System dynamics simulation modeling of health information exchange (HIE) adoption and policy intervention: A case study in the State of Maryland

Accepted Manuscript System dynamics simulation modeling of health information exchange (HIE) adoption and policy intervention: A case study in the State of Maryland Emad A. Edaibat, Jason Dever, Steven M.F. Stuban PII: DOI: Reference:

S2211-6923(16)30008-X http://dx.doi.org/10.1016/j.orhc.2017.02.001 ORHC 115

To appear in:

Operations Research for Health Care

Received date: 29 January 2016 Please cite this article as: E.A. Edaibat, J. Dever, S.M.F. Stuban, System dynamics simulation modeling of health information exchange (HIE) adoption and policy intervention: A case study in the State of Maryland, Operations Research for Health Care (2017), http://dx.doi.org/10.1016/j.orhc.2017.02.001 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

System Dynamics Simulation Modeling of Health Information Exchange (HIE) Adoption and Policy Intervention: A Case Study in the State of Maryland Emad A. Edaibat*, Jason Dever, Steven M. F. Stuban Department of Engineering Management and Systems Engineering, the George Washington University, Washington, DC 20052, USA *Corresponding author. Tel.: +1 214 517 9462; Email: [email protected]

Abstract In this paper, health information exchange (HIE) adoption barriers, challenges, influencing factors, and the impacts of policy interventions among ambulatory providers and acute-hospitals in the State of Maryland health care system are examined. The main areas discussed are HIE sustainability, financial benefits, return on investment, and the correlation between HIE and hospital readmission reductions. The proposed policies include financial incentives to adopt HIE (policy 1), awareness and education to integrate HIE into the workflow (policy 2), and a combination of policies 1 and 2 to address the most frequent barriers identified in the literature. System dynamics simulation modeling combined with statistical analyses were utilized based on monthly time-series datasets. The design for each policy was developed considering HIE adoption barriers, challenges, and influencing variables. The simulation focused on presenting the findings of many HIE adoption studies. The results suggested significant financial advantages of using HIE (an approximately $3.3 billion cumulative gain over 10 years). Ambulatory provider adoption of HIE is slow but contributes the most to the overall HIE portal queries. Three datasets were analyzed with regard to hospital readmission reductions, HIE portal query usage, and encounter-notification service alerts. Strong positive correlation coefficients of 0.75 and 0.8, respectively, were determined. Finally, the designed policy interventions (policy 1, policy 2, and combination) showed a positive effect on HIE adoption rate by 15.87%, 8.36% and 21.17% respectively. This research can be used as a framework to provide policy makers and strategic thinkers with a methodology for analyzing such complex systems and generate well informed HIE adoption policies.

Keywords: Health information exchange, health policy, system dynamics, systems engineering.

1. Introduction 1.1. Overview A health information exchange (HIE) is a mechanism to securely share and access medical records and other data electronically across health care providers. The concept of HIE is widely used and well established; however, HIE is a relatively new system that was initiated in 2009 by the Health Information Technology for Economic and Clinical Health (HITECH) Act. The HITECH Act allocated federal and state funds toward accelerating the adoption of health information technology, including HIE, to improve health care quality, efficiency, and costs [1]. The potential HIE benefits include reducing hospital readmissions, unnecessary hospitalizations, and duplicative laboratory and radiology testing; improving care management, patient satisfaction, and safety; and avoid emergency department (ED) visits and adverse drug events [2–7].

Several barriers and challenges to HIE adoption exist, including high implementation costs, integration and use in workflows, difficulty in quantifying cost savings and financial benefits, privacy and security concerns, high competition among health care providers, lack of technical standards, and lack of interoperability [8–11]. Despite these challenges, many HIE systems across the United States have been implemented and have evolved. Long-term HIE financial sustainability is a concern for the post-implementation phase [12–16]. Hence, the continuous improvement in HIE adoption and usage plays a role in reforming and shaping the quality of health care and its associated costs.

Since the initiation of HIE, a clear and rapid increase in its adoption has occurred each year. In 2014, 76% of non-federal acute-hospitals nationwide electronically exchanged health information with external health care providers compared to 44% in 2010. The rate of HIE adoption in the state of Maryland, for example, was 88% in 2014 compared to 53% in 2010 [17–19]. However, apart from drug e-prescribe data, no clear public data or reports on ambulatory-provider HIE adoption rates with nationwide external health care providers exist. Using data from the National Ambulatory Medical Care Survey in 2013, Furukawa [20] determined that 39% of ambulatory providers reported using HIE with other external providers, which is significantly lower than

hospital HIE rates. However, Furukawa did not address the frequency of HIE usage among ambulatory providers.

1.2. Research Objectives The aim of this study was to develop a system dynamics (SD) approach to build a framework for evaluating, analyzing, and presenting a solution for an HIE system with consideration of all related challenges. The proposed SD simulation model can be used to model other HIE systems across the United States because it is both replicable and adaptable to different settings. In addition, policy interventions to foster the accelerated adoption and utilization of HIE were designed.

The proposed simulation model links policy interventions, costs, variables, causality, and decision making in a collaborative information-sharing environment. The present paper reviews aspects of the HIE system considered for the creation of our model and discussed by other researchers: HIE evaluation [21], implementation [22], adoption and benefits [12, 13], usage [2, 5, 6, 25], return on investment (ROI) [3, 5, 24], sustainability [14, 15, 23], role in hospital readmission reduction [6, 24], policies [26], and financials from the establishment phase to the post-implementation phase.

Several existing SD models (causal loop and simulation models) were previously proposed for the adoption and cost benefits of health information technology (HIT) and electronic health records (EHR); SD models of HIE, HIT, and EHR are not new. The novelty of the present work is the calibration of the SD model to describe HIE use and uptake and to study HIE adoption and cost benefits for the State of Maryland.

1.3. Background Based on a review of the literature and the collected datasets (discussed in Section 2.1), we posit a set of problems that we address regarding HIE adoption and use.

First, in terms of HIE ROI, the difficulty of quantifying the financial value added by HIE is a common issue [3, 5, 8, 11, 24]. The financial benefits of the HIE system are significant to the adoption discussion. However, relationships among ROI and other HIE system elements exist that should be examined as well.

The fragmentation of information among health care stakeholders is not surprising and necessitates the creation of lean practices to eliminate waste. Waste can take the form of duplication of laboratory and radiology tests, hospital readmissions, repeated ED visits, and time required to search for medical records. Augmenting the effectiveness of health care systems and decreasing waste will ultimately increase the influence of HIE adoption. However, HIE system implementation incurs additional costs, such as investment in new infrastructures, operation, and research and development, i.e., for enabling EHR and HIE system interoperability. Moreover, training medical providers on using new platforms and the time required in using the EHR/HIE systems in the day-to-day workflow have additional related costs. We thus determine if the State of Maryland HIE will realize a significant financial gain and ROI.

Second, HIE usage in hospitals and ED settings has been the focus of previous works [2, 5, 6, 25]. Nevertheless, no adequate studies have been conducted in terms of other health care providers (such as ambulatory services). The rationale behind this problem statement is to highlight the importance of future studies in focusing on HIE adoption and usage by ambulatory providers with consideration of the large population of these providers compared to hospitals and EDs. We therefore investigate the validity of previous works using a graphical representation of a real-world dataset and provide our subsequent perceptions.

Third, the HIE effect on hospital readmission reduction is examined. A promising benefit of HIE is the reduction of hospital readmission [6, 24]; therefore, HIE can be used as a tool to provide better care management. We therefore specifically investigate if a correlation exists between HIE and hospital readmission reduction.

Fourth, HIE sustainability should be addressed. The Chesapeake Regional Information System for Our Patients (CRISP) received state funding of $10M dollars for HIE establishment (2009) to build an HIE infrastructure and core services (e.g., portal queries, the Encounter Notification Service (ENS), and the Prescription Drug Monitoring Program (PDMP)). The CRISP initial financial sustainability application for state funding asserted that CRISP would be a self-sustainable organization by the fifth year of the grant [27]. The CRISP request for application (RFA) response stated that, in its HIE sustainability business model [28], the use of participant subscription fees is a method used to reach this goal. As of May 2015, out of all participants, only payers (insurance companies and government agencies) and hospitals pay the subscription fees, which is then used to cover ongoing operation costs and core HIE services (approximately $4M). According to CRISP, the State of Maryland HIE is sustainable, and many current studies and reports concur with this assertion [12, 13, 14, 16].

However, today’s HIE requires integration within the existing health care provider workflow, i.e., interoperability of the HIE system with EHR [9]. The uncertainty of future HIE expansion projects was not considered [28]. We thus present the research question: Provided the demands for interoperability across different EHR platforms, is the State of Maryland HIE sustainable in the long term?

Finally, in terms of policy interventions, Kruse et al. provided a systematic study of HIE barriers over time by reviewing the literature from 2009 to the first half of 2014 [8]. Their study provides a summary and breakdown of each barrier frequency over that period and found that cost and efficiency/workflow are the top two barriers hindering HIE adoption. Based on their study and the available datasets, the following policies are herein proposed: Financial incentives to integrating HIE into the provider workflow, and promoting awareness and education to foster that integration.

Financial incentives (Policy 1). As discussed earlier, a holistic HIE system ROI in the State of Maryland is an important factor for its success. Comprehensive reform, including overhaul of the legislation and strategic planning, is required. If a significant ROI is realized, it should be distributed in the form of financial incentives back to the health care providers to accelerate HIE adoption and advance current HIE to HIE 2.0 and beyond

(e.g., big data concepts, data analytics, early prediction of epidemics, and improvement of patients diagnoses and prognoses) [9, 11].

Advancing awareness and education for integrating HIE into the provider workflow (Policy 2). This policy relates to the subjective response to any change, such as that of potential HIE adopters to using the new system. Addressing aspects of human nature necessitate cautious planning to eliminate and/or reduce any resistance to change. According to [29], factors that negatively influence the integration of HIE in the workflow include changing the work patterns and/or introducing new processes to existing job workflows. Health care providers may believe that HIE can add more complexity and cause workflow interruption. HIE is not intuitive in the sense that it requires additional logins and access to different platforms, thereby adding extra efforts in daily work activities [8, 30].

2. Methods In this study, we employed SD simulation and modeling techniques combined with statistical data analysis. SD is a well-established strategic planning approach that has been recently leveraged in health care applications, such as evaluations, modeling, and simulation studies [22, 31–33]. Moreover, it is a holistic approach to mapping relationships within complex systems for modeling system nonlinearities, feedback loops, and delays and for designing policy interventions [34, 35].

The system-level analysis undertaken is described herein as follows. First, the data collection method and sources are reviewed. Second, a qualitative model of the barriers, challenges, and influencing factors of HIE adoption is described. Third, a quantitative simulation model is presented. Finally, an experimental design method is proposed to address each problem statement and outline policy interventions.

2.1. Data Collection The present research involved extensive data collection to enable realistic and reliable modeling and forecasting. It was imperative to elucidate a comprehensive view by expanding the scope beyond the focus of the State of

Maryland. Thus, to comprehend HIE system causality, it was essential to study other successful and unsuccessful HIE systems nationwide and to identify the primary challenges and barriers they encountered, and the insights acquired accordingly in terms of federal funding and Center of Medicare and Medicaid (CMS) incentive programs. The data collected extended from September 2010 to February 2015 and included an immense amount of information on many HIE aspects.

To develop the simulation model of HIE adoption in the State of Maryland, two data collection methods were followed: qualitative and quantitative. Fig. 1 shows the structure of the data collection methodology. The qualitative phase included an extensive and comparative analysis of the literature: state HIE reports, surveys, and analytical studies; HIT reports; 2013/2014 reports to US Congress on HIT adoption to model the qualitative HIE adoption barriers, challenges, and influencing factors; and HIE establishment and sustainability reports from several states, including Connecticut, Indiana, Louisiana, Maryland, New York, Texas, Virginia, and Wisconsin.

The quantitative data collection began in June 2014 and included the communication and requisition of many sources, such as the Center of Medicare and Medicaid Services (CMS) [36], HIT [37], the Maryland Health Care Commission (MHCC) [38], and CRISP [39]. The chief of HIE at MHCC provided statewide monthly timeseries datasets (September 2010 to February 2015) on HIE adoption and usage. This dataset contained querybased HIE portals, ENS and PDMP adoption by facility, HIE usage volumes of portal queries (i.e., HIE web portal patient information lookups and inquiries made by health care providers), ENS, and the volume of radiology and laboratory reports uploaded to HIE. A CRISP monthly dataset (January 2012 to March 2015) was provided for non-federal acute-hospital inpatients, including the patient count and cost of readmissions and ED visits. Data were also extracted and processed from the evaluation report [26,27], dissertations [31,40], and literature [8, 13, 15, 18].

  Fig. 1. H HIE data collecction (MHCC = Maryland H Health Care Commission, C CMS C = Centerr of Medicare and Medicaidd Services, CR RISP = Chesap peake Regionnal Information n System for Our O Patients, R Rad.= Radiolo ogy, Lab.= Laaboratory).

    2.2. Qu ualitative: Cau usal Loop Mo odel ptual causal lo oop model (CLLM) was develloped, as show wn in Fig. 2. This model hass been adopte ed and  A concep enhanced d from previo ous work in the literature [332, 40] and pu ublished by fro om current woork authors [4 41]. A  CLM has two main loo ops: reinforcing loops, whic h have a posittive (increasin ng) effect on tthe system denoted by  (R), and b balancing loop ps, which have a negative ( damping) effe ect on the system denoted  by (B). (R) and (B)  loops dep pict barriers, cchallenges, an nd influencingg variables thaat affect HIE adoption. The  causality loop ps used in  the simulation model are hereafter defined as staated below.   B2: The lack of interop perability and standards beetween differe ent EHR/HIE platforms negaatively impactts HIE  adoption n. However, th he increase in demand for H HIE will call fo or a more robu ust body of staandards to go overn  EHR/HIE information eexchange, which will minim mize the intero operability gap p.     B3, B4: TThe cost of HIEE implementation, researchh and develop pment (R&D), and sustainabbility are key  challengees in short and d long terms tthat negativelly affect HIE adoption. Fede eral and state  e funds provide initial 

seed funding to establish and implement the HIE system. However, without significant financial HIE benefit  value‐added, the HIE future will be questioned in terms of ongoing operations and development costs.     B5: The integration of HIE into health care provider workflows will slow HIE adoption and usage. However,  awareness and education campaigns can accelerate HIE integration in provider workflows.    R2: The increase in HIE benefits and value‐added will have a positive impact on HIE adoption and will solve the  underlying issue of long‐term HIE sustainability.     R3: Federal and state funds serve as the prime accelerator of HIE establishment and incentivize health care  providers to use HIE systems, i.e., stage 1 and 2 programs of EHR meaningful use.     R4: The increase in HIE demand and use will have a positive holistic impact on HIE adoption and will improve  other aspects of the HIE system, such as value‐added, sustainability, and interoperability.    The interrelations between the above variables are dynamic, nonlinear, and can have an infinite cause and  effect relationship on each other. CLM is not the focus of this paper; nonetheless, it provides the necessary  groundwork for the SD simulation model in which variables are analyzed. Some of the CLM loops have been  omitted in the final simulation model because of the lack of datasets or because they are not applicable to this  case study. 

  Fig. 2. US U nationwide HIE adoptionn CLM (R = reinforcing r loo op, B = balanccing loop)

  2.3. Quantitative: Siimulation Mo odel Complexx systems, such as those of the t U.S. healthh care compleex, display cou unterintuitive behaviors [42 2, 43]. The compplexity of HIE E systems thuss requires systtematic evaluation for deveeloping SD moodels that sim mulate HIE adoption over the longg term. Develo oping new leg islation and policies requirees an understaanding of the interrelattions and interrdependenciess in the system m to produce better b informed decisions annd strategies. This T simulatioon model enabbles assessmen nts of the effe ctiveness of various v policy interventionss that are inten nded to improve the adoption of o HIEs. Acco ordingly, the pproposed SD simulation s mo odel is based oon the Bass diffusion model, w which is a set of o differential equations forr modeling pop pulations that adopt innovat ations [35]. Am mbulatory providerss and acute-hoospitals Marylland are the evvaluated popu ulations (N).

The simuulation model is based on th he effective addoption rate, which w is a resu ult of barriers,, challenges, and a influencing factors. We W introduce a high-level moodel that we developed d with h which the reeader can visu ualize the variabless affecting the HIE effectivee adoption ratee, as shown in n Fig. 3. This simulation moodel uses the core c of the diffussion adoption elements in Fig. F 4, which aare based on th he Bass diffussion word-of-m mouth (WOM M) concept [35]. Thee boxes repressent stocks, wh hich dynamicaally accumulaate or deplete over o time, whhile the doublee-line arrows coonnecting the stocks represent in-flow orr out-flow dep pending on thee direction of tthe arrows. Th hese

flows aree controlled byy valves (adop ption rates), w which ultimateely are boundeed by the effecctive adoption n rate. Finally, tthe blue singlee-line arrows represent r the ccauses and efffects between the modeled vvariables.

  Fig. 3. High-level H model structure ((WOM = word-of-mouth Bass diffusion concept).

2.3.1 Con nceptualized Bass B Diffusio on Model

PA  E

i



Ii N (t ) N

(1)

The abovve equation exxplains the bassic concept off the Bass diffu fusion model for f new produucts and innovations, where PA denotes potenntial adopters, N(t) is the nuumber of popu ulations who have h already aadopted HIE, N representts the populatiion, Ei is the coefficient of iinnovation or exogenous influence, such as HIE barrieers, and Ii denotes tthe coefficientt of imitation or o endogenouus influence, su uch as HIE fin nancial incenttives.

 

Ad doption Rate Hospitals

Hosspitals Using HIE Barriers & Challenges Mulltipliers

Adoption Rate A Ambulatory

Effective Ad doption Rate

A Ambulatory Pro oviders Using HIE

Fig. 4. Siimulation moddel core-element overview (hospitals usin ng HIE = acutte- hospitals aadopting HIE; ambulatoory providers using u HIE = ambulatory a prooviders adopting HIE).

The detailed simulation model consisting of six major sections is shown in Fig. 5. First, the “main stocks and flows” section of the studied population is conceptualized in the form of the Bass diffusion model (Fig. 4). Herein, the model variables are the Ambulatory Providers Not Yet Adopted HIE stock (i.e., the total number of ambulatory providers that could potentially adopt HIE, N = 5,153) and the Ambulatory Providers Adopted HIE stock (i.e., providers that adopted HIE). Both stocks are interconnected with the Ambulatory Providers Adoption flow bounded by the Ambulatory Providers Norm Rate (i.e., the HIE adoption rate in normal scenarios) and Effective Adoption Rate (i.e., HIE adoption rate influenced by the HIE barriers and challenges). In addition, the Acute-Hospitals Adoption Real-World Data are imported from the monthly time-series dataset (N = 47). When the model was finalized and employed, all acute- hospitals in Maryland had adopted HIE systems. Accordingly, their adoption rate no longer required modeling, and the variable was eliminated from the model, as shown in Fig. 4. Hence, the acute-care hospitals stock and associated flow are omitted in Fig. 5.

Second, the HIE cost section includes the following variables: the Cost of HIE Establishment Real-World Data, which is the overall cost of CRISP establishment using state and federal funds, as stated in the MHCC report [27], and the annual operational cost of CRISP core services; ENS Cost, which is the cost of health care providers receiving and processing notifications; Portal Queries Cost, which is the cost of health care providers consulting any information on the portal; and Interoperability and Standards Cost, which is the cost of integrating the top ten EHR vendors in the HIE system.

Third, the HIE benefits and value-added section includes the following variables: Acute-Hospital Readmissions Reduction, which is the saving results from potential hospital readmission reduction using HIE; Duplicative Radiology Reduction and Duplicative Laboratory Reduction, which are savings from potential duplicate radiology and laboratory test reductions; Successful Queries of HIE, which are the time savings from a successful portal query on any medical record instead of using conventional means, such as printing, faxing, mailing, emailing, or telecommunicating; ENS Time Saving, which is the savings of encounter notifications

instead of the health care provider consulting the information by portal or telephone; and Avoidable Repeated ED Visits, which is the savings from potentially preventable ER visits by using HIE.

Fourth, the “HIE sustainability” section includes the following variables: Cost of HIE Establishment OPEX Real-World Data, Interoperability and Standards Cost, and Annual Fees from Payers, which are the annual subscription fees paid to HIE by payers, such as health insurance or health organizations; Annual Fees from Acute-Hospitals, which are the subscription fees paid to HIE by acute-hospitals; Ambulatory Provider Fees, which are the subscription fees paid to HIE by ambulatory providers; and HIE State Funds, which are the state and federal funds been awarded to HIE over a certain number of years prior to depletion.

Fifth, the “HIE demands and usage” section includes the following variables: HIE Demands Volume RealWorld Data, which is the volume of portal queries generated by all types of health care providers; and HIE Use and Integration in Workflow, which measures the pressure of integrating HIE into the health care provider workflow.

Effective Adoption Rate

HIE ROI Multiplier Ambulatory Providers Norm Rate

Ambulatory Providers Adoption

WorkFlow Multiplier

Security and Privacy Multiplier Security Privacy Lookup

Adopted HIE Lookup

Ambulatory Providers Not Yet Adopted HIE

Adopted HIE Multiplier

HIE ROI Lookup

HIE ROI

HIE Interoperability Cost with Top 10 EHR Vendors + Annual Interoperability Subscription Fees

Cost of HIE Establishment OPEX Real-World Data

Ambulatory Providers Adopted HIE

WorkFlow Lookup

Normalization Constant

HIE Use & Integration into Workflow HIE Demands Normalized

HIE System Cost Interoperability & Standards Cost

ENS Cost HIE Sustainability

Total Ambulatory Provider Fees

Fee per Ambulatory Provider

Security & Privacy

PQ Reduction HIE Demands Volume Real-World Data

Portal Queries Cost

Successful PQ Percentage

Cost per ENS look-Ups

Portal Queries Real-World Data

Total Acute-Hospitals Fees

ENS Volume Real-World Data

Total Payers Fees Fractional HIE Adoption Rate Total N

Fee per Acute-Hospitals

Acute-Hospitals Adoption Multi Cost

Successful HIE Portal Queries Time Saving

Labs Average Cost Value $

Successful Queries of HIE Duplicative Laboratory Reduction

ENS Time Saving

Acute-Hospitals Readmissions Reduction

HIE Benefits & Value-Add

AH Reduction Acute-Hospitals Adoption Real-World Data

Labs Volume Real-World Data

Lab Reduction

ENS Time Saving Value $

The HIE State Funds



Cost of Acute-Hospitals Readmission Real-World Data Acute-Hospitals Adoption Multi Emergency Department Cost

Avoidable Repeated ED Visits

ER Reduction



Duplicative Radiology Reduction

Rad Reduction

Radiology Volume Real-World Data Radiology Average Cost Value $ Cost of ER Readmissions Real-World Data

Fig. 5. HIE adoption simulation model use case in the State of Maryland (detailed in Appendix A).

Finally, the “barriers, challenges, and influencing-factor multipliers” section includes the following variables: HIE ROI Multiplier; Adopted HIE Multiplier; Workflow Multiplier; and Security and Privacy Multiplier. These variables are barriers, challenges, and influencing factors in lookup multipliers derived from the literature and datasets. They are then calibrated to reflect real-world scenarios. Please refer to Appendix A for all parameter/variable descriptions, equations, and units, as well as the model validation and calibration.

2.4. Experiment Design Here, the experimental design method is presented for each problem statement and policy intervention. In addition, we introduce the model’s core variables and assumptions and show how we exercised the simulation model to test the hypothetical claims discussed in Section 1.3. We use two methods. First, we implement the simulation model to address HIE ROI, HIE sustainability, and policy interventions. Second, we use statistical analysis and a data representation to analyze HIE usage and HIE effect on hospital readmission reductions.

The model’s main variable and assumption figures are listed in Tables 1 and 2. These values are based on the literature associated with HIE adoption, challenges, value-added, and potential financial savings. The full description of the variable and assumption selection bases are discussed in Appendix B. Table 1 lists the variable assumptions of hospital readmission reduction [24, 44], avoidable repeated ED visits [4], duplicative radiology and laboratory test reduction [2, 25, 45], successful HIE portal-query time savings [46], successful HIE queries [38], and ENS time-saving value [38]. Table 2 lists variable assumptions of HIE State of Maryland funds, HIE interoperability costs, annual fees, and an HIE operational cost escalator. We next discuss the method for each problem statement. The details of the sensitivity analysis method and results are provided in Appendix B.

2.4.1 HIE ROI The HIE ROI was simulated using directly imported real-world data. The assumptions are listed in Table 1; the values are in Table 2. The ROI section consists of two parts: the HIE system benefits value-added, and the HIE

system cost. The HIE benefits and value-added were simulated using the base run assumptions in Table 1. These were combined with real-world data on acute-hospital adoption; the respective costs of acute-hospital readmission, ER repeated visits, and ENS; and portal query, radiology, and laboratory data volumes. The HIE system cost included the costs associated with implementation, operation, interoperability, R&D, and health care provider person hours (values shown in Table 2). The ENS and portal query costs (i.e., person-hour cost) were simulated in combination with real-world data.

In addition, the sensitivity analysis was simulated using the “sensitivity analysis range” shown in Table 1 to test the robustness of the model under extreme conditions (see Appendix B).

2.4.2 HIE Usage To analyze HIE usage, we used a graphical representation of the real-world dataset. First, we graphed HIE adoption for both ambulatory providers and acute-hospitals to show the gap in adoption between the two settings for providing a more effective representation of HIE adoption among ambulatory providers (n = 373 adopted HIE out of the total population N= 5,153) and acute-hospitals (N = 47 total population that fully adopted HIE).

To measure HIE usage across health care provider settings, acute-hospitals were divided into two categories (in addition to ambulatory providers): acute-hospital ED, and acute-hospital non-ED (i.e., ED is not included). HIE data were collected for acute-hospitals (N = 47), including non-ED hospitals, ED hospitals, radiology centers, laboratories, cancer registries, and pharmacies. In addition, ambulatory provider usage (n = 373) was also collected. These data were analyzed to investigate the HIE usage volume per month in different heath care provider settings. Table 1. Model Parameters and Assumptions Parameter Description Acute-hospital readmission reduction Avoidable repeated ED visits

Parameter in the Model AH reduction

[Ref] [24, 44]

Base-Run Value 10%

Sensitivity Analysis Range4 10%–60%

ER reduction

[4]

10%

10%–30%

Duplicative radiology and laboratory test reduction

[2, 25, 45]

15%

5%–90%

[46]

40%1

30%–50%

[38]

25%1

10%–40%

[38]

$251,2

$15–$35

Ambulatory providers, N

[38]

51533

Hospitals, N

[38]

473

Successful HIE portal queries Time savings Successful queries of HIE ENS time saving value

Radiology reduction Lab reduction PQ reduction Successful PQ percentage ENS time saving value

1

Estimated 25% of all portal queries returns with a successful hit from personal communication. 15-min time savings over $100 per hour average cost. 3 Source: MHCC. 4 Refer to Appendix B for sensitivity analysis results. 2

    2.4.3 HIE Effect on Hospital Readmission Reduction The effect of HIE on hospital readmission reduction was investigated using three datasets: acute-hospital HIE usage (HIE portal inquires), ENS alerts generated, and the cumulative monthly acute-hospital readmission reduction. The HIE portal inquiry dataset consisted of the volume of HIE patient data and records accessed using HIE. The ENS alerts dataset represented the volume of real-time alerts that physicians were sent if active patients in their respective practices were hospitalized, discharged, and visited ED. After receipt of an ENS alert, a health care provider could then follow up with the patient and provide better care management, which can help avoid another ED visit and thereby reduce hospital readmission.

In the cumulative monthly acute-hospital readmission reduction dataset, we extracted the cumulative readmission reduction based on monthly variations. The readmission count was aggregated month-over-month; for example, the number of hospital readmissions in January 2012 was 7,259; in February 2012, it was 7,039; and in March 2012, it was 7,262. The cumulative readmission reduction at the end of March 2012 was –3, which was calculated as: 7,259

7,039

7,039

7,262

3

We graphically present datasets and the correlation between cumulative monthly acute-hospital readmission reduction and HIE portal inquiries. Furthermore, generated ENS alerts are analyzed to assess the prior claims that HIE will reduce hospitals readmissions.

Parameter

Table 2. HIE Sustainability Simulation Parameters and Assumptions Value

HIE state funds

$10.5M

HIE interoperability cost with top 10 EHR vendors Annual interoperability subscription fees Interoperability/ R&D simulated years

$12M

Annual fees

$4M3

HIE operational cost escalator

3%1

$600K1 5 Years2

1: Starts after the fifth year. 2: Five years simulated starting after the fifth year as the state funds deplete and pressure increases to innovate new HIE services 3: Annual fees paid to HIE by acute-care hospitals and payers as HIE; there are no charges to other providers. Total $4M achieved by 100% adoption of payers and acute-hospitals of this amount is constant over the entire time span.

2.4.4 HIE Sustainability We simulated the HIE sustainability by considering the values shown in Table 2. These parameters were based on the CRISP evaluation reports [26, 28] and published HITECH program funding [1]. If access to CRISP balance sheets and annual financial reports had been available, we would have used precise figures as inputs to the simulation model.

After a full financial sustainability investigation of the CRISP report, it was evident that the state funds received for HIE implementation and establishment were depleted in 2015 as the grant period came to an end [28]. The CRISP RFA response identified the financial model path to sustainability, with startup state funding allocated to HIE infrastructure hardware and software, platforms, license fees, human wages, and case services.

However, the additional R&D cost required to ensure that the HIE system was interoperable with the top-ten EHR platforms used in Maryland has not yet been considered. Interoperability between HIE and EHR is required to improve the use of HIE. For example, the health care provider does not need to add more steps to their existing workflow by accessing patient health records through HIE. Instead, the EHR platform will directly connect to the HIE system.

2.4.5 Policy Interventions As discussed earlier, we designed two policies to accelerate HIE system adoption. The policy interventions are individually exercised, and their combination serves to identify their joint effect on HIE adoption. The base run was set on the parameters and assumptions shown in Tables 1 and 2.

We ran different permutations of policy interventions concerning the effective policy date. The realistic assumption of the starting date was based on current realistic data, unlike in previous studies [33, 40], in which the policy starting date was assumed to be twenty to fifty years in the past. All technological aspects of life were considered as vibrant and agile, which forces the HIE system to quickly adapt to new changes. January 2016 was selected to imitate more practical real-world scenarios of a possible policy implementation. The details of the policy intervention equations and variables are listed in Appendix A (Table 4).

Financial Incentives (Policy 1): A feedback loop from the HIE benefits and value-added variable is used to influence the financial incentives policy, which will dynamically increase or decrease alongside the HIE benefits and value-added. This policy will affect the “Adopted HIE Multiplier” variable, which represents the internal Bass-diffusion influencer coefficient (Ii), as shown in Eq. (1). The effect of Policy 1 will drive the “Adopted HIE Multiplier,” which will subsequently change the effective adoption rate.

The following qualitative variables are affected: HIE benefits, HIE demands, HIE sustainability, adoption internal influence, and time. The concept of the modeled policy states that the policy will influence the system if and only if there are perceived financial savings from the HIE. The financial incentive should be coupled with better care management and higher health care performance.

For example, provider compensation should be based on pay-for-performance. Prior to the effective start date, all policies were thus normalized to eliminate the effect of the policy leveler; an initial value of the policy

leveler of 10% was used, which then gradually increased on account of the feedback loop from financial savings of the HIE system.

Awareness and education for HIE integration in the workflow (Policy 2): The non-linear dynamics simulation of this policy intervention is based on the following qualitative variables that are interrelated and affect the HIE system adoption: HIE use and integration in workflow, HIE demands, HIE interoperability and standards, adoption external influence, and time.

Similar to the policy multiplier discussed in Policy 1, this policy leveler directly affects the “Workflow Multiplier,” which represents both endogenous and exogenous Bass-diffusion influencer coefficients (Ii and Ei), as shown in Eq. (1). The effect of Policy 2 will drive the “Workflow Multiplier,” which will subsequently change the effective adoption rate. The policy leveler value was set to 25%.

3. Results and Discussion We qualitatively and quantitatively reviewed the parameters to develop informed assumptions, which enabled us to test, validate, and build significant confidence in the model. We now address and discuss results of the problem statements in quantifiable figures and charts, and we show the effect of policy interventions on ambulatory provider HIE adoption.

3.1. HIE ROI The cumulative HIE system ROI curve shows the aggregate of the cumulative HIE system benefits and valueadded, as well as the cumulative HIE system cost month over month. As shown in Fig. 6, the Maryland HIE shows significant financial gains and ROI from using HIE based on the parameters and values outlined in Table 1. The increasing trend of the cumulative HIE ROI (Fig. 6) shows a significant estimated gain of $3.3 billion over ten years. The HIE ROI shows a slow start in the early months (before March 2015), and then a rapid increase over the following years. The quick increase in ROI relates to the increase in HIE adoption; higher HIE usage and more readily available patient data can be reused to improve health care system quality and

efficiencyy, i.e., hospitaal readmission n reduction, avvoidable repeaated ER visits, duplicative llab and radiolo ogy test reductionn, and health record r lookup--time savings..

  Fig. 6. HIE system cum mulative finan ncial ROI for Maryland. M E Usage 3.2. HIE Fig. 7 illuustrates the work performed d for the colleected health caare providers (acute-care ( hoospitals and am mbulatory providerss) that adoptedd the HIE systtem and begann querying pattient informattion. Ambulatoory providers show an increasinng trend with a steeper slopee after Septem mber 2012. Before the last quarter q of 20111, the numberr of acutehospitals adopting the HIE system iss higher than tthat of ambulatory providerrs. From Apriil 2014 to Marrch 2015, the end oof the data colllection period d, acute-hospittals show a co onstant flat rate of HIE adopption compareed to ambulatoory providers. This is becau use all 47 acutee-hospitals in Maryland hav ve adopted HIIE. It is expected that the trend of ambulatoryy providers will w show an inncreasing slope in coming years. y This is bbecause the gaap between the existing ambulatory pro ovider adopterrs and potential adopters is considerable.. Thus, many potential p adopters remain beforee the gap is fillled and the trrend slope leveels.

Fig. 8 shoows the trend of HIE usagee for the givenn health care provider p settings. This trendd is not eviden nt before Decembeer 2013. The curves c resonatte below 4,0000 portal queriees per month, as expected fr from a slow-sttart Sshaped addoption of thee new product and lack of aw awareness and d benefits of ussing the HIE ssystem. The su um of the HIE systeem usage rapiidly escalates above 10,0000 queries in Feebruary 2015 for f acute-hosppitals and overr 15,000

queries fo for ambulatoryy providers. Th his figure is pprojected to increase even more m in future years. The deemand and usage in ambulatory seettings are significantly highher than in oth her acute-hosp pitals. Therefoore, we suggest that researcheers conduct more m empirical studies on HIIE benefits and efficiency in n ambulatory provider settings.

  Fig g. 7. HIE adopption by healtth care provideers.

d of HIE ussage (portal quueries = HIE portal p usage by b health care providers and d acuteFig. 8.. Real-world data hospitals; non-ED = hospital settiings, excludin ng ED).

  3.3. HIE E Effect on Hospital H Read dmission Red duction As shownn in Fig. 9, thee month-over--month cumullative readmisssion reduction and increasee at the end off the dataset inn March 2015 are 1,744 reaadmission reduuctions compaared to 220 red ductions in M March 2012. Tw wo

correlatioon coefficientss were calculaated using com mmercially av vailable softwaare (Microsoft ft Excel 2016). A correlatioon coefficient equal to 0.75 is noted betw ween monthly acute-hospital HIE usage aand cumulativee monthly readmission reduction. This positivee cyclic correllation curve has an ascendin ng trend; the m more hospitalss there are that uuse HIE, the sttronger is the relationship inn the readmission reduction n. Related to th the above exam mple, the number oof acute-hospiitals recorded in March 201 3 utilizing thee HIE system is 10, which ggenerate 212 queries, q versus 477 hospitals in March M 2015 generating g 13,,400 queries.

A correlaation coefficieent of 0.8 was calculated forr the monthly ENS alerts geenerated and ccumulative mo onthly acute-hosspital readmisssion reduction n, as shown inn Fig. 10. To the t best of ourr knowledge, M Maryland hass recently passed leegislation for hospital h reimb bursements baased on performance instead d of service. A noteworthy period in Fig. 10 iss July 2014 to March 2015; this period iss characterized d by the ENS alerts followinng the same fluctuating fl trend of tthe readmissioon reduction (e.g., October 2014 shows an a increasing readmission r reeduction, which is the same behhavior for ENS S alerts; the behavioral trennds of the two curves are sim milar to a greaat extent). Thee readmission reduction dip in January y 2015 is mosst likely relateed to seasonal effects, such aas occurrencees of the flu and coold weather.

Fig. 9. Accute-hospital HIE H usage corrrelation with acute-hospital readmissionn reduction.

Owing too the lack of empirical study y, we cannot ddraw a conclu usion regarding g the causalityy between the above variabless. Nevertheless, the above findings fi show strong correlaations among datasets. In reecent years, accutehospital rreadmission reeductions hav ve been a focuus area of policcymakers in Maryland M and nnationwide. In n addition, other initiativves have been launched to aaddress this issue. Howeverr, based on EN NS, HIE usagee, and readmission reduction trends, we caan conclude thhat the Marylaand health caree system is mooving toward a more effective health care syystem. This is evident from m the data show wn in Figs. 9 and a 10.

Fig. 10. Correlation C of generated EN NS alerts with acute-hospital readmission reduction.

  3.4. HIE E Sustainabillity It is evideent from Fig. 11 that CRISP P HIE is not ssustainable in the long term. However, thhis research is not meant to address the soolution for thee sustainabilityy of CRISP HIE H or design new n strategiess for additionaal revenues or subscriptioon fee policiess [4, 15]. Unliike Stage 1 off the CMS EHR Incentive PPrograms, Stag ges 2 and 3 containn financial incentives for thee adoption andd meaningful use of HIE. These T financiaal incentives, which w also incoorporate the paayment of subscription fees , will encouraage more healtth care providders, including g ambulatoory providers, to actively paarticipate in H HIE. We highliight this point to compel deecision makerss to allocate aadditional funnds or generatee new revenuee streams towaard HIE sustaiinability.

 

Fig. 11. CR RISP HIE susstainability.

3.5. Fin nancial Incenttives (Policy 1) The first policy interveention simulattes the effect oof financial in ncentives on am mbulatory proovider HIE ad doption and exam mines the moddeled system’ss response. Thhe nature and response r of th his policy was performed ussing a non-lineaar dynamic sim mulation. Fig. 12 shows thee influence of Policy 1 on am mbulatory proovider HIE ad doption with resppect to the effeective policy start s date. Thee effect of Poliicy 1 on the ad doption rate iss apparently th he highest oof the two poliicies.

Ambulatoory provider HIE H adoption increased by 15.87% from the base run simulation s to tthe end of Sep ptember 2020. Cuurrently, EHR meaningful use u (as articulaated in the CM MS EHR Incen ntive Program ms) is one of th he main drivers sttimulating HIT T adoption. However, H the rresults of this model m should compel policy cymakers to co onsider innovativve financial inncentive plans to promote caare and perforrmance quality y in HIE. The applicability and repeatabiility of this poolicy are not ex xclusive to M Maryland; ratheer, they provid de a frameworrk for other HIE initiatives nationwide with w consideration of exogeenous and end dogenous dynaamics of the H HIE system an nd such policy intterventions.

  Fig. 12. A Ambulatory provider p adopttion base-run ssimulation of policy interveentions, financcial incentivess, and awarenesss and educatiion to incorporrate HIE in w workflows (pollicy effect starrting January 22016, N = 5,153)

  3.6. Aw wareness and Education fo or Integratingg HIE in Worrkflows (Policcy 2) The secoond policy inteervention was applied to sim mulate the effeect of awareneess and educat ation on using the HIE and integgrating it in daaily workflows. Fig. 12 show ws the influen nce of Policy 2 on the ambuulatory provid der HIE adoption with respect to t the effectiv ve policy startiing date. The Policy 2 effecct on ambulatoory provider HIE H adoption is evident, as the rate is inccreased by 8.336% from the base run simu ulation to the ssimulation end d in Septembeer 2020.

Based onn the results, itt is evident thaat the HIE sysstem platform m is a promising one. Moreov over, greater effforts in advancinng the system by b all stakeho olders will leadd to increasing gly better outccomes. Many innovative so olutions can be inntegrated in the HIE workflo ow to increasee adoption, su uch as inventin ng a user-frienndly system th hat provides “one click” access to patien nt health recorrds, and prom moting awareness and educaation campaign ns focusing on publicizinng HIE benefitts through worrkshops, confferences, broch hures, and offi fice visits. In addition, a state-of-tthe-art technollogy aimed at easing the poortability betw ween the HIE platform p and tthe end-user can c be highly efffective.

mbination off Policy 1 and d2 3.7. Com

In terms of policy interventions, there is no panacea policy to solve the HIE system adoption problem. The adoption of individual policies will only yield short-term temporary solutions; however, multiple policy options will likely result in even better system outcomes. For example, without a proper awareness campaign focused on a means of adopting and using HIE, potential adopters may not be interested in adopting HIE and may eschew the financial incentives. The results depicted in Fig. 12 show the benefits of policy amalgamation on ambulatory provider HIE adoption, which increased by 27.17% compared to the base run simulation.

4. Limitations The Maryland HIE system is a patient opt-out system rather than an opt-in system. Accordingly, our model does not account for the effect of patient adoption of the system. In other states, HIE is an opt-in system in which the patient is required to register to the system. In that case, the patient adoption must be modeled as a stock and flow that includes all barriers and influencing factors affecting the patient adoption flow. The total population number (N) is fixed, and the adopter population will not abandon the HIE system. This model is limited to ambulatory providers and acute-hospitals; however, it can be expanded to include the remaining health care system providers and entities, such as federally qualified health centers, long-term care facilities, pharmacies, laboratories, and radiology centers.

Additionally, competition barriers were omitted from this case study model. However, considering the pay-forservice nature of the U.S. health care system, we believe that the competition is a barrier nationwide and will subsequently slow HIE adoption and usage. The simulated ROI only addresses a set of benefits identified based on the available dataset. However, more synergies should be accounted for in using the HIE system, such as time savings for patients, and processing of redundant laboratory and radiology tests. Finally, the challenge remains to model all variables and parameters affecting the HIE system on account of data limitations.

5. Conclusions In this study, we investigated HIE adoption from the following perspectives: HIE sustainability, financial benefits, ROI, and the correlation between HIE and hospital readmission reduction. This research used a holistic

systematic analysis approach using the SD method, which can be replicated in other health care system applications to address the immense need for improved health care quality in the U.S. The proposed policy interventions discussed were financial incentives (Policy 1), awareness and education for integrating HIE in workflows (Policy 2), and a combination of Policies 1 and 2. Additional policy interventions can be designed to help decision makers achieve the most optimal HIE system performance and obtain better outcomes. Despite the progress in HIE adoption, as well as the earmarking of federal and state funds to incentivize its adoption, our simulation model highlights persistent long-term challenges discussed in previous studies [8]. A careful review of policy interventions and effective starting dates was presented based on the most recent available dataset and realistic assumptions. The rate of HIE adoption is expected to increase following the proposed policy interventions.

Acknowledgements We would like to acknowledge both the Maryland Health Care Commission (MHCC) and the Chesapeake Regional Information System for Our Patients (CRISP) for providing detailed datasets and provided insightful knowledge about the State of Maryland HIE.

Funding The author is a Ph.D. candidate and received no financial compensation to conduct this research.

References [1] HealthIt.gov, HITECH Programs & Advisory Committees. http://www.healthit.gov/policy-researchersimplementers/hitech-programs-advisory-committees, 2014 (accessed 25.07.15). [2] N. Yaraghi, An empirical analysis of the financial benefits of health information exchange in emergency departments., J. Am. Med. Inform. Assoc. (2015). [3] M.B. Buntin, M.F. Burke, M.C. Hoaglin, D. Blumenthal, The benefits of health information technology: a review of the recent literature shows predominantly positive results, Health Aff. 30 (2011) 464–471. [4] S. Sridhar, P.F. Brennan, S.J. Wright, S.M. Robinson, Optimizing financial effects of HIE: a multi-party linear programming approach., J. Am. Med. Inform. Assoc. 19 (2012) 1082–1088.

[5] M.E. Frisse, K.B. Johnson, H. Nian, C.L. Davison, C.S. Gadd, K.M. Unertl, et al., The financial impact of health information exchange on emergency department care., J. Am. Med. Inform. Assoc. 19 (2012) 328– 333. [6] J.E. Richardson, E.L. Abramson, R. Kaushal, The value of health information exchange, J. Healthc. Leadersh. 4 (2012) 17–23. [7] D.C. Kaelber, D.W. Bates, Health information exchange and patient safety, J. Biomed. Inform. 40 (2007) 40–45. [8] C.S. Kruse, V. Regier, K.T. Rheinboldt, S.C. Kruse, V. Regier, T.K. Rheinboldt, Barriers over time to full implementation of health information exchange in the United States, JMIR Med. Inf. 2 (2014) e26. [9] G.J. Kuperman, Health-information exchange: why are we doing it, and what are we doing?, J. Am. Med. Inform. Assoc. 18 (2011) 678–682. [10] J.R. Vest, T.R. Campion, R. Kaushal, Challenges, alternatives, and paths to sustainability for health information exchange efforts, J. Med. Syst. 37 (2013) 1–8. [11] J.R. Vest, L.D. Gamm, Health information exchange: persistent challenges and new strategies, J. Am. Med. Inform. Assoc. 17 (2010) 288–294. [12] V. Bostick, M. Renee, G. Crayton, E. Fishman, E. Peters, V. Smith, Sustaining state health information exchange: a state toolkit, Natl. Governors Assoc. Cent. Best Pract. State Alliance E-Health. (2011) 46, http://www.nga.org/files/live/sites/NGA/files/pdf/1103sustaininghietoolkit.pdf. [13] P. Dullabh, J. Adler-Milstein, L. Hovey and A.K. Jha, Key challenges to enabling health information exchange and how states can help, NORC U. Chicago (2014). [14] G. Morris, S. Afzal, M. Bhasker and D. Finney, Query-based exchange: key factors influencing success & failure, Offic. Natio. Coord. Health Info. Tech. (2012). [15] Weaver and Tidwell, LLP, Literature Review of Health Information Exchange Sustainability Texas Health Services Authority (2012). [16] J. Adler-Milstein, D.W. Bates, A.K. Jha, Operational health information exchanges show substantial growth, but long-term funding remains a concern, Health Aff. 32 (2013) 1486–1492. [17] M.F. Furukawa, V. Patel, D. Charles, M. Swain, F. Mostashari, Hospital electronic health information exchange grew substantially in 2008-12, Health Aff. 32 (2013) 1346–1354. [18] Office of the National Coordinator for Health Information Technology (ONC), Report to Congress: update on the adoption of health information technology and related efforts to facilitate the electronic use and exchange of health information, http://healthit.gov/, 2014. [19] Office of the National Coordinator Health Information Technology, Health information exchange among U.S. non-federal acute care hospitals: 2008-2014, http://healthit.gov/sites/default/files/databrief/ONC_DataBrief24_HIE_Final.pdf, 2014. [20] M.F. Furukawa, J. King, V. Patel, C.-J. Hsiao, J. Adler-Milstein, A.K. Jha, Despite substantial progress in EHR adoption, health information exchange and patient engagement remain low in office settings, Health Aff. 33 (2014) 1672–1679. [21] L. Lander, D.G. Klepser, G.L. Cochran, M. Morien and M.F. Fache, Health Information Exchange Evaluation, U. Neb. Med. Cen. (2014).

[22] J.A Merrill, M. Deegan, R.V Wilson, R. Kaushal, K. Fredericks, A system dynamics evaluation model: implementation of health information exchange for public health reporting., J. Am. Med. Inform. Assoc. 20 (2013) e131–e138. [23] J. Adler-Milstein, A.P. McAfee, D.W. Bates, A.K. Jha, The state of regional health information organizations: current activities and financing, Health Aff. 27 (2007) 60–69. [24] J.R. Vest, L.M. Kern, M.D. Silver, R. Kaushal, The potential for community-based health information exchange systems to reduce hospital readmissions, J. Am. Med. Inform. Assoc. 22 (2015) 435–442. [25] S.E. Ross, T.A. Radcliff, W.G. Leblanc, L.M. Dickinson, A.M. Libby, D.E. Nease, Effects of health information exchange adoption on ambulatory testing rates, J. Am. Med. Inform. Assoc. 20 (2013) 1137– 1142. [26] Maryland Health Care Commission, Maryland HIE Evaluation Results Report, Off. Nat. Coord. Health Info. Tech. State Health Info. Exch. Coop. Agr. Prog. (2014). [27] Maryland Department of Health and Mental Hygiene, Statewide health information exchange. http://mhcc.dhmh.maryland.gov/hit/hie/Pages/Statewide-Health-Information-Exchange.aspx,2014 (accessed 15. 06.15). [28] Chesapeake Regional Information System for Our Patients, The CRISP Response to the Request for Application for a Consumer-Centric Health Information Exchange for Maryland, Columbia, 2009. [29] Center for Urban Transportation, Resistance to Change, U. Wis. Milwaukee, https://www4.uwm.edu/cuts/bench/change.htm (accessed July 2015). [30] K.M. Unertl, K.B. Johnson, N.M. Lorenzi, Health information exchange technology on the front lines of healthcare: workflow factors and patterns of use, J. Am. Med. Inform. Assoc. 19 (2012) 392–400. [31] A. Burbano Collazos, Modeling the adoption of identification standards within the healthcare supply chain, U. Ark. (2012). [32] N. Erdil, C.R. Emerson, Simulation modeling of electronic health records adoption in the U.S. healthcare system, Proc. 27th Int. Conf. Syst. Dyn. Soc. (2009). [33] P. Otto, D. Nevo, Electronic health records: A simulation model to measure the adoption rate from policy interventions, J. Enterp. Inf. Manag. 26 (2013) 165–182. [34] J.D. Sterman, System dynamics modeling: tools for learning in a complex world, Calif. Manage. Rev. 43 (2001) 8–25. [35] J.D. Sterman, Business dynamics: systems thinking and modeling for a complex world, Irwin/McGrawHill, 2000. [36] Center of Medicare & Medicaid Services. http://www.cms.gov/ (accessed July 2015). [37] Health Information Technology. www.healthIT.gov (accessed May 2015). [38] Maryland Health Care Commission. http://www.mhcc.maryland.gov/ (accessed September 2015). [39] Chesapeake Regional Information System for Our Patients. https://crisphealth.org/ last (September 2015). [40] N.O. Erdil, Systems analysis of electronic health record adoption in the U.S. healthcare system, Binghamton: State U. NY Binghamton, 2009.

[41] E.A. Edaibat, J. Dever, B. Tanju and S.M.F. Stuban, A system dynamics simulation modeling: health information exchange adoption in the U.S. healthcare system, Proc. 2014 Win. Sim. Conf. IEEE (2014). [42] D.C. Lane, J.D. Sterman, Profiles in operations research: Jay Wright Forrester, Profiles Oper. Res. Pioneers Innov. (2011) 363–386. http://jsterman.scripts.mit.edu/docs/Lane-2011, Profiles in Operations Research.pdf. [43] J.W. Forrester, Industrial Dynamics, 1961. [44] State of Connecticut, Connecticut Statewide Health Information Exchange (HIE) Financial Sustainability Study Final Report: Recommended Funding Methods and Formulas for HIE Financial Sustainability, Gartner, Inc, 2010. [45] HEALTHeLINK, Study of duplicate CT scans shows promise of health information exchange, wnyhealthelink.com, Western New York, 2014. [46] Iowa Department of Public Health, Return on Investment Potential EHRs and Iowa HIE, Office of Health Information Technology, Des Moines, 2015.