Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 https://doi.org/10.1186/s12911-021-01710-8 RESEARCH Open Access Safer prescription of drugs: impact of the PREFASEG system to aid clinical decision‑making in primary care in Catalonia M. Àngels Pons‑Mesquida1,3*, Míriam Oms‑Arias1, Eduard Diogène‑Fadini2,3 and Albert Figueras3 Abstract Background: In 2008, the Institut Català de la Salut (ICS, Catalan Health Institute) implemented a prescription deci‑ sion support system in its electronic clinical workstation (ECW), which automatically generates online alerts for gen‑ eral practitioners when a possible medication-related problem (MRP) is detected. This tool is known as PREFASEG, and at the time of beginning a new treatment, it automatically assesses the suitability of the treatment for the individual patient. This analysis is based on ongoing treatments, demographic characteristics, existing pathologies, and patient biochemical variables. As a result of the assessment, therapeutic recommendations are provided. The objective of this study is to present the PREFASEG tool, analyse the main alerts that it generates, and determine the degree of alert acceptance. Methods: A cross-sectional descriptive study was carried out to analyse the generation of MRP-related alerts detected by PREFASEG during 2016, 2017, and 2018 in primary care (PC) in Catalonia. The number of MRP alerts generated, the drugs involved, and the acceptance/rejection of the alerts were analysed. An alert was considered "accepted" when the medication that generated the alert was not prescribed, thereby following the recommendation given by the tool. The MRP alerts studied were therapeutic duplications, safety alerts issued by the Spanish Medicines Agency, and drugs not recommended for use in geriatrics. The prescriptions issued by 6411 ICS PC physicians who use the ECW and provide their services to 5.8 million Catalans through 288 PC teams were analysed. Results: During the 3 years examined, 67.2 million new prescriptions were analysed, for which PREFASEG generated 4,379,866 alerts (1 for every 15 new treatments). A total of 1,222,159 alerts (28%) were accepted. Pharmacological interactions and therapeutic duplications were the most detected alerts, representing 40 and 30% of the total alerts, respectively. The main pharmacological groups involved in the safety alerts were nonsteroidal anti-inflammatory drugs and renin-angiotensin system inhibitors. Conclusions: During the period analysed, 28% of the prescriptions wherein a toxicity-related PREFASEG alert was generated led to treatment modification, thereby helping to prevent the generation of potential safety MRPs. How‑ ever, the tool should be further improved to increase alert acceptance and thereby improve patient safety. Keywords: Clinical decision support system, Primary care, Clinical safety, Electronic prescription, Pharmacovigilance, Medicines use Background *Correspondence: aponsmesquida@gencat.cat According to a European Commission report, 3–10% of 1 Unitat de Coordinació i Estratègia del Medicament (UCEM), Institut hospital admissions between 2012 and 2014 were caused Català de la Salut, Barcelona, Spain by adverse drug events (ADEs), totalling 2.5–8.4 million Full list of author information is available at the end of the article © The Author(s) 2021. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/. The Creative Commons Public Domain Dedication waiver (http://​creat​iveco​ mmons.​org/​publi​cdoma​in/​zero/1.​0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 2 of 14 cases annually. In addition, approximately 2.1–6.5% of of Catalonia. In 2008, in line with promoting the clinical hospitalised patients experienced an ADE, correspond- safety of the patient, it designed and integrated a CDSS ing to 1.8–5.5 million annually [1]. Thus, since the late into its primary care electronic clinical workstation 1990s, patient safety has become a priority of health (ECW) that made it possible to detect certain medica- systems [2, 3], and several initiatives have identified the tion-related problems (MRPs) online. This CDSS, which need for a new culture of safety in the health and policy is known as PREFASEG (PREscripción FArmacéutica environment [4–6]. To achieve this, and according to SEGura, i.e., safe pharmaceutical prescription), is a com- the definition of clinical safety, it is essential to define puter tool that acts interactively to alert clinicians to any actions to avoid, prevent, and improve adverse effects or potential drug use-related problems during the process of injuries from healthcare processes where possible, since deciding the most appropriate treatment for their patient. it should be acknowledged that some adverse events are To understand the means by which PREFASEG func- inherent in treatment, and cannot always be avoided or tions, we consider the prescription of a new drug to a minimised. specific patient. At the point at which the prescription is In this context, the 1999 technical report ‘To err is requested by the clinician, PREFASEG is activated and human’ by the Institute of Medicine (OIM) highlighted performs an assessment of the prescription to verify that the need to develop new information and communica- it is safe for the patient, and that it does not pose a poten- tions technologies to reduce medical errors [2], and, tial risk to their health. This evaluation is carried out beyond this, prescriptions which could increase the risk based on the different MRPs detected by the tool, which of developing adverse effects. Subsequent reports later include: (1) Drug interactions; (2) Therapeutic duplica- affirmed that the electronic record of healthcare activ- tions; (3) Drugs advised against for use in geriatrics; (4) ity that is typical of an electronic health record (EHR), Contraindications with a safety alert published by the together with the integration of clinical decision support Spanish Agency for Medicines and Health Products systems (CDSSs) into these EHRs, should contribute to (AEMPS, Agencia Española del Medicamentos y Produc- guaranteeing quality in the healthcare system [7, 8] by tos Sanitarios); (5) Contraindications due to health prob- helping to reduce preventable adverse effects. lems and/or clinical variables; (6) Drugs that are known In the scientific literature we find different definitions to be teratogens during pregnancy; (7) Anticholinergic of a CDSS [9, 10]. According to Kawamoto et al. [11], a drug combinations; (8) Patient history of hypersensitiv- CDSS can be considered any electronic system designed ity or suspected hypersensitivity reactions (suspected, to help clinical decision making, which takes into account not confirmed); and (9) Adverse drug events. To carry the characteristics of the patient to generate a specific out this evaluation, a number of factors are taken into evaluation and provide a recommendation to be evalu- account, such as any active prescriptions that the patient ated by the practicing clinician. The design and function- already has on their record, other medical diagnoses or alities of these CDSSs can be very varied. Some authors active health problems, the presence of any clinical vari- consider that CDSSs aimed at the initial prescription ables with altered values, and the age and/or sex of the phase may have the greatest impact on improving patient patient. In the event of a safety alert being generated fol- safety [12], while others discuss the fact that integration lowing the above evaluation, the corresponding warnings of a CDSS into the HER renders it possible to provide are shown to the clinician (e.g., the risk to the patient patient histories along with interactive signals that alert and any therapeutic alternatives) so that he can decide professionals to situations of risk for the patient [13]. As a whether to continue with the prescription or change the result, the prescription process can be improved and the medication. These safety alerts are displayed in a simple clinical safety of patients enhanced [12, 14, 15]. manner on a single screen to permit their rapid consul- CDSSs have been found to bring multiple benefits to tation and understanding, as shown in Fig. 1, which pre- patient care [16–19], wherein it has been reported that sents an example relating to the prescription of a product they contribute to improving the dosage and selection of that is not recommended for patients over 75 years of drugs, while also encouraging patients to take part in pre- age. The information displayed includes the severity icon ventive activities, improving test results, decreasing mor- (two degrees, moderate or severe), the drug or active bidity, and improving the quality of care [20]. In contrast, ingredient causing the alert, the cause of conflict (i.e., the main risk of CDSSs is the alert fatigue experienced medication or active ingredient conflict, patient age, or by physicians who are faced with a multitude of prompts pre-existing health problem), the risk to the patient, and and reminders on-screen, which can lead to important any therapeutic alternatives. Each MRP alert is classified alerts being ignored [21, 22]. as either high (red indicator) or medium–low (orange The Catalan Health Institute (ICS) is a public entity indicator) clinical relevance, according to previously that provides health services to 80% of the population described recommendations [23]. Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 3 of 14 Fig. 1 Example PREFASEG screen showing the various informative elements presented by the tool (screenshot obtained from the PREFASEG ECW and translated into English) It is also possible that more than one alert is generated The contents of the MRP alerts are defined and main- by the tool, and in such a case, the clinician is informed tained as described by a multidisciplinary group of expert of all warnings associated with the different potential professionals from the ICS (i.e., primary care (PC) physi- safety issues, as can be seen in Fig. 2. cians, pharmacists, and clinical pharmacologists) accord- Each alert generated by PREFASEG is recorded ing to previous literature [9, 24]. The MRP alerts are as "audit data” along with the information related to reviewed and updated each year according to the avail- whether it has been accepted or whether prescription able scientific evidence, and as a result, the clinical con- of the product was continued. The evaluation of these tent undergoes some changes from 1 year to the next. data allows us to determine if certain types of alerts are For example, references to the more current STOPP/ accepted to a greater degree than others, thereby improv- START and Beers criteria were included in updates for ing the clinical content definitions to adapt the tool to drugs advised against for use in geriatrics. Importantly, the healthcare reality. PREFASEG consists of a calcula- the clinical content can be revised at any time, and are tion core in the Oracle PL/SQL, in which the calculations updated from a specific maintenance platform known as that access the tables of clinical contents have been opti- ’Know How.’ mised, and in which there is a minimum visual interface The purpose of this study is therefore to describe the for the communication of safety alerts. This interface was principal characteristics of the PREFASEG tool, the main developed using Developer Forms, which is the same safety alerts generated by PREFASEG in the Catalan PC technology employed to produce the ECW. As a result, system, the degree of acceptance of these alerts by phy- the look and feel of the alert screens are comparable to sicians, and the main pharmacological groups implicated those of the original ECW, and so maximum integration in the alerts. Furthermore, three of the most frequent is achieved. alerts are also described in greater detail. Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 4 of 14 Fig. 2 Example screen of the PREFASEG system with various included safety alerts and therapeutic recommendations (screenshot obtained from the PREFASEG ECW and translated into English) Methods ferent MRPs addressed by PREFASEG, wherein A descriptive, cross-sectional study was designed, which national and international articles that were consid- began in January 2016 and continued until December ered to be the most relevant and best adapted to our 2018. This study was developed within the scope of the healthcare environment were reviewed. The safety PC system of the ICS, which is the main entity that pro- alerts included in the ICS Clinical Practice Guide- vides health services in Catalonia, and which covers a lines [25] were also considered. population of 5.8 million inhabitants of the different Cat- 2. Consensus with a group of experts. Following a lit- alan territories through a network of 288 PC teams and erature review, the pharmacological groups to be 8 hospitals. The ICS is a public company with a total of included were selected and the messages to be pre- 42,374 professionals who provide services to 80% of the sented to the prescribing physicians were defined. population of Catalonia. Each MRP alert was classified as high (red) or medium-low (orange) clinical relevance, as men- Study sample tioned above [23]. The red alerts reflected situations The sample studied consisted of all prescriptions issued of absolute contraindications, while the orange alerts by the 6411 ICS PC physicians who used the EHR during were considered precautionary. the study period. 3. Adaptation of the clinical content to the table for- mats necessary for the PREFASEG computer pro- Development of PREFASEG gram. The clinical content was transferred into a In general terms, to provide the PREFASEG with clinical computer-readable language from various configura- and pharmacological content, the following methodology tion tables presented in Excel, and for this purpose, it was followed during its development: was necessary to code the active ingredients accord- ing to the Anatomical Therapeutic Chemical (ATC) classification. Similarly, the various health issues were 1. Bibliographic search. Initially a bibliographic search coded according to the International Classification was carried out in the PubMed database for the dif- of Diseases ICD-10 system. To produce the clinical Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 5 of 14 contents, ATC groups or groups of health problems some changes from 1 year to the next. Each MRP alert were built. Each alert type was then defined and con- was classified as either high or medium–low clinical rel- figured using a combination of various attributes, as evance, as described above [23]. outlined in Table 1. The MRP alerts corresponding to “therapeutic dupli- cations” detected patients with a non-beneficial pre- scription of two or more medicines based on the same Variables and indicators active ingredient (alone or in combination) and/or with The main variable of the study was the number of MRP the same pharmacological action (further details can be alerts generated by PREFASEG. Another of the variables found in Additional file 1). Duplications of more than 60 studied was the number of accepted alerts. An alert was different pharmacological groups commonly used in PC considered "accepted" when the medicine that generated were addressed. In each group, “clinically relevant dupli- the safety alert was not prescribed. cations” and “dose adjustments duplications” (combina- Some PREFASEG alerts are associated with recom- tions sought with a therapeutic objective) were clearly mendations for clinical follow-ups or dose reductions. differentiated. Depending on their relevance, alerts Therefore, following these recommendations does not marked with different colours were generated, as indi- entail the withdrawal of the treatment that has generated cated above. the alert. Consequently, these alerts are not considered During the study period, MRP alerts associated with “accepted” alerts. “AEMPS safety alerts” reported contraindications for the “Triple Whammy,” COXIBS, diclofenac, aceclofenac, Description of the MRPs included in PREFASEG cilostazol, ivabradine, agomelatine, escitalopram, citalo- The various MRPs that are defined in the PREFASEG pram, trimetazidine, raloxifene/bazedoxifene, strontium system were outlined previously in the introduction (see ranelate, aliskiren, and canagliflozin (further details can also Table 1). Further details regarding these MRPs can be found in Additional file 1). These alerts were consid- be found in Additional file 1. ered to be highly relevant because they were absolute The global MRP alerts generated and accepted by PRE- contraindications, in addition to having a specific safety FASEG were analysed. More specifically, the safety MRPs alert originating from the AEMPS, and so they were indi- related to therapeutic duplications, medicines not rec- cated in red. ommended for use in geriatrics, and safety alerts from The MRP alerts corresponding to “medicines the AEMPS were examined in greater detail. not recommended for use in geriatrics” detected The contents of the MRP alerts were defined and patients ≥ 75 years of age who had been prescribed inap- maintained as described by a multidisciplinary group propriate medication that posed a more unfavourable of expert professionals from the ICS according to previ- risk–benefit profile due to their age (see Additional file 1). ous literature [9, 24]. The MRP alerts were reviewed and The selection of medications considered inappropriate updated each year according to the available scientific for this age group was based on the Beers (2015) [26], evidence, and as a result, the clinical content underwent EU-PIM (European Consensus) [27], STOPP/START Table 1 Combinable attributes in the configuration of each PREFASEG notice Type of alert Combinable attributes in the PREFASEG message configuration Interactions ATC drug groups Duplicate therapies ATC drug groups AEMPS safety alerts Age ATC drug groups Grouping based on health problems Dose of the active ingredient that generates the warning Advised against for use in geriatrics Age Contraindications due to health issues ATC drug groups Contraindications due to clinical variables Grouping based on health problems Teratogens in pregnancy Labelling of clinical variables (e.g., glomerular filtration and potassium levels) Combinations of anticholinergic drugs ATC drug groups Suspicions of hypersensitivity ATC drug groups Adverse drug reactions ATC drug groups Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 6 of 14 [28], and PRISCUS [29] criteria (further details can be during the years of study. Analysis of the accumulated found in Additional file 1). These alerts were displayed number of alerts issued over the 3-year study period (i.e., on-screen as alerts of medium–low relevance (i.e., orange 4.38 million alerts) gave a 28% degree of acceptance (i.e., colour) since the literature indicates that they should be 1.22 million accepted alerts). administered with caution. When analysing the alerts generated from the different MRPs throughout the study period (2016–2018), it was Data collection and analysis observed that those related to drug interactions, thera- The analysed data were obtained from the ECW that peutic duplications, and drugs advised against for use in stores the active prescriptions of all patients; however, geriatrics were the most common. Taking the data col- data from specific patients were not analysed. The study lected over the 3 years, 39% (1,691,886) of the 4,379,866 was restricted to drugs prescribed and financed by the million alerts generated were for drug interactions, 33% National Health System for use in the PC setting. (1,436,721) were for therapeutic duplications, and 10% In January 2016, information began to be extracted (441,920) were for the use of drugs advised against in ger- regarding the different types of MRP alerts gener- iatrics. Thus, these three types of MRP alerts accounted ated by PREFASEG, which were internally identified in for more than three-quarters of the PREFASEG alerts the patient’s EHR. Thus, the number of advisories for (3,570,527; 82%). Of these, 27% were accepted and 73% each MRP generated, the medicines involved in each were ignored. In addition, of the 34,063 alerts related to alert, and the acceptance or rejection of the alert were teratogens in pregnancy, 22,324 (66%) were ignored. recorded. Each month, the alerts generated by the sys- The types of alerts with the highest percentage of tem and accepted by the clinicians were accumulated in acceptance were those related to a history of suspected a computer repository. The data set was analysed annu- (unconfirmed) drug hypersensitivity, with 35% (89,279) ally through computerised extractions from the ECW of these alerts being accepted over the 3 years stud- databases. The alert traceability was stored and organ- ied. Detailed analyses indicated that four non-steroidal ised on computer servers according to the organisational anti-inflammatory drugs (NSAIDs, i.e., ibuprofen, nap- structure of the ICS, i.e., with differentiation between the roxen, desketoprofen, and diclofenac) represented 45% health territories in which the institution is organised. (113,936) of the suspected hypersensitivity reactions A descriptive analysis was carried out of the generated reported by PREFASEG, with ibuprofen generating the and accepted alerts of the different MRPs from January highest number of alerts (61,026). In addition, during the 2016 to December 2018. Initially, the analysis was car- study period, the number of suspected hypersensitivity ried out on an annual basis because the clinical contents reaction alerts for β-lactam antibiotics alone or in com- changed annually. These content changes occurred for a bination fell into the second largest group, with 39,622 number of reasons, including the inclusion of new mar- alerts and an acceptance level of 63% (25,153). keted drugs, modifications in the definitions of existing In contrast, the alerts related to potential teratogenic MRP alerts to render them more specific, and the inclu- compounds during pregnancy had a lower degree of sion of additional pharmacological groups. Despite these acceptance (i.e., 34%, 11,739). It was observed that the content changes, the data were accumulated, and a global active ingredients that generated the most alerts were analysis of the alerts generated and accepted during the ibuprofen (10,864) and acetylsalicylic acid (4336) out of 3-year study period was also carried out. a total of 34,063. Overall, the alerts with the lowest degree of acceptance Results were those attributed to interactions between treatments, General analysis of the MRP alerts generated by PREFASEG with 25% (423,884) of a total of 1,691,886 being accepted. During the period of study, 22.5, 22.3, and 22.4 million More specifically, the interactions of NSAIDs with ace- new prescriptions were issued in the ICS PC system in tylsalicylic acid generated the greatest number of alerts, 2016, 2017, and 2018, respectively, while the number of reaching 220,507 alerts with an acceptance level of 18% alerts generated by PREFASEG were 1.17 million in 2016, (40,227). Table 3 outlines the 10 main interactions at the 1.43 million in 2017, and 1.77 million in 2018. Thus, the active ingredient level, which represent 30% (506,082) of percentage of MRP alerts generated by the tool with all alerts of this type. respect to the number of new prescriptions issued were 5% in 2016, 6% in 2017, and 8% in 2018. Analysis of the PREFASEG alerts related to therapeutic The global acceptance of these alerts varied throughout duplication the 3 years studied, ranging from 31% (362,732) in 2016 In the 3 years studied, the four groups of duplications to 26% (457,976) in 2018 (see Table 2), which corresponds that generated the most alerts were the NSAIDs, par- to 69–74% of the MRP alerts generated by PREFASEG acetamol-type analgesics, renin-angiotensin system Table 2 MRP alerts generated and accepted by PREFASEG between January 2016 and December 2018 PREFASEG 2016 2017 2018 Sum 2016–2018 Type of MRP Alerts Alerts % Alerts Alerts % Alerts Alerts % Alerts Alerts % alert generated accepted Accepted generated accepted Accepted generated accepted Accepted generated accepted Accepted alerts alerts alerts alerts Interactions 439,507 118,485 27 550,692 138,947 25 701,687 166,452 24 1,691,886 423,884 25 Duplicate 426,506 141,371 33 463,418 136,097 29 546,797 148,943 27 1,436,721 426,411 30 therapies Advised 108,974 32,345 30 144,807 41,740 29 188,139 50,299 27 441,920 124,384 28 due to age Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (> 75 years) AEMPS safety 59,146 18,301 31 86,308 21,958 25 84,158 18,720 22 229,612 58,979 26 alerts Contraindica‑ 35,164 13,771 39 67,452 21,330 32 105,749 28,390 27 208,365 63,491 30 tions due to health issues (2021) 21:349 Teratogens in 11,721 4200 36 10,938 3830 35 11,404 3709 33 34,063 11,739 34 pregnancy Combinations 1077 423 39 2619 863 33 3171 877 28 6867 2163 31 of anticholin‑ ergic drugs Suspicions of 76,883 28,952 38 84,099 29,759 35 93,852 30,568 33 254,834 89,279 35 hypersensi‑ tivity Adverse drug 15,397 4884 32 23,712 6927 29 36,489 10,018 27 75,598 21,829 29 reactions Totals 1,174,375 362,732 31 1,434,045 401,451 28 1,771,446 457,976 26 4,379,866 1,222,159 28 Page 7 of 14 Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 8 of 14 Table 3 Top 10 alerts related to drug interactions between January 2016 and December 2018 Original active ingredient Conflicting active ingredient Alerts generated Acceptance (%) Ibuprofen Acetylsalicylic acid 97,847 19 Amlodipine Simvastatin 85,358 24 Naproxen Acetylsalicylic acid 64,913 16 Simvastatin Amlodipine 59,453 26 Dexketoprofen Acetylsalicylic acid 32,455 16 Acenocoumarol Simvastatin 31,349 29 Simvastatin Acenocoumarol 31,349 26 Tramadol and paracetamol Citalopram 27,479 24 Tramadol and paracetamol Sertraline 26,872 24 Diclofenac Acetylsalicylic acid 25,292 24 Enoxaparin Acetylsalicylic acid 23,715 16 Fig. 3 Degrees of detection and acceptance of the main types of duplication from January 2016–2018 (RAS) inhibitors and gastric protectors (see Fig. 3). Out i.e., 76,409 over the 3 years studied, and their acceptance of a total of 65 groups, these 4 duplication groups rep- rate was relatively low at 23%. Upon the analysis of other resented 42% (600,930) of the total alerts generated groups of potentially dangerous duplications, such as (1,436,721). Duplications related to analgesics and gastric those related to oral anticoagulants, it was observed that protectors had the highest levels of acceptance during the PREFASEG generated 14,903 alerts of duplications for study period, reaching 42% (88,435) and 32% (33,245), this group, of which 42% (6314) were accepted, and the respectively. In contrast, duplications related to the SSRI prescriptions were not continued. antidepressants and the RAS inhibitor antihypertensives had the lowest degrees of acceptance, i.e., 20% (10,753) Analysis of PREFASEG alerts related to AEMPS safety alerts and 21% (20,940), respectively. It was found that the number of alerts related to AEMPS Furthermore, the pairs of active ingredients with the safety alerts increased throughout the study period, i.e., greatest numbers of alerts generated for duplications dur- from 59,146 in 2016 to 84,158 in 2018. This was accompa- ing the 3 years studied were paracetamol–paracetamol, nied by a reduction in the degree of acceptance of these paracetamol–paracetamol with tramadol, omeprazole– alerts from 31% (18,301) to 22% (18,720), respectively. omeprazole, ibuprofen–ibuprofen, and metamizole–met- Analysing the details of these AEMPS alerts (see amizole. These 5 pairs represented 22% (317,390) of the Table 4), it was apparent that the Triple Whammy, which duplications, of which 144,450 (46%) were accepted. considered the concomitant therapy of NSAIDs, diuret- Moreover, the duplications related to antibiotics such ics, and RAS inhibitors, generated the highest degree of as amoxicillin generated a considerable number of alerts, alerts, representing 85% (195,987) of the total alerts. The Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 9 of 14 Table 4 Degree of detection and acceptance of the AEMPS safety alerts in 2016, 2017, and 2018 2016 2017 2018 AEMPS alert Alerts Alerts accepted % Accepted Alerts Alerts accepted % Accepted Alerts Alerts accepted % Accepted "TRIPLE WHAMMY"* 46,020 13,867 30 76,107 18,992 25 73,860 15,988 22 DICLOFENAC 7288 2588 36 5734 1774 31 5137 1511 29 COXIBS 3131 888 28 2535 616 24 2986 644 22 ACECLOFENAC 805 204 25 665 127 19 487 101 21 ESCITALOPRAM 645 204 32 270 95 35 305 72 24 CITALOPRAM 495 189 38 217 67 31 302 104 34 CILOSTAZOL 193 70 36 181 66 36 289 65 22 AGOMELATINE 133 54 41 175 54 31 249 72 29 IVABRADINE 130 51 39 153 56 37 200 48 24 CANAGLIFLOZIN – – – 128 42 33 177 40 23 TRIMETAZIDINE 115 62 54 82 36 44 111 47 42 RALOXIFENE and 106 64 60 23 11 48 30 20 67 BAZEDOXIFENE STRONTIUM RENAL‑ 53 39 74 21 11 52 22 5 23 ATE ALISKIREN 32 21 66 17 11 65 3 3 100 Total 59,146 18,301 31 86,308 21,958 25 84,158 18,720 22 *Triple Whammy: NSAIDs + RAS inhibitors + diuretics degree of acceptance of this type of alert varied, rang- The two pharmacological groups that generated the ing from 30% (13,867) in 2016 to 22% (15,988) in 2018, highest number of alerts within this category were the thereby indicating a decrease in acceptance over this benzodiazepines and the NSAIDs, representing 39% 3-year period. Reviewing the main anti-inflammatory (172,574) and 21% (47,966) of a total of 441,920 alerts drugs that generated the Triple Whammy alerts, it was in the 3 years studied (see Table 5). More specifically, observed that in 65% of the cases, the NSAIDs involved the alerts related to benzodiazepine usage in geriat- were ibuprofen and naproxen. rics increased by 55% during the study period, i.e., from As indicated in Table 4, the number of alerts gener- 30,662 in 2016 to 68,974 in 2018. Alprazolam represented ated by diclofenac decreased over time. More specifically, 44% (65,910) of the alerts generated for benzodiazepines in 2016, diclofenac generated 7288 alerts, while by 2018 in this group of patients (see Table 6), and overall, the this number had reduced to 5137. However, the degree benzodiazepine group showed an acceptance rate of 27% of acceptance also decreased from year to year, dropping over the 3 years. from 36% (2588) in 2016 to 29% (1511) in 2018. Within the NSAID alerts, it was observed that desketo- The alerts with the highest degree of acceptance were profen represented 48% (45,293) of the total alerts in this those related to strontium ranelate, raloxifene/bazedox- group, and this group showed one of the lowest levels of ifene, and aliskiren; however the number of such alerts acceptance (i.e., 22%). was low since these are drugs that are being gradually In general, analysis of the degrees of acceptance in this withdrawn from the market, or tend to be unused in daily class of alerts shows a significant level of variation (see practice. Table 5), and the pharmacological groups with the high- est degree of acceptance (i.e., where no prescription was Analysis of the PREFASEG alerts related to medicines issued for the corresponding treatment) were the mus- not recommended for use in geriatrics between January cle relaxants (41%, 7521) and the peripheral vasodilators 2016 and December 2018 (39%, 6621). The alerts generated by PREFASEG that were related to drugs advised against for use in geriatrics increased Discussion throughout the study period, i.e., from 108,974 in 2016 The main finding of this study was that the PREFASEG to 188,139 in 2018. However, a reduction in the degree system appears to adopt the role of a CDSS that assists in of acceptance from 30% (32,345) to 27% (50,299), respec- preventing potential safety MRPs for patients by generat- tively, was also observed. ing online alerts when starting a new treatment. During Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 10 of 14 Table 5 Pharmacological groups not recommended in the elderly that generated PREFASEG alerts from January 2016–December 2018 Not recommended pharmacological groups Alerts generated Total accepted % Accepted Benzodiazepines, hypnotics, and sedatives 172,574 47,966 27 Anti-inflammatory and anti-rheumatic (NSAIDs, COXIBS) 94,034 20,627 22 Antihypertensives 34,542 10,310 30 Digestive system (otilonium, metoclopramide, glibenclamide, chlorporpa‑ 31,496 8945 28 mide) Chronic obstructive pulmonary disease treatments (theophylline) 19,285 4758 25 Central action muscle relaxants (cyclobenzaprine) 18,170 7521 41 Tricyclic antidepressants and Fluoxetine 17,289 5654 33 Peripheral vasodilators (pentoxifilline, nicergoline, nafthydrofuril) 16,958 6621 39 Respiratory system (systemic antihistamines) 14,219 4917 35 Hormone therapy (megestrol) 10,776 2631 24 Urinary antispasmodics (oxybutynin) 7018 2495 36 Antithrombotics (cilostazol) 3270 1100 34 Beta-blockers (sotalol) 1105 392 35 Opioid and anti-migraine pain relievers 868 342 39 Antiparkinsonian drugs 316 105 33 the period studied, it was observed that 28% of the gen- was reported [18], while other studies described omis- erated security alerts led to a modification of the pre- sions of recommendations in between 49 and 96% of the scription (i.e., acceptance of the alert). In absolute terms, cases [31]. In general, subsequent systematic reviews [32, between 2016 and 2018, a total of 1,222,159 recommen- 33] concluded that online notification systems had only a dations were accepted globally, which likely led to the small or moderate effect. avoidance of numerous potential MRPs in patients. Over- Our study presents a number of characteristics that all, PREFASEG reported a safety MRP in 1 out of every 15 could explain the relatively high acceptance rate of alerts. new prescriptions. The degrees of acceptance of the rec- For example, the observed degree of acceptance could ommendations were relatively high when compared with be accounted for by considering that the tool takes into a similar study into a different online preventive alert sys- account the success characteristics of CDSSs described tem, where the percentages of acceptance ranged from 12 by various authors previously [10, 34]. More specifically, to 14% [30]. However, the variability between studies was it is integrated into the workflow of clinicians, it gener- considerable; in a 2009 Cochrane review on the effects of ates an alert automatically during the parent consulta- online prompts/reminders, an improvement of only 4.2% tion and in the context of their medical history, and it provides a specific therapeutic recommendation in each case. In addition, PREFASEG produces different types of safety alerts based the clinical situation of each patient Table 6 Top 10 alerts for drugs advised against for use in and any medication that they may be taking, and it gives geriatrics between January 2016 and December 2018 alerts related to the safe prescription of medication for Active ingredient Alerts generated Acceptance various pathologies. To date, few studies have analysed responsible for the alert (%) such a diversity of alerts simultaneously [35–37]. It Alprazolam 65,910 24 should also be emphasised here that PREFASEG is a tool Dexketoprofen 45,293 24 whose clinical contents are constantly being updated and Doxazosin 33,867 30 that has been in use for more than 12 years, during which Clonazepam 23,887 35 time it has exhibited a good degree of acceptance by a Zolpidem 23,314 25 large number of medical professionals, since it is used Pentoxifylline 20,395 31 by > 6400 PC physicians on a daily basis. Indeed, both ICS Metoclopramide 19,098 28 pharmacists and PC clinical pharmacologists promote Hydroxyzine 18,328 29 the use of this tool. Etoricoxib 15,729 18 Despite the high number of potential MRPs avoided in Potassium clorazepate 15,468 29 patients in Catalonia, it should be noted that in the case Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 11 of 14 of safety alerts relating to the use of medicines, it was number of hospitals are now working on a project to de- striking that globally, > 70% of the alerts generated by label patients with a supposed hypersensitivity reported PREFASEG were ignored, which amounts to 3.16 million in their clinical history unless it is confirmed by the cor- over the 3-year study period. However, it must be con- responding tests. sidered that, on occasions, the recommendations given Another type of MRP that drew significant attention by PREFASEG did not necessarily imply a change, or the due to its severity was that of teratogenic drugs, for which change was not recorded as such. In addition, some rec- 66% of the generated alerts were ignored. However, it ommendations involved a dose reduction or a clinical fol- must be considered that not all medicines act as terato- low-up of some variable, and therefore the prescription gens in all trimesters of pregnancy, and PREFASEG is was continued. In such circumstances, the system did unable to distinguish between such cases. It is also possi- not count the recommendation as accepted, even if the ble that some alerts were generated for women who were dose was lowered or an analysis was requested. If these no longer pregnant but who, by some registration error, modifications had been registered, the degree of accept- maintained a pregnancy status in their health records. ance would be higher, as the presented acceptance values Regarding the alerts related to therapeutic duplications, relate only to cases where the prescription was contin- it was observed that approximately 70% of these alerts ued. It should also be considered that not all alerts had were ignored by clinicians. However, many such alerts the same degree of clinical relevance, with alerts being were related to adjustment of the daily dose of treatment, accompanied by either an orange or a red icon, depend- and so it was necessary to combine presentations at dif- ing on the importance of the recommendation, as also ferent doses; this was common in the groups of antihy- described in a previous study [23]. pertensive RAS inhibitors and antidepressants, and in the Over the course of the 3 years studied, the number of replacement of amoxicillin with amoxicillin-clavulanate. generated alerts increased. This was partly related to the In terms of the NSAIDs and paracetamol-type analgesics, fact that new and updated content was introduced into it was observed that prescriptions were authorised for the PREFASEG system on an annual basis. The existence issuing on demand if necessary, which often generated of a direct relationship between the increased consump- alerts related to duplication if an attempt was made to tion of certain drugs (e.g., the benzodiazepines) was also prescribe a drug from the same pharmacological group. considered, in addition to the probability that greater Another group of duplications that drew attention due numbers of MRP alerts could be generated from such to their association with a high risk of serious adverse drugs due to their increased use among the population. effects that motivate hospital admissions were the oral At the same time, it should be noted that the degree of anticoagulants [42]. During the 3-year study period, PRE- acceptance of the alerts tended to decrease over time, FASEG produced 14,903 alerts related to duplications in with practitioners gradually ignoring the recommenda- this group of drugs, which translated to an acceptance of tions. This decline in acceptance could be partly attrib- 42% (6314), wherein the prescription was not continued. uted to alert fatigue; however, it will be necessary to In the case of the AEMPS safety alerts, an unexpected further investigate the reasons behind the rejection of low degree of acceptance was recorded considering alerts, in addition to separately analysing the high and that these constituted specific alerts from a regulatory medium–low relevance alerts, while also considering agency [43]. In fact, throughout the 3-year study period, the cases where the recommendation does not suggest the degree of acceptance of the AEMPS safety alerts a change of drug. It will also be essential to collect the decreased, and in 2018 they reflected the lowest percent- opinions of the professionals who use the PREFASEG age of acceptance (22%) of all alerts throughout that year. tool. According to various reports, the main reasons for Among these notices, the Triple Whammy, which is asso- low acceptance by clinicians are the large number of low- ciated with a significant increase in the risk of kidney fail- relevance alerts they receive and their poor content [21, ure [44], represented the largest number of alerts. 22, 38]. To reduce the risk of fatigue, it is therefore neces- Upon examination of the alerts related to the use of sary to increase the specificity of the alerts, provide clear drugs advised against in geriatrics, it was observed that and concise information, and not impact on the clini- the degree of acceptance ranged from 30% in 2016 to 27% cian’s workflow. in 2018. Despite the fact that this alert category is con- In relation to the ignored alerts regarding suspicions of sidered of low clinical relevance, wherein use of a specific a history of hypersensitivity to certain drugs, it is known drug may not be recommended in older patients but is that general practitioners tend to register cases of hyper- not totally contraindicated, it produced similar or even sensitivity that are reported by patients, despite the fact superior acceptance results compared to the AEMPS that such hypersensitivity has not been confirmed, and in safety alerts. It was therefore considered that this level many cases, are not real [39–41]. To address this issue, a of acceptance was due to physicians being somewhat Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 12 of 14 more sensitive to safety alerts related to elderly patients. On the other hand, essential future work should also However, we must not lose sight of the fact that > 70% focus on analysing the acceptance of MRP alerts based of these alerts were discarded and the corresponding on their clinical relevance and the type of recommenda- prescriptions was generated, which could lead to poten- tion, which are key aspects to consider in the case of drug tial adverse reactions in patients. In this context, it is interactions. In addition, a satisfaction survey should be estimated that drug-associated adverse effects produce carried out to request feedback and suggestions from approximately 6.5% of hospital admissions, of which practitioners with regards to improving the PREFASEG more than half of these could be prevented [45–48]. system in terms of its clinical content and technological The pharmacological groups that generated the highest aspects. number of alerts in geriatric patients were the benzodiaz- The future development of PREFASEG also involves epines and the NSAIDs, which are also widely used drugs the inclusion of medicines that can only be prescribed throughout the population. The significant increase in the in hospitals and their corresponding contraindications, number of alerts for benzodiazepines (i.e., from 30,662 in which will allow the program to be extended to different 2016 to 68,974 in 2018) was particularly surprising, and levels of care. The technological evolution of the tool is these were mainly attributed to alprazolam. It is known also necessary to render it more specific when generat- that both an advanced age, which is linked to metabolic ing alerts. For example, this could be achieved using the and pharmacokinetic changes, and the number of drug terminology common to all SNOMED CT systems (Sys- treatments that a patient is receiving, are two of the situ- tematised Nomenclature of Medicine—Clinical Terms) ations that increase the risk of adverse drug effects to the that determine the active ingredient, the dose, the phar- greatest extent [49–51]. In addition, it must be consid- maceutical form, and the number of packaging units [57– ered that the world population is constantly aging, which 59]. To optimise the use of PREFASEG and improve the is accompanied by a greater degree of pathologies, and an management of clinical information, intelligent systems increase in the use of pharmaceuticals [52–54]. such as natural language processing could be applied Analysing the percentages of acceptance for alerts that would allow the clinician to obtain and interact with related to the use of drugs advised against in geriatrics, the information recorded in text format in the patient’s significant variation was observed between the different clinical history [60, 61]. An improved follow-up and pharmacological groups. More specifically, muscle relax- monitoring of the PREFASEG alerts would also be desir- ants and peripheral vasodilators were the groups with the able, wherein details regarding the professional receiving highest degrees of acceptance. According to a previous the alert are registered and made visible, in addition to study, physicians tend to prioritise alerts that are more whether this alert is ignored, and the level of care of the clinically relevant, or that can be resolved with the least corresponding professional. amount of time or effort [55]. In summary, PREFASEG appears to be a feasible and In a classic study looking at hospitalisations caused by efficient strategy to improve some aspects of clinical adverse effects, it was found that the majority occurred in safety related to the prescription of drugs, and as a result, the elderly, and were due to commonly used drugs with in the health care received by patients. well-known safety profiles [56]. Considering this point, which can likely be extrapolated to other countries, it Conclusions would be interesting to analyse the situation of patients Our study demonstrated that the PREFASEG (PREscrip- for whom PREFASEG detected a possible MRP that was ción FArmacéutica SEGura, i.e., safe pharmaceutical pre- not addressed. scription) clinical decision support system contributes to In terms of limitations to the current study, it should be the prevention of potential safety medicine-related prob- noted that the moderate percentage of alert acceptance lems in patients. In 28% of the cases in which the tool highlights the need to investigate the causes that lead generated a safety alert, primary care physicians modi- clinicians to discard such a high number of recommen- fied their prescriptions by some means. The main drug dations. Thus, to maximise the usefulness of PREFASEG groups implicated in the PREFASEG alerts were the non- and to avoid possible alert fatigue, it will be necessary to steroidal anti-inflammatory drugs, the benzodiazepines, carry out a detailed review into the traceability data of and the renin-angiotensin system inhibitors; groups that the tool to eliminate low-relevance alerts that are gener- frequently cause adverse effects and motivate hospital ated but not accepted, and to highlight any alerts related admissions. to therapeutic orientations while providing one or more In future, it will be necessary to study in detail the rea- alternative active ingredients. The introduction of a block sons behind the fact that > 70% of the generated alerts to prevent the continuation of a prescription associated were ignored by physicians. In addition, the possibility with a severe MRP could also be considered. of reducing the number of alerts should be assessed to Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 13 of 14 avoid alert fatigue. Moreover, it is evident that strategies Salut, Hospital Universitari Vall d’Hebron, Barcelona, Spain. 3 Departament de Farmacologia, Terapèutica i Toxicologia, Universitat Autònoma de Barcelona, must be designed to make the prescriber aware of the Barcelona, Spain. importance of patient safety, as well as to technologically improve the tool and render it more robust and specific. Received: 13 July 2021 Accepted: 21 November 2021 Abbreviations ADE: Adverse drug event; AEMPS: Spanish Agency for Medicines and Health Products (Agencia Española del Medicamentos y Productos Sanitarios); CDSS: References Clinical decision support system; ECW: Electronic clinical workstation; EHR: 1. European Commission. Commission staff working document. Phar‑ Electronic health record; ICS: Institut Català de la Salut (Catalan Health Insti‑ macovigilance related activities of Member States and the European tute); MRP: Medication-related problem; NSAIDs: Non-steroidal anti-inflamma‑ Medicines Agency concerning medicinal products for human use (2012– tory drugs; PC: Primary care; PREFASEG: PREscripción FArmacéutica SEGura, i.e., 2014). https://​ec.​europa.​eu/​healt​h//​sites/​health/​files/​files/​pharm​acovi​ safe pharmaceutical prescription; RAS: Renin-angiotensin system. gilan​ce/​pharm​acovi​gilan​ce-​report-​2012-​2014_​annex​en.​pdf. Accessed 19 Oct 2019. 2. Institute of Medicine (US) Committee on Quality of Health Care in Ameri‑ Supplementary Information can. In: Kohn LT, Corrigan JM, Donaldson MS, editors. To err is human: The online version contains supplementary material available at https://​doi.​ building a safer health system. Washington, DC: National Academies Press org/​10.​1186/​s12911-​021-​01710-8. (US); 1999. 3. Council of Europe Expert Group on Safe Medication Practices. Creation of a better medication safety culture in Europe: building up safe medication Additional file 1. Annex Table 1. Description of the types of PREFASEG practices. Strasbourg: Council of Europe; 2006. alerts; Annex Table 2. Groups of pharmacological duplications included 4. National Patient Safety Agency. Seven steps to patient safety for primary in PREFASEG in 2018; Annex Table 3. Definitions of the AEMPS safety care. London: National Patient Safety Agency; 2006. alerts included in PREFASEG in 2018; Annex Table 4. List of drugs not 5. World Health Organization (WHO). World alliance for patient safety. recommended for use in geriatrics in 2018. Geneva: World Health Organization; 2007. 6. Food and Drug Administration. Safe use initiative: collaborating to reduce preventable harm from medications. Geneva: US Department of Health Acknowledgements and Human Service, Food and Drug Administration; 2009. This study has been prepared within the framework of the Doctorate in 7. Institute of Medicine (US) Committee on Quality of Health Care in Pharmacology at the Universitat Autònoma de Barcelona (UAB). We would like America. Crossing the quality chasm: a new health system for the 21st to thank ICS Information Systems for supporting us during acquisition of the century. Washington, DC: National Academies Press (US); 2001. information required to conduct this study. We would also like to thank the 8. Institute of Medicine (US) Committee on Identifying and Preventing PREFASEG working group for helping to maintain a useful and up-to-date tool. Medication Errors. Quality chasm series. Preventing medication errors. Washington, DC: National Academies Press (US); 2007. Authors’ contributions 9. Bates DW, Kuperman GJ, Wang S, Gandhi T, Kittler A, Volk L, et al. Ten MAP extracted, analysed, and interpreted the data on the use of the PREFASEG commandments for effective clinical decision support: making the tool. She also wrote the majority of the manuscript. MO helped to interpret practice of evidence-based medicine a reality. J Am Med Inform Assoc. the data on the use of the PREFASEG tool, in addition to giving support and 2003;10(6):523–30. making contributions to the writing of the manuscript. AF and ED reviewed 10. Payne TH. Computer decision support systems. Chest. 2000;118(2 the data analysis process, reviewed the various parts of the manuscript, and Suppl.):47S-52S. made the relevant contributions to give clarity and understanding to the 11. Kawamoto K, Houlihan CA, Balas EA, Lobach DF. Improving clinical prac‑ study. All authors read and approved the final manuscript. tice using clinical decision support systems: a systematic review of trials to identify features critical to success. BMJ. 2005;330(7494):765–8. Funding 12. Bates DW, Cullen DJ, Laird N, Petersen LA, Small SD, Servi D, et al. The primary care component of this study was funded by the Institut Uni‑ Incidence of adverse drug events and potential adverse drug events: versitari d’Investigació en Atenció Primària (IDIAP Jordi Gol) Scholarship to implications for prevention. JAMA. 1995;274(1):29–34. complete a PhD thesis in primary care. 13. McCoy AB, Waitman LR, Lewis JB, Wright JA, Choma DP, Miller RA, et al. A framework for evaluating the appropriateness of clinical decision support Availability of data and materials alerts and responses. J Am Med Inform Assoc. 2012;19(3):346–52. The datasets used and/or analysed during the current study are available from 14. Bates DW, Spell N, Cullen DJ, Burdick E, Laird N, Petersen LA, et al. The the corresponding author on reasonable request. costs of adverse drug events in hospitalized patients. Adverse Drug Events Prevention Study Group. JAMA. 1997;277(4):307–11. 15. Bobb A, Gleason K, Husch M, Feinglass J, Yarnold PR, Noskin GA. The Declarations epidemiology of prescribing errors: the potential impact of computerized prescriber order entry. Arch Intern Med. 2004;164(7):785–92. Ethics approval and consent to participate 16. Hunt DL, Haynes RB, Hanna SE, Smith K. Effects of computer-based Not applicable. This manuscript does not report studies involving human clinical decision support systems on physician performance and patient participants, human data or human tissue. This manuscript does not report outcomes: a systematic review. J Am Med Assoc. 1998;280(15):1339–46. studies involving animals. 17. Wasylewicz ATM, Scheepers-Hoeks AMJW. Clinical decision support systems. In: Kubben P, Dumontier M, Dekker A, editors. Fundamentals of Consent for publication clinical data science. Cham: Springer International Publishing; 2018. p. Not applicable. 153–69. 18. Shojania KG, Jennings A, Mayhew A, Ramsay CR, Eccles MP, Grimshaw J. Competing interests The effects of on-screen, point of care computer reminders on processes The authors declare that they have no competing interests. and outcomes of care. Cochrane Database Syst Rev. 2009;3:CD001096. 19. Moja L, Kwag KH, Lytras T, Bertizzolo L, Brandt L, Pecoraro V, et al. Effec‑ Author details tiveness of computerized decision support systems linked to electronic 1 Unitat de Coordinació i Estratègia del Medicament (UCEM), Institut Català de health records: a systematic review and meta-analysis. Am J Public la Salut, Barcelona, Spain. 2 Servei de Farmacologia Clínica, Institut Català de la Health. 2014;104(12):e12-22. Pons‑Mesquida et al. BMC Medical Informatics and Decision Making (2021) 21:349 Page 14 of 14 20. Coma E, Medina M, Méndez L, Hermosilla E, Iglesias M, Olmos C, et al. from The Health Improvement Network. Pharmacoepidemiol Drug Saf. Effectiveness of electronic point-of-care reminders versus monthly feed‑ 2013;22(2):176–82. back to improve adherence to 10 clinical recommendations in primary 43. Agencia Española de Medicamentos y Productos Sanitarios. Notas care: a cluster randomized clinical trial. BMC Med Inform Decis Mak. informativas de la AEMPS. Madrid (España): Agencia Española de Medi‑ 2019;19(1):245. camentos y Productos Sanitarios (AEMPS). https://​www.​aemps.​gob.​es/​ 21. Judge J, Field TS, DeFlorio M, Laprino J, Auger J, Rochon P, et al. Prescrib‑ accio​nes-​infor​mativ​as/​notas-​infor​mativ​as-​de-​la-​aemps/. Accessed 19 Apr ers’ responses to alerts during medication ordering in the long term care 2021. setting. J Am Med Inform Assoc. 2006;13(4):385–90. 44. Lapi F, Azoulay L, Yin H, Nessim SJ, Suissa S. Concurrent use of diuretics, 22. van der Sijs H, Mulder A, van Gelder T, Aarts J, Berg M, Vulto A. Drug angiotensin converting enzyme inhibitors, and angiotensin receptor safety alert generation and overriding in a large Dutch university medical blockers with non-steroidal anti-inflammatory drugs and risk of acute centre. Pharmacoepidemiol Drug Saf. 2009;18(10):941–7. kidney injury: nested case-control study. BMJ. 2013;346:e8525. 23. Paterno MD, Maviglia SM, Gorman PN, Seger DL, Yoshida E, Seger AC, 45. Aranaz-Andrés JM, Aibar-Remón C, Vitaller-Murillo J, Ruiz-López P, Limón- et al. Tiering drug–drug interaction alerts by severity increases compli‑ Ramírez R, Terol-García E. Incidence of adverse events related to health ance rates. J Am Med Inform Assoc. 2009;16(1):40–6. care in Spain: results of the Spanish National Study of Adverse Events. J 24. Sittig DF, Wright A, Osheroff JA, Middleton B, Teich JM, Ash JS, et al. Epidemiol Community Health. 2008;62(12):1022–9. Grand challenges in clinical decision support. J Biomed Inform. 46. Aranaz-Andrés JM, Aibar C, Limón R, Mira JJ, Vitaller J, Agra Y, et al. A study 2008;41(2):387–92. of the prevalence of adverse events in primary healthcare in Spain. Eur J 25. Institut Català de la Salut ICS Clinical Practice Guidelines. http://​ics.​gencat.​ Public Health. 2012;22(6):921–5. cat/​ca/​assis​tencia/​conei​xement-​assis​tenci​al/​guies-​de-​pract​ica-​clini​ca/. 47. Howard RL, Avery AJ, Slavenburg S, Royal S, Pipe G, Lucassen P, et al. Accessed 28 Oct 2021. Which drugs cause preventable admissions to hospital? A systematic American Geriatrics Society 2019 updated AGS Beers ­Criteria® for poten‑ 26. Fick DM, Semla TP, Steinman M, Beizer J, Brandt N, Dombrowski R, et al. review. Br J Clin Pharmacol. 2007;63:136–47. 48. Pirmohamed M, James S, Meakin S, Green C, Scott AK, Walley TJ, et al. tially inappropriate medication use in older adults. J Am Geriatr Soc. Adverse drug reactions as cause of admission to hospital: prospective 2019;67(4):674–94. analysis of 18,820 patients. Br Med J. 2004;329(7456):15–9. 27. Renom-Guiteras A, Meyer G, Thürmann PA. The EU(7)-PIM list: a list 49. Cullinan S, O’Mahony D, Fleming A, Byrne S. A meta-synthesis of of potentially inappropriate medications for older people consented potentially inappropriate prescribing in older patients. Drugs Aging. by experts from seven European countries. Eur J Clin Pharmacol. 2014;31(8):631–8. 2015;71(7):861–75. 50. Gavilán-Moral E, Villafaina-Barroso A, Jiménez-de Gracia L, del Gómez 28. O’Mahony D, O’Sullivan D, Byrne S, O’Connor MN, Ryan C, Gallagher P. Santana MC. Ancianos frágiles polimedicados: ¿Es la deprescripción de STOPP/START criteria for potentially inappropriate prescribing in older medicamentos la salida? Rev Esp Geriatr Gerontol. 2012;47(4):162–7. people: version 2. Age Ageing. 2015;44(2):213–8. 51. Page RL, Linnebur SA, Bryant LL, Ruscin JM. Inappropriate prescribing in 29. Holt S, Schmiedl S, Thürmann PA. Potentially inappropriate medications the hospitalized elderly patient: defining the problem, evaluation tools, in the elderly: the PRISCUS list. Dtsch Arztebl. 2010;107(31–32):543–51. and possible solutions. Clin Interv Aging. 2010;5(1):75–87. 30. Dexheimer JW, Talbot TR, Sanders DL, Rosenbloom ST, Aronsky D. Prompt‑ 52. Turgeon J, Michaud V, Steffen L. The dangers of polypharmacy in elderly ing clinicians about preventive care measures: a systematic review of patients. JAMA Intern Med. 2017;177(10):1544. randomized controlled trials. J Am Med Inform Assoc. 2008;15(3):311–20. 53. Onder G, Marengoni A. Polypharmacy. JAMA. 2017;318(17):1728. 31. Bright TJ, Wong A, Dhurjati R, Bristow E, Bastian L, Coeytaux RR, et al. Effect 54. Payne RA. The epidemiology of polypharmacy. Clin Med J R Coll Phys of clinical decision-support systems: a systematic review. Ann Intern Med. Lond. 2016;16(5):465–9. 2012;157(1):29–43. 55. Loo TS, Davis RB, Lipsitz LA, Irish J, Bates CK, Agarwal K, et al. Electronic 32. Cheung A, Weir M, Mayhew A, Kozloff N, Brown K, Grimshaw J. Overview medical record reminders and panel management to improve primary of systematic reviews of the effectiveness of reminders in improving care of elderly patients. Arch Intern Med. 2011;171(17):1552–8. healthcare professional behavior. Syst Rev. 2012;1(1):36–43. 56. Budnitz DS, Lovegrove MC, Shehab N, Richards CL. Emergency hospi‑ 33. Ivers N, Jamtvedt G, Flottorp S, Young JM, Odgaard-Jensen J, French SD, talizations for adverse drug events in older Americans. N Engl J Med. et al. Audit and feedback: effects on professional practice and healthcare 2011;365(21):2002–12. outcomes. Cochrane Database Syst Rev. 2012;6:CD000259. 57. Jiang G, Solbrig HR, Chute CG. ADEpedia: a scalable and standardized 34. Osheroff JA. Improving medication use and outcomes with clinical deci‑ knowledge base of adverse drug events using semantic web technology. sion support: a step by step guide. Chicago: HIMSS; 2009. AMIA Annu Symp Proc. 2011;2011:607–16. 35. Van Wyk JT, Van Wijk MAM, Sturkenboom MCJM, Mosseveld M, Moorman 58. Farfán Sedano FJ, Terrón Cuadrado M, García Rebolledo EM, Castellanos PW, Van Der Lei J. Electronic alerts versus on-demand decision support Clemente Y, Serrano Balazote P, Gómez DÁ. Implementation of SNOMED to improve dyslipidemia treatment: a cluster randomized controlled trial. CT to the medicines database of a general hospital. Stud Health Technol Circulation. 2008;117(3):371–8. Inform. 2009;148:123–30. 36. Clyne B, Bradley MC, Hughes C, Fahey T, Lapane KL. Electronic prescribing 59. Guía de Conjuntos de Referencias de SNOMED CT para el SNS. Centro and other forms of technology to reduce inappropriate medication use Nacional de Referencia para SNOMED CT: Ministerio de Sanidad, Servicios and polypharmacy in older people: a review of current evidence. Clin Sociales e Igualdad; 2016. Geriatr Med. 2012;28(2):301–22. 60. Wei WQ, Cronin RM, Xu H, Lasko TA, Bastarache L, Denny JC. Develop‑ 37. Upperman JS, Staley P, Friend K, Neches W, Kazimer D, Benes J, Wiener ES. ment and evaluation of an ensemble resource linking medications to The impact of hospitalwide computerized physician order entry on medi‑ their indications. J Am Med Inform Assoc. 2013;20(5):954–61. cal errors in a pediatric hospital. J Pediatr Surg. 2005;40(1):57–9. 61. Fung KW, Jao CS, Demner-Fushman D. Extracting drug indication infor‑ 38. Tamblyn R, Huang A, Taylor L, Kawasumi Y, Bartlett G, Grad R, et al. A mation from structured product labels using natural language process‑ randomized trial of the effectiveness of on-demand versus computer- ing. J Am Med Inform Assoc. 2013;20(3):482–8. triggered drug decision support in primary care. J Am Med Inform Assoc. 2008;15(4):430–8. 39. Trubiano JA, Adkinson NF, Phillips EJ. Penicillin allergy is not necessarily Publisher’s Note forever. JAMA. 2017;318(1):82–3. Springer Nature remains neutral with regard to jurisdictional claims in pub‑ 40. van Dijk SM, Gardarsdottir H, Wassenberg MW, Oosterheert JJ, de Groot lished maps and institutional affiliations. MC, Rockmann H. The high impact of penicillin allergy registration in hospitalized patients. J Allergy Clin Immunol Pract. 2016;4(5):926–31. 41. Wilcock M, Powell N. Challenges of delabelling penicillin allergy. Droga Ther Bull. 2021;59(11):163–4. 42. Gaist D, Wallander MA, González-Pérez A, García-Rodríguez LA. Incidence of hemorrhagic stroke in the general population: validation of data