Geriatric factors affecting the safety of direct oral anticoagulants in atrial fibrillation patients aged 80 years and older: An observational prospective cohort study

INTRODUCTION

Atrial fibrillation (AF) is a common condition that occurs in 1–2% of the general population, with the number of AF patients increasing with age, from <0.5% at ages 40–50 to 5–15% at the age of 80. The number of AF patients is expected to nearly double in the next 50 years. This pathology is associated with a high risk of systemic thromboembolism and ischemic stroke, which can be prevented via treatment with direct oral anticoagulants (DOACs) [1][2]. However, their use may cause bleeding, especially in elderly patients due to a number of age-related changes in the body. In particular, decreased hepatic perfusion and reduced activity of some cytochrome P450 isoforms affect the pharmacokinetics of DOACs, while the age-related decrease in the number of functioning nephrons increases the elimination half-life of these drugs. The receptor sensitivity and distribution density are modified, which directly affects the pharmacodynamic properties of DOACs¹ [3–5].

Old age is known to be associated with the presence of geriatric syndromes, which include chronic pain syndrome, urinary/fecal incontinence, high fall risk, cognitive impairments, dementia, decreased mobility, frailty, etc. Frailty constitutes a major geriatric syndrome causing a decrease in the physiological reserves of the body.²

In the 2020 clinical guidelines of the European Society of Cardiology (ESC) on AF diagnosis and treatment [6], the HAS-BLED scale (hypertension, abnormal renal/liver function, stroke, bleeding history or predisposition, labile international normalized ratio, elderly (> 65 years), drugs/alcohol concomitantly) assessed patient frailty, and risk factors for bleeding complications included certain geriatric syndromes, e.g., cognitive impairments or dementia and high fall risk. The ESC clinical guidelines for AF treatment revised in 2024 [2] abandon HAS-BLED bleeding risk assessment since systematic reviews and controlled trials in comparison groups showed highly inconsistent results and little prognostic value while focusing on the assessment and management of modifiable and potentially modifiable risk factors for bleeding, which include a high fall risk and cognitive impairments or dementia. The effect of frailty on the development of bleeding is mentioned neither in the 2024 EOC guidelines for AF treatment [2] nor in the clinical guidelines for the treatment of atrial fibrillation and flutter of the Russian Society of Cardiology [1].

Nevertheless, the results of studies in elderly AF patients receiving DOACs show that the risk of falls and frailty increase the risk of bleeding [7–11]. These studies primarily focus on major and intracranial bleeding, with clinically relevant non-major bleeding (CRNMB) not considered as a separate group. Nevertheless, the incidence of CRNMB is much higher than that of major bleeding, reaching 17.4% and increasing with age [12][13]. This complication should not be underestimated, as it is a reason for DOAC discontinuation, which increases the risk of thromboembolism and, consequently, disability or a fatal outcome. In addition, CRNMB can lead to major bleeding or patient hospitalization and a medical intervention [12]. We found no information on the contribution of other geriatric syndromes to bleeding events in patients receiving DOACs.

Current studies, therefore, lack data on geriatric syndromes as predictors of CRNMB on DOACs. Furthermore, the mean age of patients in studies examining the relationship between geriatric syndromes and DOAC safety ranges from 75 to 78 years, while the number of patients aged over 80 years is rising worldwide with increasing life expectancy [14][15]. This fact prompted the present study.

The present study aims to assess the potential of geriatric factors for predicting the risk of clinically relevant non-major bleeding on direct oral anticoagulants in nonvalvular atrial fibrillation patients aged 80 years and older.

METHODS

Study design

The observational prospective cohort study of actual clinical practice included 367 AF patients on DOACs aged ≥80 years who were treated at a multi-specialty inpatient clinic in Moscow from January 2019 to December 2022.

Study conditions

The study was conducted at the Department of Clinical Pharmacology and Therapy named after Academician B.E. Votchal, Russian Medical Academy of Continuous Professional Education, Ministry of Health of the Russian Federation. The patients were included in the study from January 2019 to December 2022. After that, outpatient and inpatient medical records were prospectively analyzed to determine the incidence of CRNMB during DOAC therapy. The observation period lasted one year.

Eligibility criteria

Inclusion criteria

Nonvalvular atrial fibrillation patients of both sexes; aged 80 years or older at study entry; regular DOAC use for at least one year from the time of study entry; signed voluntary informed consent for participation in the study.

Exclusion criteria

Age of <80 years; clinically significant heart disease (including cardiogenic shock; recent, i.e., less than one month prior to the study, complicated myocardial infarction; third-degree atrioventricular block without artificial pacemaker; hypertrophic cardiomyopathy; severe aortic and mitral stenosis); significant liver diseases (including liver cirrhosis with ascites); patients receiving renal replacement therapy (long-term hemodialysis; peritoneal dialysis; history of kidney transplant); clinically significant immunologic disease; neurological diseases (including acute cerebrovascular disorders and transient ischemic attack less than three months prior to the study); surgical procedure (except for dental or cosmetic surgeries), injuries, and fractures within the previous three months; presence of clinically significant changes in laboratory parameters indicating undiagnosed disease and requiring further examination.

Removal criteria

Violation of the examination and treatment procedures by the patient; refusal of the patient to participate in the study and withdrawal of informed consent.

Description of eligibility criteria (diagnostic criteria)

The diagnosis of AF was made according to the criteria of clinical guidelines for the treatment of atrial fibrillation and flutter of the Russian Society of Cardiology [1]. In patients receiving DOACs, CRNMB was diagnosed using criteria developed by the International Society on Thrombosis and Haemostasis (ISTH) [12]. The list of interventions performed on patients is regulated by the order of the Ministry of Health of the Russian Federation of January 29, 2016 (Order No. 38n) “On Approval of the Geriatric Medical Care Protocol” as amended on December 20, 2019 (Order No. 1067n).³

Selection of group members

The study cohort included 367 patients aged ≥80 years who were consecutively enrolled in the study, provided they met the inclusion criteria. At the end of the one-year observation, the patients were divided into two groups according to the data on CRNMB manifestation and recurrence: the main group included patients with CRNMB (n = 195), and the control group comprised patients without CRNMB (n = 172).

Target parameters in the study

Main parameters in the study

In order to identify geriatric factors affecting the development of CRNMB in patients on DOACs, patients underwent a comprehensive geriatric assessment, which included analysis in four main domains: physical status; mental and emotional status; functional capacity; identification of social problems.⁴

Additional parameters in the study

Additional parameters are not provided in this study.

Methods for measuring the target parameters

The study used the following validated charts, questionnaires, scales, tests, and estimated indices: a chart of comprehensive geriatric assessment,⁵ the Age is No Barrier questionnaire, self-assessment of quality of life and health status on the visual analogue scale (VAS), the Barthel Index for Activities of Daily Living (ADL) [16], the Lawton Instrumental Activities of Daily Living (Lawton IADL) Scale [17], Mini Nutritional Assessment [17][18], Short Physical Performance Battery, Timed Up and Go Test, and single-leg stance test.⁶ Also, all patients underwent dynamometry to determine handgrip strength; the criteria for low handgrip strength were established depending on sex and body mass index as per the clinical guidelines for frailty.²

The frailty index was calculated on the Rockwood scale [19] and on the scale used in the SPRINT (Systolic Blood Pressure Intervention Trial) study [20].

Mental and emotional status was assessed using the following tests, scales, and questionnaires⁷: Mini-Cog screening test, Mini-Mental State Examination, Montreal Cognitive Assessment Scale, Word List Recall, Digit Symbol Substitution Test, 10-word-list recall test, verbal fluency test (literal fluency/letters and category fluency/animals), Boston Naming Test, Functional Activities Questionnaire, Alzheimer’s Disease Assessment Scale-Cognitive, as well as Geriatric Depression Scale (GDS), GDS-15 (short form) and GDS-30 (long form).

Variables (predictors, confounders, and effect modifiers)

In the present study, factors affecting the final outcome were categorized as exclusion criteria, thus completely excluding the possibility of their presence in the examined patients.

Statistical procedures

Principles behind sample size determination

The sample size was not determined in advance.

Statistical methods

Statistical data analysis and visualization were performed in the R software environment, version 4.4.0 (R Foundation for Statistical Computing, Austria). Descriptive statistics are presented as absolute and relative frequencies for qualitative variables; as medians (Me) (Q1; Q3), for quantitative variables. Pearson’s χ² test and Fisher’s exact test were used to compare the groups in terms of categorical indicators (with a minimum expected number of observations <5 in the cells of the contingency table). The Mann—Whitney test was used to compare the two groups in terms of quantitative and ordinal measures. Odds ratios (ORs) with the corresponding 95% confidence intervals (CIs) were used to assess the strength of quantitative and binary indicators with binary outcomes.

The stepwise selection of predictors for prognostic models was performed using the Akaike information criterion (AIC). The selected predictors were included in a multi-factor logistic regression model without interactions. The Nagelkerke pseudo-R², Somers’ D_XY coefficient, and C-index (area under the curve, AUC) were estimated as model quality metrics, including metrics adjusted for potential overfitting using a nonparametric bootstrap (B = 1000). The multicollinearity of predictors was assessed using the variance inflation factor (VIF). The optimal threshold value of predicted probability was determined employing Youden’s index; the prognostic accuracy, sensitivity, specificity, and positive and negative predictive values were assessed with the corresponding 95% CI.

RESULTS

Sampling

Continuous sampling was conducted according to the inclusion/exclusion criteria. Upon inclusion in the study, all participants underwent a comprehensive geriatric assessment. CRNMB detection during DOAC therapy was based on a prospective study of outpatient and inpatient medical records. The observation period lasted one year. The patients were divided into groups according to the presence or absence of CRNMB. No group had any patients who refused to participate in the study.

Characteristics of the study sample (groups)

Nonvalvular AF patients (n = 367) were selected according to the inclusion and exclusion criteria (median age of 84 [ 82; 88] years). In order to prevent thromboembolism, all participants were receiving DOACs: rivaroxaban (n = 238/367, relative proportion of 64.9%), apixaban (n = 95/367, relative proportion of 25.9%), or dabigatran etexilate (n = 34/367, relative proportion of 9.2%). The main group included 195 patients with CRNMB (median age of 84 [ 82; 87] years), with women accounting for 66.7%. The control group consisted of 172 patients without CRNMB (median age of 84 [ 82; 88] years), with women accounting for 72.7%. No significant differences were found between the groups in terms of sex, age, risk of ischemic stroke on the CHA2DS2-VASc (Congestive heart failure; Hypertension; Age; Diabetes mellitus; Prior Stroke or TIA or Thromboembolism; Vascular disease; Age 65–74 years; Sex category) scale, risk of bleeding on the HAS-BLED scale, body mass index, and key bleeding risk factors according to the criteria of clinical guidelines for the treatment of atrial fibrillation and flutter of the Russian Society of Cardiology [1][20].

Main study results

Analysis of geriatric factors affecting the development of clinically relevant non-major bleeding

The social domain was assessed by studying 24 parameters, including the presence of a disability; the presence of a partner or spouse; the presence of children; marital status; social circle and social contacts; living conditions; financial capabilities; work activity; profession; education; religiosity; the need for outpatient, inpatient, routine, or emergency medical care; preferential provision of medicines. The patients with CRNMB exhibited higher social engagement as compared to the patients without CRNMB. For example, it was statistically significantly more common for the patients with CRNMB to live in a family (p = 0.031); living alone was statistically significantly associated with 1.65-time lower odds (95% CI: 1.08; 2.51) of having CRNMB as compared to living in a family (p = 0.025). As compared to the patients without CRNMB, the CRNMB patients had a statistically significantly higher number of household members (p = 0.003); on average, a one-person increase in the number of household members was statistically significantly associated with 1.22-time higher odds (95% CI: 1.03; 1.47) of having CRNMB (p = 0.025). It was statistically significantly more common for the patients with CRNMB to be married (p = 0.014) and attend church (p = 0.046); regular church attendance (mosque, synagogue, Buddhist temple, or place of worship) was found to be associated with 3.75-time higher odds (95% CI: 1.23; 11.5) of having CRNMB (p = 0.025) (Table 1).

The mental and emotional domain was analyzed using ten scales that cover cognitive functions and emotional state. As compared to the patients without CRNMB, the CRNMB patients were found to be more emotionally stable. For example, the patients with CRNMB had statistically significantly lower scores on the GDS-15 and GDS-30 scales (p = 0.007 and 0.027, respectively); on average, every 1-point increase in the GDS-15 scale score was associated with 1.08-time lower odds (95% CI: 1.02; 1.14) of having CRNMB (p = 0.01), and every 1-point increase in the GDS-30 scale score was associated with 1.04-time lower odds (95% CI: 1.00; 1.07) of having CRNMB (p = 0.032) (Table 2).

Thirteen parameters were analyzed to assess functional activity and physical health, which included screening for frailty; self-assessment of quality of life and health status; basic and instrumental activities; mobility; muscle strength; nutritional status; presence of the chronic pain syndrome, orthostatic hypotension, sensory deficits, and urinary and fecal incontinence; occurrence of falls; use of assistive devices. The results revealed that compared to the patients without CRNMB, the CRNMB patients were less likely to have frailty and use assistive devices, rated their quality of life and health status higher, were more active in performing activities of daily living, and were more mobile. For example, the patients with CRNMB exhibited better mood (OR = 0.65 (95% CI: 0.43; 0.99), p = 0.043), and it was statistically significantly more uncommon for them to have frailty according to the Age is No Barrier questionnaire (p = 0.015). As compared to the patients without CRNMB, the CRNMB patients had higher VAS scores measuring quality of life and health status (p = 0.053 and 0.05, respectively); on average, every 1-point increase in the VAS quality of life score was associated with 1.09-time higher odds (95% CI: 0.99; 1.21) of having a bleeding event (p = 0.084); every 1-point increase in the health status score, 1.11-time higher odds (95% CI: 0.99; 1.26) (p = 0.07). The patients with CRNMB had statistically significantly higher Lawton IADL scores (p = 0.004); on average, every 1-point increase in the score was associated with 1.17-time higher odds (95% CI: 1.04; 1.32) of having CRNMB (p = 0.008). The presence of CRNMB was associated with a longer tandem stance time (p = 0.014); on average, for every one-second increase, the odds of having CRNMB increased by 1.05 times (95% CI: 1.01; 1.10) (p = 0.013). The CRNMB patients had statistically significantly lower scores on the 5-time chair stand test and Timed Up and Go Test as compared to patients without CRNMB (p = 0.045 and 0.025, respectively); on average, every 10-point increase in scores on these scales was associated with 1.16-time (95% CI: 1.01; 1.34) (p = 0.045) and 1.23-time lower odds (95% CI: 1.02; 1.50) of having CRNMB (p = 0.03), respectively (Table 3).

The use of assistive devices was statistically significantly less common in the patients with CRNMB as compared to the patients without CRNMB (p = 0.022) (Table 4).

The frailty index was calculated using a deficit accumulation model, which involved the assessment of 46 deficits on the Rockwood scale [18] and 34 deficits on the scale used in the SPRINT study [19]. An analysis of frailty indices revealed that the CRNMB patients were less frail compared to the patients without CRNMB according to the frailty index used in the SPRINT study. For example, a statistically significant association was found between the presence of CRNMB and lower SPRINT scores (p = 0.005); on average, every 0.1-point increase in the score was associated with a decrease in the odds of having CRNMB by 1.46 [ 1.14; 1.88] times (p = 0.003) (Table 5).

This study presents a systems approach to comprehensive geriatric assessment that relies on the analysis of four key domains: social functioning; mental and emotional status; physical health; functional capacity. The research design enables a comprehensive status analysis of elderly patients, taking the modern geriatric principles into account. The findings provide a holistic understanding of the relationships between different aspects of geriatric status, which is essential for the development of personalized approaches in geriatric practice.

Multi-factor analysis predicting the probability of clinically relevant non-major bleeding events

Table 6 presents a multi-factor model for predicting the probability of CRNMB, developed using a stepwise selection of predictors, with their exclusion according to the Akaike information criterion (AIC). The resulting model was characterized by a Nagelkerke pseudo-R² value of 0.1 (adjusted value of 0.05), a Somers’ D_XY coefficient of 0.27 (adjusted value of 0.23), and an AUC of 0.64 [ 95% CI: 0.58; 0.69] (adjusted value of 0.61) (Figs 2, 3). The coefficients of the model were used to develop a prognostic nomogram to estimate the probability of CRNMB on DOACs (Fig. 4).

At a threshold predicted CRNMB probability of 55%, the resulting model was characterized by a prognostic accuracy of 59.9% [ 95% CI: 54.6; 65.1], a sensitivity of 48.9% [ 95% CI: 41.5; 56.4], and a specificity of 72.4% [ 95% CI: 64.9; 79.1], with a positive predictive value of 66.7% [ 95% CI: 58; 74.5] and a negative predictive value of 55.7% [ 95% CI: 48.7; 62.5].

Additional study results

No additional results were obtained during the study.

DISCUSSION

Summary of the main study result

The analysis of geriatric predictors of CRNMB on anticoagulants in AF patients aged ≥80 years identified significant factors in all four domains of comprehensive geriatric assessment: social; mental and emotional; physical; functional.

The performed multiple-factor regression analysis provided a means to develop a prognostic CRNMB risk model. The obtained coefficients were used to develop a prognostic nomogram to estimate the probability of CRNMB in patients receiving DOACs. Significant prognostic factors included the number of household members, Barthel ADL score, Lawton IADL score, Timed Up and Go Test, and use of assistive devices. The developed model was characterized by a prognostic accuracy of 59.9%, a sensitivity of 48.9%, and a specificity of 72.4%. The positive predictive value amounted to 66.7%; the negative predictive value, 55.7%.

Research limitations

The main limitations of the study include the small sample size and the associated lack of statistical power. Also, this observational study of real clinical practice does not provide a means to assess patient compliance with regard to DOAC therapy and the presence of CRNMB in frail patients (or patients with frailty); however, it reflects the actual health care practice.

Interpretation of the study results

In the social domain, the study showed that it was more common for the CRNMB patients to live in a large household and to attend church. High social engagement contributes to CRNMB detection: friends, coworkers, and relatives may notice the signs of CRNMB and pay more attention to them than the patients themselves, which makes the patient seek the help of specialists who can diagnose CRNMB.

The assessment of mental and emotional status showed that the patients with CRNMB were not prone to depression and had lower scores on the GDS-15 and GDS-30 scales. This can be attributed to the fact that depressed patients may be less likely to complain, including about CRNMB, and as a consequence, may not mention something relevant at a doctor’s appointment [3].

The assessment of functional and physical status revealed that the patients with CRNMB were more active in performing their daily living activities (high Lawton IADL score), more mobile (according to balance tests, i.e., tandem stance, 5-time chair stand test, and Timed Up and Go Test), and it was less common for them to use assistive devices, which could lead to collisions and falls, resulting in CRNMB. In addition, it was less common for the CRNMB patients to have frailty according to the Age is No Barrier questionnaire, and they were less frail according to the frailty index used in the SPRINT study [19][20].

Thus, the present study showed that the group of CRNMB patients included fewer frail patients and those in need of assistance since in actual clinical practice, patients with frailty are less likely to seek the help of specialists despite the presence of CRNMB. On the other hand, self-reliant patients are under less supervision, which may lead to injuries and the development of CRNMB.

The findings differ from the results of other studies [7–11]. For example, Brook et. al [7] presented a retrospective evaluation of 658 AF patients with a high fall risk assessed using the Northern Hospital modified STRATIFY tool [21][22]. The overall median age of patients was 75 [31–96] years, which was comparable for all three DOACs (78 years for apixaban, 72 years for rivaroxaban, and 76 years for dabigatran etexilate), with an even distribution by sex One of the study endpoints was clinically significant bleeding that scored three or more points on the 2010 ISTH (International Society of Thrombosis and Haemostasis) bleeding scale [23]. The analysis showed that a high fall risk constitutes a risk factor for clinically significant bleeding (p = 0.032).

The present study examined several parameters that determine the risk of falls: the Timed Up and Go Test, falls in the previous year, and fall-related injuries on the Age is No barrier scale. The study found that none of these parameters had a significant effect on the risk of developing CRNMB. The difference in results can be attributed to the study design, the difference in the age of the examined patients, and the older age of the patients (Me 84 [ 82; 88] years vs Me 75 [ 31; 96] years). Another reason is the difference in the criteria for the definition of bleeding: this study used the 2015 ISTH definition of CRNMB, whereas the article discussed above studied clinically significant bleeding according to the 2010 ISTH scale, which required medical intervention and the use of tranexamic acid, surgical hemostasis, transfusion, infusion of clotting factor concentrates and desmopressin, i.e., major bleeding as defined by the 2015 ISTH scale.

A number of studies examined the effect of frailty on the risk of bleeding. The retrospective study by Yamamoto et al. [8] included 240 AF patients (mean age of 76.1 ± 10.0 years) on DOACs, with women accounting for 42.9%. The patients were divided into frail (5–9 points) or fit (1–4 points) as per the 2005 Rockwood scale [24]. Frail patients accounted for 50% (mean age of 81.2 ± 7.8 years) of the total number of patients; women, 55.8%. The study endpoint was the cumulative incidence of major or symptomatic bleeding in a critical area or organ according to the definitions established by the 2005 ISTH [25]. The study results showed that the cumulative incidence of bleeding was significantly higher in the frail patient group as compared to the fit patients (p < 0.001).

In [9], Søgaard et al. present the results of an observational cohort study examining the safety of DOACs as compared to warfarin in frail AF patients. The study included 32,048 patients who had not previously taken anticoagulants. The median age of patients receiving a standard DOAC dose was 75 (69; 81) years. Patient frailty was defined using the hospital frailty risk score (HFRS) [26], with all patients categorized as having “moderate” (5–15 points) or “high” (>15 points) frailty risk. Moderate frailty risk was exhibited by 87.7% of patients (14,138/16,122) receiving a standard DOAC dose, with 12.3% of patients (1984/16,122) having a high risk of frailty. With moderate frailty risk, major bleeding developed in 2.77% of patients (392/14,138) on a standard DOAC dose, and the relative risk of developing major bleeding on a standard DOAC dose amounted to 0.71 (95% CI 0.60–0.84). In the high frailty risk group, major bleeding developed in 3.53% (70/1984) of patients on a standard DOAC dose, and the relative risk for major bleeding was 0.62 (95% CI 0.41 to 0.92). A supporting analysis covering any bleeding events (including CRNMB and minor bleeding) leading to hospitalization showed comparable results. Although the incidence of major bleeding was slightly higher in the group of patients exhibiting high frailty risk than in the moderate frailty risk group, the relative risk of major bleeding was lower in the high frailty risk group.

The study by Candeloro et al. [10] reports on the association of frailty levels in AF patients with a higher risk of such adverse clinical events as clinically significant bleeding, including major bleeding and CRNMB. The patients were grouped as frail, pre-frail, and fit. Among the 236 patients included in the study (median age of 78 years, with women accounting for 44%), 156 (66%) patients had AF, and 80 (34%) patients had venous thromboembolism. Ninety-eight patients (41%) were frail, 115 patients (49%) were pre-frail, and 23 patients (10%) were fit. During a median observation period of 304 days, the incidence of clinically significant bleeding amounted to 20% in frail patients and 10% in pre-frail patients; no clinically significant bleeding occurred in fit patients.

In [11], Kim et al. present the results of a retrospective study on the effect of frailty on the relationship between DOACs (dabigatran, rivaroxaban, and apixaban) and major bleeding in AF patients as compared to warfarin. Frailty was determined using a deficit accumulation model (claims-based frailty index) [27]. The patients were divided into the following groups: fit at CFI < 0.15, pre-frail at CFI of 0.15–0.24, and frail at CFI ≥ 0.25. The mean age of patients was 76.4 ± 7.1 years in the dabigatran-warfarin cohort, 76.8 ± 7.3 years in the rivaroxaban-warfarin cohort, and 77.3 ± 7.4 years in the apixaban-warfarin cohort. The results of this study are presented as cases per 1000 person-years. The study results showed that the incidence of major, gastrointestinal, and intracranial bleeding increased with increasing frailty.

Unlike the studies described above, the present study assessed the frailty of patients using several scales: Age is No Barrier, Short Physical Performance Battery, 2010 Rockwood frailty index, and the SPRINT scale. It was shown that frailty, defined by these parameters, had no significant effect on the risk of developing CRNMB; in addition, the CRNMB patients were more fit according to the SPRINT frailty index. The difference in results can also be attributed to the study design, the age of patients, the criteria for the definition of bleeding, and the method for measuring target parameters.

CONCLUSION

This study revealed that socially active, emotionally stable, mobile, and fit patients have a higher risk of developing CRNMB. Conversely, frail patients may be less likely to experience CRNMB. Also, a lower incidence of CRNMB was linked to reduced social activity, depression, need for assistance, and use of assistive devices. The study identified geriatric factors that may serve as predictors of CRNMB in AF patients aged over 80 years on DOACs, which were used to develop a prognostic nomogram that can be used to estimate the probability of CRNMB. Significant prognostic factors included the number of household members, Barthel ADL score, Lawton IADL score, Timed Up and Go Test, and use of assistive devices. Given the active AI implementation in clinical practice, the obtained data can be integrated into clinical decision support systems. This would provide a means to assess the risk of developing CRNMB prior to its onset and select high-risk patients to continue work on the prevention of this adverse event.