11 Service Categories
A medical claims dataset may contain millions of lines and thousands of procedure codes. Before we can explain where healthcare dollars went, we need to turn that detail into a manageable set of categories.
In this chapter, we’ll discuss service category groupers: algorithms that classify claims into recognizable types of care, such as acute inpatient, emergency department, office visits, and imaging. We’ll use Tuva’s approach to explain how a hierarchy works, what evidence supports a classification, and where seemingly simple categories become difficult.
The goal is a set of categories we can use consistently. If “outpatient surgery” means something different in every analysis, we spend our time reconciling definitions instead of understanding care.
11.1 What a service category tells us
A service category describes the type of service, its setting, or a combination of the two. It is a derived analytical field. A bill type, revenue center, procedure code, and place of service are inputs to that field, each describing a different part of the billed care.
For example, a hospital outpatient claim may contain an imaging procedure. “Hospital outpatient” describes the broader setting; “outpatient radiology” is a more useful analytical category for that line. A professional claim with an inpatient place of service represents a clinician’s work in the hospital, even though it is not the hospital’s facility bill.
The category also differs from an encounter. One hospital stay can include facility services, physician services, and an ambulance trip. Service categories describe those services. An encounter grouper decides which claims belong to the same visit or stay.
11.2 Build a useful hierarchy
Tuva uses three service-category levels so we can begin with a broad view and then investigate a more specific one. The table below illustrates the structure and the principal families of care. It is a guide to the concepts, not a substitute for the category values and rules in a particular release of the Tuva service category grouper.
| Broad group | More specific categories | Useful further detail |
|---|---|---|
| Inpatient | Acute inpatient; psychiatric; rehabilitation; substance use; skilled nursing; hospice; long-term hospital care | Medical and surgical admissions; delivery; newborn care |
| Outpatient | Emergency department; observation; hospital or clinic; ambulatory surgery center; surgery; dialysis; home health; hospice; psychiatric; substance use; rehabilitation; therapy; radiology; medical-benefit drugs; urgent care | Imaging modality and other service-specific distinctions |
| Office-based | Office visits; therapy; surgery; radiology; telehealth; other office services | Imaging modality and visit type |
| Ancillary | Ambulance; durable medical equipment; laboratory services | Additional detail when the analysis needs it |
| Other | Services not assigned to a defined category | Reasons for unmatched classification |
A good hierarchy has a few practical properties.
Each line has one final category at each level. Rules can overlap, but the final assignment must resolve those overlaps. If one line appears in both surgery and imaging totals, summing categories overstates spending.
The categories cover the full input population. An explicit “other” category preserves services the grouper cannot classify. Dropping those lines makes the output look cleaner while losing part of the financial picture.
Identify a category by its full path. Tuva reuses detailed labels such as ct under both office-based and outpatient radiology. Keep the broad and intermediate categories with the detailed label when building the hierarchy. Grouping by ct alone combines settings; it does not identify one branch of the hierarchy.
The level of detail serves the question. Five broad groups may be enough for an executive cost summary. Investigating imaging prices requires more detail. A category containing hundreds of unlike procedures is not a comparable unit of care just because it has a convenient name.
11.3 Use the right evidence
Institutional and professional claims describe services differently. A reliable grouper accounts for those differences instead of expecting every field to be present on every claim.
| Input | What it contributes |
|---|---|
| Claim type | Distinguishes institutional from professional billing and determines which rules apply |
| Type of bill | Describes the institutional billing context, including facility and service classification |
| Revenue center | Identifies the department or service represented on an institutional claim line |
| Place of service | Describes the setting reported on a professional claim |
| HCPCS/CPT code and modifiers | Describe the billed procedure, service, equipment, or drug and relevant billing distinctions |
| Diagnosis codes | Add clinical context when the service definition requires it |
| DRG and DRG system | Help distinguish types of inpatient hospital cases |
| Facility or provider information | Helps identify specialized facilities and provider types |
The CMS institutional billing manual and place-of-service definitions explain the billing fields. Preserve the original codes alongside the derived categories so a classification can be traced back to its evidence.
Code systems have versions. An MS-DRG and an APR-DRG with the same numeric value are not interchangeable. A procedure grouping must be appropriate for its code system and service date. Diagnosis groupings such as AHRQ’s Clinical Classifications Software Refined answer a different question from procedure groupings; their category numbers should never be mixed.
11.3.1 Resolve overlapping rules
Consider a hospital outpatient claim containing an emergency department service, a CT scan, and laboratory testing. A setting-oriented rule might place the entire claim in emergency department. A line-oriented rule might separate the CT and laboratory services. Both approaches can be useful, but they answer different questions.
A grouper needs an explicit order of precedence. In Tuva, institutional rules may classify an entire claim while more detailed outpatient and professional rules operate at the line level. The final result remains a line assignment. To understand any particular output, inspect which rule won and whether it inherited a claim-level classification.
Tuva’s matching-stage model retains alternative matches to support encounter flags, such as emergency department involvement. When aggregating directly from this model, select only rows with duplicate_row_number = 1. Summing all matches can count the same claim line more than once.
This distinction matters for spending. A CT scan included in an inpatient category has not disappeared from the data. Its dollars are reported under the inpatient hierarchy. An analysis of all imaging, regardless of encounter setting, needs a separate procedure-based view and should not silently redefine the financial categories.
11.4 Understand the major families of care
We’ll now look at the categories analysts encounter most often. The useful distinctions are usually about clinical setting, billing context, and the unit of service, not memorizing long lists of codes.
11.4.1 Inpatient care
Acute inpatient describes hospital admissions for acute treatment. Medical and surgical subdivisions support analysis of the reason for admission and resource use. Delivery and newborn categories are also useful because maternity and neonatal care differ substantially from the rest of the inpatient population. Keep the mother’s delivery and the newborn’s care on their respective patient records.
DRGs help describe a hospitalization, but a populated DRG alone should not override contradictory claim-type or setting evidence. Use the identified DRG system and its applicable version. Delivery, newborn, and neonatal intensive care definitions may also use diagnoses, procedures, and revenue centers; a textbook code list quickly becomes stale if it omits those qualifications.
Inpatient psychiatric, rehabilitation, substance use, and long-term hospital care describe distinct settings or treatment programs. Facility information can be useful when a general hospital billing pattern does not identify a specialized unit. A psychiatric diagnosis on an acute medical admission does not, by itself, make that admission an inpatient psychiatric stay.
Skilled nursing should remain distinguishable from acute hospital care. Nursing facility services also include settings and payment arrangements that are not equivalent to a covered skilled nursing stay. Similarly, hospice can be delivered in several locations and at different levels of care. The provider’s hospice billing identity alone does not establish that every service was an inpatient hospice stay.
11.4.2 Hospital and other outpatient care
Emergency department and urgent care are separate categories. Both may treat unscheduled needs, but their settings and billing differ. For professional claims, CMS place of service 23 identifies the emergency room of a hospital, while 20 identifies an urgent care facility. The Tuva professional urgent-care rule also accepts selected procedure codes without requiring place of service 20. Its output is therefore broader than a place-of-service-20 definition and should not be treated as independent proof of the care setting.
Observation requires special attention. A patient can remain in the hospital overnight while receiving outpatient observation services. Physical location and length of stay do not establish inpatient status. Medicare’s inpatient and outpatient status guidance explains the distinction. For analytics, preserve observation evidence even when a subsequent inpatient admission becomes the primary encounter.
Outpatient surgery and ambulatory surgery center categories help distinguish where procedural care takes place. The former may describe a procedure within hospital outpatient care; the latter identifies a dedicated surgical setting. A procedure code should be interpreted together with its setting. A code’s numerical position in a broad range is not a complete surgical classification.
Dialysis is a recurring treatment category. Service dates and units matter when moving from its spending total to a count of treatments. Home health also requires a defined service and billing context: a professional visit at a patient’s home is not automatically home health agency care.
Outpatient psychiatric, substance use, rehabilitation, and physical, occupational, or speech therapy separate services that would otherwise be hidden in a large outpatient total. Psychiatric treatment and substance use treatment should not be treated as synonyms. Comprehensive rehabilitation services and individual therapy procedures are related but distinct analytical concepts.
Hospital or clinic provides a home for outpatient services that do not fit a more specific category. Medical-benefit drugs capture medications billed through medical claims. They do not represent all pharmacy spending; retail and mail-order prescription claims commonly arrive in a separate pharmacy dataset.
11.4.3 Office-based care and imaging
Office-based categories distinguish visits, procedures, therapy, and imaging performed in an office setting. Telehealth deserves its own interpretation because the patient and clinician are not necessarily together. CMS place of service 02 and 10 distinguish whether the patient is outside or inside their home. Keep historical payer billing practices in mind when analyzing telehealth trends.
Imaging is often divided into CT, MRI, PET, and broader radiology categories. MRI uses magnetic resonance; it is not nuclear medicine. PET is a form of nuclear medicine, but not every nuclear medicine service is PET. For example, the Tuva outpatient radiology rule assigns the label pet to the entire NITOS nuclear-medicine modality. That output does not identify PET alone; use a PET-specific procedure definition for a PET analysis. The Neiman Imaging Types of Service classification provides a published approach to organizing imaging procedures.
For price comparisons, distinguish the technical component of performing an examination from the professional component of interpreting it. A line containing only an interpretation should not be compared with a globally billed examination as though they were the same product.
11.4.4 Ancillary and other services
Ambulance, durable medical equipment, and laboratory categories collect supporting services that may occur independently or in connection with another visit. Their units differ: an ambulance claim may contain both transport and mileage, equipment may be rented monthly, and a laboratory claim may contain several tests. Line counts are not a common utilization measure across these categories.
The other category is an investigative starting point. It may reflect missing codes, a mapping problem, a service outside the grouper’s design, or valid new billing patterns. A nonzero total does not prove poor data quality. Examine both its share of dollars and its share of lines, then identify the providers and codes contributing most to it.
11.5 Reconcile before interpreting
Suppose a prepared dataset contains $1 million in allowed amounts. After assigning one category per line, the categories should still sum to $1 million. Check that the classification join neither duplicated nor dropped lines, and preserve adjustments and negative amounts according to the source’s financial conventions.
Then compare category distributions by source, claim type, and month. A sudden movement from outpatient surgery to other may reflect a missing procedure mapping rather than a change in care. A sharp drop in inpatient spending may reflect an incomplete facility feed.
Finally, choose the next analytical unit deliberately. Categories are a useful way to partition spending. To explain why spending changed, we usually also need encounters, comparable procedures, units, prices, and population exposure. The service category gives us an organized place to begin that investigation.