Coding as we know it

[Silvestro]

Carl Cottrell June 2007

Coding is a process that has remained mostly unchanged from 1662 when
"...antient Matrons sworn to their office" coded deaths using the sixty-
three codes specified in the London Bill of Mortalities. What follows
here is a brief introduction to coding as it is currently performed in
US hospitals and physician offices.

What is coding? Stripped of all pretenses, coding is a knowledgeable
judgment. The coder is asked to read patient records and generate
diagnoses and procedures that accurately describe the totality of the
patient's clinical experience.

Coding is translation. An inpatient record is typically one-hundred
or so pages. The 350,000 or so unique clinical concepts that might be
present in the patient record are translated into the 16,000 possible
ICD-9-CM codes that describe diagnoses and procedures. Their
translation must be consistent with the volumes of rules that govern
the use, order and combinations of codes.

Coders must glean and translate these codes from documents created by
physicians who steadfastly refuse to speak ICD-9-CM and sometimes act
as if creating accurate and complete documentation were not the most
important use of their time.

How much is spent on coders?
There are about 63,000 coders in the US in all settings. Together,
about $1.5 billion dollars is spent on coders - one-tenth of one-
percent of the entire healthcare expenditure.

Coding effects payment.

Coding, in turn, determines payment. The ~1.2 trillion dollars spent
annually on US Healthcare is almost entirely based upon coding.
Approximately 1.5 billion dollars are spent annually on coders who
manually read medical documentation and summarize the patient's
clinical experience into ICD 9 CM and HCPCS codes that are then
submitted on the patient's bill. The extent to which the payment is
correct is totally determined by the quality and supportability of
coding.

Coding controls how soon payment is received (cash-flow)

Coding 2 days sooner could save 100 bed hospital $20K/yr Billing waits
for coding. Importantly, coders are in short supply and they have a
limited capacity. The number of discharges may vary greatly from day
to day, but the HIM department's FTE capacity is fixed. If, for
reasons of extraordinary volume or lack of coder availability, the
coding staff gets far behind, it is extremely difficult for them to
catch up.

Typically, hospital billing waits 5 to 10 days for coding. At 5 days
hospitals taken together lose an unnecessary $500 million because of
this delay in billing. Cost for physician offices is similar.

A single day of unbilled revenue in a 100 bed hospital ties up about
$200K in cash. Coding two days sooner would save this hospital almost
$20K a year in interest expenses alone.

Coding alone determines payment.

Several prospective payment systems in health care - all based upon
coding.
Healthcare facilities are paid under multiple prospective payment
System (PPS) approaches by the Center for Medicare and Medicaid
Services (CMS) and even more by individual state agencies and private
payers.

Medicare inpatients are paid under the IPPS, the Inpatient Prospective
Payment System based upon Diagnosis Related Groups (DRG). Medicare
Emergency and Ambulatory Surgery patients are paid under OPPS, the
Outpatient Prospective Payment System based upon Ambulatory Payment
Classification System (APC). Medicare Inpatient Rehabilitation
Facilities are paid under IRFPPS. Skilled nursing is paid under
Resource Utilization Groups (RUG). Physicians are paid under the
Relative Value Resource Based System RBRVS. CHAMPUS, the payment
system for active duty military people and their dependents has their
own DRG system as do several states. Payment under all of these
approaches is based upon the coding of records.

Case-mix is the average DRG weight

Optimization effects case-mix

CC capture effects case-mix

A Medicare inpatient can be placed in one of 516 different payment
groups based upon their principal diagnosis, additional diagnoses,
procedures, age, disposition and gender. Each group has a weight
indicating the expected payment. For example, the lowest paying group
- ALLERGIC REACTIONS AGE 0-17, has a weight of 0.0981 and pays about
$400. The highest paying group, HEART TRANSPLANT, with a weight of
18.6081 pays close to $75,000. Case-mix is the average Diagnosis-
Related Groups (DRG) weight billed by the hospital. If these were the
only two patients this hospital treated, their case-mix index would be
9.3531 and their average payment would be about $37,500. It is
unlikely that even a non-coder would confuse an allergy with a heart
transplant; case-mix perfection results from much more subtle coding
differences. Two large predictors of case-mix are diagnosis
optimization and complications and co-morbidities (CC) capture.

Diagnosis optimization occurs when there are multiple diagnoses that
can equally explain the reason for the patient's admission to the
hospital; the patient can legitimately be assigned two different
DRGs. Any time the coder chooses the lesser paying DRG they reduce
the hospital's case-mix.

CC capture indicates the presence of complications and co-morbidities,
conditions the patient has in addition to the primary reason for
admission that initially brought the patient into the hospital for
treatment. Patients with certain complications require greater
resources to treat and earn higher PPS payments than those who don't.
Approximately one-half of the DRGs are split between those with and
those without complications. The "With CC" group pays, on average, 82%
more per case than the average "Without CC" group or about $2,800 per
case more. Where CC's exist and coders find them, case-mix improves.

A 1% increase in case-mix means $125K to a 100 bed hospital. A 1%
increase in case-mix means that, for the same effort and expense, the
hospital receives 1% more revenue. In the case of a typical 100 bed
hospital this 1% increase generates a bottom line increase of $125K.

Coding controls whether payment is retained.

Unsupportable coding costs a 100 bed hospital $100K/year All payors
conduct retrospective audits of their claims that may result in denial
of payment. Medicare denies about 5% of all claims worth close to $20B
in net revenue. Private payors can be even more aggressive. Almost
half (45%) of denials are for insufficient documentation to support
coding.

Coding determines whether or not the facility is sanctioned by
regulators. Coders and healthcare providers feel the hot breath of the
Office of the Inspector General (OIG) on their necks while they code.
This is the agency charged with ferreting out fraud and abuse in
Medicare payments and turning cases over to the FBI for prosecution.
For the first half of Fiscal Year (FY) 2004, the OIG reported savings
of over $16.8 billion, mostly from implemented recommendations, in
addition to: exclusions of 1,544 individuals from participation in
government healthcare programs for fraudulent activities, 234
convictions, 107 civil actions for False Claims Act and unjust
enrichment offenses resulting in $995 million in settlements, and a
number of administrative recoveries.

Coding quality is questionable. Almost all facilities, when asked,
will testify that the quality of their coding is superb. But, just
how good is coding? There has been little published that
dispassionately analyzes the quality of coding. What studies there are
generally concluded that coding quality and, with it, case-mix has
room for improvement. Three old studies by the VA and the Institute
of Medicine (1977-1980) found only 2/3's agreement on the coding of
medical records and even expert coders disagreed on patient principal
diagnoses 19% of the time. More recently, three studies conducted
under AHRQ grants in the year 2000 found only 84% accuracy in coding
twenty eight complications. These same studies found physician
collaboration of selected complications in only 68.4% of surgical
cases and 29.5% of medical cases with coded complications. William
Beaumont Hospital and Wayne State University found poor agreement on
Evaluation and Management coding of emergency department records. They
found that two coders agreed on the E&M coding of a set of charts less
than 40% of the time. Recall that there are only five E&M codes to
choose from and, more surprisingly, the coders providing these results
were from four different coding agencies. In an unpublished study of
inpatient coding conducted by the author, twelve expert coders
averaged less than 70% agreement coding the same eight charts.

The largest ongoing study of coding quality is conducted by CMS in
order to assess PPS payment accuracy. They routinely find that
between 2% and 9% of the charts they sample do not support the
assigned DRG.

Coding processes haven't changed in 25 years Why do we see significant
variation in coding and why has it not completely disappeared after
twenty years of intense scrutiny under PPS reimbursement? It is
precisely because the process of coding has changed little in the last
twenty-five years. Coders still work pretty much alone; they code as
fast as they can from whatever documentation is available to them;
they are under intense pressure to code as soon as possible after
discharge; they receive very little ongoing feedback on their
accuracy. If coding accuracy and, with it, case-mix are going to
improve significant parts of this process must change.

Determinants of coding quality Coding quality is completely dependent
upon four independent factors:
· documentation,
· coder skill,
· enough time to code,
· coding tools.

Coding quality depends upon documentation. The necessity for clear,
complete and unambiguous documentation cannot be overemphasized.
Coders can't code what is not clearly specified. Ambiguous
documentation leads to ambiguous coding and is a cause for reduced
case-mix or denials. Traditionally, the chart dies shortly after
discharge. The physicians have dictated all of their reports and have
moved on to new cases. The road to document clarification and
reassessment, where ambiguity exists, requires significant effort on
the part of physicians and coders alike. On paper, it is a cumbersome
process at best.

Coding quality depends upon coding skill.

Coding is a skill built upon training, experience and feedback.
Without training, coders can miss significant content. Without
experience, they can't learn to find the subtleties in real patient
records. Without ongoing feedback they can't tell whether they are
accurate and complete.

Coding quality depends upon coding time. The old saying, "Do you want
it fast or do you want it right?" applies equally to coding. Even the
most skilled coder, with perfect documentation, will not do an
adequate job if they are not allowed sufficient time to thoroughly
evaluate the chart, including documented CCs in other volumes. Time
pressure is intense to get bills out and keep DNFB down. When all
coding must be done by a hospital's own captive staff, there is little
recourse but to spend less time on each chart when backlogged - a
tactic that jeopardizes case-mix. Also, coders often do abstracting
at the same time as coding; a task that doesn't require a coder's
skills, but is done by them because they have the chart.

Quality coding requires adequate tools. Even the best worker cannot do
well without the correct tools. Such is the case with coding. Coders
require access to current coding references, optimizing groupers, and
coding data editors if they are going to be accurate and productive.

There are tremendous savings to be realized as healthcare converts to
electronic records that are machine readable and as automated encoding
applications emerge to replace much of the manual activity in the
coding process. A 1% market share could easily be worth $150M a year
by providing hospitals with immediate, accurate, auditable coding.

Table - 1
Coding in US Healthcare
Providers How Many
Facilities? Annual Patient Revenue (Billions) Annual
Patient Visits (millions) How
Many Coders? Estimated Coder Cost (millions)
Physician offices 308,000 421 824 45,000 810
Non-federal Hospitals 5,525 612 32 8,000 320
Ambulatory surgical centers 3,597 79 32 3,000 120
Emergency Departments 3,306 8 100 4,000 160
Payors
Health Insurance carriers 3,209 3,000 120
Total $ 1,119 987 63,000 $ 1,530