Liberating medical device data for clinical research


Published on

Published in: Business, Technology
  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide

Liberating medical device data for clinical research

  1. 1. Liberating medical device data for clinical research An architecture for semantic and temporal harmonizationRafael Richards, MS, MD John R Zaleski, PhD Sameer Peesapati, MSAssistant Professor IEEE Senior Member & Application Development EngineerAnesthesiology & Critical Care Medicine CTO & VP Clinical Applications Nuvon, Inc.The Johns Hopkins Medical Institutions Nuvon, Inc. 3624 Market St., Philadelphia, PA 19104Baltimore, MD 2224-2760 3624 Market St., Philadelphia, PA 19104
  2. 2. Introduction: The State of Device Data2
  3. 3. Medical Device Data: The Problem Medical device data are trapped in silos  Unique Protocols  Unique Physical Connectivity  Unique Clock Times  Unique Time frequency of output  Unique Terminology Differences must be harmonized before data can be used 3
  4. 4. Introduction We present a tool that:  Liberates device data  Integrates data from multiple devices  Semantically and temporally harmonizes data from as many as eight medical devices connected to a given patient Data are delivered to the clinical investigator in multiple formats  CSV (spreadsheet) or HL7 4
  5. 5. Architecture Data collection appliance (DCA)  Each DCA can be connected to as many as eight patient physiologic monitoring or therapeutic medical devices.  DCA can be remotely accessed via hospital network to select parameters and frequency of collection from devices connected to the patient. Data aggregation gateway (DAG)  DAG orchestrates integration of data of from multiple patients  Filters and synchronizes data such that it can be easily stored in a data warehouse (DW) or EMR. Management Server (MS)  Orchestrates coordination of data between each bedside and DAG.  Functions include time synchronization of all devices at all bedsides to a single universal clock, and use of a standard dictionary to harmonize terminology. 5
  6. 6. Nuvon VEGA ServerPhysical ArchitectureMechanical Ventilators Native Devices Data DCA DAG & MS IDC Nuvon VEGA Server Enterprise Clinical Information System IDM-MG 3000 Electronic Medical Record Systems Native Data Ad-Hoc Vitals Monitors Nuvon VEGA Server Mechanical Ventilators Native Native Data Data IDC Enterprise Clinical Information Systems / IDM-MG 3000 IDM-Specific Electronic Medical Record Systems HL7 Data Native HL7 Data Native Data Mechanical Ventilators Data Ad-Hoc Vitals Monitors Native Physiological Monitors Data . Native IDC . Data Enterprise Clinical Information Syste . IDM-MG 3000 Electronic Medical Record System IDM-Specific IDM-MG 4000 Data HL7 Native HL7 Data Native Data Data Infusion Pumps Physiological Monitors Ad-Hoc Vitals Monitors Native Data IDM-MG 4000 Infusion Pumps IDM-Specific HL7 Data HL7 Data Native Data
  7. 7. Logical Architecture MS Device monitoring Drivers` CSV or TXT Data file DAG Device DCA polled push HL7 Clinical Information System Network Time Server
  8. 8. Configuration DCA is accessible via its web page through laptop of DAG  Through this web page, parameters may be included or excluded from offered set (or completely replaced relative to the offered set).  Note that parameter description, its code, and units of measure are defined at this level with respect to specifications of manufacturer. At DAG it is possible to alter parameters  Common codebook allows parameter semantic synchronization so that user can receive these parameters in unified codes that are desired by user  Parameters are synchronized in time for output so all data from single patient aligned with one UTC time increment 8
  9. 9. Time Synchronization Each device collects data in accord with its own time clock.  Data are aggregated and are written out synchronized to a single clock at DAG.  Each medical device has a separate clock—not synchronized with any universal clock or with each other.  Each medical device is polled at a separate frequency.  As data are collected on each DCA, these clocks are aggregated locally: that is, data are packaged for transmission to DAG along with DCA time stamp—UTC. Finally, on DAG, times of individual devices are synchronized according to a universal, UTC-based clock. 9
  10. 10. Network Time ServiceMedical Device-1 UTC UTC Text DCA DAGMedical Device-2 … EMR . . .Medical Device- M 10
  11. 11. Time Sync Process DCA DAG Auto poll Manual frequency Aggregator Output PCD k = 1   Obtain Device x k (t k ) Output PCD x k (t Rk ) Output PCD  t k  t0 Measurements  Measurements such that Data such that x R (t R ) x k (t k ) tk  t Rk and tk  t Rk  t Rk t Rk  t R and t Rk  t R  t R (start of (end of flow)data query) tk  tk  tk t Rk  t Rk  t Rk t R  t R  t R Auto Manual PCD k = 2  Obtain Device x k (t k ) Output PCD t k  t0 Measurements  Measurements such that x k (t k ) tk  t Rk and tk  t Rk  t Rk (start ofdata query) tk  tk  tk . t Rk  t Rk  t Rk . Auto . Manual PCD k = M  Obtain Device x k (t k ) Output PCD t k  t0 Measurements  Measurements such that x k (t k ) tk  t Rk and tk  t Rk  t Rk (start ofdata query) tk  tk  tk t Rk  t Rk  t Rk 11
  12. 12. Selecting Common Reporting Time tR1 Device 1 Reporting Time tR2 Device 2 Reporting Time . . . tRk Device M Reporting Time Common Reporting Time12 tR
  13. 13. Reporting When No Data ExistDevice 1 Reporting Time NULLDevice 2 Reporting Time . . .Device M Reporting Time tRCommon Reporting Time t0 t1 13
  14. 14. Terminology Synchronization Master codebook is stored on DAG.  Updated by driver specifications of each device and is modified and maintained within DAG in data file format.  It is modified by user to user-defined specifications.  Dictionary is linked to individual device dictionary through codebooks that map the device-specific parameters to a common codebook. Devices have their data vectors mapped to common unified codes representing specific data elements within master codebook.  Parameters from each device may be mapped to common unified codes within codebook.  Therefore, this codebook establishes a common dictionary of terms that are then written out to file or as HL7 messages to end user clinical information systems. 14
  15. 15. Parameters: Network Time ServiceMedical Device-1 Device-1: UTC UTC A1, A2, A3, … Text DCA DAGMedical Device-2 … Device-1: B1, B2, B3, … EMR Master Dictionary: . A1 A . A2 B . A3 C … DMedical Device- M E B1 B2 F Device-1M B3 . M1, M2, M3, … … . … . M1 M2 M3 … 15
  16. 16. User-defined Mapping Device 1 HR SpO2 NBPs User-Selected Mapped Output NBPd NBPm HR-ECG ARTs HR-SPO2 ARTd ARTm SpO2-1 CO SpO2-2 PVC etCO2 NBPs … Device 2 NBPd HR NBPm SpO2 fR RR Mve Tve fRe MVe PIP etCO2 TVe …16
  17. 17. Terminology Sync Auto defined` User defined Union OutputInput ~  x cb  Y      x cb  null ~ ~ ~xR x  x cb   x cb x cb  x cb U  x cb x cb N    x cb  x cb U  x cb 17
  18. 18. Data Output Output data are written to a comma-delimited format file in a local directory on the DAG laptop. Data between the DCA and DAG are constructed using three separate elements  (1) Patient Info Table;  (2) Param Header Table;  (3) Param Values Table. 18
  19. 19. Patient Name(32)+0x2c Patient ID(32)+0x2c Bed Location(32)+0x2c Device Name(32) +0x2c 0x0d 0x0aParameter Header:DCA_TS + 0x2c Chosen Device Parameter Codes and separated by comma (n*paramlen+(n-1) 0x2c(commas)) 0x0d 0x0aParam Values: This is the protocol for each MSH Segment—DCA_TS_value+0x2c Chosen Device Parameter Values (n*value_len+(n-1) 0x2c (commas)) 0x0d 0x0a…N Parameters Device 1 DAG DCA Device 2 CSV FILE + Parameter HL7 Patient information selection and from ADT periodicity of (optional) measurements 19
  20. 20. CSV Output20
  21. 21. CSV Data Plot Time synchronized Time measured from start of data collection (seconds)21
  22. 22. Conclusions A tool has been developed that acquires all data from up to eight patient care devices connected to a patient Tool semantically and temporally harmonizes and generates a readily useable spreadsheet of time synchronized data that can be imported to any statistical analysis software. This tool should greatly enhance the capabilities of clinical investigators to do physiologic research. 22
  23. 23. NomenclaturePoC Point of Care: physical location at bedside proximity of patient.DCA Data Collection Appliance: physical device containing 8 serial ports that communicates with the medical devices at the point of care.DW Data Warehouse: repository for locating, storing, and retrieving data.EMR Electronic Medical Record: ―a longitudinal collection of health information in electronic format for and about persons‖ – derived from the 2003 IOM Patient Safety Report.DAG Data Aggregation Gateway: the system component that receives disaggregated patient care data from DCAs and aggregates, formats, synchronizes and harmonizes these data for communication to an EMR or for writing to a DW or a file. 23
  24. 24. NomenclatureMS Management Server: software that maintains a repository of device drivers used by the DCA to communicate with medical devices; maintains a web server that enables viewing individual DCAs and DAG through Web pages; manages a database server and receives health and status of DCAs and DAG for presentation in a web dashboard to facilitate general management and oversight of the system.HL7 Health Level Seven: a non-profit organization that maintains a collection of interoperability standards associated primarily with data communication and interoperability of healthcare systems.ORU^R01 Observation Request-Unsolicited Results Transaction: one of several HL7 formatted transactions used in standard practice to communicate observations such as measurement data obtained from medical devices.WiFi Term used to describe wireless communication.RJ45 A registered jack connector and wiring pattern normally used for connecting computers and similar appliances over high-speed networks. 24
  25. 25. NomenclatureO/S Operating System: based set of programs that operate the hardware of a computer hardware resources.PACU Post Anesthesia Care Unit: an area within a hospital that typically adjoins the surgical area in which patients are moved to recover from the effects of anesthesia.MB MegaByte: 220 Bytes or 1048576 bytes.DB-xx Designation of port connectors for attaching devices serially to computers. Common connectors are DB-9 (9 pin serial connector, either male or female), DB-15 (15 pin connector), etc. These connectors are used to interface RS-232 connections with said computers.UTC Coordinated Universal Time: is a time standard based on international Atomic Time.XML Extensible Markup Language: a standardized and flexible mechanism for sharing information formats following a format standard designed for Web-based documents.Vector An array of scalar values with each value representing a different attribute of a unique object, with device and UTC time stamp appended. 25
  26. 26. Symbols DAG, Medical Device DCA MS NTS Software functions Database Proprietary control Text (RS232) (TCP/IP) (TCP/IP)26