4 Claims Data Elements
A claims dictionary can contain hundreds of fields. We do not need to memorize them all to begin an analysis, but we do need to understand the fields that determine who received care, what was billed, where and when it happened, and how the bill was adjudicated.
In this chapter, we’ll work through those fields in that order. For each group, the aim is to connect the field’s meaning to an analytic decision. Knowing that a column contains a valid code is useful. Knowing whether that code supports the conclusion we want to draw is much more useful.
Column names vary by source. The names below describe common analytic concepts; they do not guarantee that a payer’s similarly named column has the same definition or grain.
4.1 Identifying the patient and the record
The patient or member identifier connects claims to the person who received care. A subscriber identifier may instead identify the employee or policyholder whose coverage includes several dependents. Confusing the two can merge the claims of an entire family into one patient.
Establish how identifiers behave across benefit changes, employers, and source systems. A person can receive a new member identifier without becoming a new person. Conversely, matching identifiers in two sources do not establish that they refer to the same individual. Preserve source context and use a documented crosswalk when linking identities.
The claim identifier, version or transaction identifier, and line number distinguish the billing records. As discussed in Headers and Lines, all three may be needed to identify a row. An original-claim reference can connect corrected bills to earlier records. Keep these fields even when the immediate analysis only needs the latest adjudicated state; they are essential for explaining how that state was selected.
4.2 Describing the setting and the bill
Administrative codes tell us about the circumstances in which services were billed. Several are especially useful when grouping claims into care settings.
4.2.1 Institutional billing codes
| Field | Meaning |
|---|---|
| Type of bill | Identifies the facility category, bill classification, and billing frequency or sequence. A leading zero may be included or omitted. |
| Admission type | Describes the priority of the admission or visit, such as emergency or elective. |
| Point of origin, often called admission source | Describes the referral source for the admission or visit. |
| Patient discharge status, often called discharge disposition | Describes the patient’s status at the end of the billing period, including discharge destination or continued care. |
These fields belong to the institutional claim header. The NUBC maintains the underlying billing specifications; CMS’s institutional claims manual explains their use in Medicare.
Treat type of bill as a code, not a number. Store a normalized representation that preserves its components, and keep the original value for traceability. Its meaning depends on the combination of components, so testing digits in isolation can misclassify a bill.
Admission type and point of origin add context, but neither should carry an entire encounter definition by itself. For example, identifying an emergency department visit generally requires evidence from service codes and other claim information. Profile these fields by source and care setting rather than assuming a universal ranking of their reliability.
Discharge status needs similar care. A bill ending on January 31 can describe a patient still receiving inpatient care. Its through date is not automatically the discharge date. Likewise, an indication of death during a stay is useful evidence, but an absence of that indication does not establish that the patient is alive. Claims are not a complete mortality registry.
4.2.2 Place of service
Place of service, or POS, describes the setting reported for a professional service. For example, code 11 identifies an office setting and 20 identifies an urgent care facility. POS 11 alone does not establish that an office evaluation occurred; the billed service also matters. CMS publishes the place-of-service code set.
POS can vary across service lines. Preserve that detail when available, and do not silently replace a missing value with a setting inferred from the billing provider’s address. A physician’s billing office and the location where care occurred can be different places.
4.3 Describing diagnoses, procedures, and resources
Several code systems describe different aspects of the same care. Diagnoses describe conditions and reasons for an encounter. Procedure and service codes describe activities, while revenue codes classify institutional charges. We need to retain these distinctions when translating codes into analytic categories.
4.3.1 Diagnoses
ICD-10-CM codes describe diagnoses and other reasons for healthcare encounters. Keep the code system, diagnosis position or role, and any present-on-admission indicator supplied by the source. For an inpatient stay, the principal diagnosis identifies the condition determined after evaluation to be chiefly responsible for the admission. Secondary diagnoses describe additional reportable conditions. Present-on-admission describes whether a condition was present when the inpatient admission order occurred; it is separate from diagnosis position.
Coding conventions also vary by setting. Certain uncertain diagnoses may be coded as established for qualifying inpatient admissions, while outpatient coding follows different rules. The official ICD-10-CM guidelines explain this distinction. We should not read every diagnosis on every claim as an equally strong confirmation of disease.
For an analytic condition definition, decide which claim settings count, which diagnosis positions count, and whether repeated evidence is needed. Those choices belong in the definition, not in an undocumented cleanup step.
4.3.2 Procedures, services, and modifiers
HCPCS includes CPT, maintained by the American Medical Association, and Level II codes, maintained by CMS. These codes identify many billed services, supplies, and products. Modifiers add information that can change the interpretation or payment of a service. CMS’s HCPCS overview describes the two levels.
ICD-10-PCS is a separate system used for hospital inpatient procedure coding. It is not another name for HCPCS, and an inpatient procedure code does not necessarily correspond one-to-one with a billed service line. CMS’s code-set overview distinguishes the systems.
Keep codes as strings, preserve meaningful leading zeros, and validate them against the applicable version. A code that is inactive today may have been valid when the service occurred. Preserve modifiers and units rather than reducing every line to its base procedure code.
4.3.3 Revenue codes and payment classifications
Revenue codes classify institutional charges into categories such as room and board, emergency services, and laboratory services. They can appear alongside HCPCS codes: the revenue code describes the charge category, while HCPCS can provide more specific service detail. Revenue codes are relevant to billing and payment rules, not merely labels for a hospital’s accounting departments. CMS’s revenue-code guidance relates them to payable bill types.
Institutional reimbursement is not universally determined by a single DRG or per diem rate. Diagnosis-related groups (DRGs) classify inpatient stays for payment. Ambulatory payment classifications (APCs) group applicable hospital outpatient items and services; one outpatient encounter can generate more than one APC payment. Medicare uses these and other systems across settings, and some services are packaged into another payment. Even under inpatient prospective payment, outlier adjustments can matter. CMS’s payment-system overview and outlier guidance explain these distinctions.
For analytics, this means that an itemized charge is not necessarily a separately paid service. A line with zero payment may still help identify the care delivered.
4.4 Separating the clocks in claims data
A healthcare event, a billing period, a payment, and a data delivery each have their own date. Choose the clock that answers the question.
| Date concept | What it represents | A common analytic use |
|---|---|---|
| Claim start and end | The service or statement period represented by the claim, as defined by the source | Identifying the span of a bill |
| Line start and end | The service period for an individual line | Ordering outpatient services and assigning them to periods |
| Admission and discharge | The start and end of a stay, where available and applicable | Length of stay and readmission analysis |
| Received and adjudication dates | When the payer received or processed a record | Studying processing lag |
| Paid date | The source’s payment-related date | Payment-period reporting and claims development |
| Extract or load date | When the record was delivered or loaded | Reproducing what was available at a cutoff |
Suppose a patient is admitted January 28 and discharged February 4. A source delivers an interim bill through January 31 and a subsequent bill through February 4. Both may carry the January 28 admission date. Treating each claim start as a new admission would create two stays; treating the interim through date as a discharge would shorten the stay.
Line dates are particularly important when a professional claim contains services on several days. Using the header end date for every line can move January services into February or change the sequence of treatment. An extract may derive header dates from line dates, so read its definition before substituting one for another.
For pharmacy data, distinguish the recorded dispensing or service date from the transaction-processing date and financial settlement date. An approved billing transaction does not by itself establish pickup or ingestion. Enrollment dates define coverage intervals; confirm whether the end is inclusive and how open intervals are represented.
Check date ordering, missingness, implausible values, and distribution changes over time. Apply those checks in context: a future enrollment end date may be a valid open-coverage sentinel, while a future service date in a completed claims extract needs investigation. Incurred, Paid, and Runout explores how these clocks affect reporting.
4.5 Understanding the amounts
The three amounts most often confused are billed, allowed, and paid.
Billed amount is the submitted charge. It is not the provider’s accounting cost or a reliable measure of what anyone ultimately paid. Allowed amount is the amount recognized under the payer’s coverage and pricing rules for the service. Paid amount is the payer payment represented by the record. The allowed amount can include patient cost sharing, so it need not equal the payer payment. HealthCare.gov’s allowed-amount definition provides the underlying benefit concept.
Consider a covered service with these illustrative amounts:
| Component | Amount |
|---|---|
| Billed charge | $200 |
| Allowed amount | $120 |
| Deductible assigned to the patient | $40 |
| Coinsurance assigned to the patient | $16 |
| Payer payment | $64 |
In this simple example, $120 equals $40 + $16 + $64. That equation describes the example, not a universal reconciliation rule. Other-payer payments, different definitions of patient responsibility, and additional adjustments can change the relationship in an actual source. Do not fill a missing allowed amount by adding fields until their definitions support that calculation.
Payer paid amounts answer questions about the payer’s claim spending. Allowed amounts can support comparisons that include patient cost sharing. Billed amounts describe charges. None automatically captures pharmacy rebates, capitation, incentive payments, or other financial activity outside the claims file.
Finally, determine whether each amount is a header total, a line amount, a signed transaction, or a replacement value. An accurate amount at the wrong grain still produces an inaccurate total.
4.6 Keeping provider roles separate
A billing provider is the individual or organization billing for the service. A rendering provider is identified as delivering the service. An attending provider has overall responsibility for the care reported on an institutional claim or encounter; the role is not limited to inpatient stays. A referring or ordering provider initiates a referral or order. These roles are not interchangeable when comparing practices, attributing patients, or studying referral patterns. CMS’s professional and institutional billing instructions describe the associated fields.
The National Provider Identifier, or NPI, can identify an individual or an organization. It does not encode specialty or location; those characteristics require additional data. CMS’s NPI guidance explains the identifier’s scope.
Preserve the source’s provider roles and distinguish a reported value from one filled through a fallback. Substituting billing NPI for missing rendering NPI may make a column look complete while changing the question the analysis answers. Across all these fields, our task is the same: carry the source’s meaning into the analytic model, and make every transformation of that meaning visible.