Beruflich Dokumente
Kultur Dokumente
Copyright ©
© Taylor &
& Francis Group, LLC
ISSN: 1753-8157 print / 1753-8165 online
DOI: 10.3109/17538157.2014.965304
Background: The relationship between clinical decision support systems (CDSS) and
quality is a relatively new, and in light of the new health information technology (HIT)
legislation, policy-relevant area. Moreover, very few studies exist examining the link
between HIT and healthcare disparities. The purpose of this article is to examine the
association between CDSS and the treatment of pneumonia care within high-minority
(29.1% non-White, non-Hispanic) and low-minority (529.1%) Zip Code Tabulation
Areas (ZCTAs). Research design: This study employed a cross-sectional design and used
2009 data from the American Hospital Association, the Centers for Medicare and
Medicaid Services and the Research Triangle Institute. Adjusted analysis controlled for
a hospital’s propensity to use CDSS. Results: In the unadjusted analysis, hospitals in
high-minority ZCTAs had lower pneumonia quality composite scores than their low-
minority counterparts. When adjusting for other hospital and ZCTA-level variables, we
found that CDSS use had stronger positive associations with quality in high-minority
hospitals. Conclusions: Results support policy directives may support higher quality
improvements by focusing CDSS adoption in high-minority hospitals.
INTRODUCTION
In 2008, pneumonia, together with influenza, were the eighth leading causes of
death in the United States (1,2). Pneumonia is also the number one infectious
disease killer in the United States, with approximately 52 000 deaths or 2.2%
of total deaths (1,3). Associated clinical symptoms typically include fever,
cough, sputum production, pleurisy and dyspnea— caused by fluid in the lungs
hypothesize that hospitals in high-minority ZCTAs that use CDSSs will not
significantly differ in pneumonia process quality from those in low-minority
ZCTAs.
METHODS
Conceptual framework
Two theoretical frameworks guide this study: Andersen’s Behavioral Health
Model and Donabedian’s Quality Model. Andersen’s Behavioral Health Model
(15) suggests that characteristics within the external environment affect care
processes and outcomes. Andersen’s model, first developed in the 1960s, was
designed to help understand why families use health services and to promote
equitable access to those services. This model suggests that families use health
services as a function of their predisposition to use services: either enabling or
impeding. Over the years, this model has been updated to include factors
within one’s environment. Within Anderson’s Model, the principal independent
variable is hospital location within a high-minority ZCTA. Additional proxies
for the hospital’s external environment include urban/rural location, percent
65+ populations living in poverty, percent 65+ populations and percent without
a high school diploma within the ZCTA. Due to data limitations, we could
not control for patient-level characteristics and behaviors. Donabedian’s
quality model (Structure – Process – Outcome) (32) was used as a framework
to examine the association between a hospital structural element, CDSS
implementation and process quality indicators. Donabedian’s model posits that
there are three classifications, which inferences can be drawn for quality
assessment: structure, process and outcome. Structure refers to the attributes
of a healthcare setting—facilities , human resources and financial elements.
Process refers to the actions of what is actually happening—did the physician
render evidence-based medicine? Finally, outcome refers to the effects of care—
did the patient’s health status improve? The parts of this model that are
important are the relationships between structure – process – and outcome. In
other words, certain structural elements increase the likelihood of processes
being carried out, which then influence health outcomes. In this study, other
hospital-level variables within the Donabedian Model include size, ownership,
JCAHO accreditation, teaching hospital status and system membership/
management structure. It is necessary to use both the Donabedian and
Andersen models in this study to control for not only hospital-level structure
and process variables (Donabedian), but also the environment in which the
hospital operates (Andersen). Specifically, using these two models in conjunc-
tion, we can determine associations between the hospital structural element of
CDSS and processes (Donabedian), within the context of the ZCTA environ-
ment (Anderson).
Variables
Dependent variable
The pneumonia quality measure used was a calculated composite score, based
on six pneumonia process indicators from the Hospital Compare dataset for
FY2009. Landrum et al. (35) assert that within cross-sectional studies,
composite scoring of process measures better summarize quality at a provider
level. These scores were calculated using the opportunity method developed by
the Hospital Core Performance Measurement Project for the Rhode Island
Public Reporting Program for Health Care Services in 1998 (36). This method
controls for individual weighting, missing data and differences in case
volumes. It is calculated by dividing the total number of instances of rendered
evidence-based care by the total number of opportunities for those treatments.
Values for the composite measure ranged from 4.00% to 100.00% (mean
92.50%, SD ¼ 6.60%). Individual quality indicators used to create the compos-
ite score included: pneumococcal vaccination, having an ER blood culture
before antibiotic administration, smoking cessation counseling, having initial
antibiotics within six hours of arrival, receiving the most appropriate
antibiotic, and receiving the influenza vaccination.
Independent variables
Independent variables were location (high-minority ZCTA versus other) and
CDSS (use versus not). Shorter travel patterns found among minority
populations suggest that hospitals in high-minority ZCTAs will disproportion-
ately serve these populations (31). Therefore, this analysis used comparisons of
process quality indicators for pneumonia between hospitals within high-
minority ZCTAs and hospitals in low-minority ZCTAs.
CDSS and minority neighborhood disparities 133
Control variables
Control variables were used to hold community and hospital characteristics
that might be anticipated to be related to pneumonia quality measures but are
not of interest for this work. Control variables used in this analysis came from
the 2011 RTI Spatial Impact Factor Database, the 2010 AHA EHR Adoption
Survey, the 2010 AHA Annual Hospital Survey, CMS Hospital Compare
Database and the Department of Agriculture’s Rural-Urban Commuting Area
Codes (RUCA).
Community variables used, which reflect the external environment portion
of Anderson’s Model, came from the RTI Database and consisted of percent 65+
population in poverty, percent 65+ population and percent without a high
school diploma, all at the ZCTA level. Rural–Urban Commuting Area Codes
(RUCAs) from the Department of Agriculture were used to control for level of
rurality; hospitals located in a metropolitan ZCTAs were defined as urban
(RUCA ¼ 1–3); others as rural (RUCA ¼ 4–10). Within rural hospitals, CAHs
were assigned as a separate designation.
Hospital-level characteristics affect pneumonia quality processes. The
researchers therefore controlled for hospital characteristics, other than
CDSS, reflecting the structural component of Donabedian’s Model. These
variables consisted of ownership (for-profit, not-for-profit and Non-Federal
Governmental), total number of beds (565, 65–149, and 150+), JC accredit-
ation, teaching status and being a system member/management style.
Characteristics that positively affect quality include being part of a hospital
system and having a centralized management structure (37). Other charac-
teristics associated with increased quality include being not-for-profit, not
being part of the Council of Teaching hospitals, being in a non-West region and
having fewer than 100 beds (20). Finally, accreditation by JCAHO positively
affects pneumonia quality (38).
134 J. Mitchell et al.
Propensity scores
Both quality of care and use of CDSS may be endogenous, in that hospitals
interested in improving quality of care may choose to invest in CDSS.
Therefore, in an effort to control for this endogeneity, we employed propensity
scores. Because of the nature of observational studies, the ‘‘treatment’’ subjects
are not randomized. Therefore, the CDSS-treated hospitals may have other
non-observed characteristics that improve quality, creating a selection bias
toward higher process measures. The propensity score technique controls for
this bias by balancing treatment and control groups on a set of factors
described below.
Rosenbaum and Rubin (39) highlight three propensity score techniques for
constructing a matched sample: pair matching on balancing scores, sub-
classification on balancing scores and covariance adjustment on balancing
scores. This study used the regression covariance adjustment propensity score
technique. This method reduces bias by adjusting for the pattern of observed
confounders (40,41). This will allow researchers to determine a more precise
(less biased) estimate of CDSS specifically on the association of disparities in
pneumonia processes. Propensity scores were calculated using a multivariable
logit model to predict use of CDSS based on ownership, bed size and urban/
rural location.
Statistical analysis
Preliminary and unadjusted analyses were performed using SAS (SAS
Institute, Cary, NC). The adjusted analysis using propensity scores was
conducted using Stata (Stata Corporation, College Station, TX) (42).
Preliminary analysis described the study population across high and low
minority ZCTAs. Unadjusted analysis estimated pneumonia composite scores
by high and low minority ZCTAs and other hospital and community charac-
teristics; significant differences were identified using Wald chi square tests.
Finally, a multivariable ordinary least squares regression analysis was
performed to determine hospital and ZCTA-level characteristics significantly
associated with pneumonia process quality.
RESULTS
Several differences exist between the 2315 hospitals that both responded to the
CDSS question in the EHR Adoption Survey and reported quality measures to
the Hospital Compared and non-responding hospitals. Those hospitals that
both responded to the CDSS question and reported quality indicators are more
likely to be accredited by JCAHO and either be an independent hospital or
have a centralized management structure. Conversely, non-respondents are
more likely to be a teaching hospital and have a decentralized management
structure.
Study hospitals are listed in Table 1. Hospitals in high-minority ZCTAs were
no different in their use of CDSS as compared to low-minority ZCTAs (33.1%
and 31.1%, respectively). Significantly more high-minority ZCTAs were located
in urban areas, as compared to rural areas. The majority of high-minority
hospitals had 150+ beds. The majority of both high-minority and low-minority
hospitals used in the analysis were not-for-profit. High-minority hospitals used
CDSS and minority neighborhood disparities 135
Significance
High Lower 3 across minority
minority quartiles status
in the analysis were mainly located in the South, while the majority low-
minority hospitals were located in the Midwest. High-minority hospitals have
a lower percentage of CAHs than low-minority ZCTAs. Finally, high-minority
ZCTAs had higher amounts of JCAHO accredited hospitals as well as teaching
hospitals, as compared to low-minority ZCTAs.
Unadjusted estimates for the pneumonia process composite scores are listed
in Table 2. Hospitals located within high-minority ZCTAs had significantly
lower composite scores as compared to those hospitals within non-
high-minority ZCTAs (91.6% versus 93.1%, respectively). Hospitals that use
CDSS had higher pneumonia process composite scores than their non-CDSS
using counterparts (93.5% and 92.0%, respectively). Other factors associated
with higher pneumonia process scores included: being in an urban ZCTA,
136 J. Mitchell et al.
Significantly
different at
Composite p50.05
Minority status
High minority 91.62% *
Low minority 93.14%
CDSS
Yes 93.52% *
No 92.01%
ZCTA
Urban 93.38% a: b, c
Rural 92.54% b: a, c
CAH 90.08% c: a, b
Size
565 (a) 91.09% a: b, c
65–149 (b) 92.71% b: a
150+ (c) 93.07% c: a
Ownership
Government (a) 89.62% a: b, c
Not for profit (b) 93.08% b: a, c
For profit (c) 94.39% c: a, b
Region
Northeast (a) 92.95% a: b, c, d
Midwest (b) 90.08% b: a
South (c) 93.57% c: a
West (d) 92.48% d: a
JCAHO accredited
Yes 93.20% *
No 89.39%
System member
Yes 93.66% *
No 91.08%
Teaching
Yes 92.52%
No 92.49%
Management style
Not a system member (a) 91.08% a: b, c
Decentralized system member (b) 93.23% b: a, c
Centralized/mod-centralized system member (c) 94.14% c: a, b
*Significant at p50.05.
both mid-size (65–149 beds) and large (150+ beds) hospitals, for profit
hospitals, hospitals in the South, JCAHO accredited hospitals, not being a
CAH, and hospitals that were part of a system and had a centralized
management style. CAHs have significantly lower composite measures than
both urban and other rural hospitals. No significant differences between
teaching and non-teaching hospitals were detected.
Table 3 represents the adjusted least squares logistic regression analysis for
the pneumonia process composite scores. Five models are offered, predicting
the pneumonia composite score based on the following: (1) high-minority
ZCTAs versus low-minority, (2) high-minority ZCTAs and other control
CDSS and minority neighborhood disparities 137
Table 3. Adjusted least squares regression analysis for pneumonia process composite
scores for hospitals with propensity to use CDSS for guidelines.
Correlation matrix
Adjusted
Score CDSS, low CDSS, high None, low None, high
CDSS present:
Low-minority location 91.87 * * * *
High-minority location 90.89 0.3039 * * *
No CDSS:
Low-minority location 91.56 0.6687 0.4068 * *
High-minority location 89.48 0.0047 0.0250 0.0017 *
Estimates adjusted for: rurality, percent 65+ poverty population, percent 65+, percent less
than high school education, region, teaching status, JCAHO accreditation, central-
izatio'n, and propensity for CDSS implementation.
DISCUSSION
The work presented in this study affirms and builds upon previous research
(27–30) by means of CDSS being associated with higher quality measures and
maintaining that relationship in high-minority hospitals. Specifically, we
found positive associations between hospital CDSS use and pneumonia process
indicator composite scores. This association was not uniform. In adjusted
CDSS and minority neighborhood disparities 139
STUDY LIMITATIONS
Our study did not have patient-level data and thus did not directly calculate
process-level quality indicators and composite scores; we relied on hospital
reported data. Furthermore, without patient data, the researchers must rely
on aggregate measures of ZCTA poverty, education and 65+ population.
140 J. Mitchell et al.
Therefore, this study may be limited by ecological fallacy (45); hospitals located
in high-minority neighborhoods may not have a higher concentration of
minority patients. Given the goals of this study, the contextual variables that
reflect local built environment may be appropriate when examining external
health risk.
Second, the data source did not document the extent that the physicians
actually use CDSS, but was limited to whether CDSS had been implemented in
all units. We make the assumption that CDSS implementation corresponds
with actual physician CDSS use, although the degree of that correspondence is
not known. Furthermore, the CDSS used in this study within support clinical
guideline reminders and alerts, we cannot confirm whether or not pneumonia
care is included. Currently, there is no public-use dataset that documents
actual physician/nurse use of CDSS. Future researchers should focus within
individual hospitals or health systems, where it may be possible to identify
physician-usage of CDSS and other health information technologies.
Finally, we had to address missing data. Specifically, some hospitals
reported on fewer than the complete six process indicators. Hospitals may
have elected to report only for those guideline elements on which their
performance ranked high.
CONCLUSION
In the research reported in this study, the implementation of CDSS was
associated with reduced quality disparities between hospitals located in high-
minority ZCTAs and those in low-minority areas. If replicated by future
research, the link suggests a strategy that hospitals may pursue to ensure that
quality goals are met even among otherwise high-risk patient populations.
From a policy perspective, results support efforts to direct policy toward
broader adoption of CDSS.
DECLARATION OF INTEREST
The authors report no conflicts of interest. The authors alone are responsible
for the content and writing of this article.
REFERENCES
1. Xu J, Kochanek K, Murphy S, Tejada-Vera B. Deaths: final Data for 2007.
Hyattsville, MD: U.S. Department of Health and Human Services, Centers for
Disease Control and Prevention, National Center for Health Statistics; 2010.
2. Heron M. Deaths: leading causes for 2008. Natl Vital Stat Reports Centers Dis
Control Prev Natl Cent Heal Stat Natl Vital Stat Syst 2012;60:1–94.
3. Bartlett JG, Mundy LM. Community-acquired pneumonia. N Engl J Med 1995;333:
1618–24.
4. Marston BJ, Plouffe JF, File Jr TM, et al. Incidence of community-acquired
pneumonia requiring hospitalization. Results of a population-based active surveil-
lance Study in Ohio. The Community-Based Pneumonia Incidence Study Group.
Arch Intern Med 1997;157:1709–18.
5. Kaplan V, Angus DC, Griffin MF, et al. Hospitalized community-acquired
pneumonia in the elderly. Age- and sex-related patterns of care and outcome in
the United States. Am J Respir Crit Care Med 2002;165:766–72.
CDSS and minority neighborhood disparities 141
6. Niederman MS, Mandell LA, Anzueto A, et al. Guidelines for the management of
adults with community-acquired pneumonia. Diagnosis, assessment of severity,
antimicrobial therapy, and prevention. Am J Respir Crit Care Med 2001;163:
1730–54.
7. Capelastegui A, España PP, Quintana JM, et al. Improvement of process-of-care
and outcomes after implementing a guideline for the management of community-
acquired pneumonia: a controlled before-and-after design study. Clin Infect Dis
2004;39:955–63.
8. Hospital Compare. Centers for Medicare and Medicaid; n.d. Available from: http://
www.raconline.org/topics/critical-access-hospitals/faqs.
9. Burton DC, Flannery B, Bennett NM, et al. Socioeconomic and racial/ethnic
disparities in the incidence of bacteremic pneumonia among US adults. Am J Public
Health 2010;100:1904–11.
10. Harrison LH, Dwyer DM, Billmann L, et al. Invasive pneumococcal infection in
Baltimore, MD: implications for immunization policy. Arch Intern Med 2000;160:
89–94.
11. Hausmann LRM, Ibrahim SA, Mehrotra A, et al. Racial and ethnic disparities in
pneumonia treatment and mortality. Med Care 2009;47:1009–17.
12. Mortensen E, Cornell J, Whittle J. Racial variations in processes of care for patients
with community-acquired pneumonia. BMC Health Serv Res 2004;4:20.
13. Fine JM, Fine MJ, Galusha D, et al. Patient and hospital characteristics associated
with recommended processes of care for elderly patients hospitalized with
pneumonia: results from the medicare quality indicator system pneumonia
module. Arch Intern Med 2002;162:827–33.
14. Hasnain-Wynia R, Baker DW, Nerenz D, et al. Disparities in health care are driven
by where minority patients seek care: examination of the hospital quality alliance
measures. Arch Intern Med 2007;167:1233–9.
15. Andersen RM. National health surveys and the behavioral model of health services
use. Med Care 2008;46:647–53.
16. Ly DP, Lopez L, Isaac T, Jha AK. How do black-serving hospitals perform on patient
safety indicators? Implications for national public reporting and pay-for-perform-
ance. Med Care 2010;48:1133–7.
17. Jha AK, Stone R, Lave J, et al. The concentration of hospital care for black veterans
in veterans affairs hospitals: implications for clinical outcomes. J Healthc Qual
2010;32:52–61.
18. Joynt KE, Orav EJ, Jha AK. Thirty-day readmission rates for medicare
beneficiaries by race and site of care. JAMA 2011;305:675–81.
19. Jencks SF, Cuerdon T, Burwen DR, et al. Quality of medical care delivered to
medicare beneficiaries a profile at state and national levels. JAMA 2000;284:1670–6.
20. Jha AK, Li Z, Orav EJ, Epstein AM. Care in US hospitals – the hospital quality
alliance program. NEJM 2005;353:265–74.
21. Schoen C, Davis K, How SKH, Schoenbaum SC. U.S. health system performance: a
national scorecard. Health Aff (Millwood) 2006;25:w457–75.
22. Ross CE, Mirowsky J. Neighborhood socioeconomic status and health: context or
composition? City Community 2008;7:163–79.
23. Zhang W, Chen Q, McCubbin H, et al. Predictors of mental and physical health:
individual and neighborhood levels of education, social well-being, and ethnicity.
Health Place 2011;17:238–47.
24. Sarrazin MV, Campbell M, Rosenthal GE. Racial differences in hospital use after
acute myocardial infarction: does residential segregation play a role? Health Aff
(Millwood) 2009;28:w368–78.
25. Kaushal R, Shojania KG, Bates DW. Effects of computerized physician order entry
and clinical decision support systems on medication safety: a systematic review.
Arch Intern Med 2003;163:1409–16.
26. Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical practice using
clinical decision support systems: a systematic review of trials to identify features
critical to success. BMJ 2005;330:765.
142 J. Mitchell et al.
27. Bates DW, Cohen M, Leape LL, et al. Reducing the frequency of errors in medicine
using information technology. JAMIA 2001;8:299–308.
28. Sepucha KR, Fowler FJ, Mulley AG. Policy support for patient-centered care: the
need for measurable improvements in decision quality. Heal Aff Proj Hope
2004;Suppl Web Exclusives:VAR54–62.
29. Ruland CM. Decision support for patient preference-based care planning. J Am
Med Inform Assoc 1999;6:304–12.
30. Buising K, Thursky K, Black J, et al. Improving antibiotic prescribing for adults
with community acquired pneumonia: does a computerised decision support system
achieve more than academic detailing alone? – a time series analysis. BMC Med
Inform Decis Mak 2008;8:35.
31. Basu J. Severity of illness, race, and choice of local versus distant hospitals among
the elderly. J Health Care Poor Underserved 2005;16:391–405.
32. Donabedian A. Evaluating the quality of medical care. Milbank Q 2005;83:691–729.
33. Critical Access Hospital (CAH) FAQs 2011. Available from: http://rhrc.umn.edu/
2011/04/critical-access-hospital-year-6-hospital-compare-participation-and-quality-
measure-results-briefing-paper-28/.
34. Casey M, Burlew M, Moscovice I. Critical access hospital year 6 hospital compare
participation and quality measure results. Minneapolis, MN: Flex Monitoring
Team; 2011.
35. Landrum M, Bronskill S, Normand S. Analytic methods for constructing cross-
sectional profiles of health care providers. Heal Serv Outcomes Res Methodol 2000;
1:23–47.
36. Scinto J, Courtney J, Krumholz H, et al. Final report: hospital core performance
measurement project. Providence, RI: Rhode Island Department of Health; 2002.
37. Hines S, Joshi M. Variation in quality of care within health systems. Jt Comm J
Qual Patient Saf 2008;34:326–32.
38. Schmaltz SP, Williams SC, Chassin MR, et al. Hospital performance trends on
national quality measures and the association with joint commission accreditation.
J Hosp Med 2011;6:454–61.
39. Rosenbaum PR, Rubin DB. The central role of the propensity score in observational
studies for causal effects. Biometrika 1983;70:41–55.
40. Imbens GW. The role of the propensity score in estimating dose-response functions.
Natl Bur Econ Res Tech Work Pap Ser 1999;237:706–10.
41. Newgard CD, Hedges JR, Arthur M, Mullins RJ. Advanced statistics: the
propensity score – a method for estimating treatment effect in observational
research. Acad Emerg Med 2004;11:953–61.
42. StataCorp. Stata Statistical Software: Release 12. College Station, TX: StataCorp
LP; 2011.
43. With Special Feature on Emergency Care. Hyattsville, MD: National Center for
Health Statistics; 2013.
44. Andersen RM. Revisiting the behavioral model and access to medical care: does it
matter? J Health Soc Behav 1995;36:1–10.
45. Greenland S. Ecologic versus individual-level sources of bias in ecologic estimates
of contextual health effects. Int J Epidemiol 2001;30:1343–50.
Copyright of Informatics for Health & Social Care is the property of Taylor & Francis Ltd
and its content may not be copied or emailed to multiple sites or posted to a listserv without
the copyright holder's express written permission. However, users may print, download, or
email articles for individual use.