Adverse drug events (ADEs) are a common and costly problem. For example, an estimated 1.5 million preventable ADEs occur in the USA annually at a cost in excess of US$4 billion Citation[1]. Older adults are often at particular risk owing to accumulated medical comorbidies, complex medication regimens, declining functional status and physiologic changes associated with less-predictable pharmacokinetics and pharmacodynamics Citation[2,3].
One proposed ADE-reduction strategy involves the development of algorithms to identify patients at high risk. A recent example is the GerontoNet score, an algorithm designed to predict ADE risk among hospitalized older adults Citation[4]. As is common in ADE risk algorithms, the number of medications taken by the patient dominates the score: receiving eight or more medications yields 4 points, and between five and seven medications garners 1 point. Then, 1 additional point is received for each of the following: heart failure, renal failure, liver disease and four or more medical conditions. Finally, patients with a history of prior ADE receive 2 points, which yields a potential score from 0 to 10.
There are at least two general applications where ADE risk algorithms, such as the GerontoNet score, could prove to be effective. The most commonly described approach involves targeting high-risk patients for a concurrent or retrospective review by another clinician, often a clinical pharmacist, who would make recommendations to the primary prescribers. A second approach is to make ADE risk scores available via real-time decision support systems, which could assist prescribers in making risk–benefit decisions for new medications, choosing optimal dosing schedules and implementing appropriate monitoring plans. So are there any reasons why we should not be rushing to implement ADE risk algorithms in practice? After all, it seems pretty straightforward. Make scores available in real-time to prescribers and dispatch clinical pharmacists to patients with high scores – and the problem is solved. While risk algorithms are promising, there are several reasons why their use may not translate into significant ADE reductions in real-world practice.
First, the utility of ADE risk algorithms in guiding clinical intervention is contingent upon the magnitude of risk stratification achieved. For example, ADE frequency was 2% with GerontoNet scores of 0–1, and gradually increased to 21.7% with scores ≥8 Citation[4]. Unfortunately, this impressive level of stratification has not been consistently replicated. Some prominent investigators have even concluded that ADE risk stratification approaches are unlikely to be productive Citation[5]. There are many potential reasons why ADE risk algorithms may not demonstrate consistent risk stratification. Perhaps the most fundamental is that risk prediction is not typically limited to one drug (e.g., warfarin) or one type of ADE (e.g., bleeding), but all ADEs caused by all drugs. Risk factor models are typically constructed for a single condition, such as cardiovascular risk or cancer survival, but imagine the challenge in building a single risk model for all disease. Yet, in looking for any ADE caused by any drug, this is a corollary of what ADE risk algorithms attempt to do.
Second, other ADE characteristics may also influence the validity and utility of ADE risk algorithms. Were all ADEs included or just those considered to be serious and preventable? Directing additional healthcare resources toward nuisance side effects or events that cannot be prevented would be wasteful. This concern is more than theoretical, as the influence of some risk factors has been observed to change when modeling all ADEs versus serious ADEs Citation[6]. The nature of candidate risk factors is also important in regard to preventability. A common feature of ADE algorithms, including the GerontoNet score, is that few of the risk factors are readily modifiable. Therefore, if a patient is targeted for intervention based on the presence of nonmodifiable factors, it is unclear precisely what prescribing changes can be made to reduce that individual’s ADE risk. Taking this idea one step further, it is important to note that ADE risk factors are generally not novel concepts. For example, the risk factors included in the GerontoNet score will not surprise any practicing clinician. Prescribers routinely consider factors such as the number of concurrent medications, history of ADEs, medical comorbidity and impaired drug clearance in the medical decision-making process. Therefore, the incremental contribution of using a mathematical algorithm is limited to providing quantification to well-recognized qualitative ideas. An intriguing test of utility would be to compare ADE prediction accuracy between mathematical algorithms and clinical judgment.
Third, treatment setting is an important consideration. Risk factors validated in the inpatient setting may not apply to outpatients, and separate risk factor models are probably required. Change also plays a key role in ADE risk prediction. Patients are at greatest risk for ADEs after some important changes, such as the initiation of a new medication, a change in dose or changes in underlying disease state. This may be less important for inpatient settings where frequent changes in medications and symptomatic status are nearly universal. However, outpatients with large stable drug regimens may be at less risk than patients with fewer medications, but more changes. Other factors, such as regimen complexity and patient noncompliance, are also more likely to be important ADE risk factors in the outpatient setting. Reducing ADE frequency and severity in the inpatient setting is probably a more readily solvable problem, but successful outpatient interventions could have greater impact.
Even if a valid and consistently replicable ADE risk algorithm is developed, there are important limitations to implementation. Algorithms will only be useful if the risk factors included in the model are widely available and easily accessible. For all practical purposes, they must be fully automated using electronic medical record data. For example, this is a limitation to the GerontoNet score, where risk factor data were gathered by study physicians from all available medical record information Citation[4]. Clearly, this approach would not be cost–effective in real-world implementation. Many of the risk factors have the potential to be translated into automated algorithms, but it is unclear if the same results would be found. Reliance on electronic medical record information will limit widespread implementation owing to variation in the availability of certain risk factor data across different healthcare systems and electronic medical record software. This is a particular barrier for outpatient settings where electronic medical records are incomplete and fragmented.
Finally, the utility of conducting interventions based solely on ADE risk, without weighing them against potential benefits, may be inherently flawed. If the only important goal was to eliminate ADEs, then the simple solution is to stop prescribing medications. Obviously, this approach can be readily dismissed because it fails to acknowledge that benefit commonly outweighs risk in the prescribing equation. Failure to consider benefit has at least two important implications for interventions based on ADE risk factor algorithms. First, patients do not reach high ADE risk at random, but have been selected by weighing this risk against potential benefits. Targeting these individuals may simply point to patients whose high risk is counterbalanced by significant benefit, but where no prescribing intervention is needed. This case provides a further example of how a valid ADE risk factor algorithm may not translate into a clinically effective intervention. The second implication is that interventions focused exclusively on ADEs represent a missed opportunity to optimize benefit. It has been estimated that only 4–21% of patients receive the optimum benefit of pharmacotherapy, owing to a variety of prescribing errors and inefficiencies Citation[7]. Therefore, developing algorithms and interventions for only one half of the risk–benefit equation appears short-sighted. Instead, research efforts may be better spent identifying patients with unfavorable risk–benefit ratios, which would include markers of unrealized benefits, as well as ADE risk. Targeted individuals would receive a comprehensive intervention designed to simultaneously optimize benefit and safety. This approach seems more likely to translate into clinically meaningful interventions compared with those solely focused on ADE risk factors.
The development of ADE risk factor algorithms is an important line of inquiry. However, the ultimate test is whether interventions tied to these algorithms will reduce ADE-related morbidity and mortality in real-world clinical practice. Interventions narrowly designed to reduce ADEs without concurrent effort to optimize benefits could be intrinsically flawed. Even if successful, ADE-focused interventions would be a missed opportunity to enhance the overall effectiveness of pharmacotherapy.
Financial & competing interests disclosure
The author has no 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. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.
No writing assistance was utilized in the production of this manuscript.
References
- Committee on identifying and preventing medication errors; Institute of Medicine. Preventing Medication Errors. Aspden P, Wolcott J, Bootman JL, Cronenwett LR (Eds). National Academy Press, DC, USA (2007).
- Spinewine A, Schmader KE, Barber N et al. Appropriate prescribing in elderly people: how well can it be measured and optimised? Lancet.370(9582), 173–184 (2007).
- Corsonello A, Pedone C, Antonelli Incalzi R. Age-related pharmacokinetic and pharmacodynamic changes and related risk of adverse drug reactions. Curr. Med. Chem.17(6), 571–584 (2010).
- Onder G, Petrovic M, Tangiisuran B et al. Development and validation of a score to assess risk of adverse drug reactions among in-hospital patients 65 years or older: the GerontoNet ADR risk score. Arch. Intern. Med.170(13), 1142–1148 (2010).
- Bates DW, Miller EB, Cullen DJ et al. Patient risk factors for adverse drug events in hospitalized patients. Arch. Intern. Med.159(21), 2553–2560 (1999).
- Evans RS, Lloyd JF, Stoddard GJ, Nebeker JR, Samore MH. Risk factors for adverse drug events: a 10-year analysis. Ann. Pharmacother.39, 1161–1168 (2005).
- Garfield S, Barber N, Walley P, Willson A, Eliasson L. Quality of medication use in primary care – mapping the problem, working to a solution: a systematic review of the literature. BMC Med.7, 50 (2009).