EMR Adoption by Alabama Physicians Reduces the Proportion of Adverse


By Venkatesh Ekambaram

Northwestern University

EMR Adoption by Alabama Physicians Reduces the Proportion of Adverse

By Venkatesh Ekambaram

The Institute of Medicine has encouraged the use of HIT for the purpose of
delivering “safe, efficient, patient-centered, and timely” healthcare (Follen et al,
2007). The consensus opinion among medical informaticians is that “HIT is
engineered to promote improved quality and efficiency of care, and reduce
medical errors” (Belletti et al, 2010). The goals of the American Recovery and
Reinvestment Act (ARRA) of 2009, with its HITECH provisions, and the
Affordable Care Act (ACA) of 2010 are the same. The latter piece of legislation
has an additional goal of providing a minimum level of healthcare to all.

A national survey conducted in 2005 demonstrated that less than 1 in 3 practices
were utilizing EMRs at that time and attributed the lower rates of adoption to
insufficient resources, particularly among small practices (Simon et al, 2005).
However, these have been offset more recently by Meaningful Use incentives
since “adoption of these systems may depend on powerful financial incentives,
such as pay for performance, that will reward organizations for using such
systems and motivate policies that provide these systems to small provider
groups at an affordable price” (Follen et al, 2007).

Universal implementation of these systems ensures that standards are being
met, thereby serving the public good, and may be achievable if localized data,
along with incentives, are made available to healthcare providers. Analysis of this
data should convey sufficient evidence of the capabilities of this software. Added
financial benefits can be expected, in turn, by clinicians who perform well across
metrics such as Clinical Quality Measures (CQMs). Thus, providers become
“meaningful users” of these EMRs which will track their performances and also
enhance the quality of care they deliver. But the case that HIT can reduce
medical errors has not yet been made. Our research seeks to demonstrate that
increased EMR adoption leads to a corresponding reduction in medical errors.


In order to create a sample of sufficient size, in this simulation, data from 30 EMR
systems were acquired, based on the ability of each system to determine the
number of adverse events and total number of patient encounters. System data
after a January 1, 2013 “go-live” date were gathered, as were data derived from
the incorporation of paper records into patients’ records, the latter of which
served as the foundations of each patient’s electronic medical records. AEs
included in the count include all 29 of those listed in the National Quality Forum’s
2011 list of serious reportable events, which the selected EMR systems were all
capable of tracking (National Quality Forum, 2011).

Using data from each practice, the proportions of AEs out of the total number of
annual patient encounters before and after the EMR adoption date were
tabulated (see Appendix: Figure 1) and used to test the hypothesis, using a
paired samples t-test to compare means, that EMR adoption by a practice
reduces its proportion of adverse events. Practice data was analyzed using
SPSS software, at the 95% confidence level.


The results of the statistical analyses of these data are shown in Appendix:
Figure 2, using the convention of 5-step hypotheses testing.


We demonstrated that EMR systems can reduce the number of medical errors
such as prescription dosage miscalculations. The choice to pool EMR system
data from physician practices in North Alabama, which is primarily a rural setting
with only two metropolitan areas, was made deliberately in order to convince
local physicians. Many of them have yet to adopt EMR systems, but we assert
that the reduction in medical errors serves their best interests from ethical,
medico-legal, and financial perspectives. Future research could continue to
examine other facts which impact EMR adoption and reduction in medical errors,
such as clinician comfort with EMR use, familiarity with the software, and
reinforcement with thorough medical staff training by HIT vendors (Follen et al,

Literature suggests that time spent training clinicians has an influence on HIT
efficacy, such as making efforts to engage in “persistent communication and
education regarding the importance of the integration of the EMR” (Follen et al,
2007). One recommended approach would be to examine the “organizational
characteristics (that) are linked to financial and professional barriers to EMR
adoption that have been identified in other qualitative studies! (since) strategies
for overcoming those barriers may (have) now (been) developed in a more
targeted fashion” (Simon et al, 2005). In another year’s time, different results
would be seen and could be compared to previous years’ data. In the interim,
perhaps “financial and technical assistance aimed specifically at smaller groups
is particularly warranted” (Simon et al, 2005).

The healthcare informatics community remains eager to “evaluate the real-world
application of these systems in terms of accurately gauging their relative
effectiveness in promoting positive outcomes (because) more widespread efforts
aimed at increasing utilization and adoption rates are warranted” (Follen et al,


FIGURE 1: Proportion of AEs Recorded in 2012 & 2013

FIGURE 2: Hypothesis Testing using SPSS Table of Paired Samples t-Test

Establishment of Hypotheses:
H0: The mean difference between these proportions is 0, i.e. MuDiff= 0
H1: The mean difference between these proportions is less than 0, i.e. MuDiff<0
Alpha= 0.05

Selection of the Test Statistic:
The paired t-test is appropriate, considering that values were compiled pre- and
post- EMR implementation. All assumptions for the t-test were met.

Generation of the Decision Rule:
Degrees of Freedom= n-1= 29 (in both samples)
Confidence Level= 95% (by convention)
Analysis of Data performed using SPSS software
Therefore, Reject H0 if p<0.001; Do not reject H0 if p>=0.001

Computation of the Test Statistic Value:

A paired sample t-test was used to determine that the proportion of adverse
events prior to EMR implementation (Mean-.004647, SD=.0028463) was greater
than post implementation (Mean-.002600, SD=.0017996; t=5.939, p<.001).


Belletti, D., Zacker, C., & Mullins, C. D. (2010). Perspectives on electronic
medical records adoption: electronic medical records (EMR) in outcomes
research. Patient related outcome measures, 1, 29.

D’Agostino, R.B., Sullivan, L.M., & Beiser, A.S. (2006). Introductory applied
biostatistics. Belmont, CA: Brooks/Cole, Cengage Learning.

Follen, M., Castaneda, R., Mikelson, M., Johnson, D., Wilson, A., & Higuchi, K.
(2007). Implementing health information technology to improve the
process of health care delivery: a case study. Disease
Management, 10(4), 208-215.

National Quality Forum (NQF), Serious Reportable Events In Health- care—2011
Update: A Consensus Report, Washington, DC: NQF; 2011.

Simon, J. S., Rundall, T. G., & Shortell, S. M. (2005). Drivers of electronic
medical record adoption among medical groups. Joint Commission
Journal on Quality and Patient Safety, 31(11), 631-639.