Abstract
IMPORTANCE:
Hospitalizations are associated with cognitive decline in older adults.
OBJECTIVE:
To determine the association between hospitalization characteristics and the trajectory of cognitive function in older adults.
DESIGN:
Population-based longitudinal study of cognitive aging.
SETTING:
Olmsted Medical Center and Mayo Clinic, the only centers in Olmsted County, Minnesota, with hospitalization capacity.
PARTICIPANTS:
Individuals without dementia at baseline, with consecutive cognitive assessments from 2004 through 2017, and at least one visit after the age of 60.
MEASUREMENTS:
The primary outcome was longitudinal changes in global cognitive z-score. Secondary outcomes were changes in four cognitive domains: memory, attention/executive function, language, and visuospatial skills. Hospitalization characteristics analyzed included elective vs non-elective, medical vs surgical, critical care vs no critical care admission, and long vs short duration admissions.
RESULTS:
Of 4,587 participants, 1,622 had ≥1 hospital admission. Before hospitalization, the average slope of the global z-score was −0.031 units/year. After hospitalization, the rate of annual global z-score accelerated by −0.051 (95% CI −0.057, −0.045) units, P<0.001, resulting in an estimated annual slope after the first hospitalization of −0.082. The accelerated decline was found in all four cognitive domains (memory, visuospatial, language, and executive, all P<0.001). The acceleration of the decline in global z-score following hospitalization was greater for medical compared to surgical hospitalizations (slope change following hospitalization=−0.064 vs.−0.034 for medical vs surgical, P<0.001), and non-elective compared to elective admissions (slope change following hospitalization=−0.075 vs.−0.037 for non-elective vs elective, P<0.001). The acceleration of cognitive decline was not different for hospitalization with ICU admission versus not.
CONCLUSIONS:
Hospitalization of older adults is associated with accelerated decline of global and domain-specific cognitive domains, with the rate of decline dependent upon type of admission. The clinical impact of this accelerated decline will depend on the individual’s baseline cognitive reserve and expected longevity.
Keywords: global cognitive z-scores; cognitive domains: memory, attention/executive function, language, visuospatial skills; Mayo Clinic Study of Aging; older adults, hospitalization admission; elective, non-elective, medical, surgical, critical care admission
Introduction
Hospitalizations for both critical and non-critical illness are associated with long-term cognitive decline in older adults.1–8 The public health implications of cognitive impairment in older adults after hospitalization are considerable. Understanding the magnitude of cognitive decline can inform patients, families, and providers regarding long-term expectations and contribute to health management and planning of independent living. If the etiology is better delineated, possible interventions could be developed to ameliorate this decline.
It is not clear whether events occurring during hospitalization are primary contributors to-post-discharge cognitive decline, whether cognitively impaired individuals are more likely to be hospitalized,9–12 or whether hospitalization is merely a marker for increased acuity of health conditions related to cognitive decline.1,2,13 A recent study utilizing administrative records to describe hospitalization characteristics demonstrated that emergent hospital admissions, compared to elective, were associated with greater cognitive decline, even after adjusting for other factors that could affect cognition.8 The type of admission could serve as a proxy for the acuity of illness or underlying disease or may be indicative of the intensity of the intervention received during hospitalization (e.g., emergency surgery or critical care admission). Furthermore, medical hospitalizations are frequently non-elective and associated with acute worsening of systemic diseases compared to surgical admissions, which are typically elective and most frequently performed to address a single anatomical issue (e.g., orthopedic surgery).
The purpose of this study was to determine the association between hospitalization and the trajectory of global and domain-specific cognitive decline in older adults and whether the characteristics of hospitalization (elective vs. non-elective, medical vs. surgical, critical care admissions vs. no critical care admission) impact this association. We used the Mayo Clinic Study of Aging (MCSA), a large prospective, population-based longitudinal study of cognitive aging among Olmsted County, Minnesota residents.14,15 The primary outcome of interest was longitudinal changes in global cognitive z-score. Secondary outcomes were changes in four cognitive domains: memory, attention/executive function, language, and visuospatial skills.
Materials and Methods
This study conformed to the requirements of the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement and was approved by the institutional review boards of Mayo Clinic and Olmsted Medical Center, Rochester, Minnesota. At enrollment, all participants provided written informed consent to participate in the study.
Participants and setting
Study participants were enrolled in a longitudinal population-based study that was initiated in 2004 in Olmsted County, MN, which studies patterns of cognitive aging, prevalence and incidence of mild cognitive impairment (MCI) and dementia, conversion rates from MCI to dementia, and risk factors associated with cognitive decline among County residents.14 The specifics of MCSA design, enrolment dynamics, participation, and follow-up are detailed in an earlier report.14 Residents of Olmsted County, MN, were randomly selected and invited to participate in the study. The original enrollment included individuals 70-89 years old. The cohort was subsequently expanded in 2012 by starting to enroll individuals age 50 and older. The present study included non-demented participants with study evaluation visits from November 1, 2004, through December 31, 2017, and at least one visit after the age of 60. We identified all hospitalizations at Olmsted Medical Center and Mayo Clinic, the only two institutions in Olmsted County with hospitalization capacity.
Assessments
In the present analysis, we included only assessments that were available for all MCSA visits, including baseline and subsequent visits at approximately 15-month intervals.14,15 Initial evaluations included assessment of demographic information, medical history, education, subjective memory complaints, family history of dementia, and APOE genotyping. Physician evaluation included the Short Test of Mental Status,16 a medical history review, and neurologic examination. Neuropsychological evaluation included nine tests to assess four cognitive domains.14,15 Cognitive scores were expressed as domain and global cognitive z-scores. To compute these scores, means and standard deviations of raw test scores for participants who were cognitively unimpaired at the baseline visit in the MCSA 2004-2012 enrollment cohort were weighted to reflect the 2013 Olmsted county population and then used as the reference distribution.14,17 Scores within a domain were averaged and scaled using the reference cohort to create domain z-scores. An expert panel reviewed each participant’s evaluation and adjudicated cognitive status as cognitively unimpaired, MCI, or dementia based on published criteria.18–20 MCSA procedures avoid testing in the temporal vicinity of procedures and acute illness to avoid the effects of transient postoperative cognitive decline.
Characteristics of hospital admissions
We reviewed the duration and nature of hospitalization, with categories including 1) elective vs. non-elective (defined as any admission through Emergency Department - even transfers from other hospitals for patients who need an elevated level of medical care occur through Emergency Department); 2) medical vs. surgical (any hospitalization where surgery occurred was classified as surgical admission), and; 3) non-critical care vs. critical care hospitalization (hospitalizations associated with ICU admission, which may be for surgical or medical indications). Surgical and medical diagnoses were identified through the Rochester Epidemiology Project medical records linkage system21–23 using ICD-9 and ICD-10-CM codes.
Statistical analyses
Baseline demographics and comorbidities were summarized using median [25th, 75th percentiles] for continuous variables and percentage for categorical variables according to the hospitalization status during follow-up.
The primary aim was to assess whether the trajectory (slope, rate of decline) in global cognitive z-score accelerated following hospitalization. Linear mixed-effects models assessed the association between hospitalization and global cognitive z-scores.
Adjustment covariates for the analyses were specified a priori based on prior publications and potential confounding and included age at baseline, sex, years of education, APOE ε4 status, marital status, smoking status, and number of cardiometabolic conditions (CMC, including hypertension, hyperlipidemia, cardiac arrhythmias, coronary artery disease, congestive heart failure, diabetes mellitus, and stroke). CMCs were proposed by the U.S. Department of Health and Human Services in 2010 as indicators of vascular health.24 Also, three binary covariates were included in the analysis indicating any hospitalization in the 5, 10, or 20 years prior to baseline (Table 1).
Table 1:
Patient characteristics at enrollment summarized by hospital admission
| Hospitalization | |||
|---|---|---|---|
| Characteristics | Overall (N=4,587) |
No (N=2,965) |
Yes (1 or more) (N=1,622) |
| Age, years | 74 [69, 81] | 73 [67, 81] | 75 [71, 81] |
| Sex | |||
| Male | 2,337 (51%) | 1,442 (49%) | 895 (55%) |
| Female | 2,250 (49%) | 1,523 (51%) | 727 (45%) |
| Marital status | |||
| Married/living together | 3,197 (70%) | 2,051 (69%) | 1,146 (71%) |
| Widowed | 848 (18%) | 523 (18%) | 325 (20%) |
| Separated/divorced | 341 (7%) | 244 (8%) | 97 (6%) |
| Single, never married | 193 (4%) | 141 (5%) | 52 (3%) |
| Missing marital status | 8 (0%) | 6 (0%) | 2 (0%) |
| Education, years | |||
| <12 [did not complete high school] | 271 (6%) | 175 (6%) | 96 (6%) |
| 12 [high school graduate] | 1,379 (30%) | 858 (29%) | 521 (32%) |
| 13–15 [some college or technical school] | 1,255 (27%) | 799 (27%) | 456 (28%) |
| ≥16 [4-year college degree or graduate studies] | 1,682 (37%) | 1,133 (38%) | 549 (34%) |
| Smoking status, n=2,964/1,622 | |||
| Never | 2,396 (52%) | 1,596 (54%) | 800 (49%) |
| Former | 1,979 (43%) | 1,218 (41%) | 761 (47%) |
| Current | 211 (5%) | 150 (5%) | 61 (4%) |
| APOE-ε4 allele, n=2,847/1,616 | 1,206 (27%) | 776 (27%) | 430 (27%) |
| No | 3,257 (71%) | 2,071 (70%) | 1,186 (73%) |
| Yes | 1,206 (26%) | 776 (26%) | 430 (27%) |
| Unknown | 124 (3%) | 118 (4%) | 6 (0%) |
| CV and metabolic conditions, n=2,964/1,622 | 2 (1, 3) | 2 (1, 3) | 3 (1, 4) |
| Most recent prior hospitalization | |||
| None within 20 years prior | 1,513 (33%) | 1,105 (37%) | 408 (25%) |
| Prior 0-5.0 years | 1,970 (43%) | 1,081 (36%) | 889 (55%) |
| Prior 5.1-10.0 years | 723 (16%) | 486 (16%) | 237 (15%) |
| Prior 10.1-20.0 years | 381 (8%) | 293 (10%) | 88 (5%) |
| Global cognitive z-score† | |||
| Among age 60-69 | 0.42 (0.90) | 0.45 (0.91) | 0.33 (0.85) |
| Among age 70-79 | −0.46 (0.99) | −0.55 (1.03) | −0.32 (0.93) |
| Among age ≥80 | −1.31 (1.10) | −1.44 (1.12) | −1.12 (1.06) |
Patient characteristics at the time of enrollment or first MCSA visit with global z-score. Values are median [Q1, Q3] for continuous variables and number (percent) for categorical variables unless otherwise noted.
Values are mean (standard deviation). For characteristics that did not have complete data for all participants, the number with complete information is given for those without/with hospitalization.
Abbreviation: CV, cardiovascular; APOE-ε4 allele, apolipoprotein E genotype
Details of the statistical methodology are summarized in Supplementary Statistical Methods. Fixed effects included the baseline covariates listed above, and time in years after the MCSA baseline visit. The interaction of baseline covariates with time after the baseline visit, and time-dependent exposure variable for time in years after a post-baseline hospitalization were also included as fixed effects. All participants were included in the analysis, including those with no subsequent hospitalization following baseline. Time after baseline visit describes the slope of global cognitive decline following baseline without (or prior to) subsequent hospitalization. Time after post-baseline hospitalization describes the change in the slope of global cognitive decline following the hospitalization. Random effects included subject-specific random intercepts and random slopes. Domain-specific cognitive z-scores were analyzed as secondary outcomes. Many participants had multiple hospitalizations. We hypothesized that repeated hospital admissions might have an additive effect. However, the functional form of this effect was unclear (whether each subsequent hospitalization has an equal effect, increased effect, or diminishing effect relative to the prior). Under the primary outcome, we assessed several functional forms and chose the approach maximizing the model likelihood as detailed in the Supplementary Statistical Methods. The model that fit the data best was an effect multiplier of 0.7, indicating a 30% geometric reduction in effect for each subsequent admission. Three pre-specified secondary analyses were assessed to examine whether the type of hospitalization (surgical vs medical; elective vs non-elective; combinations of surgical vs medical and elective vs non-elective; and requiring ICU admission vs not) modified the relationship between hospital admission and change in the trajectory of global cognitive z-score. In these models, type of hospital admission was defined based on the first admission following enrollment, and follow-up was censored when subsequently hospitalized for a different indication (e.g., medical admission after previous surgical admission). An interaction P-value was calculated to assess whether the change in slope following admission differed between admission types. Finally, a model evaluated whether duration of hospitalization (1-3 days, vs 4-6 days, vs ≥7 days) modified the relationship between hospitalization and the trajectory of cognitive z-scores.
In sensitivity analyses, we first explored the potential impact of repeated testing in participants over time (i.e. practice effect) by including an indicator variable flagging the first MCSA visit in the model to account for ‘first-time’ testing. A second sensitivity analysis assessed the association of hospitalization and cognitive z-scores considering only the first hospitalization, with subsequent hospitalizations were ignored (assumed no further association with cognitive outcomes).
Results are summarized by presenting average slope estimate and the change in slope following hospitalization. The significance level was set at P less than 0.05, with the 2-tailed test. Analyses were performed with SAS statistical software (Version 9.3; SAS Institute, Inc, Cary)
Results
Hospitalization characteristics
Of 4,938 MCSA participants with at least one study visit after the age of 60, 351 were excluded (Supplementary Figure S1), so the present study included 4,587 participants. Of these, 2,965 had no hospitalization during follow-up (median [IQR] follow-up 2.6 [0.0-5.7] years), while 1,622 had a hospital admission (total of 3,447 admissions) with median [IQR] follow up of 2.0 [0.8-3.9] years prior to the first admission, and 3.2 [1.4, 5.7] years following the first admission (Supplementary Figure S1). Of the 1,622 participants hospitalized after their baseline MCSA visit, 793 (49%) had one, 366 (23%) had 2, 232 (14%) had 3, and 231 (14%) had ≥ 4 hospitalizations. Most participants (3,074 [67%]) were hospitalized at least once in the 20-year period prior to the index date, with the majority of these hospitalizations occurring in the 5-year period prior to the index date (Table 1). As expected, the baseline global cognitive z-score decreased with age (Table 1).
The median [IQR] hospitalization duration was 3 [2, 4] days (Table 2). The majority of hospitalizations were for medical indications, elective, and did not require ICU admission (Table 2). Among the 2,207 medical admissions, 1,113 (50%) were non-elective, and 274 (12%) required ICU admission. Among the 1,240 surgical admissions, 156 (13%) were non-elective, and 239 (19%) required ICU admission. Supplementary Tables S1 and S2 present additional hospitalization characteristics separately for surgical and medical hospitalizations.
Table 2.
Hospitalization characteristics*
| Hospitalization | First hospitalization (N=1,622) |
All hospitalizations (N=3,447) |
|---|---|---|
| Type of admission by specialty | ||
| Medical | 944 (58%) | 2,207 (64%) |
| Surgical | 678 (42%) | 1,240 (36%) |
| Type of admission by urgency | ||
| Elective | 1,036 (64%) | 2,178 (63%) |
| Non-elective | 586 (36%) | 1269 (37%) |
| Intensive care unit admission | ||
| No | 1,371 (85%) | 2,934 (85%) |
| Yes | 251 (15%) | 513 (15%) |
| Length of hospitalization, days | 3 [2, 4] | 3 [2, 4] |
| Duration of hospitalization | ||
| 1-3 days | 1,071 (66%) | 2,201 (64%) |
| 4-6 days | 366 (23%) | 785 (23%) |
| ≥7 days | 185 (11%) | 461 (13%) |
Non-elective is defined as initial admission to the hospital through the emergency department. Surgical admission is defined as hospitalization episode that included surgical procedure. Length of hospitalization corresponds to number of days in hospital during the hospitalization episode.
Values are median [Q1, Q3] for continuous variables and number (percent) for categorical variables.
Cognitive scores and hospitalization
From the adjusted linear mixed model, the mean decline in global z-score was −0.031 units per year prior to hospitalization. The parameter estimate for the change in the trajectory of cognitive global z-score associated with hospitalization was −0.051 (95% CI −0.057 to−0.045 units, P<0.001). Thus, the estimated annual slope after the first hospitalization was −0.082 (−0.031 plus the change associated with hospitalization of −0.051). Using the model that fits the data best, after a second hospitalization, the average slope estimate was −0.118 (−0.082 plus an additional change of 0.7 x −0.051) units. The average slope and change in slope, following subsequent hospitalizations, are shown in Table 3. Full model results can be found in Supplementary Tables S3a and S3b.
Table 3:
Estimated annual slope of cognitive z-scores prior to hospitalization and change in slope following hospital admission*
| Association estimate† | Estimated slope following hospitalization‡ | |||||
|---|---|---|---|---|---|---|
| Scale | Estimated slope prior to first hospitalization* | Estimate (95% CI) | P value | First | Second | Third |
| Global z-score | −0.031 | −0.051 (−0.057, −0.045) | <.001 | −0.082 | −0.118 | −0.143 |
| Cognitive domain z-scores | ||||||
| Memory | 0.009 | −0.042 (−0.048, −0.036) | <.001 | −0.033 | −0.064 | −0.083 |
| Language | −0.035 | −0.028 (−0.034 −0.022) | <.001 | −0.063 | −0.082 | −0.096 |
| Visuospatial | −0.009 | −0.019 (−0.023, −0.014) | <.001 | −0.028 | −0.041 | −0.050 |
| Attention/executive | −0.060 | −0.041 (−0.048, −0.035) | <.001 | −0.101 | −0.130 | −0.150 |
Slope estimate presented is for the average slope among all participants. Although not all participants had a hospitalization during follow up, all are considered to be at risk for hospitalization, and therefore are included in the calculation of this estimate.
Results are from multivariable linear mixed effects models. All models are adjusted for age at baseline, sex, education, marital status, APOE-ε4 genotype, smoking status, and cardiometabolic conditions (see Methods) and three binary covariates were included in the analysis indicating any hospitalization in the 5, 10 and 20 years prior to baseline. Unknown marital status and APOE ε4 status were included as separate categories. Models also adjusted for the interaction of covariates and time since first z-score. A complete case analysis assuming missing completely at random was conducted with respect to other missing baseline covariates - excluding just one subject with missing cardiometabolic conditions. Longitudinal dropout from the MCSA was assumed missing at random.
As a result of the model specification, the magnitude of the change in slope following hospitalization is decreased by 30% for each subsequent hospital admission. For example, for the calculation of the estimated change in slope of the global z-score the change following first admission is −0.051, following a second admission, the change is 0.7 × −0.051= −0.036, resulting in a cumulative change in slope from pre-hospitalization of (−0.051) + (−0.036) = −0.087, and following a third admission the change is 0.72 × −0.051= −0.025 resulting in a cumulative change in slope from pre-hospitalization of (−0.051) + (−0.036) + (−0.025) = −0.112, Similar calculations can be used for 4th, 5th etc. admissions whereby for the ith hospitalization the change in slope is 0.7(i-1) × −0.051.
Hospitalization was also associated with an increased rate of decline in all four cognitive domains. The estimated average slopes prior to and following the first hospitalization were 0.009 and −0.033 units per year for memory; −0.035 and −0.063 units per year for language; −0.009 and −0.028 units per year for visuospatial; and −0.060 and −0.101 units per year for attention/executive function, all P<0.001 (Table 3).
Results from sensitivity analyses considering only the first hospitalization and accounting for potential practice effects are presented in Supplementary Tables S4 and S5. The pattern of results from these two sensitivity analyses was similar to those from the primary analysis.
Characteristics of hospital admission and changes in global cognitive z-scores
Table 4 summarizes secondary analyses assessing whether the trajectory of global cognitive z-score following hospitalization changed based on hospitalization characteristics. The acceleration of cognitive decline following hospitalization was more pronounced for medical compared to surgical hospitalization (estimated change in slope −0.064 vs. −0.034, interaction P<0.001), and non-elective compared to elective admissions (estimated change in slope −0.075 vs. −0.037, interaction P<0.001).
Table 4:
Change in annual decline of global cognitive z-score following first hospitalization according to admission characteristics*
| Change in slope following hospitalization |
|||
|---|---|---|---|
| Type of admission | Estimate (95% CI) | P value | Interaction P |
| Type of admission by specialty | <.001 | ||
| Surgical | −0.034 (−0.043, −0.025) | <.001 | |
| Medical | −0.064 (−0.075, −0.053) | <.001 | |
| Type of admission by urgency | <.001 | ||
| Elective | −0.037 (−0.044, −0.030) | <.001 | |
| Non-elective | −0.075 (−0.094, −0.057) | <.001 | |
| Intensive care unit admission | 0.896 | ||
| No | −0.052 (−0.059, −0.045) | <.001 | |
| Yes† | −0.053 (−0.077, −0.030) | <.001 | |
| Duration of hospitalization | 0.202 | ||
| 1-3 days | −0.046 (−0.054, −0.037) | <.001 | |
| 4-6 days | −0.042 (−0.060, −0.024) | <.001 | |
| ≥7 days | −0.067 (−0.045, −0.044) | <.001 | |
Results are from multivariable linear mixed effects models. P-values test whether the change in slope due to hospital admission is significantly different from 0. Interaction P-values test whether the effect of hospital admission on slope differs according to type of admission.
In further post-hoc analyses, for surgical admissions, accelerations in global cognitive decline were greater after non-elective surgical compared with elective surgical hospitalizations (estimated change in slope −0.068 [95% CI −0.107, −0.029] vs. −0.039 [95% CI −0.050, −0.027]). A similar pattern was seen for medical admissions, with greater declines after non-elective medical compared with elective medical hospitalizations (−0.064 [95% CI −0.079, −0.049] vs −0.050 [95% CI −0.065, −0.035]). The association of hospitalization with changes in cognitive decline did not differ for hospitalization requiring ICU admission vs not (interaction P=0.896) nor dependent on hospitalization duration (interaction P=0.202).
Discussion
The major findings of this study are that 1) hospitalization was associated with accelerated declines in global cognition and all four cognitive domains assessed (memory, visuospatial skills, language and attention/executive function); 2) multiple hospitalizations were associated with further accelerated cognitive loss, albeit at a reduced magnitude compared to the first admission, and; 3) medical and non-elective hospitalizations were associated with greater rates of cognitive decline compared to surgical and elective hospitalizations, respectively.
Declines were observed for all four of the assessed cognitive domains after hospitalization. However, more pronounced declines were noted for memory and attention/executive function, two domains primarily affected by normative aging, and less for visuospatial and language, which are relatively preserved with aging.25–28 These findings confirm prior reports that hospitalization is associated with accelerated global cognitive decline.5,7,8 In the present report we confirmed that the type of hospitalization was associated with the magnitude of cognitive decline. Similarly, James et al.8 examined cognitive scores after hospitalizations, and distinguished between elective and non-elective admissions, using longitudinal cognitive assessments (based on 19 tests) but following a somewhat different analytic approach. In contrast to our findings, they found that only non-elective, but not elective, hospitalizations were associated with cognitive trajectory changes. Similar results were observed for multiple hospitalizations, associated with further cognitive declines only for non-elective hospitalizations. For non-elective hospitalizations, they also found associations with several cognitive subdomains, with the exception (in discrepancy with our results) of visuospatial ability. The reasons that our results differed are not clear, but we do note that most elective hospitalizations in the James et al.8 study were surgical (82%), whereas 50% of elective hospitalizations in our study were surgical. As we found that surgical admissions were associated with less acceleration in cognitive decline, the high proportion of surgical patients in the James et al.8 elective hospitalization group could explain the difference in findings. Nonetheless, we found in secondary analyses that even within categories of surgical or medical indications for admission, non-elective hospitalization was associated with a greater acceleration in cognitive decline than elective hospitalization. The general direction and magnitude of these associations was similar in sensitivity analyses that consider only a first hospitalization and when accounting for a practice effect.
The reasons that non-elective hospitalizations were associated with greater acceleration in cognitive declines are unknown. One explanation may be that such hospitalizations presumably reflect more acute, serious disease compared with elective hospitalizations. However, we did not find an association of ICU admission, a measure of illness acuity, with changes in trajectory. It is also possible that events during hospitalization which do not necessarily require ICU admission could accelerate cognitive decline after hospitalization.3,29 For example, delirium during hospitalization is associated with increased risk for long-term cognitive impairment following discharge.29 We did not have reliable data regarding delirium for all patients, thus could not include this as a factor in our analyses.
Medical hospitalization was associated with a greater acceleration in decline compared with surgical hospitalization. The fact that the proportion of non-elective hospitalizations was greater among medical patients may be an important contributor to this observation. Also, the most common indication for medical admission was cardiovascular disease, a known risk factor for cognitive decline.30–32 Finally, most of the surgical indications were associated with correction of specific acute conditions (knee replacement, hernia repair etc.). These patients may have a lower burden of severe systemic disease compared those with chronic medical conditions who require hospitalization, such that cognitive decline may be more pronounced in patients admitted for medical indications.
Animal studies suggested that exposure to general anesthetic agents may induce neuropathology implicated in Alzheimer dementia33,34 raising concerns regarding similar effects in humans.35 Indeed, in a series of prior analyses we found that undergoing anesthesia and surgery was associated with a modest, but statistically significant, acceleration in the rate of cognitive decline.3,36,37 However, the association between exposure to anesthesia and accelerated decline was seen for both general and regional anesthesia, arguing that confounding factors other than exposure to general anesthesia (comorbidities, stress of hospitalization, etc) are responsible.37,38 The current finding that cognitive decline was greater following medical compared with surgical admission does not support the concept that exposure to anesthesia itself causes acceleration in long-term cognitive decline.
Clinical significance of cognitive decline
In our cohort, the estimated global cognitive z-score before hospitalization declined at a rate of −0.031 unit per year, corresponding to a 5-year estimated decline in global z-scores of −0.16 units (−0.031 units per year x 5). For an individual who has a single hospitalization, the 5-year estimated decline in global z-score would be increased by −0.25 (−0.051 units per year x 5). Taking into account the type of hospital admission is also important. The additional decline over a 5-year period following a medical admission would be −0.32 (−0.064 units per year x 5) compared to an additional decline of −0.17 for surgical admissions (−0.034 units per year x 5).
The clinical implications of this accelerated cognitive decline will depend on an individual’s cognitive reserve and expected years of life following hospitalization. Figure 1 A shows simulated paths of two hypothetical 75 year-old participants (Participant 1, above-average cognitive baseline function and Participant 2, below-average cognitive baseline function) with escalating risks for cognitive decline and 3 possible hospitalization scenarios (no hospitalization, single hospitalization at 2 years, hospitalization at 2 and 4 years). For each participant, the slope of cognitive decline is accelerated following the first hospitalization with an additional, albeit attenuated, decline following the second hospitalization. If hospitalizations occur for Participant 1 at 2 and 4 years, the cognitive decline trajectory will remain well above the threshold for cognitive impairment after 8 years of follow-up, at which time the participant would be 83 years of age. However, if Participant 2 has a single hospitalization at 2 years, the associated accelerated cognitive decline will reach the threshold for cognitive impairment at approximately 5 years, age 80 years, which is 3 years earlier than would have been expected without hospitalization. Similarly, Figure 1 B. shows simulated paths for global cognitive trajectories of the same two hypothetical participants under scenarios of no hospitalization, a single medical hospitalization at 2 years and a single surgical hospitalization at 2 years. The interpretation of the clinical implications of hospitalization for these patients is similar to that described in Figure 1 A, although for participant 2, the time at which he reached the threshold for cognitive impairment is more accelerated following medical vs. surgical hospitalization.
Figure 1.
A and B simulate cognitive trajectories for 2 hypothetical participants (Pt), each with and without hospitalizations during follow-up. Pt 1 is a highly functional 75-year-old college-educated female, married, with no other comorbidities. Pt 2 is a 75-year-old male with high-school diploma, former smoker, moderate comorbidity burden (cardiometabolic conditions = 3), never married, and APOE-ε4 positive. Figure 1 A shows simulated paths of 3 possible hospitalization scenarios (no hospitalization, single hospitalization at 2 years, hospitalization at 2 and 4 years). In each case, the slope of cognitive decline is accelerated following the first hospitalization with an additional, albeit attenuated, decline following the second hospitalization. It is notable that in Participant #1 (Pt 1) who has above average cognitive function at baseline if hospitalizations occurred at 2 and 4 years, the trajectory of cognitive decline remained well above the threshold for cognitive impairment at the end of 8 years of follow-up. However, for Pt 2 who had below average cognitive function at baseline (indicating less cognitive reserve), a single hospitalization at 2 years was associated with an accelerated decline in cognitive function resulting in them reaching the threshold for impairment at approximately 5 years which is 3 years earlier than would have been expected had they not required hospitalization. Similarly, Figure 1 B. shows simulated paths for global cognitive trajectories of the same two hypothetical individuals under scenarios of no hospitalization, a single medical hospitalization at 2 years, and a single surgical hospitalization at 2 years. The interpretation of the clinical implications of hospitalization for these patients is similar to that described in Figure 1 A, although for Pt 2, the time at which they reach the threshold for impairment is more accelerated if they required medical vs. surgical hospitalization.
The clinical implications of cognitive impairment after hospitalization are likely to be minimal for older adults with preserved cognitive reserve at the time of hospitalization, and will likely impose a minimal impact on their cognitive activities of daily living.
Study limitations
Our large study uses the resources of the MCSA, a prospective, population-based study of cognitive aging among Olmsted County residents. Retrospective information about hospitalizations is obtained through the REP medical records linkage system which contains accurate information of all health issues for the entire population. An important limitation is the possibility of unmeasured confounding, including time-dependent confounding arising after baseline. Although cognition is assessed every 15 months, the timing of the MCSA visits did not correspond to consistent time intervals surrounding hospitalization. Aging is associated with cognitive decline,40 and older individuals have an increased likelihood to be hospitalized.41 Therefore, hospitalization could be a contributor to cognitive decline or only a marker for comorbidities that are independently associated with cognitive loss. For modelling purposes, we estimated the slope of cognitive trajectory prior to hospitalization using data from all participants, including those who had no hospitalizations during follow-up. Although numerous covariates are included in our modelling, we cannot rule out that unmeasured confounders exist and that those who ultimately required hospitalizations may differ from those who did not. Although statistical adjusting accounts for diagnostic entities, it cannot consider the variable acuity of the given condition.
Conclusion
In conclusion, hospitalization of older adults is associated with acceleration of decline in multiple cognitive domains. This decline is more pronounced after non-elective compared to elective hospitalizations, and after hospitalizations for medical compared to surgical indications. The clinical implications of this acceleration will depend on the individual’s baseline cognitive reserve and expected longevity. Future studies are needed to better understand the factors underlying this association.
Supplementary Material
Supplementary Figure S1: Mayo Clinic Study of Aging (MCSA) participants in the present study.
Supplementary Table S1: Surgical hospitalization according to specialty
Supplementary Table S2: Medical hospitalization according to specialty
Supplementary Table S3a: Global Z-score longitudinal mixed effects model summary
Supplementary Table S3b: Variance covariance matrix estimated from longitudinal mixed effects model for global z-score.
Supplementary Table S4: Estimated annual slope of cognitive z-scores prior to hospitalization and change in slope following hospital admission when only first hospitalization is considered
Supplementary Table S5: Estimated annual slope of cognitive z-scores prior to hospitalization and change in slope following hospital admission from analysis adjusting for practice effects
Supplementary Statistical Methods
Acknowledgments
Role of the funding source:
This work was supported by NIH grants P50 AG016574, P30 AG062677, U01 AG006786, R01 AG034676, R01 AG41851, and R37 AG11378, R01 NS097495, R01 AG056366, the Elsie and Marvin Dekelboum Family Foundation, GHR Foundation, Alexander Family Alzheimer’s Disease Research Professorship of the Mayo Clinic, Liston Award, Alzheimer’s Association, Schuler Foundation and the Mayo Foundation for Medical Education and Research. Support is provided by the Rochester Epidemiology Project (R01 AG034676, PIs: WA Rocca and J St. Sauver) and the Mayo Clinic Center for Translational Sciences Activities, grant number UL1 TR000135 from the National Center for Advancing Translational Sciences.
Disclosures and declaration of interest:
Dr. Knopman served on a Data Safety Monitoring Board for the DIAN study. He serves on a Data Safety Monitoring Board for a tau therapeutic for Biogen but receives no personal compensation. He is an investigator in clinical trials sponsored by Biogen, Lilly Pharmaceuticals and the University of Southern California. He serves as a consultant for Samus Therapeutics, Third Rock and Alzeca Biosciences but receives no personal compensation. He receives research support from the NIH. Dr. Petersen is a consultant for Biogen, Inc., Roche, Inc., Merck, Inc., Genentech Inc., Eisai, Inc; has given educational lectures for GE Healthcare, receives publishing royalties from Mild Cognitive Impairment (Oxford University Press, 2003), UpToDate, and receives research support from the NIH/NIA. Dr. Mielke consults for Brain Protection Company, receives an unrestricted research grant from Biogen, and receives funding from NIA/NIH. Dr. Weingarten currently serves as a consultant to Medtronic in the role as chairman of the Clinical Endpoint Committee for the Prodigy Trial; has received research support from Respiratory Motion and unrestricted investigator-initiated grants from Merck. Dr. Vassilaki has received research funding from Roche and Biogen, receives research funding from NIH currently and has equity ownership in Abbott Laboratories, Johnson and Johnson, Medtronic and Amgen.
Dr. Vemuri receives support from NIH. Sprung, Schulte, Martin, Hanson, Schroeder, Laporta, White, and Warner have nothing to disclose.
Alphabetical List of Abbreviations:
- APOE ε4 allele
apolipoprotein E genotype (status)
- ICD
International Classification of Diseases
- ICU
intensive care unit
- MCI
mild cognitive impairment
- MCSA
Mayo Clinic Study of Aging
- REP
Rochester Epidemiology Project
- STROBE
Strengthening the Reporting of Observational Studies in Epidemiology
Footnotes
Conflicts of Interest: We, the authors, declare that we have no competing interests.
Note: JS and DRS, ACH have full access to all the study data and take responsibility for the integrity of the data and the accuracy of the data analysis. They also have final responsibility for the decision to submit for publication.
SUPPORTING INFORMATION
Additional Supporting Information may be found in the online version of this article.
REFERENCES
- 1.Ehlenbach WJ, Hough CL, Crane PK, et al. Association between acute care and critical illness hospitalization and cognitive function in older adults. JAMA 2010; 303(8): 763–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pandharipande PP, Girard TD, Jackson JC, et al. Long-term cognitive impairment after critical illness. N Engl J Med 2013; 369(14): 1306–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Schulte PJ, Warner DO, Martin DP, et al. Association between critical care admissions and cognitive trajectories in older adults. Crit Care Med 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Chen CC, Chiu MJ, Chen SP, Cheng CM, Huang GH. Patterns of cognitive change in elderly patients during and 6 months after hospitalisation: a prospective cohort study. Int J Nurs Stud 2011; 48(3): 338–46. [DOI] [PubMed] [Google Scholar]
- 5.Wilson RS, Hebert LE, Scherr PA, Dong X, Leurgens SE, Evans DA. Cognitive decline after hospitalization in a community population of older persons. Neurology 2012; 78(13): 950–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Mathews SB, Arnold SE, Epperson CN. Hospitalization and cognitive decline: Can the nature of the relationship be deciphered? Am J Geriatr Psychiatry 2014; 22(5): 465–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Brown CHt, Sharrett AR, Coresh J, et al. Association of hospitalization with long-term cognitive and brain MRI changes in the ARIC cohort. Neurology 2015; 84(14): 1443–53. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.James BD, Wilson RS, Capuano AW, et al. Cognitive decline after elective and non-elective hospitalizations in older adults. Neurology 2019; 92(7): e690–e9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Chodosh J, Seeman TE, Keeler E, et al. Cognitive decline in high-functioning older persons is associated with an increased risk of hospitalization. J Am Geriatr Soc 2004; 52(9): 1456–62. [DOI] [PubMed] [Google Scholar]
- 10.Salas M, Inť Veld BA, van der Linden PD, Hofman A, Breteler M, Stricker BH. Impaired cognitive function and compliance with antihypertensive drugs in elderly: the Rotterdam Study. Clin Pharmacol Ther 2001; 70(6): 561–6. [DOI] [PubMed] [Google Scholar]
- 11.Teeters DA, Moua T, Li G, et al. Mild cognitive impairment and risk of critical illness. Crit Care Med 2016; 44(11): 2045–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Schulte PJ, Martin DP, Deljou A, et al. Effect of cognitive status on the receipt of procedures requiring anesthesia and critical care admissions in older adults. Mayo Clin Proc 2018; 93(11): 1552–62. [DOI] [PubMed] [Google Scholar]
- 13.Iwashyna TJ, Ely EW, Smith DM, Langa KM. Long-term cognitive impairment and functional disability among survivors of severe sepsis. JAMA 2010; 304(16): 1787–94. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Roberts RO, Geda YE, Knopman DS, et al. The Mayo Clinic Study of Aging: design and sampling, participation, baseline measures and sample characteristics. Neuroepidemiology 2008; 30(1): 58–69. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Petersen RC, Roberts RO, Knopman DS, et al. Prevalence of mild cognitive impairment is higher in men. The Mayo Clinic Study of Aging. Neurology 2010; 75(10): 889–97. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Kokmen E, Smith GE, Petersen RC, Tangalos E, Ivnik RC. The short test of mental status. Correlations with standardized psychometric testing. Arch Neurol 1991; 48(7): 725–8. [DOI] [PubMed] [Google Scholar]
- 17.Vemuri P, Lesnick TG, Przybelski SA, et al. Association of lifetime intellectual enrichment with cognitive decline in the older population. JAMA Neurol 2014; 71(8): 1017–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Petersen RC. Mild cognitive impairment as a diagnostic entity. J Intern Med 2004; 256(3): 183–94. [DOI] [PubMed] [Google Scholar]
- 19.McKhann GM, Knopman DS, Chertkow H, et al. The diagnosis of dementia due to Alzheimer’s disease: recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 2011; 7(3): 263–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders. DSM-IV. 4th ed. Washington, D.C.: American Psychiatric Association. ; 1994. [Google Scholar]
- 21.St Sauver JL, Grossardt BR, Yawn BP, et al. Data resource profile: the Rochester Epidemiology Project (REP) medical records-linkage system. Int J Epidemiol 2012; 41(6): 1614–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.St Sauver JL, Grossardt BR, Yawn BP, Melton LJ 3rd, Rocca WA. Use of a medical records linkage system to enumerate a dynamic population over time: the Rochester epidemiology project. Am J Epidemiol 2011; 173(9): 1059–68. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Rocca WA, Yawn BP, St Sauver JL, Grossardt BR, Melton LJ. History of the Rochester Epidemiology Project: half a century of medical records linkage in a US population. Mayo Clin Proc 2012; 87(12): 1202–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Vemuri P, Lesnick TG, Przybelski SA, et al. Age, vascular health, and Alzheimer disease biomarkers in an elderly sample. Ann Neurol 2017; 82(5): 706–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Greenwood PM, Parasuraman R, Haxby JV. Changes in visuospatial attention over the adult lifespan. Neuropsychologia 1993; 31(5): 471–85. [DOI] [PubMed] [Google Scholar]
- 26.Harada CN, Natelson Love MC, Triebel KL. Normal cognitive aging. Clin Geriatr Med 2013; 29(4): 737–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Hayden KM, Welsh-Bohmer KA. Epidemiology of cognitive aging and Alzheimer's disease: contributions of the Cache county Utah study of memory, health and aging. Curr Top Behav Neurosci 2012; 10: 3–31. [DOI] [PubMed] [Google Scholar]
- 28.Singh-Manoux A, Kivimaki M, Glymour MM, et al. Timing of onset of cognitive decline: results from Whitehall II prospective cohort study. BMJ 2012; 344: d7622. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Sprung J, Roberts RO, Weingarten TN, et al. Postoperative delirium in elderly patients is associated with subsequent cognitive impairment. Br J Anaesth 2017; 119(2): 316–23. [DOI] [PubMed] [Google Scholar]
- 30.Zanetti M, Ballabio C, Abbate C, Cutaia C, Vergani C, Bergamaschini L. Mild cognitive impairment subtypes and vascular dementia in community-dwelling elderly people: a 3-year follow-up study. J Am Geriatr Soc 2006; 54(4): 580–6. [DOI] [PubMed] [Google Scholar]
- 31.Casserly I, Topol E. Convergence of atherosclerosis and Alzheimer's disease: inflammation, cholesterol, and misfolded proteins. Lancet 2004; 363(9415): 1139–46. [DOI] [PubMed] [Google Scholar]
- 32.Edwards JD, Ramirez J, Black SE. Unraveling the potential co-contributions of cerebral small vessel vasculopathy to the pathogenesis of Alzheimer's dementia. Alzheimers Res Ther 2015; 7(1): 49. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Xie Z, Tanzi RE. Alzheimer’s disease and post-operative cognitive dysfunction. Exp Gerontol 2006; 41(4): 346–59. [DOI] [PubMed] [Google Scholar]
- 34.Xie Z, Dong Y, Maeda U, et al. The inhalation anesthetic isoflurane induces a vicious cycle of apoptosis and amyloid beta-protein accumulation. J Neurosci 2007; 27(6): 1247–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Chen CW, Lin CC, Chen KB, Kuo YC, Li CY, Chung CJ. Increased risk of dementia in people with previous exposure to general anesthesia: a nationwide population-based case-control study. Alzheimers Dement 2014; 10(2): 196–204. [DOI] [PubMed] [Google Scholar]
- 36.Schulte PJ, Roberts RO, Knopman DS, et al. Association between exposure to anaesthesia and surgery and long-term cognitive trajectories in older adults: report from the Mayo Clinic Study of Aging. Br J Anaesth 2018; 121(2): 398–405. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37.Sprung J, Schulte PJ, Knopman DS, et al. Cognitive function after surgery with regional or general anesthesia: A population-based study. Alzheimers Dement 2019; 15(10): 1243–52. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Whitlock EL, Diaz-Ramirez LG, Smith AK, Boscardin WJ, Avidan MS, Glymour MM. Cognitive change after cardiac surgery versus cardiac catheterization: a population-based study. Ann Thorac Surg 2019; 107(4): 1119–25. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Sweet JJ, Finnin E, Wolfe PL, et al. Absence of cognitive decline one year after coronary bypass surgery: comparison to nonsurgical and healthy controls. Ann Thorac Surg 2008; 85(5): 1571–8. [DOI] [PubMed] [Google Scholar]
- 40.Nadelson MR, Sanders RD, Avidan MS. Perioperative cognitive trajectory in adults. Br J Anaesth 2014; 112(3): 440–51. [DOI] [PubMed] [Google Scholar]
- 41.Murad K, Goff DC Jr., Morgan TM, et al. Burden of comorbidities and functional and cognitive impairments in elderly patients at the initial diagnosis of heart failure and their impact on total mortality: The Cardiovascular Health Study. JACC Heart Fail 2015; 3(7): 542–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplementary Figure S1: Mayo Clinic Study of Aging (MCSA) participants in the present study.
Supplementary Table S1: Surgical hospitalization according to specialty
Supplementary Table S2: Medical hospitalization according to specialty
Supplementary Table S3a: Global Z-score longitudinal mixed effects model summary
Supplementary Table S3b: Variance covariance matrix estimated from longitudinal mixed effects model for global z-score.
Supplementary Table S4: Estimated annual slope of cognitive z-scores prior to hospitalization and change in slope following hospital admission when only first hospitalization is considered
Supplementary Table S5: Estimated annual slope of cognitive z-scores prior to hospitalization and change in slope following hospital admission from analysis adjusting for practice effects
Supplementary Statistical Methods