Clinical decision support systems: great promises for better management of patients’ drug therapy

KEYWORDS:

• Adverse drug events
• clinical decision support system
• drug-drug interactions
• software

View correction statement:

Erratum

1. Introduction

Life expectancy has increased significantly over the years. This condition favors the occurrence of polymorbidities which, per implementation of disease treatment guidelines, lead to the use of multiple drugs (polypharmacy). Polypharmacy should not be viewed only as the use of multiple drugs in one patient but rather as the use of too many inappropriate drugs, including some to counterbalance other drugs’ side effects, in a single patient. In a frail elderly population, polymorbidities and polypharmacy increase the risk of drug–drug interactions (DDIs) and associated adverse drug events.[1]

A DDI is defined as a circumstance under which two drugs or more influence each other’s pharmacokinetic and/or pharmacodynamic actions. These influences may result in reduced or increased efficacy or in reduced or increased toxicity. We and others have clearly demonstrated that the prevalence of DDIs increases as the number of drugs being prescribed augments.[2,3] For instance, the probability of at least 1 significant DDI was 50% in patients taking 5–9 drugs, 81% with 10–14 drugs, 89% with 15–19 drugs, and 100% with 20 or more drugs.[3] The addition of each medication to a 5-drug regimen conferred a 12% increased risk of potential DDIs.[3]

It is difficult to dissect the impact of DDIs to that of adverse drug events as both phenomena are intermingled. But clearly, studies have shown that DDIs and associated adverse drug events may cause up to 2.5–4.4% of all hospital admissions.[4] Furthermore, preventable adverse drug events and DDIs prolong hospital length of stay (3.37 days), increase the cost of treatment ($3511), and elevate the risk of death.[5,6] According to the 2007 Institute of Medicine report entitled Preventing Medication Errors: Quality Chasm Series, approximately 1.5 million preventable adverse drug events occur annually in the USA.[7] These events can even translate into death, as for the last 10 years, Center for Drug Evaluation and Research reports indicate that adverse drug events rank between 4 and 6 as a leading cause of mortality.

DDIs and adverse drug events are avoidable with proper recognition of interacting drug pairs and inappropriate combinations as well as appropriate action.[8] However, one has to recognize that the number of possible combinations and load of manageable information become rapidly unbearable when a patient’s drug regimen comprises 10, 15, 20, or more drugs. Not surprisingly, studies show that prescribers’ and pharmacists’ ability to recognize well-documented drug interactions is limited, if not lacking.[9,10]

2. The challenge: to manage DDIS

Several groups and organizations have deployed significant efforts to develop databases and drug interaction screening software (DISS). Per definition, DISS mostly screens the literature and report one by one drug pair risk of DDIs and side effects or unintended outcomes. There is no factual consideration of patient’s specific conditions built in DISS. Nevertheless, these technologies have been proven effective in detecting drug interactions, improving compliance and pharmacological management in high-risk patients, and improving overall clinical management and patient outcomes.[11] Studies comparing the efficiency of various DISS showed relatively good sensitivity and specificity among them (≈90%).[12] However, despite these advantages, several issues were raised concerning DISS: limited availability and under-reporting of adverse drug events (especially for locally developed databases); lack of differentiation between pharmacokinetics and pharmacodynamics interactions; and absence of a severity rating, lack of management strategy, or lack of access to reference literature. But, the major downside associated with several DISS or databases is the over-alerting (alert fatigue) of a large number of DDIs of low clinical relevance.[13] In a review of 30 million prescriptions dispensed in a community pharmacy, 70.8% of initially detected DDIs were removed when applying additional filters to increase specificity and an additional 80.6% of DDIs were removed when reviewed by pharmacists. Ultimately, only 5.7% of initially detected DDIs were considered as clinically relevant.[14]

These observations support the basic principles of the approach we and others have adopted while trying to develop meaningful clinical decision support systems (CDSS).[3,15,16] CDSS use literature information and patient’s own condition (renal function, all other concomitant drugs, time of administration and dosage regimen, pharmacogenomic data pertaining to drug metabolizing enzymes or drug transporters, specific disease genetics) to construct recommendations based on predefined algorithms. Hence, according to us:

1. A DDI analysis by a DISS based on multiple one-to-one drug pair comparisons would often generate numerous incoherent results in patients with polypharmacy (for instance: analyzing drug A effects on drug C suggests to decrease drug C dose, but analyzing drug F effects on drug C suggests to increase drug C dose);

2. No DISS system would be able to provide accurate, meaningful, and clinically relevant recommendations for patients on multidrug regimens since DISS do not consider several factors inherent to a patient’s condition, environment, time and order of dosing, etc.

3. In contrast, proper CDSS must permit at a glance, rapid access to the most complete and accurate information about all drugs in a regimen and should let clinicians elaborate meaningful recommendations considering patient’s age, background, past medical history, disease condition, time of dosing, dose being used, personal situation, genetics, etc.

4. CDSS development should involve more clinicians, and decision-support algorithms should make more use of patient-specific information from electronic records and laboratory results.

So, an ideal CDSS should:

1. Interface with the patient’s electronic medical record or pharmacy chart;

2. Interact with pharmacogenetic information pertaining to a patient’s condition;

3. Be able to assess simultaneously multiple drug interactions from a patient’s full medication list;

4. Include meaningful analysis of pharmacokinetic interactions (Cytochrome P450s (CYP450s), other enzymes, and transporters);

5. Include information and mechanisms underlying pharmacodynamic interactions;

6. Provide rapid access to up-to-date information supporting mechanisms and clinical significance of identified drug interaction;

7. Include tables and/or allow for simulations with other similar drugs (same drug class) so that clinicians could easily select alternatives to replace interacting medications;

8. Make available detailed pharmacokinetic characteristics of drugs (amount excreted unchanged, bioavailability, partial metabolic clearance, and CYP450 isozymes);

9. And ideally, give an order of magnitude of the expected changes in plasma drug levels or actions for the drug being the target of drug interaction.

Today, CDSS have been developed commercially that facilitate consideration of a patient’s condition while taking several drugs (e.g. MediQ in Europe,[16] InterMed-Rx in Canada,[3,15] Eirene-Rx/MedWise Advisor, USA). Patient characteristics and a summary of relevant information are compiled from medical records (age, gender, renal function, electrolytes and laboratory results, allergies, genetic information) while drug-related information such as anticholinergic burden, sedative load, Beers criteria, bioavailability, urinary excretion, substrate affinity, enzyme inhibition, extent of each metabolic pathway, non-cytochrome P450 interactions and pharmacodynamic interactions are included to be used in comprehensive medication risk assessments. Direct access to literature reviews and publications is also granted as well as access to detailed pharmacokinetic parameters and pharmacogenomic analysis. Using systematic screening approaches, CDSS can identify patients at risk of significant DDIs and facilitate interventions aimed at reducing adverse events.[3,15]

3. Conclusion

Adverse drug events and DDIs have become a major public health issue. Drug interaction screening software and electronic databases have been developed to augment clinicians’ ability to detect clinically significant drug interactions and improve patient safety. However, systems analyzing multiple drug pairs in a sequence often alert clinicians of irrelevant interactions and generate too many intrusive alerts that are mentally draining and time-consuming leading care providers to disable or ignore these alerts. Also, these systems often generate recommendations that are not relevant and do not account for patients’ overall conditions. A different approach is to develop clinical decision support systems (CDSSs) aimed at providing accurate and complete information about all of a patient’s medications, at a glance, in a timely fashion. CDSS let clinicians elaborate their own recommendations and treatment strategies while considering other aspects of each patient’s conditions, genetics, past medical history, and environment. Such CDSS have proven their value in elderly patients on complex medication regimens, and have been associated with favorable clinical outcomes (hospitalization rates) as well as economic impacts. In my opinion, CDSS aimed at providing accurate information that allows clinicians to elaborate meaningful recommendations integrating several of a patient’s clinical and personal variables shall become of greatest value in the near future.

4. Expert opinion

Clinical decision support system (CDSS), taking into consideration various elements of a patient’s condition and environment, should be preferred over straightforward drug interaction screening software (DISS) to decrease drug-related problems and hospitalization. On one hand, DISS create alert fatigue as they attempt to report as many as possible potential drug interactions. On the other hand, CDSS create enthusiasm and a sense of accountability as they provide to health professionals complete information on potential drug interactions while allowing clinicians to elaborate appropriate recommendations and treatment strategies according to a patient’s condition. CDSS should interface with a patient’s electronic medical record or pharmacy chart and pharmacogenetic information pertaining to the patient’s condition, be able to assess simultaneously multiple drug interactions, include meaningful analyses of pharmacokinetic and pharmacodynamic interactions, provide access to literature data supporting proposed mechanisms, offer to clinicians a list of drugs from which alternatives can be selected, and ideally give an order of magnitude of expected changes in plasma levels or actions. Ultimately, the objective of these electronic tools should remain to prevent, identify, and decrease problems caused by DDIs and adverse drug events on patient health.

Declaration of interests

J Turgeon holds stock in InterMED-Rx and Tabula Rasa Healthcare. V Michaud holds stock in InterMED-Rx. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed. Data Management was provided by Gabriel Badeh (InterMED-Rx).

Acknowledgement

The authors would like to acknowledge the participation of Dana M. Filippoli, MA in the review of the manuscript.