Increasing Prevalence of Frailty and its Association with Readmission and Mortality Among Hospitalized Patients with IBD

Abstract

Background:

Although age is often used as a clinical risk stratification tool, recent data has suggested that adverse outcomes are driven by frailty rather than chronological age.

Aims:

In this nationwide cohort study, we assessed the prevalence of frailty, and factors associated with 30-day readmission and mortality among hospitalized IBD patients.

Methods:

Using the Nationwide Readmission Database, we examined all patients with IBD hospitalized from 2010–2014. Based on index admission, we defined IBD and frailty using previously validated ICD codes. We used univariable and multivariable regression to assess risk factors associated with all-cause 30-day readmission and 30-day readmission mortality.

Results:

From 2010–2014, 1,405,529 IBD index admissions were identified, with 152,974 (10.9%) categorized as frail. Over this time period, the prevalence of frailty increased each year from 10.20% (27,594) in 2010 to 11.45% (33,507) in 2014. On multivariable analysis, frailty was an independent predictor of readmission (aRR 1.16, 95% CI: 1.14–1.17), as well as readmission mortality (aRR 1.12, 95% CI 1.02–1.23) after adjusting for relevant clinical factors. Frailty also remained associated with readmission after stratification by IBD subtype, admission characteristics (surgical vs. non-surgical), age (patients ≥ 60 years-old), and when excluding malnutrition, weight loss, and fecal incontinence as frailty indicators. Conversely, we found older age to be associated with a lower risk of readmission.

Conclusions:

Frailty, independent of age, comorbidities, and severity of admission, is associated with a higher risk of readmission and mortality among IBD patients, and is increasing in prevalence. Given frailty is a potentially modifiable risk factor, future studies prospectively assessing frailty within the IBD patient population are needed.

INTRODUCTION

Crohn’s disease (CD) and ulcerative colitis (UC) are immune-mediated diseases of the intestine that together, comprise inflammatory bowel disease (IBD).¹ The prevalence of IBD has continued to increase over time, now reaching an estimated 0.25–0.5% of the United States population.² As a result, older patients with IBD are an expanding subpopulation, driven both by new diagnoses as well as IBD patients living longer.³

Overall, managing older patients with IBD represents a unique challenge given the higher rate of polypharmacy, increased number of comorbidities, and concern over drug-safety. Although data suggests improved outcomes with early use of immunosuppressive therapy in patients with moderate to severe disease, use of biologics in older patients is limited due to safety concerns. Providers are often wary to prescribe such medications because of small studies that have suggested biologic therapy may be associated with increased risk of mortality and severe infections in older IBD patients.^4–6 As a result, older patients with IBD are often placed on inadequate therapy; either 5-ASA agents which lack efficacy in moderate to severe disease, or prolonged courses of corticosteroids which have significant adverse effects.⁷ Additionally, the use of inadequate therapies may in part be the reason why surgical rates are also higher in older IBD patients as compared to younger IBD patients, despite the more indolent course of disease.^{8, 9} This is especially detrimental given that surgery in older IBD patients has been associated with significantly higher postoperative mortality and complication rates.^{10, 11} Thus, it is crucial that we devise an improved risk stratification tool for IBD patients, aside from age, that can be used to guide treatment options within the IBD population.

More recently, data has suggested that adverse outcomes, specifically within older patients, may be driven by frailty rather than chronological age.¹² Frailty, which is defined as a loss of physiologic reserve, has been shown to be associated with adverse outcomes in patients with chronic diseases such as cirrhotics awaiting liver transplantation, chronic obstructive pulmonary disease patients, as well as AIDS patients, independent of chronological age.^13–15

In IBD patients, frailty has also recently been shown to be an independent predictor of infection after immunosuppression use, and associated with readmission and mortality independent of age and comorbidities.^{16, 17} However, given the lack of uniformity among the several frailty instruments, and that the most recent study by Qian et al. assessed frailty using a scale derived from a cohort of adults ≥ 75 years-old, we sought to assess the impact of frailty on readmission and readmission mortality using the Johns Hopkins Adjusted Clinical Groups; an assessment that has been previously validated at the population level and has shown moderate concordance with the Vulnerable Elders Survey.¹⁸

As such, the primary aim of this study was to use the Johns Hopkins Adjusted Clinical Groups frailty indicator to assess the prevalence of frailty over time, and to assess frailty as an independent predictor of readmission and readmission mortality within the hospitalized IBD patient population. Additionally, in order to build upon prior studies, we included five years of nationwide hospitalization data, and performed sensitivity analyses to further bolster our findings.

MATERIALS & METHODS

Data Source

We conducted a retrospective cohort study using weighted data from the Nationwide Readmissions Database (NRD). The NRD is an all-payer database compiled annually by the Healthcare Cost and Utilization Project, containing state-level patient information with the ability to track patients across hospitalizations within a single state during an individual year. As of 2014, the database contained data corresponding to 15 million discharges, including 22 states and 2,048 hospitals.¹⁹ The NRD contains a set of sample weights that, when applied, make the data representative of nationwide discharges. The NRD contains patient information from a variety of care settings, including public, community, and academic hospitals (excluding the VA system). Prior studies have used the NRD to evaluate the relationship between frailty and readmissions outside of IBD.^20–22

Given the de-identified and public availability of the NRD dataset, the Institutional Review Boards of Columbia University Medical Center deemed this work exempt from review. Additionally, per the Healthcare Cost and Utilization Project Data User Agreement, cell sizes smaller than 11 were suppressed and not reported.

Study Population

From 2010–2014, we identified patients in the NRD with International Classification of Diseases, 9^th revision, Clinical Modification (ICD-9-CM) diagnosis codes for UC (556.x) or CD (555.x) during an index (initial) hospital admission and readmission, and classified them as having IBD based upon this.²³ Patients were excluded if they did not survive to discharge on their index admission.

Definition of Frailty

Frailty was determined based on validated ICD-9-CM codes derived from the Johns Hopkins Adjusted Clinical Groups frailty-defining diagnoses which includes diagnoses relating to dementia, vision impairment, presence of decubitus ulcers, fecal incontinence, social support needs, as well as mobility and nutritional issues as described by Goel et al (Appendix).^{21, 24} As in previously published studies, patients with any of the aforementioned ICD-9-CM codes were considered frail.^{16, 21, 25}

Outcomes Measured

The primary outcomes were 30-day all-cause hospital readmission and readmission mortality, as well as the prevalence of frailty among hospitalized IBD patients. The index admission was defined as the first hospitalization a patient experienced in a given year. The 30-day hospital readmission included any rehospitalization within 30 days of the index hospitalization discharge date. Of note, if a patient had >1 readmission within 30 days, only the first index admission was included in our analysis. Secondary outcomes included readmission cost and length of stay.

Patient and Hospital Characteristics

The NRD contains index admission data regarding patient and hospital factors. Demographic patient factors included patient age, sex, payer type, and median income quartile by zip code. Tobacco use, known malignancy, and C. difficile infection on index admission were also noted. Comorbidities were calculated using the Agency for Healthcare Research and Quality comorbidity index, which includes comorbid conditions used in the Elixhauser Comorbidity index.¹⁹ Patients were then categorized based on the number of underlying comorbidities (0, 1, or >1). Operating room procedures performed during the index admission were stratified by type of surgery. Those patients who underwent bowel repair or resection, anorectal manipulation, or fistula surgery were classified into the “intestinal surgery” group based on ICD-9-CM coding (Appendix) and all other surgeries were considered “non-intestinal surgery.”^{26, 27} Patients who underwent flexible sigmoidoscopy or colonoscopy were also identified using ICD-9-CM coding, which was included as a surrogate measure for disease flare (Appendix).²⁷ Patient disposition and care utilization measures including length and cost of hospitalization were obtained directly from the NRD.

Data Analysis

The above demographic, patient, and hospital factors were obtained from the index hospitalization and, using conversion factors provided by the NRD, weighted to reflect national estimates. Univariable and log-linear multivariable regression models with unadjusted (RR) and adjusted (aRR) risk ratios were used to assess the relationship between the primary outcome and clinical factors described above. To capture all potential readmissions within 30 days, only index hospitalization discharges between January 1^st – November 30^th of each calendar year were included in the analysis. The mean admission costs were calculated using the NRD charge-to-cost converter and adjusted for inflation to represent costs in 2016 dollars. Subgroup analyses of UC and CD were also performed. All statistical analyses were completed using SAS version 9.4 (Cary, NC)

Sensitivity Analysis

Sensitivity analyses were also performed to assess for potential confounding. First, all patients with malnutrition, weight loss, or fecal incontinence diagnoses were excluded from the analysis since IBD patients with severe disease often have these diagnoses. Next, we performed an analysis including only patients 60 years of age and older to assess the impact of frailty within this population. Finally, a subgroup analysis was performed, stratifying patients by the presence or absence of surgery during index admission, in order to assess frailty as an independent predictor of hospital readmission among both IBD patient populations.

RESULTS

Baseline Characteristics

Among patients with IBD, a total of 1,405,529 index admissions from 2010 to 2014 were identified, with 36% (510,199) corresponding to patients with UC, and 63% (888,981) corresponding to patients with CD. A total of 152,974 patients (10.9%) met criteria for frailty. In total, 192,181 (13.7%) of the index admissions were associated with readmission within 30 days, with IBD and infection being the most likely causes of readmission. In total, 85% (163,652) of readmissions were from non-frail patients and 15% (28,529) from frail patients. Among the patients who were readmitted, a total of 2,779 (1.5%) died. Overall, the prevalence of frailty among all hospitalized IBD patients increased each year: 10.2% (27,594) in 2010, 10.8% (29,707) in 2011, 10.8% (30,247) in 2012, 11.1% (31,919) in 2013, and 11.5% (33,507) in 2014.

Among frail IBD patients, the majority were over the age of 50, had two or more comorbidities, and had a more severe index admission (Table 1). The most common frailty defining diagnosis was malnutrition (55%) followed by weight loss (20%) and presence of a decubitus ulcer (11%). Most frail patients were discharged home without services (53%), but they were more likely to be discharged to a nursing facility (21% vs 7.4%) or home with services (23% vs 12.4%) compared to non-frail patients. Frail patients also had longer readmission length of stay (9.6 vs 6.5 days) and higher hospitalizations costs ($20,916.11 vs $13,538.98).

Table 1:

Baseline Demographics for 1,405,529 patients with IBD hospitalized between 2010–2014

	Frail N [%]	Not Frail N [%]	p-value
Age (years) at time of index admission
< 18	8,974 [13.83%]	55,926 [86.17%]	<0.01
18–30	21,919 [10.00%]	197,055 [90.00%]
31–40	16,602 [8.00%]	190,783 [82.00%]
41–50	19,223 [9.11%]	191,787 [90.89%]
51–65	35,397 [10.52%]	300,947 [89.48%]
66–80	33,415 [12.81%]	227,507 [87.19%]
>80	17,443 [16.46%]	88,549 [83.54%]
Sex
Male	70,839 [11.66%]	536,443 [88.34%]	<0.01
Female	82,134 [10.29%]	716,112 [89.71%]
IBD
Ulcerative colitis	58,144 [11.40%]	452,055 [88.60%]	<0.01
Crohn’s disease	93,778 [10.55%]	795,203 [89.45%]
Indeterminate Colitis	1,052 [16.57%]	5,297 [83.43%]
Median household Zip code income quartile
Quartile 4 (highest)	35,549 [10.05%]	318,194 [89.95%]	<0.01
Quartile 3	38,014 [10.78%]	314,589 [89.22%]
Quartile 2	37,777 [11.05%]	304,090 [88.95%]
Quartile 1 (lowest)	39,169 [11.64%]	297,427 [88.36%]
Missing	2,465 [11.90%]	18,256 [88.10%]
Insurance status
Private	48,026 [8.50%]	516,786 [91.50%]	<0.01
Medicare	70,607 [13.38%]	457,249 [86.62%]
Medicaid	21,414 [11.64%]	162,621 [88.36%]
Self-pay	6,643 [10.22%]	58,369 [89.78%]
No charge	1,041 [11.09%]	8,345 [88.91%]
Other	4,950 [9.60%]	46,634 [90.40%]
Unknown	292 [10.27%]	2,550 [89.73%]
Tobacco use
No	130,550 [10.91%]	1,066,036 [89.09%]	0.02
Yes	22,423 [10.73%]	186,519 [89.27%]
Malignancy
No	146,041 [10.69%]	1,220,685 [89.31%]	<0.01
Yes	6,933 [17.87%]	31,870 [82.13%]
Clostridium difficile on index admission
No	139,888 [10.38%]	1,208,066 [89.62%]	<0.01
Yes	13,086 [22.73%]	44,489 [77.27%]
Comorbidities
None	15,553 [5.17%]	285,260 [94.83%]	<0.01
One	32,814 [9.22%]	323,178 [90.78%]
≥2	104,606 [13.97%]	644,118 [86.03%]
Depression/Anxiety
No	116,148 [10.61%]	978,818 [89.39%]	<0.01
Yes	36,826 [11.86%]	273,737 [88.14%]
Severity of index admission
No loss of function	5,082 [2.72%]	181,843 [97.28%]	<0.01
Moderate loss of function	30,259 [4.41%]	655,234 [95.59%]
Major loss of function	73,603 [16.97%]	360,134 [83.03%]
Extreme loss of function	43,866 [44.48%]	54,758 [55.52%]
Unknown	165 [21.94%]	587 [78.06%]
Operating room procedure during index hospitalization
None	115,706 [10.88%]	947,684 [89.12%]	<0.01
Intestinal surgery	17,842 [17.19%]	85,959 [82.81]
Non-intestinal surgery	19,426 [8.15%]	218,912 [91.85%]
Hospital teaching status
Metropolitan nonteaching	49,173 [10.32%]	427,279 [89.68%]	<0.01
Metropolitan teaching	92,562 [11.46%]	715,467 [88.54%]
Nonmetropolitan	11,238 [9.28%]	109,810 [90.72%]
Hospital bed size
Small	16,352 [9.55%]	154,929 [90.45%]	<0.01
Medium	35,174 [10.76%]	291,783 [89.24%]
Large	101,447 [11.18%]	805,844 [88.82%]
Flex Sig/Colonoscopy on index admission
No	122,407 [10.13%]	1,086,516 [89.87%]	<0.01
Yes	30,567 [15.55%]	166,039 [84.45%]
Discharge status
Routine Discharge	81,534 [7.70%]	976,958 [92.30%]	<0.01
Transfer to short-term hospital	2,405 [20.28%]	9,454 [79.72%]
Transfer to SNF, ICF, or other†	32,542 [25.96%]	92,797 [74.04%]
Home health care	34,808[18.27%]	155,734 [81.73%]
Against medical advice	1,463 [7.98%]	16,878 [92.02%]
Unknown	221 [23.14%]	734 [76.86%]
Readmission length of stay	9.62 days	6.47 days	<0.01
Readmission hospitalization cost ‡	$20,916.11	$13,538.98	-

^†SNF (Skilled Nursing Facility), ICF (Intermediate Care Facility),

^‡Adjusted for inflation to 2016 dollars

Factors Associated with Readmission and Readmission Mortality

On univariable analysis, we identified factors associated with readmission and readmission mortality (Table 2). When assessing patient demographics, we found that older age was associated with a lower risk of readmission, but a higher risk of readmission mortality. Female sex was also associated with a lower risk of readmission (RR 0.93, 95% CI 0.92–0.93), but not mortality.

Table 2:

Univariable analysis looking at predictors of 30-day readmission and readmission mortality

	Unadjusted 30-day Readmission Risk Ratio & (95% CI)	Unadjusted Readmission Mortality Risk Ratio & (95% CI)
Age (years) at time of index admission
< 18	0.90 (0.88–0.92)	0.16 (0.07–0.40)
18–30	Reference	Reference
31–40	0.96 (0.95–0.98)	0.94 (0.72–1.23)
41–50	0.88 (0.87–0.90)	2.52 (2.02–3.14)
51–65	0.79 (0.78–0.80)	5.53 (4.55–6.72)
66–80	0.75 (0.74–0.76)	10.26 (8.47–12.43)
>80	0.69 (0.68–0.70)	12.46 (10.17–15.26)
Sex
Male	Reference	Reference
Female	0.93 (0.92–0.93)	0.97 (0.90–1.04)
Frail
No	Reference	Reference
Yes	1.43 (1.41–1.45)	1.89 (1.74, 2.06)
Median household Zip code income quartile
Quartile 4 (highest)	Reference	Reference
Quartile 3	1.03 (1.02–1.04)	0.83 (0.74–0.92)
Quartile 2	1.06 (1.05–1.08)	0.96 (0.87–1.07)
Quartile 1 (lowest)	1.10 (1.08–1.11)	0.90 (0.81–0.99)
Missing	1.08 (1.04–1.12)	0.36 (0.22–0.58)
Insurance status
Private	Reference	Reference
Medicare	1.15 (1.14–1.17)	2.40 (2.20–2.61)
Medicaid	1.48 (1.46–1.50)	0.35 (0.29–0.43)
Self-pay	1.11 (1.09–1.14)	0.33 (0.23–0.47)
No charge	1.20 (1.14–1.27)	0.5 (0.24–1.04)
Other	1.04 (1.01–1.06)	0.49 (0.35–0.69)
Unknown	1.12 (1.01–1.24)	-
Tobacco use
No	Reference	Reference
Yes	1.07 (1.05–1.08)	0.64 (0.57–0.72)
Malignancy
No	Reference	Reference
Yes	1.19 (1.16–1.22)	3.79 (3.37–4.26)
Clostridium difficile on index admission
No	Reference	Reference
Yes	1.19 (1.16–1.21)	0.92 (0.77–1.10)
Comorbidities
None	Reference	Reference
One	1.13 (1.12–1.15)	1.69 (1.41–2.02)
≥2	1.25 (1.23–1.26)	4.17 (3.57–4.87)
Depression/Anxiety
No	Reference	Reference
Yes	1.21 (1.20–1.22)	0.85 (0.78–0.93)
Operating room procedure during index hospitalization
None	Reference	Reference
Intestinal surgery	0.88 (0.86–0.90)	0.68 (0.57–0.81)
Non-intestinal surgery	0.64 (0.63–0.65)	1.02 (0.91–1.14)
Hospital teaching status
Metropolitan nonteaching	Reference	Reference
Metropolitan teaching	1.13 (1.12–1.14)	1.14 (0.99–1.31)
Nonmetropolitan	0.89 (0.87–0.90)	1.12 (0.98–1.27)
Hospital bed size
Small	Reference	Reference
Medium	1.08 (1.06–1.1)	1.14 (0.99–1.31)
Large	1.14 (1.12–1.15)	1.12 (0.98–1.27)
Flex Sig/Colonoscopy on index admission
No	Reference	Reference
Yes	0.79 (0.78–0.80)	1.12 (1.00–1.25)
Discharge status after index admission
Routine Discharge	Reference	Reference
Transfer to short-term hospital	1.27 (1.21–1.33)	2.21 (1.64–3.00)
Transfer to SNF, ICF, or other†	1.18 (1.16–1.19)	4.03 (3.69–4.40)
Home health care	1.39 (1.38–1.41)	1.62 (1.47–1.78)
Against medical advice	1.67 (1.61–1.72)	0.54 (0.35–0.83)
Unknown	0.87 (0.72–1.05)	1.28 (0.24–6.73)
Severity of index admission
No loss of function	Reference	Reference
Moderate loss of function	1.02 (1.01–1.04)	2.32 (1.78–3.03)
Major loss of function	1.41 (1.39–1.44)	7.96 (6.15–10.29)
Extreme loss of function	1.68 (1.64–1.71)	11.58 (8.89–15.09)
Unknown	1.97 (1.69–2.29)	3.88 (0.90–16.83)
Readmission from emergency department	-	1.56 (1.41–1.72)

^†SNF (Skilled Nursing Facility), ICF (Intermediate Care Facility)

When assessing clinical factors, frailty was associated with an increased risk of readmission (RR 1.43, 95% CI 1.41–1.45) and readmission mortality (RR 1.89, 95% CI 1.74–2.06). Additional factors also associated with increased risk of readmission and readmission mortality were the presence of comorbidities, malignancy, and increased severity of index admission, with those with extreme loss of function on initial hospitalization at the highest risk of readmission (RR 1.97, 95% CI 1.67–2.29) and readmission death (RR 11.58, 95% CI 8.89–15.09). Of note, patients who had a colonoscopy/flexible sigmoidoscopy (RR 0.79, 95%CI 0.78–0.80) and those who had a surgical procedure on their index admission (RR 0.88, 95%CI 0.86–0.90 for intestinal surgery, and RR 0.64, 95%CI 0.63–0.65 for non-intestinal surgery) were less likely to be readmitted.

When considering patient disposition, those who were discharged to a short-term hospital (RR 1.27, 95% CI 1.21–1.33), skilled nursing facility (RR 1.18, 95% CI 1.16–1.19), or home with health services (RR 1.39, 95% CI 1.38–1.41) had an increased risk of readmission on univariable analysis (Table 2). These patients also had an increased risk of readmission mortality. In addition, readmission via the emergency room was associated with an increased risk of readmission mortality (RR 1.56, 95% CI 1.41–1.72).

On multivariable analysis, many of the factors that remained associated with increased risk of readmission and readmission death on univariable analysis remained independently associated. Notably, older IBD patients still had a decreased risk of readmission (age > 80 aRR 0.43, 95% CI 0.41–0.44; Table 3). Additionally, after adjusting for additional factors such as comorbidities and severity of index admission, frailty remained independently associated with an increased risk of readmission (aRR 1.16, 95% CI 1.14–1.17). When stratified by IBD subtype, frailty remained associated with readmission among both patients with UC (aRR 1.21, 95% CI 1.18–1.24) as well as patients with CD (aRR 1.13, 95% CI 1.11–1.15; Supplemental Table 1). Additionally, after removing malnutrition, weight loss, and fecal incontinence as frailty-defining diagnoses, frailty still remained independently associated with an increased risk of readmission on multivariable analysis (Supplemental Table 2). Frailty also remained an independent predictor of readmission when assessing only patients ≥60 years-old (Supplemental Table 3), and when stratifying by both surgical and non-surgical admissions (Supplemental Table 4).

Table 3:

Multivariable analysis examining frailty as an independent predictor of 30-day readmission

	Composite IBD Diagnosis (n=1,405,529)
Variable	Adjusted† Risk Ratio, (95% CI)	p-value
Age in years
< 18	0.90 (0.87–0.92)	<0.01
18–30	(Ref.)
31–40	0.93 (0.92–0.95)	<0.01
41–50	0.80 (0.79–0.82)	<0.01
51–65	0.66 (0.65–0.67)	<0.01
66–80	0.52 (0.51–0.53)	<0.01
>80	0.43 (0.41–0.44)	<0.01
Sex
Female	0.91 (0.90–0.92)	<0.01
Frailty
Yes	1.16 (1.14–1.17)	<0.01
Median household Zip code income quartile
Highest	(Ref.)
High	0.99 (0.98–1.01)	0.34
Low	1.02 (1.00–1.03)	0.02
Lowest	1.00 (0.99–1.02)	0.58
Missing	1.04 (1.00–1.08)	0.03
Insurance status
Private	(Ref.)
Medicare	1.30 (1.28–1.32)	<0.01
Medicaid	1.32 (1.30–1.34)	<0.01
Self-pay	1.05 (1.03–1.08)	<0.01
No charge	1.13 (1.07–1.19)	<0.01
Other	1.02 (1.00–1.05)	0.10
Tobacco use
Yes	0.96 (0.95–0.97)	<0.01
Malignancy
Yes	1.16 (1.13–1.19)	<0.01
Clostridium difficile
Yes	1.02 (1.00–1.04)	0.07
Comorbidities
None	(Ref.)
One	1.10 (1.09–1.12)	<0.01
≥2	1.17 (1.15–1.18)	<0.01
Depression/Anxiety
Yes	1.14 (1.13–1.16)	<0.01
Operating room procedure during index admission
None	(Ref.)
Intestinal surgery	0.77 (0.76–0.78)	<0.01
Non-intestinal surgery	0.63 (0.62–0.64)	<0.01
Hospital teaching
Metro nonteaching	(Ref.)
Metro teaching	1.10 (1.09–1.12)	<0.01
Nonmetropolitan	0.88 (0.86–0.89)	<0.01
Flex Sig/Colonoscopy
Yes	0.74 (0.73–0.75)	<0.01
Discharge status
Routine Discharge	(Ref.)
Short-term hospital	1.19 (1.14–1.25)	<0.01
SNF, ICF, or other‡	1.25 (1.23–1.27)	<0.01
Home health care	1.41 (1.39–1.43)	<0.01
Against medical advice	1.36 (1.32–1.41)	<0.01
Missing/Unknown	0.82 (0.68–0.99)	0.04
Severity of index admission
No loss of function	(Ref.)
Moderate loss of function	1.02 (1.00–1.04)	0.01
Major loss of function	1.34 (1.32–1.36)	<0.01
Extreme loss of function	1.44 (1.40–1.47)	<0.01
Unknown	1.78 (1.53–2.07)	<0.01

^†Adjusted for all variables listed in the table,

^‡SNF (Skilled Nursing Facility), ICF (Intermediate Care Facility)

When assessing predictors of readmission mortality, frailty was also independently associated with an increased risk (aRR 1.12, 95% CI 1.02–1.23) on multivariable analysis (Table 4). When stratifying by IBD subtype, frailty was a risk factor for readmission mortality among CD patients (aRR 1.32, 95% CI 1.13–1.55), but not for UC patients (aRR 1.09, 95% CI 0.97–1.22; Supplemental Table 5). When compared with a reference age of 18–30 years old, all patients >40 years old also had an increased risk of mortality (aRR 2.03–6.38, 2.03 for patients 41–50 years old and 6.38 for those >80 years old). Additional factors associated with an increased risk of readmission mortality included the presence of two or comorbidities, discharge on initial admission to a skilled nursing facility, readmission from the emergency room, and having a surgical procedure or flexible sigmoidoscopy/colonoscopy on readmission.

Table 4:

Multivariable analysis examining frailty as an independent predictor of 30-day readmission mortality

	Composite IBD Diagnosis (n=192,181)
Variable	Adjusted† Risk Ratio, (95% CI)	p-value
Age in years
< 18	0.16 (1.02–1.23)	0.02
18–30	(Ref.)
31–40	0.90 (0.69–1.17)	0.43
41–50	2.03 (1.62–2.54)	<0.01
51–65	3.44 (2.81–4.22)	<0.01
66–80	5.39 (4.34–6.69)	<0.01
>80	6.36 (5.05–8.02)	<0.01
Sex
Female	0.94 (0.87–1.02)	0.13
Frailty
Yes	1.12 (1.02–1.23)	0.02
Median household Zip code income quartile
Highest	(Ref.)
High	0.93 (0.83–1.03)	0.17
Low	1.15 (1.04–1.28)	<0.01
Lowest	1.22 (1.09–1.36)	<0.01
Missing	-	-
Insurance status
Private	(Ref.)
Medicare	0.80 (0.72–0.89)	<0.01
Medicaid	0.43 (0.36–0.53)	<0.01
Self-pay	0.43 (0.30–0.62)	<0.01
No charge	0.70 (0.33–1.48)	0.36
Other	0.49 (0.35–0.70)	<0.01
Tobacco use
Yes	1.01 (0.89–1.14)	0.88
Malignancy
Yes	2.21 (1.96–2.49)	<0.01
Clostridium difficile
Yes	0.68 (0.57–0.82)	<0.01
Comorbidities
None	(Ref.)
One	1.04 (0.87–1.25)	0.69
≥2	1.26 (1.07–1.49)	<0.01
Admitted from ER
Yes	1.49 (1.35–1.65)	<0.01
Operating room procedure during index admission
None	(Ref.)
Intestinal surgery	0.59 (0.49–0.71)	<0.01
Non-intestinal surgery	0.65 (0.58–0.74)	<0.01
Operation on readmission
Yes	1.92 (1.75–2.11)	<0.01
Hospital teaching
Metro nonteaching	(Ref.)
Metro teaching	1.08 (0.99–1.17)	0.08
Nonmetropolitan	1.05 (0.91–1.22)	0.5
Flex Sig/Colonoscopy
Yes	1.29 (1.15–1.45)	<0.01
Discharge status after index admission
Routine Discharge	(Ref.)
Short-term hospital	1.25 (0.92–1.69)	0.15
SNF, ICF, or other‡	1.39 (1.25–1.53)	<0.01
Home health care	0.82 (0.74–0.91)	<0.01
Against medical advice	0.93 (0.61–1.43)	0.76
Severity of index admission
No loss of function	(Ref.)
Moderate loss of function	1.58 (1.20–2.06)	<0.01
Major loss of function	3.32 (2.55–4.34)	<0.01
Extreme loss of function	4.38 (3.31–5.79)	<0.01
Unknown	1.98 (0.46–8.59)	0.36

^†Adjusted for all variables listed in the table,

^‡SNF (Skilled Nursing Facility), ICF (Intermediate Care Facility)

DISCUSSION

In this study, we found the prevalence of frailty within hospitalized IBD patients to be increasing over time. Frail IBD patients had longer lengths of stay on readmission, higher healthcare related costs, and accounted for up to 15% of IBD readmissions. When assessing predictors of readmission and readmission mortality, frailty was shown to be independently associated with both outcomes, even after adjusting for age, comorbidities, severity of admission, discharge status, and other relevant clinical variables. Additionally, on sensitivity analysis, frailty remained associated with readmission when excluding overlapping IBD diagnoses, when including only patients ≥60 years-old, and when stratifying by the presence or absence of surgery on index admission.

Frailty, a clinical syndrome characterized by a decline in physiologic reserve, has been shown in many patient populations to be associated with increased healthcare utilization and higher costs.²⁸ Within this nationwide study of hospitalized IBD patients, frail patients had a three day longer median length of stay on readmission, and accounted for an additional $7,400 in cost per readmission. This is in accordance with prior studies suggesting that the majority of spending across individuals with Medicare is concentrated among those who are frail.²⁹ As such, the identification of frail patients may be an important way to target interventions aimed at reducing future healthcare utilization and costs.³⁰

This may be particularly important within the IBD patient population, as patients are likely at higher risk for frailty given the chronicity of disease and underlying dysregulation of the immune system.³¹ Overall, we found a 10.9% prevalence of frailty among hospitalized patients with IBD. This is comparable to findings by Kochar et al, who found a 6% prevalence of frailty in a predominantly outpatient IBD patient population.³² Additionally, as the proportion of older individuals with IBD is increasing over time, we similarly found a parallel rise in the prevalence of frailty over the course of the study.³³ This is likely due to the fact that as patients age, they are more likely to become frail, with similar findings seen in prior studies.³²

Although advanced age is a risk factor for frailty, it is important to note that they are not interchangeable. While frailty has been shown to be an independent predictor of readmission in many different patient populations, age has often been found to be a poor predictor.^{34, 35} In 2018, Berry et al. examined over 31 million hospitalizations within the U.S., and found that as patients aged, the likelihood of readmission declined.³⁵ Within our study, we found similar results. More specifically, we found that frailty, rather than age, was independently predictive of a higher likelihood of hospital readmission when adjusting for relevant clinical factors such as comorbidities, admission severity, and discharge status. While this, in part, may be due to the fact that only healthier older individuals are surviving to discharge, it highlights the fact that frailty, rather than age alone, is an important predictor of clinical outcomes.³⁶

Additionally, when assessing predictors of readmission mortality, we similarly found that frailty was independently associated with this outcome, even after adjusting for age, index admission severity, discharge status, having a procedure on readmission, and other relevant clinical factors. It’s important to note that both comorbidity and frailty were also independent predictors of our outcomes, which has been demonstrated in prior studies.^{37, 38} Although often present together, this underscores the notion that they are distinct clinical entities, as frailty represents a loss of physiologic reserve, whereas comorbidity represents an aggregate of clinical disease. This is of particular importance as frailty is one of the few factors that is identifiable prior to hospital admission, and can be potentially modified through intervention.³⁹

Although frailty instruments have never been validated within the IBD patient population, after excluding overlapping IBD diagnoses from the frailty assessment (malnutrition, weight loss, fecal incontinence), similar results were seen. As a result, this study is among the first to further the notion that frailty may be an important risk stratification tool among the IBD patient population. The most accurate assessment of frailty, however, remains through in-person measurements, rather than reliance on diagnostic codes (Supplemental Table 6).^{40, 41}

In addition to the retrospective nature of this study which may introduce unmeasured effects of confounding, the NRD does not contain granular data regarding outpatient or inpatient medication use, laboratory values, or IBD disease severity, which limits the scope of our analysis. Furthermore, given that the NRD tracks index admissions on an annual basis, each patient could have up to one index admission per year, potentially resulting in a single patient representing up to 5 index admissions in our analysis. In order to control for this possibility, we also analyzed each year individually, and found frailty to be independently associated with hospital readmission for each year (Supplemental Table 7).

This study also has significant strengths. It is the largest nationwide study within the IBD patient population to assess frailty as a predictor of readmission and readmission mortality, and captures a sample of 15 million hospitalizations. Additionally, the use of sensitivity analyses further confirms our findings, and emphasizes the clinical importance of frailty within IBD.

As the prevalence of IBD is rising worldwide, improved risk stratification tools are needed to help inform clinical decisions.⁴² Frailty has recently been shown to be an important and independent predictor of adverse outcomes in IBD, and in this study, is also associated with readmission and readmission mortality among hospitalized IBD patients. Furthermore, given the rising prevalence of frailty, future studies prospectively assessing its impact on the IBD patient population are needed.

Abbreviations:

aRR

adjusted Risk Ratio

CD

Crohn’s Disease

HCUP

Healthcare Cost and Utilization Project

IBD

Inflammatory Bowel Disease

ICD-9-CM

International Classification of Diseases Ninth Edition Clinical Modification

NRD

Nationwide Readmissions Database

UC

Ulcerative Colitis