Using Modeling & Simulation to Mitigate Virtual Health Application Implementation Risk Additional Sources: - 3 Steps to Faster EMR Adoption with Desktop Virtualization and SSO White Paper by VM Ware and Imprivata - CDW Health Care The Healthcare IT Tipping Point Dr. Rusty Baldwin & Mr. Mark Danis 2 April 2012 Developing the Cyber Warrior Through Education and Research
Purpose Outline Industry analysis on Virtualization and improving Health Information Technology - End User Experience (EUE) Identify issues and challenges with affecting the adoption of Health Information Technology Identify strategies to overcome resistance to the adoption of Health Information Technology Provide specific recommendations on how to minimize both the cost and the risk of a Virtualization strategy 2
Agenda Overview a current Application Virtualization Health Environment (AVHE) vs the current EHR Thick-Client Analyze system performance and End User Experience Analysis from Lovell FHCC (North Chicago) Review future AVHE Architecture Environment Integrating two independent Health Facilities Introduce Network and Application Modeling Approach 3
Bottom Line Up Front THIS IS FUNDAMENTALLY A CHANGE MANAGEMENT ISSUE Preliminary analysis of End to End (E2E) data suggests AVHE implementation may not be a big network challenge Weighted average network latency difference between AVHE and AHLTA thick-client at Lovell MTF is negligible (0.8s 1 ) Typical network latency is relatively fast at just 3.1s 1 AVHE may be an End User Device (EUD) application challenge Despite relatively fast network performance data, Lovell clinician observations show user wait times as long as 90-seconds Currently analyzing AVHE EUD application performance 1 Based on E2E network latency data comparison between 344K AVHE and 293K AHLTA thick-client transactions at Lovell MTF in the month of November 2011 4
Virtualization for a Clinician For physicians, nurses, and other clinicians, desktop virtualization can improve productivity and satisfaction Allows Single Sign On (SSO)/Context Management (CM) Personalized desktop experience travels throughout their day as they move between rooms & workstation For clinicians to embrace and use the new technology, it must make their jobs easier, more efficient, not harder 5
Virtualization for a Clinician The one consistent lesson learned from EHR efforts is that you must have physician acceptance of technology The best software in the world will not facilitate clinical workflow if it is not accepted and used by clinicians In order to improve adoption of virtualization solutions - focus on what clinicians need to be effective, productive * Must move from Input/Response to clinical workflow process of Information, Process, Resolve and Move On * Dr. Jonathan Nebeker, MS, MD VHA Office of Informatics and Analytics 6
Virtualization for IT Support The actual devices in patient rooms and nursing stations can be less expensive, thin, browser-based desktops Maintenance reduction for PCs throughout the medical treatment facility IT can install, maintain, backup and manage software and data in data center. Ensures consistency of configurations Centralized control of applications and data can deliver improved availability and performance to clinicians 7
Virtualization for Security Patient data itself remains in the data center, rather than being distributed throughout a medical treatment facility From a patient consent perspective, we are exploring how virtualization can protect Personal Health Information Health information sharing initiatives between DoD/VA and commercial health care providers can be improved 8
So What Is The Problem? IT professionals need to carefully balance investment between endpoint solutions and the infrastructure The Benefits: 43% of caregivers view new solutions, with their enhanced functionality, as more useful in patient care 34% believe the new applications are more able to effectively deliver the clinical information they need However: 41% of caregivers also rate new solutions as slower 20% of caregivers rate new applications harder to use 9
Health IT Investments Sample: 200 Healthcare IT Professionals 2012 Electronic health record Computerized physician order entry (CPOE) Barcoding Patient/visitor network access Point of care Picture archiving and communications system Video conferencing/collaboration Patient kiosks Telemedicine Radio Frequency Identification RFID tracking 10
Cases For Concern Storage Infrastructure 56% Deployed an EHR in the last 18 months - 4% added no additional storage capacity Network Infrastructure 24% Added patient/visitor access to the network in the last 18 months - 10% have added no network capacity IT Security Infrastructure 56% Deployed an EHR in the last 18 months of those - 11% have added no additional IT security 11
Balance And The Caregiver Experience Sample: 101 physicians; 101 nurses 2012 Beyond the tipping point, the addition of new endpoint solutions without adequate IT infrastructure actually makes the caregiver experience worse, not better Relative to older systems, systems implemented in the last 18 months are providing a different experience 1-3 New Systems 4-6 New Systems More useful in patient care 46% 41% Available more frequently 19% 6% Faster to use 20% 9% 12
The Health IT Tipping Point Defined: The Healthcare IT Tipping Point The point at which an additional $1 of investment in any solution reduces the caregiver experience Endpoint: IT that is actually adopted/used by caregivers Infrastructure: IT that provides the storage, computing and network capacity to support activity at the Endpoint Successful implementation of Healthcare IT systems must equip caregivers to deliver better patient outcomes This requires a balance between endpoint solutions and the computing, network and storage infrastructure 13
A Painful Truth Build 1 st, Balance 2 ND When IT departments deploy additional infrastructure, it is often after-the-fact and in response to user complaints Nearly 80% of IT professionals confess to adding infrastructure this way, and without pre-planning With tens of millions of patient records going digital in the coming years, how much should we focus on security? Hint: Probably a good deal more than we are currently This must include patient privacy or econsent 14
WPAFB Ft Campbell Application Virtualization Stand-up Application Virtualization Hosting Environment (AVHE) at Ft Campbell with Citrix server farm Clinicians at WPAFB access Ft Campbell virtual instances of AHLTA via Virtual Private Network (VPN) WPAFB Local Cache Server (LCS) & Composite Health Care System (CHCS) remain at WPAFB Questions under analysis: What are the key application and network factors? What performance impacts expected by End User? 15
Notional AHLTA Virtualization Network Topology WPAFB Enclave DISA DECC Montgomery Ft Campbell MAG LCS Server CHCS Server CDR AVHE Server Virtual AHLTA Application AHLTA Thin-Client EUD WPAFB MTF Routers/Switches SDP AFNet Increment 1* (USAF BLOCK 30 Gateway) DISA GIG? OPNET Application Traffic Monitor * Happens to be a USAF BLOCK 30 Gateway at WPAFB Fort Campbell Black Box (Firewall(s), Routers/Switches, Enclave Servers) 16
Modeling and Simulation Military Health Network Modeling and Simulation Objective is simulation resource toolkit to develop network simulations for particular questions Performance of as-is, to-be architectures, network impact of new devices, Central Data Repository (CDR), architecture Develop network components, clinical software models and clinician workflow needed for OPNET simulation engine Example components: routers, bridges, AF Block 30 gateway, base/mtf infrastructure Example application models (AHLTA, VistA, clinician workflow) 17
Modeling and Simulation Approach: Assess alternatives by comparing to baseline configuration via simulation and/or measurement There is no one-size-fits-all simulation model Type/fidelity of model must be based on the questions model is to answer i.e., what is the mission of the model? Military aircraft are tailored to expected mission There is no single fighter/bomber/ close air support/troop transport/ cargo aircraft The same holds true for simulation models! 18
Modeling and Simulation Models long-haul network Objective: Measure GW to GW performance, availability, efficiency, impact of using NIPRNet failover paths Node level metrics: node throughput and delay to/from a node behind a GW and remote server in MHSi Enterprise level metrics: GW to GW throughput, delay, packet loss, utilization 19
Modeling and Simulation Model MHSi node transactions Objective: Measure resource usage/performance of MHS Enterprise node in various configurations/usage profiles Test virtualization environment Node metrics: Transactions per second, mean delay, memory/disk utilization, others 20
Lovell Architecture 21
So What Did We Find Conducted elicitation meetings with individual clinicians and IT support, witnessing both workflow and latency Observed significant frustrations with complexity of workflow with multiple EHRs (lack of interoperability) Observed discrepancy between provider expectations of performance, and actual AVHE EUE of performance Specific issues: Tier 1 errors Single sign-on (SSO) Context Management (CM) occasionally display concurrent session message 22
Transaction Top Volume Transactions Nov 2011 AVHE Response 90th Time Percentile Mean Response (Sec) Occurrences (ms) Time AVHE vs AHLTA Thick- Client DoD Full Client 90th Response Percentile Time Response (Sec) Occurrences Mean (ms) Time ORD Requirement Select Patient 59,374 0.91 2.03 0.37 46,914 0.54 1.03 6 Load Template 32,569 2.25 9.38 1.79 40,561 0.46 1.13 5 Workflow-Prev Encounters Returned 27,114 1.83 4.03 0.31 11,967 1.52 3.05 N/A Appt - Open Module 21,793 2.64 6.22-0.57 10,291 3.21 7.02 25 Patient Search - Search for Patient 17,706 0.83 2.09-0.06 13,831 0.89 2.24 6 Workflow-CHCS I Login and Patient Select 12,352 4.46 6.61 1.25 14,109 3.21 5.05 N/A Previous Encounters - Open Module 12,089 3.58 6.98 0.16 8,869 3.42 6.11 6 Workflow-InitAllTransactionsForCore 9,662 0.03 0.05 0.01 3,977 0.02 0.02 N/A Sign - Sign Encounter 8,858 5.69 9.45 2.46 10,605 3.23 5.59 25 A/P - Diagnosis - Run Search 7,571 0.17 0.45-0.02 5,138 0.19 0.41 6 Security - User Logon 7,451 18.71 21.78 1.83 4,972 16.88 22.64 25 S/O - Open Module 5,351 5.79 10.33 1.64 8,327 4.15 6.74 6 Workflow-CHCS I Get Orders 5,122 0.21 0.3 0.15 5,099 0.06 0.08 N/A Vital Signs Entry - Close Module 4,922 0.24 0.59 0.07 4,314 0.17 0.03 6 Workflow-A/P >> Disposition 4,795 8.98 14.2 5.35 9,285 3.63 7.09 N/A Telcon - Appt - Open 4,764 7.37 11.7 1.54 340 5.83 10.8 25 S/O - Close Module 4,548 2.34 4.16 1.1 1,890 1.24 2.42 6 A/P - Med - Search 4,541 0.6 0.86-0.24 3,555 0.84 1.2 6 A/P - Open Module 4,541 7.02 10.44 0.26 3,812 6.76 10.88 6 Telcon - Open Module 4,491 2.1 5.02-0.38 273 2.48 5.41 25 Vital Signs Entry - Open Module 4,158 2.61 4.88 0.6 3,326 2.01 4.11 6 Workflow-Disposition >> Sign Encounter 4,150 3 4.17 1.07 5,118 1.93 3.11 N/A Appt - Open Encounter - Auto-cites 4,021 20.1 8.58 6.57 2,988 13.53 9.27 25 Patient Search - Cancel 3,679 1.62 0.27 1.59 1,134 0.03 0.05 6
Lovell Network Performance Comparison Analyzed E2E transaction latency data for AVHE and AHLTA thick-client Data includes AVHE and AHLTA thick-client transactions for 1 to 30 November 2011 Outlier transactions of over 600s were excluded from the analysis Our Analysis AVHE AHTLA Thick-Client Weighted Average Transaction Latency Range of Transaction Latency Differences for 21% of transactions driving 80% of volume 3.1s 2.2s -0.5s to +6.5s (Negative is AVHE faster and positive is AVHE slower) 24
Lovell MTF: Preview of WPAFB Ft Campbell Virtualization Impact? Lovell presents unique opportunity to see AVHE and AHLTA thick-client applications operating side-by-side Lovell AVHE configuration vs proposed WPAFB-Ft Campbell: Similarities to WPAFB-Ft Campbell Variances from WPAFB-Ft Campbell Lovell AVHE crosses multiple enclaves (Lovell MTF is multienclave facility) as would proposed WPAFB-Ft Campbell Lovell and WPAFB are both large MTFs with robust AHLTA traffic Lovell AVHE is within final mile vs geographic separation of WPAFB to Ft Campbell... Speed of Light in Glass argument may make this variance irrelevant Lovell AVHE does not have to contend with AFNET Increment 1 (USAF BLOCK 30) 25
Recommendations Interrogate and Evaluate what is happening inside the End User Device (EUD) Investigate what applications are doing inside the EUD that might be causing user to experience latency Analyze and model EUD application elements: Operating System VPN Software 26
Next Steps Determine AVHE application within unique architectures and determine best means to analyze application performance Install OPNET devices at Lovell MTF to capture more discrete network and application performance data Conduct initial elicitation meetings at Ft Campbell Initiate coordination with Ft Campbell to gain understanding of AVHE host site architecture 27
Questions Rusty O. Baldwin, PhD, PE, CISSP Professor of Computer Engineering Research Director, Center for Cyberspace Research Department of Electrical and Computer Engineering Air Force Institute of Technology e-mail: rusty.baldwin@afit.edu Mark Danis Principal Morgan Borszcz Consulting e-mail: mark.danis@mbc360.com 28