Abstract
Background
Older adults surviving critical illness often experience new or worsening functional impairments. Modifiable positive psychological constructs such as resilience may mitigate post-intensive care morbidity.
Research Question
Is pre-ICU resilience associated with: (1) post-ICU survival; (2) the drop in functional independence during the ICU stay; or (3) the trend in predicted independence before vs after the ICU stay?
Study Design and Methods
This retrospective cohort study was performed by using Medicare-linked Health and Retirement Study surveys from 2006 to 2018. Older adults aged ≥ 65 years admitted to an ICU were included. Resilience was calculated prior to ICU admission. The resilience measure was defined from the Simplified Resilience Score, which was previously adapted and validated for the Health and Retirement Study. Resilience was scored by using the Leave-Behind survey normalized to a scale from 0 (lowest resilience) to 12 (highest resilience). Outcomes were survival and probability of functional independence. Survival was modeled by using Gompertz models and independence using joint survival models adjusting for sociodemographic and clinical variables. Average marginal effects were estimated to determine independence probabilities.
Results
Across 3,409 patients aged ≥ 65 years old admitted to ICUs, preexisting frailty (30.5%) and cognitive impairment (24.3%) were common. Most patients were previously independent (82.7%). Mechanical ventilation occurred in 14.8% and sepsis in 43.2%. Those in the highest resilience group (vs lowest resilience) had a lower risk of post-ICU mortality (adjusted hazard ratio, 0.81; 95% CI, 0.70-0.94). Higher resilience was associated with greater likelihood of post-ICU functional independence (estimated probability of functional independence 5 years after ICU discharge in highest-to-lowest resilience groups (adjusted hazard ratio [95% CI]): 0.53 (0.33-0.74), 0.47 (0.26-0.68), 0.49 (0.28-0.70), and 0.36 (0.17-0.55); P < .01. Resilience was not associated with a difference in the drop in independence during the ICU stay or a difference in the pre-ICU vs post-ICU trend in predicted independence .
Interpretation
ICU survivors with higher resilience had increased rates of survival and functional independence, although the slope of functional decline before vs after the ICU stay did not differ according to resilience group.
Key Words: critical care, critical illness, ICU, recovery of function, resilience
Take-home Points.
Study Question: Is pre-ICU psychosocial resilience associated with post-ICU survival and functional independence in older adults?
Results: In this retrospective cohort study, higher pre-ICU resilience was associated with increased survival and likelihood of functional independence in survivors.
Interpretation: Older adult ICU survivors who had higher pre-ICU psychological resilience scores had higher rates of survival and post-ICU functional independence. Resilience is a modifiable trait and may be a target to improve survivorship.
Older adults who survive critical illness often experience new or worsening functional, psychological, and cognitive impairments persisting long after hospitalization, collectively known as post-intensive care syndrome (PICS).1 The ICU admission is associated with immobility,2 malnutrition,3 swallowing dysfunction,4 and neuromuscular weakness,5 all of which carry important consequences for delayed physical recovery, functional decline,6 risk for readmission,7 and death.2,8 Hence, ICU admissions serve as inflection points for older adults,9 leading to altered trajectories of decline.10 Studies that further our understanding of survivorship after ICU discharge are essential to improving these long-term consequences.
ICU survivorship represents a dynamic process of recovery incorporating changes in self, biography, and identity based on experiences of critical illness.11 The field has largely emphasized prevalence and risk factors for PICS.12,13 Characteristics that protect patients from stressors, such as resilience,14,15 or an individual’s capacity to resist, adapt to, recover, or grow from adversity, have not been well described. Resilience, a modifiable trait,16 buffers the impact of chronic illness on disability later in life14 and improves health outcomes across many contexts.14,17,18 Resilience takes on a prominent role in other fields such as oncology,19 but it remains unknown how pre-ICU resilience affects post-ICU survivorship and function. Psychological well-being and behavioral interventions are not directly incorporated into existing ICU bundles (eg, Society of Critical Care Medicine's ABCDEF bundle20). If resilience buffers the ill effects of the ICU stay, then interventions fostering resilience could be used to facilitate ICU survivorship.
The purpose of the current study was to examine the effects of pre-ICU resilience on post-ICU survival and functional outcomes using a nationally representative longitudinal survey of older adults. We hypothesized that higher psychosocial resilience would be associated with increased survival, a smaller immediate drop in functional independence, and a smaller decline in function over time in older adult ICU survivors relative to pre-ICU functional declines.
Study Design and Methods
Study Design and Population
This retrospective cohort study was conducted by using longitudinal data from the Health and Retirement Study (HRS), which is an ongoing, nationally representative longitudinal study of US community-based adults aged ≥ 50 years, supported by the National Institute on Aging (NIA U01AG009740). The HRS has assessed physical, psychological, and economic measures through detailed biennial longitudinal surveys since 1992.21
The HRS cohort was linked to Medicare claims files to identify participants aged ≥ 65 years who had an ICU admission between 2006 and 2018. Only the first ICU admission was considered. Because our resilience measure (discussed in the following section) is found in the “Leave-Behind” survey performed every 4 years (which is a longer, more detailed survey performed alongside every other core HRS measure survey), we only included patients who had their first ICU admission within 5 years of completing a Leave-Behind survey. This study was approved by the institutional review board (#16-19185) at the University of California San Francisco, and it adhered to the Strengthening the Reporting of Observational Studies in Epidemiology guidelines for cohort-based studies (e-Appendix 1).
The ICU admission was the unit of analysis. Inclusion for ICU admissions using revenue center codes from inpatient Medicare files are found elsewhere22 and only included ICUs for critically ill patients (eg, step-down units were excluded). The cohort flow is presented in e-Figure 1.
Resilience Measure
We based our resilience definition from the Simplified Resilience Score, which was adapted, developed, and validated for the HRS.18 The Simplified Resilience Score uses a conceptualization of resilience23 encompassing perseverance, equanimity, realization that life has purpose, self-reliance, and existential aloneness. The resilience measure served as the study predictor of interest. This includes 12 questions from the HRS Leave-Behind survey using Likert-scale responses, each normalized to a scale from 0 (lowest resilience) to 12 (highest resilience) (e-Table 1). For missing items, the mean of nonmissing items was computed, and any missing values replaced with that mean. Rates of missingness for each resilience question are in e-Table 2.
No “cutoff” or predefined categories define resilience. We therefore divided the sample into four quartiles of resilience across the distribution of scores in the cohort ([R1] represented the lowest resilience quartile and [R4] the highest) (e-Fig. 2). Detail on how resilience was calculated is described in e-Appendix 1.
Outcomes
The two outcomes were: (1) survival following hospitalization; and (2) freedom from disability (functional independence). Survival data were determined via linkage to the National Death Index. We examined freedom from disability as not having any need for assistance in any activity of daily living (ADL) or instrumental activity of daily living (IADL) (“ADL/IADL independence”) (e-Appendix 1). Similar to other investigators, we considered dependence on any of these 11 ADLs or IADLs as disability.12 For patients who died during the study period, disability data from HRS exit interviews were included; these interviews elicit information from next-of-kin about the deceased participant’s function in the months prior to death.
Covariates
In the primary analysis, sociodemographic and preexisting clinical characteristics present at the time resilience was assessed in the 5 years prior to critical illness were selected for use. Self-reported HRS sociodemographic factors included: sex, self-identified race or ethnicity, net worth, and schooling. Race and ethnicity categories were defined by the HRS investigators and categorized as Hispanic or non-Hispanic, Black or White. Age at admission of the hospitalization containing the index ICU stay was also included. All preexisting clinical characteristics were defined similar to our previous work.22 Cognitive function used the validated Wu-Glymour method,20 and frailty was assessed by using HRS-adapted criteria24 based on a Functional Domains Model.25 Also included was a surrogate for multimorbidity as defined elsewhere.22 Full definitions of these covariates are provided in e-Appendix 1.
To account for differing intervals from the resilience measure to ICU admission in the adjusted models, a time variable was included representing the time from the Leave-Behind survey to the index hospitalization. For descriptive purposes, the Lee index was included to characterize overall prognosis prior to ICU admission; this index is an HRS-validated26 measure of all-cause 4-year mortality. A score ≥ 13 correlates to > 50% mortality risk. This index was not included as a covariate because many of its predictors were included in the current analysis.
In a sensitivity analysis, we added selected factors associated with the ICU admission and critical illness were added to the study model: need for mechanical ventilation, sepsis with organ dysfunction, and ICU type (definitions of these measures are provided in e-Appendix 1).
Data Analysis
This study sought to determine whether: (1) survival differed across resilience groups; (2) ADL/IADL independence drop during the index admission differed across resilience groups; and (3) whether the trend (slopes) of predicted ADL/IADL independence differed before vs after ICU stays within each resilience group. For functional outcomes, to assess for trends (slopes) of and drops in functional independence, we also calculated predicted probabilities of functional independence at 5 years prior to and following the ICU admission and immediately prior to and following the ICU admission. Survival models used a Gompertz survival function27 adjusted for the non-ICU covariates: sex, self-identified race or ethnicity, net worth, schooling, frailty, cognitive impairment, and multimorbidity. A joint disability-survival model was used based on repeated measures of ADL/IADL independence from core HRS surveys. A Cox regression was used to obtain an inverse probability weight for survival and a random effects logistic regression for ADL/IADL independence. We included a binary ICU indicator variable (for prior to vs following the ICU comparison), a time to interview variable (for changes in time for disability), and the resilience variable (for comparison across resilience groups), allowing for a three-way interaction within the model. The log odds of ADL/IADL independence were determined. As with other studies,28,29 the predicted population rates and average marginal effects were used to illustrate the effect of resilience on the probability of ADL and IADL independence. These probabilities are presented in the results and figures for ease of interpretability, and model estimates are presented in e-Appendix 1.
To determine ADL/IADL independence and compare changes within or across groups, we determined the probability of ADL/IADL independence at 5 years prior to ICU admission, just prior to the ICU admission, the change in independence per year prior to the ICU admission, the drop in independence before vs after the ICU admission (treated as a single point in time for the entire index hospitalization), the change in independence per year following the ICU admission and independence 5 years after the index hospitalization, and slopes after ICU stays across resilience groups (e-Fig 3). These factors were considered secondary outcomes.
For the sensitivity analysis, we used identical methods as the aforementioned but further adjusted for ICU-specific confounders using Medicare data that could affect post-ICU survival and ADL/IADL independence outcomes. These included need for mechanical ventilation, sepsis with end-organ dysfunction, and ICU type. A subgroup analysis was also performed to test the hypotheses on a more homogeneous subgroup, namely sepsis with organ dysfunction (sepsis was removed as a covariate in this sensitivity analysis).
To account for HRS survey design and differential participant selection, all analyses included survey weighting. ORs were computed for all regressions and 95% CIs using SEs of the average. Statistical analyses were performed in SAS version 8.3 (SAS Institute, Inc.) and Stata version 18 (StataCorp).
Results
Across 21,408 patients who performed a Leave-Behind survey between 2006 and 2018 in the HRS, 3,409 patients were aged ≥ 65 years, had an ICU admission, and met inclusion criteria. Functional analyses were performed on 2,920 patients who survived to discharge and had a post-ICU core survey or exit interview with a proxy (e-Fig 1). Sociodemographic and clinical characteristics for each resilience group are presented in Table 1,26,28 and the distribution of resilience quartiles according to resilience scores are shown in e-Figure 2. Across the entire cohort, the median age was 77.8 years. There were more women (52.0%), and the cohort self-reported predominately as White and non-Hispanic (78.3%). Most patients were married (55.7%).
Table 1.
Sociodemographic and Preexisting Clinical Characteristics Across Resilience Quartiles
| Characteristica | Resilience Quartile (R1 = Lowest, R4 = Highest) |
Overall (N = 3,409) | |||
|---|---|---|---|---|---|
| R1 (n = 853) | R2 (n = 862) | R3 (n = 844) | R4 (n = 850) | ||
| Age, yb | |||||
| Median (min, max) | 77.6 (65.0, 98.7) | 78.5 (65.0, 99.4) | 78.3 (65.0, 104) | 77.0 (65.0, 104) | 77.9 (65.0, 104) |
| Female sex | 491 (57.6%) | 445 (51.6%) | 428 (50.7%) | 408 (48.0%) | 1,772 (52.0%) |
| Race and ethnicityc | |||||
| White, non-Hispanic | 633 (74.2%) | 651 (75.5%) | 683 (80.9%) | 703 (82.7%) | 2,670 (78.3%) |
| Black, non-Hispanic | 125 (14.7%) | 133 (15.4%) | 93 (11.0%) | 89 (10.5%) | 440 (12.9%) |
| Hispanic | 82 (9.6%) | 61 (7.1%) | 54 (6.4%) | 43 (5.1%) | 240 (7.0%) |
| Educationd | |||||
| HS or less | 649 (76.1%) | 593 (68.7%) | 500 (59.2%) | 431 (50.7%) | 2,173 (63.8%) |
| Some college or more | 203 (23.8%) | 267 (31.0%) | 343 (40.6%) | 419 (49.3%) | 1,232 (36.2%) |
| Marital statusd | |||||
| Married | 396 (46.4%) | 470 (54.5%) | 482 (57.1%) | 551 (64.8%) | 1,899 (55.7%) |
| Single | 145 (17.0%) | 133 (15.4%) | 113 (13.4%) | 89 (10.5%) | 480 (14.1%) |
| Widowed | 312 (36.6%) | 259 (30.0%) | 247 (29.3%) | 209 (24.6%) | 1027 (30.1%) |
| Net worth | |||||
| Median (min, max) | 72,800 (–486,000, 10,400,000) | 136,000 (–196,000, 13,700,000) | 200,000 (–353,000, 14,600,000) | 296,000 (–52,600, 12,000,000) | 160,000 (–486,000, 14,600,000) |
| Preexisting illnesses | |||||
| ≥ 3 chronic diseasese | 613 (71.9%) | 516 (59.9%) | 484 (57.3%) | 423 (49.8%) | 2,036 (59.7%) |
| CIND | 264 (30.9%) | 248 (28.8%) | 175 (20.7%) | 143 (16.8%) | 830 (24.3%) |
| Dementia | 90 (10.6%) | 70 (8.1%) | 42 (5.0%) | 27 (3.2%) | 229 (6.7%) |
| Fraild | 374 (43.8%) | 265 (30.7%) | 246 (29.1%) | 155 (18.2%) | 1,040 (30.5%) |
| Lee index score ≥ 13f | 209 (24.5%) | 153 (17.7%) | 92 (10.9%) | 69 (8.1%) | 523 (15.3%) |
| ADL/IADL dependenciesd | |||||
| 0 | 587 (68.8%) | 704 (81.7%) | 745 (88.3%) | 782 (92.0%) | 2,818 (82.7%) |
| 1-2 | 126 (14.8%) | 85 (9.9%) | 64 (7.6%) | 47 (5.5%) | 322 (9.4%) |
| ≥ 3 | 136 (15.9%) | 70 (8.1%) | 33 (3.9%) | 20 (2.4%) | 259 (7.6%) |
| ICU characteristicsg | |||||
| Required MV | 153 (17.9%) | 136 (15.8%) | 120 (14.2%) | 95 (11.2%) | 504 (14.8%) |
| Sepsis | 385 (45.1%) | 395 (45.8%) | 357 (42.3%) | 335 (39.4%) | 1,472 (43.2%) |
| Median LOS (min, max) | 7.00 (1.00, 46.0) | 7.00 (1.00, 80.0) | 7.00 (1.00, 121) | 6.00 (1.00, 73.0) | 7.00 (1.00, 121) |
| LOS ≥ 7 d | 428 (50.2%) | 447 (51.9%) | 426 (50.5%) | 423 (49.8%) | 1,724 (50.6%) |
ADL = activity of daily living; CIND = cognitive impairment not dementia; HS = high school; IADL = instrumental activity of daily living; LOS = length of stay; min = minimum; max = maximum; MV = mechanical ventilation; R1-4 = resilience quartiles 1 to 4.
Proxy data were used in 2% to 6% of patients for functional impairment data in post-ICU survivors (proxies interviewed because patient died prior to core survey and exit survey used for ADL/IADL impairments).
Patient data represent characteristics of patients at the first ICU admission.
Other races or ethnicities include American Indian, Asian, “something else,” do not know, or declined to answer. These were collapsed to Other in the Health and Retirement Study to protect respondent confidentiality, given the small sample sizes.
Missing data not included and represents ≤ 2% of the cohort except for frailty, which was < 6.7%.
Self-reported chronic conditions include hypertension, diabetes mellitus, cancer, chronic lung disease, coronary heart disease, congestive heart failure, stroke, arthritis, and psychiatric conditions. Three or more chronic conditions was defined as “multimorbidity,” similar to previous work.28
Health and Retirement Study-validated30 measure of all cause 4-year mortality. A score ≥ 13 correlates to > 50% mortality risk.
Notable sociodemographic and clinical differences were present across resilience quartiles (Table 1). In general, the highest resilience group is younger, more educated, has higher net worth, and has a higher number of individuals in marriages/partnerships than the lowest resilience groups. They also have lower rates of multimorbidity, cognitive impairments, dementia, frailty, and disability. Rates of sepsis and need for mechanical ventilation were lowest in the highest resilience groups.
Outcomes
Adjusted hazard ratios for survival and estimated marginal probabilities of ADL/IADL independence are presented in Table 2. Model estimates are shown in e-Table 3, and marginal effects for adjusted and unadjusted models are presented in e-Table 4. For the highest resilience quartile, the adjusted hazard ratio for mortality was 0.81 (95% CI, 0.70-0.94) compared with the lowest resilience quartile (P < .01). Adjusted rates for other resilience quartiles were associated with lower mortality hazard ratio compared with the lowest resilience quartile but did not reach statistical significance. Adjusted and unadjusted survival using Gompertz models are shown in Figure 1 and e-Figure 4, respectively.
Table 2.
HRs for Survival and Marginal Probabilities for Functional Independence and Trajectory Outcomes Across Resilience Quartiles
| HR (95% CI) | P Value | |
|---|---|---|
| Adjusted risk of mortality | ||
| Resilience group | ||
| R1 | Reference | … |
| R2 | 0.97 (0.85 to 1.10) | .63 |
| R3 | 0.93 (0.82 to 1.06) | .30 |
| R4 (highest resilience quartile) | 0.81 (0.70 to 0.94) | < .01 |
| Model Outcome | Coefficient Representing Probability or Slope (95% CI) | P Value |
|---|---|---|
| Adjusted probability of ADL and IADL independence | ||
| Drop in the probability of independence at ICU admission (prior to admission vs following discharge)a | ||
| R1 | 0.19 (0.12 to 0.26) | < .01 |
| R2 | 0.20 (0.12 to 0.28) | < .01 |
| R3 | 0.23 (0.14 to 0.33) | < .01 |
| R4 | 0.20 (0.10 to 0.29) | < .01 |
| Post-ICU change of probability of independence per year (slope of the probability of independence following discharge)b | ||
| R1 | –0.02 (–0.04 to –0.01) | < .01 |
| R2 | –0.01 (–0.03 to 0) | .14 |
| R3 | –0.02 (–0.04 to 0) | .02 |
| R4 | –0.02 (–0.03 to 0) | .05 |
All models were adjusted for sex, self-identified race or ethnicity, net worth, schooling, frailty, cognitive impairment, and multimorbidity. The lowest resilience quartile (R1) is the reference group for hazard ratios (HRs). P values < .05 were defined as statistically significant. P values for individual probabilities (eg, probability of a resilience group being independent at 5 years following discharge) are not shown. For changes and differences between groups, P values represent whether the change or difference exists (ie, they represent a statistical test being performed). ADL = activity of daily living; IADL = instrumental activity of daily living; R1-R4 = resilience quartiles 1 to 4.
Estimates of the differences in drop in probability of independence between each resilience group were not statistically significant and are not presented. For example, the difference in the drops in functional independence prior to and following ICU in the highest resilience group (R4) vs the lowest resilience group (R1) was not statistically significant. Additional results and ORs are provided in e-Appendix 1.
Estimates of differences between pre-ICU slope and post-ICU slope for each resilience group was not statistically significant and not presented. For example, the difference between the pre-ICU downward slope of probability independence vs the post-ICU downward slope in the highest resilience group (R4) did not reach statistical significance.
Figure 1.
Gompertz survival model adjusted for non-ICU covariates. Survival was adjusted for non-ICU covariates: age, sex, self-identified race or ethnicity, net worth, schooling, frailty, cognitive impairment, and multimorbidity. The highest (R4) to lowest (R1) resilience groups are shown in the colored lines. Bands around survival curves represent the 95% CI. R1-R4 = resilience quartiles 1 to 4.
Figure 2 presents rates of accumulated ADL and IADL dependence stratified according to resilience across three individual surveys prior to ICU admission and three surveys following ICU admission. All resilience quartiles exhibited increasing rates of ADL or IADL dependence across surveys along with increasing rates of death. Most patients who died had three or more ADL or IADL impairments prior to death when their cases were assessed per HRS exit interviews with next-of-kin. Highest to lowest resilience quartiles showed decreasing rates of ADL/IADL independence prior to and following ICU discharge.
Figure 2.
Accumulated activities of daily living and instrumental activities of daily living dependence according to resilience quartile at each survey time point. Using Health & Retirement Study (HRS) core surveys and exit interviews from proxies in patients who died following their ICU stays, activities of daily living and instrumental activities of daily living dependence was recorded. Resilience quartile 4 represents the quartile with the highest resilience scores and 1 represents the lowest. Missing was < 1% and not shown.
At each time point analyzed (5 years prior to ICU admission, at the time of ICU admission, at the time of ICU discharge, and 5 years after ICU discharge), the adjusted probability of independence was higher among patients in the highest resilience group compared with others (all P < .05). However, slopes of independence, both before and after the stay, did not differ across resilience groups (adjusted OR, 1.07; 95% CI, 0.99-1.16) (e-Table 3, Fig 3). Notably, there were no other notable differences between or within groups for slopes of independence, drops in independence during the ICU stay, or changes between pre-ICU vs post-ICU stay slopes (e-Tables 3-7; Fig 3; e-Table 4 and Table 2 for average marginal effects for adjusted models).
Figure 3.
Disability curves with and without covariate adjustments. This figure presents an individual pane for each resilience quartile. Dotted lines represent unadjusted models, and solid lines are adjusted for non-ICU covariates: sex, self-identified race or ethnicity, net worth, schooling, frailty, cognitive impairment, and multimorbidity. Functional independence models were calculated using a Cox regression to obtain an inverse probability weight for survival and a random effects logistic regression for activities of daily living/instrumental activities of daily living independence. A binary ICU indicator variable (for prior to vs following the ICU comparison), a time to interview variable (for changes in time for disability), and the resilience variable (for comparison across resilience groups) were included, allowing for a three-way interaction within the model. The log odds of activities of daily living/instrumental activities of daily living independence were determined. The x-axis is in years, and the 0 time point represents the ICU admission.
Sensitivity and Subgroup Analyses
Resilience quartiles follow a similar direction and pattern to the primary analysis for sensitivity and subgroup analyses for survival models and independence (e-Figs 5-6 for survival; e-Fig 7 for ADL/IADL independence curves; e-Tables 3-4 for primary analysis; e-Tables 8-9 for subgroup analysis).
Discussion
In this study using national longitudinal Medicare-linked survey data, we found that increased pre-ICU resilience is associated with a 19% increased chance of survival and more than twice as likely chance of having ADL/IADL independence compared with the lowest resilience group. These interactions remain statistically significant after controlling for additional critical illness covariates and in a sepsis-only subset. However, resilience was not associated with a difference in the drop of functional decline across resilience groups or a change in the slope of functional decline pre-ICU vs post-ICU discharge within groups. Taken together, these results suggest that resilience is associated with a difference in mortality following ICU admission. Although higher resilience was associated with a greater likelihood of functional independence after ICU stays, these differences are likely explained by differences in the probability of independence prior to the ICU admission. The possibility that interventions to enhance resilience could improve post-ICU survival and morbidity merits further study.
The current work uniquely examines resilience prior to critical illness, adding to earlier studies which have shown that positive psychosocial constructs such as dispositional optimism are associated with improved outcomes in older adults.17,31,32 We build off previous work examining the interaction between post-ICU resilience and morbidity, which found that higher vs lower resilience assessed following the ICU stay is associated with improved mental health, as well as mobility impairment among survivors.33 If we assume that resilience is modifiable, at least following significant and acute stressors,34 then measuring resilience in ICU survivors following an ICU stay could represent either preexisting resilience or resilience affected by their illness and recovery. Our analyses now address the question of how pre-ICU resilience, assessed prior to the ICU stay, could be a protective trait for survival and post-ICU morbidity. Importantly, we did not show that resilience mitigates the downward slopes of functional independence in ICU survivors. It is possible that preexisting resilience is protective against overall critical illness but may not sustain it into the post-ICU course.
Resilience represents one element of interacting constructs that could affect psychosocial well-being in response to stressors. Resilience itself, while lacking a uniformly accepted definition, is a transdisciplinary concept representing the intrinsic ability for a system (in our case, the health and identity of an older adult) to partially or fully bounce back from a perturbation (eg, critical illness).15 Although we show a correlation between underlying psychosocial resilience and ADL/IADL independence, other important related concepts are not captured by our measure but are known to affect post-ICU disability outcomes. These include social support35 and neighborhood socioeconomics,30 which affect resilience in complex ways.36 Equally important to resilience is the need for further characterization of stressors,37 namely the critical illness experience. Many different interventions, medications, and traumas occur within the ICU and could confound our results. Although we control for important risk factors for post-ICU functional impairment such as age, need for mechanical ventilation, sepsis with end-organ dysfunction, and preexisting disability, other important ICU interventions such as use of systemic corticosteroids38 or physical therapy39 are not easily captured in Medicare data.
Resilience could affect comorbidities not directly measured in our study, as lower resilience in early life is associated with worse lipid profiles40 and higher likelihood of developing hypertension41 and heart failure42 in adulthood. Higher resilience is also associated with healthier lifestyle habits and physical activity, which confer a mortality benefit at a population level.43 In ICU survivors specifically, resilience is associated with lower psychological sequelae of critical illness and a greater ability to perform self-care,33 which could play a role in post-ICU recovery and survival. In our cohort, higher resilience groups had higher net worth, more advanced education, were more often married, and had better preexisting health. It remains unclear to what extent resilience stems from these opportunities or contributes to them. Nonetheless, resilience is multidimensional concept dependent on one’s environment, individual factors, and biographies, as well as one’s community.44 Perhaps most important from this study is a shift to emphasizing strengths in lieu of deficits when considering survivorship and growth following ICU-related traumas.
If resilience represents a modifiable trait, then it could serve as a target for interventions to improve and foster ICU survivorship and outcomes. Interventions to foster resilience have been shown to improve quality of life and decrease rates of depression and stress perception.34 However, substantial heterogeneity exists in types of resilience interventions. Most emphasize psychological interventions such as cognitive-behavioral therapy, mindfulness, and stress inoculation therapies, which may be difficult to perform during critical illness but might have a role following ICU discharge (after the stressor exposure).45 However, despite some promising but limited data on early ICU psychological interventions to promote survivorship and PICS prevention,46,47 such types of interventions are not well defined or well established in the ICU.
The strengths of the current study lay in our use of a longitudinal nationally representative cohort that followed up participants with serial surveys prior to and following ICU admission. This approach stands in contrast to inception cohorts, which enroll at the time of ICU admission or discharge and lack vital information on pre-ICU resilience and function. ADL/IADL independence is an important outcome to older adults, and outcomes beyond mortality are necessary to understand life following critical illness and its role as a contributor to decision-making in serious illness.
The current study has important limitations. Residual confounding could affect the results. Although adapted from a validated conceptualization of resilience (Simplified Resilience Score), the measure may not fully capture the entirety of psychosocial resilience. Since the COVID-19 pandemic, there is increasing recognition of interactions between psychological resilience and social support,48 and social support (or social resilience) is included in our measure. Notably, social isolation and individual socioeconomic disadvantage have been linked to worse outcomes in ICU survivors,30,34,49 and their relationship to psychosocial resilience in ICU patients merits further study. Future studies are needed to examine the extent to which resilience changes over time during and following critical illness. Although we chose outcomes (ie, mortality, functional independence) that are meaningful to older adults and present across all HRS core surveys, resilience could be associated with other important outcomes such as quality of life, cognition, and health expectations; these merit future study. Importantly, our cohort may not be generalizable across other ICUs. The current cohort is older, predominately female, and has lower rates of mechanical ventilation and longer length of stay.50 Further research is needed to understand how these could affect the interactions between resilience and survivorship outcomes.
Conclusions
Psychosocial resilience is associated with improvements in survival and post-ICU outcomes such as functional independence, although pre-ICU to post-ICU functional decline did not differ by resilience group. Further research is required to delineate which elements foster resilience and how other psychological constructs, such as optimism, or social support and social resilience, could affect PICS. Our findings suggest that resilience may be an important target for interventions in the ICU to improve survivorship.
Funding/Support
J. C. was supported by the UCSF Noyce Initiative for Digital Transformation in Computational Biology & Health, the Hellman Fellows Foundation, UCSF Anesthesia Department Seed Grant, and by the UCSF Claude D. Pepper Older Americans Independence Center funded by the National Institute on Aging [P30 AG044281]. Funding was provided by the National Institute on Aging [P01AG066605 and P30AG044281, A. K. S., S. L., J. B., and S. Y. J.; K24AG068312, A. K. S.; and K24AG066998, S. L.].
Financial/Nonfinancial Disclosures
None declared.
Acknowledgments
Author contributions: J. C. is the guarantor of this work. In accordance with the International Committee of Medical Journal Editors' recommendations, J. C., S. L., and A. K. S. were involved in the conception, design, analysis, interpretation of data, and drafting and had final approval of the version to be published. S. Y. J., J. B., and E. E. were involved in the design, acquisition of data, analysis, and drafting and had final approval of the version to be published. A. C. C. and J. H. M. were involved in the interpretation of data for the work and reviewing and had final approval of the version to be published. All authors agree to be accountable for all aspects of the work ensuring that questions related to the accuracy and integrity of any part of the work are appropriately investigated and resolved.
Role of sponsors: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Additional information: The e-Appendix, e-Figures, and e-Tables are available online under "Supplementary Data."
Supplementary Data
References
- 1.Mikkelsen M.E., Still M., Anderson B.J., et al. Society of Critical Care Medicine’s International Consensus Conference on prediction and identification of long-term impairments after critical illness. Crit Care Med. 2020;48(11):1670–1679. doi: 10.1097/CCM.0000000000004586. [DOI] [PubMed] [Google Scholar]
- 2.Ali N.A., O’Brien J.M., Hoffmann S.P., et al. Acquired weakness, handgrip strength, and mortality in critically ill patients. Am J Respir Crit Care Med. 2008;178(3):261–268. doi: 10.1164/rccm.200712-1829OC. [DOI] [PubMed] [Google Scholar]
- 3.Maeda K., Koga T., Akagi J. Tentative nil per os leads to poor outcomes in older adults with aspiration pneumonia. Clin Nutr. 2016;35(5):1147–1152. doi: 10.1016/j.clnu.2015.09.011. [DOI] [PubMed] [Google Scholar]
- 4.Jardine M., Miles A., Allen J. Dysphagia onset in older adults during unrelated hospital admission: quantitative videofluoroscopic measures. Geriatrics. 2018;3(4):66. doi: 10.3390/geriatrics3040066. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.De Jonghe B. Paresis acquired in the intensive care unit: a prospective multicenter study. JAMA. 2002;288(22):2859. doi: 10.1001/jama.288.22.2859. [DOI] [PubMed] [Google Scholar]
- 6.Inouye S.K., Bogardus S.T., Charpentier P.A., et al. A multicomponent intervention to prevent delirium in hospitalized older patients. N Engl J Med. 1999;340(9):669–676. doi: 10.1056/NEJM199903043400901. [DOI] [PubMed] [Google Scholar]
- 7.Prescott H.C., Langa K.M., Iwashyna T.J. Readmission diagnoses after hospitalization for severe sepsis and other acute medical conditions. JAMA. 2015;313(10):1055. doi: 10.1001/jama.2015.1410. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Brummel N.E., Balas M.C., Morandi A., Ferrante L.E., Gill T.M., Ely E.W. Understanding and reducing disability in older adults following critical illness. Crit Care Med. 2015;43(6):1265–1275. doi: 10.1097/CCM.0000000000000924. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Iwashyna T.J. Trajectories of recovery and dysfunction after acute illness, with implications for clinical trial design. Am J Respir Crit Care Med. 2012;186(4):302–304. doi: 10.1164/rccm.201206-1138ED. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Ferrante L.E., Pisani M.A., Murphy T.E., Gahbauer E.A., Leo-Summers L.S., Gill T.M. Functional trajectories among older persons before and after critical illness. JAMA Intern Med. 2015;175(4):523. doi: 10.1001/jamainternmed.2014.7889. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Kean S., Donaghy E., Bancroft A., Clegg G., Rodgers S. Theorising survivorship after intensive care: a systematic review of patient and family experiences. J Clin Nurs. 2021;30(17-18):2584–2610. doi: 10.1111/jocn.15766. [DOI] [PubMed] [Google Scholar]
- 12.Iwashyna T.J., Ely E.W., Smith D.M., Langa K.M. Long-term cognitive impairment and functional disability among survivors of severe sepsis. JAMA. 2010;304(16):1787. doi: 10.1001/jama.2010.1553. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Bagshaw S.M., Stelfox H.T., McDermid R.C., et al. Association between frailty and short- and long-term outcomes among critically ill patients: a multicentre prospective cohort study. Can Med Assoc J. 2014;186(2):E95–E102. doi: 10.1503/cmaj.130639. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Manning L.K., Carr D.C., Kail B.L. Do higher levels of resilience buffer the deleterious impact of chronic illness on disability in later life? Gerontologist. 2016;56(3):514–524. doi: 10.1093/geront/gnu068. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Brown L., Cohen B., Costello R., Brazhnik O., Galis Z. Conceptualizing a resilience research framework at the National Institutes of Health. Stress Health. 2023;39(S1):4–9. doi: 10.1002/smi.3260. [DOI] [PubMed] [Google Scholar]
- 16.Joyce S., Shand F., Tighe J., Laurent S.J., Bryant R.A., Harvey S.B. Road to resilience: a systematic review and meta-analysis of resilience training programmes and interventions. BMJ Open. 2018;8(6) doi: 10.1136/bmjopen-2017-017858. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lim K.K., Matchar D.B., Tan C.S., et al. The association between psychological resilience and physical function among older adults with hip fracture surgery. J Am Med Dir Assoc. 2020;21(2):260–266.e2. doi: 10.1016/j.jamda.2019.07.005. [DOI] [PubMed] [Google Scholar]
- 18.Coronado R.A., Robinette P.E., Henry A.L., et al. Bouncing back after lumbar spine surgery: early postoperative resilience is associated with 12-month physical function, pain interference, social participation, and disability. Spine J. 2021;21(1):55–63. doi: 10.1016/j.spinee.2020.07.013. [DOI] [PubMed] [Google Scholar]
- 19.Ang W.R., Ang W.H.D., Cham S.Q.G., de Mel S., Chew H.S.J., Devi M.K. Effectiveness of resilience interventions among cancer patients—a systematic review, meta-analysis, and meta-regression of randomised controlled trials. Eur J Oncol Nurs. 2023;67 doi: 10.1016/j.ejon.2023.102446. [DOI] [PubMed] [Google Scholar]
- 20.Wu Q., Tchetgen E.J.T., Osypuk T.L., White K., Mujahid M., Glymour M.M. Combining direct and proxy assessments to reduce attrition bias in a longitudinal study. Alzheimer Dis Assoc Disord. 2013;27(3):207–212. doi: 10.1097/WAD.0b013e31826cfe90. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Fisher G.G., Ryan L.H. Overview of the Health and Retirement Study and Introduction to the Special Issue. Work Aging Retire. 2018;4(1):1–9. doi: 10.1093/workar/wax032. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Cobert J., Jeon S.Y., Boscardin J., et al. Trends in geriatric conditions among older adults admitted to US ICUs between 1998 and 2015. Chest. 2022;161(6):1555–1565. doi: 10.1016/j.chest.2021.12.658. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Wagnild G.M., Young H.M. Development and psychometric evaluation of the Resilience Scale. J Nurs Meas. 1993;1(2):165–178. [PubMed] [Google Scholar]
- 24.Cigolle C.T., Ofstedal M.B., Tian Z., Blaum C.S. Comparing models of frailty: the Health and Retirement Study. J Am Geriatr Soc. 2009;57(5):830–839. doi: 10.1111/j.1532-5415.2009.02225.x. [DOI] [PubMed] [Google Scholar]
- 25.Strawbridge W.J., Shema S.J., Balfour J.L., Higby H.R., Kaplan G.A. Antecedents of frailty over three decades in an older cohort. J Gerontol B Psychol Sci Soc Sci. 1998;53B(1):S9–S16. doi: 10.1093/geronb/53b.1.s9. [DOI] [PubMed] [Google Scholar]
- 26.Lee S.J. Development and validation of a prognostic index for 4-year mortality in older adults. JAMA. 2006;295(7):801. doi: 10.1001/jama.295.7.801. [DOI] [PubMed] [Google Scholar]
- 27.Lee E.T., Go O.T. Survival analysis in public health research. Annu Rev Public Health. 1997;18(1):105–134. doi: 10.1146/annurev.publhealth.18.1.105. [DOI] [PubMed] [Google Scholar]
- 28.Glymour M.M., Avendano M. Can self-reported strokes be used to study stroke incidence and risk factors?: evidence from the Health and Retirement Study. Stroke. 2009;40(3):873–879. doi: 10.1161/STROKEAHA.108.529479. [DOI] [PubMed] [Google Scholar]
- 29.Parks A.L., Jeon S.Y., Boscardin W.J., et al. Long-term individual and population functional outcomes in older adults with atrial fibrillation. J Am Geriatr Soc. 2021;69(6):1570–1578. doi: 10.1111/jgs.17087. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Falvey J.R., Cohen A.B., O’Leary J.R., Leo-Summers L., Murphy T.E., Ferrante L.E. Association of social isolation with disability burden and 1-year mortality among older adults with critical illness. JAMA Intern Med. 2021;181(11):1433–1439. doi: 10.1001/jamainternmed.2021.5022. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Soliman G., Fortinsky R.H., Mangione K., et al. Impact of psychological resilience on walking capacity in older adults following hip fracture. J Am Geriatr Soc. 2022;70(11):3087–3095. doi: 10.1111/jgs.17930. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Tindle H.A., Chang Y.F., Kuller L.H., et al. Optimism, cynical hostility, and incident coronary heart disease and mortality in the Women’s Health Initiative. Circulation. 2009;120(8):656–662. doi: 10.1161/CIRCULATIONAHA.108.827642. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Maley J.H., Brewster I., Mayoral I., et al. Resilience in survivors of critical illness in the context of the survivors’ experience and recovery. Ann Am Thorac Soc. 2016;13(8):1351–1360. doi: 10.1513/AnnalsATS.201511-782OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Kunzler A.M., Helmreich I., Chmitorz A., et al. Psychological interventions to foster resilience in healthcare professionals. Cochrane Developmental, Psychosocial and Learning Problems Group, ed. Cochrane Database Syst Rev. 2020;2020(7) doi: 10.1002/14651858.CD012527.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Falvey J.R., Murphy T.E., Leo-Summers L., Gill T.M., Ferrante L.E. Neighborhood socioeconomic disadvantage and disability after critical illness. Crit Care Med. 2022;50(5):733–741. doi: 10.1097/CCM.0000000000005364. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36.Greene J.A., Loscalzo J. Putting the patient back together—social medicine, network medicine, and the limits of reductionism. N Engl J Med. 2017;377(25):2493–2499. doi: 10.1056/NEJMms1706744. [DOI] [PubMed] [Google Scholar]
- 37.Hadley E.C., Kuchel G.A., Newman A.B., et al. Report: NIA workshop on measures of physiologic resiliencies in human aging. J Gerontol Ser A. 2017;72(7):980–990. doi: 10.1093/gerona/glx015. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Herridge M.S., Cheung A.M., Tansey C.M., et al. One-year outcomes in survivors of the acute respiratory distress syndrome. N Engl J Med. 2003;348(8):683–693. doi: 10.1056/NEJMoa022450. [DOI] [PubMed] [Google Scholar]
- 39.Schaller S.J., Anstey M., Blobner M., et al. Early, goal-directed mobilisation in the surgical intensive care unit: a randomised controlled trial. Lancet. 2016;388(10052):1377–1388. doi: 10.1016/S0140-6736(16)31637-3. [DOI] [PubMed] [Google Scholar]
- 40.Bartone P.T., Valdes J.J., Sandvik A. Psychological hardiness predicts cardiovascular health. Psychol Health Med. 2016;21(6):743–749. doi: 10.1080/13548506.2015.1120323. [DOI] [PubMed] [Google Scholar]
- 41.Crump C., Sundquist J., Winkleby M.A., Sundquist K. Low stress resilience in late adolescence and risk of hypertension in adulthood. Heart Br Card Soc. 2016;102(7):541–547. doi: 10.1136/heartjnl-2015-308597. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42.Robertson J., Schiöler L., Torén K., et al. Mental disorders and stress resilience in adolescence and long-term risk of early heart failure among Swedish men. Int J Cardiol. 2017;243:326–331. doi: 10.1016/j.ijcard.2017.05.043. [DOI] [PubMed] [Google Scholar]
- 43.Bergh C., Udumyan R., Fall K., Almroth H., Montgomery S. Stress resilience and physical fitness in adolescence and risk of coronary heart disease in middle age. Heart Br Card Soc. 2015;101(8):623–629. doi: 10.1136/heartjnl-2014-306703. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 44.Krause K.D. The impact of resilience on health: lessons learned and future directions. Behav Med Wash DC. 2020;46(3-4):375–378. doi: 10.1080/08964289.2020.1790975. [DOI] [PubMed] [Google Scholar]
- 45.Chmitorz A., Kunzler A., Helmreich I., et al. Intervention studies to foster resilience—a systematic review and proposal for a resilience framework in future intervention studies. Clin Psychol Rev. 2018;59:78–100. doi: 10.1016/j.cpr.2017.11.002. [DOI] [PubMed] [Google Scholar]
- 46.Vranceanu A.M., Bannon S., Mace R., et al. Feasibility and efficacy of a resiliency intervention for the prevention of chronic emotional distress among survivor-caregiver dyads admitted to the neuroscience intensive care unit: a randomized clinical trial. JAMA Netw Open. 2020;3(10) doi: 10.1001/jamanetworkopen.2020.20807. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 47.Peris A., Bonizzoli M., Iozzelli D., et al. Early intra-intensive care unit psychological intervention promotes recovery from post traumatic stress disorders, anxiety and depression symptoms in critically ill patients. Crit Care Lond Engl. 2011;15(1):R41. doi: 10.1186/cc10003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 48.Veer I.M., Riepenhausen A., Zerban M., et al. Psycho-social factors associated with mental resilience in the Corona lockdown. Transl Psychiatry. 2021;11(1):67. doi: 10.1038/s41398-020-01150-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 49.Jain S., Murphy T.E., O’Leary J.R., Leo-Summers L., Ferrante L.E. Association between socioeconomic disadvantage and decline in function, cognition, and mental health after critical illness among older adults: a cohort study. Ann Intern Med. 2022;175(5):644–655. doi: 10.7326/M21-3086. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 50.Barrett M.L., Smith M.W., Elixhauser A., Honigman L.S., Pines J.M. Healthcare Cost and Utilization Project, Utilization of intensive care services, 2011. HCUP Statistical Brief #185. December 2014 http://www.hcup-us.ahrq.gov/reports/statbriefs/sb185-Hospital-Intensive-Care-Units-2011.pdf [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.