Database Credentialed Access


Alex Bennett Joshua Wiedekopf Hannes Ulrich Philip van Damme Alistair Johnson

Published: Feb. 20, 2024. Version: 1.0

When using this resource, please cite: (show more options)
Bennett, A., Wiedekopf, J., Ulrich, H., van Damme, P., & Johnson, A. (2024). MIMIC-IV on FHIR (version 1.0). PhysioNet.

Additionally, please cite the original publication:

Bennett AM, Ulrich H, van Damme P, Wiedekopf J, Johnson AE. MIMIC-IV on FHIR: converting a decade of in-patient data into an exchangeable, interoperable format. Journal of the American Medical Informatics Association. 2023 Apr 1;30(4):718-25.

Please include the standard citation for PhysioNet: (show more options)
Goldberger, A., Amaral, L., Glass, L., Hausdorff, J., Ivanov, P. C., Mark, R., ... & Stanley, H. E. (2000). PhysioBank, PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals. Circulation [Online]. 101 (23), pp. e215–e220.


Fast Healthcare Interoperability Resources (FHIR) has emerged as a robust standard for healthcare data exchange. To explore the use of FHIR for the process of data harmonization, we converted the Medical Information Mart for Intensive Care IV (MIMIC-IV) and MIMIC-IV Emergency Department (MIMIC-IV-ED) databases into FHIR. We extended base FHIR to encode information in MIMIC-IV and aimed to retain the data in FHIR with minimal additional processing, aligning to US Core v4.0.0 where possible. A total of 24 profiles were created for MIMIC-IV data, and an additional 6 profiles were created for MIMIC-IV-ED data. All MIMIC terminology was converted into code systems and value sets, as necessary. We hope MIMIC-IV in FHIR provides a useful restructuring of the data to support applications around data harmonization, interoperability, and other areas of research.


The FHIR standard provides a framework for structuring health data and supporting data exchange amongst disparate systems and vendors. Documentation for FHIR is richly detailed and openly available [1]. Briefly, FHIR consists of a set of resources which describe the most commonly encountered entities in healthcare. As local contexts are expected to deviate from the global standard, FHIR introduces a mechanism called "profiling". Profiles are modifications of the base FHIR resources intended for use in a local context, such as a hospital system or primary care clinic. These profiles allow for the flexibility to capture data as exchanged in the local system, while still enabling standardization as a majority of fields present in each resource will be consistent with the global standard (the base resource). FHIR also provides significant detail around all aspects of information exchange, and has a large ecosystem of tooling particularly around data exchange.

As a part of broader federal efforts to promote health information technology and interoperability, the Office of the National Coordinator for Health Information Technology (ONC) created the United States Core Data for Interoperability (USCDI) standard. The USCDI standard is a set of health data classes and data elements that are essential for nationwide, interoperable health information exchange. Although no specific exchange format is mandated, the ONC promotes the use of FHIR to exchange information according to the USCDI standard. To support FHIR based exchange of data adhering to USCDI by health systems, a set of Implementation Guides were created, commonly referred to as US Core. The US Core Implementation Guide provides extensive detail on how to format and exchange data which adheres to the USCDI standard. As a result, US Core has become a common extension of FHIR used by health systems across the US for interoperability. The latest version is available online [2].

Although FHIR is broadly used by health systems world wide, there remains a paucity of health system derived data publicly available in FHIR. Prominent publicly available datasets include MIMIC-IV, a relational database corresponding to over 60,000 patients and MIMIC-IV-ED, an extension of MIMIC-IV including data from emergency department patients [3, 4]. MIMIC-IV has gained traction in the community due to its transparent mechanism of data access, reasonably large sample size, and authentic capture of a real-world electronic health record database. A substantial portion of MIMIC-III and MIMIC-IV has already been translated into FHIR to assist in research and development in the German FHIR context [5, 6].

MIMIC-IV-on-FHIR, this PhysioNet project, aims to translate MIMIC-IV into FHIR, preserve the MIMIC-IV structure and information, and provide an easily accessible FHIR dataset for use in research and development.


MIMIC-IV-on-FHIR aims to restructure MIMIC-IV in the FHIR format with as little information loss in translation as possible. FHIR stores healthcare information in resources, encoded as either JavaScript Object Notation (JSON), eXtended Markup Language (XML), or otherwise. Thus the mapping process involved mapping each record within MIMIC-IV relational database tables to FHIR resources. We chose JSON as the serialization format. The mapping process involved five steps:

  1. Terminology Generation. Capturing the MIMIC-IV terminology in FHIR was needed to retain the rich context MIMIC provides. We did not attempt to map any MIMIC-IV concepts to standard ontologies. However we generate new codes using only standard ontologies where necessary, such as the use of a SNOMED-CT code to indicate the existence of triage at an emergency center for MIMIC-IV-ED patients. FHIR stores terminology in two resources: CodeSystems and ValueSets. CodeSystems are the source for codes and ValueSets are use-case specific and combinations of CodeSystems. Codes were pulled from the MIMIC-IV tables and then converted to CodeSystems and ValueSets. Once created, the ValueSets are bound to data elements in resources. These bindings are utilized by the FHIR server to validate proper codes are assigned to resource data elements.

  1. Implementation Guide Creation. To provide a reproducible FHIR format for MIMIC, an implementation guide (IG) was made [11, 12]. A FHIR IG is a collection of FHIR profiles and terminology aiming to achieve a task within a specific domain. The creation of FHIR IGs additionally allows for publication of a website where users can explore the resources in detail. The MIMIC implementation guide includes 22 profiles, 64 terminology resources, and 2 extensions. Where possible MIMIC profiles were aligned with the US Core R4 profiles, but they do not explicitly inherit from US Core [2].

  1. Mapping. The goal for mapping was to have as complete a picture of MIMIC-IV in FHIR. Each column in MIMIC-IV was investigated to identify potential mappings from MIMIC-IV columns to FHIR resource elements. Data from the MIMIC-IV tables were mapped to FHIR elements, and Structured Query Language (SQL) code was written to export the mapped data as JSON serialized FHIR resources. The final MIMIC-IV column to FHIR element mappings can be found in the MIMIC implementation guide. For MIMIC-IV columns without direct mappings into FHIR, extensions were made to house the information. After mapping into FHIR, the resources may be validated against the MIMIC implementation guide, as the native SQL tooling does not guarantee conformance of the exported data.

  1. Validation. A FHIR server was used to validate the mapped MIMIC data. Validation is a useful exercise for locating issues in terminology binding, inter-resource referencing, or improper element mappings. The FHIR validation effectively provides unit testing for the correctness of the MIMIC to FHIR mappings completed.

  1. Export. We chose ndjson as the format for distribution of the dataset. Exporting the validated mimic-fhir resources to ndjson required two main steps. First, the bulk export functionality of FHIR servers was leveraged to export all the resources. Second, the exported resources were then written to the ndjson format. At this stage the exported njdsons represent the full picture of MIMIC-IV-on-FHIR. We compressed the ndjson files using gzip to reduce the file size.

Note that although steps 4 and 5 are important for ensuring adherence of the data to the publicized implementation guide, it is also possible to directly export MIMIC-IV on FHIR from the source database into ndjson after it has been completely mapped in step 3.

Data Description

MIMIC-IV-on-FHIR is a collection of FHIR profiles, each of which is used to house data transformed from MIMIC-IV and MIMIC-IV-ED databases. The base FHIR resources used include Condition, Encounter, Medication, MedicationAdministration, MedicationDispense, MedicationRequest, Observation, Organization, Patient, and Procedure. As we aimed to reproduce the data within FHIR without modification, we did not perform terminology mapping. As a result, new profiles were necessary in order to bind the original terminology present in the MIMIC-IV and MIMIC-IV-ED databases.

Table 1 provides an overview of all profiles created, the original MIMIC-IV or MIMIC-IV-ED tables used to source the data, and profile specific notes. After Table 1, we describe each profile in turn, highlighting (1) the base FHIR resource it inherits from, (2) relevant terminologies which were bound in the profile, and (3) other important attributes of the profiles.

Table 1. Approximate mapping of MIMIC-IV tables with FHIR resources.

MIMIC Schema

Table source for data

New FHIR Profile



(patient tracking) 



Patient birthdate was estimated by taking the anchor_age and subtracting the transfers.intime. 







Care units are translated to Location resources. These are referenced in the main Encounter


diagnoses_icd, d_icd_diagnoses



procedures_icd, d_icd_procedures



labevents, d_labitems


LOINC was not used. Terminology consists of the original itemid values.



Microbiology data had to be divided into three separate resources to be represented in FHIR. There is a parent-child relationship going from Test->Org->Susc.

prescriptions, poe


Prescriptions was the primary source for medication requests, but poe was used if a request was made but no linking pharmacy_id was present, such as for large volumes of fluid.




emar, emar_detail


The medication referenced are primarily drug names and formulary drug codes. We did not use NDC as there are missing values.




All ICU profiles point to the this profile, which references back to the original hospital admission encounter.



















All ED profiles point to the this profile, which references to the subsequent hospital admission encounter if the individual was later hospitalized.












The existence of an emergency department triage event is considered a procedure.

vitalsign, triage


Pain and heart rhythm documentation.

vitalsign, triage


Vital sign documentation.

Patient and Organization Resources


An Organization resource records any institutions or organization associated with healthcare services.

A single MimicOrganization resource was created for all of MIMIC-IV as the Beth Israel Deaconess Medical Center is the organization for all stays in the database.


The Location resource records any physical location where services are delivered in relation to the hospital.

Each unique care unit found in transfers was translated into a MimicLocation resource. In total, there are 41 Location resources created for each of the care units. When a patient is transferred to one of these care units, the location element of their Encounter resource will contain a reference to one of the 41 MimicLocation resources.


The Patient resource records the demographic information for a patient associated with an organization.

The patients table joined with the admissions and transfers tables mapped to the MimicPatient profile. Several assumptions were made to assist in the mapping to the Patient resource. First, the birthdate was calculated since only an anchor_age is given for patients. The intime column of the transfers table was used as a base for the birthdate calculation versus using the admittime column of the admissions table, since the admittime column is only available for hospitalized patients (i.e. it excludes ED only stays). Second, the birthsex extension element uses the gender column from MIMIC-IV as there is no specific documentation of birth sex available. Finally, the MimicPatient name element is derived from the subject_id column of the patients table following format of “Patient_”.

MimicEncounter and MimicEncounterED

The Encounter resource records the full span of a hospital stay, including admission, stay, and discharge. Although both inpatient and outpatient encounters can be referenced using the Encounter resource, only inpatient encounters exist in MIMIC-IV v2.2. Data from the admissions and icustays tables were mapped to the MimicEncounter profile. MimicEncounter contains custom terminology bindings for admission class, admission type, admission source and discharge disposition. The primary information mapped from admissions was the admission start and stop time along with the context for the admission. Additional information is supplied for the encounter from the transfers table in the form of Location resources. The Location resources track the movement of the patient throughout their encounter, i.e. each inter-ward transfer in the transfers table will result in a new reference to a Location resource. The MimicEncounterED profile is similar and contains information for stays within the ED.

Specimen Related Observation Profiles


A standard way to map lab observations was created by US Core with their profile USCoreLaboratoryResultObservationProfile. The US core profile was used as reference but not as a parent resource for MimicObservatoinLabevents as we chose to not modify terminology present in MIMIC, and the US Core profiles have strict terminology bindings. MimicObservationLabevents contains terminology bindings for lab codes and interpretation plus an extension for lab priority. Rows from the labevents table were mapped to MimicObservationLabevents resources, and these resources primarily capture the item code, resulting value, interpretation, and timing for labs.

MimicObservationMicroTest, MimicObservationMicroOrg, and MimicObservationMicroSusc

We found that microbiology observations were too complex to map to a single resource in FHIR. We chose to map a single microbiology observation into three resources covering the test, organisms detected, and susceptibility results. Three Observation profiles were generated: MimicObservationMicroTest, MimicObservationMicroOrg, and MimicObservationMicroSusc.

MimicObservationMicroTest captures the test information documented in the microbiologyevents table. Test codes, timing and comments are delineated in the profile. If a microbiology test is associated with the growth of an organism, references to MimicObservationMicroOrg resources are created. There are a significant number of tests with no reference to an organism, and these tests are stored with their result pulled from the comments. Alternatively, tests with no organism growth would not have an associated organism.

MimicObservationMicroOrg captures the organism information from the microbiologyevents table including MIMIC-IV specific organism codes and their descriptions. If the organisms were tested for susceptibility to antibiotics, references to child MimicObservationMicroSusc resources were created. Additionally, all MimicObservationMicroOrg resources will have a reference to the "parent" MimicObservationMicroTest resource.

MimicObservationMicroSusc captures the susceptibility results from the microbiologyevents table including MIMIC-IV specific terminology codes for antibiotics, their descriptive name, and interpretation of the susceptibility. An extension was also added to MimicObservationMicroSusc to house the dilution values for susceptibility testing.


Microbiology and laboratory tests are conducted on samples derived from the individual, and these samples are referred to as specimens. In FHIR, the Specimen resource records information about a material sample, and the MimicSpecimen profile is based on this resource. The MimicSpecimen profile captures the specimen information from both the microbiologyevents and labevents tables in MIMIC-IV. Notably attributes of this profile include the type of fluid and a unique identifier for the specimen, which is distinct for microbiology and laboratory tests.

Medication Profiles


In FHIR, the Medication resource records medications and drugs available in healthcare settings. As MIMIC-IV does not have a single data table for medications, the relevant information was acquired from seven sources.

  1. Formulary drug codes are pulled from the prescriptions table and the emar_detail table.

  1. National Drug Codes (NDC) were sourced from the prescriptions table

  1. Generic Sequence Numbers (GSN) were sourced from the prescriptions table

  1. Generic medication names were sourced from drug and medication columns of the prescriptions table as well as the medication column of the emar table 

  1. Intravenous (IV) specific events were extracted from the poe table when order_type is set to IV therapy or total parenteral nutrition.  

  1. ICU medications were sourced from the d_items table

  1. Medication mixes were pulled from prescriptions. Medication mixes are sets of medications which are grouped together as they are typically dispensed together by the pharmacy. Medication mixes reference their component medications using an ingredients attribute. Medication mixes were included as only one Medication reference is allowed for a MedicationRequest resource (i.e. a prescription) even though multiple medications can be requested together in MIMIC-IV.

The medication sources are included in MimicMedication using terminology bindings for the medication codes. All medication codes found in MIMIC-IV were bound to a single ValueSet resource, which combined together the CodeSystem resources derived from each of the seven above listed sources.


The MedicationAdminstration resource records any administration of medication in a healthcare setting. The two main MIMIC-IV documented sources for administration come from emar and inputevents. Each of these tables were mapped to a custom profile derived from the base R4 fhir profiles.

Information documented in the emar and emar_detail tables was mapped to the MimicMedicationAdministration profile. MimicMedicationAdministration contains terminology bindings for medication site, medication method, and medication route. The primary information is pulled from each row in the emar_detail table, with the emar table supplying the medication reference only when it is not available in product_code column of the emar_detail table.


The inputevents table was mapped to MimicMedicationAdministrationICU. MimicMedicationAdministrationICU contains terminology bindings for category ICU and medication method ICU. The ICU medication administration resources will be smaller than emar medication administrations since less detailed information is available in MIMIC-IV.


The MedicationDispense resource records the supply of medication to a patient. Medications in MIMIC-IV are primarily dispensed by the pharmacy. The pharmacy table was mapped to MimicMedicationDispense, with terminology bindings for medication route and medication frequency. The MimicMedicationDispense is linked back to MimicMedicationRequest through the element authorizingPrescription.


The pyxis table was mapped to MimicMedicationDispenseED profile. MimicMedicationDispenseED contains terminology bindings for GSN medication codes and medication dispensation timing.


The MedicationRequest resource records orders for medication and the administration instructions for the medication. The resource only accepts a single Medication resource, which means the majority of referenced medication will be medication mixes coming from the prescriptions table. There were two sources for MimicMedicationRequest: the prescriptions table and the poe table. Information from the prescriptions table was supplemented by the pharmacy table to populate missing dosage information.

The poe table maps orders to MimicMedicationRequest when events documented in the emar table do not have a related pharmacy_id. This scenario is typically for IV or TPN events occurring in the hospital when a direct prescription may not have been written but the medication was ordered.


The MedicationStatement resource documents stated history of medication information. For ED derived data, this corresponds to the medicine reconciliation data present in the medrecon table. The MimicMedicationStatementED thus provides a profile for reported medication use by individuals in the ED.

Charted Observation Profiles

MimicObservationChartevents, MimicObservationDatetimeevents, and MimicObservationOutputevents

The Observation resource records any measurements made about a patient. The chartevents, datetimeevents, and outputevents tables are all measurement tables for ICU events. Each table maps to a mimic observation profile, based on the base Observation resource. The mimic profiles are of a similar format but with different terminology bindings based on the source table.

The chartevents table is mapped to the MimicObservationChartevents profile. MimicObservationChartevents enforces terminology bindings for the observation code and category. Chartevents captures the majority of documented information from the ICU, thus the primary information mapped was the timing, item code and the result of the item. Additional columns were mapped to capture reference ranges and observation category.

The datetimeevents table is mapped to the MimicObservationDatetimeevents profile. MimicObservationDatetimeevents enforces terminology bindings for observation code and category. Information in the datetimeevents table follows a datetime format, thus the primary information mapped were the datetime value and the item code.

The outputevents table is mapped to the MimicObservationOutputevents profile. MimicObservationOutputevents enforces terminology bindings for observation code and category. The outputevents table holds information on the patients’ bodily outputs, thus the primary information mapped were item codes for these events and the resulting value.


The procedureevents table is mapped to the MimicProcedureICU profile. MimicProcedureICU enforces terminology bindings for the procedure code, body site and category. The primary information mapped from the procedureevents table includes timing, procedure codes, status and the procedure location.


Observations made in the emergency department around pain and heart rhythm were documented within the MimicObservationED resource, which inherited from the base Observation resource.


Vital signs, including heart rate, respiratory rate, temperature, and oxygen saturation, were mapped to the MimicObservationVitalSignsED profile. These observations are made either on triage or during routine monitoring of individuals in the emergency department.

Billing Resources

MimicCondition and MimicConditionED

The Condition resource records conditions, diagnoses, and problems. For the MimicCondition profile, the diagnoses_icd table was the primary source of information, including terminology bindings for the diagnosis code and condition category. The MimicConditionED profile contains billed ICD diagnoses from the end of an individual's ED stay.

MimicProcedure and MimicProcedureED

The Procedure resource records an action that is performed on a patient. The procedures_icd table was the primary source of data for the MimicProcedure profile, with terminology bindings created for the ICD codes present. As the act of triage of a patient is a procedure, we created the MimicProcedureED to capture ED procedures, with the only ED procedure currently present being provision of triage to the individual.

Usage Notes

Data files are distributed as compressed NDJSON files with the extension ndjson.gz. Each line of each ndjson.gz file is a valid FHIR resource serialized as a JSON.

An open source repository, MIMIC-FHIR, was created to store the code needed to generate and use the mimic-fhir resources [7]. The repository allows for community discussion and collaboration on mimic-fhir. An archived version of the scripts for building the MIMIC-IV-on-FHIR demo dataset are also archived in Zenodo [8].

The mimic-fhir NDJSON's can be taken and loaded into any FHIR server. A jupyter notebook was developed to walk through the loading and usage of the mimic-fhir resources with the Pathling FHIR server [9, 10]. Pathling was used to demo the mimic-fhir resources due to its simple ndjson loading and optimized analytic operations.

An implementation guide with detailed information about the profiles created for the FHIR formatted dataset is provided online [11]. The code for generating the implementation guide is available in the open-source mimic-profiles repository [12].

Release Notes

The current and first release of MIMIC-IV on FHIR is v1.0. This release used data from MIMIC-IV v2.2 [3] and MIMIC-IV-ED v2.2 [4].


MIMIC-IV on FHIR was created using publicly available deidentified data.


The authors would like to thank those behind MIMIC-IV for making the data available and the FHIR community for support in answering questions. This work was supported by the Canadian Institutes of Health Research funding reference number 470397.

Conflicts of Interest

None to declare.


  1. HL7 FHIR Documentation. [Accessed: 30 January 2024]
  2. US Core HL7 FHIR Implementation Guide. [Accessed: 6 June 2022]
  3. Johnson, A., Bulgarelli, L., Pollard, T., Horng, S., Celi, L. A., & Mark, R. (2023). MIMIC-IV (version 2.2). PhysioNet.
  4. Johnson, A., Bulgarelli, L., Pollard, T., Celi, L. A., Mark, R., & Horng, S. (2023). MIMIC-IV-ED (version 2.2). PhysioNet.
  5. Ververs, S., Ulrich, H., Kock, A.-K., & Ingenerf, J. (2018). Konvertierung von MIMIC-III-Daten zu FHIR. Jahrestagung Der Deutschen Gesellschaft Für Medizinische Informatik, Biometrie Und Epidemiologie E.V. (GDMS).
  6. Ulrich, H., Behrend, P., Wiedekopf, J., Drenkhahn, C., Kock-Schoppenhauer, A.-K., & Ingenerf, J. (2021). Hands on the Medical Informatics Initiative Core Data Set — lessons learned from converting the mimic-IV. Studies in Health Technology and Informatics.
  7. MIMIC-IV-on-FHIR Code on GitHub. [Accessed: 16 Nov 2023]
  8. Bennett AM; Johnson AJ. (2022). kind-lab/mimic-fhir: MIMIC-IV-on-FHIR v1.0 (v1.0). Zenodo.
  9. MIMIC-FHIR Tutorial. [Accessed: 6 June 2022]
  10. MIMIC Implementation Guide. [Accessed: 2 Jan 2024]
  11. Pathling: Advanced FHIR Analytics Server. [Accessed: 6 June 2022]
  12. mimic-profiles on GitHub. [Accessed: 2 Jan 2024]

Parent Projects
MIMIC-IV on FHIR was derived from: Please cite them when using this project.

Access Policy:
Only credentialed users who sign the DUA can access the files.

License (for files):
PhysioNet Credentialed Health Data License 1.5.0

Data Use Agreement:
PhysioNet Credentialed Health Data Use Agreement 1.5.0

Required training:
CITI Data or Specimens Only Research

Corresponding Author
You must be logged in to view the contact information.