key: cord-0808069-89d2gzhy authors: Rodríguez-Terol, A.; Caraballo, M. O.; Palma, D.; Santos-Ramos, B.; Molina, T.; Desongles, T.; Aguilar, A. title: Quality of interaction database management systems date: 2009-12-31 journal: Farmacia Hospitalaria (English Edition) DOI: 10.1016/s2173-5085(09)70079-6 sha: b72a14fd8d8a90efaee1aef44480df5839b2348c doc_id: 808069 cord_uid: 89d2gzhy Abstract Objective To identify drug interaction databases (DID) and assess the quality of their structures. Method A search was made of the literature for DID and a series of exclusion and structural quality criteria were defined (at least 4 quality criteria: classification according to severity, classification according to level of evidence, bibliographical reference data, description of clinical management, and 11 criteria used for weighting). The level of compliance of every DID with the criteria defined was analysed, together with the level of compliance of each criteria in each DID. Results A total of 54 DID were identified, 30 of which complied with exclusion criteria and 15 of which did not meet the minimum criteria. The rest of the criteria were evaluated in 9 DID: Botplus and Medinteract (100%), SEFH Guide, Lexi-interact and Medscape (89%), Hansten (83%), Micromedex and Stockley (78%), Drug Interactions Facts (68%). Ninety-two percent of the DID describe the mechanism of action, 87% classify the information according to the active ingredient, 75% do not state they have any conflict of interest, classify according to level of severity, have electronic format, and are easy to search. A total of 67% are specific DID, 62% are classified according to level of evidence, contain bibliographical references, and describe clinical management. Conclusions A third of the DID comply with the minimum criteria. Differences were observed in the level and compliance criteria among Spanish and foreign DID. Some of the main DID used as references in the bibliography have significant structural defects: no web presentation, no multi-check function and others. Interactions between medications administered to a patient contribute to concomitant morbi-mortality and, in many cases, could be preventable. A study carried out in Denmark upon 26 337 patients with at least 2 prescribed medications detected 21 293 different combinations, of which 4.4% carried a risk of producing a severe interaction. In this same study, 1.2% of hospitalisations were related to medicinal interactions. 1 In Spain, the APEAS study 2 found that 47.8% of adverse events detected in the primary health care field were due to medications, of which 3.5% were a consequence of medicinal interaction. Another published study reveals that 9.9% of the population over 65 years of age is at risk of clinically significant interactions. The study notes that there is an exponential growth in the risk of interactions being produced with a higher number of medications. 4, 6 Polymedication could therefore present a risk of interaction. In Australia 14% of the general population uses more than 4 medications, and in the population over 75 years of age this figure increases to 40%. Data from the UK indicates that 30% of the population over 75 years takes more than 4 medicines. In Spain, a study carried out in a rural area with basic health care indicated that 11.37% of the population was over 65, with an ageing population of 65% and an average prescription rate of 4 medications, and a greater number of prescribed medicines tallying with increased age. 7 However, management of medicinal interactions in clinical consultation is not easy. The introduction of new technologies PALABRAS CLAVE Interacciones medicamentosas; Bases de datos; Evaluación de calidad in primary health care and hospitals has brought a development in the form of computerised clinical history, which has opened up the possibility of incorporating decision support systems (DSS) with regard to interactions, which alert the user at the moment of prescribing medicines and report on possible courses of action. However, the introduction of these systems is not yet widespread. According to an investigation carried out in Spain in 2007, computer-assisted prescription is in place in only 22.4% of hospitals. 8 In primary health care, the development of electronic prescription has not apparently been accompanied (thus far) by tools for the clinical management of interactions. However, many have incorporated complete databases in consultation format, in order that the clinic may utilise them at their discretion and in specific cases. In the absence of a DSS, any clinic that wishes to carry out a systematic follow-up of medicinal interactions must manage by itself the data sources and their assigned clinical relevance, ie, the influence which the data will have upon any modification of the therapeutic plan. And it is here where the range in databases and sources of information regarding interaction is such that it usually becomes impossible to manage physically. Furthermore, in a study carried out on just 5 databases, 9 it was found that the quality was very unevenly spread and the concordance was scarce, making it difficult to pinpoint real clinical importance in each of the interactions. 10 The objective of this study is to assess the structural quality of various drug interaction databases (DID) in order to be able to subsequently create a decision support system. In order to identify existing interaction databases, a bibliographic search and exploration of grey literature was carried out. The bibliographic search was performed on M E D L I N E u s i n g t h e f o l l o w i n g k e y w o r d s : " d r u g , " "database*," and "interaction*." Subsequently, all the bibliographic citations found in the works obtained were reviewed. The search of grey literature was carried out using general information Internet search engines, using the following search terms: "drug," "database*," and "interaction." Databases detailing interactions with no clinical practice, interactions with food, medicinal plants, or other products, results in languages other than English, French, or Spanish, results contained in systems covering very small localised areas, results containing information regarding only one group of medications, interactions concerning new drugs still in development or drugs which are not readily available for purchase or prescription, and medicines designed for the PDA since they compile their information from more general DIDs. For the databases included in the study which were not freely available, an access licence was obtained or the relevant book or cD was purchased, as appropriate. Given that it was not possible to locate suitable references, the researchers themselves established the evaluation criteria. The criteria used were diverse: -Descriptive criteria: date of first edition, price, language, and number of interactions described. These factors were not used for quality evaluation -criteria used for evaluation (Table 1) . Two types of criteria were used in turn: a) minimum quality criteria, ie, any database which does not meet these criteria is discarded for later evaluation (4 criteria), and b) criteria which add weighting to the evaluation (12 criteria). The latter, in turn, were divided into 2 groups according to the relative importance assigned by the research group: 7 criteria with a weighting of 10.76% (which in total counted as 75% of the evaluation) and 4 criteria with a weighting of 6.25% (making up 25% of the total evaluation) Each criterion was assigned a score, as detailed in Table 1 . Only those DIDs which met the minimum criteria were selected for the subsequent phase of the study, which consisted of assessing whether the remaining criteria were met, and assigning a score to the general level of fulfilment. Two types of analysis were performed: a) for each DID the degree of compliance with the structural quality criteria was determined, and b) for each structural quality criterion the degree of compliance in different databases. This last analysis was carried out on all the selected DIDs and for the resulting division strata according to language or compliance with the minimum criteria. A total of 54 databases were identified, 37 from citations in articles found on MEDLINE and 18 from informal searches. Twenty-four of these databases fulfilled the inclusion criteria. Those databases which were excluded are detailed in Appendix 1, 42-82 no Spanish DID was excluded. Of the selected DIDs, 6 were edited in Spain, 14 in the United States, 3 in the United Kingdom, and 1 in France. Among the Spanish databases, 1 was recovered from MEDLINE and the remaining 5 from grey literature. Nine DIDs met the minimum quality criteria (Table 2) , whereas 15 did not and, therefore, the remaining quality criteria were not applied to these. Table 3 28-41 summarises the characteristics of these unevaluated DIDs. Among the databases which did not fulfil the minimum criteria, 3 did not meet any of the 4 criteria, 2 failed to meet 3 criteria, 5 did not meet 2 criteria, and 5 failed to meet just 1 criterion. With regard to structural quality, the values obtained for the different DIDs which exceeded the minimum criteria were: Bot-plus 11 and Medinteract 12 (100%), Guía de la SEFH, 13 Lexi-interact 14 and Medscape 15 (89%), Hansten 16 (83%), Micromedex 17 and Stockley 18 (78%), and Drug Interaction Facts 19 (67%). Table 4 summarises the degree of compliance with each criteria for all of the analysed DIDs. The stratification of the degree of severity is the most common criterion overall. The Spanish DIDs have more of a tendency to include a description of severity, bibliographical reference, and description of clinical management, whereas the DIDs of other countries more frequently include classification of the level of evidence. A large number of databases concerning medicinal interactions exist internationally. More than half are inaccessible or of no clinical interest. Of the 24 selected databases, only 17 were in English and, therefore, the selection can be considered as global in character. The discovery of such a high number of databases concerning medicinal interactions apparently makes clear that, firstly, this is a high-interest area of pharmacotherapy and, secondly, there appears to be no defined international standard. 20 Six databases were available in Spanish, of which 3 met the minimum quality requirements. Although not global in character, these can be considered a reference for the vast Latin American territories. Given that the majority of these were recovered from the grey literature, it is logical to assume that there should be a similar pattern in other languages; in other words, there are databases in any given language which are not referred to in scientific articles and, therefore, they are difficult for researchers of other languages to find. 21 Along with the issue of language, the question of which drugs are included is also pertinent, not just due to their quantity but also their relevance. In other words, those DIDs which include all medications from a specific market (eg, Bot-Plus or Medinteract, which include all medicines registered in Spain), can be of more use in that particular market compared with other databases which may have more medicines listed yet exclude some which are commercialised in that country. It is surprising that only 9 of the 24 DIDs selected complied with the required minimum quality criteria. Among those which did not comply are some of those DIDs most used as a reference for works in the field of drug interaction detection, both in primary health care environments and hospitals. The Spanish databases (both included and excluded) fulfil more of the minimum criteria than others; their weakest point is classification according to level of evidence. The non-Spanish databases place more emphasis on the stratification of degree of severity. Among the 9 DIDs which met the minimum criteria, the heterogeneity in the classification format of these criteria should be highlighted. Severity is classified into 2, 3, and even 4 levels, and neither the description or the underlying concept tally in virtually any of the databases. This presents a serious problem for the standardisation of this important issue. Particularly noteworthy is the fact that some DIDs do not stratify the degree of severity (Medical Letter, the most significant due to its widespread dissemination). The same thing occurs with the classification of level of evidence. In some DIDs reference is made to the type of article supported, whereas in others a classification is made by the authors, in general not referring to the levels of Data represents n (%) of the total of each group. evidence previously published but those produced ad hoc. For example, Micromedex ® classifies as theoretical/probable; Medinteract ® classifies as well-documented/documented/ scarcely-documented, etc, and makes no reference even to the articles in which the relevant interaction is discussed. All of this, once again, indicates a high level of variability. It should be noted that all the included DIDs were specific to interactions, whereas many among the excluded DIDs belonged to databases contained a wider range of data. Perhaps in the latter there was less physical space available. The criteria defined as non-essential were met by a higher number of databases than the minimum criteria. Half the authors or promoters were public and half private; a direct correlation is found between a private background and a higher score (data not shown). One aspect to emphasise is that some DIDs offer the possibility of seeing the structural features in a preview prior to purchase. It is even possible with some databases to view a sample prior to purchase (Medinteract, Stockley, Lexi). However, others offer up very little information to enable an evaluation prior to a formal purchase (Hansten, Drug Interaction Facts) . The preferred format is a webpage, which is logical as it allows ubiquitous access and constant updating. Book form is bothersome for consulting data, as it is costly to update and is of no use when incorporating data into expert systems. However, book form is the only option for Drug Interaction Facts and Hansten, 2 very reliable and oftencited DIDs. 22.23 This could be due to cultural factors or to a failure to update the format of older DIDs, designed in an age when the criteria were less strict. The period between updates is specified only in 12 of the 24 DIDs compared, with a very wide range, from immediate updates to a period of three years for each update of the Martindale DID. 24 Update intervals of more than 1 year should not be admissible, and demands should perhaps be made for more frequent DID updates using the Internet. The description of significant interactions detected in clinical studies, concerning commercially-available medicines, and the rapid detection of others during the post-commercialisation phase, make this aspect increasingly important. Only half of the databases have a multicheck structure, ie, the introduction of several medications at once in order to produce an analysis (a higher proportion among the Spanish and included databases), which seems a low percentage in the era of informatics if they are to be used in clinical practice, where the number of polymedicated patients is constantly increasing. This option is impossible with those databases in book format and is unavailable with two of the most well-known databases, Drug Interaction Facts and Hansten. With Micromedex, this option is only available in addition to payment for the Drugdex ® DID. The only work similar to ours found in the bibliography compared 5 DIDs relating to the United States. The highestscoring DID is Walgreens, to which we had no access. Medscape and DrugReax achieved high scores, which tallies with our study. However, the second highest scoring database is DrKoop, which was excluded from our study due to not meeting all the minimum essential criteria. One of the limitations of our work is the possibility of a slant towards detection of Spanish databases in the informal search. An attempt has been made to locate different DIDs which are available on the open market, and analyse their quality from the perspective of a Spanish professional, a method which this publication tries to achieve. Another possible limitation is the fact that the criteria and weightings used were established by the authors. However, it is worth noting that up to now it has not been possible to find any generally accepted previous classification or evaluation. Minh et al 9 describe content criteria (accuracy, complete data, references, language, and interaction management) and evaluation of usefulness (ease of use, speed, multicheck, multifunctionality). The study, which analysed just 5 DIDs (Drug Pharmacology, DrKoop, Medscape, Walgreens, and DrugReax), uses 9 quality criteria, all with the same value. In our study 20 criteria have been used, of which the following tally with Minh et al: ease of searching, multicheck, multifunctionality, references, language, and interaction management. The total number of interactions was not considered as a criterion. In our judgement, it is important that a DID contains a large number of interactions. However, when evaluating the possibility of incorporating a DID into educative programmes, electronic prescription systems or a clinical task, it is possible that certainty, clinical significance and help with decision-making are more relevant. certainty makes reference to the fact that in a medical setting based on evidence, the interaction should have sufficient bibliographical references and the DID authors have classified the interactions according to some scale portraying level of evidence, as is seen in Drug Interaction Facts. The relevance assumes that some scale of severity is used, as can be found in Medinteract or Lexi. The abundance of medicinal interactions leads to, in some clinical practice environments, the need to prioritise attention towards those which are most severe. Particularly with computerised systems, it is necessary to obtain a "good" interaction signal or noise, for which criteria of severity is essential. 25 Lastly, knowledge of medicinal interaction is especially important if clinical action is to take place in order to prevent its occurrence. For this reason is seems essential for authors that the DIDs include a concrete description of the clinical management of a patient suffering with the relevant interaction, as can be found in, for example, Lexi or Micromedex. On the other hand, it is true that the scarce number of DIDs which fulfil these criteria could be evidence of an excessive strictness in definition on our part, and that other criteria, as fulfilled in 75% of databases, could be included in more databases in the analysis. However, for the reasons described above, it seemed necessary to require all the DIDs to comply with all of the four selected criteria. Another important task is to evaluate the clinical significance of each interaction, since no standard protocol could be found for the allocation of such significance. Each DID has its own protocol, as can be seen with Drug Interaction Facts or Hansten, which depends particularly on the severity and scientific evidence of the interaction in question. Recently a study was published which attempted to create a procedure for establishing the clinical significance of interactions. 26 However, the proposal has certain significant problems, such as not accounting for the idiosyncrasy of the patient, not being validated by studies on concrete groups of medicines, and proposing a final ranking based on severity and documentation. Although this is reasonable, the 2 categories require prior definition. All of this makes difficult the task of creating a standardised procedure for establishing clinical significance. 27 The wide range of information sources regarding existing medicinal interactions poses a major problem to professionals when compiling and evaluating information regarding a specific interaction coming from a specific source. We therefore consider that this study provides information which could be of interest for the practice of health professionals. This study provides a basis for a much larger project by the same research team, in which an attempt can be made to evaluate the quality of the content of DIDs, as well as the level of agreement amongst them regarding medications belonging to various therapeutic groups. The important fact is that a database of pharmacological interactions can be very well structured, but the information may be incomplete or not as relevant as it should be. As a result, this primary information, although considered to be of great value, needs to be contrasted with the information analysed regarding the content of each DID, in order to permit a complete and general vision. Polypharmacy and the risk of drug-drug interactions among Danish elderly Estudio APEAS: estudio sobre la seguridad de los pacientes en Atención Primaria de Salud. Madrid: Ministerio de Sanidad y consumo Interacciones de fármacos prescritos en población mayor de 65 años Drug prescribing for the elderly polypharmacy? NPS News n.° 13 Department of Health [accessed Interacciones farmacológicas en tratamientos crónicos: medidas correctoras para su prevención en un área básica de salud rural Aplicación de las nuevas tecnologías a la farmacia hospitalaria en España An assessment of Free, online Drug-Drug Interaction Screening Programs (DSPs) Concordance of severity ratings provided in tour drug interaction compendia Available from: www.portalfarma.com 12. XiZ comunicacion. Medinteract [accessed Available from: www.sefh.es 14. Lexi comp. Lexi-interact [accessed Drug Interaction checker [accessed Drug Interactions Analysis and Management. St. Louis: Facts and Comparisons Drug-reax [accessed Barcelona: Pharma Editores Drug Interaction Facts. San Carlos: Facts and comparisons Evaluation of drug information databases for personal digital assistans Co-prescriptions with itraconazole and fl uconazole as a signal for possible risk of drug-drug interactions: a four-year analysis from Italian general practice Prevalence of drug-drug interactions of antiretroviral agents in the private health care sector in South Africa Extent of potential drug interactions among patients receiving anti-hypertensive medications Evaluación prospectiva de interacciones entre medicamentos en pacientes ingresados mediante una aplicación informática Interacciones medicamentosas: aproximación para establecer y evaluar su relevancia Relevancia clínica de las interacciones medicamentosas American Hospital formulary Service Drug Information [accessed Available from: www.drugdigest.org 31. Drugmaster group. Druint software [accessed Drug Interaction checker [accessed coordinadores de la Agencia española del medicamento y Productos Sanitarios Available from: www.drkoop.com 38. RxList.inc Grupo Ars XXI. Stockley Alertas Adverse drug interaction program [accessed Available from: www.agemed.sante.gouv.fr 42. Drug Bank [accessed DrugBank: a comprehensive resource for in silico drug discovery and exploration Drug Interaction Knowledge-base (DIKB) [accessed Modeling drug mechanism knowledge using evidence and truth maintenance Drug Interaction Ontology [accessed Drug interaction ontology (DIO) for inferences of possible drug-drug interactions Prevalence of drug-drug interactions of antiretroviral agents in the private health care sector in South Africa General Practice Research Database (GPRD) [accessed Onset of acute myocardial infarction after use of non-steroidal antiinflammatory drugs Structure-based discovery of a family of synthetic cyclophilin inhibitors showing a cyclosporin-A phenotype in CaenorhaBDIMtis elegans Quantitative correlations among cYP3A sensitive substrates and inhibitors: literature analysis Anti-hypertensive drug database [accessed Effect of concomitant treatment with a cYP3A4 inhibitor and a calcium channel blocker Side Effects software [accessed Drug interactions software programs STITCH: interaction networks of chemicals and proteins The marine natural products database (MNPD) [accessed SARS-coV protease inhibitors design using virtual screening method from natural products libraries The traditional chinese medicines database (TcMD) [accessed Veteran Health Administration (VHA) clinical database. Department of Veterans Affairs [accessed Available from: www.va.gov Post-marketing surveillance of ischaemic optic neuropathy in male veterans co-prescribed phosphodiesterase-5 inhibitors with organic nitrates or alphablockers Pharmacoepidemiologic study of potential drug interactions in outpatients of a university hospital in Thailand Drug interactions with cholinesterase inhibitors: an analysis of the French pharmacovigilance database and a comparison of two national drug formularies (Vidal, British National Formulary) Heavy consumers of drugs -seen from the viewpoint of the community pharmacist Drug interactions in an intensive care unit Absolute contraindications in relation to potencial drug interactions in outpatient prescriptions: analysis of the first five million prescriptions in 1999 Co-prescriptions with itraconazole and fl uconazole as a signal for possible risk of drug-drug interactions: a four-year analysis from Italian general practice Interaction risk with proton pump inhibitors in general practice: signifi cant disagreement between different drug-related information sources adverse drug-related events, and potential adverse drug interactions in elderly patients presenting to an emergency department Prescribing of sulfasalazine, azathioprine and methotrexate round pregnancy -a descriptive study A2Z Drugs Evaluation of drug information databases for personal digital assistants Drug Interaction [accessed Triple Prescribing Guide [accessed We would like to extend our gratitude to Drs Antonio Romero Tabares of the Andalusian Agency for the Evaluation of Health Technologies (AETSA) and María Victoria Jiménez Espínola of the Andalusian centre for Medicine Information (cADIME) for their excellent guidance and bibliographic searches. This study was developed with the financial help of the Andalusian Autonomous Government Ministry of Health (project 190/2006, group of 2006). References in the first column correspond to the DID location. References in the third column correspond to the article which cites the DID.