A total score lets a candidate trade a bad station for a good one. A conjunctive rule stops some of that trade. The three commonly used are a minimum number of stations passed, a must-pass station, and a critical error. Each one adds a second way to fail the exam. Each one needs a number, and the number needs a reason.
What compensation buys and what it costs
Summing station marks treats the OSCE as one long test. That is what makes it reliable. Wide sampling across cases is the whole point of the format. A candidate who does well across twelve stations has shown something a single station cannot show.
The cost is that a total can hide a hole. A candidate can pass on strong communication stations while failing every practical one. If the blueprint says the exam certifies both, the total alone has not tested what it claims. Homer and Russell put the designer's case in exactly those terms. Concerns about compensation come from three places. The complexity of what the OSCE covers, the blueprint, and the limits of the scoring sheets themselves.
So the choice is not compensation against no compensation. It is how much trade the exam allows, stated in advance and on purpose.
The three rules
Rule | What it says | The number to justify | What it catches |
|---|---|---|---|
Minimum stations passed | Pass the exam total and pass at least N stations | N | Broad weakness spread across the circuit |
Must-pass station | Fail this station and fail the exam, whatever the total | Which stations, and how many | A specific skill the program will not trade away |
Critical error | A named act inside a station fails that station, or the exam | The list of acts, and their effect | Behavior that would harm a patient |
The three are independent, and they stack. A candidate has to clear the total, then the station count, then every must-pass station. Writing them in that order in the policy makes the arithmetic easier to explain later.
The wording of the third rule belongs to the critical errors and must-pass policy, which is written with the stations. This page is about the numbers the first two need.
Where does the number come from?
Usually from habit. Homer and Russell found the rule to be common. Its stated purpose is to prevent excessive compensation. They also found little discussion of the practice in the literature. Homer's follow-up study is blunter. The number is typically fixed in advance with little justification and does not vary between sittings, which is not how a high-stakes standard should behave.
The alternative is to derive it from the same borderline candidate the pass mark already describes. Homer studied 442 administrations of one large clinical exam and reported two useful findings. First, the number of stations a candidate passes is almost as reliable a measure as the total score. The two coefficients were 0.73 and 0.76. So the count is sound enough to carry a rule. Second, one of the four methods tested came out best. Predict how many stations the borderline candidate would pass at the exam-level cut score. Use that as the minimum.
That number moves between sittings, because station difficulty moves. In the exam he studied, the rule operated in a secondary way: about a quarter of the candidates who failed the exam failed because of the stations-passed rule alone.
How to set it in practice
- Decide first whether you want the rule at all. If the blueprint and the must-pass stations already protect what you care about, a station count adds little.
- Fix the principle before the sitting, not the number. The policy says the minimum equals the stations the borderline candidate is predicted to pass. The minute after the exam records the number that produced.
- If the school insists on a fixed number, look at what it does in past sittings. Count how many candidates it failed who passed the total. If the answer is nobody, the rule is decoration. If it is a quarter of the cohort, it is the real standard and should be described as such.
- Say what happens next. A candidate who passes the total but fails the count has failed. The retake rules decide what they sit again.
- Publish it. Candidates should know the count before they sit, not from a results letter.
Does a second rule fail someone twice?
This is the appeal argument, and it has a short answer. A candidate is not failed twice for one performance. They fail once for one of two reasons: too little overall, or too little in too many separate encounters. The policy has to say which reason applied.
Two practices make that defense hold. Keep the rules independent of each other, so no single station decides the exam through two routes at once. And record the count on the result, not just the outcome. Say the letter reports a score above the exam mark, and nine stations passed against a minimum of eleven. That can be checked. A letter saying "fail" cannot.
The other honest limit is the size of the effect. A stations-passed rule shifts a small number of results, and a must-pass rule can shift one. That is not an argument against them. It is an argument for knowing the size first. Run the rule over last year's data before adopting it.
Limits
None of these rules improves a weak station. They make it more expensive. Take a station with a badly set mark that is also must-pass. It will fail candidates for its own faults, and no total can rescue them. Check the station metrics of every must-pass station before the board signs the results. Expect a higher bar for the ones that carry a rule.
There is also no agreed value for N in the literature. Anyone who quotes one is quoting a local policy. The defensible answer is a stated method for producing it, recorded in the decision record.
Frequently asked questions
What does minimum stations passed mean in an OSCE?
It is a rule that a candidate must pass at least a set number of stations and reach the exam pass mark. It limits how far a strong station can compensate for a weak one. It is also called a conjunctive standard.
How many stations should a candidate be required to pass?
There is no published number to copy. The most defensible method is to predict how many stations the borderline candidate would pass at the exam-level cut score and use that. The value then changes from sitting to sitting as station difficulty changes.
Is a must-pass station the same as a critical error?
No. A must-pass station fails the whole exam if the candidate fails that station. A critical error is a named act within a station, such as an unsafe action, that can fail the station or the exam, depending on the policy. A station can carry both.
Does a conjunctive rule fail a candidate twice for one performance?
No. The candidate fails once, for one of two separate reasons: too little in total, or too little across too many stations. The result should record which rule applied and the count that triggered it, so the decision can be checked.
How do you know whether a conjunctive rule is doing anything?
Run it over previous sittings and count the candidates it would have failed who passed the total. If it is nobody, the rule is a decoration. If it is a large share of failures, it is the real standard and should be published as such.