|Home | About | Journals | Submit | Contact Us | Français|
Health information technology (HIT)-supported quality improvement initiatives have been shown to increase ambulatory care quality for several chronic conditions and preventive services, but it is not known whether these types of initiatives reduce disparities.
To examine the effects of a multifaceted, HIT-supported quality improvement initiative on disparities in ambulatory care.
Time series models were used to assess changes in racial disparities in performance between white and black patients for 17 measures of chronic disease and preventive care from February 2008 through February 2010, the first 2 years after implementation of a HIT-supported, provider-directed quality improvement initiative.
Black and white adults receiving care in an academic general internal medicine practice in Chicago.
The quality improvement initiative used provider-directed point-of-care clinical decision support tools and quality feedback to target improvement in process of care and intermediate outcome measures for coronary heart disease, heart failure, hypertension, and diabetes as well as receipt of several preventive services.
Modeled rate of change in performance, stratified by race and modeled rate of change in disparities for 17 ambulatory care quality measures
Quality of care improved for 14 of 17 measures among white patients and 10 of 17 measures among black patients. Quality improved for both white and black patients for five of eight process of care measures, four of five preventive services, but none of the four intermediate outcome measures. Of the seven measures with racial disparities at baseline, disparities declined for two, remained stable for four, and increased for one measure after implementation of the quality improvement initiative.
Generalized and provider-directed quality improvement initiatives can decrease racial disparities for some chronic disease and preventive care measures, but achieving equity in areas with persistent disparities will require more targeted, patient-directed, and systems-oriented strategies.
The online version of this article (doi:10.1007/s11606-011-1842-2) contains supplementary material, which is available to authorized users.
The US health care system suffers from the dual problems of suboptimal quality of care for the population overall1,2 and widespread disparities in health care by race, ethnicity, gender, and socioeconomic status.3,4 Health care organizations have become increasingly engaged in generalized quality improvement activities, yet relatively few organizations have undertaken efforts to specifically address disparities in care.5 While it is possible that generalized quality improvement efforts may also help to reduce health care disparities, very few empirical studies have specifically addressed this question.6–9
Health information technology (HIT) has recently been the subject of particular enthusiasm and interest as a means to improve the delivery and quality of health care, and the potential role of HIT in addressing health care disparities has also begun to receive attention.10,11 However, there is little published evidence regarding the effects of HIT-supported interventions on health care equity.12
Quality improvement initiatives, including those that are supported through HIT, may affect health care disparities by facilitating change at any of four levels: the experience of patients, the performance of microsystems of care (i.e., individual health care providers and care teams), the functioning of health care organizations (i.e., health centers and hospitals), and the operation of the health care environment (i.e., payment and regulatory systems).2,13 Provider-directed HIT-supported quality improvement tools, such as clinical decision support and provider feedback, have been associated with improvements in quality in the domains of adherence to guideline-recommended care and patient safety,14,15 and would be most likely to impact health care disparities by facilitating change at the microsystems level (i.e., changing the behavior of individual providers). Electronic clinical reminders, for example, could lead to decreased disparities by raising provider awareness of quality gaps for individual patients, and particularly for those patients for whom quality gaps are more likely to be overlooked due to competing medical and psychosocial priorities.10 Clinical decision support could lead to decreased disparities by reducing clinical uncertainty and biased decision making, both of which have been demonstrated to contribute to disparities in care.11,12 The more rapid retrieval of clinically relevant information could improve provider efficiency and potentially improve provider-patient communication by allowing providers more time to engage in shared decision making, which has been associated with reductions in health care disparities.15–18 In addition, regular provider feedback could motivate providers to change their practice patterns in order to improve the overall quality of their patient care,10 and patients who have traditionally had lower quality of care may benefit the most from these changes.
Alternatively, provider-directed HIT-supported quality improvement initiatives could negatively impact provider performance in ways that disproportionately affect patients from historically disadvantaged groups. Clinical reminders and clinical decision support tools could worsen provider communication skills19 or increase patient mistrust20—either of which could worsen provider-patient interactions, decrease the patient-centeredness of care, exacerbate existing health care disparities, or even create new disparities.
Therefore, in order to better understand the effects of generalized, HIT-supported, and provider-directed quality improvement initiatives on disparities in ambulatory care, we examined the effects of a HIT-supported quality improvement initiative on racial disparities within a large general internal medicine practice. In particular, this study addresses two questions: following the implementation of the quality improvement initiative, were improvements in quality seen for both blacks and whites, and were baseline racial disparities between these groups eliminated? We hypothesized that the provider-directed HIT-supported quality improvement initiative would be associated with declining disparities for quality measures that are more directly under the control of health care providers (i.e., process of care measures) and that disparities would remain unchanged for measures that require more patient engagement (i.e., intermediate outcome measures).
This study was conducted in an academic general internal medicine ambulatory care practice in Chicago that serves over 25,000 patients annually and has used a commercially available electronic health record (EHR) (EpicCare, Epic Systems Corp., Verona, Wisconsin) for over 10 years. About 49% of the patients are White, 12% Black, 5% Hispanic, and 34% other or unknown. Approximately 68% of patients are privately insured, 22% have Medicaid, 5% have Medicare, and 5% are uninsured.
The UPQUAL (Using Precision Performance Measurement to Conduct Focused Quality Improvement) initiative was implemented throughout the practice on February 7, 2008. The components of UPQUAL and the effects of the initiative on overall quality of care have been described elsewhere,21 but to summarize, the intervention includes electronic point-of-care clinical reminders, decision support tools within the EHR to promote adherence to clinical guidelines, and regular provider feedback, which includes quarterly individual performance reports on all targeted measures and monthly lists of patients who are not prescribed essential medications (e.g., beta-blocker for a patient with a history of myocardial infarction). The quality improvement initiative was applied to all patients in the practice and was not specifically targeted or tailored based on patient race or other factors.
We examined disparities in the quality of care over a 2-year period following the implementation of the UPQUAL initiative for 17 quality measures. These included process of care and intermediate outcome measures for coronary heart disease, heart failure, hypertension, and diabetes, and several preventive care measures (Table 1). The proportion of eligible patients satisfying each measure, stratified by race, was assessed on the first of every month from February 1, 2008, through February 1, 2010. Patients were eligible for a measure if they were 18 years or older, had made at least two visits to the practice in the 18 months prior to the time of the quality assessment, were assigned to an attending physician, and satisfied the denominator criteria based on age, gender, and medical diagnoses as detailed in Table 1. The proportion of patients satisfying a measure was defined as the number of patients satisfying the numerator criteria divided by the number of patients meeting the denominator criteria for the measure minus the number of patients who did not satisfy the measure but had a documented exception: number satisfied/(number eligible – number not satisfied with a documented exception). Measures were assessed by querying the enterprise data warehouse, which contains stored data from the EHR, and details regarding the queries have been published elsewhere.21
Due to the small number of patients of other race or ethnicity, we limited the analysis to patients whose race/ethnicity was recorded as black or white in the EHR. We collected data regarding the distribution of age, gender, and number of chronic medical conditions for white and black patients who were eligible for at least one of the 17 quality measures at three time points: February 1, 2008, February 1, 2009, and February 1, 2010. The number of chronic medical conditions for each patient was assessed based on a count of ten possible chronic disease categories: nine categories included in the 2008 Dartmouth Atlas of Health Care (cancer, chronic pulmonary disease, coronary artery disease, congestive heart failure, peripheral vascular disease, severe chronic liver disease, diabetes, renal failure, and dementia) and hypertension.22
Pearson’s chi-square test was used to identify measures with statistically significant differences in the proportion of black versus white patients satisfying each measure at baseline (February 1, 2008). Time series models were used to assess the significance of temporal trends in the percentage of patients satisfying each quality measure over the 2 years following the implementation of the quality improvement initiative for black and white patients, separately, and to assess the significance of changes in racial disparities.
To construct the time series models, a regression model was fit to each quality measure using a single practice-level measurement (e.g., the percentage of all eligible white patients in the practice who satisfied the colorectal cancer screening measure) as the dependent variable and time in months as the independent variable. For all but one measure, a linear model was applied with time as a continuous predictor from 0 to 24 months. For one measure, the prescription of antiplatelet therapy for patients with coronary heart disease, the temporal trend was very different for months 0 to 7 compared to months 8 to 24 (Appendix available online); therefore, the regression model included time as a continuous predictor for months 0 to 24, a dichotomous indicator variable for months 0 to 7 versus months 8–24, and an interaction term between the dichotomous indicator variable and time. Since the practice-level quality measurements could be highly correlated from month to month, each time series model included the appropriate number of autoregressive error parameters.21 The beta coefficient for the time variable was used to determine the modeled rate of change per year for each quality measure.
The study was not limited to a cohort of patients who were continually eligible for a given quality measure at all time points throughout the study period (i.e., based on the dates of her two most recent visits to the practice, a woman might be eligible for the breast cancer screening measure on January 1, 2009 and ineligible for that measure on January 1, 2010). Thus, the results could be biased if changes in patient characteristics over time differed by race. To determine whether temporal changes in patient characteristics differed between black and white patients, we used linear regression models (for age and chronic medical conditions) and a logistic regression model (for gender) with time, race, and an interaction term between time and race as the independent variables.
Analyses were done using Stata10 (StataCorp, College Station, Texas). A two-sided p-value of less than 0.05 was used to determine statistical significance. The study protocol was approved by the Northwestern University Institutional Review Board, and a waiver of informed consent was obtained.
At baseline, 8,919 black and white patients were eligible for at least 1 of the 17 quality measures. The number of patients eligible for each measure ranged from 107 to 5,294 (Table 1). Black and white patients did not differ in age, but a higher proportion of black patients were female, and black patients had more chronic medical conditions compared to white patients (Table 2). However, age and number of chronic medical conditions increased similarly among black and white patients, and the proportion of female patients did not change over time. A greater percentage of black patients who were eligible for any measure on February 1, 2008, remained eligible on February 1, 2009, and February 1, 2010, compared to white patients (52% vs 43%, p<0.001).
Quality of care improved for 14 of 17 measures for white patients and 10 of 17 measures for black patients over the 2 years following the implementation of the quality improvement initiative (Table 3). For five of the eight process of care measures, quality improved for both white and black patients, and for three of these measures the rate of improvement in quality was at least as high for black patients as for white patients. For the remaining three process of care measures, improvements in quality were seen only among white patients. Of the four intermediate outcome measures, there were no measures for which improvements in quality were seen among both white and black patients. For two intermediate outcome measures quality improved for white patients only (glycemic control and LDL control for patients with diabetes), for one measure quality improved for black patients only (blood pressure control for patients with diabetes), and for one measure quality improved for neither group (blood pressure control for patients with hypertension). Quality improved for both white and black patients for four of the five preventive care measures, and the rate of improvement was at least as high for black patients as for white patients for all four measures. The percentage of patients up to date on breast cancer screening declined among both white and black patients.
There were racial disparities in quality for 7 of the 17 measures at baseline (Table 4). This included two process of care measures, one intermediate outcome measure, and four preventive care measures. The disparities ranged in absolute magnitude from 4.0% for pneumococcal vaccination to 12.3% for osteoporosis screening or treatment.
Racial disparities narrowed for two of the seven measures: the prescription of antiplatelet therapy for patients with coronary heart disease and colorectal cancer screening (Table 4). There were no significant changes in disparities for four of the seven measures, and racial disparities increased by 2.8% per year for the achievement of LDL control among patients with diabetes.
The racial difference in quality widened for two measures that did not have racial disparities at baseline: the prescription of beta-blockers to patients with a history of myocardial infarction and glycemic control
Understanding the effects of generalized and HIT-supported quality improvement initiatives on health care disparities is important given the pervasiveness of disparities and nationwide efforts to promote the adoption of HIT, and this study provides several key insights. First, the implementation of a generalized, HIT-supported, and provider-directed quality improvement initiative was associated with the narrowing of disparities for two of the seven measures with baseline disparities. Even when baseline disparities persisted, in general, the quality of care improved for both groups. Importantly, for measures with no disparities at baseline, the direction of the change in quality following the implementation of the quality improvement initiative was the same for black and white patients in most cases, and there were few new disparities.
Second, the changes in quality following the implementation of the quality improvement initiative tended to differ for intermediate outcome measures compared to process of care and preventive care measures. Quality improved for black and white patients for five of eight process of care measures, four of five preventive care measures, but none of the four intermediate outcome measures. In addition, the improvements in quality tended to be smaller for intermediate outcome measures compared to process of care and preventive care measures. Furthermore, it is notable that two of the three measures for which disparities widened were intermediate measures of disease control: glycemic control and LDL control for patients with diabetes. In both cases, the percentage of patients satisfying the measure improved for white patients only, resulting in a new, albeit small disparity.
Taken together, these findings suggest that generalized and provider-directed HIT tools are likely to have some positive effects on health disparities, particularly for chronic disease process of care measures and preventive care measures. However, in order to achieve more substantial improvements in health care equity, particularly in regards to important measures of disease control, additional strategies, such as tools to improve patient engagement,23,24 patient-physician communication,25,26 and access to care10,20 as well as systems-oriented strategies to improve care delivery will still be needed.
The breast cancer screening measure is the only one for which there was a decline in quality, making it an anomalous case that merits discussion. At approximately the same time that the UPQUAL initiative began, we began to have very long delays (more than 6 months) in scheduling screening mammograms at our institution due to a regional shortage of trained radiologists. Despite efforts to refer women to other sites for mammography, there was a decline in the proportion of white and black women who were up to date on breast cancer screening.
This study has several important limitations. As a single institution study, the results may not be generalizable to other types of settings. We only examined differences between white and black patients, and we used a simple measure of comorbidity with no adjustment for severity of illness to explore differences in patient characteristics by race. We did not adjust our threshold for statistical significance for multiple comparisons. However, as most of the p-values for statistically significant results were <0.001, we do not believe this would have affected our results significantly. In some cases, adherence to quality measures was very high at baseline, particularly for white patients; therefore, the narrowing of disparities for some measures could have been due in part to a ceiling effect (e.g., the rate of improvement was lower for white compared to black patients because white patients had less room for improvement). Furthermore, while we have demonstrated that changes in disparities followed the implementation of the quality improvement initiative, we cannot prove causality because of the quasi-experimental study design. However, we believe that a causal link is likely for several reasons. First, although black and white patients differed in some characteristics at baseline, the changes in key characteristics were similar for black and white patients over the study period. Therefore, we do not believe that changes in patient population over time contributed significantly to the observed changes in disparities. Second, no other major quality improvement initiatives (e.g., pay-for-performance incentives, changes in care team structure) were going on during the study period. Finally, the improvements in quality observed during the study period were greater than those expected based on temporal trends for most measures.21
Nonetheless, this study has several strengths and provides important evidence regarding the effects of HIT-supported quality improvement initiatives on disparities in ambulatory care. This study evaluated a very common form of quality improvement—clinical decision support and provider feedback—delivered via a commercially available and widely used EHR system that aimed to improve a range of ambulatory quality indicators. Thus, the findings should be relevant to a wide range of practices with comprehensive EHRs. Overall, we demonstrated that generalized and provider-directed quality improvement efforts can lead to reductions in disparities across several areas of preventive and chronic disease care, but will not be sufficient for achieving health care equity. Efforts focusing on other levels of the health care system (e.g., patients, heath care organizations, and payment/regulatory systems) and efforts focusing specifically on disparity reduction will still be needed. Moreover, we demonstrated that we do not need to wait until all the challenges surrounding the collection of patients’ racial/ethnic data are resolved and the full implementation of “meaningful use” is completed in order to make meaningful steps towards advancing health care equity.27
Contributors We would like to thank Andrew J. Cooper for his assistance with the time series analyses.
Funders This study was funded in part by the Agency for Healthcare Research and Quality (grant no. 5R18HS017163) and the Aetna Foundation. Dr. Jean-Jacques was supported in part by Federal funds from the National Center for Research Resources (NCRR) and National Institutes of Health (NIH) through the Clinical and Translational Science Awards Program (CTSA) (Northwestern University UL1RR0254741). Dr. Persell was supported by career development award K08HS015647 from the Agency for Healthcare Research and Quality. Content is solely the responsibility of the authors and does not necessarily represent the official views of the AHRQ, NIH, or Aetna Foundation.
Prior presentations We presented an earlier version of this manuscript as an oral presentation at the 32nd Annual Meeting of the Society of General Internal Medicine in Miami, Florida, in May, 2009.
Conflict of Interest None disclosed.