Retrospective, propensity score–matched study examining the relationship between frailty and Clostridioides difficile infection in a national cohort of US veterans

Abstract

Background:

Frailty is often more predictive of disease and mortality compared with chronological age. This study determined the impact of frailty on Clostridioides difficile infection (CDI) risk and outcomes in a national veteran population.

Methods:

This was a retrospective cohort study of CDI and control veteran inpatients and outpatients from fiscal year 2003 to 2018. Baseline frailty was presented as the Veterans Affairs (VA) Frailty Index. Propensity score–matched analyses were conducted to compare CDI risk, CDI health outcomes, and 1-year new-onset frailty-associated conditions.

Results:

A total of 11,451 CDI and 11,451 matched control patients were included. Baseline frailty conditions were more common among CDI patients, especially involuntary weight loss (6.0% vs 3.4%, P < .001) and anemia (24.6% vs 18.7%, P < .001). VA Frailty Index was significantly higher for CDI patients (0.13 vs 0.11, P = .019). Frail CDI patients were more likely to experience 30-day mortality (11.3% vs 1.1%, P < .001) and 60-day CDI recurrence (20.4% vs 16.3%, P < .001) compared with non-/prefrail CDI patients. At 1 year, CDI patients were significantly more likely to be categorized as frail (19.6% vs 17.0%, P < .001).

Conclusions:

This study demonstrated the potential association between frailty and CDI risk and health outcomes, as well as new-onset frailty diagnoses in patients who develop CDI.

BACKGROUND

Clostridioides difficile infection (CDI) continues to place a significant burden on patients and the health care systems in the United States and worldwide. While CDI commonly manifests as diarrhea, it can lead to severe complications like megacolon, intestinal perforation, sepsis, or death. All-cause mortality at 30 days ranges from 9% to 38% with higher rates noted among older and critically ill populations.^1,2 Additionally, CDI recurrence is common. Approximately one-quarter of patients will experience at least 1 CDI recurrence after successful treatment of the initial episode.^3,4 Recurrences can lead to prolonged symptoms, increased risk for adverse events, rehospitalizations, and transmission of infection to other vulnerable patients.

Advanced age is one of the most cited risk factors for CDI and for adverse CDI health outcomes. In fact, 70% of cases occur in patients 65 years and older⁵ and older adults who develop CDI are more likely to experience severe CDI, complications, and death.^6–8 These associations have been previously described using patients’ chronological age, however, chronological age may not always accurately represent overall health status. In contrast, frailty is a measurable clinical syndrome often characterized by a decline in cognitive and physiological reserve and function associated with biological aging rather than chronological aging. Indeed, frailty status is more accurate for predicting the onset of disease and death compared with chronological age.^9,10 Additionally, CDI is often preceded by disruption of the host gut microbiome, which can persist following the episode and lead to future recurrent CDI episodes and possibly risk for other microbiome-mediated conditions. Prior studies suggest that microbiome dysbiosis may play a role in frailty development as well as a multitude of aging-related conditions^11,12 and that the magnitude of aging-related changes in microbiome diversity and community structure is greater with biological aging compared with chronological aging.⁹

Next, CDI has been associated with impaired physical, mental, and social functioning. A recent systematic literature review highlighted notable psychological and physical consequences of CDI, including anxiety, depression, persistent gastrointestinal symptoms, weight loss, and fatigue.¹³ Additional studies have described the negative impact of CDI on overall quality of life, particularly among patients who experience recurrent CDI,¹⁴ CDI patients’ ability to work and conduct activities of daily living (ADL),¹⁵ and the ability to be discharged to home following CDI hospital admission.¹⁶ Together, these findings suggest that frailty may be a more important predictor of CDI compared with age, but also that CDI may perpetuate or exacerbate frailty.

The overall prevalence of multidimensional frailty is estimated to be at least 26.8% among US adults, and increases with age.¹⁷ The associated costs of frailty were estimated to increase by at least $616 and up to $32,549 per frail adult in the community-dwelling population in the United States.¹⁸ Thus, in the setting of an aging population and growing incidence of CDI, an increasingly frail older population will have major implications for the demand for health care services, including hospital usage, home care, and long-term care. Given the paucity of data on the relationship between frailty and CDI, the objective of this study was to evaluate the interplay between baseline frailty status and the overall course of CDI in a national cohort of veterans.

METHODS

Study design and data sources

This was a retrospective cohort study of patients receiving care at any Veterans Health Administration (VHA) hospital or clinic in the United States from fiscal year 2003 to 2018. Data were obtained from the Veterans Affairs (VA) Informatics and Computing Infrastructure, which includes administrative, clinical, laboratory, and pharmacy data repositories that are linked using unique patient identifiers. The main data sources used for this study included the VA Medical SAS Datasets (both inpatient and outpatient), the VA Vital Status File, the VA Decision Support System, and the VHA Annual Enrollment Files. All data collection and analyses were performed at the South Texas Veterans Health Care System, Audie L. Murphy VA Hospital. The research protocol was reviewed and approved by the Institutional Review Boards at the South Texas Veterans Health Care System and the University of Texas Health San Antonio (approval # HSC20130473H).

Populations

All adult VHA beneficiaries (18–89 years old) were eligible for study inclusion. The cohort included adult patients who had any inpatient or outpatient visit for CDI at a VA facility from October 1, 2002 to September 30, 2018 (cohort inclusion period). CDI was defined as an ICD-9-CM or ICD-10 code for CDI (008.45 and A04.72, respectively), plus a positive stool test (eg, glutamate dehydrogenase, toxin enzyme immunoassay, and polymerase chain reaction), and active CDI therapy (oral vancomycin, fidaxomicin, or metronidazole) during the visit or within 7 days of the visit during the cohort inclusion period. To identify true index visits, we limited the cohort to first CDI episodes only by excluding patients with an ICD-9-CM or ICD-10 code for CDI in the year prior to cohort inclusion. A control group was created by randomly sampling VHA patients without an ICD-9-CM or ICD-10-CM code for CDI at any time during the study period, group matching to the CDI cohort approximately 3:1 based on fiscal year, and visit setting (inpatient vs outpatient) prior to propensity score matching as described below. For both cohorts, patients with laxative use in the 90 days preceding the visit were excluded. Patients who did not have at least 1 inpatient or outpatient VA visit each year (to ensure adequate follow-up), or those who died in the hospital during the index visit were also excluded from analyses.

Data collection

Data extraction and variable creation were conducted using SAS Version 9.4 (SAS Institute). Study-dependent variables included all-cause mortality, frailty-associated diagnoses, severe CDI, CDI recurrence, and hospital length of stay (LOS). The primary outcome was all-cause mortality at 30 days. Mortality was also assessed at 90 days and 1 year, and as a continuous variable for survival analysis. Time-to-death was defined as the date of death minus the date of cohort inclusion (date of CDI diagnosis or matching control visit) during the cohort inclusion period. For survival analyses, patients with an index visit between 2016 and 2018 were excluded to allow for adequate long-term mortality assessment. The secondary outcomes included frailty-associated diagnoses, which were assessed as dichotomous variables at specific time points (90 days and 1 year) and as continuous variables for survival analysis. Frailty-associated diagnoses included coagulopathy, involuntary weight loss, fluid and electrolyte imbalance, anemia, falls, fractures, and long-term care residence (Supplementary Material). The diagnoses were chosen based on literature review and expert opinion, as well as evidence that supports the association between frailty-related diagnoses and hospital readmissions.¹⁹ Additionally, the VA frailty index (VA-FI) was calculated for all patients.²⁰ The VA-FI was developed to estimate frailty in the veteran population using the number of age-related health deficits from cognitive and physical functioning domains and specifically identifies veterans at increased risk of mortality. The VA-FI was calculated as the number of deficits for each patient divided by 31 (the total number of possible deficits). The VA-FI was used as a continuous variable and to categorize patients as nonfrail (≤ 0.1), prefrail (> 0.1–0.2), and frail (> 0.2). Severe CDI was defined as a white blood cell count > 15 × 10⁹ cells/μL or a serum creatinine > 1.5 mg/dL during the index CDI visit.²¹ CDI recurrence was defined as a second outpatient or inpatient visit during which a patient received an ICD-9-CM/ICD-10 code for CDI, plus a minimum 3-day gap between the visit and the end of active CDI therapy for the initial episode. Hospital LOS was calculated as discharge or inpatient death day minus admission day plus 1 day.

Patient demographics, including sex, race, and Hispanic ethnicity, were defined as the most frequent reporting of each over the study period to ensure accuracy. VHA priority group was included as a marker of socioeconomic and disability status. Health care exposures in the prior year included inpatient and outpatient visits to the VA and chronic dialysis. These variables aid in determining patient acuity and healthy-user bias in each cohort. Charlson, Selim, and other relevant comorbidities, defined by ICD-9-CM or ICD-10 codes, were collected in the year prior to study inclusion. Medications ordered in the past 90 days prior to cohort inclusion were also included: antibiotics, gastric acid suppressants, opioids, motility agents, and cancer chemotherapy.

Statistical methods

All statistical analyses were conducted using JMP 17 Pro (SAS Institute) or Stata 16 (StataCorp LLC). First, unmatched CDI and control group baseline characteristics were compared using bivariable statistics (χ², Fisher exact, or Wilcoxon rank sum as appropriate). Next, CDI and control groups were propensity score–matched based on characteristics at cohort inclusion (Table 1). The probability of patients experiencing CDI was estimated using a logistic regression model with CDI as the dependent variable and the following covariates: patient demographics, comorbidities and exposures in prior year, and medications in the prior 90 days. Nearest-neighbor matching was used to match CDI patients to non-CDI controls 1:1 within a propensity score maximum caliper of 0.001. Following propensity score matching, baseline characteristics were compared between groups. To determine if baseline frailty was associated with CDI risk, the prevalence of baseline frailty–associated diagnoses and VA-FI was compared between the CDI and control cohorts (unmatched and matched cohorts). The odds of CDI among frail patients were assessed using a logistic regression model with cohort (CDI vs control) as the dependent variable and frailty category as the independent variable.

Table 1.

Baseline characteristics of the unmatched and propensity score–matched cohorts

Characteristic	Unmatched cohorts			Matched cohorts
	CDI (n = 25,755)	Control (n = 74,533)	P value	CDI (n = 11,451)	Control (n = 11,451)	P value
Age, median (IQR)	67 (60–77)	61 (52–69)	< .001	65 (58–75)	65 (57–74)	.001
Male sex, n (%)	95.1	93.4	< .001	94.5	94.4	.954
Race, n (%)			< .001			< .001
Black	16.1	25.0		16.1	21.5
Other	75.5	69.6		1.7	1.4
White	1.6	1.7		75.9	72.7
Unknown	6.7	3.7		6.2	4.5
Hispanic ethnicity, n (%)	4.4	5.5	< .001	4.6	3.7	< .001
Inpatient setting, n (%)	87.9	88.2	.138	85.9	86.1	.732
VA priority group, median (IQR)	5 (2–5)	5 (1–5)	< .001	5 (2–5)	5 (2–5)	.997
Exposure in prior year
Inpatient visits, median (IQR)	1 (0–2)	0 (0–1)	< .001	0 (0–1)	0 (0–1)	< .001
Outpatient visits, median (IQR)	26 (13–46)	24 (12–45)	< .001	22 (10–41)	24 (12–42)	< .001
Chronic dialysis, n (%)	4.3	1.6	< .001	3.1	2.9	.352
Medications in the past 90 d, n (%)
Antibiotics	49.7	22.4	< .001	35.4	35.6	.814
Gastric acid suppressants	49.7	35.8	< .001	39.2	40.4	.061
Opioids	38.6	25.8	< .001	29.9	31.0	.066
Cancer chemotherapy	5.0	2.8	< .001	4.1	4.2	.487
Comorbidities in the past year, n (%)
Myocardial infarction	7.5	3.7	< .001	5.0	5.2	.368
Congestive heart failure	20.7	11.5	< .001	15.2	15.5	.558
Peripheral vascular disease	19.1	11.1	< .001	15.2	15.4	.646
Cerebrovascular disease	16.3	9.6	< .001	12.5	13.0	.284
Dementia	5.0	1.7	< .001	3.1	3.1	.761
Chronic pulmonary disease	33.4	24.5	< .001	28.0	28.4	.500
Rheumatic disease	2.7	1.8	< .001	2.2	2.4	.482
Peptic ulcer disease	3.1	1.6	< .001	2.1	2.1	.819
Mild liver disease	11.1	8.2	< .001	9.5	9.1	.285
Moderate/severe liver disease	2.4	0.9	< .001	1.5	1.4	.741
Diabetes without complications	36.5	34.5	< .001	33.8	33.8	.978
Diabetes with complications	16.8	12.7	< .001	14.5	14.2	.546
Renal disease	22.4	10.7	< .001	17.1	17.0	.930
Cancer	22.6	12.5	< .001	18.6	18.8	.760
Cancer with metastasis	4.8	1.2	< .001	2.9	3.0	.556
HIV/AIDS	1.6	1.3	< .001	1.4	1.2	.384
Schizophrenia	3.6	8.3	< .001	3.6	4.1	.063
Depression	24.8	29.0	< .001	23.9	23.8	.951
Bipolar disorder	10.3	17.3	< .001	11.1	11.5	.338
Anxiety disorder	5.7	6.3	< .001	5.5	5.6	.583
Post-traumatic stress disorder	12.9	21.8	< .001	14.6	14.9	.467
Alcohol use disorder	3.6	2.7	< .001	3.0	3.1	.490
Hypertension	70.9	63.0	< .001	65.1	65.6	.397
Dyslipidemia	5.6	5.3	.050	6.2	6.3	.765
Obesity	17.6	21.7	< .001	18.3	18.5	.682
Gastroesophageal reflux disorder	25.5	22.4	< .001	23.5	24.2	.198
Transplant	2.0	0.8	< .001	1.6	1.5	.667
Inflammatory bowel disease	3.1	0.8	< .001	2.7	2.4	.134
Irritable bowel syndrome	1.2	0.9	.001	1.1	1.2	.322

CDI, Clostridioides difficile infection; VA, Veterans Affairs.

Next, to determine the impact of frailty on CDI health outcomes (severe CDI, 30-day mortality, and 60-day recurrence), we conducted a subgroup analysis of only patients with CDI. CDI patients were propensity score–matched based on frailty category (nonfrail/prefrail and frail) similar to the methods above. Time-to-death was assessed between groups using the log-rank test. For the recurrence model, we additionally excluded patients who died within 60 days of cohort inclusion. The final statistical objective was to compare new-onset frailty during follow-up between the CDI and control groups. To accomplish this, we first excluded patients with any of the 7 frailty-associated conditions at baseline and those who died within 1 year of cohort inclusion to establish temporality and limit survivor bias. Next, new propensity scores were generated that estimated the probability of CDI based on the previously defined variables, plus baseline frailty index, using similar methods as described above. Then, the 1-year VA-FI and frailty category were compared between matched CDI and control patients using the χ² or Wilcoxon rank-sum test. Spearman rank correlation was used to assess the relationship between VA-FI and number of CDI episodes.

RESULTS

Population baseline characteristics

A total of 40,643 CDI patients were originally screened for inclusion based on ICD-9-CM/ICD-10 codes and a positive stool test (Fig. 1). Of these, 14,888 (36.6%) were excluded for the following reasons: CDI episode in the prior year (n = 446), lack of documented CDI therapy (n = 6,355), inpatient mortality (n = 2,633), recent laxatives (n = 5,246), and < 1 day of follow-up (n = 208). A total of 81,280 control patients were originally screened for inclusion based on lack of CDI ICD-9-CM/ICD-10 codes. Of these, 6,747 (8.3%) were excluded for the following reasons: inpatient mortality (n = 435), recent laxatives (n = 6,229), and < 1 day of follow-up (n = 83). After exclusions, the unmatched cohorts included 25,755 CDI patients and 74,533 controls. CDI and control patients in the unmatched cohorts significantly differed with respect to nearly all variables assessed, including greater frequency of comorbidities, health care, and medication exposures (Table 1).

Fig. 1.

Study flowchart. CDI, Clostridioides difficile infection; LOS, length of stay; PSM, propensity score match; PSM-1, CDI risk estimated using logistic regression and covariates (Table 1) and groups matched 1:1; PSM-2, CDI patients only matched based on baseline frailty category (nonfrail/prefrail and frail); PSM-3, groups rematched similar to PSM-1 with baseline frailty index included (created with BioRender.com).

After the first propensity score matching, a total of 11,451 CDI- and 11,451 control patients were available for analyses. These groups were well-matched for baseline demographics, comorbidities, and medications. Statistical, but small numeric, differences remained for race, Hispanic ethnicity, and inpatient and outpatient visits in the prior year likely due to the large sample size. The baseline characteristics of the unmatched and propensity score–matched cohorts are presented in Table 1.

Association between baseline frailty and CDI

In the unmatched cohorts, the baseline prevalence of frailty-associated diagnoses, baseline percentage of patients categorized as frail, as well as the VA-FI were all significantly higher in the CDI cohort compared with control patients (Table 2). These trends persisted, but were numerically diminished, when comparing the matched cohorts. In the matched cohorts, CDI patients more frequently had prior involuntary weight loss, fluid and electrolyte imbalances, anemia, falls, and long-term care residence at baseline. CDI patients also had a slightly higher median VA-FI (0.13 vs 0.11, P = .019) and more CDI patients were categorized as frail (20.0% vs 16.9%, P < .001). Frail patients were 1.2 times more likely to be in the CDI cohort compared with controls (OR 1.23, 95% confidence interval [CI] 1.14–1.32).

Table 2.

Baseline frailty among CDI and control cohorts

Frailty characteristic	Unmatched			Matched
	CDI (n = 25,755)	Control (n = 74,533)	P value	CDI (n = 11,451)	Control (n = 11,451)	P value
Coagulation disorder, n (%)	2.8	1.4	< .001	2.0	2.1	.963
Involuntary weight loss, n (%)	6.6	2.7	< .001	6.0	3.4	< .001
Fluid and electrolyte imbalance, n (%)	0.7	0.2	< .001	0.5	0.2	.005
Anemia, n (%)	32.7	13.7	< .001	24.6	18.7	< .001
Falls, n (%)	4.6	1.9	< .001	3.6	2.3	< .001
Fractures, n (%)	6.6	4.8	< .001	5.4	5.0	.108
Long-term care, n (%)	11.4	2.4	< .001	6.0	3.2	< .001
VA frailty index, median (IQR)	0.13 (0.08–0.23)	0.10 (0.06–0.16)	< .001	0.13 (0.06–0.19)	0.11 (0.06–0.18)	.019
VA frailty index category, n (%)			< .001			< .001
Nonfrail	37.6	55.5		47.0	48.7
Prefrail	34.1	32.6		33.0	34.4
Frail	28.3	11.9		20.0	16.9

CDI, Clostridioides difficile infection; VA, Veterans Affairs.

CDI patients were also significantly more likely to experience short- and long-term mortality compared with matched controls: 30-day mortality (11.3% vs 1.1%), 90-day mortality (16.9% vs 2.4%), 1-year mortality (27.8% vs 6.4%) (P < .001 for all), and Kaplan-Meier log-rank P < .001 (Supplementary Material).

Association between frailty and CDI health outcomes

Among CDI patients, a total of 1,961 were included for analysis after matching based on baseline frailty group. CDI patients who were frail more often died at each follow-up period compared with the nonfrail/prefrail groups (Fig. 2) (P < .05 for all comparisons). The Kaplan-Meier curves also demonstrate the differences in the time-to-death between CDI frailty categories (log-rank P value < .001) (Supplementary Material). Among CDI patients who died during the study period, median time-to-death was the lowest among frail patients (221 days) compared with prefrail (245 days) and nonfrail (804 days) patients. CDI patients categorized as frail as compared with nonfrail/prefrail patients were also more likely to experience 60-day CDI recurrence (20.4% vs 16.3%, P < .001), but similar severe CDI (47.8% vs 48.2%, P = .750) and hospital LOS (median LOS 11 days for both groups, P = .344).

Fig. 2.

Mortality for CDI patients matched by frailty category. CDI, Clostridioides difficile infection.

CDI and frailty outcomes

After exclusion of patients with baseline frailty–associated diagnoses and those who died before the 1-year follow-up, 5,804 CDI patients and 5,804 matched control patients were eligible for 1-year outcomes. At 1 year, CDI patients were significantly more likely to have new-onset frailty-associated diagnoses of coagulation disorders, involuntary weight loss, and anemia. More CDI patients were categorized as frail at 1 year as well (19.6% vs 17.0%, P < .001). It is important to note that the numeric differences between groups were relatively small. The number of CDI episodes during 1-year follow-up was positively and significantly correlated with 1-year VA-FI (R = 0.17, P < .001), though CDI episodes explained little of the variability in 1-year VA-FI (R² = 0.03) (Table 3).

Table 3.

New-onset frailty at 1 year among matched CDI and control groups

Frailty characteristic at 1 y	CDI (n = 5,804)	Control (n = 5,804)	P value
Coagulation disorder, n (%)	1.8	1.3	.042
Involuntary weight loss, n (%)	4.4	2.6	< .001
Fluid and electrolyte imbalance, n (%)	0.4	0.4	.661
Anemia, n (%)	20.4	14.7	< .001
Falls, n (%)	4.8	4.7	.727
Fractures, n (%)	4.3	4.5	.470
Long-term care, n (%)	5.5	5.0	.227
VA frailty index, median (IQR)	0.10 (0.03–0.13)	0.10 (0.03–0.13)	.055
VA frailty index category, n (%)			< .001
Nonfrail	48.3	47.1
Prefrail	32.0	35.9
Frail	19.6	17.0

CDI, Clostridioides difficile infection; VA, Veterans Affairs.

DISCUSSION

This is one of the first studies to document the relationship between CDI and frailty status, a potential indicator of biological age, as a predictor of CDI and health outcomes rather than chronological age. In this national study of US veterans, baseline frailty was more common among CDI patients compared with matched controls and was associated with higher rates of mortality and 60-day recurrence among CDI patients. In addition, a higher percentage of CDI patients developed new-onset frailty-associated diagnoses at 1 year compared with matched controls.

These findings have significant implications for clinical practice and management of patients with CDI. For example, if patients with baseline frailty are at higher risk for CDI and worse CDI outcomes, clinicians may need to begin frailty screening as part of risk assessment for antimicrobial use. Integrating frailty assessment can improve the holistic use of antibiotics so that clinicians may have a more complete risk assessment when prescribing antibiotics. Additionally, if CDI accelerates the progression toward frailty, then patients with CDI may require enhanced/additional screening for frailty in the postacute phase of CDI. Therefore, if these results are corroborated, they may impact national guideline recommendations for the types of screenings that may be routinely incorporated post CDI.

Though prior evidence is limited, our findings are generally in-line with previously published literature.²² Frailty and decreased cognitive or functional status have been previously reported as independent predictors of CDI. Olsen et al²³ conducted a longitudinal study using Medicare data to identify risk factors for CDI. Notably, this study found that chronological age did not significantly improve CDI risk prediction when included in a statistical model. However, exclusion of frailty indications (eg, dementia, failure to thrive, and difficulty walking) and other notable risk factors (eg, infections, other acute conditions, and prior health care utilization) resulted in worsened model performance and model fit, suggesting that frailty status is a more important CDI risk factor compared with age.

In addition to CDI risk, frailty and poor baseline functional status have also been demonstrated to be associated with other important CDI health outcomes. In a study by Rao et al,²⁴ CDI patients older than 50 years were classified by their ability to perform ADL. CDI patients that were classified as needing full ADL assistance were significantly more likely to experience severe CDI when compared with the independent ADL class (adjusted odds ratio (aOR) 7.00, 95% CI 1.83–26.79). Other aging-related conditions, such as congestive heart failure, and Charlson comorbidity score, were also independent predictors of severe CDI. Interestingly, ADL class was not found to be significantly associated with secondary outcomes (eg, LOS, 90-day mortality, and 90-day readmission), though this may be related to small sample size. Another study by Kyne et al²⁵ found that among 73 patients with CDI, cognitive impairment was significantly associated with severe CDI, but age was not. While our study did demonstrate an association of baseline frailty with CDI risk, baseline frailty did not significantly predict CDI severity. We hypothesize that this may be due to our population that excluded patients with inpatient mortality. Given that frail patients were more likely to experience short-term mortality, it is possible that the many frail patients were excluded from the health outcome analyses.

Frailty has also been demonstrated to be associated with other important CDI outcomes, including mortality. A 2016 retrospective study by Leibovici-Weissman et al²⁶ analyzed 184 hospitalized CDI patients aged 80 years and older and found that patients dependent in their ADLs were much less likely to survive long-term than patients independent in ADLs (adjusted hazard ratio (aHR) 0.46, 95% CI 0.24–0.90). These findings are further supported by another 2-center, case-control study that assessed 106 CDI patients and 106 non-CDI control patients. In addition to higher-baseline functional debility in CDI patients compared with controls (84 vs 69%, P = .014), CDI patients more often experienced delirium during hospitalization (28 vs 14%, P = .028), functional decline or death (39 vs 14%, P < .001), and had higher rates of short- and long-term mortality. Further highlighting the impact that cognitive status has on CDI health outcomes, CDI patients with delirium experienced greater functional decline or death than those without delirium (57 vs 32%, P = .017) and CDI patients with dementia experienced greater functional decline or death (67 vs 32%, P = .003), 90-day mortality (43 vs 18%, P = .013), and 180-day mortality (67 vs 25%, P < .001) when compared with CDI patients without dementia.²⁷ Notably, CDI has also been associated with worse mental and physical health, work time missed, impaired work, and impaired ADLs.^15,28

While several studies have focused on functional status and markers of frailty, very few studies have included a measure of frailty and assessed the impact it has on CDI health outcomes. Chaar et al²⁹ utilized the 2017 National Inpatient Sample database to analyze 93,810 hospitalized CDI patients and compare outcomes and health care utilization by Hospital Frailty Risk Score category (eg, frail score ≥ 5, nonfrail score < 5). Frail patients had higher odds of fulminant CDI (aOR 1.9, 95% CI 1.6–2.3), colectomy (aOR 4.1, 95% CI 1.5–11.2), inpatient mortality (aOR 4.5, 95% CI 2.8–7.1), and intensive care unit admission (aOR 13.7, 95% CI 6.3–29.9). Next, a population-based cohort study by Rubak et al³⁰ classified 457 older adult CDI patients by frailty category using the Multidimensional Prognostic Index. Overall, patients with severe (aHR 10.15, 95% CI 4.06–25.36) or moderate frailty (aHR 2.70, 95% CI 1.03–7.11) were more likely to experience 90-day mortality compared with low-frailty patients. Additionally, frailty was found to predict 90-day mortality (receiver operating characteristic (ROC) area under the curve (AUC) 77%) better than age (P < .001) and CDI severity (P = .04). Venkat et al³¹ evaluated 470 patients with fulminant CDI undergoing colectomy using a modified frailty index. This study found that increasing frailty score was independently associated with overall morbidity (aOR 13.0), mortality (aOR 8.8), cardiopulmonary complications (aOR 6.8), and prolonged hospital stay (aOR 6.6). Of frail CDI patients, 56% experienced 30-day mortality and 75% experienced postcolectomy complications. Last, Boone et al³² created a frailty index for older hospitalized adults with CDI. Using 36 deficits, most CDI patients were classified as frail at baseline (89%) and frailty was significantly associated with 1-year CDI recurrence. These data in addition to ours support the use of frailty diagnoses and composite indices as potentially important factors to assess and control for in future CDI studies.

Frailty-associated conditions are commonly described in the CDI literature due to the acute nature of many of the symptoms and their sequelae that mimic frailty-associated conditions. For example, fluid and electrolyte imbalances are a common sign of severe infectious diarrhea, due to dehydration. Therefore, correction of fluid and electrolyte imbalances is a common part of CDI management, as well as monitoring weight loss while receiving care. Additionally, CDI patients often experience loss of appetite, nausea/vomiting, and abdominal bloating/tenderness following a CDI episode, particularly with repeat episodes; thus, poor nutrition may result in weight and muscle loss. This may explain why the primary drivers of the differences in frailty between groups in this study were involuntary weight loss and anemia. Furthermore, there was a positive association between the number of CDI episodes and the development of frailty-associated conditions at 1 year. This “dose-response” relationship strengthens our hypothesis that CDI is associated with the development of frailty conditions by demonstrating that the more CDI episodes a patient has, the higher their risk for frailty.

This study has several strengths. First, we used a large, nationally representative sample to assess the interplay between frailty and CDI risk and health outcomes. This allowed for rigorous group matching to limit confounding and for powerful statistical comparisons between groups. Additionally, the data used for this study included robust information pertaining to disease severity, medication use (eg, antibiotics used for CDI), and multiple encounters allowing for a more comprehensive analysis. This was also the first study to analyze CDI patient health outcomes using the VA-FI.

This study has inherent limitations due to its retrospective design; therefore, a causal relationship between frailty and CDI cannot be established without further validation studies. The data may be subject to misclassification bias and confounding due to the use of administrative coding from electronic health records. We attempted to minimize these limitations with the use of robust methods to confirm CDI classification as well as control for confounding using propensity score–matched cohorts. The use of the veteran-only data provided for a predominantly older, male population; therefore, the findings may not be generalizable to non–VHA care settings. It is possible that veterans sought care outside of the VHA system and certain outpatient or inpatient visits were missed. It is largely unknown how this outside care utilization may impact the electronic measurement of frailty using the VA-FI. Finally, although the size and scope of our cohort is a major strength, the sample size could result in overpowered statistical comparisons, whereby even small differences between groups (ie, small effect size) result in statistically significant differences between groups; therefore, statistically significant findings should be evaluated in the context of clinical significance. Although the clinical threshold for a significant increase is not well-defined, any new-onset frailty-associated condition can give an individual patient burden, and with the high incidence of CDI that can translate to a large societal burden.

CONCLUSIONS

Despite advances in infection prevention and antimicrobial stewardship, CDI continues to threaten public health. In this large retrospective study of US veterans, CDI patients were more likely to be frail at baseline compared with a matched control group. These findings suggest a possible association between frailty and CDI risk. Frailty status predicted risk for mortality and CDI recurrence. Importantly, patients who developed CDI were more likely to develop new-onset frailty diagnoses, notably involuntary weight loss and anemia, compared with matched controls over a 1-year follow-up. These findings could have major implications for public health and clinical care. Frailty as a marker of biological age, rather than chronological age alone, should be considered in the design of future studies. Future studies are needed to validate the study findings, determine the biological basis for such associations, and guide future treatment decisions for patients with CDI.

APPENDIX A. SUPPLEMENTARY DATA

Supplementary data related to this article can be found at doi:10.1016/j.ajic.2024.08.020.