One is a licensed insurance plan, individual or otherwise. The other is a cost-plus arrangement run through a third-party administrator, though the availability of that option depends on the structure and arm's length character of the employer-employee relationships involved.

That distinction matters a lot for solo operators. A sole proprietor without arm's length employees faces significant constraints in accessing a self-insured cost-plus structure, since the insurance elements are difficult to establish without a genuine employer-employee relationship involving someone other than the owner.

What a compliant PHSP plan document needs to accomplish

None of this lives in theory. CRA's technical interpretations treat the plan document itself as the primary evidence that each required element, insurance character, contractual obligation, non-discrimination, all actually exist. A plan that operates correctly in practice but says the wrong thing on paper is still a plan that can lose its tax-free status.

A document built to hold up needs several things in place at once. It needs a clear, binding statement of the employer's obligation to reimburse defined medical and hospital expenses, written as a commitment, not phrased as something the employer may choose to do. It needs covered expense categories defined precisely enough to support the 90% METC-eligible calculation, since a vague expense list makes that math impossible to verify. It needs annual spending limits organized by employment-based classes.

It needs a carry-forward clause that respects the 12-month cap, covering either flex credits or unused expenses but not both, or else a straightforward use-it-or-lose-it provision, paired with an explicit statement that funds never revert to the employer under any circumstance outside the narrow notional-credit exception. It needs a termination clause that reflects a genuine, binding commitment rather than unconstrained employer discretion. And for any corporation where a shareholder is also an employee, the document needs enough detail on duties, comparable arm's length benefits, and reasonable limits to support an employment-capacity argument if CRA ever asks the question.

Get each of these pieces into the document itself, in writing, and the plan has a real shot at holding up under CRA's technical interpretations. Leaving any one out means the plan is running on assumption rather than proof, which is the gap CRA's own guidance is built to close.