Journal List > Healthc Inform Res > v.22(2) > 1075782

Choi, Kuo, Goodwin, and Lee: Association of EMR Adoption with Minority Health Care Outcome Disparities in US Hospitals

Abstract

Objectives

Disparities in healthcare among minority groups can result in disparate treatments for similar severities of symptoms, unequal access to medical care, and a wide deviation in health outcomes. Such racial disparities may be reduced via use of an Electronic Medical Record (EMR) system. However, there has been little research investigating the impact of EMR systems on the disparities in health outcomes among minority groups.

Methods

This study examined the impact of EMR systems on the following four outcomes of black patients: length of stay, inpatient mortality rate, 30-day mortality rate, and 30-day readmission rate, using patient and hospital data from the Medicare Provider Analysis and Review and the Healthcare Information and Management Systems Society between 2000 and 2007. The difference-in-difference research method was employed with a generalized linear model to examine the association of EMR adoption on health outcomes for minority patients while controlling for patient and hospital characteristics.

Results

We examined the association between EMR adoption and the outcomes of minority patients, specifically black patients. However, after controlling for patient and hospital characteristics we could not find any significant changes in the four health outcomes of minority patients before and after EMR implementation.

Conclusions

EMR systems have been reported to support better coordinated care, thus encouraging appropriate treatment for minority patients by removing potential sources of bias from providers. Also, EMR systems may improve the quality of care provided to patients via increased responsiveness to care processes that are required to be more time-sensitive and through improved communication. However, we did not find any significant benefit for minority groups after EMR adoption.

I. Introduction

The practice of medical care in the United States varies in relation to patient race and ethnicity [1]. There a multiple roots of these healthcare disparities, ranging from differences in physician and patient perceptions and communication during clinical encounters, to differences in the ways that different racial and ethnic groups relate to the institutions and systems that finance and provide care [2]. How information is presented, managed, and exchanged contributes to healthcare disparities. Therefore, the use of Electronic Medical Records (EMRs) could potentially reduce disparities.
The purpose of EMR systems is to retain, organize, and communicate medical information in a uniform manner. EMRs have the potential to make healthcare providers' decisions more consistent and objective, while ensuring that best practices are pursued [3]. EMRs provide evidence-based decision support to the healthcare provider at the point of service and can lead to timely and informed medical decision making. Such standardization has the potential to provide benefits to members of racial and ethnic minority populations that have disproportionately experienced sub-standard care.
While a growing number of studies have examined the use of EMRs to improve the efficiency of medical care delivery, surprisingly few studies have explored the impact of EMRs on reducing health disparities between racial and ethnic groups. Thus, the aim of this study was to investigate the differential impact of EMR adoption on the outcomes of care among members of minority groups. Our hypothesis was that the adoption of an EMR system by a hospital or hospital system would provide benefits to members of disadvantaged minority groups who had historically experienced lower-quality medical outcomes compared to non-Hispanic white patients.
The hospital EMR adoption rate increased from 9.4% in 2008 to 59.4% in 2013 [4]. The Healthcare Information and Management Systems Society (HIMSS) has tracked changes during the implementation of EMR systems. Using their collected information, we investigated whether the adoption of EMR systems reduced racial and ethnic disparities in health outcome measurements by comparing EMR-adopting hospitals to hospitals that have yet to implement an EMR system.
These care quality measures were calculated using linked Medicare claims data for patients receiving care in each hospital for two years before and after the adoption of an EMR system, compared to patient data from hospitals that had not adopted an EMR over the same time period. Thus, our main methodological approach was to use a difference-indifference design to examine changes in outcome disparities. Analyses were adjusted for the characteristics of patients and hospitals using data from the Medicare Beneficiary Summary File and the HIMSS data.
Poor access to care among African Americans and Hispanics is reflected in substantially lower rates of employment-based insurance coverage, lack of a regular source of medical care, and ethnic/cultural mismatches with care providers. These factors may also reduce the continuity of care among these disadvantaged populations relative to non-Hispanic whites and may diminish the engagement and trust between patient and provider, reducing the quality of medical decision-making. The Institute of Medicine found racial disparities across a wide range of disease areas, clinical services, and clinical settings, even after clinical factors, including age, disease progression, comorbid conditions, and severity of illness, were taken into account [2]. They concluded that prejudices, biases, and negative racial stereotypes may be a potential source of the disparity and proposed "evidence based guidelines" among the solutions. In addition, there are abundant research findings that racial and ethnic disparities in health care can be explained in part by minorities' disproportionate receipt of care from certain institutions where all patients, irrespective of race, tend to experience worse outcomes [5]. There is some concern that regulatory policies aimed to reduce poor outcomes may disproportionately impact minority-serving institutions, perhaps increasing racial disparities [6].
Previous studies have demonstrated disparities in both the outcomes and process of care. Compared with white patients with diabetes, African Americans with diabetes were less likely to have received the recommended processes of care metrics, including glycated hemoglobin (HbA1c) and lipid measurements [78]. Another study [9] found that compared with white patients, black and Hispanic patients experienced substantially longer wait times from the start of dialysis to being added to the kidney transplant waiting list.
Furthermore, previous health services research suggests that there are racial disparities in health outcomes. Some studies have found that black patients experienced a higher 30-day readmission rate (24.8%) compared to white patients (22.6%) according to Medicare beneficiary data [510]. Other studies found that black patients admitted with pneumonia had a longer length of stay than white patients (incidence rate ratio: 1.19). Moreover, black patients had a higher inpatient mortality rate than white patients after cardiovascular procedures and cancer resections; the odds ratios of inpatient mortality for black patients relative to white patients are 1.21 (carotid endarterectomy), 1.21 (aortic valve replacement), 1.16 (coronary artery bypass graft), 1.32 (abdominal aortic aneurysm repair), 1.08 (resection for lung cancer), 1.32 (cystectomy of the bladder), 1.57 (esophagectomy), and 1.27 (pancreatic resection).
Despite the projection that the minority population is expected to rise to 56% of the total US population, prior research on racial and ethnic disparities in healthcare has failed to identify methods of reducing these disparities [11]. This study was an expansion of the work by Lee et al. [12], which investigated the relationship between EMR adoption and health outcomes of the general population; however, they did not focus on the outcome disparities among minorities. Our study investigated the association between EMR adoption and the health outcomes of minorities, including length of stay, inpatient mortality rate, 30-day mortality rate, and 30-day readmission rate. This study was approved by the Institutional Review Board (IRB) at the University of Texas Medical Branch (IRB No. 09-054).

II. Methods

1. Data Sources

In this study, we employed four primary data sources: 1) HIMSS data, 2) Provider of Service (POS) files, 3) Medicare enrollment files and 4) the 5% Medicare Provider Analysis and Review (MEDPAR) data from 2000 to 2007. The HIMSS data was sampled from the American Hospital Association (AHA) annual survey of hospitals, which provides information on health IT applications for more than 3,000 US hospitals. Medicare enrollment files provide patient socio-demographic characteristics, such as sex, age, insurance and race. The POS file provides information on providers, including characteristics of institutional providers. The MEDPAR file contains claims data on Medicare beneficiaries hospitalized in Medicare-certified inpatient hospitals and skilled nursing facilities (SNF).

2. Establishment of the Study Cohort

This study identified more than 2,600 unique acute-care hospitals with more than 100 beds between 2000 and 2007. We excluded hospitals if they were not consecutively observed over the entire 8-year study period, if they did not have a Medicare provider number, or if they adopted any components of a basic EMR before 2002. After excluding hospitals that did not meet the eligibility criteria, 708 acute-care hospitals were finally included in our analyses.
Using HIMSS data, we identified the year of adoption of a basic EMR, defined as a computerized patient record that is supported by a clinical data repository and has clinical decision-support capabilities [1314]. The four outcomes were length of stay, inpatient mortality rate, 30-day mortality rate, and 30-day readmission rate. The sample included patients who were older than 65 years, had not been enrolled in HMOs, and had both Medicare Parts A and B for the entire 12 months before admission.

1) Outcome variables

(1) Length of stay

This sample only included stays shorter than 365 days in which the patient was discharged alive. To correct for data skewness, we excluded admissions in which the length of stay was more than 3 standard deviations from the mean. The final sample size for the length of stay was 360,105.

(2) Inpatient mortality and 30-day mortality rate

This sample included hospital stays no longer than 365 days for all patients, including patients who died in the hospital or were discharged alive. The inpatient mortality rate was calculated by the number of patients who died during hospital stays over the total number of admitted patients. The 30-day mortality rate was defined as the number of patients who died within 30 days of admission over the total number of patients admitted. The final sample size for mortality was 403,566.

(3) 30-day readmission rate

The accumulation of claims from a beneficiary's date of admission to an inpatient hospital to the date of discharge represents one stay in the MEDPAR file. We only retained data for hospitalizations lasting less than 365 days in which the patient was discharged alive. The 30-day readmission rate was defined as the number of patients discharged and readmitted to any acute hospital within 30 days of discharge over the total number of patients discharged alive. For the multiple readmission rate, one admission per year was randomly selected. The final sample size for 30-day readmission was 237,081.

2) Independent variables

Medicare enrollment files were utilized to obtain patients' demographic information, including age, sex, and race. Information on discharge diagnosis related group (DRG) was obtained from the Center for Medicare & Medicaid Services and MEDPAR files. To construct a variable indicating comorbid conditions measured by the Elixhauser index, we utilized inpatient and physician claims from MEDPAR, Outpatient Statistical Analysis files, and carrier files measured for the 12 months prior to initial hospitalization [15]. Structural characteristics of hospitals were obtained from the POS files. We measured teaching status as a binary variable defined as either none or any. We measured hospital size based on bed size and categorized hospitals into quartiles. Ownership status of hospitals was categorized as not-for-profit, profit, or government.

3. Statistical Analysis

To examine the association of EMR adoption on health outcomes for minority patients, we used a difference-indifference research design based on observational research comparing outcomes two years before EMR adoption and two years after EMR adoption. The difference-in-difference is a statistical technique used to estimate treatment effects by comparing before and after treatment differences in outcome between a control and a treatment group [16]. In our study, the treatment group was composed of hospitals adopting EMR systems between 2002 and 2005, and the control group contained hospitals who did not adopt an EMR system during the same period.
During the study period, 425 hospitals adopted an EMR (Table 1): 159 hospitals in 2002, 77 in 2003, 46 in 2004, and 143 in 2005. Hospitals not adopting EMR systems were randomly assigned to match the distribution for the year of adoption for hospitals adopting EMR systems (Table 2): 106 in 2002, 51 in 2003, 30 in 2004, and 96 in 2005. In other words, because 159 hospitals that had adopted EMR systems accounted for 37% of the total 425 hospitals that had adopted EMR systems in 2002, we randomly matched 106 out of the 283 total (37%) non-EMR-adoption hospitals for this particular year in the non-EMR-adoption group.
We considered two groups of hospitals: hospitals adopting EMR systems (treatment group) and hospitals not adopting EMR systems (control group). In the treatment group, we selected hospitals that had adopted EMR systems in their third year to compare two years before and two years after EMR system adoption. Then, we multiplied 1) EMR group, 2) year of EMR adoption, and 3) black patients to determine whether EMR adoption could improve outcomes for black patients. Our regression model is expressed as
Yit = α + β1EMRit + β2after_EMRit + β3Blacki + β4EMRit * after_EMRit + β5EMRit * after_EMRit * Black + β6Pat_ Charac + β7Hosp_Charac + Time + εit,
where i represents hospitals, t represents time of year, Y represents outcome; EMR represents EMR-adoption hospitals coded as 1 if EMR was adopted and 0 if not; After_EMR represents years after EMR adoption; Black represents patients of African-American descent; Pat_Charac represents patient characteristics, including sex, age, and race (either white or black), and comorbidity; and Hosp_Charac represents hospital characteristics, including teaching status, bed size, and ownership.
The key independent variables are interaction terms of EMR, after_EMR, and Black. If β4 or the coefficient of EMR and after_EMR (or odds) was significantly negative (or less than 1), we could confirm that EMR adoption was negatively associated with patient outcomes. Alternatively, if β5 or the coefficient of EMR, after EMR, and Black (or odds) was significantly negative (or less than 1), we could confirm that EMR adoption could improve patient outcomes for black patients. We used generalized linear models (GLMs) to investigate the link between EMR adoption and reduction in disparities in four health outcomes for black patients (i.e., length of stay, inpatient mortality rate, 30-day mortality rate, and 30-day readmission rate) after taking clinical (i.e., comorbidities) and demographic (i.e., age, sex, and race) characteristics of patients, structural (i.e., ownership type, teaching status, and bed size) characteristics of hospitals, and time (year) into account.
We used a GLM with binomial distribution and logit link for the models for (1) inpatient mortality rate, (2) 30-day mortality rate, and (3) 30-day readmission, and a GLM with normal distribution and log link for (4) length of stay. All statistical analyses were performed using STATA 14 (Stata-Corp, College Station, TX, USA).

III. Results

The study cohort included patients admitted to 425 hospitals adopting EMR systems and 283 hospitals not adopting EMR systems over the study period as indicated in Tables 1 and 2. Table 3 presents patient, disease, and hospital characteristics. A larger proportion of patients were female (59%), aged between 76–85 (43%), and white (90.7%).
White patients accounted for 90.7% of the sample. The mean of the comorbidity index was 2.9. Slightly less than half of the hospitals were teaching hospitals, while not-for-profit hospital ownership accounted for the majority (78.1%). The number of beds in each type of hospital was equally distributed. Table 3 also shows descriptive statistics for the four outcomes: the mean length of stay was 5.62 days, inpatient mortality 4.4%, 30-day mortality 13.5%, and 30-day readmission 16.3%.
Table 4 presents the difference-in-difference GLM regression results of length of stay and inpatient mortality rate. In the first column (length of stay), the variables positively associated with length of stay are black race, older age groups (65–75 and 86 and over), higher comorbidity, and greater number of beds. As seen in the second column (inpatient mortality rate), we found that patients who were black males, had a higher comorbidity rate, and belonged to an older age group (76–85 and 86 and over) had an increased likelihood of experiencing inpatient mortality. In addition, government hospital ownership and number of beds from 261–362 were other factors linked to higher mortality rates.
Table 5 presents GLM regression results of 30-day readmission and 30-day mortality after discharge. As seen in the first column (30-day readmission rate), we found that patients who were black, males, had a higher comorbidity rate, and belonged to an older age group (76–85 and 86 and over) were more likely to have a higher 30-day readmission rate. Also, hospitals of not-for-profit ownership were more likely to have a higher 30-day readmission rate. As seen in the second column (30-day readmission rate), patient characteristics associated with higher 30-day mortality rates were male, older age, and higher comorbidity. The hospital characteristic associated with higher 30-day mortality rates was not-for-profit ownership.
However, we could not find any significant effect of EMR adoption, the years after EMR adoption, the interaction terms of the EMR adopted group, after years of EMR adoption and race in all four outcome regressions.

IV. Discussion

EMR systems have been reported to support better coordinated care [3121718], thus encouraging appropriate treatment for minority patients by removing potential sources of bias from the providers [19]. Health IT systems could improve clinical care by detecting important clinical and sociodemographic risk factors for various conditions particularly relevant to minority patients. EMR systems may improve the quality of care provided to patients via increased responsiveness to care processes that are required to be more timesensitive and through improved communication [20].
While a growing number of studies have examined the association between EMRs and health outcomes, surprisingly few studies have explored the impact of EMRs on outcome disparities among minorities. Thus, to fill this gap in the current literature, this study investigated the impact of EMR system adoption on disparities in health outcomes of black patients. However, we did not observe any significant changes in health outcomes before and after EMR system implementation. Moreover, we did not find evidence that EMR system implementation was associated with outcomes of black patients.
A possible reason for the lack of any significant association between EMR and healthcare outcomes of black patients could be the short-term effect of EMR adoption. We compared the outcomes two years before and after EMR adoption. Thus, no meaningful changes in health outcomes of black patients were observed after EMR implementation, which might reflect a short-term effect of EMR implementation on patient outcomes. It has been reported that it may take several years for hospitals adopting EMR systems to fully capitalize on the clinical benefits after EMR implementation [18]. In particular, benefits from EMR on health outcomes may take longer than for process of care for minority populations. For example, Lee [21] found that greater health IT investment leads to shorter waiting times, and the waiting time reduction was greater for non-white than for white patients. He concluded that minority populations could benefit from health IT in terms of process of care [21].
Also, eliminating healthcare disparities among minority groups may remain limited, even though the evidence of healthcare disparity among minority groups is substantial [22]. Acevedo-Garcia et al. [22] argued that an effort to eliminate disparities among specific groups is currently underway but will take more time. For example, this effort has focused on training providers to offer appropriate services through the use of EMR systems and to improve coordination of care [22].
There were some limitations in this study. First, for this work, EMR was defined as computerized patient record systems (CPRS) supported by clinical decision-support (CDS) capabilities and a clinical data repository (CDR) [1314]. Thus, the definition of EMR adoption was less restraining than those used in previous studies [232425]. For example, Miller and Tucker [23] employed HIMSS data, but defined EMR adoption as four complete components of adoption including CDR, computerized physician order entry (CPOE), CDSS, and digitized physician documentation. Also, prior studies have used a more limited definition of EMR [2425] that included IT systems, such as electronic medication administration records (eMAR) and nursing documentation. This study used a more expansive definition of EMR systems because HIMSS data was incompatible with health IT systems, such as physician documents.
Second, we did not have access to in-depth data indicating the level of care regarding provider proficiency and reluctance related to the use of EMR systems in hospitals. Hence, if care providers are resistant to using an EMR system or are not IT-savvy for various reasons after EMR adoption, our data may be underestimated.
Lastly, there may have been unobserved confounding factors that might have impacted our findings. For example, organizational and management behavior may be correlated with health IT adoption. Although we controlled for structural characteristics of the hospital and patient clinical and demographic characteristics, EMR adoption behavior may vary in ways we could not observe in the data set, leading to bias in our results.

Figures and Tables

Table 1

EMR-adoption group (treatment group)

hir-22-101-i001

EMR: Electronic Medical Record.

Table 2

Non-EMR-adoption group (control group)

hir-22-101-i002

EMR: Electronic Medical Record.

Table 3

Descriptive statistics for patients and hospitals

hir-22-101-i003
Table 4

Association between EMR and outcomes using generalized linear model regression: length of stay (360,105) and inpatient mortality (403,566)

hir-22-101-i004

EMR: Electronic Medical Record, SE: standard error, CI: confidence interval.

*p < 0.1, **p < 0.05, ***p < 0.001.

Table 5

Association between EMR and outcomes using generalized linear model regression: 30-day mortality (403,566) and 30-day readmission (237,081)

hir-22-101-i005

EMR: Electronic Medical Record, SE: standard error, CI: confidence interval.

*p < 0.1, **p < 0.05, ***p < 0.001.

Acknowledgments

This research was supported by the Hallym University Research Fund (No. HRF-201407-008) and by a grant (No. K05-CA134923) from the National Cancer Institute.

Notes

Conflict of Interest No potential conflict of interest relevant to this article was reported.

References

1. Rubio M, Williams DR. The social dimension of race. In : Beech BM, Goodman M, editors. Race and research: perspectives on minority participation in health studies. Washington (DC): American Public Health Association;2004. p. 1–25.
2. Smedley BD, Stith AY, Nelson AR. Unequal treatment: confronting racial and ethnic disparities in health care. Washington (DC): National Academies Press;2003.
3. Lee J, Dowd B. Effect of health information technology expenditure on patient level cost. Healthc Inform Res. 2013; 19(3):215–221.
crossref
4. Charles D, Gabriel M, Furukawa MF. Adoption of electronic health record systems among U.S. non-federal acute care hospitals: 2008-2013 [Internet]. Washington (DC): Office of the National Coordinator for Health Information Technology;2014. cited 2016 Mar 30. Available from: https://www.healthit.gov/sites/default/files/oncdatabrief16.pdf.
5. Joynt KE, Orav EJ, Jha AK. Thirty-day readmission rates for Medicare beneficiaries by race and site of care. JAMA. 2011; 305(7):675–681.
crossref
6. Bhalla R, Kalkut G. Could Medicare readmission policy exacerbate health care system inequity? Ann Intern Med. 2010; 152(2):114–117.
crossref
7. Virnig BA, Lurie N, Huang Z, Musgrave D, McBean AM, Dowd B. Racial variation in quality of care among Medicare+Choice enrollees. Health Aff (Millwood). 2002; 21(6):224–230.
crossref
8. Agency for Healthcare Research and Quality. National healthcare disparities report 2011 [Internet]. Rockville (MD): Agency for Healthcare Research and Quality;2012. cited at 2016 Mar 30. Available from: http://www.ahrq.gov/research/findings/nhqrdr/nhdr11/nhdr11.pdf.
9. Arce CM, Goldstein BA, Mitani AA, Lenihan CR, Winkelmayer WC. Differences in access to kidney transplantation between Hispanic and non-Hispanic whites by geographic location in the United States. Clin J Am Soc Nephrol. 2013; 8(12):2149–2157.
crossref
10. Jencks SF, Williams MV, Coleman EA. Rehospitalizations among patients in the Medicare fee-for-service program. N Engl J Med. 2009; 360(14):1418–1428.
crossref
11. Berges IM, Kuo YF, Ostir GV, Granger CV, Graham JE, Ottenbacher KJ. Gender and ethnic differences in rehabilitation outcomes following hip replacement surgery. Am J Phys Med Rehabil. 2008; 87(7):567–572.
12. Lee J, Kuo YF, Goodwin JS. The effect of electronic medical record adoption on outcomes in US hospitals. BMC Health Serv Res. 2013; 13:39.
crossref
13. McCullough JS, Casey M, Moscovice I, Prasad S. The effect of health information technology on quality in US hospitals. Health Aff (Millwood). 2010; 29(4):647–654.
crossref
14. Fonkych K, Taylor R. The state and pattern of health information technology adoption. Santa Monica (CA): Rand Corporation;2005.
15. Elixhauser A, Steiner C, Harris DR, Coffey RM. Comorbidity measures for use with administrative data. Med Care. 1998; 36(1):8–27.
crossref
16. Bertrand M, Duflo E, Mullainathan S. How much should we trust differences-in-differences estimates? Q J Econ. 2004; 119(1):249–275.
crossref
17. Drye EE, Normand SL, Wang Y, Ross JS, Schreiner GC, Han L, et al. Comparison of hospital risk-standardized mortality rates calculated by using in-hospital and 30-day models: an observational study with implications for hospital profiling. Ann Intern Med. 2012; 156(1 Pt 1):19–26.
crossref
18. Lee J, McCullough JS, Town RJ. The impact of health information technology on hospital productivity. RAND J Econ. 2013; 44(3):545–568.
crossref
19. DRG expert: a comprehensive guidebook to the DRG classification system. Reston (VA): Ingenix;2007.
20. Longhurst CA, Parast L, Sandborg CI, Widen E, Sullivan J, Hahn JS, et al. Decrease in hospital-wide mortality rate after implementation of a commercially sold computerized physician order entry system. Pediatrics. 2010; 126(1):14–21.
crossref
21. Lee J. The impact of health information technology on disparity of process of care. Int J Equity Health. 2015; 14:34.
crossref
22. Acevedo-Garcia D, Osypuk TL, McArdle N, Williams DR. Toward a policy-relevant analysis of geographic and racial/ethnic disparities in child health. Health Aff (Millwood). 2008; 27(2):321–333.
crossref
23. Miller AR, Tucker C. Can healthcare information technology save babies? J Polit Econ. 2011; 119(2):289–324.
24. Rosko MD. Performance of US teaching hospitals: a panel analysis of cost inefficiency. Health Care Manag Sci. 2004; 7(1):7–16.
crossref
25. Koenig L, Dobson A, Book R, Chen Y. Comparing hospital costs: adjusting for differences in teaching status and other hospital characteristics. Falls Church (VA): The Lewin Group;2005.
TOOLS
Similar articles