Socioeconomic Position and Health Outcomes Following Critical Illness: A Systematic Review

Objectives: Systematically review evidence examining association between preadmission socioeconomic position and physical function, health-related quality of life and survival following critical illness. Data Sources: Four electronic databases (MEDLINE, Embase, CINAHL, CENTRAL) and personal libraries were searched. Reference lists of eligible articles were cross-checked. Study Selection: Primary quantitative studies reporting association between socioeconomic position and physical function, health-related quality of life, or survival of adults admitted to the ICU were included. Data Extraction: Performed by two reviewers independently in duplicate using a prepiloted data extraction form. Quality appraisal was completed by two reviewers independently in duplicate using standardized algorithms and checklists. The Preferred Reporting Items for Systematic Reviews guidelines were followed. Data Synthesis: From 1,799 records, 10 studies were included, one examining association of socioeconomic position with health-related quality of life and five with survival. Four studies accounted for socioeconomic position in survival analyses. Patients with lower socioeconomic position were found to have higher ICU, in-hospital, 30-day, and long-term mortality and lower 6-month Short Form-12 Mental Component Summary scores. No articles examined socioeconomic position and performance-based physical function. Notable variability in methods of socioeconomic position assessment was observed. Conclusions: Lower socioeconomic position is associated with higher mortality and lower 6-month Short Form-12 Mental Component Summary scores following critical illness. Effect on performance-based physical function is unknown. We make recommendations for consistent socioeconomic position measurement in future ICU studies.

I n the years following critical illness, patients experience poor physical function and health-related quality of life (HRQoL) (1)(2)(3)(4)(5). To date no interventions delivered, including physical rehabilitation programs, report sustained improvements in long-term outcomes (6)(7)(8)(9)(10)(11). In an attempt to identify factors influencing recovery and response to interventions, rehabilitation trialists have considered the importance of patient characteristics, specifically chronic disease burden (12,13). It is hypothesized the burden of premorbid health status, exacerbated by acute illness severity (14), results in poorer physical function (4,13), HRQoL (12,15), and lower physical activity (16) after critical illness. There is growing support that other preadmission patient characteristics including older age (17), cognitive impairment (18), frailty (19), poor physical function, and muscle strength (20) may be as important as acute illness severity in determining long-term outcomes.
Social and economic factors that influence our position within society (socioeconomic position; SEP [21]) may also be important. SEP is a broad term and can be assessed according to our socioeconomic status (SES) or deprivation. Deprivation describes a lack of resources or opportunities deemed customary by society (22,23), whereas SES reflects our social and economic ranking within society (24). Deprivation and SES can be assessed according to various indicators with education, occupation, and income used most frequently (21,(24)(25)(26)(27). However, SEP assessment permitting local, national, and international comparison (22,27) with socioeconomic categories that remain consistent over time is lacking (22,(27)(28)(29).
The social gradient (30) describes the phenomenon where lower SEP groups experience poorer health and is driven by economic factors, social influences, and health risk factors (behavioral, psychosocial, environmental) (30,31). Lower SEP is associated with poorer HRQoL, functional recovery, and death in chronic obstructive pulmonary disease (COPD) (32)(33)(34)(35) and stroke (36)(37)(38). Given the social gradient observed in stroke and COPD, it may be important to establish if a social gradient exists in critical illness to inform delivery of targeted interventions to improve recovery. Therefore, we conducted a systematic review to explore association between preadmission SEP (specifically SES and deprivation) and physical function, HRQoL, and survival of adults across the critical illness recovery continuum (ICU and beyond).

Registration
Prior to registration with the International Prospective Register of Systematic Reviews (PROSPERO, CRD42017084114 https://www.crd.york.ac.uk/prospero/display_record. php?RecordID=84114), the Cochrane Database of Systematic Reviews and PROSPERO were searched to ascertain no similar systematic review had been conducted.

Data Sources
Four electronic databases (MEDLINE, 1950-2018, Embase, 1947-2018, CINAHL, 1937-2018, and CENTRAL, July 2018) were searched by one reviewer (J.R.A.J.) with the last search run July 10, 2018. Reference lists of included studies were cross-checked, and authors' personal files reviewed for additional references.

Search Strategy
One reviewer (J.R.A.J.) developed the search; no restrictions were applied. Search strategies are provided in the Supplemental Digital Content (http://links.lww.com/CCM/E473).

Study Selection
Primary quantitative studies reporting association between preadmission SEP and physical function, HRQoL or survival of adults admitted to ICU were eligible (Table E1, Supplemental Digital Content 1, http://links.lww.com/CCM/E473). One reviewer (J.R.A.J.) initially reviewed search results to remove duplicate and nonrelevant material. Two independent reviewers (J.R.A.J., J.L.W.) then screened remaining titles and abstracts against the eligibility criteria. When there was insufficient information to determine eligibility, full texts were retrieved. One reviewer (J.R.A.J.) retrieved full texts and contacted authors of conference abstracts to determine publication status. Two independent researchers (J.R.A.J., J.L.W.) screened full texts in a similar manner. When consensus on eligibility could not be reached, a third independent reviewer made the final decision (S.B., L.D.)

Data Extraction
Data extraction was performed independently in duplicate by two reviewers (J.R.A.J., J.L.W.) using a standardized prepiloted form. Disagreements were resolved by consensus and a third reviewer arbitrated if necessary (D.M.G.).

Data Items
Data extracted included study characteristics, participant characteristics, SEP assessment, outcomes (physical function, HRQoL, survival), and study results (association between SEP and outcomes of interest).

Quality Appraisal
Quality appraisal was completed independently in duplicate by two reviewers (J.R.A.J., B.C.) using standardized checklists for nonrandomized study designs (39). Disagreements were resolved by discussion; a third independent reviewer was not required.

Data Synthesis
Data were descriptively summarized and qualitatively synthesized. Meta-analysis was inappropriate due to the heterogeneity of SEP assessment and mortality time-point assessment variation.

Study Selection
Database searches yielded 1,788 records and 11 additional studies from personal files. Study selection process and exclusion reasons are presented in Figure 1. Full texts were required for 51 conference abstracts where 18 articles were retrieved in the search yield (duplicates) or located without needing to contact authors. For the remaining abstracts (n = 33), 29 authors were contacted, 21 responded resulting in inclusion of one study. Four authors had no currently active contact details.
Three authors were contacted to confirm age threshold of included subjects resulting in inclusion of one study. One author confirmed deprivation assessment was determined from zip codes, resulting in inclusion. In total, 10 studies satisfied the eligibility criteria each evaluating unique cohorts.

SEP and Physical Function
No articles were identified specifically examining SEP and performance-based physical function.

SEP and Survival
Four of five studies (41-43, 46, 48) (Table 3) for participants with lower SEP. One study reported no difference in deprivation scores between survivors and nonsurvivors 1 year post ICU discharge (42).
Deprivation and SES were important variables in the relationship between hospital mortality and insurance status (40), acute illness severity and comorbidity scores (44), and 6-month mortality and alcohol dependence (47) ( Table 3). When the relationship between race and post-ICU mortality was examined, lower income and education level were no longer significant predictors when the model included comorbidities (45).

DISCUSSION
Our systematic review examined association between preadmission SEP and outcomes of adults admitted to ICU and to our knowledge is the first to have done so. Ten articles were included with the majority focusing on survival. Only one study examined HRQoL, and none performance-based physical function. Lower SEP was associated with higher ICU (46), in-hospital (48), 30-day (43), and long-term (41) mortality and lower 6-month SF-12 MCS scores (12). SEP was identified as a key variable in the relationship between mortality and insurance status (40), alcohol dependence (47), acute illness severity, and comorbidity scores (44).

Significance of Results and Future Directions
Patients with lower preadmission SEP have higher mortality across the critical illness continuum (41,43,46,48), signaling a social gradient exists. Given similar findings in other patient populations (35,37), this trend may not reflect the critical illness insult alone. Rather, the poor health trajectory people with lower SEP experience is characterized by gradual accumulation of behavioural, environmental, and psychosocial risk factors prior to the acute insult (31). For example, increased tobacco smoking and lower physical activity (behavioral risk factors) of individuals with lower SEP are associated with higher rates of cardiovascular disease and mortality (poor health trajectory) compared with more socioeconomically advantageous groups   (49) and potentially stem from health literacy issues or psychosocial influences. Exploration of the mechanisms contributing to the relationship between SEP and critical illness outcomes is warranted, to assist in elucidating factors in recovery that can be modified with targeted interventions. One study assessed HRQoL (12), where lower 6-month SF-12 MCS scores for survivors with lower SEP is supported by the higher prevalence of poor mental health in lower SEP groups in epidemiologic studies of general populations (25,50). Specific to recovery from critical illness, financial distress has a direct effect on anxiety and depression symptoms (51). Given the lower 6-month SF-12 MCS scores for survivors with lower SEP (12) did not persist to 12 months, this may indicate a subgroup more vulnerable to mental health distress during their earlier recovery.
Survivors (52,53) and their family members (54) experience joblessness and loss of life savings following critical illness and patients who do not return to work within 6-12 months report poorer HRQoL (52). The results of Griffith et al (12) may be attributable to a transient change in patient or caregiver employment status, where survivors with lower SEP experience greater financial (and mental health) distress which could potentially be resolved with return to work plans.
Higher social integration (55) and social support (51) are associated with improved quality of life after critical illness supporting the hypothesis that social relationships can mitigate negative effects of illness. Griffith et al (12) findings may also reflect a subgroup of survivors needing increased social support early in their recovery as lower SEP is associated with social isolation in epidemiologic studies of general populations (56). The role of family and informal caregivers (57,58) and peer support networks (59-61) is a growing research area, additionally considering SEP in future research on this topic may help decipher how social and economic factors impact recovery from critical illness.
Despite a strong focus on survivorship (62,63) and growing research on importance of preadmission patient characteristics on recovery (4,12,13,(16)(17)(18)(19), no articles were identified specifically examining SEP and performance-based physical function. Although Griffith et al (12) reported no significant association between SEP and the SF-12 PCS HRQoL questionnaire at 6 and 12 months, this is a patient reported measure rather than a performance measure of function and these results are in contrast to those reported in other populations (32)(33)(34). This variation provides sound basis for future examination of SEP and physical function in ICU populations.
Physical inactivity is associated with poor health (64,65), and lower premorbid physical activity is reported for ICU patients with lower SEP (66). This subgroup may continue on a trajectory of poor health, unless addressed with targeted interventions. The reduced pulmonary rehabilitation completion rates for participants with lower SEP (67) provoke contemplation on alternative healthcare delivery systems for these patients and indeed may be of interest to rehabilitation trialists designing programs for survivors of critical illness.

Critique of Method and SEP Assessment Considerations
This review benefits from robust methods for its conduct in keeping with established guidelines (68), including a registered protocol. Searches were comprehensive and screening, data extraction, and quality appraisal conducted in duplicate.
We defined the eligibility criteria to reflect SEP assessment but did not specify measurement instruments to ensure maximum inclusivity of relevant material in this relatively underresearched area. To this end, we observed notable variability in methods employed across included studies, a well-recognized challenge (25,26,69,70), subsequently limiting data synthesis. Different approaches across international healthcare jurisdictions may contribute to this, but our data were insufficient to explore. Comprehensive assessment including multiple SEP indicators is recommended (21,26) and adopted in our methods. However, this criterion resulted in the exclusion of seven articles, three not meeting criteria for deprivation, and four assessing a single SEP indicator.
Assessment of SEP needed to be participant specific, resulting in the exclusion of one study where main supporting family member SES was measured. Highlighting the importance of cultural and contextual consideration of SEP assessment in relation to social and financial support of primary caregivers.
Findings of this systematic review are limited to generalization to higher income countries as reflected by the included studies country of origin. In lower and middle-income countries, admission to ICU and recovery from critical illness can be influenced by unique socioeconomic factors, including limitations upon number of ICUs, staff, and resources (71). Higher mortality rates and lower 6-month SF-12 MCS scores for participants with lower SEP in the current review are specific to the United Kingdom (12,46,48), Europe (43), and Australia (41). The effect of SEP internationally according to economic profile or healthcare system status (privatized or socialized) remains unknown.
Quality appraisal of included studies was conducted using a standardized checklist, with the majority of acceptable quality. However quality appraisal was limited as many criteria were deemed "not applicable" in studies which conducted retrospective analyses (40,41,(43)(44)(45)(46)48), a contributory factor to the overall grading.

Recommendations for SEP Assessment
Examination of SEP and health outcomes following critical illness is an emerging research area; routine and consistent reporting of participant SEP will help advance the field. Including multiple indicators in SEP assessment is recommended with occupation, education, and income reported most frequently (21,26,72). In the absence of consensus on SEP assessment, we made recommendations based on the merits of existing measurement instruments (Fig. 2). Implementation of which is at the discretion of the researcher, as different measurement instruments or indeed other SEP indicators may be more appropriate for the research question or study design.
Several SEP-specific occupation classification tools exist but are country specific (73). Therefore, we recommend the International Standard Classification of Occupations (74) according to the 10 major groups (Fig. 2). These data may not be readily available or if limited ability to interview participants or their proxies, occupational status can be simplified to employed or unemployed (28). The major limitation of occupation data is the exclusion of participants not actively participating in the workforce. We recommend reporting participant education level using the International Standard Classification of Education (ISCED) (75) to enable international comparison. Schnegelsberg et al (43) collected these data from national registries; however, it would not be feasible for researchers to complete the given extensive coding process. Alternatively, ISCED attainment levels ( Fig. 2) (75) could be used to structure participant or proxy interviews; however, we acknowledge interviews may be challenging for some trials.
Assessments of participant income that facilitate international comparison are lacking. Despite this, we suggest researchers report these data as a continuous variable, according to predefined categories or metrics appropriate to the dataset (i.e., tertiles, quartiles). The major obstacle in obtaining income data is participants are unwilling or uncomfortable disclosing information (21,72).
If individual-level assessment is not possible, we recommend using SEP data for the participants' residential area. Data can be acquired for specific SEP indicators, for example, percent postsecondary education from census tracts (44). Alternatively, composite measures encompassing several SEP indicators can be utilized, for example, deprivation indices according to zip code (12,42,(46)(47)(48). These data are for a geographical area and therefore may over-or underestimate the SEP of the participant. Geographical SEP assessments are also contextual, precluding comparison over time and between countries (22,26).

CONCLUSIONS
The findings of this systematic review, uniquely examining the association between SEP and health outcomes of adults admitted to ICU, signal the presence of a social gradient where patients with lower SEP have worse survival outcomes and lower 6-month SF-12 MCS scores. Effect on performancebased physical function remains unknown. Assessment of SEP lacks consistency. Recommendations for SEP measurement in future research are provided which we anticipate will progress the field by identifying contributing factors and informing healthcare delivery system design.