Assessing Health Care Equity Through Administrative Data-Derived Qquity Indicators and Socio-Economic Determinants of Health in Emergency Care: A Systematic Review


 Background: Achieving equity in health care remains a challenge for health care systems worldwide and marked inequities in access and quality of care persist. The performance assessment of health care systems is often limited to quality and efficiency indicators. Identifying indicators of health care equity is an important first step in integrating the concept of equity into assessments of health care system performance. Because emergency care serves as the interface between ambulatory and inpatient care, it appears to be a key setting in which to begin this process.Methods: We conducted a systematic review of administrative data-derived health care equity indicators and their association with socio-economic determinants of health (SEDH) in emergency care settings. Ovid MEDLINE, EMBASE, PUBMED and Web of Science were searched for relevant studies using administrative data following PRISMA-equity guidelines. The outcomes of interest were indicators of health care equity and the associated SEDH they examine.Results: Among 29 studies identified, 14 equity indicators were identified and grouped into four categories that reflect the patient emergency care pathway. Total emergency department (ED) visits and ambulatory care sensitive condition-related ED visits were the two most frequently used indicators. The studies analysed equity based on seven SEDH: social deprivation, income, education level, social class, insurance coverage, health literacy and financial and non-financial barriers. Despite some conflicting results, all identified SEDH are associated with inequalities in access to and use of emergency care. Conclusion: The use of administrative data-derived indicators in combination with identified SEDH could improve the measurement of health care equity in emergency care settings across health care systems worldwide. Using a combination of indicators is likely to lead to a more comprehensive, well-rounded analysis of health care equity than using any one indicator in isolation. Though studies analysed focused on emergency care settings, it seems possible to extrapolate these indicators to measure equity in other areas of the health care system. Further studies elucidating root causes of health inequities in and outside the health care system are needed.

settings and the choice of equity indicators and SEDH, and how best to combine them, represent some of the key challenges facing researchers in this eld.
The World Health Organization de nes health equity as "the absence of avoidable, unfair, or remediable differences among groups of people, whether those groups are de ned socially, economically, demographically or geographically or by other means of strati cation. "Health equity" or "equity in health" implies that ideally everyone should have a fair opportunity to attain their full health potential and that no one should be disadvantaged from achieving this potential". (13) Several other de nitions are used to de ne health equity, (14) but the principle of fairness is common across them all. Because of its fundamental link to the principle of justice, health equity is inherently normative and can vary according to cultural, geographical or temporal differences. (15) Traditionally, a distinction is made between horizontal and vertical social inequities. Vertical social inequities are associated to education level, professional status, income, etc. and horizontal social inequities to race/ethnicity, sexual orientation and gender identity, nationality, origin, place of residence, etc. (16) Altogether these factors associated to vertical and horizontal inequities are named social determinants of health. For the purposes of this study, we focus on vertical social inequities and their associated factors that we call: socio-economic determinants of health (SEDH). They denote the economic and social conditions that in uence the health of people and communities.
The rationale of this systematic review is based on two landmark World Health Organization (WHO) reports. The rst is the WHO Report of the Commission of Social Determinants of Health published in 2008. (17,18) It highlighted the importance of building health care system based on the principle of equity and the necessity of assessing social determinants of health to achieve this goal. (18) The second WHO report published in 2010 focused on the importance of guaranteeing access to health care for all independent of one's ability to pay and of raising su cient resources for health. (19) Both landmark reports therefore emphasize the importance of SEDH in achieving health care equity.
In this review, we focus our analysis on inequities in the health care system, and more speci cally on emergency care inequities (related to emergency department and emergency surgery). Emergency care is indeed situated at the interface of outpatient (ambulatory) care and inpatient (hospital-based) care.
Identifying indicators of health care equity in this setting make it possible to both determine access to quality outpatient care and highlight differences in quality of care within hospital structures that provide inpatient care. (8,20,21) In addition, emergency department (ED) overcrowding remains a major issue for many health care systems and is frequently source of potential inequities. (22)(23)(24) In fact, it has been suggested that the ED is a good setting to assess inequities, particularly concerning frequent users of the ED. (25,26) An important step to improving health care equity is establishing effective means of measuring it. Several approaches and data sources might be used to do so, from primary qualitative or quantitative data collection to the use of routinely collected administrative data. The latter has two fundamental advantages in the analysis of health care equity: the achievement of near complete coverage of the target population and the possibility of disaggregation in subpopulations. (27) Moreover, using administrative data minimizes cost and burden of response. (28) However, as they are not designed for the purpose of equity monitoring, administrative data also have some limitations, including lack of quality control of the collected data, a time lag in data availability, differences in concepts and de nitions used between datasets limited comparability, and the possibility of missing records.
The aim of this systematic review is to identify how health care equity is measured through the combination of administrative data-derived emergency care equity indicators and SEDH with the goal of creating a set of valuable and replicable indicators that can be used in the identi cation and analysis of health care equity in emergency care settings around the world.

Ii. Methods
The protocol of this systematic review was published in PROPSPERO at the outset of the study. (See Additional le 1) The reporting of this systematic review was based on the PRISMA-equity guidelines. (29) (See Additional le 2)

Inclusion/Exclusion Criteria
We included studies reporting on health care equity indicators, and that were analysed as such, with a focus on studies conducted in emergency care in high-income countries using administrative data. We chose to focus on high-income countries, in order to limit bias of the health care equity assessment due to the global lack of resources that could affect the health care system of low and middle-income countries and that could impact the health care use upstream of equity issues. We included studies on adults (age 18 and over). If a study included both children and adults, we limited data extraction to data pertaining only to adults. We included studies regardless of whether or not a disease-speci c focus was taken (for example cancer, chronic diseases or mental health). Searches were limited to articles in English, German, French, and Italian (due to the language skills of the authors) published between January 2010 and January 2019. We chose to focus on studies published after 2010 because of the signi cant evolution of health care equity-related literature that occurred following the two landmark WHO reports on health equity as already described. (18,19) We limited our analysis to studies looking at vertical social inequities and their associated socioeconomic determinants of health (SEDH) as de ned above, excluding studies looking at horizontal social inequities and their associated social determinants of health such as race/ethnicity, gender and place of residence, in order to ensure consistency between studies.
We excluded studies that did not focus on equity, as well as opinion papers, editorials, conference abstracts and study protocols.

Page 5/24
The search strategy was conducted with the assistance of a medical librarian using four 4 databases: Ovid MEDLINE, EMBASE, PubMed and Web of Science. We used keywords in the eld of equity, socioeconomic factors and emergency care. We combined the Medical Subject Headings (MeSH) terms "Health Services Accessibility", "Health Equity" or "Health care Disparities" with a combination of terms de ning administrative data and with text words "emergency department" or "emergencies". Initial searches were conducted in November 2018 to assess the scope of the literature. The nal search was conducted in January 2019. The full search strategy can be found in supplementary le 3. (See Additional le 3) Following the initial search, to identify any further relevant studies that were not initially captured or had not yet been published, we screened reference lists of all included studies and performed Google and Google Scholar searches using key search terms.

Study Selection
Two reviewers (KM, XL) conducted screening of articles independently and in duplicate. This was done in two stages. First by screening all titles and abstracts and second, by reviewing the full-text of all relevant articles to determine their eligibility in the nal analysis. Two other reviewers (JM, PB) provided arbitration in the event of a disagreement at both stages of screening. Reasons for exclusion of articles at the fulltext screening stage were documented.

Data Extraction
Two authors (KM, XL) extracted data independently and in duplicate from included studies using Rayyan® (free online systematic review management system) and any discrepancies were resolved by consulting the two other reviewers (JM, PB). Data on the key characteristics of the studies were extracted in a prede ned data extraction form, into an Excel® spreadsheet, including information about the design of the study, population, type of data, indicators of health care equity, SDEH addressed, main ndings and key conclusions.

Quality and Bias Assessment
In order to assess the risk of intrinsic bias inherent in this type of systematic review, we evaluated the quality of each study using an adapted Newcastle-Ottawa scale for cohort and cross-sectional studies.
1. Conceptual framework for the analysis To address equity, we based our analysis on a conceptual framework of access to health care, developed by Levesque and al.(31) This framework combines ve dimensions of accessibility (approachability, acceptability, availability/accommodation, and affordability/appropriateness) with ve corresponding abilities of the target population (ability to perceive, to seek, to reach, to pay and to engage). It provides a comprehensive approach to health care equity and the different factors that could impact it. (Fig. 1) We will use this framework to structure data extraction.
Due to the heterogeneity of study designs, indicators used and SDEH analysed, a quantitative metaanalysis of the data was not possible. Further, the signi cant heterogeneity precluded the creation of a funnel plot to assess publication bias. Quality assessment of the articles showed an overall good quality of studies with 24 (89%) considered to be good quality, two (7%) considered fair quality and one (3%) considered poor quality.
A summary description of each study is presented in Additional le 4. This group of indicators analysed access to outpatient care through differences in emergency care consumption (poor access to outpatient care leading to excess emergency care use). Therefore, they are indirect indicators of access to outpatient care. Five indicators belonged to this group.
(1) ED visits/Emergency admissions rate With 26% (n = 7) of articles using this indicator, it was the most commonly reported indicator identi ed in this systematic review. (32)(33)(34)(35)(36)(37)(38) It was used to highlight disparities of access to outpatient care. Since both re ect poor access to quality primary care, we have grouped them under the same indicator.
(2) Ambulatory care sensitive conditions (ACSCs) ED visits/ACSCs emergency admissions rate Also called Preventable ED visits/Preventable emergency admissions, this indicator, used in seven articles, is used as often as the previous indicator "ED visits/Emergency admissions rate". (8,33,35,(39)(40)(41)(42) It is deemed a more speci c indicator than "ED visits/Emergency admissions" alone to assess disparities in access to outpatient care.

(3) Frequent ED visits
One study used this indicator considering frequent ED visits when 4 or more ED visits occurred by an individual per year. (43) (4) ED-associated initial diagnosis rate This indicator compared the rate of initial diagnosis of cancer in the ED between different SEDH. (44) 1. Equity indicators of quality of emergency care (Group 2) The second group of health care equity indicators identi ed was indicators of quality of emergency care. They characterize disparities of care in the ED among targeted SEDH.
(5) Emergency speci c procedures rate Emergency speci c procedures comprised a combination of different procedures performed during emergency care, highlighting disparities in the quality or access to care for speci c emergency conditions such as a brain scan for the diagnosis of acute stroke, (45) reperfusion therapy in acute stroke, (46) and cardiac catheterization after myocardial infarction or cardiac arrest. (47,48) (6) Delay to diagnosis or treatment rate Two studies focused on disparities in time to a diagnostic procedure (CT scan for stroke)(45) and to de nitive treatment (time to permanent pacemaker implementation for emergency cases).(49)

(7) Missed diagnoses in ED rate
This indicator, used in one study, highlighted disparities of missed diagnoses of acute myocardial infarction according to insurance status or median household income. (50) This third group of indicators includes indicators of outcome disparities. We identi ed six outcome indicators.

(8) Major adverse event rate
This indicator was used in 2 studies that analysed emergency general surgery. (51,52) It represented the rate of speci c complications following an emergency general surgery including cerebrovascular accident, pneumonia, pulmonary embolus, acute respiratory distress syndrome, renal failure, urinary tract infection, myocardial infarction, sepsis, septic shock and cardiac arrest.
(9) In-hospital mortality and (10) failure to rescue rate In-hospital mortality was used to re ect the quality of care during emergency care or surgery as reported in three articles identi ed in our review. (48,51,52) One distinguishes in-hospital mortality from failure to rescue, which occurs when a patient dies as a result of a major adverse event and seems, therefore, to be more sensitive to assess differences in quality of emergency care or surgery. (51) (11) Neurological recovery rate This speci c indicator was used in one study analysing the neurological recovery over time of patients who presented to the ED with a cardiac arrest. (48) (12) Length of stay/Bed days (after emergency admission) Although these are traditional indicators of hospital care quality, they are used in one study that analysed inequities following emergency admission according to social deprivation.(53)

Global Equity indicators
As they could re ect a lack of outpatient care following a discharge post-admission and/or poor quality of care during an emergency admission, these following indicators could apply to the three different groups of indicators. Overall, indicators of access to outpatient care and particularly ED visits and emergency admission and ACSCs ED visits/ACSCs emergency admissions were the most frequent indicators used across studies.
The different emergency care equity indicators are summarized in Table 1. The articles included in this review analysed health care equity based on seven SEDH: Insurance status, social deprivation, income, education level, social class, health literacy and nancial and non-nancial barriers. (Additional le 4) They covered the ve abilities considered by Levesque et al. (Fig. 1), as mentioned in brackets and italics at the end of each paragraph.
Overall, the three main SEDH used to analyse health care equity across the 29 included studies were health insurance status, indices of social deprivation and income, and eight studies (28%) used more than one SEDH in their health care equity-focused analysis.

Insurance status
Insurance coverage as a relevant SEDH was approached in diverse ways amongst the 16 articles that used it including comparing outcomes between uninsured and insured individuals, (33,39) between publicly and privately insured individuals, (42, 47-49, 55, 58) or between uninsured, publicly and privately insured individuals. (32, 34, 44, 50-52, 56, 57) Present in more than half (55%) of the studies analysed, it is the most widely used SEDH in analyses of health care equity identi ed in this review. (It re ects the ability to pay in Levesque's framework).

Income
To measure income differences, four studies that used this SEDH used median income household (divided into quartiles or thirds) (50,52,55,56) and one used presence versus absence of a reportable income.(59) (It re ects the ability to reach and to pay).

Education level
Depending on the studies, the education level was divided into three or four categories ranging from never attended school to graduate degree.(46, 58) (Additional le 4 for details) (It re ects the ability to perceive, to seek and to engage).

Social class
This SDEH is de ned hierarchically into six classes in descending order: professional, managerial, skilled non-manual, skilled manual, semi-skilled manual, non-skilled manual. This SEDH was used in one study to analyse health care equity.(40) (It re ects the ability to reach and pay).

Health literacy
In one study, health literacy was the SEDH used in the health equity-focused analysis, based on scores obtained through the Rapid Estimate of Adult Literacy in Medicine test, a reading recognition test comprised of 66 health-related words arranged in ascending order of di culty.(41) (It re ects the ability to perceive and engage).

Financial and non-nancial barriers
In one paper, these two types of barriers were used based on subjects' responses to 14 questions (7 questions each) relating to nancial concerns and non-nancial barriers. (38) (It re ects the ability to reach and to pay).

Addressing health care equity through the association of emergency care indicators and SEDH
Although the heterogeneity of included studies in this systematic review precluded a meta-analysis, the last two columns of Additional le 4 present the statistical data and associated conclusions of the 29 reviewed studies. It can thus be seen that across all studies, all identi ed SEDH were found to be associated with statistically signi cant differences in emergency care indicators. Descriptive examples of associations between equity indicators and some of the main SEDH identi ed in this review are described below. The group of each indicator is indicated in bold and in brackets.

Health insurance
In a large retrospective study including over 2.

Education level and Health Literacy
Contrary to expectations, only two studies assess this SEDH, including one with a small sample of patients (n = 647), which found that lower education level was potentially associated with an increased risk of being an "emergency presenters" (de ned as presenting to ED around the time of a new cancer diagnosis) (group 1).(58) The other study, by Stecksen et al. highlighted that access to reperfusion therapy (group 2) for stroke is associated with higher patient education level (Odds ratio 1.14, 95%CI 1.03-1.26). (46) Only one study analysed the impact of health literacy on potentially preventable ED visits and found that patients with poor health literacy are approximately twice as likely to have preventable ED visits (group 1) than patients with adequate health literacy, even after adjustment for relevant confounding factors (Rate Ratio 1.93, 95%CI 1.55-2.40). (41) Iv. Discussion In the context of emergency care settings, ndings of this systematic review suggest that administrative data allows for a broad analysis of health care equity.  Representation of a conceptual synthesis of the assessment of health care equity in an emergency setting, through the combination of socio-economic determinants of health with emergency care equity indicators.

Emergency care Equity indicators
Emergency care equity indicators were identi ed and categorized into three main groups. They allow the assessment of health care inequities based on selected SEDH. (Fig. 3) The rst group (indicators of poor access to outpatient care) use emergency care to re ect issues of access to ambulatory care. Interestingly this is the most widely used group of indicators. Among them, the most frequently used indicator is ED visits/Emergency admissions but due to its lack of speci city, it must be interpreted with caution as there are notably many factors that could explain differences in ED visits or emergency admissions beyond health care equity (ie. differences in general health status, prevalence of disease, care-seeking behaviours, etc.).(60) ACSC ED visits/ACSC emergency admissions, which is the second most frequently used indicator in this group, is arguably more speci c from an equity perspective as it focuses on ED visits/admissions that are potentially preventable with good access to primary care. (22,61) The second group of indicators (indicators of quality of emergency care) directly analyse emergency care and are therefore more speci c in their measurement of health care equity in emergency care settings than indicators in the rst group. We found that they are used considerably less. This may re ect di culty in obtaining relevant data to measure these indicators through administrative datasets.
However, they might be useful indicators to use in future studies analysing health care equity.
Among outcome indicators, in-hospital mortality seems to be the most reproducible and available administrative data-derived indicator.
Finally, 30/90/326-day mortality and ED readmission, which are more global equity indicators assess the lack of access to outpatient care following an ED visit, but also potential issues during the emergency care leading to outcomes inequities.
Due to the inherent di culties of measuring a complex concept like health care equity and the large number of potential confounding factors, using a combination of indicators instead of one sole indicator to measure health care equity in any given health care context is more likely to result in a well-rounded assessment. As such, we suggest combining indicators from different categories when assessing health care equity. The choice of speci c indicators will depend on the context of the study, the study objectives and availability of administrative data (and relevant variables) in the health care setting of interest.

Socio-Economic Determinants of Health (SEDH)
In analysing equity in emergency care settings, seven SEDH were identi ed through this review.
Although median household income appears to be a common and reproducible measurement, many studies chose to use speci c area-level indices that account for multiple domains of deprivation experienced by the target population. These indices combine different parameters to assess deprivation, such as income, employment status, living environment deprivation, and education. These indices are less reproducible than median household income since they require many more variables for their calculation (which may not be available in all health care administrative datasets), but they are arguably better at measuring inequities due to the broad domains of deprivation they assess.
To assess the accessibility of health care services, one particular SEDH emerged: insurance status. Most of the studies analysing this SEDH were published in the US. We assume that this is due to the speci cities of the US health care system and the different health insurance reforms (most notably the Affordable Care Act), (62) which make this a very relevant SEDH in the US context. Despite most of the studies being US-based, this SEDH seems relevant to most health care systems in high-income countries, even those with universal health care coverage, where some individuals are able to access private insurance that covers additional bene ts, therefore creating potential inequities.(63) As such, this SEDH could be used more widely than currently represented in the literature.

Perspectives and implications
An important implication of our research is the identi cation of three categories of indicators and a set of speci c socio-economic determinants of health that can be used to analyse health care equity in highincome countries. As most of the indicators identi ed in this review are not speci c to emergency care settings, it seems possible to study health care equity in other areas of the health care system of highincome countries with similar administrative data-derived indicators, as for example hospitalization, (64,65) ACSCs during the total hospital admission,(66) and wait times.(61) Such information could be useful for policy makers or health equity researchers to ll the gap in data about health care equity within different health care settings, particularly in high-income countries, using available administrative data.
Our ndings suggest that SEDH have a considerable impact on health care equity. The next step would also be to better characterize root causes for differences in emergency care utilization that lie outside the In conclusion, our focus has been solely on SEDH and their associated vertical inequities. It would also be important to analyse the horizontal components of health care equity through social determinants such as race/ethnicity, gender, or place of residence, in order to have a comprehensive picture of health care equity.

Limitations
Our review has some limitations that require consideration. First, the content and quality of administrative datasets are highly variable within countries (sometimes even within regions) and between countries. As such, many of the indicators identi ed in our review might not be available in many health care settings, reducing their generalizability and widespread applicability. However, important equity indicators such as preventable ED visits are frequently used and easily replicable between countries. Secondly, administrative data do not tend to capture detailed clinical information, such as patients' physical or biological data. To address this, further studies of health equity indicators and SEDH using different types of datasets would be helpful for the researchers. Third, in order to de ne the criteria relevant to this review, it was necessary to make many normative choices before data analysis (ie. when de ning equity and SEDH, etc.). As such, these results must be interpreted in the context of the concept of health care equity and the de nitions we used. Lastly, as more than the half of the studies was conducted in the US, the extrapolation of the results should be carefully interpreted.

V. Conclusion
Measuring health care equity should be an integral component of all comprehensive assessments of a health care system's performance. However, in order to measure health care equity, indicators for making such measurements need to be identi ed, as was the goal of this review. Such indicators can be used by researchers and policy makers interested in measuring health care equity through thoughtful selection of the most relevant indicators de ned by the local context and stated objectives. Using a combination of indicators is likely to lead to a more comprehensive, well-rounded analysis of health care equity than using any one indicator in isolation. Though studies analysed focused on emergency care settings, it seems possible to extrapolate these indicators to measure equity in other areas of the health care system.
Meta-analyses focusing on speci c socio-economic determinants of health such as health insurance coverage, income or indices of social deprivation in combination with studies analysing factors that could in uence the use of emergency care related to social inequalities would help to further characterize root causes of ongoing health care inequity in health care systems around the world.