Skip to content
Zowork
All case studies
Platform studyCaché®~9 min read

The database that stores a patient record as a path, not a join.

InterSystems Caché is a high-performance, multi-model database: object, relational and multidimensional at once. Under the hood it keeps data in hierarchical globals you address directly instead of searching, which is exactly why it sits beneath some of the most demanding systems in healthcare. We spent years engineering on it, inside large-scale healthcare platforms.

InterSystems CachéHealthcareLegacy ModernizationInteroperability
multidimensional storage^Patient
ObjectScript · global set
SET ^Patient(1001,"Demographics","Name")= "Riley Carter"
SET ^Patient(1001,"Demographics","DOB")= "1986-04-12"
SET ^Patient(1001,"Clinical","Assessment")= 14
SET ^Patient(1001,"Billing","Claims")= "90837"
Hierarchical global
^Patient(1001)
Demographics
Name
DOB
Clinical
Assessment
Billing
Claims
direct addressed write · no joins
Introduction

A database that doesn't look like a database

InterSystems Caché combines database management, application development and integration in a single platform. It speaks three data models (relational, object-oriented and multidimensional) over one physical store, so a team can model entities as objects, query them in SQL, and still get the raw speed of direct multidimensional access.

That combination is why it shows up in healthcare, banking, government and telecom: places where transactions are high-volume, latency is felt by a human waiting, and downtime isn’t an option. InterSystems has built data software since 1978; Caché became the platform of record for organisations that needed performance without giving up developer productivity.

Healthcare
Banking
Government
Telecom
The problem

Healthcare data, at the speed care needs it

A healthcare provider generates an enormous amount of data every day: demographics, diagnoses, treatment plans, assessments, medications, lab results, authorisations and claims. A clinician needs the right slice of it instantly.

  • Manage large volumes of structured and unstructured clinical data.
  • Keep transaction processing fast while thousands of records change at once.
  • Integrate data across departments and outside systems without building a silo per source.
  • Stay available and consistent, because a few seconds’ delay can change a clinical decision.

Traditional relational databases can stall here: complex, high-volume workloads turn into multi-table joins and repeated disk seeks. Caché takes a different route to the same data.

The platform

One platform, cleanly layered

Applications talk to business logic; logic reaches data through either an object or a SQL lens; both resolve to a single multidimensional engine that owns physical storage. A request travels down and a response travels back up the same spine.

User applications
Web · mobile · enterprise systems
Business logic
Objects, classes & application services
SQL engine · object access
Relational and object views of one store
Multidimensional enginecore
Hierarchical globals at the core
Physical storage
Disk · backup · recovery
The core idea

Why multidimensional beats joins

A relational read fans out across tables and joins them back together. A Caché read walks a known path straight to the value. Same data, far fewer moving parts.

Traditional relational
multi-join
Application
SQL
query planner
joins across tables
PatientVisitBilling
Storage
InterSystems Caché
direct read
Application
Object access / SQL
one store, two views
Multidimensional engine
addressed, not searched
Storage

Data lives in globals: persistent, hierarchical arrays where the address is the index. There is no separate table to define and no row to look up; the path you write is the path you read. Globals are sparse, so only the data that exists consumes space, and deeply nested clinical structures come back without a single join.

FeatureWhat it buys you
Multidimensional storageDirect, sparse access with no joins to read a full record
Object + SQL on one storeModel in objects, query in SQL, no second copy of the data
ObjectScriptMUMPS-class speed with a modern development language on top
High availabilityMirroring and shadowing keep downtime low
ScalabilityGrows with users and data without a redesign
Built-in securityRow- and role-level controls over sensitive records
The foundation

MUMPS isn't legacy, it's the engine's heart

MUMPS (the M language, born in a hospital in 1966) introduced global variables and a multidimensional data model. Caché didn't replace that; it grew from it, adding indexing, transactions, security, objects and SQL on top.

MUMPS · the M language
1966, built for healthcare
Global variables
persistent
Multidimensional storage
hierarchical
InterSystems Caché engine
indexing · transactions · security
ObjectScript
OOP extensions
SQL access
relational
Enterprise applications
healthcare · banking · government

Understanding MUMPS pays off precisely because so much of Caché’s behaviour, and so much production healthcare code, still rests on globals. ObjectScript is the modern extension: it keeps M’s speed while adding the structure of a contemporary language. The same record can be a global, an object and a SQL row at once.

Why it still matters
Treating MUMPS as “old” is the expensive mistake. Knowing how the engine really works is what makes it possible to keep decades-old systems fast, debug the deep paths nobody else will touch, and extend long-lived logic safely. The performance everyone wants from Caché is inherited straight from M.
The work

What we delivered on Caché

Inside large healthcare platforms (patient records, treatment planning, clinical documentation, billing, reporting and compliance), we worked across the full breadth of the system. The common thread: outcomes clinicians and administrators feel, delivered on a platform most teams are afraid to touch.

Core clinical features

New capabilities and fixes across long-lived clinical workflows, extending decades-old logic without destabilising it.

Clinical data capture

Electronic forms that catch bad data at the point of entry, so records are right before they are ever saved.

Faster deployments

Automation that turned days of manual system setup into a repeatable process measured in minutes.

Connected systems

Integrations with hospitals, labs, pharmacies and payers so a record entered once is correct everywhere.

Reporting & compliance

Patient summaries, clinical documentation and compliance reporting the organisation could stand behind.

Performance under load

Tuning that kept clinical workflows responsive as data volumes and user counts grew.

Clinical capture

Records that are right from the start

Electronic forms are where clinicians actually enter assessments, diagnoses and treatment plans. The most valuable place to fix data quality is the moment of entry, before a bad value ever reaches the record.

Clinical form field
PHQ-9 · total score
14/ 27checking
Required
Numeric range
Cross-field dependency
Clinical rule
awaiting validation

Validation runs the instant a clinician types, and again before anything is stored, so a treatment plan can’t reference a diagnosis that was never made. Just as important, forms are configurable: organisations adapt their clinical workflows without waiting on a software release. The result is cleaner records, less rework downstream, and reporting the organisation can actually trust.

Integration

One record, reflected everywhere

A modern healthcare ecosystem is many systems that must agree. Caché is built to be the integration point: it collects from hospitals, labs and payers, standardises it, and serves a single repository the rest of the platform reads from.

Source systems
Hospital A
Hospital B
Laboratories
Insurance systems
InterSystems Caché
standardize · integrate
Unified data repository
Reports & analytics

We built and maintained the integrations that move data between the platform and external systems, so information entered in one place is reflected accurately across every connected system. Standardising into a unified repository is what makes comprehensive reporting and analytics possible, and what dissolves the data silos that slow care down.

Outcome

Why it holds up where care can't wait

1,966

The year MUMPS was created, and its model still powers the engine

six decades in production
0 joins

To read a full patient record from a global

direct addressing
3

Data models over one physical store: object, relational, multidimensional

no second copy
  • Faster retrieval of deeply nested records, with no joins to reassemble a patient.
  • Lower database overhead and high transaction throughput under real clinical load.
  • Interoperability that keeps records consistent across hospitals, labs and payers.
  • Reliability and availability that let clinicians trust the data is there when they reach for it.
Conclusion

Old roots, modern reach

Caché’s advantage is structural: store data the way you read it, and most of the work a relational database does on every query simply disappears. It asks for specialised knowledge (a smaller community than mainstream databases, and real effort to master), but in return it delivers the speed, reliability and scale that mission-critical healthcare depends on. Knowing the platform from MUMPS up is what lets a team extend it safely instead of working around it.

Running on Caché, MUMPS or another platform everyone’s afraid to touch? We’re at home in the engine room as well as the front end.

Talk to engineering