A smart hospital-driven approach to precision pharmacovigilance

Adverse drug reactions (ADRs) rank as the fifth most frequent cause of death in developed countries, with the majority of severe ADRs occurring in hospitalized patients. Some groups (notably women and children) seem to be the most affected.Pharmacovigilance is the pharmacological science relating to the collection, detection, assessment, monitoring, and prevention of adverse drug effects. It is currently based on a fragmentary and uncoordinated process of data collection that is unfit to tackle individual decisions.Precision pharmacovigilance is a new concept where pharmacovigilance 'meets' precision medicine to match patients' individual needs.A more precise data collection and effective computational methods can pave the way to precision pharmacovigilance.Smart hospitals can serve as hubs for data collection, analysis, and distribution of clinical decisions tailored for individual patients. Researchers, regulatory agencies, and the pharmaceutical industry are moving towards precision pharmacovigilance as a comprehensive framework for drug safety assessment, at the service of the individual patient, by clustering specific risk groups in different databases. This article explores its implementation by focusing on: (i) designing a new data collection infrastructure, (ii) exploring new computational methods suitable for drug safety data, and (iii) providing a computer-aided framework for distributed clinical decisions with the aim of compiling a personalized information leaflet with specific reference to a drug's risks and adverse drug reactions. These goals can be achieved by using 'smart hospitals' as the principal data sources and by employing methods of precision medicine and medical statistics to supplement current public health decisions. Researchers, regulatory agencies, and the pharmaceutical industry are moving towards precision pharmacovigilance as a comprehensive framework for drug safety assessment, at the service of the individual patient, by clustering specific risk groups in different databases. This article explores its implementation by focusing on: (i) designing a new data collection infrastructure, (ii) exploring new computational methods suitable for drug safety data, and (iii) providing a computer-aided framework for distributed clinical decisions with the aim of compiling a personalized information leaflet with specific reference to a drug's risks and adverse drug reactions. These goals can be achieved by using 'smart hospitals' as the principal data sources and by employing methods of precision medicine and medical statistics to supplement current public health decisions. Some global health issues leading to hospitalization and even death have been, and still are, poorly recognized. For instance, adverse drug reactions (ADRs) (see Glossary) represent a silent but persistent pandemic that is ranked as the fifth most frequent cause of death in developed countries [1.Edwards I.R. An agenda for UK clinical pharmacology: pharmacovigilance.Br. J. Clin. Pharmacol. 2012; 73: 979-982Crossref PubMed Scopus (24) Google Scholar]. More worrisome, according to extensive surveys conducted in the United States, the majority of severe ADRs occur in hospitalized patientsi [2.Shehab N. et al.US Emergency department visits for outpatient adverse drug events, 2013-2014.JAMA. 2016; 316: 2115-2125Crossref PubMed Scopus (376) Google Scholar] and some groups, notably women and children, seem to be the most severely affected [3.Watson S. et al.Reported adverse drug reactions in women and men: aggregated evidence from globally collected individual case reports during half a century.EClinicalMedicine. 2019; 17100188Abstract Full Text Full Text PDF PubMed Scopus (59) Google Scholar]. These groups are often excluded from randomized controlled trials (RCTs) for understandable safety concerns but they do appear to experience ADRs more frequently than would be expected [4.Zucker I. Prendergast B.J. Sex differences in pharmacokinetics predict adverse drug reactions in women.Biol. Sex Differ. 2020; 11: 32Crossref PubMed Scopus (99) Google Scholar]. Pharmacovigilance is a process that aims at managing ADRs and any other drug-related problems by means of drug surveillance and risk prevention. Beyond academic and clinical research, the detection, assessment, understanding, and prevention of ADRs are implemented via a dual surveillance system involving drug regulators and pharmaceutical companies [5.Andrews E.B. Moore N. Mann's Pharmacovigilance. John Wiley & Sons, 2014Crossref Scopus (12) Google Scholar,6.Cobert B. et al.Cobert's Manual of Drug Safety and Pharmacovigilance. World Scientific, 2019Crossref Google Scholar]. Information technologies should play a greater role in achieving the goals of pharmacovigilance by introducing data and knowledge engineering methods for safety signal detection, analysis, and management [7.Hauben M. Zhou X. Quantitative methods in pharmacovigilance.Drug Saf. 2003; 26: 159-186Crossref PubMed Scopus (123) Google Scholar, 8.Hauben M. et al.The role of data mining in pharmacovigilance.Expert Opin. Drug Saf. 2005; 4: 929-948Crossref PubMed Scopus (186) Google Scholar, 9.Harpaz R. et al.Novel data-mining methodologies for adverse drug event discovery and analysis.Clin. Pharmacol. Ther. 2012; 91: 1010-1021Crossref PubMed Scopus (236) Google Scholar, 10.Shang N. et al.Identifying plausible adverse drug reactions using knowledge extracted from the literature.J. Biomed. Inform. 2014; 52: 293-310Crossref PubMed Scopus (45) Google Scholar, 11.Koutkias V.G. Jaulent M.-C. Computational approaches for pharmacovigilance signal detection: toward integrated and semantically-enriched frameworks.Drug Saf. 2015; 38: 219-232Crossref PubMed Scopus (28) Google Scholar, 12.Caster O. et al.vigiRank for statistical signal detection in pharmacovigilance: first results from prospective real-world use.Pharmacoepidemiol. Drug Saf. 2017; 26: 1006-1010Crossref PubMed Scopus (18) Google Scholar, 13.Basile A.O. et al.Artificial intelligence for drug toxicity and safety.Trends Pharmacol. Sci. 2019; 40: 624-635Abstract Full Text Full Text PDF PubMed Scopus (54) Google Scholar, 14.Natsiavas P. et al.Computational advances in drug safety: systematic and mapping review of knowledge engineering based approaches.Front. Pharmacol. 2019; 10: 415Crossref PubMed Scopus (7) Google Scholar, 15.Ding Y. et al.An evaluation of statistical approaches to postmarketing surveillance.Stat. Med. 2020; 39: 845-874Crossref PubMed Scopus (9) Google Scholar, 16.Phillips R. et al.Statistical methods for the analysis of adverse event data in randomised controlled trials: a scoping review and taxonomy.BMC Med. Res. Methodol. 2020; 20: 288Crossref PubMed Scopus (8) Google Scholar]. To this end, various data sources of potential signals are being explored both in the literature and at the operational ground level. These new approaches include: (i) voluntary reporting systems, through which observed cases, known as individual case safety reports (ICSRs), are reported by healthcare professionals to the regulatory authorities and other bodies; (ii) observational databases [e.g., electronic health records (EHRs)], which are useful in identifying causal relationships between groups of problematic clinical conditions and suspected drugs; (iii) free-text resources, for example, the scientific literature and patient self-reports, next to the growing role played by social media data [17.Li Y. et al.A method to combine signals from spontaneous reporting systems and observational healthcare data to detect adverse drug reactions.Drug Saf. 2015; 38: 895-908Crossref PubMed Scopus (37) Google Scholar,18.Audeh B. et al.Pharmacology and social media: potentials and biases of web forums for drug mention analysis–case study of France.Health Inform. J. 2019; 26: 1253-1272Crossref PubMed Scopus (5) Google Scholar]. Due to the importance of drug safety and the limitations of each signal source (e.g., data sparsity, small samples sizes, short time horizon limiting the detection of long term ADRs), it has been argued by international authorities that there is a need for a more comprehensive adverse drug event (ADE) surveillance system that would be capable of handling all of these possible information sources [19.De Pretis F. et al.Artificial intelligence methods for a Bayesian epistemology-powered evidence evaluation.J. Eval. Clin. Pract. 2021; 27: 504-512Crossref PubMed Scopus (6) Google Scholar]. However, the heterogeneity, fragmentation, and lack of standardized/well-defined interfaces that could characterize the available data sources and the signal detection methods complicate the implementation of this kind of synthesis. At present, there is an uncoordinated process of data collection that is unable to integrate ab initio a minimum package of information related to signals (especially those stemming from ICSRs) from one side with drug consumption data from the other. This appears to be the main culprit, explaining the low reliability in estimating the true incidence not only of ADRs but also of other pharmacoepidemiologic measures in patient populations. This failure to deliver affordable statistical measures has been summarized by Edwards [1.Edwards I.R. An agenda for UK clinical pharmacology: pharmacovigilance.Br. J. Clin. Pharmacol. 2012; 73: 979-982Crossref PubMed Scopus (24) Google Scholar], 'pharmacovigilance operates without clear objectives in relation to individual decisions, [...] with obscure materials and methods used for making decisions, with very limited reasoning and discussion, and little or no follow up and audit of the results'. Despite these criticisms, the problem has been approached by: (i) analyzing one data source at a time, mainly resorting to nonparametric statistical methods [20.Candore G. et al.Comparison of statistical signal detection methods within and across spontaneous reporting databases.Drug Saf. 2015; 38: 577-587Crossref PubMed Scopus (78) Google Scholar] (for instance, disproportionality analysis based on contingency tables built on data); (ii) linking ex post two or more data sources of potential signals to reimbursed prescriptions databases or other proxies of drug consumption [9.Harpaz R. et al.Novel data-mining methodologies for adverse drug event discovery and analysis.Clin. Pharmacol. Ther. 2012; 91: 1010-1021Crossref PubMed Scopus (236) Google Scholar,17.Li Y. et al.A method to combine signals from spontaneous reporting systems and observational healthcare data to detect adverse drug reactions.Drug Saf. 2015; 38: 895-908Crossref PubMed Scopus (37) Google Scholar,21.Christiaans-Dingelhoff I. et al.To what extent are adverse events found in patient records reported by patients and healthcare professionals via complaints, claims and incident reports?.BMC Health Serv. Res. 2011; 11: 49Crossref PubMed Scopus (70) Google Scholar]. An ambitious example of the latter approach is the ORDEI drug safety projectii, launched by the French National Agency for Medicines and Health Products Safety, ANSM, which is intended to merge information coming from at least three national databases: one concerning ADEs extracted from case reports, one collecting information on reimbursed prescriptions, and one including drug safety information for a large set of authorized medicines. However, there is a major concern emerging from this kind of approach regarding the concept of therapeutic adherence: the data of prescription purchases does not guarantee per se that the patient has consumed the prescribed drug at all or at the correct dosage or time. Sometimes, this can be particularly problematic and have disastrous consequences, not only for his/her health but also on the soundness of database linking and on the robustness of information extraction. As an alternative to this kind of approach, another path has been advocated by academics [22.Donzanti B.A. Pharmacovigilance is everyone's concern: let's work it out together.Clin. Ther. 2018; 40: 1967-1972Abstract Full Text Full Text PDF PubMed Scopus (7) Google Scholar,23.Lavertu A. et al.A new era in pharmacovigilance: toward real-world data and digital monitoring.Clin. Pharmacol. Ther. 2021; 109: 1197-1202Crossref PubMed Scopus (7) Google Scholar], drug regulators [24.Arlett P. et al.Pharmacovigilance 2030.Clin. Pharmacol. Ther. 2019; 107: 89-91Crossref PubMed Scopus (13) Google Scholar], and the pharmaceutical industryiii: precision pharmacovigilance. In this case, the goal of precision pharmacosurveillance is similar to the changes occurring in medical treatment (i.e., the development of precision medicine). This strives to achieve a customization of healthcare, with medical decisions, treatments, practices, or products being tailored to each specific patient, instead of a one-drug-fits-all paradigm [25.König I.R. et al.What is precision medicine?.Eur. Respir. J. 2017; 501700391Crossref PubMed Scopus (141) Google Scholar]. Whereas precision medicine is clearly on the horizon in even routine clinical practice in many diseases [26.Christiani D. Ahasic A. Personalized critical care medicine: how far away are we?.Semin. Respir. Crit. Care Med. 2015; 36: 809-822Crossref PubMed Scopus (9) Google Scholar], pharmacovigilance is lacking such advances and the push towards a precision framework for this field is still at a very initial stage [27.Masimirembwa C. et al.Rolling out efavirenz for HIV precision medicine in Africa: are we ready for pharmacovigilance and tackling neuropsychiatric adverse effects?.OMICS. 2016; 20: 575-580Crossref PubMed Scopus (19) Google Scholar]. Nevertheless, some proposals for implementing precision pharmacovigilance have been outlined [28.Wong M.U. et al.Towards precision informatics of pharmacovigilance: OAE-CTCAE mapping and OAE-based representation and analysis of adverse events in patients treated with cancer drugs.in: AMIA Annual Symposium proceedings, November 4-8 2017. Precision Informatics for Health: The Right Informatics for the Right Person at the Right Time. American Medical Informatics Association, 2018: 1793-1801Google Scholar,29.Bate A. Stegmann J.-U. Safety of medicines and vaccines – building next generation capability.Trends Pharmacol. Sci. 2021; 42: 1051-1063Abstract Full Text Full Text PDF PubMed Scopus (5) Google Scholar], for example, attempting to cluster specific risk groups in spontaneously reported ADE databasesiv, with the goal of mitigating children's risks in paediatric pharmacosurveillance [30.Giangreco N. Tatonetti N. Using precision pharmacovigilance to detect and evaluate antiepileptic drug associated adverse reactions in pediatric patients. Poster.in: 26th International Conference on Intelligent Systems for Molecular Biology (ISMB) 2018, July 6-10 2018. F1000Research, 2018Google Scholar,31.Giangreco N.P. et al.No population left behind: improving paediatric drug safety using informatics and systems biology.Br. J. Clin. Pharmacol. 2021; 88: 1464-1470Crossref PubMed Scopus (4) Google Scholar] and with a particular focus on hospitalized patients in emergency wards [32.La Russa R. et al.Personalized medicine andadverse drug reactions: the experience of an Italian teaching hospital.Curr. Pharm. Biotechnol. 2017; 18: 274-281Crossref PubMed Google Scholar, 33.Just K.S. et al.Personalising drug safety—results from the multi-centre prospective observational study on adverse drug reactions in emergency departments (ADRED).Eur. J. Clin. Pharmacol. 2019; 76: 439-448Crossref PubMed Scopus (8) Google Scholar, 34.Just K.S. et al.Adverse drug reactions in the emergency department: is there a role for pharmacogenomic profiles at risk?—Results from the ADRED study.J. Clin. Med. 2020; 9: 1801Crossref Scopus (5) Google Scholar]. These publications highlight our viewpoint that the hospital can be considered as a privileged observatory for precision pharmacovigilance. In this respect, the data should not be restricted to registering the medical events in emergency departments. We propose that a smart hospital could resemble a good laboratory where researchers could examine, at the same time, samples of patients taking drugs and subsets of those experiencing possible ADRs. This constitutes the kernel of a smart hospital-driven approach to precision pharmacovigilance: we will outline this new concept in the next section. Afterwards, we will investigate methods that could be worthwhile implementing in the section 'A research agenda for developing precision pharmacovigilance' and end by discussing some limitations and possible extensions of this approach in the section 'Concluding remarks and future perspectives'. In this article, by the term precision pharmacovigilance, we mean providing a more comprehensive and interactive framework than that currently delivered by standard pharmacovigilance for drug safety assessment. Our approach stresses that pharmacovigilance should be at the service of the individual patient. This is an ambitious goal as it challenges the current viewpoint about pharmacovigilance. Nonetheless, before it can be achieved, several subtasks will have to be undertaken, the main ones being: (i) designing a new data collection infrastructure for precision pharmacovigilance; (ii) exploring new computational methods capable of analyzing and assessing data regarding drug safety; (iii) providing a computer-aided framework for distributed clinical decisions with the aim of compiling a personalized information leaflet (also known as a personalized package insert) with specific reference to a drug's contraindications, warnings, precautions, and ADRs; and (iv) integrating this framework into clinical practice. In addition to these four points, we believe that the feasibility of precision pharmacovigilance demands that two additional elements need to be considered. First, precision pharmacovigilance can be best achieved by positioning the hospital as the main center of this kind of research work, and second, by exploiting the growing role of the secondary use of healthcare data laws that are now in force in several nations and are expected to be more widely implemented in the future. Some recent reports have highlighted how hospitals can be considered as privileged local observatories for understanding the temporal features and causal links behind the onset of ADEs [32.La Russa R. et al.Personalized medicine andadverse drug reactions: the experience of an Italian teaching hospital.Curr. Pharm. Biotechnol. 2017; 18: 274-281Crossref PubMed Google Scholar,35.Amalberti R. et al.Adverse events in medicine: easy to count, complicated to understand, and complex to prevent.J. Biomed. Inform. 2011; 44: 390-394Crossref PubMed Scopus (41) Google Scholar]. While hospital ADEs clearly account for only a fraction of all possible events, it is evident that the most severe ADEs occur in hospitals (see earlier). Furthermore, signals stemming from hospital medications are not the only ones to be tracked in a clinic, since even over-the-counter drugs may cause ADEs and lead to hospitalization. Therefore, an assessment of ADEs as a result of hospital prescriptions and prescriptions given before the start of hospital care has to be considered. However, even with these caveats, the data quality of hospital ADEs may be viewed as superior to the average pharmacovigilance signals originating from other sources. In fact in a hospital setting, therapeutic adherence problems generally affecting patients and their own data should be minimized by properly recording drug administration/dosing, a task generally fulfilled by physicians and nurses with the assistance of the available technology. The idea of using hospitals as the main data sources of signals for computer-aided pharmacovigilance purposes has its roots in the late 1980s, when the first proposals were outlined about the concept of hospital pharmacoepidemiology and how this could be underpinned by the automated data management systems then becoming widely available [36.Platt R. et al.Pharmacoepidemiology in hospitals using automated data systems.Am. J. Prev. Med. 1988; 4: 39-47PubMed Google Scholar,37.Burke J.P. et al.Expanding roles of hospital epidemiology: pharmacoepidemiology.Infect. Control Hosp. Epidemiol. 1989; 10: 253-254Crossref PubMed Scopus (7) Google Scholar]. Nowadays, next to the already up and running computerized systems (e.g., claim databases, insurance databases, e-prescription systems, EHRs, computerized physician order entry systems, laboratory information systems, to name a few), the growing implementation of sensors [38.Triantafyllidis A.K. et al.Framework of sensor-based monitoring for pervasive patient care.Healthc. Technol. Lett. 2016; 3: 153-158Crossref PubMed Scopus (11) Google Scholar, 39.Dorj U.-O. et al.The intelligent healthcare data management system using nanosensors.J. Sens. 2017; 20177483075Crossref Scopus (3) Google Scholar, 40.Beninger P. Ibara M.A. Pharmacovigilance and biomedical informatics: a model for future development.Clin. Ther. 2016; 38: 2514-2525Abstract Full Text Full Text PDF PubMed Scopus (33) Google Scholar] and the Internet of Things (IoT) [41.Yu L. et al.Smart hospital based on Internet of Things.J. Netw. 2012; 7: 1654-1661Google Scholar,42.Zhang H. et al.Connecting intelligent things in smart hospitals using NB-IoT.IEEE Internet Things J. 2018; 5: 1550-1560Crossref Scopus (137) Google Scholar] in smart hospitals could pave the way to an even more robust and widespread data collection, leading to a much more fine-tuned representation of in-patient and out-patient states. In this scenario, a much richer picture would be acquired for each individual patient depicting the profiles of the disease path and the effects of interventions (i.e., with a higher time resolution and longer duration) allowing better opportunities to provide personalized therapies, even in real time. Furthermore, since data collected in a smart hospital would be accurate, this would help to identify the subset of patients experiencing ADEs out of the total number receiving some particular drug. To supplement this minimal setting, several covariates like the genomic background of patients/other omics data, drug dosage, polypharmacy, comorbidities, and real world evidence could also represent a more in-depth level of patient stratification (the division of one patient group into subgroups, each one representing a particular subsection of the potential patient population). This would involve the application of a wide spectrum of data analysis, based on standard and new computational methods specially devised for pharmacovigilance. In particular, Bayesian inference methods [19.De Pretis F. et al.Artificial intelligence methods for a Bayesian epistemology-powered evidence evaluation.J. Eval. Clin. Pract. 2021; 27: 504-512Crossref PubMed Scopus (6) Google Scholar,43.De Pretis F. Landes J. EA3: a softmax algorithm for evidence appraisal aggregation.PLoS One. 2021; 16e0253057Crossref PubMed Scopus (3) Google Scholar] could be employed, since they represent a natural framework for aggregating diverse types of evidence and, importantly, for updating the reliability of a working hypothesis as more evidence or information becomes available. Ultimately, the previous data infrastructure and data analysis could be beneficial in a retrospective manner, aiding in clinical decision-making, by helping to generate a personalized information leaflet with specific reference to a drug's contraindications, warnings, precautions, and ADRs for each individual patient. According to some of the recent literature in the field [44.Darabi S. et al.The feasibility and potential role of pharmacogenetics to improve drug safety in patients with advanced cancers.J. Clin. Oncol. 2021; 39e22522Crossref Google Scholar, 45.García-González X. Salvador-Martíın S. Pharmacogenetics to avoid adverse reactions in cardiology: ready for implementation?.J. Pers. Med. 2021; 11: 1180Crossref PubMed Scopus (2) Google Scholar, 46.Just K.S. et al.CYP2D6 in the brain: potential impact on adverse drug reactions in the central nervous system—results From the ADRED study.Front. Pharmacol. 2021; 12624104Crossref PubMed Scopus (3) Google Scholar, 47.Shnayder N.A. et al.Using pharmacogenetics of direct oral anticoagulants to predict changes in their pharmacokinetics and the risk of adverse drug reactions.Biomedicines. 2021; 9: 451Crossref PubMed Scopus (9) Google Scholar], this would be moving in the same direction as many pharmacogenetic approaches and not focusing solely on selecting the dose with optimal efficacy. Instead it should be appreciated that this is too simplistic; optimal dose should also consider the patient's risk of avoiding drug–drug interactions and/or ADRs. Figure 1presents what this personalized information leaflet may look like and how it differs from the current leaflets used in clinical practice. In addition to smart hospitals, the secondary use of healthcare data laws represents the foundation for the concept of precision pharmacovigilance. For more than a decade, the shift to the collection of data for secondary use has been forecast and encouraged [48.Sandhu E. et al.Secondary uses of electronic health record data: benefits and barriers.Jt. Comm. J. Qual. Patient Saf. 2012; 38: 34-40Abstract Full Text Full Text PDF PubMed Scopus (12) Google Scholar] and several international organizations and national states have now launched programs and projects to meet such a challenge (see, e.g., the directions provided by the Organisation for Economic Co-operation and Development [49.Oderkirk J. et al.International comparisons of health system performance among OECD countries: opportunities and data privacy protection challenges.Health Policy. 2013; 112: 9-18Crossref PubMed Scopus (31) Google Scholar]). In this respect, Finland represents one of the forerunners, with a specific national legislation passed by its parliament in 2019v. This Finnish act mainly aims at facilitating the effective and safe processing and access to the personal social and health data for steering, supervision, research, statistical purposes, and development in the health and social sector. A second objective is to guarantee an individual's legitimate expectations as well as their rights when processing personal data. This legislation was the result of a long reform process that has shaped the new 'secondary use'-friendly environment [50.Aula V. Institutions, infrastructures, and data friction –reforming secondary use of health data in Finland.Big Data Soc. 2019; 6205395171987598Crossref Scopus (8) Google Scholar]. Alongside this process, some preparatory work, both in terms of sketching and prototyping how data should be gathered and exploited within this kind of system, has been carried out [51.Lähteenmäki J. et al.Integrating data from multiple Finnish biobanks and national health-care registers for retrospective studies: practical experiences.Scand. J. Public Health. 2021; (Published online April 12, 2021)https://doi.org/10.1177/14034948211004421Crossref PubMed Scopus (2) Google Scholar] with the clear intention of taking advantage of the opportunities evident in this new framework. In the next section, we will provide more details regarding the methods to be employed within a secondary use healthcare data environment. The concept of precision pharmacovigilance will have to be based on three strands of research: data collection, data analysis, and data exploitation. Both theoretical and applied research will be needed to match abstract modeling to a boots-on-the-ground approach, where statistical-based models will have to handle real world evidence and cope with the clinical constraints present in a hospital setting, as detailed later. Figure 2 shows a graphical representation of this entire process taking place in a hospital. The methods concerning data collection will be mainly borrowed from health informatics [52.Maglaveras N. et al.Integrated care and connected health approaches leveraging personalised health through big data analytics.in: Maglaveras N. Gizeli E. Studies in Health Technology and Informatics. Vol. 224. IOS Press, 2016: 117-122Google Scholar] and information engineering [53.Natsiavas P. et al.A knowledge-based platform for assessing potential adverse drug reactions at the point of care: user requirements and design.in: Ohno-Machado L. Séroussi B. Studies in Health Technology and Informatics. Vol. 264. IOS Press, 2019: 1007-1011Google Scholar] on one side and hospital pharmacoepidemiology [54.Strom B.L. Schinnar R. Hospital pharmacoepidemiology.in: Strom B.L. Pharmacoepidemiology. John Wiley & Sons, 2006: 539-553Crossref Scopus (2) Google Scholar] on the other. The task of tracking one patient (administered a drug treatment) from the very first moment he/she is admitted to a hospital, till the moment he/she is discharged, will require the design and realization of a data management system based on the best scientific data [39.Dorj U.-O. et al.The intelligent healthcare data management system using nanosensors.J. Sens. 2017; 20177483075Crossref Scopus (3) Google Scholar,55.No J. et al.High-performance scientific data management system.J. Parallel Distrib. Comput. 2003; 63: 434-447Crossref Scopus (12) Google Scholar]. This system will have to be integrated with the current data management systems and practices used by the hospital (especially those for data sharing with the local regulatory agencies) and be matchable in real time and in a scalable way with the drug consumption data stored in the hospital pharmacy, the information coming from the patient's EHRs, and the reports and codifications of possible ADRs entered by physicians, drawing on current standard classifications for drug safety [56.Fang H. et al.FDA drug labeling: rich resources to facilitate precision medicine, drug safety, and regulatory science.Dr