Abstract
Objective
To analyze the prevalence of potentially inappropriate medication use in older adults and develop a nomogram for predicting the individualized risk for potentially inappropriate medication use.
Methods
A retrospective cross-sectional study was conducted using prescription data from older adults who visited the Hefei Third People’s Hospital between May 2022 and May 2024. The 2019 Beers Criteria and Chinese criteria for Determining Potentially Inappropriate Medication Use in Older Adults in China were used to identify potentially inappropriate medication use. We conducted univariate and multivariate logistic regression analyses to identify the factors associated with potentially inappropriate medication use and developed a nomogram model to predict the individualized risk of potentially inappropriate medication use.
Results
Among the 475 older adults included, 195 (41.05%) had at least one incidence of potentially inappropriate medication use (total 288 occurrences). Medications considered as potentially inappropriate were most commonly used (88.72%), followed by medications to be used with caution (6.67%), potentially inappropriate drug–drug interactions (1.54%), and medications potentially inappropriate for patients with certain diseases or syndromes (2.05%). Benzodiazepines, rapid/short-acting insulin, proton pump inhibitors, and amitriptyline were the most frequently used potentially inappropriate medications. Independent risk factors for potentially inappropriate medication use included: (a) age ≥70 years; (b) diabetes mellitus; (c) hypertension; (d) coronary heart disease; (e) sleep disorders; (f) ≥3 comorbidities; and (g) use of ≥4 medications. The nomogram showed moderate discriminative ability (concordance index =0.738) with good calibration and minimal overfitting.
Conclusion
Advanced age, multiple chronic conditions, and polypharmacy are key predictors of potentially inappropriate medication in older adults. Enhanced monitoring and personalized medication management may help reduce the risk of potentially inappropriate medication use in this population.
Keywords
Introduction
Polypharmacy is a prevalent concern among older adults. As individuals age, their pharmacokinetic and pharmacodynamic characteristics undergo significant changes, often affecting the efficacy and safety of their medications. Potentially inappropriate medication (PIM) use represents a high-risk category where the dangers outweigh the expected clinical benefits. 1 This can not only jeopardize the health of older adults but also increase healthcare resource utilization and costs due to the heightened risks associated with medication use. Thus, it is essential to scientifically evaluate and prevent PIM use to improve medication safety in this population.2,3 To ensure contextually relevant and comprehensive PIM use assessment, we combined the internationally recognized 2019 Beers Criteria 4 (American Geriatrics Society) and the Chinese criteria for Determining Potentially Inappropriate Medication Use in Older Adults in China (2017 edition). 5 The Beers Criteria provides a standardized framework for cross-study comparisons, while the Chinese criteria enhances applicability to local clinical practice and disease spectrum. This dual approach aids clinicians pinpoint medications that should be avoided or used with caution in individuals aged ≥65 years, with the goal of minimizing adverse drug reactions (ADRs), drug–drug interactions (DDIs), and irrational drug use. Traditional studies often employ logistic regression to analyze factors contributing to PIM use, typically yielding group-level insights or risk associations that can be challenging to translate into individual patient risk prediction tools that can be used in clinical practice.
A nomogram is a graphical calculating tool that integrates multiple predictive variables to generate a simple, visual estimate of an individual’s probability of experiencing a specific outcome, such as PIM use. Its clinical value lies in its ability to transform complex statistical models into an intuitive, user-friendly format, enabling clinicians to quickly assess a patient’s personalized risk at the point of care. By providing a clear, numerical probability, nomograms can directly support prescribing decisions in routine practice; for example, if the nomogram indicates a high risk of PIM use for a specific older patient, the clinician might be prompted to conduct a more thorough medication review, consider safer therapeutic alternatives, or implement closer monitoring.
To address this issue, we designed the present study with the aim of developing an advanced nomogram model that can visually calculate the probability of PIM use for each patient, facilitating rapid clinical screening for those at high risk.
Methods
Study design and ethical approval
This retrospective cross-sectional study was conducted at Hefei Third People’s Hospital in accordance with the Declaration of Helsinki, 1975 (as revised in 2013). Approval was obtained from the Ethics Committee of Hefei Third People’s Hospital (Approval No.: 2024LLWL029). Due to the retrospective nature of the study, the requirement for informed consent was waived by the ethics committee. All patient data were anonymized and deidentified prior to analyses. The reporting of this study conforms to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines. 6
Data sources and participants
We retrospectively collected data from the Clinical Pharmacy Management System of Hefei Third People’s Hospital. The study population comprised outpatient prescriptions for patients aged ≥65 years who visited the hospital between May 2022 and May 2024. Inclusion criteria were as follows: (a) age ≥65 years and (b) prescriptions with complete diagnostic and medication information. Exclusion criteria were as follows: (a) unreasonable diagnoses and (b) inappropriate routes of administration. To ensure a representative sample, we employed a stratified sampling method based on common diseases, multimorbidity status, and typical medication patterns in the older adult population. In total, 475 prescriptions were included in the final analysis.
Assessment of PIM use
Two supervising pharmacists independently assessed the data using the 2019 American Geriatrics Society Beers Criteria and Chinese criteria for Determining Potentially Inappropriate Medication Use in Older Adults in China. The 2019 Beers Criteria categorize PIMs into five domains: (a) general PIMs; (b) medications potentially inappropriate for patients with certain diseases or syndromes; (c) medications to be used with caution; (d) medications associated with potentially inappropriate DDIs; and (e) medications requiring dosage adjustment based on renal function. The Chinese criteria were applied in parallel to ensure regionally appropriate and comprehensive PIM use identification. Any disagreements were resolved by a deputy chief pharmacist or a senior professional. Each case meeting either criterion was recorded as one PIM use occurrence; thus, a single patient could have multiple PIM uses. Patients were then divided into PIM use and non-PIM use groups according to whether PIM use was identified.
Statistical analyses
Research data were compiled in Microsoft Excel and imported into the Statistical Package for Social Sciences (SPSS) software (version 28.0) for analyses. Categorical variables were presented as frequencies and percentages and analyzed using the chi-square test. Continuous variables were reported as mean ± SD (x̄ ± s) values and compared using the independent-samples t-test. To identify risk factors for PIM, variables with p < 0.10 in univariate analysis were included in multivariate logistic regression analysis. Both crude and adjusted odds ratios (ORs) with 95% confidence intervals (CIs) were calculated. Potential confounding factors, particularly age and number of comorbidities, were included as covariates in the multivariate model to control for their effects. A nomogram model was developed using R software (version 3.5.3) with the rms package. The nomogram was constructed based on the final multivariate logistic regression model. Model performance was evaluated using the concordance index (C-index) and calibration plots. Internal validation was performed using bootstrapping with 1000 resamples. A p-value <0.05 was regarded statistically significant.
Results
Basic medication information and incidence of PIM
In total, 475 outpatient older adults were included in this study. Among these patients, 195 (41.05%) experienced at least one instance of PIM use, accounting for 288 occurrences in total. The distribution of PIM use across the five Beers Criteria categories was as follows:
Use of medications considered potentially inappropriate was most common, identified in 173 patients (88.72%). Medications to be used with caution were used by 13 patients (6.67%). Use of medicines that could induce inappropriate DDIs was identified in three patients (1.54%). Use of medications potentially inappropriate in patients with certain diseases or syndromes was identified in four patients (2.05%). Use of medications for which the dosages should be adjusted based on renal function was recorded for two patients (1.03%).
The most frequently implicated medications included benzodiazepines in 66 cases (33.84%), rapid- or short-acting insulin in 41 cases (21.03%), proton pump inhibitors in 28 cases (14.36%), and amitriptyline in 20 cases (10.26%).
Associated factors
The proportion of patients in the 65–69 years age group was significantly greater in the non-PIM use than in the PIM use group. Conversely, the proportions of patients aged 70–79 years and those aged ≥80 years were significantly greater in the PIM use group than in non-PIM use group. Additionally, a greater proportion of patients with diabetes mellitus, hypertension, coronary heart disease, and sleep disorders as well as those with ≥5 concurrent diseases and those taking ≥8 medications, was observed in the PIM use group compared with that in the non-PIM use group (p < 0.05). Detailed results have been presented in Table 1.
Baseline characteristics of patients with and without PIM use.
PIM: potentially inappropriate medication.
Logistic regression analysis of factors influencing PIM use
Binary logistic regression analysis was conducted with PIM use as the dependent variable. The independent variables included the eight factors that showed statistical significance in the univariate analysis, including age, diabetes, hypertension, coronary heart disease, renal insufficiency, sleep disorders, number of diseases, and number of medications. The results indicated that age, diabetes, hypertension, coronary heart disease, sleep disorders, number of diseases, and number of medications significantly influenced PIM use in older adults (p < 0.05) (Table 2).
Multivariate logistic regression analysis of factors influencing PIM use by older adults.
Assignment of variables: age (65–69 years = 0, 70–79 years = 1, and ≥80 years = 2); diabetes (Yes = 1 and No = 0); hypertension (Yes = 1 and No = 0); coronary heart disease (Yes = 1 and No = 0); renal insufficiency (Yes = 1 and No = 0); sleep disorders (Yes = 1 and No = 0); number of diseases (1–2 = 0, 3–4 = 1, and ≥5 = 2); and number of drugs (1–3 = 0, 4–7 = 1, and ≥8 = 2).
PIM: potentially inappropriate medication; OR: odds ratio; CI: confidence interval.
Construction of a nomogram-based risk prediction model
Nomogram development and performance
Based on the multivariate logistic regression analysis, a nomogram for predicting the risk of PIM use in older adults was constructed (Figure 1). The final model included the seven predictors of age, diabetes mellitus, hypertension, coronary heart disease, sleep disorders, number of diseases, and number of drugs. The regression coefficients, adjusted ORs, and their 95% CIs for each predictor are presented in Table 2. The model intercept (constant) was 0.105 (p = 0.768), corresponding to a baseline OR of 0.900.
Nomogram for predicting the risk of PIM use in older adults. The nomogram assigns a score to each predictor based on its contribution to the risk of PIM use. To use the nomogram, locate each patient’s characteristics on the corresponding axis, draw a vertical line to the “Points” scale to obtain individual points; add all the points to obtain the “Total Points,” and then draw a vertical line from the “Total Points” axis to the “risk of PIM” axis to estimate the individual probability of PIM use. Higher total points indicate a greater risk of PIM use. PIM: potentially inappropriate medication.
Model discrimination
The nomogram demonstrated moderate discriminative ability, with a C-index of 0.738 based on the area under the receiver operating characteristic curve (AUC) (Figure 2). This indicates that the model correctly distinguished between patients with and without PIM use in 73.8% of cases.
Calibration curve validation of the nomogram model.
Model calibration
Calibration of the nomogram was assessed by comparing the predicted probabilities with the observed outcomes. The calibration plot (Figure 3) showed good agreement between the nomogram-predicted probabilities and actual diagnosed PIM proportions. The apparent calibration curve was closely aligned with the ideal diagonal line, and the bias-corrected curve (derived from bootstrapping with 1000 resamples) confirmed minimal overfitting. Detailed calibration data are presented in Table 3.
ROC curve validation of the nomogram model. ROC: receiver operating characteristic.
Calibration data of the PIM risk nomogram.
PIM: potentially inappropriate medication.
Discussion
This study systematically evaluated prescriptions for older adults using the Beers Criteria (2019 version) and Chinese criteria for Determining Potentially Inappropriate Medication Use in Older Adults in China (2017 edition). The findings revealed an incidence rate of 41.05% for PIM use in the study population, consistent with previous studies, which have reported a PIM use prevalence of 30%–60% in various healthcare settings. For example, a study by Li et al. 7 conducted in a Chinese tertiary hospital reported a PIM use prevalence of 38.7%, while Al-Dahshan et al. 8 reported a rate of 44.2% in a primary care setting in Qatar. The relatively high proportion observed in our study may be closely associated with the prevalent issue of polypharmacy among older adults, which significantly increases the risk of ADRs.
Regarding the distribution of PIM use across Beers Criteria categories, use of medications considered potentially inappropriate was most prevalent (88.72%, 173 cases), followed by that of medications to be used with caution (6.67%, 13 cases), those associated with potentially inappropriate DDIs (1.54%, 3 cases), medications potentially inappropriate for patients with certain diseases or syndromes (2.05%, 4 cases), and medications for which the dosages should be adjusted based on renal function (1.03%, 2 cases). This distribution pattern is similar to that reported by Zhao et al., 9 who also found that medications considered potentially inappropriate were used very frequently, representing the most predominant category (82.3%) in a Chinese outpatient population. However, our findings differ from those reported by Sharma et al., 10 according to which, medications potentially inappropriate for patients with certain diseases or syndromes were used more commonly (18.7%), an inconsistency that may be attributed to differences in the study population’s comorbidity profiles and healthcare settings.
Regarding specific medications, benzodiazepines (33.84%), rapid- or short-acting insulins (21.03%), proton pump inhibitors (14.36%), and amitriptyline (10.26%) were the most frequently identified PIMs. These findings align with trends reported in multiple domestic and international studies. For instance, Li et al. 7 identified benzodiazepines and proton pump inhibitors as the primary drug categories associated with PIM use, which is consistent with our findings. Li et al. 11 have reported that clopidogrel, estazolam, and insulin are the three most frequently encountered PIMs, partially supporting our observation regarding insulin use. Su et al. 12 have reported that vasodilators, diuretics, and central nervous system medications are the most commonly used PIMs. Although our study did not identify vasodilators as a major category, the prominence of central nervous system medications (e.g. benzodiazepines) was consistent with previous reports. Shu et al. 13 have reported that proton pump inhibitors, benzodiazepines, and amiodarone are associated with higher rates of PIM use, closely mirroring our findings for the first two categories.
International comparisons reveal both similarities and differences. Al-Dahshan et al. 8 have reported that 84.2% of PIM use cases involve gastrointestinal medications, which contrasts with our finding of 14.36% for proton pump inhibitors alone, suggesting possible differences in prescribing patterns or classification criteria. Sharma et al. 10 have reported that 44.6% of PIM use cases are related to the use of endocrine drugs, supporting our observation regarding insulin use, though at a higher proportion. Bhagavathula et al. 14 have demonstrated that diuretics, insulin, amitriptyline, and aspirin are the most commonly used PIMs, which aligns with our findings for insulin and amitriptyline, but differs from our results regarding diuretics and aspirin use. Zhao et al. 9 have identified insulin, clopidogrel, and eszopiclone as the top three PIMs used, partially consistent with our findings regarding insulin use. Vatcharavongvan et al. 15 have reported that amitriptyline, dimenhydrinate, chlorpheniramine maleate, and lorazepam are the most commonly used PIMs, which supports our finding for amitriptyline but not for the other drugs.
These discrepancies may be attributed to factors such as variations in study population characteristics (e.g. age distribution and comorbidity profiles), healthcare settings (inpatient vs. outpatient and tertiary vs. primary care), drug availability and prescribing formularies across different countries and regions, and differences in the PIM use evaluation criteria used (e.g. Beers Criteria vs. Screening Tool of Older Persons’ Prescriptions (STOPP) criteria vs. Chinese criteria). Additionally, temporal factors may play a role, as prescribing patterns evolve over time and newer editions of screening tools may capture different medications compared with older versions.
This study identified several significant risk factors for PIM use, including age ≥70 years, diabetes, hypertension, coronary heart disease, sleep disorders, comorbidity burden of ≥3 diseases, and the use of ≥4 medications. First, age ≥70 years is recognized as an independent risk factor for PIM use. The aging process is associated with a decline in hepatic and renal functions, which reduces their capacity for drug metabolism and excretion, thereby increasing the risk of ADRs. Second, the presence of chronic conditions such as diabetes, hypertension, and coronary heart disease often necessitates polypharmacy, which elevates the risk of drug interactions and adverse reactions, consistent with findings from previous studies. 16 Sleep disorders contribute to PIM use through various pathways, including neurobiological changes, comorbid conditions, and behavioral and cognitive factors.17,18 The inappropriate use of sedative medications, such as benzodiazepines, further amplifies the risk of PIM use due to their long half-life and tendency to accumulate in older adults, increasing the likelihood of falls and cognitive decline. Although non-benzodiazepines have shorter half-lives, they may induce complex sleep behaviors, such as sleepwalking. Additionally, the presence of ≥3 diseases and the use of ≥4 medications are crucial influencing factors for PIM use.19,20 As the burden of disease increases, the necessity for additional medications rises, potentially leading to drug interactions, duplicate prescriptions, or overdoses. This risk is further heightened by inadequate collaboration among various medical departments. Moreover, patients’ use of over-the-counter (OTC) medications may also elevate the risk of PIM use, particularly when potential interactions exist between OTC and prescription drugs.
The nomogram, a visual prediction tool grounded in multifactor regression analysis, synthesizes the outcomes of logistic or Cox regression analyses to graphically represent the risk of specific clinical events in individual patients. 21 Using the identified risk factors, this study developed a PIM use risk prediction model with an AUC value of 0.738, demonstrating robust predictive performance. By incorporating seven risk factors and calculating a total score, the model can effectively predict the PIM use risk in older adults, providing a foundation for early clinical intervention. It is advisable for clinicians to thoroughly consider individualized risk factors when prescribing medicines, optimize medication regimens to mitigate PIM use, and concurrently enhance patient medication adherence.
This study has certain limitations. First, the model lacks external validation, which may limit its generalizability; second, the study did not encompass all drugs listed in the PIMs criteria (e.g. dihydroergotoxine mesylate and glibenclamide); third, factors such as patients’ educational level, medication adherence, and regional healthcare disparities were not comprehensively assessed; fourth, although the sample size of 475 met the basic requirements for model construction, it may be insufficient for complex predictive modeling. Future research should aim to address these limitations to further improve the PIM use risk assessment system.
Conclusion
This study suggests that PIM use exerts a significant negative impact on the health outcomes of older adults, possibly increasing the risk of ADRs and potentially elevating healthcare costs. Consequently, proactive pharmacist review of prescriptions for high-risk patients and the implementation of pharmacist-led prescription intervention systems may help reduce PIM use and enhance the safety and efficacy of drug therapy.
Footnotes
Acknowledgments
We would like to thank all the patients who participated in the study. Artificial intelligence (AI)–assisted tools were used for language polishing and grammatical improvement during the preparation of this manuscript.
Authors contributions
Weiwei Qi conceived and designed the study, drafted the manuscript, and revised the manuscript critically for important intellectual content. Wang Zhang performed the research, collected and curated the data, and assisted in manuscript preparation. Jiucui Tong contributed new methods or models, provided technical support, and revised the manuscript. Zhen Chen analyzed the data, interpreted the results, and approved the final version for publication. All authors have read and approved the final manuscript and agree to be accountable for all aspects of the work. All authors shall ensure that any and all questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.
Data availability statement
The datasets generated and analyzed during the current study are not publicly available due to reasons of patient privacy and ethical restrictions imposed by the institutional review board. However, deidentified data can be made available from the corresponding author upon reasonable request and with permission from the relevant ethics committee.
Declaration of conflicting interests
The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Disclosures
The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.
Ethical consideration
This study received approval from the Ethics Committee of Hefei Third People’s Hospital (Approval No.: 2024LLWL029), and the committee deemed informed consent unnecessary. The procedures used in this study adhere to the tenets of the Declaration of Helsinki (1975, as revised in 2024). This study is not a clinical trial, and therefore does not require registration.
Funding
This study was supported by the Health Science and Technology Project of Hefei Municipal Health Commission (Project No.: Hwk2025jyy007).
