2322 articles were identified, and 57 articles were selected for full text assessment from which six journal articles were included in this review (Figure 1). Six conference abstracts were excluded due to limited information available and poor quality of reporting. No articles in any language other than English were identified. The quality of the full text articles included was good: ≥6/9 for the four non-randomised experimental studies and 10/13 for the two randomised-controlled trials RCTs.
3.1. Participants characteristics
The total number of participants in all included studies was 553, while individual sample sizes ranging from 46 to 177 and mean participant age range from 79–85 years. The Canadian Clinical Frailty Score (CFS) was the most commonly used frailty measure (n=2)[33, 34], with others including: the Edmonton Frailty Scale , Identiﬁcation of Seniors At Risk (ISAR) , the Electronic Frailty Index (e-FI) , and Fried Frailty Phenotype .
3.2. Study characteristics and deprescribing interventions
This review includes two RCTs, two pre- and post- comparison studies and two prospective interventional cohort studies (Table 2). Studies were conducted in Ireland, Belgium, New Zealand, Canada, and Israel. Study settings included: hospital (3), care home (1), primary care (1), community (1). All included articles were published between 2014-2019.
Several tools and algorithms were used to guide deprescribing. Two studies used explicit criteria (lists of drug names targeted); one used only the STOPP criteria , one used the STOPPFrail tool . Two studies utilised implicit criteria (lists of evaluative questions or a process) developed by the research teams themselves including the Garfinkel algorithm for concurrent deprescribing of multiple medications , and guidelines for targeted deprescribing of anticholinergic and sedative medicines . The remaining two studies used a combination of both explicit criteria and implicit criteria; one used STOPP and Beer’s alongside pharmacist’s judgment , and the second used an algorithm based on STOPP guidelines, Beers criteria, Choose Wisely and Choose Wisely Canada . The medications most frequently deprescribed across the studies regardless of the setting were: benzodiazepines, antidepressants, neuroleptics, opiates, lipid-lowering agents (statins), vitamin and nutritional supplements, proton pump inhibitors, and cardiovascular drugs (aspirin, antiplatelets, b-blockers, digoxin).
Due to heterogeneity of the outcome measures and study designs, studies were grouped according to interventions: pharmacist-led deprescribing interventions (n=3) and multidisciplinary team-led intervention (n=3).
3.2.1. Pharmacist-led deprescribing
Three of the six studies described pharmacist-led deprescribing interventions: one in a care home , one in primary care  and one in hospital . The three studies were non-randomised experimental studies with a good quality scoring >6/9. Follow up periods were three months (after discharge from hospital), six months (in primary care) and twelve months (in care home).
The care home and primary care studies used a person-centred collaborative pharmacist-led deprescribing medication review with the General Practitioner (GP) [35, 37]. Both interventions involved a detailed medication review process that engaged patients and their relatives in decisions about medication discontinuation; both drafted and shared a medication plan and followed patients for close monitoring. While the hospital study was a prospective interventional cohort study that employed medication reviews by a Medication Rationalization (MERA) team led by pharmacists and involved physicians among frail inpatients (CFS ≥4) and compared them with 51 patients in the control arm who did not receive the MERA review . In one study, the pharmacist used implicit guidelines of sedatives and anticholinergic medicines developed by the research team among 46 residents identified to be frail using Edmonton Frailty Scale in three residential care facilities . Whereas the two studies in primary and secondary care, the pharmacists used the STOPP and Beers criteria to identify prescribing problems among 54 community dwelling older people with som degree of frailty using the electronic frailty index (e-FI) across six practices  and 53 frail inpatients (CFS ≥4 .
3.2.2. Multidisciplinary team (MDT)-led deprescribing
Three studies implemented multidisciplinary team-led deprescribing focused medication review: two were RCTs in hospital settings (high quality score of 10/13) and one was a longitudinal prospective interventional study in community (good quality, 6/9). Follow up periods were three and twelve months in hospital studies and three years in the community study.
One RCT of 146 patients (74 in the intervention group vs 71 in the control group) living with frailty (ISAR ≥2/6) implemented medication review using STOPP criteria on admission by an inpatient geriatric team consisting of nurses, geriatricians, a dietician, an occupational therapist, a physiotherapist, a speech therapist, and a psychologist . Another RCT of 130 inpatients (65 in the intervention group vs 65 in the control group) living with frailty (CFS score ≥ 7) used a STOPPFrail-guided deprescribing intervention by a research physician pre-discharge to care home . A longitudinal prospective nonrandomized study among 177 community dwelling older people living with frailty (median Fried Frailty Phenotype=3) employed concurrent deprescribing of multiple medications based on the Garfinkel algorithm by a geriatric consultant in collaboration with GPs . The intervention group included patients who had 3 or more medications deprescribed (Poly-De-Prescribing PDP group n=122), while the control group included participants who agreed to stop only 2 medication or less (n=55).
3.3. Outcomes of deprescribing
The outcomes reported in the included studies are presented in Table 2; safety of deprescribing in three studies; clinical outcomes (n=3); medication-related outcomes (n=6); feasibility of deprescribing (n=4); acceptability (n=2) and cost-related outcomes (n=2).
3.3.1. Primary outcome
Safety of deprescribing
Three studies reported the safety of deprescribing and its impact on adverse events [34, 35, 38]. Potential adverse drug reactions using the UKU-SERS score in a pharmacist-led deprescribing intervention among 46 care home residents decreased by a mean difference of 2.8 (95% CI; p < 0.05) after three months and 4.2 (95% CI; p < 0.05) after six months of deprescribing of sedative and antipsychotic medications . In addition, adverse effects of psychotropic medications decreased significantly by a mean difference of 1.8 (95%, CI; p < 0.05) three months after deprescribing, and by a mean difference of 2.24 (95%, CI; p < 0.05) after six months of deprescribing. One RCT in hospital reported that 88% of deprescribing recommendations based on STOPPFrail were accepted and implemented and no adverse events were reported of MDT-led deprescribing during three months follow-up .
Two MDT-led deprescribing studies in hospital and community showed no significant differences in unplanned hospitalisation and mortality [34, 38]. The RCT in hospital showed no statistically significant differences between the intervention and control groups for three months unscheduled hospital presentations [0.14 intervention vs 0.08 control, 95% CI, P=0.27] and mortality [ 0.18 vs 0.28, 95% CI; P=0.22]  . Similarly a longitudinal cohort study in community reported that the incidence of hospitalizations per patient per year (0.39 intervention vs 1.02 comparator, p= 0.1006) and survival (77% intervention vs 67% comparator group, p=0.026) was comparable between the groups after three years . This study also reported that the incidence of significant complications per patient/year was significantly reduced with deprescribing [0.22 intervention vs 1.72 comparator group, p= 0.0047] .
3.3.2. Secondary outcomes
Frailty and function: Two studies reported the outcomes of deprescribing on frailty and function [35, 38]. Pharmacist-led deprescribing sedatives and anticholinergic medicines among 46 care home residents showed a significant decrease in frailty scoring (mean difference of 1.35, 95% CI, P<0.05) using the Edmonton Frailty scale after six months of deprescribing. Another pharmacist-led deprescribing study did not report whether deprescribing has led to changes in frailty status but reported a positive and statistically significant correlation between number of PIMs (STOPP and Beers criteria) and frailty using e-FI (r = 0.280, P = .040) . The impact of deprescribing on functional status defined using a 5 point scale (1= independent, 2=frail, 3= mild disability, 4=disability, 5=severe disability) was examined in another MDT-led deprescribing study . Patients in the poly-deprescribing group had less functional deterioration compared to the comparator group [38(69.1%) vs 42(34.4%), P<0.001].
Falls: The impact on falls was mixed and reported in two studies [34, 35]. Significant decreases in falls rate, defined as the number of falls in the past 90 days, was reported after pharmacist-led deprescribing psychotropic medicines among care home residents . But on the other hand, falls risk (determined using an inhouse falls risk assessment tool utilised by most residential care facilities in New Zealand) remained the same six months after deprescribing. Another MDT-led deprescribing RCT study using STOPPFrail at hospital discharge reported no significant difference in incidence of falls [0.27 vs 0.30, 95% CI, P=0.75] and non-vertebral fractures [0.02 vs 0.09, 95% , P=0.18] among patients who moved to nursing home after three months of deprescribing .
Cognition, depression and mental status: two studies reported outcomes on cognition, depression and mental status [35, 38]. No change in cognition using the interRAI cognitive performance scale was observed after three and six months of pharmacist-led deprescribing antipsychotic medications among care home residents (mean difference of 0, p=0.26), however significant improvement of depression scores using the geriatric depression scale (GDS) (mean difference of -2, p<0.05) were seen after six months . Another MDT-led deprescribing study reported that patients in the poly-deprescribing (PDP) group had improvement in mental status using the Mini Mental State Examination (MMSE) test (3 comparator vs 63 intervention, p<0.0001) and cognitive status (0 comparator vs 7 intervention, P=0.0004) . These improvements occurred within three months after deprescribing in 83% and persisted for ⩾2 years in 68%.
Quality of life (QoL): Two studies assessed QoL of participants and showed no significant differences after implementing deprescribing interventions [34, 35]. QoL among care home residents was assessed using EQ-5D-3L and showed no significant different pre and six months after deprescribing . QoL in one RCT in hospital using QUALIDEM or ICECAP-O scores showed deterioration in both the intervention and the control group from baseline to three months follow-up, but no statistically significant differences were found in the mean change between groups from baseline to follow-up .
All six studies reported medication-related outcomes [33-38]. Four studies reported a significant reduction in the number of medications taken by patients living with frailty after implementing deprescribing, ranging from a mean of 2-3 medicines stopped per patient [33-35] across the different settings to unsurprisingly 7 medications per patient when poly-deprescribing of three or more drugs was implemented in people’s home . Two studies also reported significant decreases of potentially inappropriate medications associated with deprescribing interventions [36, 37]; for example the deprescribing interventions in two hospital studies reported a mean decrease in PIMs number of 2.2 (p<0.01) in a pharmacist-led study and a reduction in PIMs was twice as high for the intervention group compared to the control group in a MDT-led deprescribing intervention. One care home study reported a significant decrease in the drug burden index (by 0.34) six months after deprescribing in care home residents with pharmacist-led deprescribing intervention .
Feasibility of deprescribing
Four studies reported the feasibility of deprescribing among older people with frailty [33-35, 38]. They displayed that 72-91% of the suggestions to deprescribe medications made by either pharmacists or the MDT were accepted and implemented across the different settings [33-35, 38]. For example, in care homes, forty-five PIMs were identified and suggested to be stopped by pharmacists, of which 82% were agreed upon by the residents’ GP and 96% were agreed upon by the resident or their relatives/family resulting in implementing 72% of the recommendations . Similarly in two hospital studies, 72% and 81%of the recommendations made were accepted and implemented by the admitting physician and then patients [33, 34]. And in one community study, 91% of the recommendations made by a geriatrician were accepted by GPs .
Deprescribing was also reported to be well tolerated as most medications stopped were not restarted. For example, in care homes, medicines were re-prescribed by the GP in only five instances (15%); stopping medication was not completed in 13 residents (28%) due to mood changes, increased pain levels or overall health deterioration . Similarly in hospital, of the 162 medications that were stopped only 40 (25%) were restarted during hospital admission or at time of discharge and 81% of medications stopped during hospitalisation remained discontinued after three months . Another RCT study among hospitalised older patients discharged to care homes showed that only three medications stopped at discharge by the MDT were restarted .
Acceptability of deprescribing
Two studies evaluated the acceptability of their MDT-led deprescribing interventions and showed that patients and healthcare professionals were happy to stop unnecessary medication [33, 38]. For example, following a pharmacist-led intervention 87% participants felt comfortable stopping medications as recommended by the team and only a very small number found the experience stressful or confusing (5% and 11% respectively) . In the poly-deprescribing intervention in community, the overall satisfaction of patient/family from the changes was defined as high/very high in 89 % (n = 109) .
Two studies reported the cost implications of deprescribing [33, 34]. A pharmacist-led intervention reported a total saving of $1508.47 or $94.28 per 100 patient-days when STOPP criteria were implemented in hospital . Using STOPPFrail by MDT at discharge from hospital also led to a mean change in monthly medication cost of –$74.97 compared to –$13.22 in the control group (mean difference $61.74; 95% CI; P = .02) .