Reply to “From Evidence to Impact: Bridging the Implementation Gap in Geriatric Deprescribing”

We appreciate the recognition of Chen et al. [1] of the importance of our findings of associations between potentially inappropriate prescribing (PIP) and long-term mortality among community-dwelling older adults [2]. In their letter, Chen et al. [1] highlight the need to translate the growing body of evidence in this field into systematic deprescribing actions in clinical practice, for example by integrating advanced Clinical Decision Support Systems (CDSS) with electronic health records. The informed translation of observational findings into clinical practice mandates careful review of the scientific evidence. A recent review of 16 RCTs across diverse healthcare settings underscores the persistent gap between knowledge and practice: while CDSS can improve prescribing practices, their impact on clinical outcomes remains less conclusive. The review identified two key factors influencing the success of CDSS interventions, pharmacist involvement and adherence to the intervention protocol, and concluded that broader clinical outcomes may rely on complementary care processes or require longer follow-up to become evident [3]. These findings are consistent with the checklist developed by Armando et al. [4], which outlines key considerations for optimizing the design of CDSS evaluation studies, which also include employing a pre-post design, enrolling participants with complex morbidity but with adequate life expectancy for follow-up, and integrating a user-friendly CDSS platform with the facilities' electronic systems, with data-sharing capabilities across professionals. Importantly, as in many other studies in this field [3, 5], Chen et al. [1] focus primarily on deprescribing. However, our results [2] demonstrate that under-prescribing in older adults should not be overlooked. Incorporating tools such as the START criteria [6], which identify prescribing omissions, into future CDSS is crucial to ensure comprehensive and appropriate medication management and to improve long-term outcomes in this vulnerable population. Several concerns were also raised regarding the methodology of our investigation. First, as noted in our paper [2], misclassification of potential prescribing omissions (PPOs) is possible, since omitting certain medications may be clinically justified, and some of these PPOs may have been corrected during follow-up, diluting the strength of associations observed by us toward the null. Nevertheless, significant, most probably conservative, associations with mortality were still observed, despite these limitations. Misclassification could have also occurred for potentially inappropriate medications (PIMs), for instance, as noted with respect to renal function [1]. However, since such misclassification is expected to be nondifferential, this has likely biased results toward the null as well. Importantly, deprescribing after an adverse event occurred would not eliminate risk if the event already initiated downstream complications that contribute to later mortality. Indeed, future research should examine time-varying PIP and its dynamic effects on outcomes [2, 7]. Chen et al. [1] list several potential contributing factors to residual confounding in our analysis, such as frailty and cognitive decline. As noted in our paper [2], the cohort included relatively few cognitively impaired individuals, and our conclusions apply primarily to relatively healthy, community-dwelling older adults. Frailty was not directly measured, but 51.6% of our participants reported no leisure-time physical activity (pertaining to one of Fried's frailty criteria [8]), and 24.4% rated their health as “poor” or “very poor.” This subjective measure of health was shown to capture severity of both diagnosed and undiagnosed conditions [9]. Importantly, associations between PIP and mortality remained statistically significant even after adjusting for a comprehensive set of baseline health-related factors, including not only polypharmacy and chronic conditions, but also subjective health perception, lifestyle behaviors and apolipoprotein E genotype. Regarding socioeconomic status, the Israel Study of Glucose Intolerance, Obesity and Hypertension (GOH Study) is a nationwide cohort sampled from the Central Population Registry, designed to equally represent sex, ethnic origin and age groups among Jewish individuals born between 1912 and 1941 [10]. Most participants immigrated during Israel's early statehood (1940s–1950s), resulting in a socioeconomically diverse cohort that retained its heterogeneity over follow-up. In our study sample (third follow-up), 40.1% had fewer than 9 years of education and 43.8% held blue-collar occupations in their past [2]. All sociodemographic variables, including education, occupation, and ethnic origin, were accounted for in our multivariable analysis. We appreciate the opportunity to expand the discussion on the importance of effectively translating current evidence on adequacy of drug treatment in older adults into clinical practice. Such efforts should address both aspects of PIP, overprescribing and under-prescribing, be both realistic and practical, and consider not only the optimal development of CDSS but also the appropriate design of studies evaluating their implementation and effectiveness. All authors have read and approved of the submission of this manuscript. R.D. contributed to the acquisition of data. L.O., A.C., and R.D. contributed to the study concept and design. L.O., A.C., R.D., and K.L. contributed to the analysis and interpretation of data. L.O. and A.C. drafted the manuscript. L.O., A.C., R.D., and K.L. contributed to critical revisions of the manuscript for important intellectual content and gave final approval of the final draft. This work was supported by the Tel-Aviv University, Tel-Aviv, Israel (through the Albert and Alba Cuenca Institute for Treatment Approaches to Age Related Diseases). The sponsor had no influence on the design, data collection, analysis, or writing of this article. The authors declare no conflicts of interest. This publication is linked to a related editorial by Chen et al. To view this article, visit https://doi.org/10.1111/jgs.70207.