Since the Institute of Medicine report, To Err Is Human, reducing medical errors and improving patient safety have become a national priority.1 Crossing the Quality Chasm further explored quality improvement by outlining objectives for the U.S. health care system. It called for redesign by changing “systems of care” to be timely, efficient, effective, patient centered, safe, and equitable.2 Furthermore, the Agency for Healthcare Research and Quality (AHRQ) developed the Patient Safety Indicators (PSIs), which are most frequently used in estimating the incidence of adverse events in hospitals.3–5 These indicators facilitate the analysis of administrative data and suggest specific areas for improvement in hospitals.
Racial/Ethnic Disparities in Patient Safety
Two schools of thought have emerged regarding the existence of racial and ethnic disparities in patient safety. Some authors have documented disparities that are driven by characteristics of hospitals in which minorities seek care.6–10 Thus, all patients, regardless of their racial and ethnic backgrounds, are at increased odds of experiencing an adverse event in these hospitals. Others have proposed that individual-level factors influence the quality of care received and are a result of discrimination and biases toward minority groups.11,12 The literature on implicit attitudes suggests that providers, like the rest of the society, have an inherent preference toward white patients.13 Green et al14 explored these preferences in specialties such as internal and emergency medicine and found strong preferences for white patients, which resulted in an increased likelihood of thrombolytic treatment among white patients with acute coronary syndrome than nonwhite patients with the same condition. Although these studies have provided some clues to sources of implicit and explicit biases and their impact on nonverbal expressions and behaviors, there has been limited evidence linking these preferences to specific outcomes or adverse events.
There is a paucity of research in patient safety that includes race/ethnicity and how its interaction with other demographic factors such as education, income, and occupation impacts quality of care. Our objectives in this study, carried out through systematic examination of published studies on racial and ethnic disparities in patient safety since 1991 in the United States, were to (1) explore differences in reporting race/ethnicity in studies on disparities in patient safety; (2) assess adjustments for socioeconomic status, comorbidity, and disease severity; and (3) make recommendations on how to include race/ethnicity data in future publications on adverse events.
We searched PubMed for articles published from 1991 through May 1, 2013, using the following search strategy based on Medical Subject Headings (MeSH) terms: medication errors or medical errors or iatrogenic disease or malpractice or negligence or patient safety or diagnostic error or drug induced disease or adverse drug reaction or reporting systems or drug therapy/adverse effects or iatrogenic disease/drug therapy and race or African Americans or blacks or Hispanic Americans or Mexican American or Asian/Pacific Islanders or Indian North American or ethnicity. A comprehensive list of MeSH terms used to identify quantitative studies on adverse events and race/ethnicity is shown in Appendix 1.
A total of 174 publications were initially identified including a manual search of references of relevant articles. One hundred forty-four studies were excluded on the basis of their titles and abstracts. Two independent authors (J.S.O., E.F.U.) extensively reviewed 30 full-text articles. Additional studies that did not meet selection criteria were excluded, and 24 publications were finally selected for further analysis.
The included articles from PubMed were published beginning from 1991 to May 1, 2013. We also limited the search to articles and studies conducted in the United States. We included prospective and retrospective studies, chart reviews, and articles conducted with admission and discharge records. We included articles on adverse events that resulted from diagnostic errors, adverse drug events, and medical errors. Articles written on adverse events and malpractice claims or negligence were also included. An adverse event was defined as “an event that results in unintended harm to the patient by an act of commission or omission rather than by the underlying disease or condition of the patient.” Studies on adverse events that met all inclusion criteria without reference to race/ethnicity were also included.
Articles were excluded according to their titles, according to their abstract, and if these were not available (Fig. 1). Full-text articles were subsequently reviewed and excluded on the basis of study design. Articles that did not specify use of database or records were excluded. Literature reviews, technical notes, and case reports were excluded. Publications that did not specify an adverse event or patient safety event were excluded. Studies using self-report or surveys were not considered. Studies conducted outside the United States were also excluded.
Data collection was completed independently by 2 reviewers (J.S.O., E.F.U.) using a standard data extraction form. Data on study design, methods, clinical specialty, clinical setting, outcomes, patient demographic characteristics, payer, and type of events were extracted. Information on authors’ inclusion of race/ethnicity in baseline characteristics and outcomes was limited to tables with percentages or numbers only. Studies discussing in text without stratification in tables or multivariate analysis with specified covariates were considered as not reported. We used the following standard definition of comorbidity: a medical condition existing simultaneously but independently with another condition; we looked for the indication of presence on admission. We did not assess the methodological quality of the adjustments, that is, how information on the presence of comorbidities, teaching status, or technological sophistication was gathered, or the validity of the indices used. Data on the presence or absence of adjustments for comorbidities and hospital-level factors were pooled in a binary format. The data were then checked for any discrepancies and collated. Both authors agreed on 73% of studies included (κ score was 0.80).
Twenty-four studies published between 1991 and May 1, 2013, met the inclusion criteria (Table 1). Eleven studies specified age distribution of the study population: men and women were equally represented in the studies that included sex (n = 8), whereas a few studies reported close to 100% male enrollment (n = 3). National administrative databases such as the National Inpatient Sample and Veterans Health Administration (VA) were most commonly used in studies with population data. Two studies used states with greater than 90% race/ethnicity data collected. Studies examined a wide variety of specialties including obstetrics, pediatrics, medicine, and surgery as well as subspecialties such as nephrology and vascular surgery. The payer mix was quite variable, but 13 studies did not report a payer. Six studies indicated Medicare or Medicaid as the primary payer, 2 studies used the VA, whereas 4 studies considered all types of payers including private insurance and self-pay.
Consideration of Race/Ethnicity in Baseline Characteristics
Eight studies stratified baseline characteristics by race/ethnicity (Table 1). Three different studies were from Utah/Colorado, and both states did not collect these data as of 1992. African Americans represented less than 20% of the study population. The Hispanic population was the least represented, at less than 10% in reporting studies (n = 5). Three studies included socioeconomic status, which was obtained by median household income from the patient’s zip code, and none of the studies included information on educational attainment. Authors used a variety of methods to analyze adverse events. Thirteen studies used the PSIs, 3 studies used the Institute for Healthcare Improvement Trigger Tool for detecting adverse events, and 7 studies used a manual chart review with trained physicians and nurses.
Multivariate Adjustments for Comorbidities and Severity
Eight studies stratified outcomes by race/ethnicity, of which 6 conducted a multivariate analysis adjusting for comorbidity and patient severity (Table 2). Three studies conducted a binary regression comparing odds ratios of safety outcomes between racial/ethnic groups, using whites as the reference group. Because potential confounders were analyzed separately, we did not consider these to be adjusted results. One study reported univariate analysis for whites and nonwhites clustered in 4 categories defined by common risk factors and race-specific risk factors. Sixteen studies did not report outcomes by race/ethnicity or include in results tables. Regression models included covariates on patient characteristics such as age, sex, hospital length of stay, presence of cardiovascular disease, diabetes, hypertension, coagulopathy, and renal failure as well as severity such as Injury Severity Score in trauma. Hospital-level factors included teaching status, number of beds, hospital annual discharges, clustering of hospitals, technology sophistication index, number of discharges and procedures coded, nurse-patient ratio, as well as expertise. Four studies controlled for teaching status of the hospitals, and 1 study included geographic region of the country.
Among the few studies that were reviewed on adverse events, authors rarely explained the reason for the exclusion of race/ethnicity in baseline characteristics and results. By 2042, minorities will represent close to 50% of the entire U.S. population; the most dramatic change, however, is expected in Hispanics.18 However, only 20% of studies reported information on this population (n = 5) (Table 3). Contrary to belief, we observed that studies that stratified outcomes by race/ethnicity did control for variables related to patient conditions using a mixture of comorbidities or composite indices such as the Charlson and Elixhauser; however, only 1 study indicated the presence of comorbidities on admission, which would be the most accurate way to document their existence, rather than comorbidities on discharges.15,19 In contrast, we found that hospital-level variations beyond teaching status were least accounted for, which has been described as a possible explanation why disparities in quality of care received by minority groups exist (n = 2, 8%).20 Only 1 study included nursing expertise, technological sophistication, as well as number of diagnoses and procedures coded.16
In reporting of disparities in patient safety events, the evidence was mixed. Prior studies suggest that there are disparities in failure to rescue rates as well as increased mortality with hospitalizations complicated by pneumonia, sepsis, acute renal failure, and thromboembolic events in minority populations. However, understanding the mechanisms driving the disparities has been limited. Shimada et al17 reported fewer disparities in PSI in the Veterans Affairs population. Only 2 PSIs (13%) reached statistical significance controlling for 27 comorbidities: African Americans had a 30% increased odds of decubitus ulcers and 23% increased odds of developing a postoperative deep venous thrombosis (DVT)/pulmonary embolism (PE). The authors noted that there were no significant differences between blacks and whites in the experiences of PSI because risk factors were somewhat evenly distributed among the 2 groups. A plausible explanation is that the VA population is relatively homogenous compared with the non-VA population; however, disparities in the quality of care received by minority patients in the VA have been described.21–23
Metersky et al24 did not find African Americans at increased risk for pressure ulcers in an analysis of randomly selected charts from the National Medicare Patient Safety Monitoring System. This can be due to either suboptimal reporting or a true null association. The rationale for increased risk for ulcers in African Americans has been attributed to difficulty in diagnosing stage 1 decubitus ulcers in pigmented skins, which can lead to an underestimation of the risk or vice versa.25–28 There have been no plausible explanations for the increased risk for PE/DVT in African American patients in either the trauma or the VA population; some have suggested that genetics may potentially play a role; however, evidence on this remains inconclusive apart from in sickle cell disease.29–34
Although chart review is more sensitive for examining adverse events, we presume that administrative data will continue to be used by authors because it is readily available and cheaper. However, administrative databases can underestimate and overestimate comorbidities depending on clinical symptoms and impact of the comorbidity on resource use during an episode of hospitalization.35,36 In addition, secondary diagnosis codes sometimes fail to differentiate between existence of comorbidities before admission and complications due to medical care, although many agencies including Center for Medicaid and Medicare Services and state agencies have policies on reimbursements and separation of hospital-acquired conditions from comorbidities present on admission.37,38 Appropriate study designs and methods are needed to generate the evidence on the existence of racial/ethnic disparities in patient safety.
Despite the limitations of administrative databases, (1) studies should stratify baseline characteristics by race/ethnicity. In the event data on racial/ethnic groups, authors should indicate a priori intentions to consider but limitations based on data availability. If the exclusion of racial/ethnic groups was the intention of the study, authors should discuss in the limitations the generalizability of the studies to all populations. (2) Studies should adjust for factors related to disease and access to care by reporting multivariate analysis with covariates in a regression model. Sensitivity analysis should be done considering adjustments for different measures of comorbidities and hospital quality indicators as well as variability in other factors that can affect patient safety measures. (3) Studies should discuss the inclusion of comorbidities before admission and control for hospital-level variations in populations served such as socioeconomic status, percentage of minorities served, teaching status, nurse-patient ratio acuity, and technological capacity, which typically impact the incidence of PSI in hospitals.
Our study’s primary objective was to examine all quantitative studies since 1991 that examine racial and ethnic disparities in patient safety events based on predetermined search criteria. Although we have not captured all studies on patient safety based on our search, we limited our analysis to offer possible explanations for inconclusive evidence particularly in studies whose primary objectives were to report disparities in patient safety. Secondly, we did not focus our methods on specific types of events; further subgroup statistical analysis on specific events is warranted to explore the association of race/ethnicity and specific outcomes.
Poor inclusion of race/ethnicity in baseline characteristics, adjustments for comorbidities with indications of presence before admission, as well as poor consideration of geographic and hospital-level variations are plausible explanations for the mixed evidence on racial and ethnic disparities in patient safety. Our recommendations are based on the evidence gathered and do not represent the totality of all studies, however unique in informing future investigations on this topic. It is imperative that improvements in access to care be made through expansion of coverage, comparative effectiveness, and pay for performance and significant improvements be made in capturing race/ethnicity data, statistical analysis, and reporting of race/ethnicity specific outcomes by authors. This is important for examining the system-level factors that impact the quality of care received by minority populations in the United States and can further suggest areas for targeted solutions.
The authors thank Dr John Jackson at Harvard School of Public Health for his technical assistance and Dr Smetana at Beth Israel Deaconess Medical Center for his contributions to the study design.
1. Kohn T, Corrigan JM, Donaldson MS. To Err Is Human: Building a Safer Health System. Institute of Medicine
. Washington, DC: National Academy Press; 1999.
2. Institute of Medicine. Crossing the Quality Chasm: A New Health System for the Twenty-First Century
. Washington, DC: National Academy Press; 2001.
3. Patient Safety
Indicators, 2006 [AHRQ Web site]. December 14, 2012. Available at: http://http://www.ahrq.gov
/legacy/qual/psoact.html. Accessed July 25, 2013.
4. Doing what counts for patient safety
: federal actions to reduce medical errors and their impact. Report of the Quality Interagency Coordination Task Force (QuIC) to the President, February 2000. Available at http://archive.ahrq.gov/quic/report/fullreport.htm/. Accessed July 2014.
6. Katritsis DG, Ioannidis JP. Percutaneous coronary intervention versus conservative therapy in non-acute coronary artery disease: a meta-analysis. Circulation
7. Romana HA, Baker DA, Nerenz DA, et al. Disparities in health care are driven by where minority patients seek care. Arch Intern Med
8. Lucas FL, Stukel TA, Morris AM, et al. Race and surgical mortality in the United States. Ann Surg
9. Gaskin DA, Spencer CH, Richard PA, et al. Do minority patients use lower quality hospitals?. Inquiry
10. Coffey RM, Andrews RM, Moy E. Racial, ethnic, and socioeconomic disparities in estimates of AHRQ Patient Safety
Indicators. Med Care
11. Schulman KA, Berlin JA, Harless W, et al. The effect of race and sex on physician’s recommendations for cardiac catheterization. N Engl J Med
12. Van Ryn M, Fu SS. Paved with good intentions: do public health and human service providers contribute to racial/ethnic disparities in health? Am J Public Health
13. Sabin JA, Nosek BR, Greenwald AG, et al. Physicians’ implicit and explicit attitudes about race by MD race, ethnicity, and gender. J Health Care Poor Underserved
14. Green AR, Carney DR, Pallin DJ, et al. Implicit bias among physicians and its prediction of thrombolysis decisions for black and white patients. J Gen Intern Med
15. Miller MA, Elixhauser AN, Zhan CH. Patient safety
events during pediatric hospitalizations. Pediatrics
16. Chang DC, Handly N, Abdullah F, et al. The occurrence of potential patient safety
events among trauma patients: are they random? Ann Surg
17. Shimada SL, Montez-Rath ME, Loveland SA, et al. Racial disparities in Patient Safety
Indicator (PSI) rates in the Veterans Health Administration. In: Advances in Patient Safety: New Directions and Alternative Approaches (Vol. 1: Assessment)
. Rockville, MD: Agency for Healthcare Research and Quality; 2008.
18. United States Census Bureau. Population projections 2008. Available at: http://http://www.census.gov
/population/projections/. Accessed July 17, 2013.
19. Pine M, Norusis M, Jones B, et al. Predictions of hospital mortality rates: a comparison of data sources. Ann Intern Med
20. Silber JH, Rosebaum PR, Romano PS, et al. Hospital teaching intensity patient race and surgical outcomes. Arch Surg
21. Saha S, Freeman M, Toure J, et al. Racial and ethnic disparities in the VA Health Care System: a systematic review. J Gen Intern Med
22. Trivedi AN, Grebla RC, Wright SM. Despite improved quality of care in the Veterans Affairs Health System, racial disparity persists for important clinical outcomes. Health Aff
23. Halanych JH, Wang F, Miller DR, et al. Racial/ethnic differences in diabetes care for older veterans. Med Care
24. Metersky ML, Hunt DR, Kliman R, et al. Racial disparities in the frequency of patient safety
events: results from the National Medicare Patient Safety
Monitoring System. Med Care
25. Rosen J, Mittal V, Degenholtz H, et al. Pressure ulcer prevention in black and white nursing home residents: A QI initiative of enhanced ability, incentives, and management feedback. Adv Skin Wound Care
26. Scanlon E, Stubbs N. Pressure ulcer risk assessment in patients with darkly pigmented skin. Prof Nurse
27. Lapane KL, Jesdale W, Zierler S. Racial differences in pressure ulcer prevalence in nursing homes. J Am Geriatr Soc
28. Li Y, Yin J, Cai X. Association of race and sites of care with pressure ulcers in high-risk nursing home residents. JAMA
29. White RH, Zhou H, Murin S, et al. Effect of ethnicity and gender on the incidence of venous thromboembolism in a diverse population in California in 1996. J Thromb Haemost
30. Stein PD, Hull RD, Patel KC, et al. Venous thromboembolic disease: comparison of the diagnostic process in blacks and whites. Arch Intern Med
31. Ibrahim SA, Stone RA, Obrosky DS, et al. Racial differences in 30-day mortality for pulmonary embolism. Am J Public Health
32. Aujesky D, Long JA, Fine MJ, et al. African American race was associated with an increased risk of complications following venous thromboembolism. J Clin Epidemiol
33. Horlander KT, Mannino DM, Leeper KV. Pulmonary embolism mortality in the United States, 1979–1998: an analysis using multiple-cause mortality data. Arch Intern Med
34. Hooper WC. Venous thromboembolism in African-Americans: a literature-based commentary. Thromb Res
35. Romano PS, Roos LL, Luft HS, et al. A comparison of administrative versus clinical data: coronary artery bypass surgery as an example. Ischemic Heart Disease Patient Outcomes Research Team. J Clin Epidemiol
36. Iezzoni LI, Foley SM, Daley J, et al. Comorbidities, complications, and coding bias. Does the number of diagnosis codes matter in predicting in-hospital mortality? JAMA
37. Goldman E, Chu P, Prothro C, et al. Accuracy of conditions present on admission, do not resuscitate and e-codes in California patient discharge data [California OSHPD Web site]. Spring 2011. Available at: http://http://www.oshpd.ca.gov
/HID/Products/PatDischargeData/ResearchReports/PDDValidation/PDD_Validation_Study.pdf. Accessed August 9, 2013.
38. Snow LC, Holtzman L, Waters H. Accuracy of coding in the hospital-acquired conditions—Present on Admission Program [CMS Web site] June 16, 2012. Available at: http://http://www.cms.gov
/Medicare/Medicare-Fee-for-ServicePayment/HospitalAcqCond/Downloads/Accuracy-of-coding-Final-Report.pdf. Accessed May 11, 2013.