| کلیدواژههای انگلیسی مقاله |
Drug interactions, Electronic prescribing, Prevalence, Cross-sectional studies, Iran, What&,rsquo s Known Drug-drug interactions can cause serious adverse effects and are a common prescribing error. Studies in various countries, including Iran, have documented the prevalence of potential drug-drug interactions, but recent, large-scale data across multiple medical specialties in Iran is lacking. What&,rsquo s New This study of over one million prescriptions reveals that 38.77 percent contain a potential drug-drug interaction, with 12.98 percent of all prescriptions involving high-risk combinations. This research identified dexamethasone, ketorolac, and quetiapine as the drugs most frequently involved in potential interactions and presented the most common interactions, distinguished by specialty. IntroductionThe treatment of complex diseases often requires the simultaneous use of multiple drugs, but this method may cause interactions between the drugs that can lead to side effects and even failure of the treatment. Potential drug-drug interactions (DDIs) are one of the common errors in prescribing medications, and studies show that a significant number of these DDIs is severe or moderate in therapeutic terms. 1, DDI is a distinct form of adverse drug event that arises when one medication alters the action of another, potentially leading to heightened toxicity or diminished therapeutic benefit. DDIs are especially prevalent in hospital environments, where patients frequently receive multiple medications simultaneously. 2, Potential DDIs are a common cause of adverse drug events, significantly contributing to patient morbidity and rising healthcare costs. They can reduce medication effectiveness, increase toxicity, and undermine patient adherence to prescribed treatment plans. 3, , 4, It is estimated that this results in increased hospital stays, which incurs an approximate annual cost of US$ 1 billion to the healthcare system. 5, Potential DDIs are avoidable, 4, and tools such as Lexicomp&,reg , Medscape&,reg , and Drugs.com exist for checking DDIs. In this regard, studies indicate that Lexicomp&,reg , which is a subsidiary of UpToDate&,reg , has higher accuracy. 6, , 7, Extensive research on DDIs has been conducted both internationally and domestically. International studies&,mdash from countries such as Greece, Denmark, Sweden, Finland, Nepal, and others&,mdash have highlighted the common occurrence of DDIs in clinical settings. 8, - 15, Similarly, domestic investigations in Shiraz have demonstrated the clinical significance of DDIs among diverse patient populations. 5, , 16, , 17, However, a gap remains in the literature, only one study, published in 2011, has explored the frequency of DDIs across various medical specialties and among general practitioners in Iran, with limited studies worldwide focusing on medical specialties and the most frequent DDIs in each specialty. 15, , 18, This gap underscores the need for further research to fully understand the distribution and impact of DDIs in different clinical contexts, particularly in Iran. To the best of our knowledge, no recent studies have examined the frequency of potential pharmacological DDIs across various medical specialties, including general practitioners in Iran, since the 2011 study. This study aims to address this gap by identifying the most prevalent DDIs overall, the top contributing drug specialties for each DDI, the specialties with the highest rates of potential DDIs, and the most prevalent DDI within each specialty, using a dataset of 1 million electronic prescriptions (e-prescriptions).Materials and MethodsA total of 1,049,769 anonymous e-prescription records were collected from treatment centers in Shiraz, comprising 479,770 from Salamat Insurance and 569,999 from Tamin Insurance, over the period from November 2021 to February 2024. All e-prescriptions issued during this timeframe were included in the study no random sampling was performed. The data were provided by the Information Technology Center at Shiraz University of Medical Sciences in a fully pre-anonymized format. In other words, all direct personal identifiers (such as the names of patients and prescribing doctors) and indirect identifiers (such as date of birth, residential area, and unique prescription codes) had been removed from the dataset before our access. Consequently, the research team never had access to identifiable patient information. Due to the retrospective design and the use of pre-anonymized data, obtaining individual consent from each patient was not feasible. The study protocol was reviewed and approved by the Research Ethics Committee of the School of Medicine, Shiraz University of Medical Sciences, which confirmed that the project was conducted in accordance with ethical principles and the national norms and standards for medical research in Iran (reference number, IR.SUMS.MED.REC.1403.642). The prescriptions were analyzed using Lexicomp&,reg DDI checker software (version 2023 Wolters Kluwer, the Netherlands) for potential DDI assessment. This tool was selected because previous studies have demonstrated that it provides a higher level of accuracy than other similar tools. 6, , 7, Since levels A and B DDIs are not clinically significant, they are not included in this report. According to guideline recommendations, these categories represent interactions that either lack meaningful clinical effects or have such a low likelihood of causing harm that no intervention is necessary. 3, , 19, , 20, The classification of DDIs (A, B, C, D, and X) and their definitions are presented in table 1,. 3, ClassificationInteractionDefinitionANo intervention specifiedEvidence shows no interaction affecting pharmacodynamics or pharmacokinetics.BNo action neededEvidence shows drug interactions occurring simultaneously without clinical concern.CMonitor therapyEvidence shows that drug interactions can result in clinical symptoms, but the benefits of using these drugs together outweigh the potential risks. Close monitoring is required to identify any adverse effects, and dosage modifications for one or both drugs may be necessary.DConsider therapy modificationEvidence shows potential clinical interactions. Each patient should be assessed individually to see if the benefits outweigh the risks. Steps may be needed to reduce toxicity, including intensive monitoring, dosage changes, or alternative treatments.XPreventing the interactionEvidence shows interactions with clinical side effects. The risks generally outweigh the benefits.Table 1.Classification of DDIs and their definitionsAll data processing and descriptive statistical analysis were performed using Python (version 3.10.9, Python Software Foundation, United States), Pandas library (version 1.5.3, NumFOCUS/PyData, United States), 21, and NumPy (version 1.23.5, NumPy Developers/NumFOCUS, USA). 22, Our Python script uses Iran&,rsquo s generic drug codes and then searches for them in the Lexicomp&,reg software. Due to the complexity of mapping local Iranian drug codes to an international database and the large volume of data, a custom computational pipeline was developed. The key stages of this pipeline are detailed below. Initial Data Cleaning and Standardization Raw prescription data, containing drug details and physician specialties, were loaded into Pandas DataFrames. This involved, a) Creating a unique list of Iranian generic drug codes from the entire dataset to serve as a master reference. b) Standardizing drug names by systematically replacing local variations and abbreviations with consistent terminology. This process was guided by a manually curated replacement list (e.g., standardizing &,ldquo valproate Sodium&,rdquo to &,ldquo valproic Acid&,rdquo ). This list was manually created to handle exceptions and drug names that failed to be identified by the primary search algorithm described below. This pre-processing step was crucial for improving the accuracy of subsequent database lookups. Drug Name Parsing and Ingredient Extraction A significant challenge was that Iranian drug names often contain multiple pieces of information (e.g., active ingredient, salt form, dosage) within a single string. A rule-based parsing code was built to deconstruct these names, a) Regular expressions were used to identify and separate the core ingredient from information in parentheses (e.g., extracting &,ldquo as hydrochloride&,rdquo from &,ldquo dapoxetine [as hydrochloride]&,rdquo ). b) The parenthetical text was further classified into categories such as salt forms (e.g., &,ldquo as metformin hydrochloride&,rdquo ), formulation numbers (e.g., &,ldquo cold adult 4-2&,rdquo ), or protein sources (e.g., &,ldquo recombinant&,rdquo ). c) For combination drugs, identified by a &,ldquo /&,rdquo separator, the script splits the string into a list of individual active ingredients (e.g., parsing &,ldquo losartan potassium/hydrochlorothiazide&,rdquo into &,ldquo losartan potassium&,rdquo and &,ldquo hydrochlorothiazide&,rdquo ). d) The route of administration (e.g., topical, systemic, ophthalmic) was also programmatically extracted from the full drug name string to aid in disambiguation during the mapping stage. Mapping Iranian Drugs to the Lexicomp&,reg Database The core of our analysis involved mapping the extracted Iranian drug ingredients to their corresponding generic drug entries in the Lexicomp&,reg database. This was a multi-step, hierarchical search process, a) Direct search, The script first attempted to find an exact, case-insensitive match for the extracted ingredient in the Lexicomp&,reg generic table. b) Wildcard search, If no exact match was found, a LIKE structured query language (SQL) query was used to find ingredients starting with the same name. c) Brand name search, If the generic search failed, the script then searched the Lexicomp&,reg brand table, as some local generic names correspond to international brand names. d) Salt-stripping search, For drugs with identified salt forms, if the full name (e.g., &,ldquo metformin hydrochloride&,rdquo ) failed to yield a match, the script would search again using only the core ingredient (&,ldquo metformin&,rdquo ). e) Route disambiguation, In cases where a search returned multiple Lexicomp&,reg entries for the same drug (e.g., systemic vs. topical formulations), the previously extracted route of administration was used to select the correct database entry. For each successfully mapped Iranian drug, its corresponding Lexicomp&,reg category_id(s) were retrieved. These IDs are essential for querying the interactions table. DDI Analysis For each prescription containing two or more successfully mapped drugs, the following analysis was performed, a) The script compiled a list of all Lexicomp&,reg category_ids for all drugs in the prescription. b) Using Python&,rsquo s combinations function, it generated every possible two-drug pair within the prescription. c) For each pair, a query was executed against the Lexicomp&,reg monograph table to find any documented interactions, checking for the drugs in both the object_id and precipitant_id columns. d) It is possible for a single drug pair to have multiple documented interactions with varying risk levels in the Lexicomp&,reg database. When multiple interactions were found for a single pair, the one with the highest risk level (where X=5, D=4, C=3, B=2, A=1) was selected as the primary interaction. e) The results, including the drug pair, risk level (C, D, or X), summary, and management advice, were appended to a final results DataFrame. As mentioned, DDI levels A and B were excluded from the final analysis. Descriptive Report Using the obtained DDI results, we employed the Pandas library to compute the descriptive statistics for this study. The data were grouped by medical specialty and by specific DDI pairs to determine overall frequencies, prevalence rates, and the most common DDIs (along with the specialties that most frequently prescribed them). The top DDIs within each specialty were identified and presented in a separate table. Data visualization was conducted using the Seaborn (version 0.13.2, Michael Waskom/PyData, United States) and Matplotlib (version 3.9.1, Matplotlib Development Team/NumFOCUS, United States) Python packages. 23, , 24, ResultsThe details of how many prescriptions each specialty has, the number of prescriptions with four or more drugs (considered polypharmacy), the portion of injectable drugs, and the most frequently prescribed drugs in each specialty are shown in table 2, (total count and top most frequently prescribed drugs across different physician specialties). The results showed that 136,242 prescriptions (12.98%) included at least one D or X DDI, and 407,031 prescriptions (38.77%) had at least one C, D, or X DDI from all 1,049,769 prescription records. Figure 1, (counts and percentages of C, D, and X DDIs across all prescriptions) presents a bar chart showing the counts of A, B, C, D, and X level DDIs across all prescriptions, as well as the percentage of each relative to the total number of DDIs. Figure 2, (top 20 drugs most frequently involved in DDIs) illustrates the most frequently occurring drugs in all DDIs. While figure 3, (average D or X DDIs per prescription by specialty) presents the D or X DDI count per prescription by specialty. The most common potential DDIs, along with the specialties of the five types of physicians most frequently responsible for these errors, are listed in table 3, (the top 50 most common X or D interactions along with the top 5 physician specialties). These DDIs are sorted by their frequency of occurrence in prescriptions, and due to the large number of DDIs, only the top 50 most frequent are included. Additionally, the table lists the top 5 most frequent D or X DDIs within each specialty. Among the identified DDIs, certain drugs and drug pairs appeared frequently across prescriptions, revealing notable patterns in prescribing practices. The most common DDIs involved dexamethasone, ketorolac, quetiapine, aspirin, and valproic acid. In addition, the most common DDIs were between aprepitant and dexamethasone, ketorolac and naproxen, aprepitant and doxorubicin, prednisolone and tacrolimus, and diclofenac sodium and ketorolac. The majority (33 out of 50, or 66%) were classified as having moderate severity, while 17 interactions (34%) were categorized as major severity according to Lexicomp&,reg . Regarding the reliability of the evidence supporting these interactions, 36 (72%) were rated as fair, 9 (18%) as good, 4 (8%) as excellent, and 1 (2%) as poor. In table 4, (each specialty&,rsquo s total prescriptions, D or X interaction count, and the top 5 most frequent D or X interactions for each specialty), each specialty of doctor expertise is displayed along with the total number of prescriptions and the count of X or D DDIs. The highest incidence of D- or X-level DDIs per prescription was observed in rheumatology, endocrinology, orthopedics, oncology, internal medicine, emergency medicine, and psychiatry. The average number of D- or X-level DDIs per prescription in these specialties was 0.53, 0.41, 0.40, 0.40, 0.26, 0.24, and 0.23, respectively. The complete version and a detailed list of all specialties presented in Supplementary Tables 1-3,. SpecialtyPrescription countPrescriptions with polypharmacyDrugs countInjectable drugTop most frequent drugsTop most frequent injectable drugsAll physicians1049769 (100.0%)431604 (41.1%)3895071 (100.0%)131100 (3.4%)1) Sodium chloride (Parenteral),114745 (2.9%)2) Pantoprazole (As sodium sesquihydrate) (Oral),85329 (2.2%)3) Dexamethasone (as disodium phosphate) (Parenteral),77978 (2.0%)4) Acetaminophen (Oral),65292 (1.7%)5) Famotidine (Oral),60583 (1.6%)1) Acetaminophen (Intravenous),50524 (38.5%)2) Granisetron (Intravenous),31600 (24.1%)3) Dextrose / Sodium chloride (Intravenous),18087 (13.8%)4) Carboplatin (Intravenous),6057 (4.6%)5) Docetaxel (Intravenous),5637 (4.3%)2) Pantoprazole (As sodium sesquihydrate) (Oral),85329 (2.2%)2) Granisetron (Intravenous),31600 (24.1%)3) Dexamethasone (as disodium phosphate) (Parenteral),77978 (2.0%)3) Dextrose / Sodium chloride (Intravenous),18087 (13.8%)4) Acetaminophen (Oral),65292 (1.7%)4) Carboplatin (Intravenous),6057 (4.6%)5) Famotidine (Oral),60583 (1.6%)5) Docetaxel (Intravenous),5637 (4.3%)General practitioner176746 (16.8%)113410 (64.2%)827631 (21.2%)50914 (6.2%)1) Sodium chloride (Parenteral),76767 (9.3%)1) Acetaminophen (Intravenous),45038 (88.5%)2) Ketorolac trometamol (Parenteral),46029 (5.6%)2) Dextrose / Sodium Chloride (Intravenous),4570 (9.0%)3) Acetaminophen (Intravenous),45038 (5.4%)3) Dextrose (Intravenous),571 (1.1%)4) Azithromycin (Oral),40099 (4.8%)4) Trifluoperazine (Intramuscular),249 (0.5%)5) Ondansetron (Parenteral),33488 (4.0%)5) Furosemide (Intravenous),179 (0.4%)Oncology109940 (10.5%)53499 (48.7%)498664 (12.8%)58761 (11.8%)1) Dexamethasone (as disodium phosphate) (Parenteral),34819 (7.0%)1) Granisetron (Intravenous),27601 (47.0%)2) Granisetron (Intravenous),27601 (5.5%)2) Dextrose / Sodium Chloride (Intravenous),10002 (17.0%)3) Sodium chloride (Parenteral),24211 (4.9%)3) Carboplatin (Intravenous),4917 (8.4%)4) Aprepitant (Oral),19453 (3.9%)4) Docetaxel (Intravenous),4874 (8.3%)5) Filgrastim (Parenteral),18095 (3.6%)5) Dextrose (Intravenous),3666 (6.2%)Neurologists104968 (10.0%)37997 (36.2%)361923 (9.3%)1674 (0.5%)1) Gabapentin (Oral),21498 (5.9%)1) Ibuprofen (Intravenous),1099 (65.7%)2) Meloxicam (Oral),11876 (3.3%)2) Immune Globulin (Intravenous),342 (20.4%)3) Famotidine (Oral),11716 (3.2%)3) Acetaminophen (Intravenous),140 (8.4%)4) Propranolol hydrochloride (Oral),11432 (3.2%)4) Dextrose / Sodium Chloride (Intravenous),50 (3.0%)5) Pantoprazole (as sodium sesquihydrate) (Oral),10285 (2.8%)5) Dextrose (Intravenous),9 (0.5%)(, sign separates interacting drugs, and / sign is used for compounded drugs). *The summaries were gathered from Lexicomp&,reg . If the specialty was unknown, it was categorized as &,lsquo Others.&,rsquo Prescriptions with four or more, which four drugs are considered as polypharmacy. |