Health Informatics & Inform System - Assignment 4

496

Authors’ Note: This article, submitted to Medical Care Research and Review on November 30, 2007, was revised and accepted for publication on December 11, 2007.

Medical Care Research and Review Volume 65 Number 4 August 2008 496-513

© 2008 Sage Publications 10.1177/1077558707313437

http://mcr.sagepub.com hosted at

http://online.sagepub.com

Do Hospitals With Electronic Medical Records (EMRs) Provide Higher Quality Care? An Examination of Three Clinical Conditions Abby S. Kazley Medical University of South Carolina Yasar A. Ozcan Virginia Commonwealth University

This study investigates how hospital electronic medical record (EMR) use influences qual- ity performance. Data include nonfederal acute care hospitals in the United States. Sources of the data include the American Hospital Association, Hospital Quality Alliance, the Healthcare Information and Management Systems Society, and the Centers for Medicare and Medicaid Services case-mix index sets. The authors use a retrospective cross-sectional format with linear regression to assess the relationship between hospital EMR use and quality performance. Quality performance is measured using 10 process indicators related to 3 clinical conditions: acute myocardial infarction, congestive heart failure, and pneu- monia. The authors also use a propensity score adjustment to control for possible selection bias. After this adjustment, the authors identify a positive significant relationship between EMR use and 4 of the 10 quality indicators. They conclude that there is limited evidence of the relationship between hospital EMR use and quality.

Keywords: electronic medical records (EMRs); health information technology (HIT); hospital quality; hospital compare

The recent environment for hospitals has been uncertain and challenging.Hospitals have felt pressure to increase both their efficiency and quality perfor- mance, and competition has made these demands more pressing. One proposed method for increasing hospital quality is the use of health information technology (HIT). HIT may improve health care quality through the use of standardized clinical pathways; e-prescribing systems, which would detect drug interactions; and the better and more complete documentation of care (Miller & Sim, 2004; Thompson & Brailer, 2004). According to the Institute of Medicine (IOM; Crossing the Quality Chasm: A New Health System for the 21st Century, 2001), “health care delivery has

Kazley, Ozcan / Hospitals With Electronic Medical Records 497

been relatively untouched by the revolution in information technology that has been transforming nearly every other aspect of society” (p. 15). This same report calls for dras- tic changes to the health care system, including increased and improved use of HIT (Crossing the Quality Chasm: A New Health System for the 21st Century, 2001). The pur- pose of this analysis is to determine if electronic medical records (EMRs) are associated with higher quality care in a national sample of hospitals in three clinical conditions.

Since 2001, many efforts have been made to increase HIT use, but the prevalence has increased only modestly (G. F. Anderson, Frogner, Johns, & Reinhardt, 2006; Burt & Hing, 2005; Health Information Management Systems Society [HIMSS] Foundation, 2004, 2005; Poon et al., 2006). An example of such efforts is the Leapfrog practice of encouraging hospital Computerized Physician Order Entry (CPOE) use and ambulatory EMR use because of their potential use as a structural mea- sure of hospital quality (Leapfrog Group for Patient Safety, 2007). Despite these efforts, only approximately 31% of hospital emergency departments and 29% of hospital out- patient departments use EMRs (Burt & Hing, 2005). The HIMSS Analytics Annual Report from 2004-2005 first quarter (HIMSS Foundation, 2005) confers claiming that “hospitals continue to under-invest in IT spending as represented in the low percent- ages of installed electronic medical record (EMR) and clinical applications” (p. 3).

Furthermore, Thompson and Brailer (2004) released a report that outlined the framework for strategic action to increase the use of EMRs in the United States. EMRs are individual computerized documents of patient medical history, treatments, lab results, X rays, and MRIs, which are often coupled with other tools such as CPOE, electronic prescribing, and a multitude of administrative and fiscal functions (Schmitt & Wofford, 2002; Shortliffe, 1999). Up until this point, individual medical records have been primarily maintained in paper format, which poses a number of problems. First, transcribing, storing, and retrieving these files are challenging and expensive endeavors, and paper medical records can be lost. Many patients do not transfer their medical records between their health care providers, which leads to an incomplete view of one’s medical history, a fragmentation of care, and possible dupli- cation of services if diagnostic test results are not shared (G. F. Anderson et al., 2006). Brailer (2005) claims that this lack of integration, based on multiple care providers per patient, leads to errors and duplication, both of which can increase health care costs. In addition to the possible benefits in connecting clinicians in different institu- tions, the use of an EMR in a single facility may increase communication and coor- dination among clinicians in that facility, as they could all access and add to the record simultaneously (Thompson & Brailer, 2004).

Another problem associated with the use of traditional paper records is their qual- ity. Some have indicated that traditional paper records are inaccurate, illegible, and incomplete (Shortliffe, 1999). In relation to this, some have claimed that paper records cannot adequately deal with the complex health care that is provided today (Shortliffe, 1999; Varon & Marik, 2002). Even before the advent of HIT, Donabedian (1980) wrote about the downfalls of relying on traditional paper records, which do

not always reflect the true nature of the interaction between a patient and provider. According to a report from the IOM, each year, between 44,000 and 98,000 deaths occur in hospitals as the result of medical errors (Kohn, Corrigan, & Donaldson, 1999).

Hospital EMR use may reduce medical errors by omitting poorly handwritten patient charts, physician orders, or provider notes and through other EMR tools such as e-prescribing, screen prompters, CPOE, and standardized clinical pathways or procedures (Shortliffe, 1999; Thompson & Brailer, 2004). In addition, dangerous drug interactions and inappropriate testing or treatment may be avoided, and proper documentation may increase with screen prompters (Linder, Ma, Bates, Middleton, & Stafford, 2007; Miller & Sim, 2004; Thompson & Brailer, 2004). Although not all EMR systems have each of these features, many offer some possible benefits to the quality of care through improved documentation and communication. In support of the EMR, the IOM (Crossing the Quality Chasm: A New Health System for the 21st Century, 2001) calls for “the elimination of most handwritten clinical data by the end of the decade” (p. 5).

Quality health care has been defined as “doing the right thing at the right time in the right way to the right person and having the best possible results” (Agency for Healthcare Research Quality, 2004). Quality health care has many components, including error reduction and patient safety. These two important components of health care quality, error reduction and patient safety, may be especially affected through EMR use, which can apply standardized clinical pathways, CPOE, and other clinical reminders and indicators of possible prescription interactions and potentially dangerous medical procedures (Miller & Sim, 2004; Thompson & Brailer, 2004). Based on the automation and available features of EMRs, the quality of health care stands to benefit from the use of EMRs (Crossing the Quality Chasm: A New Health System for the 21st Century, 2001). G. K. Anderson (2004) suggests that EMRs may increase the quality of health care by drastically reducing many of the patient deaths caused by patient errors, whereas Miller and Sim (2004) have stated that “of all the health infor- mation technology in use, the electronic medical record (EMR) has the most wide- ranging capabilities and thus the greatest potential for improving quality” (p. 116).

EMRs could improve many of the flaws of paper medical records. EMRs can be transportable, transferable, complete, automated, standardized, and typed, as well as connected to standardized clinical pathways and other tools, which could reduce the number of medical errors while increasing communication and coordination between physicians (Fonkych & Taylor, 2005). Because of these characteristics, EMRs have been predicted to increase the quality of health care services (Miller & Sim, 2004; Thompson & Brailer, 2004). This study attempts to determine if EMRs are associated with higher quality care in treating three certain clinical conditions. The three clinical conditions include acute myocardial infarction (AMI), congestive heart failure, and pneumonia. The research questions are as follows: Do hospitals with EMRs provide higher quality care? And in which conditions can EMRs make an impact on quality?

498 Medical Care Research and Review

New Contribution

Although some have suggested that EMR use will drastically change and improve health care in the United States, a majority of previous research in this area has focused on case studies, small sample sizes, and qualitative or anecdotal evidence with limited generality (Asch et al., 2004; Kinn et al., 2001; Spencer, Swanson, Hueston, & Edberg, 1999). This study attempts to fill this gap by examining a large, national sample size of hospitals to compare those that use EMRs to those that do not in the three clinical conditions using process measures. Although hospitals have been encouraged to consider EMR adoption and use, there is little rigorous evidence to indicate that it will change health care for the better.

One challenge for researchers has been finding or defining what hospital EMR use entails; in this study, it is defined by the HIMSS Foundation (2004) as a “com- prehensive database system used to store and access patients’ health care informa- tion electronically. The computer-based patient record replaces the paper medical record as the primary source of information for healthcare” (p. 152). The use of this clear and strong definition allows for the comparison of providers that otherwise may refer to very different systems and features in EMR use. In addition, this study examines quality using a new data set of process measures, which allow for the com- parison of a large group of hospitals based on the large number of respondent orga- nizations involved with the Hospital Quality Alliance (HQA). Because EMR use may change the processes and outcomes of care through the better automation and coordination among clinicians, the examination of this practice with these data allow the researchers to examine the relationship between hospital HIT and compliance with clinical process guidelines associated with greater quality patient outcomes.

Previous Research

Some previous research has examined the relationship between EMR use and quality. One study compares the quality of care for patients at the Veteran’s Health Administration (VHA), which has used EMRs for more than a decade, to patients in a national sample treated at community hospitals (Asch et al., 2004). According to this study, which examined 596 VHA patients’ care and 996 community hospital patients’ care, the VHA’s EMR system is associated with higher levels of patient care quality in the areas of overall quality, chronic disease management, and preventative care, but not in the area of acute care. Because the VHA’s system is interoperable, it is likely that screen prompters and electronic tracking and reminders increase clini- cian and patient attention to the management of chronic disease and preventative care without influencing the quality of acute care.

Also contributing to quality, certain health care screenings appear to be carried out more effectively with the use of EMRs. Spencer et al. (1999) claim that EMRs used in conjunction with continuous quality improvement lead to drastic improvements in

Kazley, Ozcan / Hospitals With Electronic Medical Records 499

documentation and screening for smoking status at a clinic in Eau Claire, Wisconsin. According to this study, the smoking screening rate increased from 18.4% of patients to 80.3% of patients in 2 weeks, while the rate of smoking cessation counseling with appropriate documentation rose from 17.1% to 48.3%.

Another study examined how EMR use affects patient populations with specific conditions. Kinn et al. (2001) compared two similar groups of patients at an outpa- tient cardiology practice to assess the impact of an EMR on patient outcomes. After 12 months, significant differences between patients with the traditional paper chart and those with the new EMR systems emerged. According to this study, patients with the EMRs were more likely than those without EMRs to be on an appropriate drug, to reach lipid-related goal levels, to have better medical documentation of treatment and symptoms, and to be receiving lipid-lowering therapy. These benefits are attrib- uted to the availability of real-time data, which allowed physicians to better monitor and more quickly respond to patient progress.

A similar study by Linder et al. (2007) examined the relationship between orga- nizational EMR use and quality. The authors in this study examined the ambulatory care setting and 17 related process indicators that were believed to relate to the qual- ity of medical management of common diseases, antibiotic use, preventive counsel- ing, screening tests, and avoiding inappropriate prescribing in elderly patients. They found that EMRs in ambulatory care settings did not improve clinical adherence con- sistently in all measures; rather, they report that EMR use was associated with increased performance in two measures (avoiding benzodiazepine use for patients with depression and avoiding routine urinalysis during general medical examina- tions), but EMR use actually decreased quality in one of the measures (statin pre- scribing to patients with hypercholesterolemia).

Theoretical Framework and Conceptual Model

Some have credited Donabedian with providing the first framework for examin- ing health care quality (Mick & Wyttenbach, 2003). Donabedian’s (1980) structure, process, outcome model claims that organizations implement structures, which influence processes and outcomes. Although this framework does not actually pro- vide information about the direction of relationships, it does emphasize the fact that organizational structure will affect the flow and practice procedures in an organiza- tion, which will in turn affect the performance of organizations. The roots of this framework in General Systems Theory, however, are evident, and Donabedian indi- cates that his model is intended to be “a guide, not a straightjacket” (p. 89). Because of the dynamic framework of Donabedian’s model, it is herein joined with previous literature and rationale to create the conceptual model used here.

According to Scott (2003), structure includes organizational inputs and provides organizations with the capacity to do work. Structures include human, physical, and financial resources, including human resources, facilities, equipment, organization,

500 Medical Care Research and Review

information systems and records, financing, management, and amenities (Starfield, 1992). Donabedian (1980) states that “good structure, that is, a sufficiency of resources and proper system design is probably the most important means of pro- tecting and promoting the quality of care” (p. 82). Processes include “the set of activities that take place between providers and patients,” such as clinical decision making and the protocols and procedures used in health care (Paganini, 1993, p. 7). Because of their automated nature, it is likely that EMRs as a structure will influ- ence the processes of health care through provider requirements for care (mandatory fields) and detection of inappropriate diagnoses, pharmaceuticals, practices, or screening. In turn, outcomes, defined by Donabedian (1980) as “a change in a patient’s current and future health status that can be attributed to antecedent health care” (p. 82), are influenced by the processes of care associated with EMRs. Because EMRs are an organizational structure, their potential influence on processes and out- comes of care is supported with this framework. Additionally, the quality process measures in this study could be prompted by screen requirements of an EMR system or field, and these practices have been developed because of their linkage to quality patient outcomes for each of the conditions (Jha, Li, Orav, & Epstein, 2005).

EMRs are predicted to make the processes of health care more standardized and automated through the presence of screen prompters, mandatory patient information fields to be entered, and tools to catch prescription interactions or inappropriate diag- noses (Miller & Sim, 2004; Shortliffe, 1999). These improved automated processes are expected to lead to fewer medical errors and oversights, thus improving health care quality outcomes. Hospitals that use EMRs have likely adopted them as a struc- tural feature because they believe they will lead to better performances in processes and outcomes. The proper structure can allow an organization to thrive in perfor- mance, increasing both efficiency and quality (Donabedian, 1980). Proponents of EMR use claim that they will reduce medical errors through standardized clinical pathways to ensure that evidence-based medicine practice directs care, through the decrease of adverse prescription use based on an automated interaction detection and through the automated access physicians may have to patient laboratory and other diagnostic results (Ash & Bates, 2004; Crossing the Quality Chasm: A New Health System for the 21st Century, 2001; Thompson & Brailer, 2004). Also, EMRs will eliminate mistakes that occur based on illegible handwritten medical records. EMRs allow providers immediate access to patient information, are connected to a library of medical information, and often generate reminders or indications of important or time-sensitive clinical information (Linder et al., 2007; Miller & Sim, 2004; Thompson & Brailer, 2004). These EMR systems can also allow clinicians in the same facility to add to and access medical records for a patient at any time instead of having to wait for the paper record. This feature alone could drastically change provider communication and coordination of care for teams of physicians. In this way, the changes in the processes of care achieved through EMR (structure) use will lead to improved health care quality (outcome).

Kazley, Ozcan / Hospitals With Electronic Medical Records 501

Hypothesis: Hospitals with EMRs will report higher quality than those without EMRs in each of the quality measures.

Because we believe that hospitals do not randomly select to use EMR systems, we realize that there is the potential for selection bias in this study. We will control for the possibility that hospitals that adopt EMRs are already performing at a higher level through the use of a propensity score to predict the likelihood of hospital EMR use based on organizational factors. This likelihood is included in our conceptual model and will be further discussed in the method section.

Research Design

This study is conducted using a retrospective, nonexperimental design with one observation. The observation takes place in 2004 and compares hospitals with EMRs to those without EMRs. The data come from several sources, including the HIMSS, the American Hospital Association (AHA), the Centers for Medicare and Medicaid Case Mix Index (CMS CMI), and the HQA. The unit of analysis is the acute care hospital, and the sample includes all nonfederal hospitals that have submitted valid quality scores in at least one measure area.

Although quality is a highly complex construct, this study measures it using newly available data. The HQA is a group of organizations working together to collect vol- untary information from hospitals. The organizations include the CMS, the Joint Commission on Accreditation of Healthcare Organizations, the AHA, the American Medical Association, the Association of American Medical Colleges, the Chamber of Commerce, and others. The information collected is publicly available for researchers and consumers through the Medicare Modernization Act and includes “ten indicators of the quality of care for acute myocardial infarction, congestive heart failure, and pneumonia” in 2004 (Jha et al., 2005, p. 265). These data are updated and audited quarterly and have since added additional measures (Hospital Quality Alliance Overview Summary, 2006).

The quality measures are process measures that have been selected as indicators of hospital quality based on their widely endorsed validity and account for 15% of all Medicare admissions, though these measures are available and included for the entire patient population in this study (Jha et al., 2005). The measures of quality of care for AMI include the use of aspirin and a beta blocker at arrival at the hospital and at dis- charge and the use of angiotensin-converting enzymes for left ventricular systolic dys- function. The measures of quality of care for congestive heart failure include assessment of the left ventricular function and use of an angiotensin-converting- enzyme inhibitor. The measures of quality of care for pneumonia include the time of antibiotic therapy, the availability of the pneumococcal vaccine, and the assessment of oxygenation. Each of the indicators is dichotomous (was the guideline followed?), and the measure is a percentage representing the number of qualifying patients who

502 Medical Care Research and Review

received the care described. We especially believe that these measures are appropriate for use in this study because EMRs’ potential contribution to quality likely lies with increased clinician adherence to standardized process measures associated with greater quality outcomes rather than an EMR’s direct influence on quality outcomes.

Although each of the hospitals included in this study submitted quality measure data, not all of these hospitals submitted valid data for each of the 10 measures. Each of the hospitals has a continuous value for each measure that indicates the percentage of patients for whom the process was followed as required. Data are not considered valid if a hospital treats fewer than 25 qualifying patients in each measure (Jha et al., 2005). For this reason, hospitals that treat fewer than 25 patients in any measure are excluded from this analysis of that particular measure. The identification of data mea- sures with fewer than 25 patients as being potentially unreliable and invalid is consis- tent with the CMS recommendation for use of these data. The exclusion of hospitals with fewer than 25 patients for any condition prevents possible misinterpretation of hospital performance based on a limited and perhaps unrepresentative sample size.

The HIMSS data, formerly the Dorenfest Data, provide information about HIT use and staffing for thousands of hospitals, chronic care facilities, and ambulatory practices (Hillestad et al., 2005). These data will be used to measure hospital EMR use. According to the HIMSS data (HIMSS Foundation, 2004), EMRs are called Computerized Patient Records (CPRs) and are a

comprehensive database system used to store and access patients’ health care informa- tion electronically. The computer-based patient record replaces the paper medical record as the primary source of information for healthcare meeting all clinical, legal, and administrative requirements. It is seen as a virtual compilation of non-redundant health data about a person across a lifetime, including facts, observations, interpreta- tions, plans, actions, and outcomes. The CPR is supported by a system that captures, stores, processes, communicates, secures, and presents information from multiple dis- parate locations as required. (p. 152)

Using this definition, hospital EMR use is categorized as automated, not automated, or contracted. In this study, we identify hospitals with automated EMRs as those with EMR use and do not consider not automated or contracted in this measure. Automated EMRs are those with which “software and hardware is used to automate a process” (HIMSS Foundation, 2004, p. 159).

This model includes several other variables to control for potentially confound- ing factors. The variables include bed size, ownership, system affiliation, CMI, and payer mix. Previous research has indicated that the variables may potentially influ- ence hospital behavior and performance (Ginn & Young, 1992; Jha et al., 2005; Milcent, 2005; Zinn, Proenca, & Rosko, 1997). Bed size comes from the AHA data and is the number of beds set up and staffed. To allow for better interpretation of this coefficient, it was rescaled by dividing its value by 100 to prevent the coefficient in

Kazley, Ozcan / Hospitals With Electronic Medical Records 503

the analyses from being a zero value. Ownership also comes from the AHA data and includes for-profit, nonprofit, and publicly owned hospitals. The publicly owned hospitals serve as the reference group. System affiliation is a dichotomous measure that comes from the AHA data. The CMI is provided by CMS and represents the severity of patient illness and conditions in a hospital. Payer mix is calculated from the AHA data and is the number of Medicare and Medicaid inpatient days divided by the total number of inpatient days per hospital.

Method

We use a multivariate regression approach to consider the effects of EMR use on hospital adherence to indicators related to the three conditions. We regressed 10 indi- cators of quality on EMR use and organizational control variables. A significant coefficient on EMR use indicates it is related to the quality indicator. The following equation represents the models:

Quality = f(EMR use, size, ownership, case mix, system membership, payer mix)

In addition, we use a propensity score approach to control for possible selection bias. This may be a problem because hospitals that are already high-quality providers may be more likely to adopt EMRs than those with lower quality. The propensity score is the likelihood of hospital EMR adoption based on the organiza- tional characteristics in the model (ownership, system membership, case mix, payer mix, and size). A propensity score approach enables us to control for the potential confounding bias of hospital EMR use by using the predicted likelihood of EMR use from a regression that then divides the sample into five equal strata based on this likelihood (Coyte, Young, & Croxford, 2000; Rubin, 1997; Stearns, Dalton, Holmes, & Seagrave, 2006). According to Coyte et al. (2000), “five strata suffice to remove over 90% of the bias” (p. 914). Once the five strata are calculated, they are included in the models to control for the possible effects of endogeneity. The strata are added to the model through the inclusion of four dummy variables representing the strata. The comparison of the results for the regressions both with and without the propen- sity scores’ five strata groups included will be presented and evaluated. We use this approach in the absence of a suitable instrumental variable.

A multicolinearity analysis revealed reasonable correlations among the variables included in the model, and it was determined that there are missing values for the CMI of many hospitals. In the correlation matrix, the strongest correlation among the independent variables is between public ownership and system membership, yet the strength of this relationship (B < .400) does not present any violations of assump- tions needed for this analysis. Several regression equations were attempted to esti- mate the missing CMIs, but none of the correlations were strong enough to allow for

504 Medical Care Research and Review

Kazley, Ozcan / Hospitals With Electronic Medical Records 505

this approach. For the missing CMI values, the 2003 value of the hospital’s CMI was substituted for the absent 2004 value when available (n = 44). For the remaining missing values (n = 300), the average of the CMIs given for all hospitals with exactly the same number of beds was used. This imputation technique was selected in an effort to include the entire population of interest in this study and with the rationale that hospitals of the same size offer similar services and see patients of similar sever- ity based on the relationship between hospital volume and outcomes (Lynk, 2001). Sensitivity analyses excluding the hospitals with missing CMIs produced similar results. These hospitals were included with the estimated CMI in an effort to increase the external validity of the study.

Results

The sample in this study includes 4,605 nonfederal, acute care hospitals in the United States. Of these hospitals, 479 use fully automated EMR systems. We include only 2,969 hospitals in the quality analysis because of limitations in the reporting and validity of the quality measures as discussed previously. Of these 2,969 hospi- tals, 348 used EMRs in 2004. It is important to note that the sample sizes for each of the models vary based on the number of hospitals that reported for the measures with valid sample sizes (> 25 qualifying patients). The number of hospitals report- ing for the measures ranges from 1,701 to 2,969 and is reported in Table 1 with the range, mean, and standard deviation for each of the quality measures.

Using the continuous measure of quality, 10 models were regressed. The results of this analysis are in Table 2. Each of the models is significant at the p < .001 level. Of these 10 models, 5 show a significant relationship between hospital EMR use and quality based on the measures. There is a statistically significant relationship between hospital EMR use and the following measures: (heart attack) use of beta blocker at arrival to the hospital, use of aspirin at discharge from hospital, and use of beta blocker at discharge from hospitals; (congestive heart failure) assessment of left ventricular function; and (pneumonia) patients given antibiotic within 4 hours. The significant relationships between hospital EMR use of the heart attack and heart failure measures is positive. These results indicate that patients in hospitals with EMRs were more likely to receive the care as indicated by clinical guidelines for heart attack and heart failure patients. The significant relationship between hospital EMR use and the measure of patients with pneumonia receiving antibiotics within 4 hours is negative with a coefficient of –0.016, indicating that pneumonia patients in hospitals with EMRs are less likely than patients in hospitals without EMRs to get antibiotics within 4 hours.

To control for the possible selection bias already mentioned in this article, we use a propensity score matching approach. The frequency of hospitals with EMR use in

Ta bl

e 1

D es

cr ip

ti ve

S ta

ti st

ic s

of Q

ua lit

y P

ro ce

ss M

ea su

re s

A cu

te M

yo ca

rd ia

l I nf

ar ct

io n

C on

ge st

iv e

H ea

rt F

ai lu

re Pn

eu m

on ia

U se

o f

U se

o f

U se

o f

U se

o f

B et

a U

se o

f A C

E A

ss es

sm en

t U

se o

f Pa

tie nt

s G

iv en

A va

ila bi

lit y

A sp

ir in

a t

B et

a B

lo ck

er A

sp ir

in B

lo ck

er a

t fo

r LV

of L

V A

C E

A nt

ib io

tic s

W ith

in of

P ne

um on

ia A

ss es

sm en

t o f

A rr

iv al

at

A rr

iv al

at D

is ch

ar ge

D is

ch ar

ge D

ys fu

nc tio

n Fu

nc tio

n In

hi bi

to r

4 H

ou rs

V ac

ci ne

O xy

ge na

tio n

n 2,

17 0

1, 73

2 1,

70 1

1, 73

2 1,

78 7

2, 85

9 2,

05 3

2, 93

2 2,

79 5

2, 96

9 M

.9 4

.9 0

.9 1

.9 0

.7 6

.8 1

.7 6

.7 3

.4 6

.9 8

M in

. 0.

54 0.

17 0.

36 0.

17 0

0 0.

18 0.

04 0

0. 31

M ax

. 1

1 1

1 1

1 1

1 1

1 SD

.0 6

.1 0

.0 9

.1 0

.2 3

.1 7

.1 3

.1 3

.2 5

.0 5

N ot

e: A

C E

= an

gi ot

en si

n- co

nv er

tin g

en zy

m es

; L V

= le

ft v

en tr

ic ul

ar .

506

Kazley, Ozcan / Hospitals With Electronic Medical Records 507

each of the five strata of hospitals is presented in Table 3. We estimate the propensity score by considering the likelihood of EMR adoption for all 4,605 hospitals identi- fied as the sample, though only 2,969 have reported at least one quality measure. We use this approach to provide control for factors that may influence hospital EMR adoption and use. We have run each of the models with the dummy variables for the propensity score strata and present the results in Table 4. According to these analy- ses, the statistically significant relationships between hospital EMR use and each of the indicators remain. This indicates that there is no evidence of major selection effects in this model.

We also examine the treatment effects for each of the strata to consider the bene- fit of EMR use for each of the measures. We estimate the treatment effects by running a separate model for each strata. These results do not indicate a clear pattern of strong benefit for hospitals using EMRs in the area of quality. The results are presented in Table 5. Although there are significant relationships in this portion of the analysis, they do not indicate that any particular measure is significantly improved by EMR use across all strata. One area of potential benefit is in the area of congestive heart fail- ure; both measures for this condition indicate improvement in quality associated with EMR use for hospitals in Strata 4. When comparing the increased quality between the strata, we find two significant relationships in Strata 1, one that presents evidence of

Table 2 Multiple Regression Models

Variance Correlation With EMR Use Explained

Unstandardized R Condition and Measure Coefficient (SE) Square

Acute myocardial infarction Use of aspirin at arrival to hospital .005 (.004) .102 Use of beta blocker at arrival to hospital .0160** (.006) .087 Use of aspirin at discharge from hospital .010* (.006) .175 Use of beta blocker at discharge from hospital .012* (.007) .104 Use of angiotensin-converting enzymes for .021 (.017) .05

left ventricular systolic dysfunction Congestive heart failure

Assessment of left ventricular function .023*** (.009) .235 Use of angiotensin-converting enzymes inhibitor .010 (.009) .035

Pneumonia Patients given antibiotic within 4 hours –.016** (.006) .206 Availability of pneumococcal vaccine .019 (.015) .023 Assessment of oxygenation .002 (.003) .049

Note: EMR = electronic medical record. Each model controls for bed size/100, ownership, case mix index, system affiliation, and payer mix. *p < .10. **p < .05. ***p < .01.

508 Medical Care Research and Review

a decrease in quality associated with EMR use. We do not find any significant qual- ity differences among EMR users in Strata 2. We find two significant relationships with quality in Strata 3, five significant relationships with quality in Strata 4, and one significant relationship with quality in Strata 5. Based on the improved quality in Strata 4, it appears that EMRs may offer more benefits for the hospitals that are more likely to be adopters rather than those hospitals that are less likely to be adopters based on organizational characteristics. However, these significant effects do not carry over into Strata 5 or the other strata, suggesting that the selection effects are not consistently present for likely EMR adopters. Future research may wish to further explore the selection bias of hospital HIT use and outcomes.

Table 3 Strata and EMR Use

Strata 1 Strata 2 Strata 3 Strata 4 Strata 5 Total

No EMRS 907 906 849 764 700 4,126 EMRs 14 15 72 157 221 479 Total 921 921 921 921 921 4,605

Note: EMR = electronic medical record.

Table 4 Regression Models With Propensity Score Adjustments

Variance Correlation With EMR Use Explained

Unstandardized R Condition and Measure Coefficient (SE) Square

Acute myocardial infarction Use of aspirin at arrival to hospital .005 (.004) .108 Use of beta blocker at arrival to hospital .016** (.006) .097 Use of aspirin at discharge from hospital .010* (.006) .177 Use of beta blocker at discharge from hospital .012* (.007) .107 Use of angiotensin-converting enzymes for left .020 (.017) .053

ventricular systolic dysfunction Congestive heart failure

Assessment of left ventricular function .022** (.009) .24 Use of angiotensin-converting enzymes inhibitor .010 (.009) .038

Pneumonia Patients given antibiotic within 4 hours –.015** (.006) .210 Availability of pneumococcal vaccine .021 (.015) .026 Assessment of oxygenation .002 (.003) .054

Note: EMR = electronic medical record. Each model controls for bed size/100, ownership, case mix index, system affiliation, payer mix, and EMR use propensity. *p < .10. **p < .05. ***p < .01.

Kazley, Ozcan / Hospitals With Electronic Medical Records 509

Implications and Conclusions

According to this analysis, hospital EMR use has a limited effect on quality as measured through process adherence to three clinical conditions. For this reason, there is limited support for the hypothesis that hospitals with EMRs provide higher quality care in all areas. Rather, higher quality care associated with EMR use in hos- pitals tends to be sporadic and appears to have a positive effect in only four indica- tors of quality. Furthermore, we were surprised to see that EMR use is associated with decreased adherence to clinical guidelines in one measure. This is similar to the findings of Linder et al. (2007). We find no consistent evidence that EMR use can increase care in any particular condition over another.

Donabedian’s framework, which recognizes that an organization’s structure, processes, and outcomes as interrelated and influences to one another, offers a simple framework for examining the use of EMRs, but it may not be able to account for other unmeasurable factors, such as provider attitude toward HIT, patient factors, and the dynamics of working groups within hospitals. In addition, there may be sig- nificant differences among the EMR systems considered in this study that these data are not sensitive enough to detect. We are unaware of any more specific EMR mea- sures for a national sample. In this case, hospital EMR use, a structural feature, has

Table 5 Treatment Effects of EMR Use by Strata and Quality Measure

Treatment Effects

Condition and Measure Strata 1 Strata 2 Strata 3 Strata 4 Strata 5

Acute myocardial infarction Use of aspirin at arrival to hospital .012 –.014 .000 .014 .004 Use of beta blocker at arrival to hospital .031 –.030 –.008 .033** .015** Use of aspirin at discharge from hospital –.036 .041 .024 .015 .008 Use of beta blocker at discharge from hospital –.009 .012 .015 .039* .006 Use of angiotensin-converting enzymes for left .041 .097 –.055 .088* .017

ventricular systolic dysfunction Congestive heart failure

Assessment of left ventricular function .146* .022 –.027 .056*** .005 Use of angiotensin-converting enzymes inhibitor .067 .022 –.058** .052** .007

Pneumonia Patients given antibiotic within 4 hours –.150** .018 .000 .004 –.006 Availability of pneumococcal vaccine .040 .060 .089** –.003 .005 Assessment of oxygenation –.001 .006 –.001 .004 .003

Note: ΕΜR = electronic medical record. *p < .10. **p < .05. ***p < .01.

510 Medical Care Research and Review

offered little evidence to show that it can automate and standardize the processes of care and lead to consistently better quality performance in each clinical process mea- sure and condition.

The recent environment for health care organizations has focused attention on providing high-quality care while still containing costs. Since medical errors and poor coordination among clinicians can conflict with both of these goals, there is incentive for health care providers to prevent medical errors. One way to do so may be through the automation and standardization of the processes of care associated with a structural feature such as EMRs, but the best practices in HIT must first be indentified and used as benchmarks. Furthermore, the clear identification of consis- tent quality benefits related to EMR use should be a priority.

Since hospital EMR use is not necessarily associated with higher quality care, its implementation and practice should be further explored. By standardizing and automating the processes of care, EMR use may improve the outcomes of care, but it is not clear that this will happen in any particular diagnosis group or treatment type. If EMR use was determined to be associated with better compliance to clinical guidelines and decreased medical errors as determined by the process measures of the HQA, the practice should be more widespread. However, some have indicated that the cost of EMRs prevents many hospitals from using them (Burt & Hing, 2005; Cushman, 1997). Since the cost of EMR implementation is significant, it may not be cost effective to use such systems until clearer benefits are demonstrated in the area of quality.

Based on this study, we do not find consistent and conclusive evidence that hos- pital EMR use can improve the quality of care in all hospitals. For this reason, it may be suggested that the EMR systems and the individuals that use them should be examined more to identify any best practices or impediments to quality and progress. By identifying, addressing, and promoting the structures and processes that are nec- essary for high-quality performance, the expected benefits of EMRs may be realized.

Limitations

There are three specific limitations with this study. First, quality is a complex and multifaceted construct. It is likely that there are many unobservable elements of quality, such as patient satisfaction and long-term outcomes, which are not included in the measures used in this study. However, the measures have been validated and used in previous research and provide one of the largest information sources of hos- pital quality in the United States (Jha et al., 2005; Werner & Bradlow, 2006). As the number of measures continues to grow, the quality of the measure itself may also continue to grow.

Another limitation lies in the identification of hospitals with EMRs. While the HIMSS data provide a strong definition and a concrete identification of fully auto- mated EMR use, there is still variability in the features associated with each EMR system. Additionally, the level of use and amount of staff buy-in likely also vary

Kazley, Ozcan / Hospitals With Electronic Medical Records 511

based on the hospital. Because of this, EMR use may not be a homogeneous prac- tice; rather, hospitals that have EMRs may be using different types of systems for different types of functions.

Finally, EMR use is a structural feature of hospitals that is not randomly assigned. It is possible that hospitals that use EMRs were providing high-quality care before implementation of an EMR system. This study attempts to control for this selection effect by using logic, theory, methodology, statistical control, and previous research to design a conceptual model examining hospital EMR use and quality. It is, how- ever, possible that we have failed to identify an influencing factor of hospital EMR use in the propensity score adjustment approach.

References

Agency for Healthcare Research Quality. (2004). Introduction to health care quality. Retrieved March 13, 2006, from http://www.achp.org/page.asp?page_id=445

Anderson, G. F., Frogner, B. K., Johns, R. A., & Reinhardt, U. E. (2006). Health care spending and use of information technology in OECD countries. The United States is an outlier in both its health spend- ing and its use of health information technology. Health Affairs, 25(3), 819-831.

Anderson, G. K. (2004). Preventive medicine and the electronic health record. Medscape Public Health and Prevention, 2.

Asch, S. M., McGlynn, E. A., Hogan, M. M., Hayward, R. A., Shekelle, P., Rubenstein, L., et al. (2004). Comparison of quality of care for patients in the Veterans Health Administration and patients in a national sample. Annals of Internal Medicine, 141(12), 938-945.

Ash, J. S., & Bates, D. W. (2004). Factors and forces affecting EHR system adoption: Report on a 2004 ACMI discussion. Journal of the American Medical Informatics Association, 12(1), 8-12.

Brailer, D. G. (2005). Interoperability: The key to the future health care system interoperability will bind together a wide network of real-time, life-critical data that not only transform but become health care. Health Affairs WebExclusive, W5-19-W5-21.

Burt, C. W., & Hing, E. (2005). Use of computerized clinical support systems in medical settings: United States, 2001-2003. Advance Data From Vital and Health Statistics of the CDC, 353, 1-8.

Coyte, P. C., Young, W., & Croxford, R. (2000). Costs and outcomes associated with alternative discharge strategies following joint replacement surgery: Analysis of an observational study using a propensity score. Journal of Health Economics, 19(6), 907-929.

Cushman, R. (1997). Serious technology assessment for health care information technology. Journal of the American Medical Informatics Association, 4(4), 259-265.

Donabedian, A. (1980). Explorations in quality assessment and monitoring: Volume I the definition of quality and approaches to its assessment. Ann Arbor, MI: Health Administration Press.

Fonkych, K., & Taylor, R. (2005). The state and pattern of health information technology adoption. Santa Monica, CA: Rand Corporation.

Ginn, G. O., & Young, G. J. (1992). Organizational and environmental determinants of hospital strategy. Hospital and Health Services Administration, 37(3), 291-302.

Health Information Management Systems Society Foundation. (2004). HIMSS analytics user guide and data dictionary final release. Retrieved May 22, 2007, from http://www.himss.org/DorenfestInstitute/ default.aspx

Health Information Management Systems Society Foundation. (2005). HIMSS annual report of the US hos- pital IT market. Retrieved May 22, 2007, from http://www.himss.org/DorenfestInstitute/docs/Himss AnalyticsAnnualReport.pdf

512 Medical Care Research and Review

Hillestad, R., Bigelow, J., Bower, A., Girosi, F., Meili, R., Scoville, R., et al. (2005). Can electronic med- ical record systems transform health care? Potential health benefits, savings, and costs. Health Affairs, 24(5), 1103-1117.

Hospital Quality Alliance Overview Summary. (2005). Retrieved March 4, 2006, from www.cms.hhs.gov /HospitalQualityInits/

Institute of Medicine. (2001). Crossing the quality chasm: A new health system for the 21st century. Washington, DC: National Academy Press.

Jha, A. K., Li, Z., Orav, E. J., & Epstein, A. M. (2005). Care in U.S. hospitals–The hospital quality alliance program. The New England Journal of Medicine, 353(3), 265-274.

Kinn, J. W., O’Toole, M. F., Rowley, S. M., Marek, J. C., Bufalino, V. J., & Brown, S. A. (2001). Effectiveness of the electronic medical record in cholesterol management in patients with coronary artery disease (Virtual Lipid Clinic). The American Journal of Cardiology, 88(2), 163-165.

Kohn, L. T., Corrigan, J. M., & Donaldson, M. S. (1999). To err is human. Washington, DC: National Academy Press.

Leapfrog Group for Patient Safety. (2007). Retrieved May 22, 2007, from http://www.leapfroggroup.org/ for_hospitals/leapfrog_safety_practices/cpoe

Linder, J. A., Ma, J., Bates, D. W., Middleton, B., & Stafford, R. S. (2007). Electronic health record use and the quality of ambulatory care in the United States. Archives of Internal Medicine, 167(13), 1400-1405.

Lynk, W. J. (2001). One DRG, one price? The effect of patient condition on price variation within DRGs and across hospitals. International Journal of Health Care Finance and Economics, 1(2), 111-137.

Mick, S. S., & Wyttenbach, M. E. (2003). Advances in health care organization theory. San Francisco: Jossey-Bass.

Milcent, C. (2005). Hospital ownership, reimbursement systems and mortality rates. Health Economics, 14(11), 1151-1168.

Miller, R. H., & Sim, I. (2004). Physicians’ use of electronic medical records: Barriers and solutions. Health Affairs, 23(2), 116-126.

Paganini, J. M. (1993). Quality and efficiency of hospital care: The relationship between structure, process, and outcome. Washington, DC: Pan American Health Organization, The World Health Organization.

Poon, E. G., Jha, A. K., Christino, M., Honour, M., Fernandopulle, R., Middleton, B., et al. (2006). Assessing the level of healthcare information technology adoption in the United States: A snapshot. BMC Medical Informatics and Decision Making, 6(1), 1-9.

Rubin, D. B. (1997). Estimating causal effects from large data sets using propensity scores. Annals of Internal Medicine, 127(8 Pt. 2), 757-763.

Schmitt, K. F., & Wofford, A. D. (2002). Financial analysis projects clear returns from electronic medical records. Healthcare Financial Management, 56(1), 52-57.

Scott, W. R. (2003). Organizations: Rational, natural, and open systems (5th ed.). Upper Saddle River, NJ: Prentice Hall.

Shortliffe, E. H. (1999). The evolution of electronic medical records. Academic Medicine, 74(4), 414-419. Spencer, E., Swanson, T., Hueston, W. J., & Edberg, D. L. (1999). Tools to improve documentation of

smoking status, continuous quality improvement and electronic medical records. Archives of Family Medicine, 8(1), 18-22.

Starfield, B. (1992). Primary care: Concept, evaluation, and policy. New York: Oxford University Press. Stearns, S. C., Dalton, K., Holmes, G. M., & Seagrave, S. M. (2006). Using propensity stratification to

compare patient outcomes in hospital-based versus freestanding skilled-nursing facilities. Medical Care Research and Review, 5(5), 599-622.

Thompson, T. J., & Brailer, D. G. (2004). The decade of health information technology: Delivering consumer-centric and information-rich health care framework for strategic action. Washington, DC: U.S. Department of Health and Human Services.

Varon, J., & Marik, P. E. (2002). Clinical information system and the electronic medical record in the intensive care unit. Current Opinion in Critical Care, 8(6), 616-624.

Werner, R. M., & Bradlow, E. T. (2006). Relationship between Medicare’s hospital compare performance measures and mortality rates. Journal of the American Medical Association, 297(22), 2694-2702.

Zinn, J. S., Proenca, J., & Rosko, M. D. (1997). Organizational and environmental factors in a hospital alliance membership and contract management: A resource-dependence perspective. Hospital and Health Services Administration, 42(1), 67-86.

Kazley, Ozcan / Hospitals With Electronic Medical Records 513