The Effect of Social Isolation on 1-Year Outcomes After Surgical Repair of Low Energy Hip Fracture

Abstract

OBJECTIVES:

To evaluate if social isolation or loneliness is associated with outcomes 1 year after low energy hip fracture.

METHODS:

Design:

Prospective inception cohort study.

Setting:

Academic Level I Trauma Center.

Patient Selection Criteria:

Participants were ≥ 65 years of age and enrolled 2–4 days after surgery for a first low energy hip fracture. Exclusion criteria were bilateral or periprosthetic hip fracture, previous hip fracture, non-English speaking, international address, active cancer, stage 4 cancer in the past 5 years, radiation to the hip region and cognitive impairment. Participants were followed longitudinally for one year.

Outcome Measures and Comparisons:

The PROMIS-29 was elicited 2–4 days post-operatively and 1 year later. Patient reported risk factors included the Lubben Social Networks Scale and the UCLA Loneliness Scale, which were compared to the Lower Extremity Activity Scale and PROMIS-29 domains.

RESULTS:

Three hundred and twenty-five patients were enrolled. Participants had a median age of 81.7 years, were 70.9 % female and were 85.9% white. 31.6% of patients were socially isolated at time of fracture. At 1 year, 222 of the 291 subjects who were confirmed alive at one year provided data. Multivariable linear models were performed separately for each outcome, including Lower Extremity Activity Scale and PROMIS-29 domains. Controlling for age, sex, education, and body mass index, those who were socially isolated at time of fracture had worse PROMIS-29 function (β= −3.83 p= 0.02) and ability to participate in social roles (β= −4.17 p= 0.01) at 1 year. Secondary analyses found that pre-fracture loneliness was associated with clinically meaningfully worse function, anxiety, depression, fatigue, sleep, pain and ability to participate in social roles at 1 year, (all p< 0.01).

CONCLUSIONS:

Pre-fracture social isolation was associated with worse outcomes 1 year after surgical repair of low energy hip fracture. This data suggests loneliness may be more strongly associated with important patient centric metrics than pre-fracture social isolation. Given the dearth of modifiable risk factors in this population, future studies are needed to evaluate whether improving social connections could impact outcomes in this rapidly growing demographic.

INTRODUCTION

Hip fractures are common events, with over 318,335 annually in the U.S.¹ Hip fractures are harbingers of increased mortality and can lead to significant decrements in quality of life: up to one-half of survivors are unable to live independently, and only 39% return to pre-injury functional status.^2–4 These limitations can be devastating for both patients and caregivers. Most known risks for poor outcomes, (e.g. male gender, age, and cognitive impairment) are non-modifiable.⁴ Social isolation, how integrated a person is into their community, and loneliness, the feeling of being alone regardless of the amount of social contact, are related but distinct constructs which are both associated with poor quality of life.^5,6 Importantly, both are modifiable.

The goal of this study was to evaluate the association of pre-fracture social isolation with patient reported outcomes one year after surgical repair of first low energy hip fracture. A secondary goal was to explore association of pre-fracture loneliness with patient reported outcomes one year after surgical repair of first low energy hip fracture.

METHODS

Between 01/22/2016 and 05/9/2019, patients undergoing hip fracture surgery (OTA/AO 31) were identified from operating room schedules at a Level 1 trauma center with approval obtained from the Institutional Review Board at Hospital for Special Surgery. Subjects ≥ 65 years with a first low energy fracture, defined as falling from standing height or less, were recruited 2–4 days post-surgery and followed prospectively for one year. The mechanism of fracture was ascertained via chart review. Consent was obtained in hospital before 5PM to avoid the “sundowning” effect.

Exclusion criteria were bilateral or periprosthetic hip fracture, previous hip fracture, (to exclude those who may have changed behaviors or living situation due to previous fracture), non-English speaking, international address, active cancer, stage 4 cancer in the past 5 years, radiation to the hip region and cognitive impairment. Cognitive impairment was defined as “dementia” or “Alzheimer’s Disease” in the electronic medical record (EMR) or ≤ 3 on the “6-Item Cognitive Screen,” a validated instrument developed to identify cognitive impairment among potential clinical research participants.⁷ The cognitive screen was administered at time of consent.

Initially, subjects were often excluded for not knowing the day of the week on the 6-Item Cognitive Screen. On review, this metric seemed inappropriate for elderly surgical patients who had been hospitalized for 2–4 days. Therefore, the definition of cognitive impairment was broadened to include subjects who scored a 2 or 3 on the 6-Item Cognitive Screen if they also scored ≥10 on the “University of California, San Diego Brief Assessment of Capacity to Consent” (UBACC). The UBACC is a detailed, time intensive tool developed to evaluate individuals’ capacity to consent to research; scoring ≥10 indicates an ability to understand the consenting process.⁸

Age, sex, body mass index (BMI kg/m2), and insurance status were recorded from the EMR. Participants self-reported education, living situation, race and ethnicity was collected at time of enrollment. Baseline patient reported outcome measures (PROMs) were elicited in-hospital 2–4 days post-operatively after enrollment; participants were specifically asked to reference the week prior to fracture. Retrospective measurement of health-related PROMs within 15 days of orthopedic surgery has been shown to be valid and reliable.⁹

PROMs include the Lower Extremity Activity Scale (LEAS),¹⁰ a validated 18-item measure of function, and PROMIS-29,¹¹ a validated instrument measuring physical function, fatigue, anxiety, depression, sleep, ability to participate in social roles, pain interference and pain intensity. PROMIS-29 pain intensity is measured on a visual analogue scale of 0–10. All other PROMIS-29 domains are measured as T-scores, with 50 being the population mean, and higher scores signifying more of the domain being measured (i.e. higher depression scores mean worse depression, and higher physical function scores mean better physical function).

Participants were called at 30 days, 3 months and 1 year after enrollment to elicit follow up data. At 1 year, questionnaires were completed via phone or mailed paper surveys. Participants were confirmed alive if contacted by study staff, confirmed alive by a family member or caregiver, or if the participant attended a medical appointment one week prior or any time after their scheduled 1-year call. Participants were confirmed deceased if death was documented in the EMR, confirmed by a family member or caregiver, or an obituary obtained. If none of these criteria were met, the participant’s 1-year status was “unknown.” The exposures in this study were the independent variables which were hypothesized to be associated with outcomes at one year. The primary exposure was social isolation measured using the Lubben Social Networks Scale-18 (LSNS-18), a validated 18-item self-report instrument designed to measure social isolation in the elderly. Based on the literature, an LSNS-18 score < 36 has been used to identify social isolation.^12,13

Loneliness was measured using the UCLA 3-item Loneliness Scale. This scale has a range of 3–9, with higher scores indicating more loneliness. Based on the literature, scores of 6–9 identify those who are “lonely.”¹⁴ Loneliness is a different construct than social isolation, and reflects the difference between desired social interaction and actual relationships. The Loneliness Scale was incorporated midway through the study and only administered to 65% of participants; thus, it is considered a secondary exposure.

Primary outcomes were the PROMIS-29 domains of pain, function, and ability to participate in social roles at 1 year. PROMIS-29 T-score differences of 5-points [0.5 standard deviation] are considered clinically meaningful.¹⁵ The clinically meaningful difference of VAS scores for pain, such as the PROMIS-29 pain intensity scores, is usually considered 3-points.¹⁶ Secondary outcomes were other PROMIS-29 domains and LEAS at 1 year. Higher scores of the LEAS denote better function. Descriptive analyses of baseline characteristics were stratified by social isolation, and differences evaluated using chi-square, Fisher’s exact or T-tests as appropriate. Since all outcome measures were continuous, linear models were run to produce estimates and 95% confidence intervals. Each model was adjusted for age, sex, education, and BMI. Lack of variability among racial and ethnic groups precluded statistical adjustments for confounding.

As an exploratory analysis, Spearman Rank correlations were performed to test the crude association between pre-fracture loneliness and each of the 1-year outcomes. Similarly, correlation analyses were performed using the continuous measure of social isolation with 1-year outcomes to determine differences between these associations. Based on these results, multivariable analysis was conducted on subjects with loneliness scores.

RESULTS

Patients:

877 patients undergoing surgery for low energy hip fractures were identified; 462 met inclusion criteria and 325 consented to participate (See Figure 1). Prior to discharge, 4 participants withdrew consent and 1 died; therefore 320 subjects were followed longitudinally. Because there was less than 5% of missing data, results are presented without any imputation. Median age was 81.7 years, 85.9% were white, 70.9% female, 65.9% were college graduates and 48.4% lived alone. Almost all had Medicare, 62.2% had additional commercial insurance and only 6.6% had Medicaid (Table 1). The median American Association of Anesthesia Score was 3, indicating most subjects had substantive functional limitations.¹⁷

Figure 1.

Selection of subjects who had low energy hip fracture for the study

Table 1.

Baseline characteristics and patient reported outcomes stratified by pre-fracture social isolation

		Overall (N=320)	Pre-Fracture Not Socially Isolated (N=219)	Pre-Fracture Socially Isolated (N=101)	p-value*
Demographics
Age (years) Median [IQR]		81.7 [75.0–87.2]	81.1 (74.9–86.3)	83.5 (75.2–88.4)	0.07
White N (%)		275 (85.9%)	187 (85.4%)	88 (87.1%)	0.67
Hispanic N (%)		8 (2.5%)	8 (3.7%)	0	0.06
Female N (%)		227 (70.9%)	156 (71.2%)	71 (70.3%)	0.89
Body Mass Index (BMI kg/m2) N (%)					0.68
	Normal (<25)	55 (17.2%)	34 (15.5%)	21 (20.8%)
	Overweight (25 ≤30)	169 (52.8%)	117 (53.4%)	52 (51.5%)
	Obese (30.1 ≤40)	64 (20.0%)	46 (21.0%)	18 (17.8%)
	Super obese (>40)	24 (2.5%)	17 (7.8%)	7 (6.9%)
Living Situation N (%)					0.06
	Live Alone	155 (48.4%)	95 (43.3%)	60 (59.4%)
	With others/ immediate family/partner or spouse	162 (51.0%)	122 (55.7%)	40 (39.6%)
	Retirement or nursing home	2 (0.6%)	1 (0.5%)	1 (0.5%)
Insurance N (%)
	Medicare	306 (95.6%)	207 (94.5%)	99 (98.0%)	0.24
	Commercial	199 (62.2%)	143 (65.3%)	56 (55.4%)	0.05
	Medicaid	21 (6.6%)	10 (4.6%)	11 (10.9%)	0.11
Education Level N (%)					0.09
	Some college or less	108 (33.8%)	72 (32.9%)	36 (35.6%)
	College graduate or above	211 (65.9%)	146 (66.7%)	65 (64.4%)
Lubben Social Networks Scale (LSNS)a Median [IQR]		44.0 [32.0–53.5]	50.0 [43.0–57.0]	28.0 [23.0–31.0]	<0.001
American Society of Anesthesia Physical Status (Risk) Score Median [IQR]		3.0 [2.0–3.0]	3.0 [2.0–3.0]	3.0 [2.0–3.0]	0.03
Lonelyb N (%)f		45/208 (21.6%)	23/157 (14.6%)	22/51 (43.1%)	<0.001
Patient Reported Outcomes (PROMs)
Lower Extremity Activity Score (LEAS)c Median [IQR)		9.0 [7.0–11.0]	9.0 [8.0–12.0]	9.0 [7.0–10.0]	0.002
PROMIS Physical Functiond Median [IQR)		45.3 [37.9–56.9]	48.0 [40.4–56.9]	41.8 [35.6–56.9]	0.001
PROMIS Anxiety Median [IQR)		40.3 [40.3–55.8]	40.3 [40.3–53.7]	48.0 [40.3–55.8]	0.045
PROMIS Depression Median [IQR)		41.0 [41.0–51.8]	41.0 [41.0–51.8]	41.0 [41.0–55.7]	0.10
PROMIS Fatigue Median [IQR)		43.1 [33.7–53.1]	43.1 [33.7–53.1]	46.0 [33.7–57.0]	0.12
PROMIS Sleep Disturbance Median [IQR)		48.4 [41.1–52.4]	46.2 [41.1–52.4]	48.4 [43.8–54.3]	0.01
PROMIS Ability to Participate in Social Roles Median [IQR)		64.1 [51.6–64.1]	64.1 [55.6–64.1]	55.6 [44.6–64.1]	<0.001
PROMIS Pain Interference Median [IQR)		41.6 [41.6–53.9]	41.6 [41.6–49.6]	41.6 [41.6–55.6]	0.04
PROMIS Pain Intensitye Median [IQR)		0.0 [0.0–4.0]	0.0 [0.0–4.0]	1.0 [0.0–5.0]	0.03

^aSocially Isolated = Lubben Social Networks Scale (LSNS) score <36

^b3 Item Loneliness Scale-Range 3–9. Lonely = 6–9.

^cLEAS: Lower Extremity Activity Scale-Range 0–18. Higher = better function

^dPatient Reported Outcomes Measurement Information System (PROMIS29): T-scores, 50 = population mean. Higher = more of the domain being

^eVisual analogue scale 0–10. Higher = worse pain

^f208 participants evaluated for loneliness

^*bold p-value = p < 0.05

Baseline

Characteristics:

31.6% were socially isolated pre-fracture and 21.6% were lonely. While loneliness and social isolation were strongly associated, (${X^2}_1$ = 18.4 p< 0.0001), they were not synonymous: 14.6% of those who were not socially isolated were lonely, and 56.9% of those who were socially isolated were not lonely. Those who were socially isolated were more likely to be living alone (59.4% vs. 43.3%; p= 0.06) and more likely to be lonely (43.1% vs 14.6%; p <0.001). Before hip fracture, most participants had little pain (PROMIS-29 Pain Intensity median score:0 [IQR 0–4] and pain interfered with their activities less than the mean of the US population, (PROMIS-29 Pain Interference T-score 41.6, US mean= 50).

Outcomes:

At one year, participants who were socially isolated were more likely to have clinically meaningful worse PROMIS-29 physical function (41.8, [IQR:35.6, 56.9] vs. 48.0, [IQR:40.4, 56.9], worse ability to participate in social roles (55.6, [IQR:44.6, 64.1] vs. 64.1, [IQR:55.6, 64.1] and more anxiety, (48.0, [IQR:40.3, 55.8] vs. 40.3, [IQR:40.3, 53.7], (a change ≥ 5 is considered clinically meaningful;¹⁵ See Table 1).

In multivariable models, adjusted for age, sex, education and BMI, those who were socially isolated pre-fracture had statistically significantly worse LEAS (β= −0.88 p= 0.008), PROMIS-29 physical function (β= −3.60 p= 0.002), PROMIS-29 ability to participate in social roles (β= −4.67 p= <0.001), PROMIS-29 sleep disturbance (β= 3.21 p= 0.01) and PROMIS-29 depression (β= 2.32 p= 0.044), though these differences were not clinically meaningful (See Table 2). At 1 year, 90.9% (291/320) of participants were confirmed alive; 20 had died and 9 had an unknown mortality status. Of those confirmed alive, 76.3% (222/291) provided 1-year data. Of the 69 participants who were confirmed alive and did not provide data: 26 withdrew, 36 did not respond after 3 reminders, and 7 were too cognitively impaired to participate. Those who provided 1-year data had higher pre-fracture LSNS-18 scores (i.e. were less socially isolated) than those who did not, (see Table, Supplemental Digital Content 1).

Table 2.

Adjusted association of pre-fracture social isolation^a with pre-fracture and one-year scores^b

	Estimate (β)c	95% Confidence Limits	p-value*
Pre-Fracture Scores
Lower Extremity (LEAS)d	−0.88	−1.54 to −0.23	0.01
PROMIS Physical Functione	−3.60	−5.91 to −1.30	0.002
PROMIS Depression	2.32	0.07 to 4.56	0.044
PROMIS Anxiety	2.26	−0.11 to 4.64	0.06
PROMIS Fatigue	2.37	−0.47 to 5.20	0.10
PROMIS Pain Interference	1.67	−0.72 to 4.07	0.17
PROMIS Sleep Disturbance	3.21	0.73 to 5.69	0.01
PROMIS Ability to Participate in Social Roles	−4.67	−7.22 to −2.13	<0.001
One-Year Scores
Lower Extremity (LEAS)	−0.68	−1.70 to 0.33	0.19
PROMIS Physical Function	−3.83	−7.03 to −0.63	0.02
PROMIS Depression	1.04	−1.55 to 3.63	0.43
PROMIS Anxiety	2.58	−0.55 to 5.71	0.11
PROMIS Fatigue	2.94	−0.71 to 6.59	0.11
PROMIS Pain Interference	1.43	−1.83 to 4.69	0.39
PROMIS Sleep Disturbance	2.73	0.10 to 5.36	0.04
PROMIS Ability to Participate in Social Roles	−4.17	−7.37 to −0.98	0.01

^aSocially Isolated = Lubben Social Networks Scale (LSNS) score <36

^bAll models adjusted for age, sex, education and BMI

^cβ Estimate is the difference in outcome score between 0 (not socially isolated) to 1 (socially isolated)

^dLEAS: Lower Extremity Activity Scale-Range 0–18. Higher = better function

^ePatient Reported Outcomes Measurement Information System (PROMIS29): T-scores, 50 = population mean. Higher = more of the domain being measured

^*bold p-value = p < 0.05

At 1 year, those who were socially isolated pre-fracture were more likely to have worse LEAS (p= 0.04), PROMIS-29 physical function (p= 0.02), and PROMIS-29 ability to participate in social roles (p= 0.01), (see Table 3). In multivariable modes, adjusted for age, sex, education and BMI, those who were socially isolated pre-fracture had statistically significantly worse PROMIS-29 physical function (β= −3.83 p= 0.02), sleep disturbance (β= 2.73 p= 0.04), and ability to participate in social roles (β= −4.17 p= 0.01; see Table 2).

Table 3.

One-year outcomes stratified by pre-fracture social isolation^a

	Overall Median [IQR] (N=222)	Not Socially Isolated Median [IQR] (N=164)	Socially Isolated Median [IQR] (N=58)	p-value*
One-Year Scores
Lower Extremity (LEAS)	9.0 [7.0–11.0]	9.0 [7.0–12.5]	8.0 [6.0–9.0]	0.04
PROMIS Physical Function	41.8 [35.6–48.0]	43.4 [35.6–56.9]	39.1 [32.1–45.3]	0.02
PROMIS Depression	41.0 [41.0–51.8]	41.0 [41.0–51.8]	41.0 [41.0–51.8]	0.78
PROMIS Anxiety	40.3 [40.3–55.8]	40.3 [40.3–53.7]	44.2 [40.3–55.8]	0.24
PROMIS Fatigue	48.6 [39.7–55.1]	48.6 [36.7–54.1]	48.6 [39.7–57.0]	0.20
PROMIS Pain Interference	41.6 [41.6–57.1]	41.6 [41.6–55.6]	41.6 [41.6–59.9]	0.90
PROMIS Sleep Disturbance	48.4 [43.8–54.3]	46.2 [43.8–52.4]	50.5 [43.8–56.1]	0.08
PROMIS Pain Intensity (0–10 score)	2.0 [0.0–5.0]	2.0 [0.0–5.0]	2.0 [0.0–5.0]	1.0
PROMIS Ability to Participate in Social Roles	64.1 [46.4–64.1]	64.1 [48.1–64.1]	55.6 [43.2–64.1]	0.01

^aSocially Isolated = Lubben Social Networks Scale (LSNS) score <36

^bLEAS: Lower Extremity Activity Scale-Range 0–18. Higher = better function

^cPatient Reported Outcomes Measurement Information System (PROMIS29): T-scores, 50 = population mean. Higher = more of the domain being measured

^dVisual analogue scale 0–10. Higher = worse pain

^*bold p-value = p < 0.05

Correlation analyses show that both more social isolation and more loneliness pre-fracture were statistically significantly correlated with worse LEAS (p= 0.02 and p<0.001, respectively) and worse PROMIS-29 physical function (p= 0.04 and p= 0.004, respectively), anxiety (p= 0.004 and p<0.001, respectively), fatigue (p= 0.01 and p=0.02, respectively), and ability to participate in social roles at 1 year (p<0.001 and p=0.01, respectively). In addition, pre-fracture loneliness scores were also significantly correlated with worse PROMIS-29 depression (p<0.001), pain interference (p= 0.02), sleep disturbance (p<0.001), and pain intensity scores at 1 year (p= 0.01) (See Table 4). In multivariable analysis, after adjustment for age, sex, education and BMI, those who were lonely pre-fracture showed statistically significantly and clinically meaningfully worse pain interference (β= 9.11; p <0.001), physical function (β= −9.26; p <0.001), ability to participate in social roles, (β= −8.78; p <0.001), depression (β= 7.50; p <0.001), anxiety (β= 8.38; p <0.001), fatigue (β= 7.14; p= 0.007), sleep disturbance (β= 5.87; p= 0.001), and LEAS function, (β= −2.62; p <0.001), (see Table, Supplemental Digital Content 2).

Table 4.

Spearman rank correlation analysis of pre-fracture Lubben Social Networks Scale score^a and pre-fracture loneliness score^b with one-year outcomes

One-year Outcomes	Pre-Fracture LSNS Score (Higher score= LESS isolated) N = 222		Pre-Fracture Loneliness Score (Higher score = MORE loneliness) N = 148
	r f	p-value*	r	p-value
Lower Extremity (LEAS)c	0.16	0.02	−0.29	<0.001
PROMIS Physical Functiond	0.19	0.04	−0.24	0.004
PROMIS Depression	−0.10	0.13	0.31	<0.001
PROMIS Anxiety	−0.19	0.004	0.40	<0.001
PROMIS Fatigue	−0.18	0.01	0.19	0.02
PROMIS Pain Interference	−0.03	0.62	0.19	0.02
PROMIS Sleep Disturbance	−0.12	0.07	0.33	<0.001
PROMIS Pain Intensity (0–10 score)e	−0.02	0.80	0.21	0.01
PROMIS Ability to Participate in Social Roles	0.25	<0.001	−0.23	0.01

^aSocially Isolated = Lubben Social Networks Scale (LSNS) score <36

^b3 Item Loneliness Scale-Range 3–9. Lonely = 6–9

^cLEAS: Lower Extremity Activity Scale-Range 0–18. Higher = better function

^dPatient Reported Outcomes Measurement Information System (PROMIS29): T-scores, 50 = population mean. Higher = more of the domain being measured

^eVisual analogue scale 0–10. Higher = worse pain

^fr represents the correlation statistic

^*bold p-value = p < 0.05

DISCUSSION

Being socially isolated or lonely prior to surgery for a low energy hip fracture was associated with worse physical function and worse ability to participate in social roles one year later. In addition, loneliness was associated clinically meaningfully worse depression, anxiety, fatigue, sleep and pain 1 year.

Despite improvements in perioperative care, poor outcomes in the elderly after low energy hip fractures are common.^2,18 Given that the U.S population > 65 years will skyrocket to over 98 million by 2060,¹⁹ it will become increasingly important to identify novel modifiable risk factors to improve outcomes among hip fracture survivors. Since 46% of women aged >75, (a group at high risk of hip fracture) live alone, social isolation and loneliness are particularly appealing domains to target in this population.²⁰ Notably, social isolation and loneliness are both potentially modifiable, with several interventions having been shown to be effective, including creative use of the internet, facilitating support groups, physical exercise, social activities, and counseling.^21–24 Other promising interventions include the development of intergenerational partnerships, programs to facilitate volunteering, and non-human companions.^25–28

To date, there has been little research on social isolation or loneliness as predictors of hip fracture outcomes. Among community dwelling patients ≥65, having no social contact with friends during the 2 weeks prior to fracture was associated with a 5x greater risk of death at 2 years compared to those who had daily contact with friends.²⁹ A Dutch study reported that pre-fracture loneliness was associated with a worse recovery and increased anxiety and depression after hip fracture.³⁰ However, in that study 27% of participants had dementia, a strong predictor of poor outcomes. This study showed that amongst relatively high functioning, cognitively intact individuals, pre-fracture loneliness was associated with clinically meaningfully worse anxiety, pain, depression, sleep and physical function 1 year later.

Similar to others studies, median physical function in this cohort had decreased at one year compared to baseline.³¹ Although systematic reviews have shown that interventions can lead to clinically important improvements in mobility after hip fracture, these interventions need to be multipronged (i.e. gait, balance and functional tasks etc.),³² with less comprehensive rehabilitation programs being unlikely to show meaningful benefits. While it is encouraging that hip fracture patients can benefit from focused rehabilitation, such comprehensive programs are expensive, not universally available, and may not be acceptable or generalizable to patients outside of those who qualify for clinical trials. In fact, some multicomponent programs appear no more effective than active control interventions.³³ Social isolation and loneliness may be more feasible targets for intervention, particularly as improving social isolation and loneliness likely has benefits beyond mobility and physical function, (i.e. improving depression, anxiety, and cognitive decline).³⁴ These additional benefits may make hip fracture patients more amenable to participation in programs to improve social connections rather than programs focusing only on physical rehabilitation.

While 25% of community dwellers > 65 years old are socially isolated, 32% of our study participants were socially isolated and 21.6% were lonely.³⁵ An English study surveyed 215 participants who had surgical repair of a unilateral hip fracture, and found that 14% and 15% were social isolated and lonely, respectively.³⁶ That survey was 10–20 years prior to this study, so whether these higher prevalences reflect secular changes, or difference between the US vs. England are unknown. The relatively high prevalence of social isolation in this sample is a little surprising, given that some of the strongest risks for social isolation include being male and low educational attainment, both infrequent in this cohort. Given that very few of these subjects had Medicaid, and most had commercial insurance in addition to Medicare, it is also unlikely that many of these subjects were low income, another known risk factor for social isolation. Whether standard risk factors adequately predict social isolation and loneliness in those at risk for hip fractures is an area of future research.

How social isolation and loneliness lead to poor health outcomes is not clear, though there are several hypothesized mechanisms. These may be instrumental: fewer social contacts may provide less peer support for healthy behaviors such as not smoking, being active, or adhering to prescribed medical regimens.³⁷ Social cognitive theory provides a conceptual framework which explains how social support and interaction can improve self-efficacy and outcome expectations, both of which are strongly related to positive health behaviors.³⁸ Social connections may also act as a buffer against stress, and thus have direct physiologic effects; for example, social isolation has been associated with higher blood pressure and higher levels of inflammatory mediators such as C-reactive protein.^39,40 In addition, while there have been few studies of social isolation in diverse populations, some data suggest social isolation may have a particularly large impact on Black individuals.⁴¹

While it is clear that social isolation and loneliness are different constructs and that both are strongly associated with poor health outcomes such as mortality, cardiovascular disease, depression and anxiety,^37,42–45 results are mixed as to whether loneliness has effects independent of social isolation.⁴⁶ This study found that while both pre-fracture social isolation and loneliness were associated with worse physical function, sleep disturbance, and ability to participate in social roles at 1 year, only pre-fracture loneliness was associated with worse 1-year depression, anxiety, fatigue and pain. Social isolation and loneliness may have different downstream effects on the elderly after major orthopedic trauma. Future studies should prioritize evaluating broad populations and include robust measures of both social isolation and loneliness, to better understand the impact of these domains on the increasingly diverse aging population. The a priori goal of this study, which was achieved, was to assemble a cohort of patients likely to survive to 1 year, who would be representative of patients able to derive benefit from interventions to improve social isolation. Therefore, enrollment was limited to cognitively intact subjects with a first unilateral hip fracture, without a malignancy, who survived at least 2 days after surgery. While this limits generalizability, this cohort reflects the relatively healthy elderly population which will continue to grow over the coming decades. Generalizability may have also been limited in other ways. Low energy hip fractures are unanticipated events which occur during regular life activities, and subjects are often transported to the closest emergency room. This cohort was on average White, highly educated, with commercial insurance, reflecting the demographic mix of our catchment area. Loss to follow up may have impacted generalizability as well, as those without follow up data were more socially isolated at baseline. Based on existing literature, social isolation was evaluated as a dichotomous variable. However, there is evidence that the effect of social isolation is continuous, with a dose-response effect seen along the spectrum of social isolation; larger studies are needed to explore the impact of any change in social isolation on hip fracture surviors.⁴⁷ The UCLA loneliness instrument was added after study inception, and thus loneliness data was only collected on 65% of this cohort.

Strengths include a large, well characterized cohort of cognitively intact participants with excellent follow-up: only 9 participants (2.8%) had an unknown status at 1 year, and 76.3% of those known to be alive provided 1-year self-report data. Pre-fracture information was elicited 2–4 days post-operative, minimizing recall bias. Validated instruments were used for social isolation, loneliness, and other PROMs. Enrollment was completed in 2019, with last follow up in May 2020, making it unlikely these results were meaningfully impacted by the COVID-19 pandemic.

In conclusion, being socially isolated or lonely at time of first low energy hip fracture were associated with worse physical function, sleep, and less social participation in those who survive to 1 year. Exploratory analyses suggest that pre-fracture loneliness, but not social isolation, is also associated with clinically meaningfully worse depression, anxiety, fatigue and pain 1 year later. Given the dearth of modifiable risk factors in elderly hip fracture patients, studies are needed to evaluate whether improving social connections could lead to better outcomes in this rapidly growing demographic.

Supplementary Material

Table, Supplemental Digital Content 1

Table, Supplemental Digital Content 1.doc. Pre-fracture LSNS scores stratified by one-year data obtainment

Table, Supplemental Digital Content 2

Table, Supplemental Digital Content 2.doc. Adjusted association of pre-fracture 3 item loneliness score with one-year patient reported outcomes