Incidence and outcomes of critical illness in Indigenous peoples: a systematic review and meta-analysis

The search strategy was developed in consultation with the Alberta Research Centre for Health Evidence (ARCHE) at the University of Alberta and conducted by an information specialist. The search strategy included the following two groups of terms (key words with similar characteristics): ‘Indigenous Peoples’ and ‘critical care’ (Additional file 1: File S1). We systematically searched the following electronic databases from inception to October 2022: Ovid MEDLINE/PubMed, Ovid EMBASE, Google Scholar, and Cochrane Central Register of Controlled Trials (Additional file 1: File S1). We limited results to human studies that were published in English. We also complemented this search by scanning potentially relevant websites for grey literature (National Collaborating Centre for Indigenous Health, First Nations Health Authority, Canadian Institutes of Health Research Institute of Indigenous Peoples’ Health, National Association of Friendship Centres, the First Nations Information Governance Centre, Métis Nation of Alberta; Australia Institute of Health and Welfare [Indigenous Health and Wellbeing]; New Zealand Ministry of Health [Manatū Hauora and Māori Health]; United States Department of Health and Human Services [Indian Health Service]). Lastly, we hand-searched bibliographies of included studies and relevant reviews for additional citations. We exported bibliographic records into EndNote X9 (Thomas Reuters, Philadelphia, PA, USA) database for screening and removal of duplicate citations.

Retrospective and prospective observational cohort studies, case series reporting aggregate data on > 100 patients, arms of clinical trials (e.g., usual care, control, or placebo arm), and analytical data from grey literature reports of Indigenous adults (≥ 18 years) either without critical illness (i.e., general population) or with critical illness (i.e., admitted to an ICU) were all eligible for inclusion. Descriptive studies, cross-sectional studies, case-reports, and articles that do not present original data (e.g., editorials, commentaries, narrative reviews) were excluded.

Our primary outcome measures were ICU admission and ICU mortality. For ICU admission, we only included studies with a non-Indigenous ‘general population’ or ‘hospitalized’ comparator. Studies in the general or hospitalized population that did not have a non-Indigenous comparison group were only included if they also had information on the following factors within the Indigenous population: age, sex, and illness acuity. For ICU mortality, we included studies with and without a non-Indigenous comparator. Secondary outcomes of interest included ICU length of stay, ICU re-admission, receipt IMV, duration of IMV, receipt of tracheostomy, receipt of vasoactive support, duration of vasoactive support, acute kidney injury (AKI), receipt of renal replacement therapy (RRT), duration of RRT, and quality of life.

Both abstract title (Level 1) and full text (Level 2) reviews had pre-determined eligibility criteria. Level 1 criteria were broader than Level 2 to ensure all potentially relevant studies were captured. Two reviewers independently, and in duplicate, reviewed the titles and abstracts of the retrieved citations. Full-text articles were obtained for all titles and abstracts identified by one or both reviewers as potentially relevant. Subsequently, two reviewers independently, and in duplicate, identified full-text articles that met the inclusion criteria. Any eligibility disagreements between reviewers were resolved through discussion and a third reviewer with clinical expertise, when required (Fig. 1) [22].

For each study, one reviewer extracted data into a standardized electronic form, while a second reviewer verified the extracted data. Discrepancies at any stage were resolved through discussion and referred to the third reviewer, when required.

Two reviewers independently assessed the risk of bias (RoB) using the Newcastle–Ottawa Scale (NOS) across the following domains: selection of the exposed and unexposed cohorts, comparability of the cohorts, and outcome ascertainment (Additional file 1: File S2) [24]. We required that groups were comparable, or controlled for age, sex/gender, and level of acuity, at a minimum. We summed scores the primary outcomes (ICU admission and ICU mortality); studies were rated as high risk of bias (< 4/9), moderate risk of bias (4–6/9), or low risk of bias (> 6/9). Any disagreement in ratings between reviewers were resolved by discussion or by consulting a third reviewer.

We classified studies first according to their risk of bias for the primary outcomes of ICU admission and ICU mortality. There was adequate clinical and methodological homogeneity to perform a meta-analysis for ICU mortality, ICU length of stay, and receipt of IMV (Review Manager version 5.4, the Nordic Cochrane Centre, the Cochrane Collaboration, Copenhagen, Denmark). We did not pool data for the outcome of ICU admission, due to substantial differences in the presentation of data across the four studies that reported this outcome. Therefore, findings for this outcome are summarized using a descriptive synthesis approach for systematic reviews [25].

Dichotomous outcomes (ICU mortality and IMV) were analyzed using risk ratio and random effects, whereas continuous outcomes (ICU length of stay) were analyzed using mean difference. Data that were reported as medians and interquartile ranges were imputed as means and converted to standard deviations, respectively. Denominators reported as number of admissions were imputed as number of people, where appropriate, to permit pooling. Heterogeneity was quantified using the I² statistic and was explored using between-study subgroup analyses (i.e., by case-mix). We also performed sensitivity analyses to understand the effects of variably defined exposures and/or outcomes. For meta-analyses that included at least eight studies of varying size (i.e., ICU length of stay), we tested for small study bias by interpreting funnel plots and statistically using Egger’s test [26]. Statistical analyses were performed using Stata SE version 13.1 (Stata Corp, LP, College Station, TX).

Two reviewers independently assessed the certainty of evidence for the primary outcomes using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach [27, 28]. As the best evidence for prognostic factors originates from observational studies [29], evidence from these started at high certainty, and were rated down for concerns about risk of bias, inconsistency, indirectness, imprecision, and other concerns. Inconsistencies between reviewers were resolved by discussion or the involvement of a third reviewer, if needed. We used “probably”, “may be”, or “uncertain” to reflect level of certainty in the evidence based on GRADE of moderate, low, or very low, respectively.

In addition to scientifically appraising evidence through a Western lens, we aimed to describe and evaluate the quality of research from an Indigenous perspective. We used the 14-item Aboriginal and Torres Strait Islander Quality Assessment Tool (ATSI QAT) to appraise included studies (Additional file 1: File S3) [30]. The ATSI QAT items focus on understanding, from an Indigenous perspective, whether the research responds to a community need, has community and Indigenous leadership engagement, has negotiated agreements for access to and protection of Indigenous intellectual and cultural property, provides benefits to Indigenous participants and communities, and enables Indigenous ownership of data collection and management [30]. Two reviewers independently assessed the included studies using the ATSI QAT, with each item assessed as “yes”, “partial”, “unclear”, or “no”. Any discrepancies between reviewers were resolved by discussion or the involvement of a third reviewer. We considered studies to have a higher quality, from an Indigenous perspective, if they had a higher number of “yes” assessments. We summed the assessments for the 14 ATSI QAT items for each study. We ranked the 15 studies to identify the highest and lowest quality from an Indigenous perspective.

Ethics approval was not required for this study. We invited our Indigenous Peoples and Critical Care Advisory Committee (IPCCAC), which is comprised of Métis (n = 2) and First Nation (n = 3) individuals, to review, appraise and offer perspective to the findings of this systematic review. This exchange was facilitated through dialogue and aimed to consider alternative interpretation and incorporation of feedback from the IPCCAC. This systematic review represents a foundational project within a larger program of culturally appropriate, respective, and mutually beneficial work aimed at understanding the lived experience of Indigenous Peoples with critical care, to be co-designed with Indigenous partners.

The definition and description of ICU admission across studies was heterogeneous: number of ICU admissions rather than number of patients [31]; estimated ICU incidence with age-adjusted incidence rate ratio [33]; annual incidence of ICU admission with relative risk [39]; and estimated ICU admission with incidence ratio [40] (Additional file 1: Files S4 and S5; Table 2). Among cohorts that compared general populations or hospitalized patients, there was suggestion of increased risk of ICU admission for Indigenous when compared with non-Indigenous populations (low certainty evidence). In the study by Laupland et al., the annual incidence of critical illness among Indigenous Peoples was significantly higher (620.6 admissions per 100,000 population) compared with the general Canadian population (302.6 admissions per 100,000 population) (RR 2.1, 95% CI, 1.78 to 2.35; P < 0.0001), which was consistently observed across all age groups [39]. Three studies examined ICU admission among specific patient subgroups, including sepsis [31], pandemic H1N1 influenza A [33], and trauma-related injuries [40]. In these examples, Indigenous patients had higher incidences of ICU admission compared to non-Indigenous patients [31, 33, 40].

Nine studies reported on ICU mortality [34, 35, 38, 39, 41‐45], of which, eight were pooled (Fig. 2; Table 2; Additional file 1: File S6A and S7A). We found no statistical difference between Indigenous and non-Indigenous populations for ICU mortality (8 studies; RR 1.14, 95% CI, 0.98 to 1.34; absolute RD 16 more per 1,000; 95% CI, 2 fewer to 40 more; I² = 87%; low certainty evidence), regardless of clinical indication for admission. Two studies reported on patients admitted emergently (i.e., unplanned) or electively to ICU [35, 44]. Among emergent patients, we observed a significantly higher ICU mortality among the non-Indigenous patients (2 studies; RR 1.18, 95% CI, 1.03 to 1.35) (Additional file 1: File S6A).

The study by Maiden et al. reported on a subgroup of obstetric critically ill patients aged 15–49 years old and was not included in any of our pooled analyses [41]. In this subgroup, we found no statistically significant difference in ICU mortality between Indigenous Australian Torres Strait Islander patients in Australia (RR 1.11, 95% CI, 0.51 to 2.40) or Māori (Indigenous) patients in New Zealand (RR 1.16, 95% CI, 0.32 to 4.26), compared to non-Indigenous patients (Additional file 1: File S7A) [41].

Twelve studies reported on ICU length of stay [32, 35‐45] (Fig. 3; Table 2; Additional file 1: File S6B and 7B). Two studies were omitted from pooled analysis, due to one reporting on obstetric critically ill patients [41] and the other not reporting a measure of variance [39]. We found no significant difference between Indigenous and non-Indigenous populations for ICU length of stay (10 studies; MD 0.03, 95% CI, -0.22 to 0.28; I² = 95%), regardless of clinical indication for admission (Fig. 3). We found no evidence of small study bias for ICU length of stay (Egger’s test, p = 0.19; Additional file 1: File S8).

Seven studies reported on receipt of IMV [34, 36, 39, 40, 42‐44] (Fig. 4; Table 2). We found that non-Indigenous patients were significantly more likely to receive IMV in ICU compared to Indigenous patients (7 studies; RR 1.10, 95% CI, 1.06 to 1.15; I² = 81%). This was similarly shown in sensitivity analyses of subgroups (Additional file 1: File S6C and S7C).

We were not able to perform pooled analyses on other pre-specified secondary outcomes (Additional file 1: File S4). Kidney outcomes were variably reported across four studies: RRT [34]; AKI and AKI treated with RRT, among patients undergoing redo aortic and/or mitral valve surgery [37, 43]; and AKI on day 1 [39]. Hanson et al. was the only study reporting on use of vasopressors [34]. There was no difference in vasopressors use between Indigenous and non-Indigenous patients. Two studies reported rates of ICU readmission. Secombe et al. found Indigenous patients were more likely to experience readmission to ICU in subsequent hospitalizations compared with non-Indigenous patients [43]. The study by Dunlop et al. compared Indigenous and non-Indigenous patients with kidney failure receiving chronic RRT [32]. While the reported rate of ICU admission was higher among Indigenous compared with non-Indigenous patients receiving chronic RRT, rates of ICU readmission within the index hospitalization were similar [32]. No studies reported on the receipt of tracheostomy or on quality of life among survivors.

We performed a rigorous systematic review and evidence synthesis to describe the incidence of critical illness and associated outcomes among Indigenous compared with non-Indigenous populations. Importantly, we found that Indigenous consultation, involvement, and perspectives were rarely considered in the design, methodology or interpretation of the results across these studies. Few studies compared the incidence of critical illness among Indigenous and non-Indigenous populations; however, among these, Indigenous populations were generally found to have higher rates of critical illness and receive support in ICU settings comparatively [31, 33, 39, 40]. Our evidence synthesis did not find differences in ICU mortality between Indigenous and non-Indigenous populations. While our evidence synthesis suggested that Indigenous critically ill patients had shorter stays in ICU and were less likely to receive IMV, inferences from these findings may be limited due to significant heterogeneity across studies, likely due to bias and residual confounding.

An important consideration in systematically evaluating the published literature is to not only use established Western methodologies, but also appraise the literature through an Indigenous perspective. As such, we applied the validated ASTI QAT to qualitatively appraise the included studies through an Indigenous lens. We found that the current understanding of Indigenous Peoples experiences with critical illness and with critical care is limited and poorly characterized. Importantly as well, the existing literature has largely adopted a Western research approach. Moreover, the existing literature appears to have largely failed to seek Indigenous consultation and perspectives in the research process, has not adopted co-design or Indigenous research methodology, has not provided details on Indigenous Data Governance (such as Ownership, Control, Access, and Possession (OCAP™), has not sought an Indigenous lens for interpretation, and has not commented on whether the research output had any direct impact on Indigenous communities [30, 46].

As such, our evidence synthesis should be contextualized with prior work that has focused on Indigenous Peoples experience in other acute care settings, particularly, the emergency department (ED). A population-based study in Alberta, Canada found that First Nations Peoples were three-fold more likely to visit the ED compared with non-First Nations people, despite only comprising 4% of the total population [47]. These authors further showed that status as First Nations was associated with lower odds of receiving higher acuity triage scores across several different diagnoses and conditions upon presentation to the ED [48]. Further follow-up work has suggested that overt systemic racism and stereotyping of First Nations patients occurs in the ED [6]. Numerous barriers to care were identified, including issues related to communication, health system access (e.g., access to primary care), and the cultural safety of the ED environment. While this may partly explain the higher rate of ED utilization among First Nations Peoples, this does not account for the enduring systemic cultural or structural barriers propagating health inequity among Indigenous patients (i.e., triage acuity scores).

Similar findings have been observed in EDs in Australia and New Zealand. A systematic review found that Aboriginal and Torres Strait Islanders visited EDs twice as often as non-Indigenous Australians, with Indigenous patients also more likely to leave the ED prior to being evaluated [49]. In a cohort study from New Zealand, Māori patients were found to have worse outcomes after visiting the ED, including higher mortality and ED re-presentation, compared with non-Indigenous patients [50]. The authors concluded that these health inequities were not driven by differences in process measures (i.e., assessment and disposition times) in the ED [50]. The findings of our evidence synthesis may align with these experiences in ED settings. The observed shorter ICU stays and lower rates of receipt of IMV among Indigenous compared with non-Indigenous patients raises important questions of whether similar issues of systemic and structural racism and pedagogy exist in critical care settings.

First, our systematic review is strengthened by a priori publication of a protocol, inclusion of a rigorous and comprehensive peer-reviewed search strategy, and systematic evaluation of the quality and risk of bias of included studies, all using established Western methodologies [23]. Second, we applied the ASTI QAT to qualitatively appraise the included studies for an Indigenous perspective. Third, we further invited members of local Indigenous communities, who formed an Advisory Committee, to offer perspective and an Indigenous-specific lens to the co-design and to the interpretation of our findings.

Our systematic review also has limitations to consider. First, the definition of “Indigenous” and the availability of “identifiers” in health administrative varied substantially, driven by differences in legal definitions across countries and by academic discussions on culturally appropriate terminology. Indeed, the lack of suitable or validated identifiers for Indigenous Peoples in health data represents a barrier to understanding the impact of structural racism, complex post-traumatic stress disorder and inter-generational trauma on health access and outcomes [13]. Second, as shown by our analyses of the ATSI QAT scores, included studies were largely led by non-Indigenous researchers using Western methodologic approaches, which may have contributed to implicit biases that impact both the analysis and interpretation. Third, studies were appraised as being at low-to-moderate risk of bias, all studies were focused in only three high-income countries, were all observational, and showed marked heterogeneity across outcomes of interest. As such, generalizability is limited, and any inferences and interpretation should be conservative. Further, health systems and access likely differ across the included studies (Australia, New Zealand, and Canada); therefore, we have been cautious in comparing the findings from the different countries. Lastly, we acknowledge that the general population comparison groups were likely highly variable and heterogenous among studies. However, it was not within the scope of this systematic review and meta-analysis to tease out these effects.

Our evidence synthesis would strongly imply there is a narrow and incomplete understanding of Indigenous Peoples risk of critical illness and their experiences with critical care (i.e., ICU environment). The scope and magnitude of health inequities in access to ICU support and outcomes after critical illness, if existing, remains poorly described and represents a barrier to action. These observations imply further work is urgently needed. Moreover, this work would align directly with key recommendations for health within the Truth and Reconciliation Commission of Canada: Calls to Action report and further honors the principles of the United National Declaration on the Rights of Indigenous Peoples [13, 51]. Ideally, Indigenous together with non-Indigenous researchers and their communities would partner to co-develop and co-learn to better characterize and understand Indigenous Peoples’ (i.e., patients, families, and communities) experiences and outcomes with critical care, to identify knowledge and care gaps, and to work toward ensuring culturally appropriate and safe space.