Sage Journals: Discover world-class research

Abstract

Background

Potentially inappropriate prescribing (PIP) is common in primary care and is associated with adverse outcomes. Knowledge of the risk factors of PIP can be critical in designing PIP interventions, especially in guiding our understanding on how PIP occurs in primary care.

Objective

This systematic review examined factors associated with PIP, specific to primary care.

Methods

We searched PubMed, Embase, CINAHL, Web of Science, Scopus and PsycINFO for studies related to ‘older persons’, ‘primary care’ and ‘inappropriate prescribing’. Two reviewers conducted study selection, data extraction and quality appraisal. Factors associated with PIP were narratively synthesized.

Results

Of the 1017 articles identified, we included 25 articles and a total of 2,893,925 participants, with average age of 70.4–84.0 years. Risk factors of PIP could be classified into patient, physician and system factors. Patient factors were related to patient demographics (advanced age, lower education level and lower socioeconomic status), medical comorbidities (polypharmacy and multimorbidity) and lifestyle factors (unhealthy habits and use of over-the-counter medications). Physician and system factors included older, male, solo general practitioner (GP), higher number of visits of pharmaceutical sales representatives to GP, centrally located GP practice, and smaller number of older patients following up with GP, and medication source from public health system.

Conclusions

The findings provide contextual information to guide our understanding of PIP in primary care. Factors identified in this review can inform the design of complex interventions for PIP, as well as be used to develop risk prediction tools to identify high-risk groups who may require further interventions related to PIP.

Keywords

Potentially inappropriate prescribing older person primary care risk factors systematic review

Introduction

Potentially inappropriate prescribing (PIP) refers to prescribing, or under-prescribing, of medications in older persons that may potentially cause significant harm.¹ It can be identified in clinical practice using either implicit (judgement-based) or explicit (criterion-based) tools.^1–3 Implicit tools refer to quality indicators that clinicians can apply to a prescription to judge the prescribing appropriateness and include tools such as the Medication Appropriateness Index.^1–3 Explicit tools comprise lists of drugs or drug classes (developed from literature reviews, expert opinion and consensus techniques), which when prescribed or under-prescribed can potentially cause harm in older persons. Such tools are more commonly adopted⁴ due to the ease of administration and include the Beers criteria, STOPP (Screening Tool of Older Persons’ potentially inappropriate Prescriptions) and START (Screening Tool to Alert to Right Treatment).^1–5

As shown in recent meta-analyses,^2,3 PIP is common in primary care setting (affecting one in three older persons in this setting)³ and is associated with 1.3–1.6 times higher risks of adverse outcomes (including emergency room visits, adverse drug events, functional decline, poorer quality of life and hospitalizations).² PIP was also shown to explain 7.7–17.3% of adverse outcomes related to older persons in primary care, with up to 79 cases of adverse outcomes (per 1000 adverse outcomes) in primary care being potentially preventable if the current prevalence of PIP is halved.³ The relationship between PIP and adverse outcomes may be related to the changes in pharmacokinetics and pharmacodynamics with rising age, where there is limited physiological reserve and poorer tolerability to inappropriate medication use.⁵ The issue of PIP in primary care is expected to accentuate in future, given the rapid ageing of populations around the world,⁶ the expected rise in the number of older persons with chronic diseases who require regular prescriptions from primary care,⁷ and the increasingly central role that primary care plays in coordinating care across healthcare sectors.^8,9 As such, efforts to address and prevent PIP can be especially pertinent in the primary care setting. In particular, the interventions for PIP need to be guided by the known risk factors of PIP, using these risk factors to inform our understanding of how PIP can occur in primary care as well as identify those who may require further assessment for the presence of PIP.

Previous reviews in the literature have reported various factors that may increase the risk of PIP, such as comorbidities, lower socioeconomic status, depression, cognitive impairment and lower functional status.¹⁰ However, these previous reviews have included a heterogeneous range of populations, with much of the focus on participants from specialized populations such as those from tertiary hospitals or nursing home settings. We are uncertain about the applicability of these findings to primary care settings, given that ambulatory patients in primary care can differ in their health profiles and disease comorbidities compared to those in tertiary hospitals and nursing homes. Therefore, we sought to conduct a systematic review to identify the risk factors of PIP that have been specifically reported in primary care settings, with the intention of using these risk factors to improve our understanding of PIP in primary care and guide future interventions to address PIP in this setting.

Methods

Search strategy and selection criteria

This systematic review was developed in accordance with the Preferred Reporting Items for Systematic Review.¹¹ It is part of a larger project to consolidate the evidence on PIP in primary care – the protocol of this systematic review was previously registered with International Prospective Register of Systematic Reviews (registration number CRD42016048874) and published in BMJ Open,¹² while two companion systematic reviews on the prevalence and adverse outcomes of PIP were recently published as separate papers.^2,3 Details on the conduct of current systematic review are briefly described here.

We systematically searched PubMed, Embase, CINAHL, Web of Science, Scopus and PsycINFO, from inception to seventh of January 2017, for original articles related to the keywords of ‘older persons’, ‘primary care’ and ‘inappropriate prescribing’. A sample of the search strategy based on PubMed is shown in Supplementary Appendix 1. Similar search strategies were used for the other databases. Additionally, we also hand-searched the references of six review articles related to PIP^5,10,13–16 to further identify relevant articles. We included studies that met the following criteria: (1) recruited participants from primary care settings; (2) either had ≥90% of the participants who were 60 years old and above, or reported subgroup analyses based on participants who were 60 years old and above; (3) conducted using observational study designs, such as cross-sectional, case–control or cohort studies; (4) reported the risk factors related to PIP; and (5) reported in the English language. Studies were excluded if they (1) recruited participants from non-primary care settings, such as those from the tertiary hospitals or nursing homes; (2) did not assess PIP based on published criteria; or (3) focused only on PIP related to specific classes of medications, such as antibiotics or analgesics.

Data extraction, quality assessment and data synthesis

Two reviewers (SKLG and ZYC) independently selected eligible articles, extracted the relevant data from the included articles and assessed the risk of bias. Any discordance between the two reviewers was resolved by discussion with a third independent reviewer (CSL or TML).

The extracted data included year of publication, study design, sampling method, sample size, inclusion criteria, age, sex, data source of medication prescriptions, criteria of PIP, reported risk factors of PIP and data source of risk factors. Risk of bias was assessed using the original 8-item Newcastle-Ottawa Scale (NOS)^17,18 for case–control and cohort study, and modified 5-item NOS for cross-sectional study.^19,20 NOS assesses three potential areas of bias in the included studies, namely, selection of participants, comparability of groups and measurement of exposure. A sample of NOS is shown in Supplementary Appendix 2. Studies were deemed to have low risk of bias if they achieved maximum or near-maximum scores on NOS (i.e. scores of 5–6 for cross-sectional studies; and scores of 8–9 for case–control and cohort studies).

Findings from the systematic review were synthesized qualitatively, as the data were deemed to be too heterogeneous for quantitative synthesis using meta-analytic methodologies.

Results

Of the 1017 articles that were initially identified, we included 25 articles comprising 24 cross-sectional studies and 1 case–control study. The flowchart on the selection process is shown in Figure 1. The 25 included studies involved a total of 2,893,925 participants, with the average age of study participants ranging from 70.4 to 84.0 years. In risk of bias assessment, 76% of the included studies were identified to be of low risk of bias. Key characteristics of the included studies, as well as detailed results on the quality assessment, are presented in Table 1.

Figure 1.

Flowchart of the study selection.

Table 1.

Key characteristics of the studies included in the systematic review.

Author (year)	Study design	Sampling method	Sample size	Inclusion criteria	Mean age in year	% female	Criteria of PIP	Data source of PIP	Data source of risk factor	Risk of bias (Newcastle-Ottawa Scale)^a
Amos (2015)²¹	Cross-sectional	Whole registry	865,354	≥65 years	77.0	57.0	Maio criteria 2011	Secure record	Secure record	Low
Barry (2016)²²	Cross-sectional	Whole registry	6826	Person with dementia	79.6	64.4	STOPP	Secure record	Secure record	Low
Berger (2009)²³	Cross-sectional	Whole registry	975	≥65 years Generalized anxiety disorder	75.0	72.0	Beers' 1997	Secure record	Secure record	Higher
Blanco-Reina (2014)²⁴	Cross-sectional	Random sampling	407	≥65 years	79.3	57.2	Beers' 2003 Beers' 2012 STOPP	Structured interview	Structured interview	Low
Blanco-Reina (2016)²⁵	Cross-sectional	Random sampling	225	≥65 years	73.1	56.9	STOPP V1 and V2 START V1 and V2 Beers’ 2012	Structured interview	Structured interview	Low
Ble (2015)²⁶	Cross-sectional	Whole registry	5262	≥65 years	NA	55.8	Beers’ 2012	Secure record	Secure record	Low
Bradley (2012)²⁷	Cross-sectional	Whole registry	166,108	≥70 years	NA	59.2	STOPP	Secure record	Secure record	Low
Bradley (2014)²⁸	Cross-sectional	Whole registry	1,019,491	≥70 years	NA	58.0	STOPP	Secure record	Secure record	Low
Buck (2009)²⁹	Cross-sectional	Whole registry	37247^b 24004^b	≥65 years	75.6 76.4	59.5 60.4	Beers’ 2003 Zhan 2001	Secure record	Secure record	Low
Cahir (2014)³⁰	Cross-sectional	Whole registry	338,375	≥70 years	NA	57.4	STOPP	Secure record	Secure record	Low
Carey (2008)³¹	Cross-sectional	Whole registry	218,567	≥65 years	NA	57.0	Beers’ 2003	Secure record	Secure record	Low
Castillo-Paramo (2014)³²	Cross-sectional	Random sampling	272	≥65 years	76.0	60.6	STOPP START	Secure record	Secure record	Low
Dalleur (2015)³³	Cross-sectional	Consecutive sampling	567	≥80 years	84.0	63.0	STOPP START Beers 2012	Structured interview	Structured interview	Low
Fiss (2011)³⁴	Cross-sectional	Whole registry	744	≥65 years Immobile patient receiving home visit	80.5	73.1	Beers’ 2003	Structured interview	Structured interview	Low
Fiss (2013)³⁵	Cross-sectional	Convenient sampling	111	≥65 years Immobile patient receiving home visit and with suspicion of dementia	80.9	78.4	PRISCUS	Structured interview	Structured interview	Higher
Howard (2004)³⁶	Cross-sectional	Random sampling	777	≥65 years	74.1	64.8	Beers’ 1997	Secure record	Structured interview	Low
Kovacevic (2014)³⁷	Cross-sectional	Consecutive sampling	509	≥65 years	74.8	57.4	STOPP START	Structured interview	Structured interview	Low
Meid (2016)³⁸	Cohort	Convenient sampling	989	≥65 years	70.4	75.3	START V2	Structured interview	Structured interview	Low
Moriarty (A) (2015)³⁹	Cohort	Random sampling	2051	≥65 years	74.8	54.0	STOPP Beers’ 2012 ACOVE START	Secure record	Structured interview	Low
Moriarty (B) (2015)⁴⁰	Cross-sectional	Whole registry	539,752	≥65 years	NA	NA	STOPP	Secure record	Secure record	Low
Oliveira (2012)⁴¹	Cross-sectional	Random sampling	139	≥60 years	NA	61.9	Beers’ 2003	Structured interview	Structured interview	Higher
Oliveira (2015)⁴²	Cross-sectional	Random sampling	139	≥60 years	NA	61.9	Beers’ 2012 STOPP	Structured interview	Structured interview	Higher
Projovic (2016)⁴³	Case–control	Consecutive sampling	690	≥65 years	NA	NA	STOPP START	Structured interview	Structured interview	Low
Ryan (2009)⁴⁴	Cross-sectional	Whole registry	1329	≥65 years	74.9	60.9	Beers' 2003 STOPP START	Secure record	Secure record	Higher
Wang Y (2016)⁴⁵	Cross-sectional	Convenient sampling	367	≥65 years Atrial fibrillation	77.8	44.7	Beers' 2012 PRISCUS	Secure record	Secure record	Higher

PIP: potentially inappropriate prescribing; IQR: inter-quartile range; STOPP: Screening Tool of Older Persons’ potentially inappropriate Prescriptions; START: Screening Tool to Alert to Right Treatment; ACOVE: Assessing Care of Vulnerable Elders; SD: standard deviation; NA: not available; V: version.

^bResult separated into two centres.

^aCross-sectional study can have a quality score ranging from 0* to 6* (with 5–6* indicating low risk of bias and <5* indicating higher risk of bias), while each case–control study can have a quality score ranging from 0* to 9* (with 8–9* indicating low risk of bias and <8* indicating higher risk of bias).

In general, PIP could be further classified into two categories, namely: (1) misuse or overuse of medicine (denoted as Potentially Inappropriate Medication, PIM) and (2) underuse of medication (denoted as Potential Prescribing Omission, PPO). PIM was evaluated in 24 studies, with majority of the studies evaluating PIM using STOPP (8 studies), Beers criteria (6 studies), or both STOPP and Beers criteria (6 studies). PPO was evaluated in eight studies, of which seven studies evaluated PPO using START and one study used both START and Assessing Care of Vulnerable Elders (ACOVE) indicators.⁴⁶ The risk factors that were significantly associated with PIM and PPO are summarized in Tables 2 and 3, respectively, as well as further described in the ensuing paragraphs. There are also other factors that have been evaluated but not found to be significant, and they are separately shown in Supplementary Appendix 3.

Table 2.

Factors associated with Potentially Inappropriate Medication (PIM) that have been reported to be significant in at least one study.

	Number of studies reporting significant finding	Total number of participants	Risk or protective factor
Patient factors
Polypharmacy	21^{21,22,24–32,34–37,39–41,43–45}	3,235,148	Risk
Multimorbidity	4^26,33,43,44	7848	Primarily risk^a
Psychiatric disorder	3^23–25	1607	Risk
Digestive disorder	1²³	975	Risk
Musculoskeletal and connective tissue disorder	1³⁵	342	Risk
Age	8^{21,26,28–30,39,40,44}	2,840,135	Primarily risk^b
Female	11^{21,22,26–31,35–36,39}	2,691,674	Primarily risk^c
Lower education	1⁴¹	139	Risk
Black ethnicity	1⁴¹	139	Risk
Use of over-the-counter medicine	1⁴¹	139	Risk
Unhealthy lifestyle	1⁴³	366	Risk
Physician and system factors
Older GP	1²¹	865,354	Risk
Male GP	1²¹	865,354	Risk
Solo GP	1²¹	865,354	Risk
GP who provide care to smaller pool of older patients	1²¹	865,354	Risk
More frequent visits by pharmaceutical sales representatives to GP	1⁴³	366	Risk
Higher number of primary care visits by patients	1²⁹	61,251	Risk
Medication source from public healthcare system	1⁴¹	139	Risk

GP: general practitioner.

^aThree studies reported multimorbidity as a risk factor, while one study reported as protective factor.

^bFive studies reported age as a risk factor, while three studies reported as protective factor.

^cNine studies reported female as a risk factor, while two studies reported as protective factor.

Table 3.

Factors associated with Potential Prescribing Omission (PPO) that have been reported to be significant in at least one study.

	Number of studies reporting significant finding	Total number of participants	Risk or protective factor
Patient factors
Multimorbidity	4^32,33,39,43	3214	Risk
Polypharmacy	2^32,39	2323	Risk
Age	2^32,39	2323	Risk
Female	1⁴⁴	1329	Risk
Cardiovascular disorder	1²⁵	225	Risk
Cognitive impairment	1³⁸	989	Risk
Poorer self-perceived health status	1³⁸	989	Risk
Physician and system factors
Higher number of primary care visits by patients	1³⁸	989	Risk
Lower number of specialist visits by patients	1⁴³	366	Risk
Larger GP practices	1⁴³	366	Risk

GP: general practitioner.

Polypharmacy

Polypharmacy – in association with PIM – was evaluated by 22 studies, with 21 studies showing polypharmacy as a risk factor (of which 16 had low risk of bias). Several studies also demonstrated that the risk estimate of polypharmacy rises with the increasing number of medications.^{21,26,32,35,43} For example, Ble et al.²⁶ reported OR 8.44 (95% CI 5.92–12.0) among individuals with five to nine medications and OR 25.46 (95% CI 17.7–36.5) among those with more than 10 medications, in reference to individuals with less than five medications.

Polypharmacy – in association with PPO – was evaluated by five studies, with two studies showing it as a significant risk factor of PPO (both studies had low risk of bias). Castillo-Paramo et al.³² reported RR 2.18 (95% CI 1.27–3.73) in the presence of more than six medications, while Moriarty et al.³⁹ reported OR 1.04 (95% CI 1.01–1.07) in higher number of medications.

Multimorbidity

Multimorbidity – in association with PIM – was evaluated by eight studies, with three studies showing multimorbidity as a risk factor (of which two had low risk of bias). Projovic et al.⁴³ demonstrated that higher scores on the geriatric version of Cumulative Illness Rating Scale (CIRS) were associated with higher PIM (OR 1.14, 95% CI 1.01–1.30), while Dalleur et al.³³ reported that those with lower comorbidities (CIRS <4) were associated with lower PIM (OR 0.40, 95% CI 0.30–0.70). Similarly, Ryan et al.⁴⁴ reported the significant association between higher scores on Charlson Comorbidity Index (CCI) and higher PIM. It should also be noted that one study, by Ble et al.,²⁶ demonstrated an atypical finding of multimorbidity as a protective factor of PIM. However, the association was no longer significant after the authors further accounted for four cardiovascular conditions (hypertension, diabetes, coronary artery disease and stroke), suggesting the presence of spurious association due to confounding by indication.

Multimorbidity – in association with PPO – was measured by five studies, with four studies reported multimorbidity as a risk factor. All four studies were of low risk of bias. Projovic et al.⁴³ demonstrated that the higher scores on CIRS-Geriatric were associated with higher PPO (OR 1.30, 95% CI 1.13–1.49), while Castillo-Paramo et al.³² reported that those with CCI>2 had RR of 2.42 (95% CI 1.54–3.80), and Moriarty et al.³⁹ reported that higher number of comorbidities had OR of 1.47 (95% CI 1.39–1.56). Similarly, Dalleur et al.³³ reported that those with lower comorbidities (CIRS <4) were associated with lower PPO (OR 0.20, 95% CI 0.10–0.30).

Three categories of diseases (namely, psychiatric disorder, digestive disorder, and musculoskeletal or connective tissue disorder) were identified as risk factors of PIM. Psychiatric disorder was measured by six studies, with three studies reporting it as a significant risk factor. Two studies by Blanco-Reina et al.^24,25 reported OR of 2.22 (95% CI 1.13–4.37) and 2.91 (95% CI 1.83–4.66), respectively, while another study by Berger et al.²³ showed that patients with comorbid depression were more likely to receive PIM (51.6% vs 35.5% for those without comorbid depression; p < .01). Digestive disorder was reported as a risk factor by Berger et al.²³ (43.0% vs 36.6% for those without the disorder; p = .04), while musculoskeletal disorder was reported as a risk factor by Fiss et al.³⁵ (OR 4.20, 95% CI 1.10–16.01). However, all the studies were assessed to be of higher risk of bias – Berger et al.²³ sampled population with general anxiety disorder and Fiss et al.³⁵ sampled the population with cognitive impairment; both of which may not be representative of the general population. Two categories of diseases (cardiovascular disorder and cognitive impairment) were identified as risk factors with PPO. Blanco Reina et al.²⁵ reported OR 2.50 (95% CI 1.16–5.81) in cardiovascular disorder, while Meid et al.³⁸ reported OR 0.93 (95% CI 0.87–0.99) in person with higher cognition scores on Mini-Mental State Examination.

Age

Age – in association with PIM – was evaluated by 18 studies, with 5 studies showing age as a risk factor (of which 4 had low risk of bias) and 3 studies showing age as a protective factor (all 3 had low risk of bias). Among studies that identified age as a risk factor, Moriarty et al.³⁹ reported OR 1.03 (95% CI 1.02–1.04) and Amos et al.²¹ reported, OR 1.25 (95% CI 1.23–1.26) (comparing 75–84 to 65–74 years) and OR 1.53 (95% CI 0.1.50–1.55) (comparing 85 years to 65–74 years). Among studies that identified age as a protective factor, Bradley et al.²⁸ reported OR 0.9 (95% CI 0.9–0.9) (comparing 75–80 years to 70–74 years), OR 0.8 (95% CI 0.8–0.8) (comparing 81–85 to 70–74 years) and OR 0.40 (95% CI 0.40–0.40) (comparing ≥85 years to 70–74 years), Ble et al.²⁶ reported OR 0.90 (95% CI 0.82–0.99) (comparing ≥85 years to 65–84 years), and Cahir et al.³⁰ reported OR 0.95 (95% CI 0.93–0.96) (comparing ≥75 years to 70–74 years).

Age – in association with PPO – was evaluated by seven studies, with two studies showing it as a risk factor (both studies had low risk of bias). Of the two studies, Moriarty et al.³⁹ reported OR 1.03 (95% CI 1.02–1.04) and Castillo-Paramo et al.³² reported relative risk (RR) 1.02 (95% CI 1.00–1.04).

Gender

Female gender – in association of PIM – was evaluated by 20 studies, with 9 studies showing female gender as a risk factor (of which 6 studies had low risk of bias) and 2 studies showing female gender as a protective factor (both were of low risk of bias). Among studies that identified female gender as a risk factor, Barry et al.²² reported OR of 1.30 (95% CI 1.2–1.4), Ble et al.²⁶ reported OR of 1.18 (95% CI 1.04–1.33), Bradley et al.²⁷ reported OR 1.26 (95% CI 1.23–1.29), Fiss et al.³⁵ reported OR of 10.36 (95% CI 1.28–83.87) and Howard et al.³⁶ reported OR 1.60 (95% CI 1.00–2.40), Moriarty et al.³⁹ reported OR 1.27 (95% CI 1.07–1.5), Amos et al.²¹ reported OR of 0.98 (95% CI 0.97–1.00) comparing male to female, Carey et al.³¹ reported OR of 0.84 (95% CI 0.81–0.86) comparing male to female and Buck et al.²⁹ reported statistical significant association of female to PIM (p < .001). Among studies that identified female gender as a protective factor, Cahir et al.³⁰ reported OR 0.94 (95% CI 0.91–0.97) and Bradley et al.²⁸ reported OR 0.90 (95% CI 0.90–0.90). Female gender – in association with PPO – was evaluated by six studies, with only one study by Ryan et al.⁴⁴ showing it as a risk factor. Ryan et al.⁴⁴ reported that the incident of PPO was significantly higher in female (27.8%) than in male (14.8%) (p < .001). However, this study is of higher risk of bias, while the rest of five studies that have low risk of bias did not demonstrate significance association between female gender and PPO.

Other patient-related factors

Several other patient factors have also been reported to be significant risk factors of PIM, albeit only in single studies. They include lower educational attainment,⁴¹ black ethnicity,⁴¹ unhealthy lifestyle (defined as meeting at least two of the four predefined unhealthy behaviours, namely, sedentary lifestyle, improper nutrition, active smoking or heavy alcohol consumption)⁴³ and the use of over-the-counter medicine.⁴¹ In the context of PPO, one study reported the association between poorer self-perceived health status and PPO, but the study authors cautioned that the exact causal relationship between the two remained unclear.³⁸

Physician-related factors

Several characteristics of the general practitioners (GP) were reported as risk factors of PIM, including older age (OR 1.05, 95% CI 1.00–1.11),²¹ male gender (OR 1.29, 95% CI 1.22–1.36),²¹ solo practice (OR 1.16, 95% CI 1.16–1.25),²¹ providing care to smaller pool of older patients (OR 1.25, 95% CI 1.16–1.35)²¹ and more frequent visits by pharmaceutical sales representatives (OR 2.28, 95% CI 1.10–4.75).⁴³ These characteristics were reported in single studies that were assessed to have low risk of bias. On top of these characteristics, one study that had low risk of bias also reported the association between the size of GP practice and PPO. Projovic et al.⁴³ found that bigger GP practice (in contrast to smaller centres or branches) had a higher risk of PPO, with the reported OR of 1.99 (95% CI 1.06–3.74). The authors suggested that the size of GP practice may be a reflection of GPs’ familiarity with their patients, whereby those in smaller centres may possibly have less patients and hence more familiarity with the patients that they are serving.

Systems-related factors

Several factors related to health systems were identified as risk factors of PIM or PPO, including higher number of primary care visits by patients, lower number of specialist visits by patients and medication source from public health healthcare system. Buck et al.²⁹ reported the significant association of higher number of primary care visits with PIM, while Meid et al.³⁸ reported the association between PPO and higher frequency of GP consultation in the preceding 3 months (OR 1.07, 95% CI 1.00–1.15), and Projovic et al.⁴³ reported the reduced risk of PPO associated with higher number of specialist visits (OR 0.88, 95% CI 0.81–0.95). These three studies had low risk of bias.

Meanwhile, medication source from the public health system (in contrast to medications from other sources such as private health system) was reported by Oliveira et al.⁴¹ as a significant risk factor of PIM (PR 1.42, 95% CI 1.10–1.81). This study has higher risk of bias. The author suggested that the higher PIM in public sector may be explained by administrative factors in the study population (where the choice of safer therapeutic alternatives were limited), as well as explained by demographic factors (where more black people in the study population utilized the public health system compared to other ethnicity).

Discussion

Summary of findings

This systematic review identified multiple factors that could be associated with PIP in primary care, and these key factors could be further summarized in Table 4, with broad classifications into those related to patients (e.g. demographic, medical condition and lifestyle factor), as well as physicians and systems (e.g. characteristic of general practitioners and factors related to health systems).

Table 4.

A summary of the risk factors associated with Potentially Inappropriate Medication (PIM) and Potential Prescribing Omission (PPO).

Patient factors	Physician and system factors
Patient demographics	Characteristics of general practitioners
• Female	• Older age^a
• Older adult	• Male gender^a
• Lower education^a	• Solo practice^a
• Black ethnicity^a	• Providing care to smaller pool of older patients^a
Medical comorbidities	• More frequent visits by pharmaceutical sales representatives^a
• Polypharmacy	• Larger GP practices^b
• Multimorbidity	Factors related to health systems
• Psychiatric disorder^a	• Higher number of primary care visits by patients
• Digestive disorder^a	• Lower number of specialist visits by patients^b
• Cardiovascular disorder^b	• Medication source from public healthcare system^a
• Cognitive impairment^b
• Musculoskeletal or connective tissue disorder^a
Lifestyle factors
• Unhealthy lifestyle^a
• Use of over-the-counter medicine^a
• Poorer self-perceived health status^b

^aFactor associated with PIM only.

^bFactor associated with PPO only.

Comparison with existing literature

Polypharmacy is among the most commonly reported risk factor of PIP (both PIM and PPO), as seen in the current study. Polypharmacy is commonly defined as the concurrent use of multiple medication (generally ≥4 or 5 medications).⁴⁷ In the literature, polypharmacy has been known to impact about 30% of older person in primary care.^48,49 The association of polypharmacy with PIM may possibly be explained by the complex phenomenon of ‘prescribing cascade’,⁵⁰ which involves the situation whereby a clinician is not being able to link the complaints back to the adverse effects of medication and hence unnecessarily prescribes even more medications to manage the adverse effects. An example of ‘prescribing cascade’ can be seen in a recent cohort study, whereby older adults prescribed with calcium channel blocker were more likely to be prescribed with loop diuretic as a result of lower limb swelling caused by the calcium channel blocker.⁵¹ Many factors can contribute to the phenomenon of ‘prescribing cascade’, with a recent qualitative study delineating three key reasons from the patients’ and healthcare providers’ point of view: (1) varying awareness of medications and cascades; (2) varying feelings of accountability for making decisions about medication-related care; and (3) accessibility to an ideal environment and relevant information.⁵² While the relationship between polypharmacy and PIM may be easier to understand, the association between polypharmacy and PPO has been less clear in extant literature. One recent study⁵³ suggested that in the presence of polypharmacy, healthcare professionals may be reluctant to prescribe additional medications even when indicated, especially among those who are older or more frail. This phenomenon may have plausibly explained the unintentional omission of appropriate medications (i.e. PPO) in the presence of polypharmacy.

Multimorbidity is another commonly reported factor associated with both PIM and PPO. Multimorbidity is most frequently defined as co-occurrence of two or more chronic conditions,⁵⁴ with reported prevalence ranging from 12.9% to 95.1% in primary care population aged 65 years and older.⁵⁵ In the presence of multimorbidity, older patients often have to take multiple medications to treat the respective diseases.⁵⁶ Yet, it may not be an easy task for physicians to make therapeutic decisions in the presence of multimorbidity. There is a paucity of scientific evidence and guidelines for this group of patients, given that older patients are often excluded from clinical trials, and most of the recommendations from various clinical guidelines often only focus on the treatment of single disease rather than considering multiple diseases holistically.^56,57 Consequently, some patients may be over-prescribed with inappropriate medications when drug-drug and drug-disease relationships are not carefully considered.⁴³ At the same time, when patients have higher number of comorbidities, it may also be easier for physicians to miss out the prescribing of appropriate medications even in the presence of clinical indication, which may inadvertently result in PPO.^32,39,43

As seen in this study, various diseases (e.g. psychiatric, digestive, musculoskeletal and cardiovascular disorders) have been reported to be associated with PIP. While it is not yet clear on why there are such associations, the literature has reported the links of these diseases to specific classes of drugs. For example, PIM in psychiatric disorders has been linked to the over-prescribing of psychotropic drugs such as antidepressant, sedatives (anxiolytics, hypnotics) and antipsychotic,^{23–25,35,58,59} possibly due to the lack of prescribers’ knowledge on the use of psychotropics.^60,61 Meanwhile, PIP in digestive disorders have been linked to the overuse of acid suppressing treatments such as proton pump inhibitors in the absence of clear clinical indications⁶²; PIMs in musculoskeletal disorders are linked to inappropriate use of analgesics such as opioid and non-steroidal anti-inflammatory drugs^63,64; and PPO in cardiovascular disorders are linked to under-prescribing of aspirin, statin and angiotensin-converting enzyme inhibitor.^25,32,44 Notwithstanding our finding of association between PIP and various diseases, it is pertinent to note that most of the studies included in this systematic review utilized explicit criteria such as the Beers’ and STOPP/START to assess PIP, and these explicit criteria follow a strict rules on what constitute PIP. While explicit criteria can be useful to improve consistency in clinical practice, such approach can be less flexible and may not take into account unique clinical scenarios that require clinical judgement and weighing of the risk–benefit ratio to the patients. As a case in point, there are clinical scenarios where short-term benzodiazepine may rarely be required, especially among those with severe agitation not amenable to other interventions or medications. In such situation, the use of benzodiazepine may still be necessary, but would have been deemed as a PIP under Beers’ criteria regardless of the duration of prescription and the risk–benefit ratio. Clinicians who prefer more flexibility in the assessment of PIP can consider using implicit criteria such as the Medication Appropriateness Index (MAI), which accounts for clinical judgement in deciding the presence of PIP. However, such implicit criteria have their inherent limitations, as they are time consuming and require specialized clinical knowledge to administer.⁴

This study also identified several physician and systems related factors of PIP, which is not inconsistent with what has been reported in the literature. A recent meta-synthesis described four key themes that may explain why physicians resort to inappropriate prescribing: (1) the need to please patient; (2) feeling of being forced to prescribe due to lack of access to non-pharmacological interventions or allied health professionals; (3) prescriber’s reluctance to adopt new recommendations in prescribing guidelines, due to lack of awareness on potential harm of PIP; and (4) prescriber’s fear related to lack of clinical experience, knowledge gap in using certain drugs and discomfort in changing medications prescribed by another physician.⁶⁵ Another systematic review demonstrated that inadequate prescribing knowledge among physicians, as well as pharmaceutical marketing strategy (e.g. by actively promoting specific drugs or through engagement of pharmaceutical representatives), can contribute to inappropriate prescribing of antibiotics.⁶⁶

Implications of the findings

Findings from the current systematic review can be useful to improve our understanding on the mechanism of PIP and inform the design of future intervention studies on PIP. Traditionally, various strategies have been proposed to target at both the prescribing milieu and the systemic factors that contributed to PIP in primary care settings.^1,67 For example, the prescribing milieu can be effectively addressed by deprescribing⁶⁸ or medication review by clinical pharmacists,⁶⁹ while the systemic factors can be addressed by improving prescriber’s knowledge through educational interventions (e.g. academic detailing)⁵ or using technology such as computerized clinical decision support systems⁷⁰ to guide the prescribers. However, most of the studies focused on isolated intervention in improving PIP and as a result, although these intervention studies have shown success in reducing the number of PIP, these successes have not translated into robust evidence related to improvements in clinically relevant outcomes such as mortality, and hospital admissions or healthcare utilization.^71,72 The design of complex interventions often requires much contextual considerations, that is, by conceptualizing the problem (i.e. classifying and evaluating the factors associated with PIP), developing an optimal intervention, and evaluating the outcome measures.⁷³ Findings from the current study may allow a glimpse into the independent and interdependent factors associated with PIP, which may be critical when designing complex multifaceted interventions targeting the patient, physician and system levels.⁷⁴

The identified factors of PIP from this study may potentially be used to develop a risk prediction tool to capture high-risk groups who may benefit from further interventions related to PIP. In the current age of technology advancement, machine learning has been used in multiple healthcare domains to investigate its potential for prognosis, diagnosis or differentiation of clinical groups.⁷⁵ In particular, deep learning, which is a subfield of machine learning, has been applied to the development of risk prediction models, using large and complex electronic health data which allow computational models that are composed of multiple processing layers based on neural networks to learn representations of data with multiple levels of abstraction.⁷⁶ In the context of PIP, deep learning methods may be applied (via integration of broad and complex information from the electronic health data to capture the risk factors associated with PIP) in developing a risk prediction model for risk stratification and future event prediction. Through such prediction approach, we can possibly identify high-risk patients who may benefit from further interventions related to PIP.

Limitation

Several limitations should be considered. First, we included only studies reported in the English language and may have missed other relevant evidence in the non-English literature. Second, meta-analysis of the individual risk factors could not be conducted in this study due to the heterogeneity of the included studies (e.g. the vast variations in the definition of polypharmacy, measurement of comorbidities and definition of advanced age). As such, this study could not provide conclusive evidence on the effect sizes of various risk factors of PIP, and instead, could only qualitatively synthesize the evidence based on studies that have reported significant findings on each risk factor. Third, we did not include studies of PIP that are specific to certain drug classes (e.g. analgesics and antibiotic) and therefore may not be able draw a strong conclusion on the association of the specific classes of drug with PIP. Fourth, due to the large heterogeneity of the included studies, extended time was needed to analyse the data which resulted in the lag time between when the literature search was conducted and the writing up of the findings in the current report.

Conclusion

This systematic review summarized the evidence on factors associated with PIP among the older persons in primary care. The findings provide contextual information to guide our understanding of PIP in primary care. The factors identified in this review can inform the design of complex interventions for PIP, as well as be used to develop risk prediction tools to identify high-risk groups who may benefit from further interventions related to PIP. Given the central role that primary care plays in coordinating healthcare, the findings call for prioritization of initiatives to reduce iatrogenic medication-related harm targeting at multiple level (individuals, prescribers and systems level) in primary care settings.

Supplemental Material

Supplemental Material - Factors associated with potentially inappropriate prescribing among older persons in primary care settings: Systematic review

Supplemental Material for Factors associated with potentially inappropriate prescribing among older persons in primary care settings: Systematic review by Cia Sin Lee, Ngiap Chuan Tan, Kuan Liang Shawn Goh, Zi Ying Chang, and Tau Ming Liew in Proceedings of Singapore Healthcare.

Footnotes

Author contributions

All authors have contributed significantly, and all authors agree with the content of the article.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Cia Sin Lee

Ngiap Chuan Tan

Data availability statement

Data will be made available upon reasonable request to the corresponding authors, Associate professor Dr Liew Tau Ming, email: liew.tau.ming@singhealth.com.sg, or Dr Lee Cia Sin, email: .

Supplemental Material

Supplemental material for this article is available online.

References

O’Connor

Gallagher

O’Mahony

. Inappropriate prescribing: criteria, detection and prevention. Drugs Aging 2012; 29: 437–452.

Liew

Lee

Goh Shawn

, et al. Potentially inappropriate prescribing among older persons: a meta-analysis of observational studies. Ann Fam Med 2019; 17: 257–266.

Liew

Lee

Goh

SKL

, et al. The prevalence and impact of potentially inappropriate prescribing among older persons in primary care settings: multilevel meta-analysis. Age Ageing 2020; 49(4): 570–579.

Kaufmann

Tremp

Hersberger

, et al. Inappropriate prescribing: a systematic overview of published assessment tools. Eur J Clin Pharmacol 2014; 70: 1–11.

Spinewine

Schmader

Barber

, et al.

Appropriate prescribing in elderly people: how well can it be measured and optimised?

Lancet 2007; 370: 173–184.

World Health Organization . Global health and aging 2011 [Internet]. Geneva, Switzerland: World Health Organization, [cited 2022 Nov 22]. Available from: https://www.nia.nih.gov/sites/default/files/2017-06/global_health_aging.pdf

Gong

, et al. Epidemiology of chronic noncommunicable diseases and evaluation of life quality in elderly. Aging Med 2018; 1: 64–66.

World Health Organization . Medication errors: technical series on safer primary care. Geneva, Switzerland: World Health Organization, 2016.

van Weel

Kidd

. Why strengthening primary health care is essential to achieving universal health coverage. Cmaj 2018; 190: E463–E466.

10.

Tommelein

Mehuys

Petrovic

, et al. Potentially inappropriate prescribing in community-dwelling older people across Europe: a systematic literature review. Eur J Clin Pharmacol 2015; 71: 1415–1427.

11.

Moher

Shamseer

Clarke

PRISMA-P Group , et al. Preferred reporting items for systematic review and meta-analysis protocols (PRISMA-P) 2015 statement. Syst Rev 2015; 4: 1.

12.

Lee

Liew

. Inappropriate prescribing among older persons in primary care: protocol for systematic review and meta-analysis of observational studies. BMJ Open 2017; 7(2): e015395.

13.

Hyttinen

Jyrkka

Valtonen

. A systematic review of the impact of potentially inappropriate medication on health care utilization and costs among older adults. Med Care 2016; 54: 950–964.

14.

Tesfaye

Castelino

Wimmer

, et al. Inappropriate prescribing in chronic kidney disease: a systematic review of prevalence, associated clinical outcomes and impact of interventions. Int J Clin Pract 2017; 71(7): e12960.

15.

Muhlack

Hoppe

Weberpals

, et al. The association of potentially inappropriate medication at older age with cardiovascular events and overall mortality: a systematic review and meta-analysis of cohort studies. J Am Med Dir Assoc 2017; 18: 211–220.

16.

Opondo

Eslami

Visscher

, et al. Inappropriateness of medication prescriptions to elderly patients in the primary care setting: a systematic review. PLoS One 2012; 7: e43617.

17.

Wells

Shea

O’Connell

, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomised studies in metaanalyses [internet]. Ottawa, ON: The Ottawa Hospital, [cited 2022 Nov 17]. Available from: https://www.ohri.ca/programs/clinical_epidemiology/oxford.asp

18.

Lewis-Mikhael

A-M

Bueno-Cavanillas

Ofir Giron

, et al. Occupational exposure to pesticides and prostate cancer: a systematic review and meta-analysis. Occup Environ Med 2016; 73: 134–144.

19.

Stang

. Critical evaluation of the Newcastle-Ottawa scale for the assessment of the quality of nonrandomized studies in meta-analyses. Eur J Epidemiol 2010; 25: 603–605.

20.

Mata

Ramos

Bansal

, et al. Prevalence of depression and depressive symptoms among resident physicians: a systematic review and meta-analysis. JAMA 2015; 314: 2373–2383.

21.

Amos

Keith

Del Canale

, et al. Inappropriate prescribing in a large community-dwelling older population: a focus on prevalence and how it relates to patient and physician characteristics. J Clin Pharm Ther 2015; 40: 7–13.

22.

Barry

Cooper

Ryan

, et al. Potentially inappropriate prescribing among people with dementia in primary care: a retrospective cross-sectional study using the enhanced prescribing database. J Alzheimers Dis 2016; 52: 1503–1513.

23.

Berger

Mychaskiw

Dukes

, et al. Magnitude of potentially inappropriate prescribing in Germany among older patients with generalized anxiety disorder. BMC Geriatr 2009; 9: 31.

24.

Blanco-Reina

Ariza-Zafra

Ocana-Riola

, et al. 2012 American geriatrics society beers criteria: enhanced applicability for detecting potentially inappropriate medications in European older adults? a comparison with the screening tool of older person’s potentially inappropriate prescriptions. J Am Geriatr Soc 2014; 62: 1217–1223.

25.

Blanco-Reina

Garcia-Merino

Ocana-Riola

, et al. Assessing potentially inappropriate prescribing in community-dwelling older patients using the updated version of STOPP-START criteria: a comparison of profiles and prevalences with respect to the original version. PLoS One 2016; 11: e0167586.

26.

Ble

Masoli

Barry

, et al. Any versus long-term prescribing of high risk medications in older people using 2012 beers criteria: results from three cross-sectional samples of primary care records for 2003/4, 2007/8 and 2011/12. BMC Geriatr 2015; 15: 146.

27.

Bradley

Fahey

Cahir

, et al. Potentially inappropriate prescribing and cost outcomes for older people: a cross-sectional study using the Northern Ireland Enhanced Prescribing Database. Eur J Clin Pharmacol 2012; 68: 1425–1433.

28.

Bradley

Motterlini

Padmanabhan

, et al. Potentially inappropriate prescribing among older people in the United Kingdom. BMC Geriatr 2014; 14: 72.

29.

Buck

Atreja

Brunker

, et al. Potentially inappropriate medication prescribing in outpatient practices: prevalence and patient characteristics based on electronic health records. Am J Geriatr Pharmacother 2009; 7: 84–92.

30.

Cahir

Fahey

Teljeur

, et al. Prescriber variation in potentially inappropriate prescribing in older populations in Ireland. BMC Fam Pract 2014; 15: 59.

31.

Carey

De Wilde

Harris

, et al. What factors predict potentially inappropriate primary care prescribing in older people? analysis of UK primary care patient record database. Drugs Aging 2008; 25: 693–706.

32.

Castillo-Paramo

Claveria

Verdejo Gonzalez

, et al. Inappropriate prescribing according to the STOPP/START criteria in older people from a primary care setting. Eur J Gen Pract 2014; 20: 281–289.

33.

Dalleur

Boland

De Groot

, et al. Detection of potentially inappropriate prescribing in the very old: cross-sectional analysis of the data from the BELFRAIL observational cohort study. BMC Geriatr 2015; 15: 156.

34.

Fiss

Dreier

Meinke

, et al. Frequency of inappropriate drugs in primary care: analysis of a sample of immobile patients who received periodic home visits. Age Ageing 2011; 40: 66–73.

35.

Fiss

Thyrian

Fendrich

, et al. Cognitive impairment in primary ambulatory health care: pharmacotherapy and the use of potentially inappropriate medicine. Int J Geriatr Psychiatry 2013; 28: 173–181.

36.

Howard

Dolovich

Kaczorowski

, et al. Prescribing of potentially inappropriate medications to elderly people. Fam Pract 2004; 21: 244–247.

37.

Kovacevic

Simisic

Rudinski

, et al. Potentially inappropriate prescribing in older primary care patients. PLoS One 2014; 9: 7.

38.

Meid

Quinzler

Groll

, et al. Longitudinal evaluation of medication underuse in older outpatients and its association with quality of life. Eur J Clin Pharmacol 2016; 72: 877–885.

39.

Moriarty

Bennett

Fahey

, et al. Longitudinal prevalence of potentially inappropriate medicines and potential prescribing omissions in a cohort of community-dwelling older people. Eur J Clin Pharmacol 2015; 71: 473–482.

40.

Moriarty

Hardy

Bennett

, et al. Trends and interaction of polypharmacy and potentially inappropriate prescribing in primary care over 15 years in Ireland: a repeated cross-sectional study. BMJ Open 2015; 5: e008656.

41.

Oliveira

Amorim

de Jesus

, et al. Factors associated with potentially inappropriate medication use by the elderly in the Brazilian primary care setting. Int J Clin Pharm 2012; 34: 626–632.

42.

Oliveira

Amorim

de Jesus

, et al. A comparison of the Beers and STOPP criteria for identifying the use of potentially inappropriate medications among elderly patients in primary care. J Eval Clin Pract 2015; 21: 320–325.

43.

Projovic

Vukadinovic

Milovanovic

, et al. Risk factors for potentially inappropriate prescribing to older patients in primary care. Eur J Clin Pharmacol 2016; 72: 93–107.

44.

Ryan

O'Mahony

Kennedy

, et al. Potentially inappropriate prescribing in an Irish elderly population in primary care. Br J Clin Pharmacol 2009; 68: 936–947.

45.

Wang

Singh

Bajorek

. Old age, high risk medication, polypharmacy: a ‘trilogy’ of risks in older patients with atrial fibrillation. Pharm Pract 2016; 14: 706.

46.

Wenger

Roth

Shekelle

ACOVE Investigators . Introduction to the assessing care of vulnerable elders-3 quality indicator measurement set. J Am Geriatr Soc 2007; 55(Suppl 2): S247–S252.

47.

Masnoon

Shakib

Kalisch-Ellett

, et al. What is polypharmacy? a systematic review of definitions. BMC Geriatr 2017; 17: 230.

48.

Junius-Walker

Theile

Hummers-Pradier

. Prevalence and predictors of polypharmacy among older primary care patients in Germany. Fam Pract 2007; 24: 14–19.

49.

Acosta

. Prevalence of polypharmacy and interactions in older adults in primary care. Int J Family Community Med 2018; 2(5): 310–313.

50.

Rochon

Gurwitz

. The prescribing cascade revisited. Lancet 2017; 389: 1778–1780.

51.

Savage

Visentin

Bronskill

, et al. Evaluation of a common prescribing cascade of calcium channel blockers and diuretics in older adults with hypertension. JAMA Intern Med 2020; 180: 643–651.

52.

Farrell

Jeffs

Irving

, et al. Patient and provider perspectives on the development and resolution of prescribing cascades: a qualitative study. BMC Geriatr 2020; 20: 368.

53.

Khawla Abu Hammour

Abu Farha

Alhabeis

et al. Potential prescribing omissions of medications in older adults based on START version 2 criteria in a tertiary Jordanian teaching hospital. Journal of Applied Pharmaceutical Science 2021; 11(11): 046–052.

54.

Almirall

Fortin

. The coexistence of terms to describe the presence of multiple concurrent diseases. J Comorb 2013; 3: 4–9.

55.

Violan

Foguet-Boreu

Flores-Mateo

, et al. Prevalence, determinants and patterns of multimorbidity in primary care: a systematic review of observational studies. PLoS One 2014; 9: e102149.

56.

Cahill

. Prescribing for patients with multimorbidity: aiming to tailor to patient-set goals. Br J Gen Pract 2015; 65: 114–115.

57.

Van Spall

Toren

Kiss

, et al. Eligibility criteria of randomized controlled trials published in high-impact general medical journals: a systematic sampling review. JAMA 2007; 297: 1233–1240.

58.

Aparasu

Mort

Sitzman

. Psychotropic prescribing for the elderly in office-based practice. Clin Ther 1998; 20: 603–616.

59.

Mort

Aparasu

. Prescribing potentially inappropriate psychotropic medications to the ambulatory elderly. Arch Intern Med 2000; 160: 2825–2831.

60.

Fraser

Oyama

. Knowledge of psychotropics and prescribing preferences of family physicians: a preliminary study. Acad Psychiatry 2013; 37: 325–328.

61.

Svensson

Hedenrud

Wallerstedt

. Attitudes and behaviour towards psychotropic drug prescribing in Swedish primary care: a questionnaire study. BMC Fam Pract 2019; 20: 4.

62.

Ryder

O'Reilly

Miller

, et al. Long term acid suppressing treatment in general practice. BMJ 1994; 308: 827–830.

63.

Ingrasciotta

Sultana

Giorgianni

, et al. Analgesic drug use in elderly persons: a population-based study in Southern Italy. PLoS One 2019; 14: e0222836.

64.

Soprano

Hennessy

Bilker

, et al. Assessment of physician prescribing of muscle relaxants in the United States, 2005–2016. JAMA Netw Open 2020; 3: e207664.

65.

Cullinan

O’Mahony

Fleming

, et al. A meta-synthesis of potentially inappropriate prescribing in older patients. Drugs Aging 2014; 31: 631–638.

66.

Md Rezal

Hassali

Alrasheedy

, et al. Physicians’ knowledge, perceptions and behaviour towards antibiotic prescribing: a systematic review of the literature. Expert Rev Anti Infect Ther 2015; 13: 665–680.

67.

Anderson

Stowasser

Freeman

, et al. Prescriber barriers and enablers to minimising potentially inappropriate medications in adults: a systematic review and thematic synthesis. BMJ Open 2014; 4: e006544.

68.

Scott

Hilmer

Reeve

, et al. Reducing inappropriate polypharmacy: the process of deprescribing. JAMA Intern Med 2015; 175: 827–834.

69.

Lenander

Elfsson

Danielsson

, et al. Effects of a pharmacist-led structured medication review in primary care on drug-related problems and hospital admission rates: a randomized controlled trial. Scand J Prim Health Care 2014; 32: 180–186.

70.

Marasinghe

. Computerised clinical decision support systems to improve medication safety in long-term care homes: a systematic review. BMJ Open 2015; 5: e006539.

71.

Huiskes

VJB

Burger

van den Ende

CHM

, et al. Effectiveness of medication review: a systematic review and meta-analysis of randomized controlled trials. BMC Fam Pract 2017; 18: 5.

72.

Kaur

Mitchell

Vitetta

, et al. Interventions that can reduce inappropriate prescribing in the elderly: a systematic review. Drugs Aging 2009; 26: 1013–1028.

73.

Campbell

Murray

Darbyshire

, et al. Designing and evaluating complex interventions to improve health care. BMJ 2007; 334: 455–459.

74.

Campbell

Fitzpatrick

Haines

, et al. Framework for design and evaluation of complex interventions to improve health. BMJ 2000; 321: 694–696.

75.

Triantafyllidis

Tsanas

. Applications of machine learning in real-life digital health interventions: review of the literature. J Med Internet Res 2019; 21: e12286.

76.

Miotto

Wang

, et al. Deep learning for healthcare: review, opportunities and challenges. Brief Bioinform 2018; 19: 1236–1246.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.19 MB