Cognitive Impairment, Anesthesia, and Critical Illness: Learning From the Past to Gain Perspective on the Future

In the present issue of Mayo Clinic Proceedings, Shulte et al⁷ report the results of a longitudinal cohort study exploring the primary hypothesis that established cognitive impairment in elderly patients is associated with an increased likelihood of undergoing procedures requiring anesthesia. The authors’ secondary hypothesis was that cognitive impairment is also associated with increased rates of postoperative ICU admission. To answer this question, the authors linked data on cognitive function from the Mayo Clinic Study on Aging to electronic health record (EHR) data from the Rochester Epidemiology Project. Using multivariable Cox proportional hazards modeling to analyze data from 1977 cognitively normal patients, 387 patients with mild cognitive impairment (MCI), and 72 patients with established dementia, the authors’ primary finding was that compared with patients with normal cognition, those with MCI were equally likely to undergo procedural anesthesia (hazard ratio [HR], 0.98; 95% CI, 0.82-1.16; P=.78), and those with dementia were less likely to undergo procedural anesthesia (HR, 0.50; 95% CI, 0.28-0.89; P=.02). This finding runs counter to the authors’ primary hypothesis, which was that patients with cognitive impairment would be more likely to undergo surgical procedures. In their evaluation of the association between prevalent cognitive impairment and incident ICU admissions, the authors found that both MCI and dementia were associated with increased rates of ICU admission. In post hoc analyses, the authors found that this observation was driven by increases in non–procedure-related ICU admissions, and procedure-related ICU admissions were similar across the patient groups. The study results were qualitatively similar in sensitivity analyses restricted to general anesthetic procedures, repeated anesthetic procedures, and censoring based on longitudinal changes in cognitive status or changes in residency that could affect outcome ascertainment.

This study has several important strengths. First, the authors used high-quality data sets with robust documentation of the exposure (cognitive function) and outcomes (anesthesia and ICU admission), increasing the internal validity of their findings; the Mayo Clinic Study on Aging data set used validated instruments to define categories of baseline cognitive impairment, and the Rochester Epidemiology Project data set relies on EHR data rather than on administrative coding alone. Second, by using a longitudinal cohort study design the authors were better able to propose causal links between preexisting cognitive impairment and incident procedural anesthesia and ICU admission than could be determined from cross-sectional data. Third, the study’s results held across several sensitivity analyses, increasing the level of confidence in the observed association between cognitive impairment and procedural anesthesia or ICU admission.

A few limitations may affect the implications of this study’s findings. First, although the authors used available covariates to institute appropriate risk adjustment models, it is likely that residual unmeasured confounders partially explain the fact that procedural anesthesia was less common in patients with dementia. For example, the EHR data did not include physiologic information from the time of admission that might have informed risk-benefit discussions surrounding surgery. Second, by virtue of the data sets examined, this study included a selected patient population from a geographically restricted area; it is unclear whether these findings would generalize to a more diverse patient population. Third, the distribution of patients was disproportionate across the 3 patient groups, and the small number of patients with dementia makes it difficult to know whether any nonsignificant results were simply the result of insufficient statistical power. Ultimately, however, none of these limitations seriously undermines the primary finding that patients with dementia were less likely to undergo procedural anesthesia and more likely to experience non–procedure-related ICU admission.

What might explain the observed association between dementia and fewer anesthetic procedures? Due to limitations in the available data, the authors’ analyses do not provide mechanistic insight into this question. However, it is easy to imagine how a diagnosis of dementia might affect preoperative decision making. In the face of conditions with both operative and nonoperative management options, patients, families, and physicians may perceive patients with dementia to be at higher risk for postoperative complications and perhaps less able to derive long-term benefits from an operation, leading to a decision not to pursue surgery. This decision-making process may be beneficial to the degree that it results in care that is consistent with patients’ preferences and values; however, inappropriate risk aversion by physicians or families could also impede surgical interventions that would help elderly patients with cognitive impairment.

It is not necessarily clear from this study whether physicians, patients, and family members should find solace in the observation that postoperative ICU admission was not more common in patients with cognitive impairment. An optimistic interpretation of this finding is that, following a process of preoperative patient selection, cognitive impairment did not lead to worse postoperative outcomes. However, ICU admission for elderly patients is not necessarily a marker of differences in mortality or functional status,^8, 9 making ICU admission alone a suboptimal proxy for other patient-centered outcomes. Although an analysis of functional status, quality of life, or mortality was beyond the scope of the present study, information about differences in these outcomes is essential to understand how preoperative cognitive impairment influences the risk of adverse postoperative outcomes.

Ultimately, this study adds to the growing literature on the bidirectional relationship between cognitive impairment and serious illness.^3, 4, 10 Future work on this important topic must move beyond describing the problem. We need to develop better strategies to identify and intervene on behalf of our patients who are at the greatest risk for physical and cognitive deterioration in advance of and during serious illness or surgery. In some cases, this may lead to a decision not to perform surgery at all. In other cases, it may alter perioperative and intraoperative management or prompt interventions to prevent the development of critical illness and further cognitive impairment. In all cases, greater awareness of cognitive impairment as a driver and consequence of serious illness should add depth to the conversations that occur between elderly patients, their families, and physicians about prognosis and shared medical decision making.