Perceived Social Support among Acute Respiratory Failure Survivors in a Multicenter Prospective Cohort Study

To the Editor:

Social isolation is associated with an increased risk of mortality comparable to smoking (1). The term “social support” refers to real or perceived resources—informational, emotional, and tangible—provided via relationships that promote adaptive responses to life stressors (2, 3). Nationally representative cohort studies have shown that being able to identify a friend or relative who could help with personal care is protective against prolonged nursing home stays for adults >65 (4), and social isolation before an intensive care unit admission is associated with increased disability burden and mortality in the year following critical illness (5). Despite being widely recognized as a risk factor for poor health outcomes, there is little quantitative research on social isolation and perceived social support following critical illness. Herein, we present data on perceived social support as part of the APICS-01 (Addressing Post-Intensive Care Syndrome) prospective, multicenter, cohort study (6, 7).

Methods

APICS-01 enrolled acute respiratory failure (ARF) survivors who were discharged directly to home from six hospitals at five U.S. academic medical centers between January 2019, and August 2020. ARF was defined as ⩾24 consecutive hours of 1) mechanical ventilation via endotracheal tube; 2) noninvasive ventilation; or 3) high flow nasal cannula with FIO₂ ⩾50% and flow rate ⩾30 L/min. Exclusion criteria included respiratory support provided for airway protection, preexisting dementia, and barriers to telephone-based follow-up. Informed consent was obtained directly from participants.

Perceived social support was measured using the 12-item, theoretically driven Multidimensional Scale of Perceived Social Support (MSPSS) (8). The MSPSS is validated in a range of populations, including healthy older adults, people with heart failure, and stroke survivors (8). In addition to a total score (range 12–84; higher scores indicate greater support), three subscale scores (range 4–28) can be calculated for perceived support from each of family, friends, and significant other. The MSPSS was administered at hospital discharge and during 3- and 6-month telephone-based assessments.

Baseline characteristics were summarized with descriptive statistics. Differences in the characteristics of participants who completed the 6-month social support assessment versus survivors who missed 6-month assessments were described using the standardized difference statistic (9). MSPSS scores across follow-up were visualized using a spaghetti plot. The Wilcoxon rank sum test was used to test for differences in MSPSS scores between subpopulations, and the Wilcoxon signed-rank test (for paired comparisons) for comparing MSPSS scores in the same participants at different assessment times.

Inverse probability-of-censoring weights were used to estimate perceived social support for all patients who were alive at 6 months (10). Mean differences and Kolmogorov-Smirnov statistics were reviewed to assess covariate balance after weighting. Finally, baseline characteristics were summarized stratified by trajectory of perceived social support during follow-up. The study was approved by a centralized Institutional Review Board at Vanderbilt University with additional oversight by the Human Research Protections Office of the Department of Defense.

Results

Two-hundred ARF survivors were enrolled. At 6 months, 129 (65%) completed the MSPSS, 45 (23%) were successfully contacted for follow-up but did not complete the MSPSS (a “missed assessment”), 17 (9%) had died, and 9 (5%) were lost to follow-up. Among the 154 (77%) completing 3-month MSPSS assessment, there was a slight decrease in the median total score compared with discharge (70 [interquartile range (IQR) 60–74] versus 72 [IQR 62–82], P = 0.001). For the 129 completing the MSPSS at 6 months, median scores were, again, slightly lower than at hospital discharge (72 [IQR 60–78] vs 73 [IQR 64–83], P = 0.005) (Figure 1). After weighting, the estimated median MSPSS score at 6 months for all survivors was substantially lower than at hospital discharge (56 [IQR 40–84] vs 73 [IQR 64–83], P < 0.001). Participants with missed MSPSS assessments at 6 months were more likely to be White, and less likely to test positive for the novel coronavirus SARS-CoV2 (COVID-19) during hospitalization (Table 1). For survivors who completed versus not completed the 6-month MSPSS assessment, the median score at hospital discharge was higher (73, [IQR 64–83] versus 69 [IQR 54–75], P = 0.002). MSPSS scores at discharge for patients with versus without COVID-19 were similar (72 [IQR 58–81] versus 72 [IQR 64–77], P = 0.70). Survivors with relatively stable MSPSS scores during follow-up were younger and more likely to be female (Table 2).

Figure 1.

Perceived Social Support following critical illness. Unadjusted Multidimensional Scale of Perceived Social Support (MSPSS) scores at enrollment, 3- and 6-month follow-up assessments for the 129 survivors of acute respiratory failure who completed the 6-month follow-up assessment. MSPSS scores range from 12–84 with higher scores indicating greater perceived social support. Black dots represent the median of observed scores at each time-point, and black lines represent the interquartile range.

Table 1.

Participant demographics

Demographics	Completed 6 mo Assessment (N = 129)	Missed Assessment or Lost to Follow-up (N = 54)	Standardized Difference*
At enrollment†
Age, yr, median (IQR)	53 (43–64)	55 (37–66)	0.007
Female, n (%)	70 (54%)	26 (48%)	0.04
Race, n (%)	—	—	0.32
Black	36 (28%)	10 (19%)	—
Non-Hispanic White	79 (61%)	39 (72%)	—
Other/multiple	14 (11%)	5 (9%)	—
Resided at home before admission, n (%)	127 (98%)	53 (98%)	0.35
Tested positive for COVID-19 during admission, n (%)	26 (20%)	6 (11%)	0.33
Multidimensional Scale Perceived Social Support at hospital discharge, median (IQR)	73 (64–83)	69 (52–75)	0.47
Significant other subscale	26 (22–28)	24 (9–28)	0.35
Family subscale	26 (24–28)	27 (22–28)	0.16
Friends subscale	24 (22–28)	23 (16–28)	0.43
6-month outcomes
Multidimensional scale perceived social support, median (IQR)	72 (60–78)	—	—
Significant other subscale	24 (18–28)	—	—
Family subscale	24 (23–27)	—	—
Friends subscale	24 (21–26)	—	—

Definition of abbreviations: COVID-19 = coronavirus disease; IQR = interquartile range; MSPSS = Multidimensional Scale of Perceived Social Support.

^*The difference in means or proportions divided by standard error. It is indifferent to sample size. Absolute value ranges from 0.0–1.0.

^† Inverse probability of censoring weights were estimated via a logistic regression model for loss to follow-up that included all demographic data available at enrollment: age, sex, race, residing at home before critical illness, COVID-19 status, and baseline MSPSS score.

Table 2.

Participant demographics by trajectory of perceived social support following critical illness

Attribute	Stable*	Improving*	Declining*	Volatile*
	(N = 42)	(N = 23)	(N = 24)	(N = 23)
Age, years, median (IQR)	50 (42–60)	55 (48–62)	53 (50–65)	64 (49–67)
Female, n (%)	27 (64%)	12 (52%)	9 (38%)	14 (61%)
Race, n (%)
Black	10 (24%)	5 (22%)	10 (42%)	7 (30%)
Non-Hispanic White	25 (60%)	16 (70%)	13 (54%)	14 (61%)
Other/multiple	7 (17%)	2 (9%)	1 (4%)	2 (9%)
Resided at home before admission, n (%)	41 (98%)	23 (100%)	23 (96%)	23 (100%)
Tested positive for COVID-19 during admission, n (%)	10 (24%)	5 (22%)	4 (17%)	4 (17%)
Multidimensional Scale Perceived Social Support, median (IQR)	73 (68–79)	72 (57–79)	84 (71–84)	72 (55–82)
Significant other	26 (24–28)	25 (13–28)	28 (23–28)	26 (18–28)
Family	26 (24–28)	25 (24–27)	28 (28–28)	25 (24–28)
Friends	24 (22–26)	24 (23–26)	28 (27–28)	24 (19–28)

Definition of abbreviations: COVID-19 = coronavirus disease; IQR = interquartile range; MSPSS = Multidimensional Scale of Perceived Social Support.

^*Participants’ total MSPSS scores at 3- and 6-month assessments were compared to their score at the previous assessment, with 3-month scores compared to hospital discharge. Scores were only treated as higher or lower than the previous assessment if they changed by >1 standard deviation from the patient’s score at the previous assessment. This created 9 groups of survivors, all but one of which contained ⩽15 people. For example, there were 8 people whose total MSPSS score was higher at the 3-month assessment, and effectively the same at the 6-month assessment, termed: “higher-same”. These groups were then categorized as follows: Stable = same–same (N = 42); Improving = same–higher (15) + higher–same (8) +  higher–higher (0); Declining = same–lower (8) + lower–same (15) + lower–lower (1); Volatile = higher–lower (15) + lower–higher (8).

Discussion

To our knowledge, this is the first quantitative assessment of perceived social support in ARF survivors. Although observed decreases in MSPSS scores during follow-up were statistically significant, the minimal clinically important difference for MSPSS is unknown, and observed declines may not have been clinically meaningful. The trajectory of perceived social support we observed is consistent with systematic reviews of stroke survivors for whom perceived social support remains relatively stable, even as measures of social isolation, such as social network size, number of friends, and the frequency of involvement in social activities, generally decrease (11).

Among APICS-01 survivors, 95% were assessed at 6-month follow-up, but 23% did not contribute MSPSS data, possibly due to participant time limitations since the MSPSS assessment occurred at the end of the survey battery. In this study, lower perceived social support at hospital discharge was associated with an increased risk of missed MSPSS at 6-month follow-up. Further work is needed to understand this association, but one hypothesis is that some survivors who perceived themselves as lacking social support wished to avoid repeatedly speaking critically about their family, friends, and significant other. If future studies demonstrate that perceived social support is associated with both missed assessments and study outcomes (i.e., informative censoring), this issue could introduce bias.

Finally, despite alarming rates of loneliness reported by American adults early in the COVID-19 pandemic (12), participants who survived COVID-19 did not report lower perceived social support at discharge than participants who were enrolled before the pandemic or who tested negative during the pandemic. It may be that loneliness, defined as the subjective feeling of being lonely, was not strongly correlated with perceived social support (3). Alternatively, COVID-19 may not have effected loneliness in this study population.

Limitations of this study include missed MSPSS assessments and few patients with COVID-19. While it is premature to make clinical recommendations, our findings suggest a mechanism by which professionally moderated support groups could provide benefit following critical illness. We encourage future studies to expand on our findings by studying related concepts like social isolation, loneliness, and family functioning after ARF. Given the robust association between social support and life satisfaction in other populations, exploring the potential role of social connection is an important new frontier for ARF survivorship research.