10 Healthcare Metrics Every Self-Funded Employer Should Track

Self-funding transfers the risk from an insurer's balance sheet to yours. Once you are paying the claims, the difference between a plan you manage and a plan that manages you comes down to whether you can see what is happening inside it while there is still time to act. The question is no longer whether to examine your data, but which numbers deserve a standing place on the report.

The ten self-funded health plan metrics below form a working scorecard. Each entry gives the definition and the signal that should prompt a conversation with your broker or TPA.

Cost and financial exposure

Cost is where a scorecard has to begin. Sixty-seven percent of covered workers are now in self-funded plans, rising to 80% at large firms, average family premiums reached $26,993 in 2025, and renewals are being quoted near 9.1% for 2026. Most employers are absorbing that trend directly.

1. Total cost of care PMPM. Total plan spend divided by member months, reported with medical and pharmacy split out. Use PMPM rather than aggregate spend so headcount changes do not disguise the trend. Watch for pharmacy growing faster than medical, which is now the common pattern.

2. Plan paid versus member paid share. The percentage of allowed charges the plan covers against what members pay through deductibles, copays, and coinsurance. A rising member share can flatter your PMPM while quietly suppressing necessary care, so read this alongside utilization rather than on its own.

3. High-cost claimant count and concentration. The number of members exceeding a fixed threshold, typically $100,000, and the share of total spend they represent. Industry-wide, 5% of members generate 56% of spending and the top 1% account for 28%. A concentration far above that range signals either a real severity problem or a data issue worth resolving.

4. Large claim severity against your stop-loss attachment point. The distribution of your largest claims relative to where reinsurance begins. Claims clustering just beneath the attachment point mean you are absorbing severity that your policy was priced to cover, and that pattern should inform your renewal negotiation.

Utilization and site of care

5. Avoidable emergency room rate per 1,000 members. Non-emergent emergency department visits, normalized per 1,000 members so quarters remain comparable as enrollment moves. The economics are stark. UnitedHealth Group put the average emergency visit at $2,032 against $167 in a physician's office and $193 at urgent care, which makes redirection one of the few levers that improves cost and experience together.

6. Thirty-day readmission rate. The share of inpatient discharges followed by an unplanned readmission within thirty days. Readmissions usually point to a breakdown in discharge planning or follow-up rather than clinical severity, which means they are among the more addressable items on this list.

7. Network leakage. The proportion of spend flowing to providers outside your preferred network or direct contracts. Leakage tends to be invisible in a standard carrier report and is one of the clearest cases where plan design, not renegotiation, is the remedy.

Price

8. Price paid for identical services. Your allowed amount for the same procedure across facilities, ideally expressed as a percentage of Medicare. Private plans reimburse hospitals at 254% of Medicare on average, with state-level figures ranging from under 170% to above 300%, and your internal spread across facilities is usually wider than that.

Pharmacy

9. Generic dispensing rate and specialty share. The percentage of prescriptions filled as generics, alongside what specialty and brand drugs consume. The concentration is severe: brand and specialty medications are 14% of prescriptions filled but 87.6% of pharmacy spend, while GLP-1s alone reach 20.3% of prescription spend at roughly $7,400 per member annually. Pharmacy overall has climbed to 29.5% of total claims.

Clinical

10. Care gap closure rate. The share of members with an open, evidence-based gap who close it within the year, covering diabetes monitoring, blood pressure control, and cancer screening. This is the only metric on the list that predicts next year's cost rather than explaining last year's, which is precisely why it tends to get dropped from the report.

Making the scorecard trustworthy

Employer health plan metrics only mean something if the underlying data is handled properly, and three habits separate a scorecard you can act on from one you argue about. Risk-adjust before comparing yourself to any benchmark, since an older workforce will always look expensive in absolute terms. Allow at least three months of claims run-out before drawing conclusions from a closed period. And insist that medical, pharmacy, eligibility, and laboratory data resolve to one member record, or the same person will appear several times and every count you produce will be wrong.

Frequently asked questions

How many metrics should a self-funded employer actually track? Ten self-funded health plan metrics make a workable standing scorecard. Reviewing PMPM, high-cost claimants, and utilization monthly, with the full set quarterly, gives you enough signal to act without generating reports nobody reads.

What is the single most important health plan metric? Total cost of care PMPM, because it captures everything else in one number. It tells you the direction of travel but never the cause, which is why the other nine exist.

Can we get these metrics from our TPA? Some of them. TPA reporting usually covers spend and basic utilization well. It rarely supports price comparison across facilities, network leakage, or care gap tracking, since those require joining claims to eligibility, pharmacy, and clinical data.

Where to start

A scorecard is only as good as the data feeding it. Health Compiler consolidates claims, eligibility, pharmacy, laboratory, and EHR data into a single member record, then reports cost, utilization, price, and quality from one source. Schedule a call to see these ten metrics built from your own plan data.