• emergency service;
  • hospital;
  • insurance claim review;
  • economic analysis;
  • length of stay


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

ACADEMIC EMERGENCY MEDICINE 2010; 17:824–833 © 2010 by the Society for Academic Emergency Medicine


Objectives:  The objective was to evaluate the use of a payer-based electronic health record (P-EHR), which is a clinical summary of a patient’s medical and pharmacy claims history, in an emergency department (ED) on length of stay (LOS) and plan payments.

Methods:  A large urban ED partnered with the dominant health plan in the region and implemented P-EHR technology in September 2005 for widespread use for health plan members presenting to the ED. A retrospective observational study design was used to evaluate this previously implemented P-EHR. Health plan and electronic hospital data were used to identify 2,288 ED encounters. Encounters with P-EHR use (n = 779) were identified between September 1, 2005, and February 17, 2006; encounters from the same health plan (n = 1,509) between November 1, 2004, and March 31, 2005, were compared. Outcomes were ED LOS and plan payment for the ED encounter. Analyses evaluated the effect of using the P-EHR in the ED setting on study outcomes using multivariate regressions and the nonparametric bootstrap.

Results:  After covariate adjustment, among visits resulting in discharge (ED-only), P-EHR visits were 19 minutes shorter (95% confidence interval [CI] = 5 to 33 minutes) than non-P-EHR visits. Among visits resulting in hospitalization, the P-EHR was associated with an average 77-minute shorter ED LOS (95% CI = 28 to 126 minutes), compared to non–P-EHR visits. The P-EHR was associated with an average of $1,560 (95% CI = $43 to $2,910) lower total plan expenditures for hospitalized visits. No significant difference in total payments was observed among discharged visits.

Conclusions:  In the study ED, the P-EHR was associated with a significant reduction in ED LOS overall and was associated with lower plan payments for visits that resulted in hospitalization.

In his weekly address on January 24, 2009, President Barack Obama announced: “To lower health care costs, cut medical errors, and improve care, we’ll computerize the nation’s health records in five years, saving billions of dollars in health care costs and countless lives.”1 Twenty-four days later the American Recovery and Reinvestment Act of 2009 (ARRA) was signed into law, which includes an estimated net investment of $19 billion for health information technology (HIT).

The fragmentation of health care delivery without the parallel distribution of critical health care information lays the groundwork for an error-prone and inefficient health care delivery system. Previous advances in HIT have focused on improving quality, efficiency, and affordability in localized settings;2 however, a recent advance targets health information exchange (HIE) and involves the electronic sharing of patient information across health care settings while protecting patient privacy.3–6 Within the emergency department (ED), electronic health records (EHRs) have the potential to increase patient flow and enhance the quality, efficiency, and safety of care.7 By providing a more complete and comprehensive clinical picture of the patient, the interoperability of HIE has the potential to incrementally improve the positive effect of EHRs.4,8 Recommendations by the Committee on the Future of Emergency Care include, among other technologies, HIE systems such as regional health information organizations that link emergency physicians (EPs) with other providers, for whom “immediate access to medical records can mean the difference between lifesaving intervention and life-threatening medical errors.”7

The ED is an exceptionally high-risk, high-stress environment prone to errors and quality concerns, in part due to the discontinuity in the provision of care, which results in part from problematic information flow across the continuum of the health care system.9–15 Specific concerns in this setting include multiple providers, high acuity, distractions from noise and overcrowding, the need for rapid decision-making, and communication barriers.7 Poor information exchange results in EPs being frequently deprived of critical patient information, typically without a medical record available for patients who may be unconscious or uncommunicative on arrival.7 Patients presenting to the ED are especially susceptible to having information gaps that have resulted in decreased care quality,16 as well as inefficiencies in care such as redundant testing, care delays, and less effective treatments,17,18 all of which may serve to result in increased expenditures.

The literature on outcomes associated with HIE is sparse,7 with even fewer studies that evaluate HIE in the ED setting.2,17,19,20 Although administrative health insurance claims were developed for provider reimbursement, they can provide a view of a member’s medical and pharmacy encounter experience across the continuum of care, which may serve to inform providers in the ED. The study hypothesis was that the use of a payer-based electronic health record (P-EHR), composed of summary information based on health insurance claims from all paid medical encounters incurred across the spectrum of care, is associated with a reduction in ED length of stay (LOS) and total health plan payment for the ED visit in a health-insured population.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

Study Design

This study was a retrospective analysis of ED visits to a single institution using a pre–post design. This study was approved by the Christiana Care Health System (CCHS) Institutional Review Board.

Study Setting and Population

Christiana Care Health System is a large, not-for-profit health care provider in the Mid-Atlantic region, serving all of Delaware and portions of seven counties bordering the state in Pennsylvania, Maryland, and New Jersey. The Christiana Hospital ED is a designated Level I trauma center with 76 ED treatment bays and a census of 91,391 visits in 2005.

Study Protocol

Beginning in September 2005, CCHS partnered with a health plan in the Mid-Atlantic region and a private software and services company (MEDecision, Wayne, PA) to implement a P-EHR within the Christiana Hospital ED. The P-EHR provided a clinically meaningful electronic summarization of a member’s medical history based on health insurance claims for all paid services rendered by physicians, laboratories, pharmacies, hospitals, and other sources. The components of the P-EHR included a list of medical problems, prescription refill history, and physician and laboratory services (see Appendix). The work flow within the ED was such that, after triage and during insurance status verification, if the registration clerk recognized that the patient had insurance coverage by the P-EHR partner health plan, the P-EHR was accessed from a secure website. If the P-EHR existed, the clerk would print the summary along with the chart face sheet and place it on the patient chart. This procedure either occurred during the full registration process or most often at the patient’s bedside via portable computer. Because the procedure of determining if the P-EHR was available online served as one form of verifying insurance, the implementation was consistent with the usual ED work flow.

Data Sources.  This study used health insurance claims from the Mid-Atlantic health plan that partnered with CCHS. Supplemental data from Christiana Hospital included LOS in the ED, the Emergency Severity Index (ESI) measure used at triage, the census of the ED at the time of each ED visit, and whether or not the visit resulted in hospital admission or discharge to home. The ESI score was available for 86% of all study ED visits. ED LOS and ED census were measured using data gathered from the CCHS automated passive tracking system, which was implemented in November 2004, and data were available for 76% of the remaining visits after November 2004. All data elements obtained from the administrative claims database and CCHS were merged using date of service and member health plan identification number by approved users of protected health information. All data sets were then deidentified and analyzed in a HIPAA-compliant manner.

Study Sample.  The unit of analysis for this study was an individual ED visit. Visits were identified by the presence of facility charges (Health Care Financing Administration [HCFA] Uniform Bill-92 [UB-92] codes 450–452) and claims for ED evaluation and management (Current Procedural Terminology [CPT] codes 99281–99285). Unique patients were not restricted from contributing multiple ED visits from the analysis. Each visit was assessed for study inclusion independently.

Payer-based EHR retrievals occurring between September 1, 2005, and February 17, 2006, were identified electronically from the P-EHR provider’s secure website and made available for this research. Retrieval dates were then linked to corresponding insurance claims data to identify claims for ED encounters within 1 day of the P-EHR access. This allowed for situations such as late-night ED visits and related early morning P-EHR retrievals. Overall, 71% of the P-EHR retrievals were linked to paid claims for ED visits.

Historical comparison ED visits from the Christiana Hospital ED were identified from medical claims from the same health plan between November 1, 2004, and March 31, 2005. This date range was chosen to minimize the history threat to internal validity (compared to going back more than 1 year), bias due to seasonal variation, and potential biases associated with the residency training cycle, as Christiana Hospital is a teaching facility and medical problem solving skills have been shown to be quite different between experienced and novice decision-makers.21 Furthermore, November 2004 served as the beginning of the intake period for comparison visits because this was the earliest month that historical electronic data on arrival and departure times (for calculation of LOS) and electronic capture of ED census for all ED visits were available due to implementation of a new electronic tracking system with the ED. After implementation of the ED tracking system, there were no major technological or physical changes made within the ED and no benefit package design changes between this time period and the time period over which the P-EHR visits occurred. Comparison visits were not sampled during the same time period as the P-EHR visits (concurrent control) because it was known that not all ED visits were associated with P-EHR access and it was likely that systematic differences (i.e., selection bias) between groups existed that could not be measured from the data available (e.g., decision made at triage to not access the P-EHR or it may not have been available for download). Figure 1 illustrates the time frame used to identify P-EHR visits and comparison visits.


Figure 1.  Time frame for ED visit identification. LOS = length of stay in the emergency department; P-EHR = payer-based electronic health record.

Download figure to PowerPoint

All study ED visits were required to be associated with continuous health plan eligibility over the 6 months prior to the visit for measurement of baseline information. ED visits were excluded if there was evidence of another ED visit within the previous 30 days, to avoid inclusion of return visits. Study visits were also required to have ESI, LOS, and ED census data available. Furthermore, because patients classified as ESI 1 (most urgent) do not flow through the normal triage process, and consequently were not associated with P-EHR use, comparison visits with ESI 1 were excluded from analysis.

Study Variables.  Principle outcomes included ED LOS and total paid amount by the health plan for the ED encounter. ED LOS was calculated as the difference between the arrival time and the departure time of the patient (either transferred to a different ward or discharged from the ED). Total plan payments included total health plan paid amount for all health care services and facility claims submitted for reimbursement for the ED visit. Because ED charges and inpatient charges are combined into a single bill submitted for reimbursement for patients admitted into the hospital, plan payments specifically for ED services rendered in the ED are indistinguishable from payments for services rendered in other hospital departments during the same visit. For this reason, health plan paid amounts for only the first day (i.e., the day of the ED visit) were included for ED encounters that resulted in hospitalization. This method is consistent with previous research.17 Payment data were adjusted to constant 2006 U.S. dollars using the consumer price index for all items averaged across the Philadelphia-Wilmington-Washington, DC, region.22

Control Variables.  Plan payments and ED LOS are generally determined as the result of patient and/or provider decisions to use additional services (although other factors such as hospital bed availability may play a large role among admitted visits). Variables chosen to be included as control variables and potential confounders were based on Andersen’s model23,24 and previous ED research.25–28 Predisposing factors included age, sex, prior occurrence of an ED visit, and total plan paid amount for all claims in the prior 6 months. Enabling factors included health plan type, day of week (weekday vs. weekend), and ED census at ED admission. Need factors included the ESI triage severity score, the ED primary diagnosis (identified via ICD-9-CM code in the first diagnosis field on the ED claim), three comorbidity burden measures, arrival by ambulance, and whether or not the ED visit resulted in a hospitalization or discharge. The ESI is a five-point triage system (1 = most urgent; 5 = least urgent) designed to identify those cases most likely to require hospitalization. The ESI and similar systems have been demonstrated to be an accurate and reliable predictor of ED resource consumption29–31 and LOS.27,29,31,32 Three comorbidity burden measures were quantified over the 6-month baseline period and included the Deyo-Charlson Index (DCI),33 the count of unique drug classes, and the number of physician office visits. The combination of multiple measures of comorbidity has been demonstrated to be a better method of adjustment than any single measure when using administrative claims data.34,35 Ambulance service use was identified by the presence of HCFA UB-92 codes 540–549 on the hospital facility claim.

Data Analysis

Descriptive statistics included means and standard deviations (±SDs) for continuous variables and number and percents for categorical variables. Bivariate comparisons between P-EHR and comparison visits were made with two sample t-tests and chi-square tests for continuous and categorical variables, respectively. Unadjusted comparisons on ED LOS and plan payments were also made with the two-sample t-test. Although health care payment (cost) data are typically highly skewed, the t-test has been shown to be robust to violations of the normality assumption.36 However, to test the difference in plan payment means without the parametric assumption of normality, nonparametric bootstrap methods were used.36

To adjust for baseline differences between P-EHR visits and comparison visits, LOS was modeled using ordinary least squares (OLS) regression. Total plan payments during the ED visit were modeled using a multivariable generalized linear model with gamma distribution and log link (GLM-gamma).37,38 Regression coefficients were exponentiated for interpretation in terms of economic ratios (adjusted mean plan payment if P-EHR = 1/adjusted mean plan payment if P-EHR = 0). The modified Park’s test was used to confirm that the data followed the gamma distribution.39 The method of recycled predictions was used to obtain predicted payment differences to avoid the reintroduction of covariate imbalance that arises due to nonlinear retransformation.39 Data were initially examined to determine the extent of multiple ED visits from the same patients. Because it was desired to adjust for the potential for within-patient correlation, multivariable regression models were estimated specifying the clustered sandwich estimator (robust variance estimator), which relaxes the independence assumption within clusters.40,41

Evaluations of model fit and regression assumptions were performed, including residual plot analysis and calculation of the variance inflation factors. As previously described, the sampling time frame for comparison visits did not include exactly the same calendar months as the P-EHR visits. September and October 2004 were not included because the technology for electronically capturing ED census and LOS was not available until November 2004. As a sensitivity analysis, the final economic regression models were estimated with the addition of qualifying comparison ED visits from September and October 2004, but without ED census as a control variable. All analyses were conducted using Stata version 10.0 (StataCorp, College Station, TX).


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

The total sample size included 2,288 unique ED encounters from 2,166 patients (779 P-EHR visits from 771 patients and 1,509 comparison visits from 1,458 patients). Five percent of the patients contributed more than one ED visit.

Table 1 presents the comparison of sample characteristics. Significant differences between P-EHR and comparison groups were observed on the distribution of the health plan line of business (p < 0.001) and ED census at triage (p < 0.001). Significant differences related to need factors were observed with respect to the count of unique drug classes over the baseline period (p < 0.001), the ESI triage score (p = 0.014), and the distribution of the primary diagnosis of the ED visit (p = 0.007). Overall, P-EHR visits appeared to be associated with higher percentages of visits with more severe triage scores, fewer unique drug classes, and fewer number of prescription claims over the baseline period compared to comparison visits. In addition, no significant difference was observed on the frequency of visits resulting in hospital admission (p = 0.059).

Table 1.   Sample Characteristics
VariableP-EHR Visits (n = 779)Comparison Visits (n = 1,509)p-value*
  1. DCI = Deyo-Charlson Index; ESI = Emergency Severity Index; GU = genitourinary; IPA = independent practice association; P-EHR = payer-based electronic health record; POS = point of service; PPO = preferred provider organization.

  2. *Pearson chi-square test for categorical data and the two-sample t-test for continuous data.

  3. †Two-sample t-test with unequal variances used.

  4. ‡Measured over the 6 months prior to the ED visit date.

  5. §Symptoms category included diagnoses associated with ill-defined signs and symptoms (e.g., abdominal pain, fever of unknown origin).

 Age (yr), mean (±SD)  37.3 (±17.2)  38.7 (±19.1)0.06†
 Men, No. (%)370 (47.5)707 (46.9)0.77
 ED visit, No. (%) with ≥ 1 visit‡119 (15.3)247 (16.4)0.50
 Office visits (No.), mean (±SD)‡3.0 (±3.3)  3.3 (±3.8)0.12†
 Total health care expenditures, mean (±SD)‡$5,201 (±$19,753)$4,874 (±$14,624)0.68†
 Line of business, No. (%)
  IPA304 (39)673 (44.6)<0.001
  PPO130 (16.7)308 (20.4)
  POS253 (32.5)244 (16.2)
  Other92 (11.8)284 (18.8)
 Weekend (vs. weekday), No. (%)221 (28.4)408 (27)0.50
 ED census, mean (±SD)  69.2 (±21.6)  73.8 (±23.6)<0.001†
 DCI category, No. (%)‡
 No score589 (75.6)1,080 (71.6)0.10
 1109 (14)231 (15.3)
 235 (4.5)101 (6.7)
 ≥346 (5.9)97 (6.4)
 Unique drug class count, mean (±SD)‡2 (±3.9)4 (±4.8)<0.001†
 ESI, No. (%)
  5 (least urgent)19 (2.4)26 (1.7)0.014
  4171 (22)387 (25.6)
  3412 (52.9)826 (54.7)
  2 (more urgent)177 (22.7)270 (17.9)
ED primary diagnosis, No. (%)
  Nervous system43 (5.5)112 (7.4)0.007
  Circulatory system107 (13.7)247 (16.4)
  Respiratory system67 (8.6)183 (12.1)
  Digestive system72 (9.2)115 (7.6)
  GU system67 (8.6)91 (6.0)
  Injury/poisoning201 (25.8)371 (24.6)
  Symptoms§107 (13.7)194 (12.9)
  Other115 (14.8)196 (13)
Ambulance, No. (%)103 (13.2)213 (14.1)0.56
Hospital admission, No. (%)161 (20.7)263 (17.4)0.06

The initial multiple regression model for LOS revealed poor fit caused by combining ED visits that resulted in discharge with ED visits that resulted in hospital admission in a single model. As a result, the effect of the P-EHR use on LOS was examined for discharged and admitted visits separately. Results of the unadjusted group comparisons on LOS are presented in Table 2. Without adjustment, the mean (±SD) LOS was 274 (±204) minutes overall; 258 (±184) minutes for P-EHR visits and 283 minutes (±213) for comparison visits (p = 0.004). Among discharged (ED only) visits, P-EHR visits were 23 minutes (±7) shorter than comparison visits (p = 0.002); among admitted visits, P-EHR visits were 76 minutes (±24) shorter than comparison visits (p = 0.002). The complete results of the stratified OLS regression models are included in the Data Supplement S1 (available as supporting information in the online version of this paper). Among discharged visits, the P-EHR was associated with a shorter ED duration by 19 minutes (p = 0.008). Among ED visits resulting in hospitalization, the P-EHR was associated with a shorter ED duration by 77 minutes (p = 0.002).

Table 2.   Unadjusted Comparison of ED LOS* Overall and Stratified by Discharge Status
 P-EHR VisitsComparison VisitsDifference in Means (95% CI)p-value†
  1. LOS = length of stay in the ED; P-EHR = payer-based electronic health record.

  2. *Length of stay was calculated as the difference between the arrival time and the departure time of the patient in the ED.

  3. †p-value from two-sample t-test with unequal variances.

Overall LOSn = 779n = 1,509−25 (−42 to −8) 0.004
 Minutes, mean (±SD)258 (±184)283 (±213)
Discharged visitsn = 618n = 1,246−23 (−37 to −8) 0.002
 Minutes, mean (±SD)212 (±147)235 (±156)
Admitted visitsn = 161n = 263−76 (−123 to −28)0.002
 Minutes, mean (±SD)432 (±205)507 (±289)

As with LOS, the analysis of total plan paid for the ED encounter was also stratified by hospital admission status due to poor fit of the combined regression model. Unadjusted results of P-EHR and comparison visits on total paid amount for the ED visit overall and stratified by hospital admission status are displayed in Table 3. The results of the stratified GLM-gamma models are displayed in Table 4. Among discharged visits, no statistical difference was observed between the paid amount for P-EHR visits compared to comparison visits (economic ratio = 1.04, 95% confidence interval [CI] = 0.95 to 1.14). Among visits resulting in hospital admission, P-EHR visits were associated with 18% lower total paid amounts than comparison visits (economic ratio = 0.82, 95% CI = 0.69 to 0.98). The modified Park’s test indicated that the gamma distribution was the best fit for the data. Table 5 displays the adjusted mean plan payment on the dollar scale using the method of recycled predictions from the multivariable GLM-gamma models and the standard errors (SEs) and 95% CIs of the predicted mean differences from bootstrapped sampling. Among ED visits that resulted in a hospital admission, the predicted mean paid amount was $7,019 for P-EHR visits and $8,579 for comparison visits for a savings of $1,560 (p = 0.047).

Table 3.   Unadjusted Comparison of Plan Paid* for the ED Visit Overall and Stratified by Discharge Status
 P-EHR VisitsComparison VisitsDifference in Means (95% CI)p-value†Bootstrap p-value‡
  1. P-EHR = payer-based electronic health record.

  2. *Total plan paid amounts were adjusted to constant 2006 US dollars using CPI-U for Philadelphia-Wilmington-Washington, DC, region.

  3. †p-value from two-sample t-test with unequal variances.

  4. ‡p-value from nonparametric bootstrap.

Overalln = 779n = 1,509−202 (−586 to 182)0.300.29
 Plan paid, mean (±SD)$2,097 (±$3,650)$2,299 (±$5,671)
Discharged visitsn = 618n = 1,24615 (−163 to 193)0.870.89
 Plan paid, mean (±SD)$915 (±$1,072)$900 (±$2,813)
Admitted visitsn = 161n = 263−2,294 (−3,778 to −810)0.0030.003
 Plan paid, mean (±SD)$6,635 (±$5,851)$8,929 (±$9,699)
Table 4.   Economic Ratios (95% CI) From Multivariable GLM-gamma Models for Total ED Paid Amount Stratified by Discharge Status
 Model for Discharged Visits (n = 1,864)Model of Admitted Visits (n = 424)
  1. DCI = Deyo-Charlson Index; ESI = Emergency Severity Index; GLM-gamma = generalized linear model with gamma family and log link; GU = genitourinary; IPA = independent practice association; P-EHR = payer-based electronic health record; POS = point of service; PPO = preferred provider organization.

  2. *Symptoms category included diagnoses associated with ill-defined signs and symptoms (e.g., abdominal pain, fever of unknown origin).

ComparisonReferent groupReferent group
P-EHR1.04 (0.95–1.14)0.82 (0.69–0.98)
Age (× 10 yr)1.03 (1.00–1.06)1.13 (1.04–1.22)
Male sex1.11 (1.00–1.23)1.08 (0.91–1.28)
Previous ED visit0.93 (0.84–1.03)1.12 (0.88–1.42)
Previous hospitalization0.93 (0.75–1.15)0.81 (0.63–1.05)
Previous office visits0.99 (0.97–1.00)1.01 (0.99–1.03)
Previous total plan paid1.01 (1.00–1.01)1.00 (0.99–1.01)
Line of business
 IPAReferent groupReferent group
 PPO0.97 (0.86–1.10)1.00 (0.78–1.29)
 POS0.98 (0.87–1.11)0.82 (0.67–1.00)
 Other0.67 (0.59–0.76)0.86 (0.69–1.07)
Day of week
 WeekdayReferent groupReferent group
 Weekend1.07 (0.95–1.21)1.06 (0.87–1.27)
Triage census (× 10 people)1.01 (0.99–1.03)0.95 (0.92–0.99)
DCI category
 No scoreReferent groupReferent group
 21.08 (0.95–1.24)1.02 (0.81–1.28)
 3 or more0.69 (0.55–0.86)1.00 (0.74–1.34)
Unique drug class count1.01 (0.99–1.02)1.00 (0.98–1.01)
ESI score
 5 (least urgent)Referent groupReferent group
 40.96 (0.72–1.30)1.12 (0.51–2.48)
 31.43 (1.06–1.94)0.88 (0.42–1.86)
 2 (more urgent)1.92 (1.40–2.63)0.99 (0.46–2.15)
ED primary diagnosis
 Injury/poisoningReferent groupReferent group
 Nervous system1.20 (0.99–1.47)1.06 (0.65–1.72)
 Circulatory system2.04 (1.75–2.38)1.40 (1.03–1.90)
 Respiratory system1.00 (0.88–1.14)0.79 (0.55–1.14)
 Digestive system3.19 (1.44–7.08)1.10 (0.81–1.49)
 GU system2.13 (1.83–2.49)0.93 (0.61–1.42)
 Symptoms*1.65 (1.44–1.90)0.80 (0.53–1.19)
 Other1.20 (1.04–1.38)0.90 (0.65–1.23)
Ambulance1.06 (0.94–1.20)0.90 (0.71–1.15)
Constant1.20 (0.99–1.47)1.06 (0.65–1.72)
Table 5.   Predicted Plan Paid From GLM-gamma Models and Bootstrapped 95% CIs for Total ED Paid Amount Stratified by Discharge Status
ModelPredicted Means from GLM-gammaDifferencep-value†Nonparametric Bootstrap of Predicted Differences
P-EHRComparison95% CIp-value
  1. GLM-gamma = generalized linear model with gamma family and log link; P-EHR = payer-based electronic health record.

  2. *Total plan paid amounts were adjusted to constant 2006 US dollars using CPI-U for Philadelphia-Wilmington-Washington, DC, region.

  3. †p-value based on Wald z-statistic from GLM model.

Discharged*$ 923$ 886$ 370.37−39 to 1210.37
Admitted*$ 7,019$ 8,579$ 1,5600.027   43 to 2,9100.047

To evaluate the effect of not including visits from September and October 2004 in the comparison group, the stratified regressions were reestimated with the addition of comparison visits from the 2 additional months. An additional 990 comparison ED (140 admitted, 850 discharged) were included with the inclusion of 2 additional months. Note that the apparent average number of ED visits per month is larger for these 2 months since having LOS and ED census captured was not applied as an inclusion criterion. Among ED visits that resulted in hospital admission, EHR visits were associated with 24% lower total paid amounts than visits in the expanded comparison group (economic ratio = 0.76; 95% CI = 0.66 to 0.90). Using recycled predictions, this corresponds to predicted mean paid amounts of $6,075 for EHR visits and $7,995 for comparison visits resulting in a savings in plan payments of $1,920.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

Health information exchange technologies, with varying similarities to the one evaluated in this study, are currently being developed and implemented in many practice settings across the United States; however, published literature on their actual benefits remains sparse.6 Providing clinicians with timely and clinically relevant information should improve both the quality and the efficiency of clinical decision-making.42 Efficiency is gained by decreasing the time and effort required for obtaining needed clinical information. In this study, a greater time savings was observed among ED visits that resulted in hospital admission. This finding is consistent with theories of decision making and problem solving. For example, the admitting physician may engage in more thorough or detailed data gathering for patients who are ultimately admitted into the hospital. With the potential for a more complex medical and/or drug history, the admitting physician and clinical staff may have more time for completing histories, following up with providers of recent medical care (laboratory tests, pathology reports, diagnostic scans, procedures, etc.), or engaging in diagnostic hypothesis testing. According to the cognitive continuum theory of decision-making, as either more time is available or the complexity of the case increases, the decision-maker is more likely to engage in less intuitive and more analytical decision-making.43 In these instances, analytical reasoning (e.g., data gathering and diagnostic hypothesis testing for rule-out or rule-in) may be more efficient when the P-EHR is available. On the other hand, there may be less of an observed effect on ED LOS for discharged visits because these visits may be more routine, the patients may be able to communicate the information that would have been in the EHR, and the medical history contained in the P-EHR may potentially be less pertinent to the decision-making process.

The United Hospital Fund lists reduced LOS as a primary outcome that is important in the evaluation of HIE technologies.6 LOS in the ED is not only important as an indicator for efficiency, as longer lengths of stay may be associated with inefficient use of resources, but it is also related to patient satisfaction.44,45 Longer waiting times can lead to patients leaving without receiving care and may, therefore, increase the concern for patient safety.46,47 Implementing the use of a P-EHR can lead to a substantial and clinically meaningful reduction in ED duration. Gorelick et al.48 observed a 15-minute decrease in ED duration after changing a single process of care by implementing a parallel, in-room registration system. Forster et al.49 found a statistically significant 18-minute increase (95% CI = 11 to 24 minutes) among admitted ED visits for every 10% increase in hospital occupancy. The authors of both studies considered the magnitude of the observed time savings clinically and practically meaningful. Longer ED durations have been observed in previous research among admitted visits.49

We also hypothesized that the P-EHR would be associated with a reduction in the total amount the plan paid for the ED visit. Our rationale was that the EHR would be associated with improvements in efficiency through reductions in redundant medical services and an improvement in the quality of decision-making and fewer errors. In a randomized controlled trial of HIE using a longitudinal computerized patient record system from an urban hospital to two EDs in the Indianapolis metropolitan area, Overhage et al.17 observed a $26 (95% CI = $2 to $51) savings per encounter in one urban ED. Other studies have observed improvements in test ordering and test charges in the ED associated with patient clinical summaries.19,20 The results of this analysis agree with previous research, although the savings among admitted ED visits is much larger in the current study. A statistically significant difference in the distribution of health plan line of business between P-EHR and comparison visits was observed. This difference cannot account directly for the differences in plan payments observed in this study, because health plan reimbursement and management of facility and professional services do not differ by line of business. It is possible that younger, healthier patients may select into preferred provider organization plans as opposed to the other plans. To control for possible selection bias, all regressions included line of business, age, comorbidity burden, and severity at ED presentation.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

Overall, P-EHR visits occurred within 6 months of implementation of the P-EHR. Evaluating the effectiveness of a new technology is challenging during the implementation phase as there is limited sustained use, and the users are new and inexperienced with the technology.44 The newness of the technology may have served to bias the comparison toward the null (i.e., underestimation of the true benefits that may be quantified during sustained use). Although this is a limitation, evaluation of the use of technologies, such as the one under study, is critical at all stages of system development to inform users and developers of potential issues and limitations. Furthermore, it was observed that the P-EHR was accessed for about half of the total ED visits from health plan members between September 1, 2005, and February 17, 2006. The main contributor to this observation was perceived time and likely familiarity and comfort level with the technology by the registration clerk (again because the evaluation was during the implementation phase). It is likely that for these cases, the clerk either forgot to access the P-EHR or was in too much of a hurry and may have been less likely to take the time to access the P-EHR. The latter may be suggested by the fact that the census at the time of admission was lower by five patients, on average, for visits with the P-EHR accessed than for comparison visits. This highlights the importance of controlling for ED census for this type of observational study and specifically for not using concurrent “non P-EHR-accessed” visits as the comparison.

Other limitations related to the P-EHR itself include the fact that results from laboratory tests, imaging procedures, pathology, etc., are not included. Also, since the information contained in the P-EHR is based on payer data, the data are subject to errors as well as a lag time from a service encounter to the time such service appears in the EHR. This time is dependent upon provider submission for reimbursement and may lag up to 60 days. Further, although the P-EHR was accessed during each of the P-EHR visits, there is no assurance that the information contained in the record was actually used or deemed useful by the ED staff. Predicting the percentage of time HIE-obtained clinical information is useful to ED staff is very challenging and has been roughly estimated to be about 25% of the time.50

Limitations related to study design stem from the retrospective, observational nature of the study, which is subject to several threats to internal validity. Selection is a threat due to the lack of randomization; however, differences on observed covariates were controlled using regression methods. Furthermore, this study used historical controls. Although the comparison visits were sampled over the same months in the previous year, history remains a salient threat to internal validity. Maturation is also a threat, as the same patient may present to the ED during the period prior to and after P-EHR implementation and thus contribute a P-EHR visit and a comparison visit; however, this occurred for less than 3% of the study population.

Another potential limitation related to study design is the period of time over which the comparison visits were included. ED census was only available beginning in November 2004. As a result, the time frame for the selection of comparison visits was not identical to the months from which the EHR visits came from in the following year and were shifted forward by a month and a half (through March 2005). To examine the effect of not including September–October 2004 in the comparison visit sample, the economic analysis was repeated using these two additional months (but not including the census variable). The results were consistent with the main study.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

Based on health insurance claims from all health plan-covered medical and pharmacy encounters, the payer-based electronic health record studied was associated with a significant reduction in ED length of stay among both admitted and discharged visits. Technologies that can reduce ED lengths of stay can have a substantial effect on the care provided to patients and their satisfaction. The results also suggest that the P-EHR is associated with lower health plan paid amounts among admitted visits. Additional research should be conducted to confirm these findings. This study highlights that, when provided to EPs, electronic health records based on a summary of patients’ medical and pharmacy claims can have a meaningful effect on the improvement of ED throughput and plan payments.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

The authors acknowledge Christy Fang, MS, Research Analyst, HealthCore, for providing data programming support; Edward Armstrong, PharmD, and Terri L. Warholak, PhD, University of Arizona College of Pharmacy, for providing study design advice; and Duane Sherrill, PhD, University of Arizona College of Public Health, for providing statistical analysis advice.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information
  • 1
    Obama B. Weekly Address. Washington, DC: The White House Briefing Room; January 24, 2009. Available at: Accessed Apr 4, 2009.
  • 2
    Chaudhry B, Wang J, Wu S, et al. Systematic review: impact of health information technology on quality, efficiency, and costs of medical care. Ann Intern Med. 2006; 144:74252.
  • 3
    Overhage JM, Evans L, Marchibroda J. Communities’ readiness for health information exchange: the National Landscape in 2004. J Am Med Inform Assoc. 2005; 12:10712.
  • 4
    Kaelber DC, Bates DW. Health information exchange and patient safety. J Biomed Inform. 2007; 40(6 Suppl):S405.
  • 5
    Marchibroda JM. Health information exchange policy and evaluation. J Biomed Inform. 2007; 40(Suppl 6):S11S16.
  • 6
    Hripcsak G, Kaushal R, Johnson KB, et al. The United Hospital Fund meeting on evaluating health information exchange. J Biomed Inform. 2007; 40(Suppl 6):S310.
  • 7
    Institute of Medicine, Committee on the Future of Emergency Care in the United States Health System. Hospital-based emergency care: at the breaking point. Washington, DC: National Academies Press, 2007, pp 180, 129–208.
  • 8
    Shapiro JS, Kannry J, Kushniruk AW, Kuperman G. Emergency physicians’ perceptions of health information exchange. J Am Med Inform Assoc. 2007; 14:7005.
  • 9
    Leape LL, Brennan TA, Laird N, et al. The nature of adverse events in hospitalized patients. Results of the Harvard Medical Practice Study II. N Engl J Med. 1991; 324:37784.
  • 10
    Chisholm CD, Collison EK, Nelson DR, Cordell WH. Emergency department workplace interruptions: are emergency physicians “interrupt-driven” and “multitasking”? Acad Emerg Med. 2000; 7:123943.
  • 11
    Burstin H. Crossing the quality chasm in emergency medicine. Acad Emerg Med. 2002; 9:10747.
  • 12
    Goldberg RM, Kuhn G, Andrew LB, Thomas HA Jr. Coping with medical mistakes and errors in judgment. Ann Emerg Med. 2002; 39:28792.
  • 13
    Cosby KS. A framework for classifying factors that contribute to error in the emergency department. Ann Emerg Med. 2003; 42:81523.
  • 14
    Chamberlain JM, Slonim A, Joseph JG. Reducing errors and promoting safety in pediatric emergency care. Ambul Pediatr. 2004; 4:5563.
  • 15
    Selbst SM, Levine S, Mull C, Bradford K, Friedman M. Preventing medical errors in pediatric emergency medicine. Pediatr Emerg Care. 2004; 20:7029.
  • 16
    Stiell A, Forster AJ, Stiell IG, Van Walraven C. Prevalence of information gaps in the emergency department and the effect on patient outcomes. Can Med Assoc J. 2003; 169:10238.
  • 17
    Overhage JM, Dexter PR, Perkins SM, et al. A randomized, controlled trial of clinical information shared from another institution. Ann Emerg Med. 2002; 39:1423.
  • 18
    Hripcsak G, Sengupta S, Wilcox A, Green RA. Emergency department access to a longitudinal medical record. J Am Med Inform Assoc. 2007; 14:2358.
  • 19
    McDonald C, Blevins L, Glazener T, Haas J, Lemmon L, Meeks-Johnson J. Data base management, feedback control, and the Regenstrief medical record. J Med Syst. 1983; 7:11125.
  • 20
    Wilson GA, McDonald CJ, McCabe GP Jr. The effect of immediate access to a computerized medical record on physician test ordering: a controlled clinical trial in the emergency room. Am J Public Health. 1982; 72:698702.
  • 21
    Schmidt HG, Norman GR, Boshuizen HP. A cognitive perspective on medical expertise: theory and implication. Acad Med. 1990; 65:61121.
  • 22
    Bureau of Labor Statistics, U.S. Department of Labor. Consumer Price Index. All Urban Consumers: Philadelphia-Wilmington-Atlantic City, PA-NJ-DE-MD. Available at: Accessed Apr 7, 2008.
  • 23
    Andersen R, Newman JF. Societal and individual determinants of medical care utilization in the United States. Milbank Mem Fund Q Health Soc. 1973; 51:95124.
  • 24
    Aday LA, Andersen R. A framework for the study of access to medical care. Health Serv Res. 1974; 9:20820.
  • 25
    Kyriacou DN, Ricketts V, Dyne PL, McCollough MD, Talan DA. A 5-year time study analysis of emergency department patient care efficiency. Ann Emerg Med. 1999; 34:32635.
  • 26
    Rathlev NK, Chessare J, Olshaker J, et al. Time series analysis of variables associated with daily mean emergency department length of stay. Ann Emerg Med. 2007; 49:26571.
  • 27
    Gardner RL, Sarkar U, Maselli JH, Gonzales R. Factors associated with longer ED lengths of stay. Am J Emerg Med. 2007; 25:64350.
  • 28
    McCarthy ML, Aronsky D, Jones ID, et al. The emergency department occupancy rate: a simple measure of emergency department crowding? Ann Emerg Med. 2008; 51:1524.
  • 29
    Tanabe P, Gimbel R, Yarnold PR, Adams JG. The Emergency Severity Index (version 3) 5-level triage system scores predict ED resource consumption. J Emerg Nurs. 2004; 30:229.
  • 30
    Tanabe P, Gimbel R, Yarnold PR, Kyriacou DN, Adams JG. Reliability and validity of scores on The Emergency Severity Index version 3. Acad Emerg Med. 2004; 11:5965.
  • 31
    Dong SL, Bullard MJ, Meurer DP, et al. Predictive validity of a computerized emergency triage tool. Acad Emerg Med. 2007; 14:1621.
  • 32
    Sibbritt D, Isbister GK, Walker R. Emergency department performance indicators that encompass the patient journey. Qual Manag Health Care. 2006; 15:2738.
  • 33
    Deyo RA, Cherkin DC, Ciol MA. Adapting a clinical comorbidity index for use with ICD-9-CM administrative databases. J Clin Epidemiol. 1992; 45:6139.
  • 34
    Schneeweiss S, Seeger JD, Maclure M, Wang PS, Avorn J, Glynn RJ. Performance of comorbidity scores to control for confounding in epidemiologic studies using claims data. Am J Epidemiol. 2001; 154:85464.
  • 35
    Farley JF, Harley CR, Devine JW. A comparison of comorbidity measurements to predict healthcare expenditures. Am J Manag Care. 2006; 12:1109.
  • 36
    Barber JA, Thompson SG. Analysis of cost data in randomized trials: an application of the non-parametric bootstrap. Stat Med. 2000; 19:321936.
  • 37
    Manning WG, Mullahy J. Estimating log models: to transform or not to transform? J Health Econ. 2001; 20:46194.
  • 38
    Barber J, Thompson S. Multiple regression of cost data: use of generalised linear models. J Health Serv Res Policy. 2004; 9:197204.
  • 39
    Glick H, Doshi J, Sonnad S, Polsky D. Economic Evaluation in Clinical Trials. New York, NY: Oxford University Press, 2007.
  • 40
    Rogers W. sg17: Regression standard errors in clustered samples. Stata Tech Bull Repr. 1993; 3:8894.
  • 41
    Williams RL. A note on robust variance estimation for cluster-correlated data. Biometrics. 2000; 56:6456.
  • 42
    Elson RB, Faughnan JG, Connelly DP. An industrial process view of information delivery to support clinical decision making: implications for systems design and process measures. J Am Med Inform Assoc. 1997; 4:26678.
  • 43
    Hammond KR, McClelland GH, Mumpower J. Human Judgment and Decision Making: Theories, Methods, and Procedures. New York, NY: Hemisphere, 1980.
  • 44
    Johnson KB, Gadd C. Playing smallball: approaches to evaluating pilot health information exchange systems. J Biomed Inform. 2007; 40(6 Suppl):S216.
  • 45
    Magaret ND, Clark TA, Warden CR, Magnusson AR, Hedges JR. Patient satisfaction in the emergency department--a survey of pediatric patients and their parents. Acad Emerg Med. 2002; 9:137988.
  • 46
    Arendt KW, Sadosty AT, Weaver AL, Brent CR, Boie ET. The left-without-being-seen patients: what would keep them from leaving? Ann Emerg Med. 2003; 42:31723.
  • 47
    Kennedy J, Rhodes K, Walls CA, Asplin BR. Access to emergency care: restricted by long waiting times and cost and coverage concerns. Ann Emerg Med. 2004; 43:56773.
  • 48
    Gorelick MH, Yen K, Yun HJ. The effect of in-room registration on emergency department length of stay. Ann Emerg Med. 2005; 45:12833.
  • 49
    Forster AJ, Stiell I, Wells G, Lee AJ, Van Walraven C. The effect of hospital occupancy on emergency department length of stay and patient disposition. Acad Emerg Med. 2003; 10:12733.
  • 50
    Greene J. The trials and tribulations of health information sharing: the turbulent rise of the RHIO. Ann Emerg Med. 2007; 50:54951.


  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

Summary of Payer-based Electronic Health Record Elements

• Basic patient information
 ○ Age
 ○ Sex
 ○ Date of birth
 ○ Address
• Primary care physician information
• A list of major conditions and severity
• A health status measure
• Previous inpatient admissions and principal diagnoses
• Previous ED visits
• Previous laboratory services (actual test results/images not provided)
• Previous outpatient medication use (including number of fills and date of last fill)
• Name, specialty, and contact information for recent providers seen
• Clinical flags
• Active and closed care management summaries

Supporting Information

  1. Top of page
  2. Abstract
  3. Methods
  4. Results
  5. Discussion
  6. Limitations
  7. Conclusions
  8. Acknowledgments
  9. References
  10. Appendix
  11. Supporting Information

Data Supplement S1. Coefficients (95% CI) for multivariable ordinary least squares regression models for ED length of stay.

Please note: Wiley Periodicals Inc. are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article.

ACEM_816_sm_DataSupplementS1.pdf37KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.