Digital Health Interventions to Improve Medication Adherence Among Older Adults: A Systematic Review Authors: Kaleab Taye Haile, Wejdan Shahin, Eyob Alemayehu Gebreyohannes, Abdella Birhan Yabeyu, Amina Hareem, Mulugeta Hayelom Kalayou, Xun Yi, Kate Wang Published: 2026 ABSTRACT ABSTRACTObjectiveTo review evidence on the type, characteristics and effect of digital health interventions (DHIs) on medication adherence among older people.MethodsArticles were searched from inception to May 2025 in PubMed, Embase, CINAHL, Scopus and Web of Science. Randomised and non‐randomised studies were included if they: involved older people aged 65 years or older; applied any DHI(s); compared the intervention with a comparator group or baseline; and reported medication adherence as an outcome. Risk of bias was assessed using the Cochrane risk of bias tools, and certainty of evidence using GRADE (Grading of Recommendations, Assessment, Development and Evaluation). A narrative synthesis was conducted.ResultsA total of 35 articles were included, most of which were randomised controlled trials (n = 22). Risk of bias ranged from low to high, and certainty from very low to moderate. Nearly half of the studies (n = 17) reported DHIs improved medication adherence compared with control or baseline. Mobile apps (3/5), electronic reminders (3/3), social assistive robots (1/1), telenursing (1/1) and combined DHIs (2/2) showed the most promise. Interventions that used multiple functionalities or strategies to support behaviour change (reminders or prompts) were most likely to improve adherence. The other DHIs had mixed results or no significant effects.ConclusionsWhile findings across DHIs varied, interventions incorporating tailored reminders, multicomponent features and interactivity may have the potential to be more effective in improving medication adherence among older adults. Further research is needed to identify usage patterns and investigate the factors underlying differences in effectiveness.Keywords: aged 65 and older, digital health, digital health technology, medication adherence Introduction The global population is ageing, with the share of people aged 65 years or over continuing to rise, and expected to reach 1.4 billion in 2030 [1, 2, 3]. Ageing is accompanied by a sharp increase in chronic disease and comorbidities [1, 4]. Consequently, polypharmacy has become common, with about half of older adults taking five or more drugs, and around 10% taking 10 or more [5]. These medications are prescribed to reduce the risk of disease progression, comorbidity and mortality, which can be achieved only when patients adhere to them [6]. Medication adherence is a crucial part of disease management and treatment outcomes [7, 8], and is defined as ‘the degree to which patients follow their medication regimen directed by their health care provider’ [9](p696). Despite its importance, poor medication adherence is prevalent among older adults [10, 11], ranging from 26% to 59% [2, 5]. The contributing factors include patient, treatment and social factors [5, 12]. Poor adherence worsens patient outcomes, accelerates disease progression, increases hospitalisation rates and raises healthcare costs [13, 14, 15]. These challenges highlight the need for innovative, evidence‐based and targeted strategies to improve medication adherence among older adults [6]. Digital health interventions (DHIs) have emerged as promising tools to support medication adherence [8, 16]. Digital health interventions encompass a diverse range of information technologies, including mobile health (mHealth), telemedicine, wearable devices and other digital platforms [8, 17]. They range from simple methods to more comprehensive, integrated strategies that may utilise automation and artificial intelligence. Digital health interventions offer education, behavioural counselling, reminder systems or a combination of these, while also promoting patient engagement, self‐monitoring and communication with healthcare providers and/or caregivers [6, 17, 18]. Consequently, DHIs are well‐suited to address the complex, multifaceted nature of medication‐taking behaviour, typically neglected by traditional strategies (which often lack continuous monitoring, rely heavily on caregivers and are restricted by distance) [6, 13, 19, 20]. Numerous systematic reviews have examined the effectiveness of DHIs in improving medication adherence. Some evidence supports beneficial effects, whereas others report conflicting results; most reviews have focused on specific long‐term conditions [5, 8, 21, 22], or particular types of DHIs, such as mHealth [13] or interactive mHealth [6]. Among studies targeting older adults, evidence has been synthesised narratively rather than systematically [23, 24]. To date, strong and conclusive evidence focusing exclusively on older adults remains lacking. Evidence from broader populations may not apply to this group, whose needs are influenced by comorbidities, polypharmacy and age‐related barriers [2]. In response, this systematic review aimed to examine the evidence on the type, characteristics and effects of DHIs designed to support medication adherence among older adults. Methods Protocol and Registration This systematic review adhered to the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta‐Analyses) guidelines and was registered on PROSPERO, the International Prospective Register of Systematic Reviews (registration number CRD420251065096). Search Strategy and Selection Process Five electronic databases, PubMed, Embase, CINAHL, Scopus and Web of Science, were searched from inception to 14 May 2025. Search terms covered three concepts: ‘older people’, ‘digital health interventions’ and ‘medication adherence’, using keywords, synonyms, MeSH terms, Emtree and CINAHL subject headings combined with Boolean operators (File S1). Additionally, reference lists of related articles and reviews were hand searched. Only peer‐reviewed original articles in English were included. Citations were imported into Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia); duplicates were removed, and the screening process occurred in two stages. Two reviewers independently screened titles and abstracts, followed by full texts. Disagreements were resolved through discussion. Eligibility Criteria Studies were included according to the PICO (Population, Intervention, Comparison and Outcome) framework: Population: older adults aged 65 years or older, or in mixed‐age studies, when data for older adults could be extracted separately. Population: older adults aged 65 years or older, or in mixed‐age studies, when data for older adults could be extracted separately. Interventions: DHIs delivered via information and communication technologies aimed at improving medication adherence [8, 17]. Interventions: DHIs delivered via information and communication technologies aimed at improving medication adherence [8, 17]. Comparison: usual care, standard care, other active intervention/s (e.g., behavioural or educational intervention) or pre‐post assessments. Comparison: usual care, standard care, other active intervention/s (e.g., behavioural or educational intervention) or pre‐post assessments. Outcomes: medication adherence. Outcomes: medication adherence. Study design: randomised controlled trials (RCTs), quasi‐experimental, observational, pilot or longitudinal studies. Study design: randomised controlled trials (RCTs), quasi‐experimental, observational, pilot or longitudinal studies. Exclusion criteria were as follows: Studies including only adults younger than 65 years or with unclear age criteria. Studies including only adults younger than 65 years or with unclear age criteria. Studies in which digital health was not the central intervention, or only played a supportive role, to ensure the effect of the digital component on medication adherence could be meaningfully isolated. Studies in which digital health was not the central intervention, or only played a supportive role, to ensure the effect of the digital component on medication adherence could be meaningfully isolated. Studies without a comparator group. Studies without a comparator group. Study protocols, conference abstracts, opinion pieces, editorials, commentaries or reviews. Study protocols, conference abstracts, opinion pieces, editorials, commentaries or reviews. Data Extraction The data extraction was conducted using Covidence. A data extraction template was developed, reviewed by three researchers (KW, WS and EAG), and pilot‐tested by two reviewers. Subsequently, two reviewers independently extracted data from each article and reached consensus after discussion. Extracted information included: study details (first author, publication year and country); methodological characteristics (study design, study population and follow‐up duration); participant characteristics (sample size and patient age); intervention type, description and duration; adherence measurement method, corresponding adherence thresholds or cut‐off points and medication adherence outcomes. Medication adherence outcomes were extracted as reported in the studies, including changes from baseline following the intervention and differences between intervention and comparison groups using quantitative measures. For studies with multiple adherence measures, time points or analyses, all the results were extracted and reported. Risk of Bias (RoB) Assessment Cochrane's risk of bias tools were used to assess the methodological quality of the studies, specifically the RoB 2 tool was applied to RCTs, including the version for cluster‐randomised trials [25], and ROBINS‐I (version 2) was used to assess non‐randomised studies [26]. Two reviewers independently assessed each study, and disagreements were resolved through discussion. The Cochrane RoB 2 Excel tool was used to assess RCTs, whereas a modified Excel tool based on previous studies was applied to ROBINS‐I assessments [27]. A diagrammatic representation of the RoB evaluations, the traffic plot, was generated using the robvis Shiny App [28]. Certainty Assessment The certainty of the overall evidence was assessed using the GRADE (Grading of Recommendations, Assessment, Development and Evaluation) approach and classified as high, moderate, low and very low [29, 30]. One reviewer (KTH) initially conducted the GRADE assessment separately for RCTs and observational studies, and the ratings were then independently reviewed by three researchers. Data Synthesis Descriptive statistics were used to summarise quantitative data, and results were presented in tables and figures to facilitate comparison across studies. Where available, effect sizes (odds ratios, risk ratios, hazard ratios, mean difference, Cohen's d and regression coefficients) were extracted. For studies without explicit effect sizes, comparative statistics such as means and standard deviations (SD), median and interquartile ranges (IQR) (with or without p value or confidence intervals (CI)), χ2 statistics, t‐test results and F statistics were extracted and used to assess differences or associations between groups. Narrative synthesis was conducted to qualitatively summarise the findings and highlight variations in the effects of DHIs on medication adherence. Interventions were classified as combined DHIs when they incorporated two or more types of DHIs and as DHIs combined with non‐DHIs when they integrated with non‐digital methods. Additionally, each study's strategies (functionalities) were categorised using Lowe's taxonomy of interventions for consumer medicine use [31], selected for its practical application to medication adherence. Results Search Results The database search yielded a total of 11,993 articles. After removing duplicates and title and abstract screening, 417 articles were eligible for full‐text screening, and 34 studies met the eligibility criteria. One additional article was identified through a hand search, and ultimately, 35 studies were included in this review (Figure 1). PRISMA flow chart describing study identification and selection process. Study Characteristics The included studies were published between 1999 and 2024, with 13 articles published in the last 5 years [32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44]. Most studies were conducted in the United States (n = 15) [33, 34, 41, 42, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55], followed by studies in European countries (n = 9) [32, 35, 36, 39, 40, 56, 57, 58, 59]. More than half were RCTs (n = 22) [32, 36, 37, 38, 40, 43, 44, 45, 46, 48, 49, 50, 55, 57, 58, 59, 60, 61, 62, 63, 64, 65] whereas eight studies used a quasi‐experimental design [33, 34, 35, 47, 52, 53, 66, 67]. A total of 21,453 participants were included across 34 studies, with one additional study analysing 38,041 prescriptions. Gender was reported in 22 studies, with women comprising 55% (n = 4200) and men 45% (n = 3465) of participants. Intervention follow‐up duration ranged from 1 week [60] to 60 months [59], most commonly 12 months (n = 11) [39, 48, 49, 53, 55, 56, 57, 62, 64, 65] and 6 months (n = 5) [38, 44, 45, 47, 52], (Table 1). TABLE 1Summary of study characteristics.First author, year and countryStudy designStudy populationSample sizeAge eligibility criteria. Mean age (SD), yearsDuration (length) of follow‐upBasoglue et al. (2024) Turkey [30]RCTOlder people taking oral anticancer agents for at least one cycle Intervention = 23 Control = 25 65 years or older * Intervention‐ 72(7.9) Control‐ 70.5(5.3). 12 weeksYoon et al. (2024) South Korea [36]RCTPatients diagnosed with AF and prescribed Edoxaban. Intervention = 101 Control = 112 65 years or older* 6 monthsSolmaz et al. (2024) Turkey [34]RCTOlder people with hypertension who were taking at least one antihypertensive drug Training with a reminder wristwatch group = 30 Training group = 30 Control = 30 65 years or older12 weeksWang et al. (2024) Spain [35]RCTCHF patients 18 years or older on HF medication regimen. Intervention: 39 Control: 45 75 years or older* 52 weeksLemos et al. (2024) Taiwan [33]Quasi‐experimental (single group, pre‐post interventional study)Older people aged 65 years or olderSingle group = 26 65 years or older 71.7(5.4) 1 monthChen et al. (2024) United States [31]Quasi‐experimental (single group, pre‐post interventional study)Patients with schizophrenia who are prescribed long‐acting injectable.265 years or older* 2–12 weeksLee et al. (2024) United States [32]Quasi‐experimental (single group, pre‐post interventional study)Older adults (low‐income and solo living), willing to use a socially assistive robotSingle group = 30 65 years or older 82.6(6.36). 4 monthsMartini et al. (2023). Italy [38]Pilot RCTOlder people older than 70 years, taking at least four drugs Group—N = 19 Group—n = 19 Group—n = 18 70 years or older30 daysMohan et al. (2023) United States [39]Prospective cohort studyOlder people with diabetes and hypertension, prescribed ACEIs/ARBs and identified as nonadherent Intervention =247 Control = 480 65 years or older12 monthsGunderson et al. (2022) United States [40]Retrospective observational analysisPatients 18 years or older prescribed with hydroxymethyl glutaryl‐coenzyme A reductase inhibitors, renin‐angiotensin system antagonists and non‐insulin type 2 diabetes mellitus medications. ATC: 16,110 prescriptions ATM: 10,354 prescriptions Control: 11,577 prescriptions 65 years or older* 14 daysPoonprapai et al. (2022) Thailand [41]RCTOlder people diagnosed with type 2 diabetes and taking oral antihyperglycemic medications Intervention = 78 Control = 79 65 years or older Intervention—67.4(5.7) Control—67.8(6.2). 9 monthsLiang et al. (2021) Taiwan [42]RCTOlder people who were at high risk for readmission with a LACE index ≥ 7 s Intervention = 100 Control = 100 65 years or older Intervention—79.8(6.5) Control—81.51(7.91). 6 monthsQvist et al. (2020) Denmark [57]RCTOlder people diagnosed with AAA, PAD or HB Intervention = 725 Control = 721 65 years or older60 monthsRedfern et al. (2020) Australia [63]RCTPatients with moderate and high CVD risks Intervention = 292 Control = 301 65 years or older* 12 monthsIvers et al. (2020) Canada [60]RCTAdults diagnosed with MI and evidence of obstructive coronary artery disease. Mail and telephone calls—n = 450 Mail‐outs—n = 482 Control—n = 474 65 years or older* 12 monthsKim et al. (2019). South Korea [61]RCTOlder people with hypertension, taking antihypertensive medications, living in senior welfare centres. Health‐coaching group = 30 LMS group = 32 Health‐coaching and LMS group‐ n = 31 Control—n = 31 65 years or older Health‐coaching group —77.7(6.6), LMS group—77.2(6.7), Health‐coaching with LMS group—78.2(6.6), Control group—77.9(6.4) 8 weeksChoudhry et al. (2018) [46]RCTPatients with hypertension, hyperlipidaemia or diabetes, who had poor disease control and were nonadherent to prescribed therapy.Total number of participants = 132165 years or older* 12 monthsBellido et al. (2018) Spain [37]Descriptive observational studyAdults with type 2 diabetes initiating or switching to insulin glargine 300 U/mL (Toujeo)Total number of participants = 26065 years or older* 12 months (outcome was reported for 6 months follow‐up)Pandey et al. (2017) Canada [62]RCTPost‐MI patients aged 18 years or older, discharged from the hospital and enrolled in a structured rehabilitation program. Additionally, patients had to take antiplatelet, beta‐blockers and ACEIs/ARBs once daily. Intervention =7 Control = 8 65 years or older* 12 monthsHuang et al. (2017). Portugal [58]RCTOutpatients who were prescribed medication for 14 days Intervention = 145 Control = 58 65 years or older* 1 weekAbugosh et al. (2017). United States [43]Prospective cohort studyOlder people diagnosed with both diabetes mellitus and hypertension, taking ACEIs/ARBs and identified as nonadherent (6 months PDC < 0.8) Intervention = 495 Control = 248 65 years or older Intervention = 68.5(10.11) Control = 72(10). 6 monthsBiese et al. (2014) United States [44]RCTOlder people aged 65 years or older discharged from the emergency department Intervention = 39 Placebo = 35 Control = 46 65 years or older10 weeksMira et al. (2014). Spain [56]RCTOlder people with multimorbidity, taking multiple medications Intervention = 48 Control = 51 65 years or older Intervention = 72.9(6) Control = 70.9(8). 3 monthsKooy J., et al. (2013) Netherlands [55] RCT Older people who have been prescribed statins and were nonadherent with 365 PDC between 50% and 80% Counselling with ERD = 130 ERD = 123 Control = 128 65 years or older Counselling with ERD—73.3(6.6) ERD—73.2(5.8). Control—76.5(6.3). 12 monthsChrischilles et al. (2013) United States [47]RCTCommunity‐dwelling older adults aged 65 years or older who used a computer in the past month to visit a website or send/receive email and completed the baseline questionnaire. Intervention = 802 Control = 273 65 years or older6 monthsSolomon et al. (2012) United States [53]RCTOlder people receiving new prescription medication for osteoporosis Intervention = 1046 Control = 1041 65 years or older Intervention—77.8(6.4) Control—77.7(6.6). 12 monthsOwnby et al. (2012) United States [52]RCTOlder people older than 50 years with memory impairment. Tailored intervention = 8 Automated telephone call = 8 Control = 11 65 years or older* 24 monthsCastle et al. (2012) [45]Quasi‐experimental cohort interventional studyPatients with newly prescribed antidepressant medications.Total number of participants = 826165 years or older* 6 months Moczygemba et al. (2012) United States [51] Quasi‐experimental designOlder people who were 65 years or older Intervention = 60 Control = 60 65 years or older* Intervention—71.2(7.5) Control—73.9(8). 12 monthsKamimura et al. (2012) Japan [64]Quasi‐experimental (single group, pre‐post interventional study)Older people with a CDR scale of 0.5 or 1, and a history of missed medication dosesSingle group = 18 65 years or older 81.1(6.2). 3 monthsParker et al. (2012) New Zealand [65]Quasi‐experimental (single group, pre‐post interventional study)Older people 70 years or older taking three or more medications daily (≤ 5 dosing times), living independentlySingle group 31 participants70 years or older8 weeksHayes et al. (2009). United States [49]Three‐period within‐subject repeated‐measures designOlder people living independently and alone, managing their own medication. Selected from a prior cohort as poor adherers (missed > 20% of prescribed doses)Ten participants were enrolled in a single group and received each intervention sequentially65 years or older10.7 ± 4.4 weeks (no prompting), 10.1 ± 3.3 weeks (time‐based prompting) and 8.4 ± 4.2 weeks (context‐aware prompting)Antonicelli et al. (2008). Italy [54]RCTCHF patients 70 years or older with NYHA Class II, III and IV Intervention = 28 Control = 29 70 years or older12 monthsHuddleston et al. (2004) United States [50]Quasi‐experimental designOlder veterans with multiple chronic diseases Intervention = 243 Control = 1120 65 years or older12 monthsFulmer et al. (1999). United States [48]RCTOlder people with a primary or secondary diagnosis of CHF IG 1 = 17 IG 2 = 15 Control = 29 65 years or older IG1—73.1(6.5) IG2—76.2(8.8) Control = 73.7(5.3) 6 weeksAbbreviations: AAA, abdominal aortic aneurysm; ATC, automatic telephone call; ATM, automatic text message; CHF, congestive heart failure; CVD, cardiovascular disease; HP, high blood pressure; LACE index, Length of stay, acuity of admission, comorbidity and visits to ED; LAI's, long‐acting injectables; LMS, long message service; MI, myocardial Infarction; NYHA, New York heart association; RCT, randomised controlled trials.*Studies included mixed‐age participants (both younger and older than 65 years), but only data specific to participants aged 65 years or older were extracted separately. Summary of study characteristics. Intervention = 23 Control = 25 65 years or older * Intervention‐ 72(7.9) Control‐ 70.5(5.3). Intervention = 101 Control = 112 Training with a reminder wristwatch group = 30 Training group = 30 Control = 30 Intervention: 39 Control: 45 65 years or older 71.7(5.4) 65 years or older 82.6(6.36). Group—N = 19 Group—n = 19 Group—n = 18 Intervention =247 Control = 480 ATC: 16,110 prescriptions ATM: 10,354 prescriptions Control: 11,577 prescriptions Intervention = 78 Control = 79 65 years or older Intervention—67.4(5.7) Control—67.8(6.2). Intervention = 100 Control = 100 65 years or older Intervention—79.8(6.5) Control—81.51(7.91). Intervention = 725 Control = 721 Intervention = 292 Control = 301 Mail and telephone calls—n = 450 Mail‐outs—n = 482 Control—n = 474 Health‐coaching group = 30 LMS group = 32 Health‐coaching and LMS group‐ n = 31 Control—n = 31 65 years or older Health‐coaching group —77.7(6.6), LMS group—77.2(6.7), Health‐coaching with LMS group—78.2(6.6), Control group—77.9(6.4) Intervention =7 Control = 8 Intervention = 145 Control = 58 Intervention = 495 Control = 248 65 years or older Intervention = 68.5(10.11) Control = 72(10). Intervention = 39 Placebo = 35 Control = 46 Intervention = 48 Control = 51 65 years or older Intervention = 72.9(6) Control = 70.9(8). RCT Counselling with ERD = 130 ERD = 123 Control = 128 65 years or older Counselling with ERD—73.3(6.6) ERD—73.2(5.8). Control—76.5(6.3). Intervention = 802 Control = 273 Intervention = 1046 Control = 1041 65 years or older Intervention—77.8(6.4) Control—77.7(6.6). Tailored intervention = 8 Automated telephone call = 8 Control = 11 Moczygemba et al. (2012) United States [51] Intervention = 60 Control = 60 65 years or older* Intervention—71.2(7.5) Control—73.9(8). 65 years or older 81.1(6.2). Intervention = 28 Control = 29 Intervention = 243 Control = 1120 IG 1 = 17 IG 2 = 15 Control = 29 65 years or older IG1—73.1(6.5) IG2—76.2(8.8) Control = 73.7(5.3) Abbreviations: AAA, abdominal aortic aneurysm; ATC, automatic telephone call; ATM, automatic text message; CHF, congestive heart failure; CVD, cardiovascular disease; HP, high blood pressure; LACE index, Length of stay, acuity of admission, comorbidity and visits to ED; LAI's, long‐acting injectables; LMS, long message service; MI, myocardial Infarction; NYHA, New York heart association; RCT, randomised controlled trials. Studies included mixed‐age participants (both younger and older than 65 years), but only data specific to participants aged 65 years or older were extracted separately. Medication Adherence Measures Out of 35 studies included, nearly half (n = 18) used objective medication adherence measures, such as proportion of days covered (PDC) [37, 41, 45, 48, 57, 59, 62, 64, 65], pill count [38, 43], medication possession ratio (MPR) [53, 55], medication event monitoring system (MEMS) [50, 54], prescription abandonment [42], refill pattern [52], discontinuation [41, 45] and medication persistence [55]. The remaining studies (n = 17) reported subjective measures using standardised self‐reported questionnaires, such as the Morisky Medication Adherence Scale (MMAS 4‐item [49, 63] or 8‐item [40]), medication adherence rating scale (MARS) [34, 35], medication adherence behaviour scale (MABS—Chinese version) [44], oral chemotherapy adherence scale (OCAS) [32], self‐administration medication rate (SAMR) [66] and the hypertension compliance assessment scale (HCAS) [36]. Non‐standardised self‐reported measures included patient compliance with prescribed therapeutic regimens [56], medication delays and forgetting to take medication [60], the motivation questionnaire [39] and others [33, 46, 47, 67]. One study combined objective and subjective medication adherence measures [58]. Characteristics of DHIs and Effects on Medication Adherence A variety of DHIs were implemented to improve medication adherence among older adults. Most studies (n = 10) employed telephone‐based interventions (telephone calls through direct interactive calls, or automated calls, such as an interactive voice response system [IVR]) [41, 42, 45, 46, 47, 50, 53, 54, 55, 59], followed by five studies that utilised mobile or tablet applications [35, 37, 38, 43, 58]. Three studies employed text messages [33, 60, 64] or electronic devices [51, 66, 67], and two each explored telemonitoring [44, 56] and web‐based applications [49, 65]. Two studies combined DHIs [52, 63], and four integrated DHIs with non‐digital methods, such as counselling, training and informational mail [36, 40, 57, 62]. Additionally, telemedicine [39], telepharmacy [48], telenursing [32] and a socially assistive robot [34] were employed in one study, respectively (Figure 2). Types of digital health interventions in the included studies (n = 35). Among the included studies, 27 reported an effect size or tested for statistical significance to assess the effect of DHIs on medication adherence [33, 39, 40, 42, 47, 52, 53, 66], whereas eight reported results descriptively [33, 39, 40, 42, 47, 52, 53, 66]. Types of DHIs and Effects on Medication Adherence Mobile/Tablet Apps Five studies implemented mobile applications to improve medication adherence. DigiAdherence delivered tailored video education [35], whereas another app provided push alerts and vital‐sign prompts with optional data sharing with the physician [38]. Two studies employed interactive models: one used pharmacist‐moderated WeChat education and a reminder message [37], and the other used pharmacist‐led, family‐mediated education delivered through a mobile app [43]. In another study, a medication self‐management app, called ALICE, focused on reminders and medication information [58]. Three trials reported significant improvements in medication adherence: The ALICE app led to higher MMAS‐4 scores compared with controls (MD: 0.7; p < 0.001) [58]. Yoon et al. (2024) reported a higher proportion of participants achieving adequate adherence (≥ 95%) at 6 months in the app group than in standard care (81% vs. 59%; p = 0.001) [38]. Poonprapai et al. documented progressive gains in the intervention group compared with usual care at 3 (MD 1.41; 95% CI: 0.72, 2.09, p = 0.001), 6 (MD 2.27; 95% CI: 1.57, 2.96, p < 0.001) and 9 months (MD 2.94; 95% CI: 1.96, 3.91, p < 0.001) [43]. In contrast, no significant effects were observed for the tailored video app (baseline mean (SD); 8.1 (1.1)) vs. first assessment (8.5 (0.8), p = 0.46); first vs. second assessment (8.6 (0.9), p = 0.23) [35], and a pharmacist‐moderated social media app compared with usual care (RR = −0.06; 95% CI: −0.23 to 0.10) [37]. Electronic Devices Three studies evaluated electronic devices to support medication adherence among older adults. Two studies utilised time‐based audio‐visual reminders, including an automatic pill dispenser to deliver the correct dose [66] and a personal response unit medication reminder [67]. The third study assessed context‐aware reminders, triggered when a dose was due, had not been taken and was still safe to consume, and compared these with time‐based alerts and no reminders [51]. Context‐aware prompting outperformed fixed‐time reminders and no reminders (χ2 = 17.0; p < 0.001) [51]. A personal response unit reminder improved adherence from 52% to 81% (p = 0.01) [67]. The automatic pill dispenser increased the proportion of patients achieving a self‐administered medication rate > 80% at 1 and 3 months compared with baseline [66]. Text Messages/SMS Text messages were the DHIs in three studies, delivered to participants as reminders to take their medications. In two studies, the SMS reminders were sent daily [60, 64], whereas in one study, they were scheduled at specified intervals [33]. Results were mixed. One study reported a higher 12‐month PDC in the SMS group compared with usual care (91% vs. 73%; p = 0.01) [64]. Others found no significant group differences: for example, in Huang et al. (2017) for medication delays (RR = 0.60;95% CI: 0.31–1.14; p > 0.05) or missed doses (1.28 (0.28–5.78, p > 0.05)) [60]; and in Chen et al., data were extracted for two older adults, and a slight decline in compliance with long‐acting injectables was observed from baseline to post‐intervention (88% to 83%) [33]. Telephone‐Based Interventions Ten studies examined telephone‐based DHIs, most involved interactive calls (n = 7). Motivational interviewing delivered by pharmacists, pharmacy students or nurses was used in three studies to address adherence barriers and reinforce behaviour [41, 45, 55]. Three studies delivered telephone consultations: (i) pharmacist‐led medication therapy management (MTM) following electronic medical record (EMR) review [53]; (ii) nurse follow‐up to resolve medication issues [46] and (iii) nurse‐provided counselling on refills and side effects [59]. Three studies used reminder/prompting calls (one interactive via videophone/telephone [50] and two automated IVR [47, 54]). Gunderson et al. compared automated telephone calls (ATC) with automated text messages (ATM) and no intervention for prescription pickup [42]. Evidence for telephone‐based interventions was mixed. Motivational interviewing approaches showed some benefit: Abughosh et al. reported higher PDC among call recipients compared with non‐recipients (0.67 vs. 0.56; p < 0.001) [45], whereas Mohan et al. observed no effect at 6 months (OR 1.31; 95% CI 0.94–1.83, p = 0.10) but a significant improvement at 12 months (OR 1.46; 95% CI 1.15–2.04, p = 0.02) [41]. Solomon et al. found no overall difference in MRP between motivational interviewing calls and informational mail (49% vs. 41%, p = 0.07), although participants aged 65–74 years benefited (48% vs. 31%, p = 0.045) [55]. Fulmer et al. reported significant improvements in telephone and videophone reminder groups compared with usual care from baseline to post‐intervention (F[2.34] = 4.08, p < 0.05) [50]. In contrast, simpler or automated telephone approaches were less effective: Qvist et al. found no significant difference in composite PDC at 6 months (mean [SD]) (telephone counselling; 52.8 (45.2) vs. no counselling; 48.9(45.3), p = 0.20) [59]; Biese et al. reported comparable filled prescription rates (%) across groups (telephone call: 96%, placebo: 94%, no call: 94%; p = 0.96) [46]; and Moczygemba et al. found no difference in post‐intervention MRP(mean [SD]) (telephone MTM plus EMR review: 0.51[0.18] vs. EMR review: 0.57[0.17]) [53]. Automated calls had minimal impact: Gunderson et al. reported slightly lower prescription abandonment with automated calls (2.1%) compared with automated text messages (2.5%) and no intervention (2.3%) [42], and Castle et al., using an IVR system, yielded no meaningful effect [47]. Another study comparing automated calls (mean = 65; SE = 5.6; 95% CI: 53.5–76.0) with tailored information (82; 3.7, 95% CI: 74.5–89.2, p = 0.03) reported lower adherence, though adherence remained higher than participants attending regular monthly visits (60; 6.1, 48.1–72.0, p = 0.48) [54]. Web‐Based Applications Two studies implemented web applications; in the CONNECT study, the application was integrated into the electronic health records (EHR) and offered tailored lifestyle and medication adherence advice, monitoring and reminders, along with other services [65]. In the other study, the Iowa personal health record (PHR) allowed users to track current and past medications and health events [49]. Neither study found a significant difference compared with usual care [49, 65]. In CONNECT, the proportion of patients with more than 80% days covered was 34% in the web group versus 32% in the usual care group. In the Iowa PHR study, modified Morisky medication adherence scores were similar between the Iowa PHR and usual care groups (18.1 ± 1.9 vs. 13.9 ± 1.9; p = 0.98). Telemonitoring, Telemedicine, Telenursing and Telepharmacy Telemonitoring was evaluated in two studies. In a study by Antonicelli et al., home telemonitoring was delivered by a specialised CHF team through weekly telephone contact to collect health and medication‐related information and adjust therapy as necessary [56]. In the other study, nurse‐led telemonitoring involved continuously monitoring vital signs and symptoms via a smartphone [44]. In the telenursing study, health education was provided using WhatsApp and telephone calls [32]. In the study that employed telepharmacy, the clinical pharmacist provided a tailored telephone consultation based on the motivational interviewing principle to identify barriers to medication adherence and disease control and developed an improvement strategy [48]. Specialist‐led home telemonitoring improved medication compliance compared with standard care (26 vs. 10, p < 0.03) [56]. In contrast, nurse‐led continuous telemonitoring showed no significant difference in adherence relative to home visit during follow‐up (at 3 months: β = 0.64, SE = 0.42, p = 0.13; at 6 months: β = −0.08, SE = 0.28, p = 0.77) [44]. Telenursing education was associated with significantly higher adherence scores than usual care (F = 37.09, p < 0.001) [32], whereas telepharmacy produced statistically significant but small improvements in mean PDC compared with usual care(adjusted absolute difference (95% CI): 4 (1.5–6.4)) [48]. Robots In Lee et al.'s study, socially assistive robots called Hyodol embedded with sensors and artificial intelligence were used to support daily lives and health needs, including medication reminders, cognitive stimulation and emotional companionship. Socially assistive robots significantly changed MARS scores from baseline to post‐intervention (t(−0.51), df (29), p < 0.001, Cohen's d = 0.82) [34]. Combined DHIs Digital health interventions were combined in two studies. Kim paired nurse‐led phone‐based coaching with long message service for health coaching and compared this combined intervention with each modality stand‐alone and with usual care [63]. A study by Huddleston et al. integrated a home dialogue box (Health Buddy) with a web‐based risk management application to challenge patients through asking a series of health‐related questions and using their responses to monitor their health and identify individuals at risk [52]. Combining modalities showed promise. Huddleston et al. reported adherence rising from 64% at baseline to 90% post‐intervention with the use of an automated phone and web‐based platform (the usual care outcome was not reported) [52]. Kim found that nurse phone‐coaching plus long message service produced higher post‐intervention MMAS as compared to the interventions stand‐alone and usual care (F = 16.21; p < 0.001) [63]. DHIs Combined With Non‐DHIs Four studies integrated DHIs with non‐digital methods. In a multicomponent study by Solmaz et al. (2024), there were two experimental groups: one combined face‐to‐face training with telephone counselling (T) and the other added a medication wrist reminder (W), both compared with no intervention [36]. In a study from Martini et al. an electronic device (pill box, with sound and LED reminders), followed by SMS if doses were missed, was integrated with counselling and compared with counselling alone and a control group [40]. Ivers et al. assessed automated telephone calls paired with mailed booklets versus mail alone and usual care [62], whereas Kooy et al. examined an electronic reminder device (ERD) alongside pharmacist counselling and compared it with ERD alone and usual care [57]. Solmaz et al. reported significantly better antihypertensives compliance in both experimental groups compared with no intervention group (W: 4.7; T: 7.1; C: 9.1, p < 0.001) [36]. Martini et al. found high adherence in both the electronic pillbox plus counselling group (100%) and counselling‐only group (88%), with relatively lower MMAS for the usual care group (90%) [40]. In contrast, Kooy et al. reported no significant differences in medication adherence (PDC 365, PDC ≥ 80%, complete discontinuation; p > 0.05) between ERD plus counselling, ERD alone and usual care [57]. Ivers et al. observed no differences in composite PDC ≥ 80% between mail‐outs plus automated calls and mail‐outs alone (0.99(0.75–1.30), p = 0.95) [62]. (File S2). Intervention Strategies The studies included in this review employed either a single or combined strategy to improve medication adherence, and were categorised according to Lowe's taxonomy [31]. Most studies (n = 14) combined two or more functionalities, commonly informing and educating, with supporting behaviour change strategies (reminders) (n = 8) [36, 37, 40, 43, 57, 58, 63, 65]. Additional combined strategies included minimising harm and teaching skills (n = 2) [52, 56]; informing, educating and providing support (n = 1) [62]; supporting and facilitating communication and shared decision‐making (n = 2) [34, 48]; and supporting behaviour change and providing support (n = 1) [34]. Among single strategy interventions, the majority targeted to support behaviour change, primarily using reminder systems (n = 10) [33, 38, 42, 50, 51, 54, 60, 64, 66, 67]. Five focused on information and education delivery [35, 39, 47, 49, 59], four aimed to support older adults through assistance, encouragement and monitoring [41, 45, 46, 55], and one study each targeted healthcare quality improvement [53] and teaching skills [44] as methods to enhance medication adherence. Regarding the effectiveness of intervention strategies, DHIs that supported behaviour change, specifically through reminders and prompts (6/10) [38, 50, 51, 64, 66, 67] demonstrated the most consistent improvements in medication adherence, followed by combined or multiple approaches (8/14) [32, 34, 36, 43, 52, 56, 58, 63]. Whereas education‐only strategies showed limited effectiveness (0/5) [39, 47, 49, 59, 68]. (Table 2). TABLE 2Digital health intervention strategies categorised by Lowe's taxonomy of consumer medicine use.Lowe's taxonomyFirst author & yearType of DHIs: purposeStatistical significanceTo inform and educateLemos et al. (2024) [33]Mobile app (Digiadherence): health‐related information−Castle et al. (2012) [45]Telephone call (IVR): antidepressant educationNon‐negligible difference (SS not determined)Chrischilles et al. (2013) [47]Web‐based app (Iowa PHR): health and medication information tracking−Bellido et al. (2018) [37]Telemedicine: diabetes educationNon‐negligible difference (SS not determined)Qvist et al. (2020) [57]Telephone call: counselling and education on medication−To support behaviour change—patient reminders/reminder devicesYoon et al. (2024) [36]Mobile app: medication and vital‐sign reminder+Chen et al. (2024) [31]Text message: appointment reminderCompliance lower than baseline (SS not determined)Huang et al. (2017) [58]Text message: medication reminder−Pandey et al. (2017) [62]Text message: medication reminder+Parker et al. (2012) [65]Electronic device: medication reminder+Kamimura et al. (2012) [64]Electronic device: pill dispenser coupled with a medication reminderImproved from baseline (SS not determined)Hayes et al. (2009) [49]Electronic device: medication reminder+Gunderson et al. (2022) [40]Automated telephone call/automated text message: medication reminderNon‐negligible difference (SS not determined)Ownby et al. (2012) [52]Telephone call (automated): medication reminder+ (non‐DHI was effective)Fulmer et al. (1999) [48]Telephone/videophone call: medication reminder+To support—counselling, encouragement and motivational interviewingMohan et al. (2023) [39]Telephone call: motivational interviewingMixed effectAbugosh et al. (2017) [43]Telephone call: motivational interviewing+Biese et al. (2014) [44]Telephone call: discharge assistance−Solomon et al. (2012) [53]Telephone call: motivational interviewing, counselling and tailored problem‐solving−To teach skills—self‐monitoringLiang et al. (2021) [42]Telemonitoring: monitoring of vital signs and symptoms−To improve healthcare quality, collaborative care and a pharmaceutical care planMoczygemba et al. (2012) [51]Telephone call: Medication therapy management and medication adherence action planNon‐negligible difference (SS not determined)Combined strategiesWang et al. (2024) [35]Social media app: medication‐related education and medication reminder−Poonprapai et al. (2022) [41]Mobile app: family education, medication and appointment reminder+Mira et al. (2014) [56]Tablet app: medication information and medication reminder+Redfern et al. (2020) [63]Web app (CONNECT): health and medication‐related information and reminder alerts−Antonicelli et al. (2008) [54]Telemonitoring: minimise risk and self‐monitoring+Kim M (2019) [61]Phone/SMS coaching: education and motivation+Kooy et al. (2013) [55]ERD with counselling: medication reminder and counselling−Martini et al. (2023) [38]Pill box: reminder and auto‐dosing, and educationNon‐negligible difference (SS not determined)Lee et al. (2024) [32]Socially assistive robot: companionship and reminders+Huddleston et al. (2004) [50]Automated phone and web app: risk identification and monitoringImproved from the baseline (SS not determined)Solmaz et al. (2024) [34]Training and counselling with or without a reminder watch+Ivers et al. (2020) [60]IVR call with booklet: encouragement and health‐related information−Choudhry et al. (2018) [46]Telepharmacy: motivational interviewing and consultation−Basoglue et al. (2024) [30]Telenursing: education and encouragement+ Note: + Statistically significant; − no significant difference; SS, statistical significance (for the studies where the effect of size was not determined). Digital health intervention strategies categorised by Lowe's taxonomy of consumer medicine use. Note: + Statistically significant; − no significant difference; SS, statistical significance (for the studies where the effect of size was not determined). Risk of Bias Assessment Of the 21 parallel RCTs assessed using Cochrane's RoB 2, three articles were rated low risk [37, 62, 65], 10 had some concerns [32, 43, 46, 49, 54, 55, 57, 58, 59, 63] and eight were considered high risk [36, 38, 40, 44, 50, 56, 60, 64]. A significant number of studies (n = 14) did not describe allocation concealment or lacked a pre‐specified statistical analysis plan because of the absence of a pre‐registered protocol or failure to provide the necessary information in the clinical trial registration [36, 38, 43, 44, 49, 50, 54, 55, 56, 57, 58, 60, 63, 64]. Six studies were judged to be at high risk because of self‐reported measures of medication adherence coupled with the unblinding of study participants and/or outcome assessors [36, 40, 44, 56, 60, 64]. (Figure 3). Risk of bias summary for RCT studies. A single study, a cluster RCT by Choudhry et al., had a high risk of bias due to post‐cluster allocation identification and recruitment of participants [48]. (Figure 4). Risk of bias assessment for cluster RCT study. From 13 non‐randomised studies, nine were judged a serious risk of bias [33, 34, 35, 39, 42, 47, 51, 66, 67], three moderate risk [41, 45, 53] and one critical [52]. The main concern was the lack of reported measures or attempts to control confounding (n = 11) [33, 34, 35, 39, 40, 41, 42, 47, 51, 52, 53, 66, 67]. (Figure 5). Summary of risk of bias assessment for non‐randomised studies. Certainty of Evidence Based on the GRADE assessment, the certainty of evidence across the included studies ranged from very low [33, 34, 35, 39, 41, 42, 45, 47, 51, 52, 53, 66, 67] to moderate [32, 36, 37, 38, 40, 43, 44, 46, 48, 49, 50, 54, 55, 56, 57, 58, 59, 60, 62, 63, 64, 65]. (Table 3 and File S3). TABLE 3Summary of findings GRADE assessment: Digital health interventions compared with usual care/standard care/another active intervention/before and after the study for medication adherence among older adults.OutcomeEffectStudy designNumber of participants (studies)Risk of biasInconsistencyIndirectnessImprecisionPublication biasCertainty in the evidenceMedication adherenceHalf of the included studies showed positive effects, whereas the other half reported no effect.RCTs9857 (22 RCTs)Serious a Not seriousNot seriousNot seriousNot seriousModerate ⊕ ⊕ ⊕OObservational studies11,591 (12 observational studies)* Serious b Not seriousNot seriousNot seriousNot serious Very low ⊕OOO Note: Inconsistency: The heterogeneity of findings could largely be explained by methodological differences, including the types of digital health interventions, their design and components, the duration or follow‐up of the intervention, the medication adherence measures used, and variations in study design and sample size. Therefore, inconsistency was not judged serious. Indirectness: In 13 of 35 studies, mixed‐age populations were included, but data specific to older adults were extracted separately, and reflected the population of interest for this review [32, 33, 37, 38, 39, 42, 48, 53, 54, 60, 62, 64]. In six of 35 studies, more than one active intervention was included (e.g., combined DHIs and separate DHIs, DHIs integrated with non‐digital interventions, DHIs alone, non‐DHIs alone) alongside a control group, and the effects of the DHIs could still be meaningfully attributed to the outcome of interest [36, 40, 52, 57, 62, 63]. Even though various types of medication adherence measures were used, medication adherence was reported as an outcome in all studies. Therefore, concerns about indirectness were considered unlikely, and indirectness was not judged serious. Imprecision: The total number of participants is substantial, and there are studies with a precise effect size estimate. Although five RCT studies showed variability in effect size estimates, with confidence intervals spanning both no effect and potential benefit, or with wide standard deviations or interquartile ranges, and rated low for imprecision in individual assessment (File S3: Telephone‐based and text message interventions, respectively), their representation is small relative to the total number of studies (5/22) [37, 55, 59, 60, 62]. Imprecision was therefore not considered serious. Publication bias: No clear indications of publication bias were identified across the studies; both positive and negative results were reported in the published studies. a The RCTs, nine of 22 studies were assessed as high risk of bias, primarily due to bias in outcome measurement (6/9) [36, 40, 44, 50, 56, 60, 64], and 10 were assessed as having some concerns, mainly related to randomisation, selection or selective reporting [32, 43, 46, 49, 54, 55, 57, 59, 63]. b Among the observational studies, 10 of 13 were judged to have a serious risk of bias, mostly due to confounding [33, 34, 35, 39, 42, 47, 51, 52, 66, 67]. Therefore, the authors were downgraded by one level for risk of bias.*Participant numbers are based on 12 observational studies; one additional study analysed prescription‐level data only. Summary of findings GRADE assessment: Digital health interventions compared with usual care/standard care/another active intervention/before and after the study for medication adherence among older adults. Very low ⊕OOO Note: Inconsistency: The heterogeneity of findings could largely be explained by methodological differences, including the types of digital health interventions, their design and components, the duration or follow‐up of the intervention, the medication adherence measures used, and variations in study design and sample size. Therefore, inconsistency was not judged serious. Indirectness: In 13 of 35 studies, mixed‐age populations were included, but data specific to older adults were extracted separately, and reflected the population of interest for this review [32, 33, 37, 38, 39, 42, 48, 53, 54, 60, 62, 64]. In six of 35 studies, more than one active intervention was included (e.g., combined DHIs and separate DHIs, DHIs integrated with non‐digital interventions, DHIs alone, non‐DHIs alone) alongside a control group, and the effects of the DHIs could still be meaningfully attributed to the outcome of interest [36, 40, 52, 57, 62, 63]. Even though various types of medication adherence measures were used, medication adherence was reported as an outcome in all studies. Therefore, concerns about indirectness were considered unlikely, and indirectness was not judged serious. Imprecision: The total number of participants is substantial, and there are studies with a precise effect size estimate. Although five RCT studies showed variability in effect size estimates, with confidence intervals spanning both no effect and potential benefit, or with wide standard deviations or interquartile ranges, and rated low for imprecision in individual assessment (File S3: Telephone‐based and text message interventions, respectively), their representation is small relative to the total number of studies (5/22) [37, 55, 59, 60, 62]. Imprecision was therefore not considered serious. Publication bias: No clear indications of publication bias were identified across the studies; both positive and negative results were reported in the published studies. The RCTs, nine of 22 studies were assessed as high risk of bias, primarily due to bias in outcome measurement (6/9) [36, 40, 44, 50, 56, 60, 64], and 10 were assessed as having some concerns, mainly related to randomisation, selection or selective reporting [32, 43, 46, 49, 54, 55, 57, 59, 63]. Among the observational studies, 10 of 13 were judged to have a serious risk of bias, mostly due to confounding [33, 34, 35, 39, 42, 47, 51, 52, 66, 67]. Therefore, the authors were downgraded by one level for risk of bias. Participant numbers are based on 12 observational studies; one additional study analysed prescription‐level data only. Discussion This review examined how DHIs affect medication adherence in older adults. Various DHIs were used, either as standalone tools, combined with other digital tools or integrated with non‐digital methods, employing different strategies to support medication adherence. Overall, this review suggested that DHIs hold considerable promise in enhancing medication adherence among older adults. Approximately half of the studies reported positive effects, particularly those involving mobile apps, electronic devices, combined DHIs, telenursing, telepharmacy and socially assistive robots, consistently demonstrating beneficial outcomes. However, findings varied; 15 studies found no significant difference [33, 35, 37, 39, 42, 44, 46, 47, 49, 55, 57, 59, 60, 62, 65], one reported mixed result [41] and two favoured non‐DHIs or control conditions [53, 54]. These mixed outcomes may be partly due to differences in intervention types, adherence measures, study designs and comparison groups [6]. Additionally, issues like technology acceptance, adoption, sustained engagement and the ability to use technology‐assisted services among the older population, shaped by factors such as gender, socioeconomic status, education, residence and health status, may also influence effectiveness by creating unique patterns of digital inclusion and exclusion [69, 70]. The evidence reveals a clear pattern in the effectiveness of interventions related to their primary functionality. Providing medication information or health education alone through DHIs appears insufficient to improve medication adherence among older adults. A recent review on adherence challenges and strategies in older adults similarly concluded that education programs alone are inadequate [11]. This limited impact may be attributable to age‐related cognitive decline that affects information processing and retention, digital health literacy gaps and limited independent learning experiences among older adults [71]. In contrast, strategies supporting behaviour change through reminders to take medications were common and often effective. Interventions using electronic devices and mobile apps were successful, aligning with previous reviews showing that reminder‐based technologies help reduce forgetfulness [15, 24]. However, their effectiveness often declines over time as patients may tune them out, suggesting periodic updates may be needed to maintain efficacy [72]. Additionally, the design of reminder systems appears critical; simple, one‐way methods, such as automated calls or single texts, often fail to improve adherence. Conversely, context‐aware reminders that adapt to the user's behaviour and prompt only when a dose is due or missed tend to produce better adherence scores than basic time‐based alerts. This implies that personalised, adaptive interventions tailored to individual needs and preferences could be more successful [13]. Interventions focused on counselling, encouragement and motivational interviewing showed mixed but promising results. Interactive telephone calls using motivational interviewing and goal‐setting strategies showed positive impacts [45], whereas other telephone‐based counselling programs reported non‐negligible [46] or mixed effects [41]. This inconsistency indicates that the quality of interaction may be crucial, as patient‐provider communication, characterised by emotional understanding, motivational support and collaborative decision‐making, can improve attitude towards disease management and enhance older adults' sense of participation in their own care [20]. These findings also underscore that digital technology should facilitate patient‐provider engagement rather than merely replacing human interaction [73]. Multicomponent approaches were common and favourable, indicating many successful DHIs integrated two or more functions. Studies combining education with reminders [36, 43, 58] or layering education with motivational support consistently reported positive outcomes [32, 63]. These results appear to overcome the limitations of education‐only interventions (limited impact) and support the recommendation that educational strategies be paired with behaviour‐based modalities [11]. Overall, integrated models show stronger, more consistent benefits for older adults. Similar conclusions have emerged from reviews of patients on long‐term medications and in nonhospital settings, where multiple functional health technologies produced significant improvements [6, 37]. Non‐digital studies also found multimodal interventions effective [74]. This likely reflects the ability of multiple component programs to offer more comprehensive support, addressing the multifactorial and complex behavioural nature of nonadherence in older people [15, 19, 20, 75]. Another notable finding is that most DHIs (31/35) targeted the patient; one focused on a family caregiver, and three involved patients and caregivers. This pattern may reflect a shift from in‐person support to remote care, with older adults increasingly relying on self‐care supported by DHIs to adhere to treatment [24]. However, the success depends on their willingness and ability to adopt and engage with these technologies. This review has some limitations, and caution is advised when interpreting its findings. Grey literature and unpublished studies were excluded, and only English‐language articles were included, which may have led to the omission of relevant findings. Although careful consideration was given to classify the DHIs and their respective strategies, misclassification could have occurred due to the diverse components and settings of digital technologies. Heterogeneity in medication adherence measures prevented pooled estimates and restricted synthesis to narrative methods. The reliance on self‐reported medication adherence measures in almost half of the studies, some of which were not validated (n = 7), increases the potential for social desirability bias. Lastly, although evidence from RCTs provides moderate certainty, the overall confidence in the review findings is tempered by the very low certainty of observational studies. This can limit confidence in causal interpretations and affect the strength of the conclusions. Conclusions This review highlighted that DHIs varied widely in their ability to enhance medication adherence among older adults, and the available evidence did not identify a single clearly superior approach. Greater efficacy was more frequently associated with interventions that went beyond informational content alone, incorporating tailored reminders, multicomponent strategies and interactive elements such as goal setting. Interventions that shared information with clinicians and enabled follow‐up also appeared promising, although evidence was limited to fewer studies. Future research should more explicitly report patterns of use and investigate the factors behind differences in DHI success, including barriers and facilitators to their adoption. Long‐term evaluations and more high‐quality studies are also warranted. Funding The authors have nothing to report. Conflicts of Interest The authors declare that Associate Professor Kate Wang is an Associate Editor of the Australasian Journal on Ageing. No other conflicts of interest are declared. Supporting information File S1: ajag70206‐sup‐0001‐FileS1.docx. File S2: ajag70206‐sup‐0002‐FileS2.docx. Table 1. Characteristics of the DHIs, medication adherence measures, measurement time points and medication adherence results of the studies reviewed. File S3: ajag70206‐sup‐0003‐FileS3.docx. Data S1: PRISMA 2020 Checklist. Acknowledgements The authors thank Karen Macvean, librarian at RMIT University, for her assistance with the literature search, and Yu Cui, Piechang Yang and Zijun Li for their contribution to full‐text screening and data extraction. Open access publishing facilitated by RMIT University, as part of the Wiley ‐ RMIT University agreement via the Council of Australasian University Librarians. Data Availability Statement Data supporting this study are available in the article's Supporting Information. References Associated Data Supplementary Materials File S1: ajag70206‐sup‐0001‐FileS1.docx. File S2: ajag70206‐sup‐0002‐FileS2.docx. Table 1. Characteristics of the DHIs, medication adherence measures, measurement time points and medication adherence results of the studies reviewed. File S3: ajag70206‐sup‐0003‐FileS3.docx. Data S1: PRISMA 2020 Checklist. Data Availability Statement Data supporting this study are available in the article's Supporting Information.