Poor Performance on a Preoperative Cognitive Screening Test Predicts Postoperative Complications in Older Orthopedic Surgical Patients

APPROXIMATELY one of every three surgical procedures nationally is performed on a patient 65 yr of age or older. There is intense interest in identifying predictors of adverse outcomes in this age group, given that they have a high complication rate and often do poorly.^1–4  Preoperative assessment of major vital organs has been a routine part of preparation for surgery for decades^5,6  but brain function is typically not formally evaluated.⁷  Yet cognitive impairment is common in older persons, including those living independently. Five percent of Americans aged 70 to 79 yr, 24% of those aged 80 to 89, and nearly 40% of those 90 or older are affected with dementia.⁸  In epidemiologic surveys, the prevalence of impairment is 35 to 50% in those 65 yr of age or older and higher still in those 85 yr of age or older if milder forms of cognitive impairment (e.g., MCI [mild cognitive impairment] or cognitive impairment, not dementia) are included, although estimates vary with the age structure of the population and definition and assessment methods used.^9–12  Consequently, it is reasonable to assume that many seniors without a diagnosis of dementia scheduled for elective surgery have cognitive impairment at baseline. In fact, using the Mini-Cog, a brief, validated, structured cognitive screening tool with high interrater reliability and patient acceptance, we demonstrated recently that 25 to 33% of elective surgical patients 65 yr of age or older score in a range consistent with probable cognitive impairment preoperatively¹³  and, using the same test, others report that 44% of geriatric surgical patients with planned admission to an intensive care unit (ICU) postoperatively are impaired before surgery.¹⁴ 

One key question in the geriatric surgical setting is whether baseline cognition predicts medical complications and other adverse outcomes. Previous work demonstrates that a chronic dementing illness or a clouded sensorium (i.e., acute or chronic delirium) before surgery is associated with a greater risk of postoperative cognitive and noncognitive (medical) morbidity and that a low preoperative Mini-Cog score predicts adverse outcomes in older surgical patients requiring postoperative care in an ICU.^14–16  However, few persons suffering from dementia or an acute change in cognition have elective surgery and the vast majority of elective procedures performed on older persons (e.g., elective joint replacements, spine surgery) do not typically require admission to an ICU postoperatively. Unresolved, therefore, is whether preoperative cognitive screening, as recommended by the American College of Surgeons and the American Geriatrics Society in jointly published guidelines,¹⁷  can help identify those at risk for an adverse outcome when the procedure is common and elective. We hypothesized that even in that situation poor preoperative cognition will be associated with suboptimal surgical outcomes. To test this hypothesis, we cognitively screened older patients without a diagnosis of dementia with the Mini-Cog prior to scheduled elective lower extremity joint replacement surgery and examined the relationship of a low preoperative Mini-Cog score to postoperative morbidity and outcomes.

The Partners Institutional Review Board (Boston, Massachusetts) approved this prospective observational study (clinicaltrials.gov No. NCT02570451). Between September 30, 2014, and July 27, 2015, study staff members approached patients 65 yr of age and older scheduled for a primary lower extremity (hip or knee) joint replacement procedure, who presented to the Weiner Center for Preoperative Evaluation at the Brigham and Women’s Hospital (Boston, Massachusetts). We selected this group because lower extremity joint replacements are relatively homogeneous, do not share a risk factor with cognitive impairment (beyond age), and do not affect the central nervous system directly. All eligible patients were identified from the preoperative evaluation center tracking system on the day prior to surgery. Exclusion criteria included concurrent enrollment in another study; a prior diagnosis of dementia noted on the patient chart or reported to the investigator by the patient or a surrogate; planned outpatient surgery; planned postoperative ICU stay; history of stroke or brain tumor; uncorrected vision or hearing impairment (unable to see pictures or read or hear instructions); limited use of the dominant hand (limited ability to draw); and/or inability to speak, read, or understand the English language.

A power calculation of the number of patients required for 80% power to detect a 25% difference in discharge destination at the P = 0.05 level (primary outcome) using a logistic regression model with a baseline incidence of discharge to place other than home being 53% and our expectation of a 20% loss to follow up in this older patient population would require 192 patients. After obtaining written informed consent, 211 patients participated in the study and completed a survey about their perceptions of preoperative cognitive screening and their primary outcome goals for their surgical procedure (table 1) and were tested on the Mini-Cog. The Mini-Cog involves a three-item recall test for memory and a clock drawing test that serves in part as a distractor; it tests visuospatial representation, recall, and executive function, and takes just minutes to complete.^18,19  The Mini-Cog is validated in community-based populations; it has minimal education, language, or ethnic bias, high sensitivity and specificity for cognitive impairment, and good interrater reliability.^20,21  Investigators were trained to grade the tests by reviewing information easily accessed via the Internet (https://www.alz.org/documents_custom/minicog.pdf; accessed August 18, 2017.) and education sessions provided by the geriatrician (H.J.). The Mini-Cog is graded on a 5-point scale, where 5 is considered a perfect score and a score of 2 or less is considered probably impaired.¹⁸  Accordingly, we used a score of 2 or less as the cutoff in the current study. Two investigators scored each test independently. The first scored it during the preoperative evaluation and the second investigator scored it later and was blinded to patient identity. In the event of a disagreement, a third investigator scored the test and served as a tie-breaker. Patients also completed the (1) short form 36 health survey,²²  an index of quality of life across eight domains (physical functioning, limitations due to physical health or emotional problems, energy/fatigue, emotional well-being, social functioning, pain, general health); (2) geriatric depression scale short form;²³  (3) activities of daily living;²⁴  and (4) instrumental activities of daily living.²⁵  We also measured grip strength as an index of frailty using a Jamar Dynamometer²⁶  (JLW Instruments, USA) and obtained baseline data on age, weight, gender, highest level of education, American Society of Anesthesiologists (ASA) functional status, Metabolic Equivalent of Task,²⁷  and type of surgical procedure from the medical record.

The a priori primary outcome was discharge to place other than home; those living elsewhere prior to surgery were excluded from the discharge location analysis. Secondary outcomes were delirium and complications involving the cardiac (myocardial infarction, congestive heart failure, cardiac arrest, new onset arrhythmia); pulmonary (pneumonia, reintubation); immune (wound infections); circulatory (pulmonary embolism, deep venous thrombosis); renal (acute renal injury), or cerebrovascular (stroke) systems. Additional secondary outcomes were postanesthesia care unit length of stay, hospital length of stay, 30-day readmission, and 30-day mortality. Delirium was identified both by chart review using published criteria²⁸  and by direct, independent assessment with the Confusion Assessment Method (CAM).²⁹  The CAM was administered once per day on postoperative days 1 to 3, or until discharge if the patient was discharged early, by an investigator trained by the geriatrician (H.J.) and blinded to chart review information. We used both methods because they are complementary and well-established. The Confusion Assessment Method typically is administered once or twice a day, but delirium waxes and wanes so this test will miss episodes of delirium if they occur at other times. Conversely, chart review reflects events over an entire day but may miss hypoactive delirium (the most common form) because it may be mistaken for sedation. We gathered most of the other patient information by systematic chart review or examination of discharge diagnoses in the Brigham and Women’s Research Patient Data Registry. Study data were collected and managed using Research Electronic Data Capture (REDCap) tools hosted at Partners Healthcare.³⁰ REDCap is a secure, web-based application designed to support data capture for research studies, providing (1) an intuitive interface for validated data entry; (2) audit trails for tracking data manipulation and export procedures; (3) automated export procedures for seamless data downloads to common statistical packages; and (4) procedures for importing data from external sources.

Data were analyzed by several methods. We used Fisher exact test to examine patient responses to the survey questions by Mini-Cog score and Krippendroff α (KA) was calculated using the “kripp.alpha” function in “irr” package in R software (https://cran.rproject.org/web/packages/irr/irr.pdf; accessed September 5, 2017) to evaluate the agreement between the two initial raters of the Mini-Cog. The CIs of KA were calculated using a bootstrapping method by random sampling the data points with replacement.

We used logistic regression to estimate the odds ratios (ORs) for dichotomous outcomes and a Cox’s proportional hazard model to estimate the hazard ratio of length of hospital stay (time to discharge) by Mini-Cog score. We first performed age-adjusted univariate analyses between covariates (Mini-Cog score, gender, weight, education level, ASA, and metabolic equivalents of task) that, based on a priori background knowledge, could modify the outcomes. Subsequently, all the covariates were entered into a backwards stepwise algorithm, retaining variables with P < 0.1 in the multivariate models. Age and Mini-Cog score were forced into the multivariate model. For the primary and secondary outcomes, the significance threshold was set at P < 0.05. The Hosmer-Leesha goodness of fit test was performed to evaluate model-fitting of the logistic multivariable models. The proportional hazards assumption was tested using scaled Schoenfeld residual. All analyses were performed with statistical software R version 3.1.2 (R Foundation, Vienna, Austria).

During the study period, our preoperative center evaluated 368 patients 65 yr of age or older scheduled for elective total knee or total hip replacement surgery. Of these, 43 were ineligible, 14 refused to front desk staff and were not approached by study personnel, and 30 were missed because the study staff was occupied with a concurrent subject. Study personnel approached 281 eligible patients; 70 declined to participate and 211 patients were enrolled (fig. 1). Among those enrolled, eight did not have their surgical procedure and were eliminated from outcome analysis.

Overall, 50 of 211 (24%) patients scored 2 or lower on the preoperative Mini-Cog, suggesting probable cognitive impairment. Interrater reliability in Mini-Cog scoring was similar to that found in our prior experience with a Krippendroff α of 0.906 (95% CI = 0.857 to 0.950). Characteristics associated with a Mini-Cog equal to or less than 2 included advanced age (P < 0.001) and less education (P = 0.02); low metabolic equivalents of task (P < 0.001), instrumental activities of daily living (P = 0.03), and basic activities of daily living (P = 0.02); physical function limitations on the short form 36 health survey (P = 0.015) and having a knee rather than a hip replacement procedure (P = 0.03; table 1). Patients with a Mini-Cog score less than or equal to 2 were also less likely to live in their own home (P = 0.004) and more likely to be accompanied by someone to the preoperative evaluation appointment (P = 0.02) (table 2). Ninety-four percent of subjects supported the idea of performing a short memory test (table 2). Pain and use of pain medications were common but did not vary by Mini-Cog score. Thus, based on pain scores reported on the short form 36, there was no difference in preoperative pain between patients with a Mini-Cog score less than or qual to 2 versus those with a score greater than or equal to 3 (51 [95% CI 45 to 56] vs. 50 [95% CI 47 to 53], respectively; P = 0.88). Likewise, 84% of patients were taking pain medication (opioids, NSAIDS, acetaminophen, gabapentin) at the time of the preadmission testing visit, but there were no differences in the type of pain medications used between those with a Mini-Cog score less than or qual to 2 versus those with a score greater than or equal to 3 (P = 0.999). Accordingly, it is unlikely pain or the medications used to treat it biased the Mini-Cog results.

Eighty-eight patients (42%; table 3A) living at home prior to surgery were discharged to a place other than home after surgery (primary outcome measure). This outcome was more likely if they had a preoperative Mini-Cog score less than or equal to 2 (67% vs. 34%; OR = 2.97 [95% CI = 1.43 to 6.18]; P = 0.004) in the age-adjusted univariate analysis and remained a predictor of discharge location after multivariate adjustment (OR = 3.88 [95% CI = 1.58 to 9.55]; P = 0.003). The average hospital length of stay was 2.6 ± 0.9 days, with a low preoperative Mini-Cog score predicting longer hospital stay by both univariate (P = 0.018) and multivariate analysis (hazard ratio = 0.63 [95% CI 0.42 to 0.95]; P = 0.026) (table 3B).

Four patients were discharged from the hospital less than 24 h after surgery and prior to delirium screening by CAM. Of the remainder, 14 (6.9%) developed CAM+ delirium postoperatively and 21 (10%) were delirium positive by comprehensive chart review. Of the 14 patients positive by CAM, 11 were also positive by chart review. A preoperative Mini-Cog score less than or equal to 2 was associated with development of postoperative delirium diagnosed by the Confusion Assessment Method on both age-adjusted univariate (P = 0.003) and multivariate analysis (18% vs. 4%; OR = 4.52 [95% CI 1.3 to 15.68]; P = 0.017) (table 4A). A preoperative MiniCog score less than or equal to 2 was likewise associated with postoperative delirium identified by chart review on both age-adjusted univariate (P = 0.021) and multivariate analysis (21% vs. 7%; OR = 3.41 [95% CI 1.26 to 9.23]; P = 0.016) (table 4B). A post-hoc age-adjusted analysis revealed that patients with delirium stayed in the hospital 1.12 days longer than those without delirium (95% CI 0.67 to 1.58; P < 0.001).

Seventeen patients (8.1%) had postoperative cardiac complications, with the majority (N = 15) being onset of new arrhythmias, mainly atrial fibrillation. A low preoperative Mini-Cog score was associated with cardiac events on age-adjusted univariate (OR = 3.14 [95% CI 1.07 to 9.18]; P = 0.037) but not multivariate analysis (17% vs. 6%; OR = 2.87 [95% CI 0.89 to 9.23]; P = 0.077). Other adverse events identified by chart review or discharge diagnosis codes, including pneumonia, reintubation, pulmonary embolism, deep venous thrombosis, stroke, coma, wound infection, sepsis, renal failure, urinary tract infection, reoperation, and unanticipated ICU admission, occurred too infrequently to be analyzed as independent outcomes. The only predictor of 30-day emergency room visits was metabolic equivalents of task (P = 0.017 and 0.013 by univariate and multivariate analysis, respectively), and 30-day mortality was too rare (N = 2) to be analyzed statistically.

These data confirm that poor preoperative cognition as assessed by Mini-Cog screening is both prevalent among geriatric patients scheduled for elective major joint replacement surgery and predictive of adverse outcomes including postoperative delirium, a longer hospital stay, and greater likelihood of being discharged to a place other than home. Importantly, this was true even though we excluded patients with a known diagnosis of dementia. In contrast, age, ASA functional status, grip strength, preoperative geriatric depression scale scores, and functional state (short form 36 health survey, Instrumental Activities of Daily Living and Activities of Daily Living) were not associated with the prespecified outcomes and/or complications by multivariable modeling. Metabolic equivalents of task was the exception; it predicted delirium diagnosed by chart review (but not Confusion Assessment Method) and the likelihood of being discharged to a place other than home. Taken together, these data show that a remarkably high percentage of seniors electing to undergo a total hip or knee replacement procedure have probable, but previously undetected, cognitive impairment at baseline and that preoperative cognitive screening with a simple, brief test can help identify those at risk of postoperative cognitive and medical complications.

That about one in four geriatric patients scheduled for elective major joint replacement surgery have probable cognitive impairment preoperatively is not surprising given the prevalence of dementia and milder forms of cognitive impairment in community samples.^9–11  Much of this is undetected because, by definition, MCI can be present with no functional deficit and only a minority of people with dementia have a clinical cognitive evaluation that leads to a diagnosis.³¹  Our results compare well with our prior data on geriatric patients scheduled for a variety of elective noncardiac, non-neurosurgical procedures¹³  and with results of studies in hospitalized patients or other surgical populations.^14,32–34  For instance, depending upon age and type of cognitive testing, the prevalence of cognitive impairment in patients 65 yr of age or older presenting to an emergency department, an ambulatory urogynecology clinic, or having surgery with planned admission to the ICU ranges from 5 to 63%.^14,32,34  Nor is it surprising that people with cognitive impairment are more likely to develop delirium. Poor cognitive status, typically defined as dementia in population studies, is a well-known risk factor for in-hospital delirium and also appears to be an independent predictor of morbidity and mortality in geriatric patients having major elective operations.^14,35  The problem, however, is that in both primary care and hospital settings cognitive impairment, and even dementia, often go unnoticed without structured screening because routine clinical interactions are insensitive.^19,36,37  Accordingly, as we demonstrate, a formal, yet simple and brief, cognitive screening procedure can be useful both to identify probable cognitive impairment before surgery and, in conjunction with other information gathered routinely preoperatively, to forecast which patients are most likely to have undesirable postoperative outcomes. Moreover, most subjects endorsed use of a brief memory test preoperatively.

There are numerous abridged cognitive screening tests but few have been used in the preoperative setting. We chose the Mini-Cog because it is brief, freely available, requires no specialized personnel or technology, has minimal education and cultural/language bias, and is validated against standardized cognitive measures in community samples.^38–43  Designed for primary care, the Mini-Cog has been used in surgical settings, including by us,^13,14  and has high interrater reliability and is easy to administer. The Mini-Cog involves a three-item recall test for memory and a clock drawing test that serves as a distractor. It tests visuospatial representation, recall, and executive function and detects dementia with a sensitivity and specificity of 0.91 and 0.86, respectively.^14,18,19,21  We used 2 or less as the cutoff for probable cognitive impairment because it identifies with reasonable sensitivity and specificity the level of impairment found in individuals who might present to a memory clinic for evaluation of MCI or dementia,¹⁸  but others have used a higher cutoff and found a correspondingly higher prevalence of probable cognitive impairment preoperatively.¹⁴  Category fluency also has been used as a cognitive screening test in this setting with similar results in terms of prevalence of probable cognitive impairment preoperatively and association with delirium postoperatively, but selection bias is possible because about half of eligible patients were not screened.⁴⁴  It is important to emphasize in this context that no single cognitive test, administered at a single time, can diagnose MCI or dementia. Therefore, by itself, a low preoperative Mini-Cog score is not enough to diagnose or label a patient as having a memory disorder. As we demonstrate, however, what it can do is help identify a subpopulation of geriatric surgical patients at risk for postoperative delirium and poor outcomes and, as such, potentially guide and enhance the care of these patients.

This study has multiple limitations. First, the stress of being in the preoperative evaluation center could confound the performance of seniors on the cognitive screening test, leading to a high false-positive rate for cognitive impairment and, potentially, hesitation among patients about undergoing elective surgery for fear of having cognitive impairment afterward. Few experiences, however, are as stressful as surgery and hospitalization. As such, testing in a busy preoperative clinic may reveal more about an individual’s likely response to surgery and hospitalization than if testing were done in the quieter, artificial environment of a neuropsychology laboratory. Second, other brief cognitive screening instruments may work as well or better than the Mini-Cog in the presurgical setting, and noncognitive screening measures might be equally useful. Indeed, frailty, walking speed, functional dependency, and self-reported diminished mobility or history of falls have all been linked to postoperative complications and mortality in geriatric patients.^45–49  Third, we assessed patients for delirium only once per day, typically around noon, but clinical delirium waxes and wanes throughout the day. Thus, we may have underestimated the incidence of delirium. Likewise, because we used grip strength as the only marker of frailty, we may have underestimated the prevalence of this syndrome in our population and made it difficult to detect the relationship between frailty and adverse postoperative outcomes observed by others.⁴⁸  Also, because we cannot entirely exclude confounding by covariates (e.g., age, comorbidity) and the significance threshold for the primary and secondary outcomes was set at P < 0.05, the results should be considered preliminary and in need of confirmation in larger studies.⁵¹  Lastly, our study was limited to orthopedic patients having elective major joint replacement procedures, so the results may not generalize to all geriatric surgery patients. Studies involving general surgical patients, however, suggest the link between poor cognition and medical-surgical morbidity is not unique to older orthopedic patients.^14,52,53 

Based on limited evidence, the American College of Surgeons and the American Geriatrics Society recently published joint guidelines that recommend preoperative cognitive assessment of older surgical patients with a screening tool such as the Mini-Cog.¹⁷  Cognitive screening requires time, however, and can trouble older adults,^54,55  so it is not a trivial matter to adopt it in a preoperative clinic, and results must be interpreted cautiously. Yet, because data from this and other studies show that preoperative cognitive screening is practical and that poor performance is associated with adverse postoperative events (delirium, surgical complications), cognitive screening may be a valuable adjunct to traditional preoperative risk assessment practices for this demographic. There are as yet no data to show targeting poor cognitive performers for special attention before, during, and after surgery improves surgical outcomes, but recent evidence that prehabilitation, specialized units, and comprehensive geriatric care may enhance outcomes of older surgical patients provides reason for optimism that outcomes can be improved.^56–58  Preoperative cognitive risk stratification may help identify those at greatest risk for adverse surgical outcomes so interventions designed to mitigate complications can be targeted to those most likely to benefit.

Supported by Anesthesia Patient Safety Foundation (Rochester, Minnesota) and grant No. AG048522 from the National Institutes of Health (Bethesda, Maryland; to Dr. Culley); grant No. AG048637 from the National Institutes of Health (Bethesda, Maryland; to Dr. Crosby); and the Department of Anesthesiology, Perioperative, and Pain Medicine, Brigham and Women’s Hospital, Boston, Massachusetts.

Dr. Culley serves as director of the American Board of Anesthesiology, chair of the American Board of Medical Specialties (Chicago, Illinois) Committee on Continuous Certification, Accreditation Council for Graduate Medical Education, Anesthesiology Residency Review Committee, ad-hoc member (Chicago, Illinois), executive editor of Anesthesiology, American Society of Anesthesiologists committee member; receives grant funding from CRICO (Boston, Massachusetts), NIA; has lectured at the Department of Anesthesiology, The State University of New York, Department of Anesthesiology, Maine Medical Center, Washington State Society of Anesthesiology, Virginia Mason Medical Center, University of Florida (Jacksonville), Mayo Department of Anesthesiology, and the University of Alabama. Dr. Crosby receives grant funding from CRICO; serves as editor of Anesthesia and Analgesia, as an American Society of Anesthesiologists committee chair and member, and on the Scientific Advisory Board of The Medicines Company; has lectured at New York Postgraduate Assembly, California Society of Anesthesiologists, American Society of Anesthesiologists, and International Anesthesia Research Society. All other authors declare no competing interests.