1 Introduction to Claims
A claim begins with a practical question: who should pay for a healthcare service, and how much should they pay? The answer produces a record of the patient, the provider, the services billed, and the payer’s decision. Put these records together over time and we can study how people use healthcare and what it costs.
But the billing process leaves its fingerprints on every analysis. One hospital stay can produce many claims. One claim can appear in several versions. A service can have an allowed amount even when the insurer pays nothing. Understanding these details is the starting point for working with claims data.
In this chapter, we’ll follow a claim from the delivery of care to the files an analytics team receives. Along the way, we’ll separate the healthcare event from the administrative records it creates. That distinction will come up throughout this book.
1.1 The organizations behind the data
Several organizations may participate in a single claim. Knowing their roles helps explain whose information we have, whose information we don’t have, and why two datasets about the same patient can look different.
Providers deliver services and supplies. They include people, such as physicians and physical therapists, and organizations, such as hospitals, laboratories, pharmacies, and durable medical equipment suppliers. The organization submitting the bill is not necessarily the person who delivered the service. A physician group may bill for the work of an individual physician; a hospital may bill separately for the facility resources used during that physician’s care.
Health plans and claims administrators receive claims and apply coverage and payment rules. Sometimes an insurance company both administers the benefit and bears the financial risk. In a self-funded employer plan, an employer may bear that risk while an insurer or third-party administrator processes claims. Throughout this book, we use payer for the organization adjudicating a claim, while recognizing that it may not be the organization ultimately funding the payment.
Clearinghouses help route electronic transactions between providers and payers. They can check whether a submission meets technical requirements before sending it onward. A provider may also submit directly to a payer. A clearinghouse’s acceptance tells us that a submission passed that stage of processing; it does not tell us whether the payer will cover the service. CMS’s EDI overview describes these exchange relationships.
Revenue cycle management companies help providers code services, submit bills, work denials, and collect payments from payers and patients. Pharmacy benefit managers, or PBMs, administer pharmacy benefits for many plans. Their systems may be the source of a pharmacy file even when the medical file comes from the health plan.
The patient’s perspective matters too. The patient receives care and may owe a deductible, copayment, or coinsurance. A payer’s claims file usually describes the amount assigned to the patient, not whether the provider actually collected it.
1.2 From care to an adjudicated claim
Consider a patient who visits a physician for knee pain. The clinician documents the visit, orders an X-ray, and discusses treatment. Those clinical activities and the subsequent billing activities are related, but they are not the same record.
1.2.1 The provider submits a bill
The provider translates the documented services into billing codes and submits a claim. A professional medical claim typically uses an electronic 837P transaction; a hospital facility claim typically uses an 837I. These are standardized exchange formats, not warehouse table schemas. Retail pharmacy submissions use a separate family of NCPDP standards. CMS’s adopted transaction standards distinguish these transactions.
A submission may be rejected because it cannot be processed, for example because required information is missing. This is different from an adjudicated denial: the payer has not necessarily reached a coverage or payment decision. Some analytic extracts include these early submissions; others begin only after adjudication.
1.2.2 The payer adjudicates the claim
Adjudication is the process of applying the payer’s rules to the submitted bill. The payer considers questions such as whether the patient was enrolled on the service date, whether the benefit covers the service, whether authorization requirements were met, and which contract or payment methodology applies. Other insurance and patient cost sharing can affect the result.
The result can differ across lines on the same claim. One service may be payable, another may be included in payment for a related service, and another may be denied. A single claim-level label may conceal these differences.
1.2.3 The payer explains its decision and settles payment
An electronic remittance advice, commonly an 835 transaction, explains adjudication and financial adjustments. It is distinct from the original claim submission. One remittance can cover multiple claims, and the funds transfer can also include adjustments that do not belong to a specific claim. CMS’s remittance guidance explains these levels of reporting.
For the knee visit, suppose the provider bills $200, the payer allows $120, and the full $120 is assigned to the patient’s deductible. The payer’s payment is $0. The service was still processed as covered; the zero payment does not establish a denial. We will return to these relationships in Claims Data Elements.
1.2.4 The record can change
The provider may correct a bill. The payer may recover an overpayment or reconsider a decision. A pharmacy may reverse a transaction for a prescription that was not picked up. The resulting records describe changes in billing and adjudication, not necessarily additional care.
An extract might contain every transaction, the latest version of each claim, or a mixture. This is why we need to understand adjustments, denials, and reversals before counting services or adding payments.
1.3 Three billing structures to recognize
The paper forms remain useful teaching tools because they make the structure of medical claims visible. Electronic transactions carry the corresponding concepts, often with additional detail and different repetition limits.
| Billing structure | Common use | What it emphasizes |
|---|---|---|
| Professional: CMS-1500 and 837P | Clinician services and many supplier services | Procedure or service codes, dates, units, rendering providers, and place of service |
| Institutional: CMS-1450, also called UB-04, and 837I | Hospital and other institutional services | Type of bill, facility charges, revenue codes, and information about an admission or billing period |
| Retail pharmacy: NCPDP transactions | Prescriptions billed through a pharmacy benefit | The dispensed product, quantity, days supply, pharmacy, and transaction outcome |
These are billing categories, not a complete taxonomy of organizations or care settings. The same organization may submit different kinds of claims. Drugs administered in a physician’s office, for example, may appear on medical claims rather than in the retail pharmacy file. Consult the NUCC professional form instructions and CMS institutional form guidance to understand the underlying layouts.
1.4 What arrives in the warehouse
An analytics team usually receives files derived from a payer’s operational systems rather than untouched 837 and 835 transactions. The payer or data vendor has already made choices: which records to include, how to represent adjustments, where to put diagnoses, and how to combine submitted and adjudicated amounts.
Those choices matter as much as the column names. A field called paid_amount might hold a payment transaction or the current total for a claim. A row might represent a claim line, an entire claim, or a version of either. Two files with identical schemas can require different analytic treatment.
Start by establishing the population, benefits, time period, record grain, and processing history represented in the delivery. Obtain the source dictionary and worked examples of corrected claims. Keep the source identifiers and extract dates so that a result can be traced back to the delivered records.
1.5 What claims can tell us
Claims are especially useful when we need to follow covered care across many providers. They support analyses of spending, utilization, diagnoses, procedures, and treatment patterns. Enrollment data supplies the population and coverage periods needed to interpret those observations.
Their limits follow from the same billing process. Claims generally do not explain the clinician’s full reasoning, establish whether a patient took a dispensed medication, or capture every service received outside the covered benefit. A diagnosis records information used in billing; it does not independently confirm every clinical detail. An absence of claims may mean no care, missing coverage, or missing data.
Good claims analysis makes these boundaries explicit. Before asking what the data says about healthcare, we first ask what the records represent. Headers and Lines takes that question down to the level where most analytic mistakes begin.