Fall risk factors among poly-medicated older Lebanese patients in primary care settings: a secondary cross-sectional analysis of the “MGPIDP-L project”

This study is a secondary analytical cross-sectional study, designed to examine falls and their associated risk factors following a medication review. This study is part of the project entitled “Management of potentially inappropriate drug prescribing among Lebanese patients in primary care settings” (MGPIDP-L). The MGPIDP-L project aimed to assess potentially inappropriate drug prescribing (PIDPs) in Lebanese older patients in primary care settings (upon hospital discharge, at the patient’s home, and in community pharmacies).

Participants: Patients were systematically recruited over a 12-month period from a wide range of primary care settings, such as hospitals upon discharge, through home visits, and within community pharmacies. Contact was established with community and hospital pharmacies in Lebanon through conferences, phone calls, and mails to explain the study's rationale and methodology. Hospital pharmacists selected recently discharged patients meeting inclusion criteria, while community pharmacists displayed posters to encourage patient’s participation. Patients aged 65 years or older who were taking 5 or more medications per day, and who could provide all necessary medical information directly or with the assistance of a caregiver or family member in certain cases such as dementia were included. Patients not fulfilling the mentioned criteria were excluded.

During this project, a Medication review was conducted by two experienced clinical pharmacists over 12 months. Following the patient’s interview, an analysis of the patient’s medical treatment was performed. The analysis included the list of medications incorporating both those prescribed by the physician for the patient and any over-the-counter medications self-administered by the patient. Then, a formulation of justified pharmaceutical interventions was proposed for the patient’s general practitioner. After the physician’s approval, treatment modifications were communicated to the patient [28].

In the Original MGPIDP-L project, a sample size was calculated using Epi Info 7.2.5.0. For a total population size of around 600,000 Lebanese older adults, a confidence level of 99.9%, and a prevalence of 60% of inappropriate prescribing among the elderly population [31], 1083 patients were needed. Out of 1,038 patients involved in the study, 188 (18.11%) records were excluded due to incomplete, insufficient, or conflicting data. This left a sample of 850 patients enrolled with complete data during the 12-month inclusion period. In this study, we used this convenience sample for secondary analysis. Based on the following calculation: Using the G-Power software, version 3.0.10, this sample was sufficient to detect a calculated difference in fall rate of 0.0526, an alpha error of 5%, and a power of 80%. Given the expected fall rate of 0.05, 25 predictors could be included in the model.

The data collected directly from the patients included information on patients’ sociodemographic characteristics, chronic diseases, presence of geriatric syndromes including previous falls, medical treatment, and its management, tolerability, and compliance. The collected data were analyzed using a combination of both implicit and explicit approaches. The implicit criteria are based on the patient's global assessment while explicit criteria consist of predefined indicators of potentially inappropriate medication use in the geriatric population, established by expert consensus. To enhance the detection of potentially inappropriate drug prescriptions (PIDPs) and optimize medical prescribing, we utilized five distinct lists—Beers criteria, EU7-PIM, Laroche, STOPP/START, and STOPP/Frail [31‐35]. Given the variation in PIDP prevalence based on the criteria employed and considering the diverse origins of medications in the Lebanese market, we combined these lists to increase the likelihood of PIDP detection. Two pharmacists assessed each medication taken by the patient to determine if it matched any of the criteria listed in the five sets. They also evaluated the appropriateness of each medication based on the patient's clinical condition and determined whether a pharmaceutical intervention should be recommended to the general practitioner. In the current study, the medical prescriptions of the included subjects were also analyzed to detect the presence of FRIDs according to the Swedish National Board of Health. The Swedish National Board of Health and Welfare defined FRID classes and further classified them into drug categories that cause a high risk of falling, orthostatic hypotension, and/or systemic hypotension [24].

The variable of interest “the previous fall” was self-reported by the study participants. In the questionnaire, participants were asked whether they had fallen at least once in the past 3 months. The period of 3 months was chosen to avoid bias due to memory loss and to associate the fall with medication use recorded at baseline [36, 37]. These subjects were compared with participants who had not fallen to identify variables associated with increased fall risk.

Potential risk factors for falls were divided into 3 categories:

Age (young old (65–74); middle-old (75–84); old-old (≥ 85)), biological gender, weight, living arrangement (patient lives alone, with a spouse, or with family), residence area (North Lebanon, Beirut, Mount Lebanon, Beqaa and South), and type of assistance needed (nurse, housekeeper, physical therapy). The choice of these forms of assistance was influenced by the Lebanese context, where hiring a housekeeper was feasible for many individuals prior to the economic downturn. This served as a significant alternative to nursing homes, particularly for families with members who go out to work or are residing abroad. Registered nurses are typically required for older adults with specific healthcare needs. Additionally, in instances of falls, patients needing specialized medical attention are often hospitalized in tertiary care facilities due to the increased expenses associated with home-based nursing care and rehabilitation.

Existing medical conditions and/or geriatric syndromes [38] (Frailty, urine incontinence, insomnia, daytime drowsiness, involuntary weight loss, loss of appetite).

Total number of PIDPs/medical prescriptions, presence of fall risk-increasing drugs (FRIDs)/medical prescriptions, and FRID classes used according to the Swedish National Board of Health [24], dichotomous questions related to the patient’s behavior toward his medical treatment. These questions covered various aspects such as medical treatment management (e.g., pill organizer usage, medication pick up, self-dependent medication management), prescription of drugs (e.g., perception of medication quantity, adaptability to lifestyle), stock management (e.g., medication availability, excess, sharing), preparation and administration challenges, medication utility, side effects, treatment follow up, and self-medication tendencies.

SPSS version 26 was used to analyze the data. The dependent variable was the self-reported fall history (fallers vs. non-fallers). The chi-square test was used to assess bivariate associations between categorical variables and fall history. The student-independent t-test and Wilcoxon-Mann–Whitney tests were used when appropriate to compare means between fallers and non-fallers for continuous variables (weight, total number of medications, total number of diseases, total number of identified (PIDPs)/prescriptions, and total number of (FRIDs)/prescription) after checking for normality using the Kolmogorov–Smirnov test. Given the exploratory nature of our analysis, we opted not to adjust for multiple testing in the bivariate analyses as an initial step to identify variables that might warrant further investigation in subsequent multivariate analyses.

Multivariate logistic regression analysis was performed to identify independent factors associated with a fall in the past 3 months using a backward likelihood ratio method. Adjusted odds ratios (OR) with 95% confidence intervals (CI) were calculated. To mitigate the risk of chance findings due to multiple tests, we applied the Bonferroni correction which adjusts the significance threshold for each test to maintain an overall experiment-wise error rate. The Hosmer & Lemeshow test was used to ensure sample adequacy. The evaluation of collinearity in our analysis was conducted by computing Variance Inflation Factors (VIF) for each predictor variable incorporated in the different models. The main explanatory variables that were associated with fall in the bivariate analysis (p < 0.05) were included in the models to decrease potential confounding. Considering the case-to-independent variables (IVs) ratio of 40:1 [39], three consecutive regression models were conducted. Sociodemographic and clinical predictors variables from Table 1 were included in the first model. Medication-related variables from Tables 2, and 3 were introduced into the second model. The final model (model 3) was run with significantly associated variables from the previous two models as explanatory variables. The total number of drugs listed as FRID /prescription and the total number of PIDPs/prescription were also forced. In all cases, a P-value of less than 0.05, adjusted for the number of tests using the Bonferroni correction, was considered significant. This adjustment ensured a more conservative interpretation of statistical significance, guarding against inflated Type I error rates associated with multiple comparisons.

In total, four hospitals (located in Beirut, Mount Lebanon, North and South) and ten community pharmacies (two in each region) participated in the study. 850 patients were included from the MGPIDP-L project. Of those, 166 (19.5%) had experienced a recent fall in the past three months. No difference between fallers and non-fallers was found in terms of age, gender, living arrangement, and residence area in our sample. However, fallers relied significantly more on different types of assistance when compared to non-fallers: 26.5%(n = 44) of them were helped by a housekeeper versus (vs.) 19.3%(n = 132) of non-fallers, 6.6% (n = 11) needed assistance of a registered nurse vs. 3.3% (n = 21) of non-fallers, and 7.8%(n = 13) necessitated physiotherapy sessions compared to 2.8%(n = 19) of non-fallers, Table 1.

Geriatric syndromes were more prevalent among fallers. 26.5% (n = 44) were frail and 47%(78) were dependent according to the Find survey [40]. Insomnia had a high overall prevalence in fallers, thus leading to experiencing more daytime drowsiness. 61.4%(n = 102) and 67.5% (n = 112) experienced insomnia and daytime drowsiness compared to 37.5%(n = 257) and 45.5% (n = 311) in non-fallers. Urinary incontinence was also more common among fallers (35.2%(n = 58) vs. 20.7%(n = 142)). Loss of appetite was more frequent in fallers (48.2%(n = 80) vs.18.5%(n = 127)) as well as weight loss (39.8%(n = 66) vs. 20.3%(n = 139)), Table 1.

Our results showed that fallers suffered from a higher number of diseases than non-fallers. The most frequent diseases among fallers and non-fallers were hypertension, followed by diabetes and dyslipidemia. Osteoporosis (29.8% (n = 48). vs. 16.8%(n = 115)), depression (15.7%(n = 26) vs. 9.8%(n = 67)), Alzheimer’s disease or dementia (6.6% (n = 11) vs. 1.5%(n = 10)), and epilepsy (4.8%(n = 8) vs. 1.6%(n = 11)) were significantly higher in fallers than non-fallers, Table 1.

The multivariate analysis was done in three consecutive models as described in Table 4.

In the second model, the use of beta-blockers increased the risk of falls. Self-dependent medication management was negatively associated with falls. In contrast, patients who ran out of medication from time to time were more likely to fall. The adjusted P-value after the Bonferroni correction for model 2 was 0.0055.

The final model (Model 3) showed that having experienced a fall in the last three months was significantly associated with several risk factors considering an adjusted Bonferroni P-value of 0.0055. Loss of appetite increased the odds of falling by 3 (ORa = 3.020, CI [2.074–4.397]). The risk of fall almost tripled for dependent subjects (ORa = 2.877, CI [1.787–4.632]) and, it was doubled for patients suffering from drowsiness (ORa = 2.093, CI [1.440–3.043]). Among the FRIDs, using beta-blockers increased the odds of falling by 1.9 (ORa = 1.943, CI [1.339–2.820]), Table 4.