Pay equity analysis: a step-by-step guide for health systems

What a pay equity analysis is (and how it differs from benchmarking)

A pay equity analysis looks inward. It compares employees with each other inside your organization. Compensation benchmarking looks outward. It compares your pay with the external market. You need both, and they answer different questions.

The two connect. Ranges built from a stale benchmark push new hires above incumbents, and that compression often shows up later as an internal equity problem. For the outward-looking half, see compensation benchmarking, step by step.

The legal context, at a high level

Several overlapping laws shape why and how organizations run these reviews. In broad terms:

• Equal Pay Act of 1963 (federal). According to the EEOC's guidance on equal pay, the Act requires equal pay for men and women doing substantially equal work in the same establishment, judged by skill, effort and responsibility under similar working conditions. Job content decides this, not job titles. Differences are allowed when based on seniority, merit, quantity or quality of production, or a factor other than sex, and the employer has to prove the exception applies. The EEOC also notes that when a gap is corrected, no employee's pay may be reduced to do it.
• Title VII of the Civil Rights Act, the ADEA and the ADA (federal). The EEOC explains that these laws prohibit pay discrimination based on race, color, religion, sex, national origin, age or disability, and unlike the Equal Pay Act they don't require the jobs to be substantially equal.
• State pay equity laws. Many states go further. The California Equal Pay Act (Labor Code section 1197.5), as summarized by the California Department of Industrial Relations, covers sex, race and ethnicity, uses a “substantially similar work” standard, and says prior salary can't justify a disparity. Massachusetts' Equal Pay Act, per the state Attorney General's guidance, gives employers an affirmative defense if they completed a good-faith, reasonable self-evaluation within the previous three years and made reasonable progress on any gaps it found.
• Pay transparency and reporting laws. Several states now require pay ranges in job postings. California requires employers with 15 or more employees to include the pay scale in any job posting (California DIR), and Colorado's Equal Pay for Equal Work Act requires a pay range and a general description of benefits in postings (Colorado Department of Labor and Employment). California also requires private employers with 100 or more employees to file annual pay data reports (California Civil Rights Department). Published ranges make internal gaps far easier for employees to spot.

Step 1: Define scope and protected categories

Decide up front, in writing, what the analysis will cover:

• Population. Which entities, facilities and employee groups. Most systems start with employed clinical staff, then extend to non-clinical, leadership and physicians in later cycles.
• Protected categories. At minimum sex and race/ethnicity. Add others your state laws name, and age if you have the data and counsel agrees.
• Pay elements. Base hourly rate is the core. Decide separately whether to test differentials, bonuses, sign-on and retention payments, and starting pay at hire.
• Snapshot date. One effective date for every record, ideally before your merit cycle so findings can feed it.
• Exclusions. Agency and travel clinicians are not your employees and their rates reflect a different market. Leave them out and document why.

Step 2: Build comparable-work groups

Every comparison happens inside a group of employees doing comparable work. Get this wrong and the rest of the analysis is noise. Groups that are too broad blend jobs that legitimately pay differently. Groups that are too narrow leave too few people to compare. In healthcare, these distinctions usually matter:

• Job family and credential. RN, LPN/LVN and CNA are different jobs, even on the same unit. So are APRNs, PAs and RNs. Group by licensed role, not by department.
• Unit and specialty. ICU, ED, OR and med-surg nurses often sit on different pay structures. Whether they belong in one group or several depends on whether the work is genuinely comparable in skill, effort and responsibility.
• Union vs non-union. Represented staff are paid on a collectively bargained scale. Analyze them separately from non-union staff, and inside the union group check that step placement follows the contract consistently.
• Employment type. Per diem rates are often set higher in place of benefits and guaranteed hours, so per diem and benefited staff belong in separate groups. See what per diem means in healthcare.
• FTE status. Compare hourly rates, not annual earnings, so part-time and full-time staff can share a group. Keep FTE as a field to check whether it correlates with rate.
• Shift. If night and weekend premiums are paid as differentials, shift shouldn't change the base rate. If some nurses have a higher base because they work nights, document that and treat it as a structural question.

Write down the grouping logic with the reason for each split. An HR, compensation and nursing leader should be able to read it and agree that the people in each group do comparable work.

Step 3: Assemble clean data

The most common data mistake in healthcare is analyzing total hourly earnings with differentials mixed in. A nurse who works every night and weekend will look “overpaid” on total pay even if her base rate is low. Test the base rate, and test differentials on their own. Use this checklist for each employee on the snapshot date:

Identity and group

• Employee ID (pseudonymized for the analysis file)
• Job code, job title and job family
• Comparable-work group (from step 2)
• Facility, department and unit
• Union status and bargaining unit; contract step where applicable
• Employment type (staff, per diem, PRN pool) and FTE

Pay

• Base hourly rate (annual salary converted at the scheduled hours)
• Pay range minimum, midpoint and maximum for the job
• Shift, weekend, charge, preceptor and certification differentials, each as a separate field
• Starting rate at hire and date of last increase
• Sign-on, retention and incentive payments in the past 12 months

Legitimate pay factors

• Hire date and time in current job (tenure)
• Total years of relevant licensed experience, including prior employers
• Highest relevant degree (for example ADN vs BSN) where your pay policy uses it
• Specialty certifications (for example CCRN, CEN) where your pay policy uses them
• Clinical ladder or career level
• Most recent performance rating, if ratings drive pay

Protected characteristics

• Sex, race and ethnicity as self-reported in your HRIS
• Date of birth or age band, if in scope

Only include a factor if it is actually part of how you set pay, and if it is recorded consistently. Prior-experience credit entered by hand at hire is often the messiest field, so audit a sample against offer letters before you rely on it.

Step 4: Descriptive analysis (raw gaps and compa-ratios)

Start simple. For each comparable-work group, compare average and median base rates by protected category, and calculate compa-ratio (pay ÷ range midpoint) for each group. This shows where gaps exist before you try to explain them.

Illustrative: a group of 140 non-union staff RNs on acute care units.

A raw gap is a signal, not a conclusion. It may be fully explained by legitimate factors, or it may hide a larger problem in one unit. For more on reading compa-ratios, see compa-ratio explained, or calculate them with the compa-ratio calculator.

Step 5: Regression analysis, in plain English

A regression answers one question: if two employees in the same group had the same experience, unit, certification and so on, would their pay still differ by sex or race? It estimates how much each legitimate factor moves pay, then reports what is left over. That leftover difference is the adjusted gap, sometimes called the unexplained gap.

Which controls are legitimate?

• Usually legitimate: relevant experience, tenure, credential, certification, unit or specialty, clinical ladder level, union step, and performance, provided your pay policy actually uses them and they are applied consistently.
• Handle with care: starting pay, prior salary and “market adjustment” flags. These can carry the very bias you're testing for. California, for example, says prior salary can't justify a sex, race or ethnicity-based disparity. Ask counsel before including them.
• Not a control: anything that isn't a real pay rule, like “negotiated harder” or a manager's discretion that isn't documented.

What “statistically significant” means

A statistically significant adjusted gap is one that is unlikely to be random noise given the number of people in the group. It doesn't prove discrimination, and it doesn't say why the gap exists. It says the gap is real enough to investigate. The reverse also holds: a gap that isn't significant in a small group may simply mean there are too few people to tell, not that pay is fair.

You don't need to run the model yourself, but you should be able to explain the controls and the result to a nurse leader in two minutes. If you can't, the model is too complicated to defend.

Step 6: Cohort review for small groups

Many clinical jobs have only a handful of incumbents in a facility: perfusionists, a small cath lab team, a single wound care nurse. Regression can't say much about them. Review these groups person by person instead:

1. List everyone in the comparable-work group, sorted by base rate.
2. Next to each person, show the legitimate pay factors: experience, tenure, credential, step or ladder level.
3. Flag anyone whose pay position doesn't match their factors, for example a nurse with more experience and the same certification paid below a newer colleague.
4. For each flag, find the documented reason. If there isn't one, treat it as a candidate for adjustment.

Cohort review is also the right way to look at the outliers a regression flags in larger groups. A model points you to people; the review decides what to do about each one.

Step 7: Remediation

Findings without money behind them create risk, not equity. Plan remediation in three parts.

Size the adjustment budget.

The usual approach is to raise each flagged employee to at least the pay the model predicts for them, or to a defined point in the range, and add up the cost. Run a few scenarios (for example, closing all unexplained gaps vs only significant ones) so leadership can choose with the cost in view. Remediate by raising pay, not by cutting the higher-paid group. Under the Equal Pay Act, no employee's pay may be reduced to correct a differential.

Prioritize.

• Largest unexplained gaps first, and anyone below the range minimum.
• Groups where the adjusted gap is statistically significant.
• Off-cycle adjustments for the most serious cases rather than waiting for merit.

After adjustments, re-run the model to confirm the gaps closed. Keep equity adjustments separate from merit increases in your records so each is visible.

Prevent recurrence.

Most gaps come from a few places: discretionary starting pay, inconsistent experience credit at hire, and new-hire rates that rise faster than incumbent pay. The fixes are structural:

• Clear salary bands with written rules for placement in the range at hire.
• A consistent experience-credit table for licensed clinical roles.
• A plan for pay compression in nursing, so market adjustments for new hires also reach experienced staff.
• Merit guidelines that are reviewed for equity before increases are approved. See merit increases in healthcare.
• An equity check on every offer outside the normal range placement rules.

Step 8: Documentation and cadence

Record enough that someone could repeat the analysis next year and get the same answer:

• Scope, snapshot date and exclusions
• Comparable-work groups and the reason for each split
• Data sources, cleaning steps and any fields you corrected
• Controls used, controls rejected and why
• Results by group, flagged individuals, adjustments made and their cost
• Process changes adopted to prevent recurrence

Agree with counsel which documents sit under privilege and which are shared more widely. On cadence, most organizations run a full analysis once a year, timed before the merit cycle, and a targeted check after events that move pay quickly: a new union contract, a round of market adjustments, an acquisition, or a new pay structure.

Pay equity work also depends on having current ranges. If a role's range is out of date with the market, new hires will be placed above it and internal gaps will grow. ClinicalRate for hospitals and health systems gives compensation teams current pay rates for 400+ clinical roles by state and metro, refreshed every business day, to keep those ranges anchored to the market.

Frequently asked questions