Intrabases divergences in the mHealth era: a drug interaction investigation in an infectious-diseases hospital setting

Introduction: Information on potential drug interactions (PDI) are obtained from databases available on the web or through mobile healthcare applications (mHealth), and can prevent unfavorable clinical outcomes for patients. This study compared PDI information available in Micromedex® drug interaction checker, its web version and its mHealth app. Method: A cross-sectional study realized based on a retrospective review of drug prescriptions in a reference hospital in infectology in the Midwest Region of Brazil, 2018. We selected all prescriptions containing two or more drugs. Drugs were classified according to the first level of the Anatomical Therapeutic Chemical (ATC) classification, according to the route of administration and the number of drugs prescribed. PDIs were classified according to the severity system and four-level evidence classification system. Results: This study selected 72 patients, predominantly male, median age of 38 years, average length of stay of 15.8 days, and most diagnosed with HIV/AIDS. The most frequently prescribed anatomical groups according to ATC were digestive system and metabolism (22.1%) and general anti-infectives for systemic use (21.6%). The average number of drugs per prescription was 10.8 (SD±6.7). The Micromedex® mHealth app found 381 PDIs while its web version detected 502 PDIs, with an average of 5.3 and 7.0 and frequency of 61.1% and 72.2%, respectively. According to the severity classification in mHealth and web versions, the following stood out, respectively: 221 and 321 severe; 139 and 149 moderate. The majority (>65%) of identified PDIs had their documentation classified as reasonable. Conclusion: Digital tools although they aid decisionmaking, are not unanimous and consistent in detecting such interactions.


Introduction
In 2018 infectious diseases were responsible for more than 700 thousand hospitalizations in Brazil (Brasil, 2019).
World Health Organization (WHO) data show that hospitalized patients tend to be in polypharmacy regimens-use of five or more comedications -which is considered one of the main risk factors for drug-drug interactions (DDI) (WHO, 2017a).
In this context, the identification of potential DDI (PDDI) is essential to avoid unfavorable clinical outcomes.
Currently, information about PDDI, traditionally available on websites, can be also accessed through mobile health applications (mHealth). These tools make access to drug information easier and more versatile in the dynamic setting of bedside-care.
Annually, 44,000 to 98,000 deaths in the United States occur due to medication errors (Committee on Quality of Health Care in America, 2000). In this context, the dynamism provided by mHealth apps can contribute to the safe use of medicines.
Although mHealth apps could enhance contemporary health practice, important divergences in PDDI identification, severity classification and its clinical management have been shown by comparative studies (Pauly, et al., 2014).
In this context, we conducted a study on potential drug-drug interactions in prescriptions of hospitalized patients admitted into an infectious disease referral hospital and we compared PDDI information's available in web drug interactions checker and their related mHealth apps.

Material and Methods
We conducted cross-sectional study based on a retrospective review of drug prescriptions, available in electronic medical records, from a referral infectology hospital in Center-West Region of Brazil, 2018. This study was approved by the A census was conducted on a date determined by electronic draw, data collection took place on a single day, in March 2018. We selected all prescriptions containing two or more medications. Individuals sociodemographic data and clinical characteristics were extracted from the SOUL MV Hospitalar® software (version 2000), which is the main software system comprising all the hospital's patient data, using an own standard author form. Data were analyzed using the Epi info® platform (version 7.2.2.6 CDC, Atlanta Georgia -USA), Excel® (version 2013) and OpenEpi® (version 3.01). We calculated mean, median, standard deviation for continuous variables and frequency for categorical variables.
The drugs were classified according to the first level system (anatomical or pharmacological groups) of the World Health Organization's Anatomical Chemical Therapeutic Classification (ATC) (WHO, 2017b). Additionally, the drugs were classified according to the route of administration and the number of prescribed drugs: a) two to four medications; b) polypharmacy (5-9 comedications); c) excessive polypharmacy (≥10 comedications). The platform IBM Micromedex® Drug Interaction Checking (IBM Micromedex® Drug Interaction Checking, 2018) platform (Micromedex®) was elected for this study because of its widespread use in the literature and its recognized sensitivity and specificity in identifying PDDI (Jodlowski et al., 2011;Reis & Cassiani, 2011;Cedraz & Santos, 2014). Thus, the PDDI were classified according to the fivelevel severity system and the four-level documentation rating system, as described in Table 1. Good Documentation strongly suggests interaction, but robust studies are lacking.

Fair
Available documentation is limited, but there is pharmacological evidence.

Contraindicated
Drugs are contraindicated for concomitant use.

Major
Interaction may be life-threatening and may require medical intervention to reduce or prevent serious adverse reactions.

Moderate
Interaction may result in aggravation of the health problem or require change in treatment.

Minor
Interaction may result in limited clinical effects. There may be increased frequency of side effects.

Results
We selected 72 individuals and their respective drug prescriptions. Among these, 76.4% (n=55) were hospitalized in wards, 5.5% (n=4) in the emergency room and 18% (n=13) in an intensive care unit. The median age was 38 years, ranging from 1 to 82. Most individuals were male, 80% of them lived in urban areas and the average length of stay was 15.8 days (range 1 to 200). Approximately half of the individuals were hospitalized due to AIDS (Table 2).  Among 779 most commonly prescribed drugs from 126 active substances, sodium chloride and dipyrone were used by more than 80% of study participants (Table 4). We could not find any information about dipyrone and bromopride, the second and fourth most commonly medication prescribed, in the mHealth version of Micromedex®. Dipyrone and Bromopride are widely prescribed in Brazil, unlike most foreign countries, including the United States.

Analysis of Potential Drug-Drug Interactions
In this study analyzed prescriptions, at least one PDDI was found in 73.6% (n=53) of its total. From a total of 505 PDDIs, 75% (n=378) represents the intersection of identified ones in both Micromedex® versions. The web and mHealth version individually identified 502 and 381 PDDI, respectively. The PDDI average per prescription varied according to database version, being 7 (Dp±8.3) in web version and 5.3 (Dp ± 6.8) in mHealth version (p=0.52).
Among the 126 distinct active substance identified in this study, five of them (tenoxicam, ringer lactate, dipyrone, nitrazepam and bromopride) were not available in the mHealth database, resulting in 124 PDDI, classified as contraindicated (n=11), major (n=103), moderate (n=10) that could not be detected by this version. In contrast, three severe PDDI, related to the combination of codeine-acetaminophen with comedications, were uniquely identified by mHealth app.
The severity rating showed that more than 60% of PDDI were contraindicated or severe and most of them (> 65%) were classified with documentation rated as reasonable (Figure 1). Figure 1 illustrates PPIs classified as severe, with frequency ≥0.6%, represented by constantly prescribed drugs. The arrows indicate IMP between the communicating parts. Bold are drugs with high frequency of IMP, which are only in the web version database., are they: Trimethoprim+Sulfamethoxazole : SXT; Rifampicin, Isoniazid and Pyrazinamide: RHZ. There were 31 types of ART-related-PDDIs, the most common of them were found between dipyrone and the combination of tenofovir + lamivudine 1.4% (n=7/502).

Discussion
Our study analyzed drugs prescribed to individuals admitted to a reference hospital in infectology in the central-west of Brazil in order to investigate the occurrence of PDDI but also their information provided by two different versions (web and mHealth app) of the same database. Most of the participants in this study were young people or adults, who were in use of analgesics and antimicrobials, and half of the overall investigated population were individuals living with HIV/AIDS. More than 80% of individuals were under polypharmacy (p<0.01) and a prevalence of 73.6% of PDDI was identified.
Two versions of the Micromedex® database were used in order to identify and classify PDDI and important differences among those data were detected. These differences are mainly explained by the lack of information in the mHealth version on interactions involving the drugs dipyrone, bromopride, tenoxicam, nitrazepam and ringer lactate solution. In this setting, the mHealth version of Micromedex® was in disadvantage, since it did not got updates that contemplated information related to drugs commonly used in Brazil.
Dipyrone, the second most prescribed drug in our study, was identified in 76 PDDIs comedications pairs, about 87% of which were classified as severe. Dipyrone relevant clinical data could not be evaluated in the mHealth version scenario since this application not presented this drug in the available list.
Dipyrone is classified as a non-steroidal analgesic and anti-inflammatory drug (NSAID) and its effects as analgesic and antipyretic are unquestionable (Rang, et al., 2007). In Brazil dipyrone is included in the list of over-the-counter drugs (OTC) (ANVISA, 2020). However, safety information about the use of this medicine diverges, mainly because of the rare but serious adverse reactions such as aplastic anemia, Stevens-Johnson syndrome, toxic epidermal necrosis and agranulocytosis (Magni, et. al., 2010).
Drug interactions related to bromopride, which could not be identified in the mHealth version, accounted for 9% (n=44) of total PDDI identified, 10 of them were considered contraindicated and 34, severe; both due to the risk of extrapyramidal reactions. The incidence of these reactions may be even higher when intravenous doses of bromopride are required, a warning alert, since more than a half of the drugs analyzed in this study were administered by this route (Tonini, et. al., 2004).
We highlighted that 116 antimicrobial-related-PDDI were identified in both versions of Micromedex® database which is a relevant event since the study setting was a referral infectology hospital. The most common antimicrobial-related-PDDIs were: sulfamethoxazole + trimethoprim (6.8%), azithromycin (4.8%) and rifampicin + isoniazid + pyrazinamide (4.6%). Yet, among those, severe PDDI identified were: azithromycin and fluconazole, azithromycin and ondansetron as well as sulfamethoxazole + trimethoprim and fluconazole, which were related to the risk of ventricular arrhythmias (QT interval prolongation) and ventricular fibrillation (Roden, et al., 2016).
It is worth highlight that azithromycin, among others antimicrobials, show electrophysiological effects similar to class III antiarrhythmic drugs (MARTINS et al., 2015) and fluconazole may prolong the QT interval, either directly or by inhibiting the hepatic metabolism of other agents that have direct action under this signal, such as atazanavir an antiretroviral drug involved in treatment of HIV infections (Molloy, et al., 2018).
About a half of the analyzed population was living with HIV/AIDS and almost 20% of individuals were on antiretroviral therapy (ART). In these setting, DDI may be an adjuvant factor in ART failure and an important cause of prejudice in treatment adherence (Molas, et al., 2018). The most frequent ART-related interaction of dipyrone and the combination of tenofovir + lamivudine 1.4% (n = 7/502) is related to the risk of renal failure, so patients should be adequately monitored for glomerular filtration rate (GFR), especially patients with GFR> 60 mL /min. Additionally, it is noteworthy that this PDDI could not be detected in the mHealth version (Machado, et. al., 2014).
Monitoring ART-related-PDDIs is essential to improve the treatment of people living with HIV (PLHIV) (BRASIL, 2013). In this study, information provided by The Micromedex® web version indicated that all patients on ART had at least one interaction associated with antiretroviral drugs, while the mHealth version showed at least one ART-related-PDDI in 70% of prescriptions. Although the present study was conducted with prescriptions of hospitalized patients, where the chances of self-medication are minimized, it was found that the scenario regarding the occurrence of PDDIs is not more optimistic than that in an outpatient treatment setting, whose PDDIs prevalence varied from 23.6% to 52.2% according studies conducted in different regions of Brazil (Cascao, et al., 2017;Santos, et al., 2016).
The majority of participants underwent to the main risk factor for PDDIs since according to the literature, polypharmacy is the main risk factor for drug interactions (Monegat et al., 2014). Approximately 80% of the individuals in our study received prescriptions with 5 or more medications, and a half of the young and adults individuals received prescriptions containing 10 or more comedications (excessive polypharmacy).
This study innovates since it brings highlights to intrabases divergences in the mHealth era trough comparative information about PDDIs in different versions (mHealth and web) of a largely database (Micromedex®) applied in the literature. In addition, the present study also contributes in a better comprehension of the occurrence of PDDIs in hospitalized patients in a infectious diseases scenario, an unusual theme in the PDDIs investigation setting. Our study presents as its main limitation the impossibility of verifying if, in fact, such identified PDDIs occurred in clinical practice, though we did not have the propose to investigate it, since patient follow-up was not within the scope of this study.
Access to computerized databases, electronic prescriptions and alert programs are mechanisms of remarkable importance for the prevention and management of drug interactions (Correr & Otuki, 2013). However, it is necessary to standardize the contents of these databases and their different versions, so that any doubts on the appropriate clinical management are properly addressed.

5.Conclusion
This study demonstrated a high frequency of polypharmacy and a high frequency of contraindicated and severe pDDIs in prescriptions of hospitalized patients at a referral hospital in infectious diseases, Center-West Region of Brazil. Despite the clear advantages related to the versatility, mobility and optimization in the use of Micromedex® mHealth version in clinical practice, we have identified that pDDIs information related to frequently prescribed drugs in Brazil, such as dipyrone and bromopride, was only available in the web version of the database.