More than Words: Race, Language Preference, and Restraint Use in the Intensive Care Unit

The use of physical restraints in the intensive care unit (ICU) is controversial at the minimum. Although intended to protect patients and staff (1), physical restraints often do not achieve their purpose (2) and expose patients to harm through increased risk of delirium, longer ICU stay, and post-traumatic stress disorder among survivors (3, 4). By curtailing patient autonomy, restraints also undermine dignity and may be perceived as coercive, diminishing trust in clinicians and the health system (5, 6).

These harms can be amplified for marginalized patients. If patients of Black race or those who prefer a non-English language—populations that may already have reasons to mistrust the healthcare system—are restrained at higher rates, trust can erode further, jeopardizing shared decision making about issues such as end-of-life care in the ICU (5, 7). Unfortunately, constructs of “agitation” and “imminent danger”—justifications for restraint use—are susceptible to individual and systemic biases that lead to inequitable care. On the patient’s end, unfamiliarity with the English language can limit understanding of treatments, leading to agitation. On the provider’s side, cognitive biases against patients from minoritized racial groups can exaggerate the perception of danger, and language barriers can thwart verbal deescalation attempts that might prevent restraint use (8). It is therefore plausible that patients belonging to minoritized racial or ethnic groups and those preferring languages other than English are restrained more frequently.

In this issue of AnnalsATS, McGowan and colleagues (pp. 1729–1737) test this hypothesis in a large, retrospective cohort using electronic health record data from five ICUs in a major health system over the course of a decade (2013–2022) (9). Among 15,460 admissions of nonintubated patients, nearly one in five had a restraint order. Patients identifying as Black and those who preferred Chinese or a language other than Chinese, English, or Spanish were more likely to be restrained than those identifying as White and those with English as the preferred language (odds ratio [95% confidence interval], 1.51 [1.27, 1.79] for Black vs. White; 1.57 [1.31, 2.87] for Chinese vs. English, and 1.60 [1.36, 1.89] for other vs. English). There was no modification of the effect of race and language on each other and no collinearity between the two, suggesting independent effects of each on restraint use.

Whenever differences by race or language preference are identified, the key question is whether they constitute a disparity. The Institute of Medicine defines healthcare disparities as differences that persist after accounting for clinical needs and patient preferences (10). Because patients and families are unlikely to “prefer” physical restraints, rigorous adjustment for clinical needs is crucial to determine whether observed differences represent inequity rather than appropriate care. For example, Black patients often present with greater sepsis severity (11), which may necessitate vasopressors or renal replacement therapy requiring arterial lines or large-bore central venous catheters that might justify restraint use. The present study addresses this issue in several ways. First, by focusing on nonintubated patients, it removes the most common reason for restraint use in the ICU—the endotracheal tube—from consideration (2, 4). Second, it uses treatments as proxies for invasive devices that commonly require restraints: dialysis for large-bore central venous catheters, vasopressors for central venous and arterial catheters, and tube feeds for nasogastric tubes. These adjustments strengthen causal inference by isolating the effect of race, ethnicity, and language preference on restraint use unexplained by need for device protection. Finally, sensitivity analyses accounting for body mass index and excluding hospitalizations during the COVID-19 pandemic speak to the generalizability of these findings beyond select populations or periods.

The authors rightfully acknowledge study limitations. First, delirium—both a trigger for restraints and a consequence of being restrained—can function as a potential confounder or mediator in the association between race, ethnicity, language preference, and restraint use. Differences by race and ethnicity in sedation practices, preexisting cognitive impairment, and illness severity can influence delirium prevalence (11–13), and language preference (and consequently patient–provider language concordance) can alter delirium detection (14). Because delirium was assessed only at the ICU admission level in this study and not matched to shift-level restraint use, the possibility of residual confounding and reverse causation remains. Second, sedation, another factor influencing restraint use (2), could not be characterized in the absence of Richmond Agitation-Sedation Scale scores. Third, the use of a restraint order as a proxy for actual restraints may reduce specificity, although it is unlikely to miss patients who were restrained, given strictness around physician orders for restraints. Finally, the single health system setting limits generalizability.

Despite these limitations, the findings of increased restraint use among patients of Black race and those with non-English and non-Spanish language preference align with prior work and are unlikely to be artifacts (15, 16). Combined with a recent ICU study that found nurse–patient language concordance and not Spanish language preference itself to be associated with restraint use (16), the present study highlights that interpreter availability, staff language skills, and cultural familiarity can mitigate restraint use. That Spanish language was not associated with higher restraint odds in the present study suggests that local capacity matters: When interpreter services are scarce or specific dialects (e.g., Cantonese) are unavailable, clinicians may resort to restraints when verbal deescalation in the patient’s preferred language might have contained agitation. The racial disparity for Black patients, independent of language, echoes broader evidence that minoritized patients are more often labeled as agitated or “combative” and are more likely to receive coercive measures across care settings (15).

The present study highlights the need for improved practices around safe and equitable restraint use in the ICU. First, in the absence of data establishing the superiority of restraints to other approaches in preventing removal of invasive devices or promoting safety, it emphasizes the importance of minimizing exposure to restraints through approaches such as verbal and nonverbal deescalation measures, shorter default duration of restraint orders, and nudges to attempt alternative strategies before ordering restraints. Second, it implores us to closely examine language justifying restraint use, such as “fighting the ventilator” or “pulling lines,” that can shift blame to patients, normalizing restraints as the “safe” response and simultaneously introducing room for our cognitive biases in whom we perceive as “aggressive” or “threatening” (17). Third, it urges us to integrate a public health race praxis at the provider level and to create space for reflection and analysis necessary for organizational culture change to mitigate racial disparities stemming from systemic racism. Fourth, it highlights the need for mechanistic research investigating the relationship between restraint use and shift- and provider-level language concordance, family engagement, delirium assessment, and use of interpreter services to direct interventions to appropriate targets. As these interventions are designed, tracking changes by measuring language-concordant encounters, interpreter response times by language, hours spent in restraints, and impact on patient outcomes can inform cycles of change and quality improvement.

In summary, this study identifies an important opportunity to improve quality and equity in ICU care. Restraints can be lifesaving in rare moments, but when historically marginalized patients experience a greater burden of their use not explained by clinical need, we must interrogate the systems and cognitive biases driving clinician decisions. Narrowing indications for physical restraints, defaulting to alternative approaches, and measuring processes and outcomes with an equity lens are crucial to prevent unnecessary harm from this imperfect tool.