Two of the five classification variables are not clinical
Every 30-day period under the Patient-Driven Groupings Model is classified on five things: admission source, timing, clinical grouping, functional impairment level, and comorbidity adjustment. Clinical grouping comes from the principal diagnosis. Functional level comes from OASIS items. Comorbidity comes from the secondary diagnoses on the claim.
Admission source and timing are different. Neither is a judgment about the patient. Admission source is a question about where the patient was receiving care in the 14 days before home health took over. Timing is a question about where this 30-day period sits in the sequence of periods the agency has already billed. Both are answered from claim data, which means both are wrong for administrative reasons when they are wrong at all — a date entered from the referral fax instead of the hospital discharge summary, a gap nobody noticed, a template that carried last period's answer forward.
That matters because these two variables set the first character of the HIPPS code: 1 for community early, 2 for institutional early, 3 for community late, and 4 for institutional late. A period grouped on the wrong combination is paid on the wrong case-mix weight, and the error is invisible in the chart because the chart was never the source.
Admission source: a 14-day lookback, not a referral source
A period is classified as institutional when the patient received acute or post-acute inpatient care in the 14 days before the home health admission date — the from date on the claim. Acute means an inpatient acute care hospital. Post-acute means a skilled nursing facility, inpatient rehabilitation facility, long-term care hospital, or inpatient psychiatric facility. If none of those applies, the period is community.
The detail that causes most of the errors is that the lookback runs against a calendar, not against an intention. A patient discharged from a skilled nursing facility 20 days before start of care is a community admission, however recently the agency received the referral. A patient discharged from an acute hospital nine days before start of care is institutional even if the referral came from a physician who saw them at home afterwards.
So the evidence for this field is a discharge date on a facility document — the hospital discharge summary, the SNF discharge notice, the IRF paperwork. Intake work that records the referral date, the fax date, or the date the order was signed is recording the wrong date, and any of the three can sit on the wrong side of a 14-day line.
Timing: only the first period in a sequence is early
The first 30-day period in a sequence of adjacent periods is early. Every period after it is late. A sequence continues for as long as there is no gap of at least 60 days between the end of one period and the start of the next, so a patient who stays on service for months generates one early period and a run of late ones.
This is where agencies misread the model. Under the episode-based system that preceded PDGM, early meant the first and second 60-day episodes. Carrying that habit into PDGM produces a second early period that does not exist, and since early carries more weight than late, the error is not symmetric — it inflates one period and misprices every period after it if the sequencing logic is rebuilt incorrectly.
The mirror-image error is just as common and costs the agency rather than the payer: a discharge and readmission treated as a fresh start when the gap was shorter than 60 days. Reopening a patient after a two-week break does not restart timing. The period continues the sequence, which means it is late, and a claim built as early has to be corrected or it will be adjusted downstream.
Late periods default to community, with one exception
This is the rule that catches people who have learned the 14-day lookback but not the interaction. A late period is always classified as community unless there was an acute hospitalization in the 14 days before that late period. A post-acute stay in those 14 days does not make a late period institutional unless the patient had been discharged from home health before the post-acute stay began.
Read that twice, because it is the difference between two similarly documented claims. A patient on service who goes into a skilled nursing facility for rehabilitation and comes home three weeks later is, in most cases, a community late period — the SNF stay is a post-acute event and the patient was never discharged from home health. The same patient, formally discharged before the SNF admission and reopened afterwards, can be institutional.
There is an operational choice buried in that rule. Medicare allows the agency to decide whether to discharge a patient during a short hospitalization. That decision changes admission source for the following period as well as the certification trail behind it, so it should be made deliberately and recorded, not left to whoever processes the transfer.
Where the classification goes wrong in practice
The failures repeat, and almost all of them are process failures at intake or in the billing system rather than clinical errors.
The lookback date is the referral date. Intake records when the referral arrived. If the hospital discharged the patient two weeks before that, the 14-day window was missed and an institutional period was billed as community.
Discharge and readmission are treated as a new sequence. The patient came off service, then returned. If nobody checks the number of days between the end of the last period and the new start of care, a late period is billed as early.
A hospitalization mid-period is handled inconsistently. One coordinator discharges, another keeps the patient open. Both can be correct, but the admission source on the following period depends on which was done, and a mixed practice produces a mixed — and unauditable — claim file.
The electronic record carries the previous period forward. Admission source is a field in most systems, and a field with a default is a field that is sometimes never deliberately answered. The result is a community period for a patient who came straight from a SNF.
Nobody recomputes the grouping. The claim groups, the payment posts, and the first character of the HIPPS code is never read back against the dates. An error found here costs a correction; an error found in a post-payment review costs a repayment and an error-rate finding.
What to audit, and how cheaply
Because both variables are administrative, they can be audited without a clinical review — which makes them the cheapest accuracy work in the revenue cycle.
Start with the claims already billed. Pull every institutional period and compare the from date on the claim with the facility discharge date in the record: is the discharge inside 14 days, and is the document the facility's rather than the referral's? Then pull every early period and confirm the period genuinely begins a sequence — no earlier period for that patient ended within 60 days of this start of care.
Next, pull every late period with an institutional admission source, because that combination is the one the default rule restricts. Each should carry an acute hospitalization in the 14 days before the period, not a post-acute stay. Finally, recompute the first character of the HIPPS code from the evidence and compare it with what was submitted. Any mismatch is either a billing error or a data error, and both are fixable the same week.
Sampling on volume alone will not find these. Sample by exception: admissions from facilities, readmissions, transfers mid-period, and any period where the admission source field was left at its default.
Why fixing it is worth the small effort
Functional scoring and comorbidity capture both need clinicians and coders to change how they write and read a record. They are worth the effort and they are slow to move. Admission source and timing need a date field, a discharge-timing decision, and a check in the billing system. That is a workflow fix, and it holds once it is made.
The measurement is equally simple. Track the share of periods billed institutional, the share billed early, and the exception rate from the audit above against your own history. A number you can explain beats a benchmark you cannot source, and in this case the explanation is visible on the claim.
Two of the five PDGM variables are free money to get right. The other three are where the real revenue work is — see how the comorbidity adjustment is decided for the diagnosis side, and OASIS Section GG scoring for the functional side.
Bottom line
Admission source and timing are the two PDGM variables an agency controls with paperwork rather than clinical judgment, which makes them the fastest place to start an accuracy program. Get the 14-day lookback from the facility discharge date, keep the period sequence honest against the 60-day gap rule, remember that late periods default to community, and recompute the first character of the HIPPS code on a sample every month.
The complete PDGM guide sets both variables inside the full grouping model, and revenue cycle management is where the intake and billing controls that keep them accurate are run end to end.