The impact of racism on subsequent healthcare use and experiences for adult New Zealanders: a prospective cohort study

This study demonstrates using a high-quality prospective design that ever experiencing racism is linked to subsequent higher unmet need, lower satisfaction with healthcare and generally poorer experiences with visits to a GP. It is important to understand these findings in the context of ethnic inequities in exposure to racism, whereby people from non-European ethnic groups (e.g. Māori, Pacific, Asian) report higher experience of racial discrimination than people from the broad European grouping (Additional file 2; [13]) alongside the potential contribution of this to inequities in healthcare use and experience.

These findings are consistent with the broader international literature that shows links between racism and negative healthcare measures, particularly measures of unmet need for healthcare and poorer satisfaction and patient experiences [9]. This study addresses a gap in current literature, with most studies being cross-sectional [9]. An important strength of this study is the longitudinal prospective design, particularly the measurement of exposure to racism (T1:2016/17 NZHS) prior to healthcare measures (T2: 1–2 year follow-up), adding to the strength of evidence that racism is a potentially causal factor in determining poorer access to healthcare and poorer quality of care.

While this study demonstrates a prospective association between experience of racism and adverse healthcare measures of unmet need, lower satisfaction and worse experiences, it does not show the specific mechanisms by which racism impacts healthcare. As a complex system, the mechanisms by which racism may impact healthcare are likely to be multiple, with both direct and indirect pathways that can operate at both institutional and individual provider levels [9, 21, 30], with the potential for racism to “… impact the quality of healthcare, how individuals access and use health services, and experiences and perceptions of healthcare” [9]. For example, experiences of racism (both within healthcare and more broadly) may lead to higher unmet need via mechanisms such as socioeconomic status, increased health need and reduced access to healthcare providers [9, 21, 31]. In this study, we adjusted for health status (e.g. self-rated health or mental health) in the examination of racism and unmet need for healthcare, and still show an independent association, suggesting mechanisms other than increased health need. We also demonstrated an association between experience of racism and lower confidence and trust in GPs, which has the potential to affect access to and experience of healthcare. Benjamin [32] reminds us that measures of trust among individuals reflect the “trustworthiness” (or not) of providers/systems and in NZ is supported by extensive evidence of ethnic healthcare inequities [33].

The mechanisms linking experience of racism and lower satisfaction and negative experiences are similarly likely to be complex and multidimensional, encompassing poorer quality care as well as patient perceptions of care [9, 30]. For example, discriminatory treatment may lead to lower satisfaction and poorer patient experiences; racism experienced outside the health system may lead to lower trust in healthcare providers; experience of racism may influence patient perceptions of care as negative; and, experience of racism more broadly can elicit stereotype threat that can negatively influence healthcare encounters [9, 30].

The complex and multifaceted potential mechanisms by which experience of racism can impact healthcare cannot be determined from quantitative studies like this one that only include measures of exposure and outcome. These individual experiences also exist within the broader system of racism that operates at institutional and societal levels, as well as at the level of individuals. We are conscious of not wanting to over-interpret findings or focus too heavily on individuals, particularly those exposed to racism, as this may inadvertently shift the framing of the problem to the level of individuals exposed to racism rather than the broader system of racism, which has previously been described as deflecting “discourse from the ways in which group stereotypes and institutional arrangements are products of racism and serve to reinforce racial inequality” [31, p202]. Instead, interventions that address the distribution of power implicit in racism require that anti-racism actions seek to eliminate racism and that health systems and providers understand their roles and responsibilities in relation to this. Such interventions also need to operate at multiple levels [34], and across different sectors to realise meaningful and long-lasting change.

In addition to the strengths of a prospective study, this study also adds to the limited international studies undertaken at a national population level, with sufficient data to understand the impacts of racism for Indigenous populations. Methodologically, the propensity matching approach provided an efficient and effective means to minimise potential confounding. However, this approach is limited to adjusting for pre-planned variables that existed in the baseline NZHS data, and there is still potential for residual confounding if confounders were inadequately measured or not measured at all. While there may be other factors impacting on access to care that do not relate to experience of racism, our design for comparison of those exposed/unexposed to racism likely accounts for major confounders that might be expected to drive differential access to healthcare. Secondly, the multimodal survey design allowed us to maximise response rates, although opening the potential for differential responding by modality. This was mitigated prior to follow-up data collection through careful questionnaire design to ensure alignment of questions and consistent handling of responses across modalities e.g. in-built skips in online and telephone modalities were applied to paper questionnaire responses. The distribution of modality was also similar across exposure groups. Considering outcome variables ordinally also allowed for examination of more subtle impacts of racism on variables where this was appropriate. This analysis of HSU outcomes with ordinal logistic regression returned similar patterns to the binary outcome analysis, but avoided the potential to be determined by (sometimes) arbitrary decisions about where to place the cut-point in making a binary categorisation (e.g. the boundary between satisfied and dissatisfied response categories). This has additional utility as the distribution of these outcome variables may reflect a general downgrading of responses in the exposed group (shifting more towards lower satisfaction) without necessarily crossing these arbitrary boundaries.

There are several limitations that should be considered in interpreting these findings. Firstly, the timeframe to follow-up was relatively short, between one to two years. It is possible that the impacts of racism on healthcare measures may be different at different lead-times. This is seen in health status outcomes, where the lead-time of racism on mental health outcomes is shorter than physical health outcomes [9, 35]. In addition, the follow-up time is for a composite one to two year follow-up period, rather than a fixed interval, yielding variable follow-up time among participants (although this is not different between exposed and unexposed). We were unable to undertake analyses stratified by ethnicity as participation across ethnic groups was largely driven by who had participated in the NZHS from which we sampled, limiting our ability to report results further stratified by ethnicity. We note however that a previous cross-sectional study suggests that different strengths of associations were not a major feature for individual ethnic categories [13]. Furthermore, the ethnic categories that we stratified for in sampling do not represent individual ethnicities (apart from Māori) but rather are broad groupings of diverse populations e.g. the Pacific, Asian and European categories are made up of multiple ethnic groups. Additionally, it is unclear how these relationships may be affected by the COVID-19 pandemic and the increase in telehealth consultations that occurred after the data collection period of this study. Further research could examine whether these associations are found in analyses of recent experiences of racism (rather than ever) and for racism experienced in healthcare settings specifically. This was not possible in this study due to relatively small numbers for these potential analyses, and the propensity score sampling step matching groups based on ever experience of racism. It was also not possible to examine impact of experience of racism by a health professional separately to other forms of racism due to small numbers of participants reporting this exposure. Cross-sectional analysis of the NZHS has shown racism by a health professional to have stronger associations with negative healthcare measures [31], and so we might have expected a similar phenomenon in prospective results as well.