Emergency Department Information System Implementation and Process Redesign Result in Rapid and Sustained Financial Enhancement at a Large Academic Center

Authors

  • Jason S. Shapiro MD, MA,

    1. From the Department of Biomedical Informatics, Columbia University (JSS), New York, NY; and the Department of Emergency Medicine (KMB, CH, LDR), the Department of Community Medicine (JG), and the Departments of General Medicine (FY) and Emergency Medicine (JSS), Mount Sinai School of Medicine, New York, NY. Dr. Shapiro is currently with the Department of Emergency Medicine, Mount Sinai School of Medicine, New York, NY. Dr. Chawla is currently with the Department of Emergency Medicine, Inova Fairfax Hospital, Falls Church, VA.
    Search for more papers by this author
  • Kevin M. Baumlin MD,

    1. From the Department of Biomedical Informatics, Columbia University (JSS), New York, NY; and the Department of Emergency Medicine (KMB, CH, LDR), the Department of Community Medicine (JG), and the Departments of General Medicine (FY) and Emergency Medicine (JSS), Mount Sinai School of Medicine, New York, NY. Dr. Shapiro is currently with the Department of Emergency Medicine, Mount Sinai School of Medicine, New York, NY. Dr. Chawla is currently with the Department of Emergency Medicine, Inova Fairfax Hospital, Falls Church, VA.
    Search for more papers by this author
  • Neal Chawla MD,

    1. From the Department of Biomedical Informatics, Columbia University (JSS), New York, NY; and the Department of Emergency Medicine (KMB, CH, LDR), the Department of Community Medicine (JG), and the Departments of General Medicine (FY) and Emergency Medicine (JSS), Mount Sinai School of Medicine, New York, NY. Dr. Shapiro is currently with the Department of Emergency Medicine, Mount Sinai School of Medicine, New York, NY. Dr. Chawla is currently with the Department of Emergency Medicine, Inova Fairfax Hospital, Falls Church, VA.
    Search for more papers by this author
  • Nicholas Genes MD,

    1. From the Department of Biomedical Informatics, Columbia University (JSS), New York, NY; and the Department of Emergency Medicine (KMB, CH, LDR), the Department of Community Medicine (JG), and the Departments of General Medicine (FY) and Emergency Medicine (JSS), Mount Sinai School of Medicine, New York, NY. Dr. Shapiro is currently with the Department of Emergency Medicine, Mount Sinai School of Medicine, New York, NY. Dr. Chawla is currently with the Department of Emergency Medicine, Inova Fairfax Hospital, Falls Church, VA.
    Search for more papers by this author
  • James Godbold PhD,

    1. From the Department of Biomedical Informatics, Columbia University (JSS), New York, NY; and the Department of Emergency Medicine (KMB, CH, LDR), the Department of Community Medicine (JG), and the Departments of General Medicine (FY) and Emergency Medicine (JSS), Mount Sinai School of Medicine, New York, NY. Dr. Shapiro is currently with the Department of Emergency Medicine, Mount Sinai School of Medicine, New York, NY. Dr. Chawla is currently with the Department of Emergency Medicine, Inova Fairfax Hospital, Falls Church, VA.
    Search for more papers by this author
  • Fen Ye MS,

    1. From the Department of Biomedical Informatics, Columbia University (JSS), New York, NY; and the Department of Emergency Medicine (KMB, CH, LDR), the Department of Community Medicine (JG), and the Departments of General Medicine (FY) and Emergency Medicine (JSS), Mount Sinai School of Medicine, New York, NY. Dr. Shapiro is currently with the Department of Emergency Medicine, Mount Sinai School of Medicine, New York, NY. Dr. Chawla is currently with the Department of Emergency Medicine, Inova Fairfax Hospital, Falls Church, VA.
    Search for more papers by this author
  • Lynne D. Richardson MD

    1. From the Department of Biomedical Informatics, Columbia University (JSS), New York, NY; and the Department of Emergency Medicine (KMB, CH, LDR), the Department of Community Medicine (JG), and the Departments of General Medicine (FY) and Emergency Medicine (JSS), Mount Sinai School of Medicine, New York, NY. Dr. Shapiro is currently with the Department of Emergency Medicine, Mount Sinai School of Medicine, New York, NY. Dr. Chawla is currently with the Department of Emergency Medicine, Inova Fairfax Hospital, Falls Church, VA.
    Search for more papers by this author

  • Presented at the Society for Academic Emergency Medicine annual meeting, San Francisco, CA, May 18–21, 2006.

  • Dr. Shapiro is supported in part by NLM 1K99LM009556-01/02 and 4R00LM009556-03, which has run from July 16, 2007, to the present. There is/was no direct support for work on this project.

  • In 2006 Dr. Baumlin became President and CEO of SunriseSolutions Group (SSG) Inc. During late 2007 and 2008, PICIS contracted SSG to perform services related to assessing processes and performing application training for physicians and other staff members. PICIS in no way sponsored nor offered financial or other support for this research and the consulting engagement ended by late 2008.

Address for correspondence and reprints: Jason S. Shapiro, MD, MA; e-mail: jason.shapiro@mssm.edu.

Abstract

Objectives:  The objectives were to measure the financial impact of implementing a fully integrated emergency department information system (EDIS) and determine the length of time to “break even” on the initial investment.

Methods:  A before-and-after study design was performed using a framework of analysis consisting of four 15-month phases: 1) preimplementation, 2) peri-implementation, 3) postimplementation, and 4) sustained effects. Registration and financial data were reviewed. Costs and rates of professional and facility charges and receipts were calculated for the phases in question and compared against monthly averages for covariates such as volume, collections rates, acuity, age, admission rate, and insurance status with an autoregressive time series analysis using a segmented model. The break-even point was calculated by measuring cumulative monthly receipts for the last three study phases in excess of the average monthly receipts from the preimplementation phase, corrected for change in volume, and then plotting this against cumulative overall cost.

Results:  Time to break even on the initial EDIS investment was less than 8 months. Total revenue enhancement at the end of the 5-year study period was $16,138,953 with an increase of 69.40% in charges and 70.06% in receipts. This corresponds to an increase in receipts per patient from $50 to $90 for professional services and $131 to $183 for facilities charges. Other than volume, there were no significant changes in trends for covariates between the preimplementation and sustained-effects periods.

Conclusions:  A comprehensive EDIS implementation with process redesign resulted in sustained increases in professional and facility revenues and a rapid initial break-even point.

ACADEMIC EMERGENCY MEDICINE 2010; 17:527–535 © 2010 by the Society for Academic Emergency Medicine

Emergency department (ED) adoption of comprehensive information systems has been slow1,2 despite perceived advantages in the efficiency and safety for health information technology (IT) in other areas of the hospital.3–5 Unclear costs and concerns about workflow disruption during the transition are reasons frequently ascribed to this reluctance to adopt new technology.3,6

The adoption of IT solutions may be of particular benefit to EDs due to the pressures of increasing volumes, crowding, and patient acuity and the difficulty of obtaining complete histories from patients at presentation. ED tracking systems alone have been shown to improve patient throughput times,7 and comprehensive ED information systems (EDISs) have been shown to improve retrieval of medical and radiology records, laboratory turnaround times, and patient throughput times.8,9 EDISs also represent an opportunity to improve patient safety10,11 and clinical documentation12,13 and, through the use of decision support, to improve compliance with standard medical guidelines.5,14 EDIS implementation at Mount Sinai Medical Center has led to markedly decreased turnaround times for test results and improved ED throughput times8,9 and has facilitated research initiatives, quality assurance activities, and teaching. To the best of our knowledge, no comprehensive financial analysis of an EDIS implementation has previously been described. Authors and working groups have called for more data regarding the financial impact of implementing an EDIS.12

This article describes the implementation of an EDIS in a large academic ED and the process redesign that was undertaken to optimize the implementation. The study presents a financial analysis calculating the time it took to break even on the initial investment for the EDIS implementation and the total revenue enhancement at the end of a 5-year study period in a large academic ED.

Methods

Study Design

This was a pre- and postimplementation study conducted using monthly averages of aggregate financial data to perform a time series analysis.15 A time series analysis examines the change in a measure over time and, through calculation of a slope, whether there is a change in the rate of change of that measure at the time that an intervention takes place. This investigation was deemed exempt from human subjects review by the Mount Sinai Institutional Review Board.

Study Setting and Population

The Mount Sinai Medical Center (New York, NY) is an 1,171-bed tertiary care academic medical center located in Manhattan, bordering on the Upper East Side (an affluent New York neighborhood) and East Harlem (a working class and underserved area). The facility is known for cardiac care and transplant services, has a world-renowned geriatrics program, and is a designated heart hospital and primary stroke center.

Mount Sinai has 45 licensed ED beds, which increased from 41 in 2005, and includes separate pediatric, fast-track, and observation areas with an annual combined volume of 88,140 visits in 2007. There is a 4-year emergency medicine residency program that includes 60 residents, and there are house staff rotating through the ED from other departments, including internal medicine, psychiatry, and obstetrics and gynecology. There are also fellowship tracks in pediatrics, emergency ultrasound, clinical research, and informatics. During the study period, the total number of full-time-equivalent (FTE) hours decreased slightly for attending and resident coverage, while the hours of nursing, technician, and clerical staff support remained unchanged. The patient population has changed little over the years, with a payer mix that includes 25% commercial (with 12% health maintenance organization/preferred provider option), 30% Medicaid, 25% Medicare, and 20% self-pay or uninsured. The admission rate during the study period ranged from 23.5% to 25.9%, and the mean (±SD) patient age ranged from 30.28 (±1.38) to 32.3 (±1.10) years (Table 1). Our prenegotiated rates with payers did not vary significantly over the 5-year study period, with the exception of New York State Medicaid, which increased professional reimbursement from $17 to $25 in 2006 and underwent a transition to Medicaid managed care programs for a portion of those who had been on traditional Medicaid.

Table 1. 
Revenues and ED Trends (Covariates)
 Preimplementation 07/02–09/03Peri-implementation 10/03–12/04Postimplementation 01/05–03/06Sustained Effect 04/06–06/07
  1. *Professional refers to cases assessed a professional or physician’s fee and generally applies to all patients including those admitted and those treated and released from the ED.

  2. †E&M = evaluation and management: the five-point scale used for coding ED charts for billing purposes.

  3. ‡Facility refers to cases assessed a facility charge by the hospital and generally refers to only those treated and released from the ED.

  4. §ESI = Emergency Severity Index; data not collected until November 2003.

Revenues
 Average professional* E&M† level3.173.433.663.73
 Total professional critical care cases14809341,614
 Total professional charges$21,104,013$28,496,273$33,853,661$41,610,935
 Total professional receipts$4,445,221$5,006,412$7,171,488$8,545,243
 Professional fee collection rate21.06%17.57%21.18%20.54%
 Average facility‡ E&M level2.852.472.142.24
 Total facility critical care cases451931,046974
 Total facility charges$46,590,732$52,061,306$65,705,669$73,063,284
 Total facility receipts$10,776,247$12,164,834$15,540,093$17,339,872
 Facility fee collection rate23.13%23.37%23.65%23.73%
 Total charges (professional + facility)$67,694,745$80,557,579$99,559,330$114,674,219
 Total receipts (professional + facility)$15,221,468$17,171,246$22,711,581$25,885,115
 Average collections per patient (professional)$54$62$82$90
 Average collections per patient (facility)$131$151$177$183
 Average collections per patient (professional + facility)$186$214$259$274
 Total excess receipts compared to preimplementation (unadjusted)N/A$1,949,778$7,490,114$10,663,647
 Total excess receipts compared to preimplementation (adjusted for volume) $1,860,730$7,011,804$9,202,975
Trends
 Total volume of billed visits81,97680,35587,74694,527
 Total collection rate22.56%21.33%22.90%22.56%
 Average acuity (as ESI score)§N/A§ 3.30 3.22 3.17
 Average age30.1830.6631.8432.34
 Average admission rate25.47%25.90%24.19%23.53%
 Average rate of uninsured patients20.47%18.25%18.17%18.98%

Study Protocol

Theoretical Model.  Evaluations of health IT implementations are often conducted using some variation of pre- and postimplementation quasi-experimental design.15 Because this implementation had a prolonged roll-out with multiple phases that took place over approximately 15 months, a framework of analysis was adopted using the following four 15-month time frames that extended prior to and following the implementation: 1) preimplementation July 2002–September 2003 when the vast majority of ED processes were still on paper; 2) peri-implementation October 2003–December 2004 while the phases of implementation were taking place; 3) postimplementation January 2005–March 2006, which was a period of adjustment immediately following the completion of EDIS implementation; and 4) sustained effects April 2006–June 2007, which was included to examine the lasting financial impact of EDIS implementation.

Interventions.  Beginning in November 2003, The Mount Sinai Hospital implemented a comprehensive EDIS (Picis, ED Pulsecheck, Wakefield, MA, formerly IBEX). Prior to this, patient care in the ED involved numerous processes supported by disparate systems, requiring multiple logins and passwords by users. The EDIS provides triage, patient tracking, physician and nurse documentation, retrieval of charts from prior ED encounters, one-click access to more extensive historical hospital data from an enterprise data repository, computerized provider order entry, results review, discharge instructions, and prescription writing. Physician and nurse documentation within the system are template-driven. History of present illness, procedure, nursing assessment, results interpretation, and follow-up care templates are chosen by the user, which in turn drive the workflow within the documentation system. All data entered into the EDIS are time-stamped. There was a phased EDIS implementation beginning with nursing documentation in the fourth quarter of 2003 (Q4 2003), followed by physician documentation in Q1 2004. Computerized order entry and integration with other hospital systems (including registration, laboratory, and hospital electronic data repository) proceeded through Q4 2004. Patient care documents from other departments or facilities are scanned into the patient’s electronic chart and are simultaneously viewable by all personnel caring for the patient. To create a single sign-on application for users, and to maximize the EDIS’s efficiency, multiple electronic interfaces were built to other hospital systems such as admission, discharge, and tracking; radiology; pharmacy; laboratory; bed management; and the hospital’s data repository (Table 2). These interfaces and single sign-on capability obviated the need for personnel to train on multiple systems and radically reduced the number of steps required to perform common functions.

Table 2. 
Interfaces Built as Part of the EDIS Implementation
InterfaceDescription
  1. EDIS = emergency department information system.

Admission discharge and trackingInbound
Physician master contact databaseInbound
Laboratory systemOutbound order entry; inbound results
Radiology systemOutbound order entry; inbound status update for orders and report (preliminary and final)
Hospital data repositoryContext-specific pass-through (historical data); real-time display of ED report for non-EDIS users
Hospital bed management systemOutbound request; inbound bed location assignment with status updates
Professional billing companyOutbound report (flat file)
Facility billing systemOutbound interface
Department of healthOutbound surveillance data

Process redesign, which involved detailed process mapping and changes to clinical and administrative workflows to maximize efficiency and synergy with the EDIS, took place throughout the peri-implementation phase and included 1) the initiation of a “quick-reg” process that allows patients to be entered into the EDIS and given a visit and medical record number in a much more timely fashion than had been done previously, thereby allowing the initiation of orders much sooner; 2) changes in the way that laboratory orders and results are processed, which previously required paper orders and several steps of human transcription with various electronic systems and now is done by bar-coded labels generated by the EDIS, time-stamped orders sent directly to the laboratory system, and results returned directly back into the EDIS via a bidirectional interface; 3) changes in the way that radiology orders are processed, which previously required faxing a hand-written request, and now an electronic order is routed to printers in the various radiology areas depending on the type of test ordered, and transcribed results are routed directly back into the EDIS via an interface; and 4) changes in the way that facility charges are captured for nursing procedures, which previously required the nurse to fill out and submit a paper charge slip for each procedure separately from their documentation workflow in the medical record and now are automatically captured by the EDIS when the nurse documents the procedure within it through a “charge-by-documentation” feature. Flow charts describing the before and after states in the Mount Sinai ED are depicted in Data Supplements S1 and S2 (available as supporting information in the online version of this paper).

Methods of Measurement.  Cost data were provided by the Mount Sinai Department of Information Technology. Financial revenues data were provided by the ED’s professional services billing agency (McKesson, Grand Rapids, MI; formerly Per-Sé Technologies, Alpharetta, MI) and The Mount Sinai Hospital finance department.

Data Collection and Processing (Cost).  Cost data were divided into two categories—capital and operating (Table 3). Capital costs include all items incurred as a one-time charge, such as contracting, training, certain licensing fees, and hardware. Operating costs include items such as salaries plus fringe for staff supporting the EDIS, ongoing software licensing fees, and ongoing hardware fees for server hosting and data storage. Because cost data were available as monthly fees or one-time charges during a particular calendar year, all costs were added together for each year, divided by 12 to calculate a monthly average, and then the corresponding monthly amounts were added for each of the 15-month periods of the four study phases. This number, the period total, was then divided by 15 and used to determine the straight-line depreciation for costs during each of the four study phases.

Table 3. 
Total Capital and Operating Costs Over the Five-year Study Period
  1. FTE = full-time equivalent; IT = information technology.

Capital
 One-time implementation expenses
  Training$86,869
  Travel during vendor selection process$36,242
 Other capital expenses
  Contract execution payment$144,786
  Sequel server license $11,741
  Workstations$76,453
  Plasma displays$52,082
  Plasma wall mounts$849
  Document scanners$4,005
  Contingency (other hardware/expenses)$14,889
  IT salary for support of  preimplementation vendor  selection process$31,500
 Total capital costs$336,305
Operating
 IT salary (1.25 FTEs/year to  support implementation)$409,500
 Software lease$962,970
 Server maintenance$49,500
 Data storage$86,670
 Total operating$1,508,640
Total cost (capital + operating)$1,936,556

Data Collection and Processing (Revenue).  Revenue data were provided as both charges and receipts in two categories—professional and facility. Professional revenue derives from patients both discharged home and those admitted to the hospital from the ED, because both receive a professional bill for physicians’ services. Facility revenue derives from patients treated and discharged from the ED, because admitted patients do not incur a separate facility charge from the ED. Total charges and receipts were calculated by adding the values for professional and facility together for a corresponding month, and then sums were tallied for each study phase. Since December 2003, professional billing and coding services have been provided by McKesson. They, along with the ED administration, provide ongoing biannual training and feedback to all faculty members on billing and coding issues. This training began in the peri-implementation phase and includes chart reviews with specific deficiencies and potential remedies and individualized remediation for some practitioners who lag behind their peers on specific measures. Similar training had been conducted internally prior to the EDIS implementation. Facility coding was outsourced in 2004, with the same vendor in place since.

Outcomes Measures

The primary outcomes measures were 1) time to an initial “break-even point” and 2) total revenue enhancement at the end of the sustained-effects period. Secondary outcome measures included changes in chart completion rates, the evaluation and management (E&M) levels (the five-point scale used for coding ED charts for billing purposes), the number of charts billed for critical care, professional charges, professional collections, facility charges, and facility collections.

Data Analysis

Data were analyzed by calculating simple averages and percentages from aggregate monthly data using Microsoft Excel (Microsoft Corp., Redmond, WA). The break-even point was calculated by determining the difference between the monthly receipts for each month in the later three phases of the study and the average monthly receipts from the preimplementation phase. Excess cumulative monthly receipts were then plotted against cumulative monthly EDIS costs for each of the later three phases in the study. To determine total revenue enhancement at the end of the study period, the total cumulative cost of the implementation was subtracted from the amount of the total cumulative receipts from the later three phases of the study that exceeded the cumulative average monthly receipts from the preimplementation period.

Statistical Methods.  Time series data for the months encompassing the preintervention, peri-intervention, postintervention, and sustained-effects periods were analyzed using autoregressive integrated moving average (ARIMA) models as implemented in SAS software (SAS Institute, Cary, NC).16 The first stage in the analysis was to see if each observed series was a stationary series and, if not, to detrend the observed series to obtain a stationary series. In all instances detrending was required, and simple differences (between response at time “t” and response at time “t – 1”) were calculated for each variable to obtain a stationary series. The four response series of interest were modeled as functions of the intervention periods, yielding adjustments to the intercept for each period. In these models other covariates, including average age, average collection rate, average rate of uninsured patients, and total volume of billed charts were forced into the model regardless of the p-values. After fitting the models to the data, tests were done to examine how well the models fit the data. This was done using the Dickey-Fuller unit root test,17 which examines the residuals to see if they are a stationary series—the first underlying assumption of the model. The p-value for the Dickey-Fuller test was p > 0.05, indicating that the residuals were stationary. The Ljung-Box portmanteau test18 was carried out to see if the residuals were uncorrelated, i.e., if they represented white noise. The p-value of >0.05 indicated that the data were consistent with the white noise assumption required by the model.

Results

Costs

Up-front capital costs, involving one-time charges such as training and hardware, totaled $427,916, with over 90% of this cost incurred in the peri-implementation period (Table 3). All initial training was rolled into the capital budget and kept at a minimum by training a set of “super users” and then ensuring that there were one or two of them in the ED during every shift. Two-hour training classes for faculty, residents, and staff were included. Operating costs, covering items such as salaries, ongoing software licensing fees, and hosting and storage fees, totaled $1,508,640 across the latter three 15-month periods. Although EDIS software is often purchased outright, in this implementation the software was financed, and therefore the associate expenses were categorized as operating costs, causing operating costs to be relatively higher compared to capital costs than what might be expected.

Revenue

Gross charges and receipts rose significantly beginning in the peri-implementation phase. The discrepancy between charges and receipts allows the calculation of collection rates and average collections per patient, a convenient indicator of ED financial health18 (Table 1). These rates are multifactorial and are affected by the percentage of uninsured patients and negotiated contracts with and collections from a number of large commercial payers. There was a transient 16.59% drop in the professional collection rate during peri-implementation, but both professional and facility collection rates otherwise remained stable through all four study periods. Average collections per patient rose 47.48% (from $186 to $274) between preimplementation and sustained effects, total charges rose 69.40% during the same period (from $67,694,745 to $114,674,219), and total receipts rose 70.06% (from $15,221,468 to $25,885,115).

Costs Versus Revenue

By calculating the amount by which monthly receipts in the later three study phases exceeded the preimplementation monthly receipt average, correcting for actual monthly volume and comparing this to the monthly total cumulative costs (capital + operating), the break-even point for the initial EDIS investment was determined to have taken place in August 2004. This was within 8 months of the beginning of the peri-implementation phase and only 5 months after physician documentation went live (Figure 1). After implementation was completed, costs remained relatively fixed while revenues continued to rise. Comparing the increased amount of total excess receipts at the end of the sustained-effects phase with overall EDIS costs through all four phases, the total revenue enhancement at the end of the study period was $16,138,953.

Figure 1.

 A comparison of cumulative monthly receipts and costs from the latter three phases of the study to the average monthly receipts and costs from the preimplementation period. The circles represent the cumulative amount by which receipts from the corresponding month exceeded the average monthly receipts from the preimplantation period. The triangles represent the same thing after adjustment for changes in ED volume. The diamonds represent the average total cumulative monthly cost, which began with the up-front capital costs and was then calculated using straight-line depreciation for the cumulative operational costs from each of the four study phases.

Our secondary outcome, physician documentation, improved with the implementation of the EDIS. End-of-month chart completion rates by attending physicians rose from 65% in 2003 to 95% in 2005, while lost or illegible charts decreased from 4,992 in 2003 to zero in 2005. The average professional E&M level rose from 3.17 during the preimplementation period to 3.73 during the sustained-effects period. The number of charts meeting professional fee criteria for critical care billing increased from just one chart during the entire 15-month preimplementation period to 1,614 charts in the sustained-effects period and from 45 to 974, respectively, for facility critical care billing. Although the mean E&M level for facility billing actually decreased from 2.85 during preimplementation to 2.14 during postimplementation, it was trending back up with an average of 2.24 during the sustained-effects period. Despite this overall decline in the facility E&M level, net facility receipts increased 60.91%, from $10,776,247 to $17,339,872, between the preintervention and sustained-effects phases. Table 1 describes the professional and facility billings and collections for each phase of the study.

ED Trends and Analysis of Covariates

During the four phases of the study period, the volume of billed ED visits initially dropped 1.98%, from 81,976 during preimplementation to 80,355 during peri-implementation, but then rose 15.31%, to 94,527 by the sustained-effects period (Table 1). Total collection rates remained relatively flat, with rates over the course of the four study phases of 22.56, 21.33, 22.90, and 22.56%, respectively. Average monthly acuity, using a five-point scale (1 = most acute) employed by the ED triage nurses (Emergency Severity Index [ESI] increased from 3.30 during peri-implementation, to 3.22 during postimplementation, to 3.17 during sustained effects). The ESI was not used in the ED prior to the peri-implementation phase; therefore, the outcomes could not be adjusted for ESI use. There were no significant changes in trends for the average monthly age of patients, admission rates, or rate of uninsured patients over the study period.

Time Series Analysis

A change in the slope between study phases was calculated for all measures, which corresponds to the amount and direction of change in the measured trend for that variable when comparing one phase of the study to the other, while adjusting for ED volume, collection rates, patient age, admission rates, and percentage of uninsured patients (Table 4). Significant results from this analysis include the change in slope between the pre- and peri-implementation phases for professional charges and receipts of 7.06 × 105 (p < 0.0001) and 1.21 × 105 (p = 0.0216), respectively, and the change in slope between the preimplementation phase and sustained-effects phases for professional charges and receipts of 6.76 × 105 (p = 0.0026) and 1.99 × 105 (p = 0.0144). Additionally, there was a significant change in the slope for facility charges between the peri- and postimplementation phases of 3.72 × 105 (p = 0.0447). Figure 2 represents the change in slopes for professional charges between each time period and shows that the slope in the preintervention phase is virtually flat with a sharp up-turn in the peri-intervention phase that is maintained across the two subsequent study phases, with minor changes from period to period once the upturn takes place. Although the slopes for some measures yielded a negative result as they transitioned from one study phase to another, none were significant (all results are listed in Table 4).

Table 4. 
Results of Time Series Analysis
 Δ Slope Pre/Perip-valueΔ Slope Peri/Postp-valueΔ Slope Post/Sustp-valueΔ Slope Pre/Sustp-value
  1. Slope is the change in the stated measure as a function of time and is adjusted for the covariates volume, age, rate of self-pay, and total collection rate.

  2. Explanation of notation: −1.06 × 105 = 0.0000106; 1.21 × 105 = 121,000.

  3. Pre = preimplementation; Peri = peri-implementation; Post = postimplementation; Sust = sustained effects.

  4. *p < 0.05.

Professional charges7.06 × 105<0.0001*−4.80 × 1030.9654−2.54 × 1040.82216.76 × 1050.0026*
Facility charges−1.06 × 1050.63953.72 × 1050.0447*−3.29 × 1040.8522.33 × 1050.5495
Professional receipts1.21 × 1050.0216*6.84 × 1040.09829.76 × 1030.81461.99 × 1050.0144*
Facility receipts5.82 × 1040.4830−3.40 × 1030.95742.43 × 1040.70747.91 × 1040.5532
Figure 2.

 The changes in slopes for professional charges between each time period. This graph shows that the slope in the preintervention phase was virtually flat with a sharp up-turn in the peri-intervention phase that was maintained across the subsequent study phases. Following the initial up-turn, changes in slope from period to period are visually imperceptible due to the scale of the graph, but are presented in Table 4.

Discussion

Although total ED volume increased by 15.31% over the study period, and acuity increased by 3.9% from the peri-implementation to sustained-effects phases, the significance of the increases in professional charges and collections, while adjusting for volume, demonstrated that these changes alone would not have resulted in the observed increases in revenue. Furthermore, an increase in total ED volume while maintaining a flat collections rate could not, on its own, have produced an increase in the collections per patient as was seen. It is the increased collections per patient that led to the actual increase in revenue, and this coincides with the implementation of the EDIS and consequent improvements in charting and billing. Improved chart completion rates may account for some of the net financial enhancement and could have been increased through process redesign alone without EDIS implementation. Other factors, such as increased number of critical care cases and increased professional E&M levels, are also major contributors to the effect seen. Figure 3 depicts the change in E&M level distribution over the 5-year study period and shows a clear “shift to the right” phenomenon, with an increase in Level 5 and critical care charges and a corresponding decrease in Levels 1–3. Because fee schedules are nonlinear (i.e., a charge for a Level 5 visit can be two to four times that for a Level 3), this shift represents a sizable increase in charges. In this case, the average professional E&M level rose from 3.17 to 3.73, corresponding with a 36% increase in charges based on the department’s fee schedule. Prior to EDIS implementation, concerted training efforts to improve revenue capture through physician documentation failed to produce this shift to the right.

Figure 3.

 The change in professional E&M level distribution over the 5-year study period. This shows a clear “shift to the right” phenomenon, with an increase in Level 5 and critical care charges and a corresponding decrease in Levels 1–3. This distribution shift accounts for the reported increase in average professional E&M level from 3.17 to 3.73. E&M = evaluation and management.

One of the reasons that charting likely improved is that the physician and nurse documentation components, which come bundled with the EDIS system, contain templates for the completion of various chart elements such as history of present illness, review of systems, physical exam, and procedure notes. There is also a score feature that shows the current E&M level on the tracking board and allows the clinicians to see which components of the chart are deficient. These EDIS features prompt clinicians to more completely document the services that they provide, which in turn leads to enhanced billing, charges, and receipts.

Although there was an overall decline in the facility E&M billing level during the study period, the net facility receipts increased 60.91%, from $10,776,247 to $17,339,872. It is possible that the decreased E&M level was due, in part, to poorer overall documentation by nurses in the EDIS as a result of increased ED volume without any increases in nurse staffing levels. Although several contracts for global fees were negotiated with large managed care organizations during the study period, and this may have helped offset some of the potential decline in receipts due to the decline in E&M level, most of the increase in receipts is likely due to the charge-by-documentation feature that automatically sends a charge to the billing system each time a billable procedure (such as placing an intravenous line or giving a medication) is documented. Prior to the implementation of the EDIS, charge capture for these routine nursing procedures on the facility side required the nurse to fill out and submit a separate paper charge slip on top of their procedure documentation, which was a significant workflow impediment and often did not occur. Although a change in the rate at which these procedures were being documented was not measured as part of this study, an increase in billable procedures combined with the increase in the number of cases billed for critical care time under the facility charges could certainly account for the increase in overall facility receipts despite the decline in E&M level.

The time series analysis allowed us to examine the revenue increases with adjustment for changes in the covariates over time using statistical methods. In general, although not all of the transitions from one study phase to the next yielded a significant p-value, there was an increase in charges and receipts between preimplementation phases, with a subsequent decrease during the peri- to postimplementation transition and a resumption of increases during the transition from postimplementation to sustained-effects phases (Figure 2 and Table 4). This initial increase is likely due to effects of the process redesign during peri-implementation, in which workflow problems were fixed through the implementation, followed by a period of adjustment and a series of enhancements to the EDIS that likely caused a transient decrease in charges and receipts. The transition to the sustained-effects period, however, shows a clear trend toward continued financial benefit due to the implementation when correction is made for the covariates.

The data presented here demonstrate that implementation of a comprehensive EDIS with process redesign resulted in rapid break even on the initial investment and a significant total revenue enhancement. Examination of potential confounders, such as acuity, collection rates, age, and rate of uninsured patient visits, suggests that the effects measured are likely due to the EDIS implementation and not other secular trends that may have affected the results.

Limitations

This analysis includes those costs directly related to the EDIS implementation. There are other costs that may have had affected the results presented in this article, but were too difficult to measure and financially quantify. These other costs included disruptions to workflow from physically installing the system; in-kind time contributed to the implementation by various members of the ED faculty, staff, and residents; and in-kind time contributed by members of the IT department in addition to the quantified 1.25 FTEs in the study, including activities such as pulling network cable, installing new network jacks, server refresh, and some of the in-house work done by the institution’s interface team. These undocumented potential costs, however, must be balanced by other unmeasured potential benefits, such as decreased cost of paper charts, decreased adverse drug events, decreased time spent on record retrieval, and improved ED throughput times.

There are some data that could have been helpful in performing a more granular analysis, such as the number of billed procedures on the facility side, which were not available in our data set. Other data could not be shared because they represent proprietary business information, such as the details of the institution’s negotiated fee schedules. Although using relative value units (RVUs) as a measure of the change in physician productivity, as was done by Kahn et al.,19 could have been used, and would have accounted for changes in the fee schedule or billable procedures, RVUs were not available in the study data set, represent a measure of productivity, and would not have allowed measurement of the direct financial outcomes presented in this study.

Initially the cost analyses included formal accounting techniques for the determination of the opportunity cost of noncapital hardware expenses and the net present value of capital hardware using a 5% discount rate. The total cost using these methods was only $77,796 (4.02%) higher than the total cost using the simple straight-line depreciation method ($2,014,352 vs. $1,936,556). Because the reporting of results using these more formal methods would have been much more cumbersome and of little added value due to the negligible difference on the overall effect size of $16,138,953, the decision was made to use the simpler, straight-line depreciation method.

There were some changes in Medicaid and Medicare reimbursement over the 5-year course of the study, including the conversion of regular Medicaid patients to Medicaid managed care beginning in 2005, but the data from these changes could not be segregated for analysis. Furthermore, these changes took place after the initial break-even point and did not affect charges, and the impact on collections per patient and receipts would have been small in light of the overall effect size.

In this single-institution study, there were certain local events, such as interruptions in billing due to changes in vendors, which may have affected some of the measured outcomes. However, since such interruptions would tend to decrease enhanced revenues, they are likely not responsible for the strongly positive study results.

Conclusions

The Mount Sinai ED information system experience echoes similar findings from previous studies of a variety of electronic systems in other clinical settings, including smaller EDs.19–26 This is the first detailed published study to show a rapid break-even point and large overall revenue enhancement for an ED information system implementation at a major urban academic ED of this size. This successful implementation included an extensive number of interfaces to other hospital systems that were built with a concurrent and complimentary process redesign to optimize ED workflow. These results should help support the decision for other large EDs to adopt comprehensive information systems.

Ancillary