Executive guide · Length of stay

Hospital length of stay: how to read LOS against expectation

Executive evidence mapInterpret LOS through separate lenses
01Observed LOS
02Case mix
03Throughput
04Discharge timing
05Placement and access
06Data coverage
The lenses can guide investigation; none alone proves why observed LOS differed from expectation.

Start by defining both sides of the comparison

Observed LOS needs an encounter cohort and a consistent start-and-end convention. The expectation also needs a name. If the reference is the CMS geometric mean LOS associated with an MS-DRG, state the CMS fiscal-year table used and report what share of encounters matched an eligible DRG. A health-system model or another benchmark has different assumptions and should be labeled accordingly.

The phrase expected LOS is not self-defining. A comparison can look precise while mixing fiscal years, unmatched encounters, different service-line definitions, or changed discharge windows.

Keep associated lenses distinct

Case mix may change the expected resource profile. Throughput measures may reveal where process time accumulates. Discharge timing can show weekday, time-of-day, placement, transport, or coordination patterns. Documentation and coding can affect classification and coverage. Each lens can refine the next question without being treated as the cause.

  • Case mix: did the distribution of valid, fiscal-year-matched MS-DRGs change?
  • Throughput: where is elapsed time accumulating across defined operating milestones?
  • Discharge timing: do patterns differ by day, time, service, or documented barrier?
  • Coverage: did the matched denominator or missing-data pattern change?

A clearly synthetic example

Suppose a synthetic medical cohort has an observed mean LOS of 5.1 days and a selected matched benchmark of 4.6 days. The 0.5-day difference is a description, not a diagnosis. Leadership should first inspect cohort stability, matching coverage, fiscal-year alignment, and distribution—not immediately attribute the difference to discharge planning or staffing.

The strongest output is a bounded investigation plan: which stratification to review, what additional evidence is needed, who owns the question, and when the result returns to the operating review.

What leaders should ask next

  1. Which encounters are in the denominator, and how is LOS calculated?
  2. What does expected mean here, and which fiscal-year reference is in use?
  3. What percentage of the cohort has a valid matched benchmark?
  4. Did case mix, coverage, or data completeness change before the reported movement?
  5. Which associated lens should leadership investigate next without assuming causation?

Limits to keep visible

  • A difference between observed LOS and a benchmark is not itself avoidable time or proof of inefficiency.
  • CMS geometric mean LOS is a reference attached to an MS-DRG table, not a patient-level prediction.
  • sanalytics does not promise that its use will reduce length of stay.

Sources and method notes