Prescription drugs commonly evoke debates about value. Defining and measuring drug value, broadly or specifically, is complex. Recent and increasing discussion of drug price and price transparency in the US highlights the importance of measuring and communicating drug value (Califf and Slavitt, 2019). The need to explicitly measure value is most apparent for new, expensive drugs, but for every individual, each prescription use decision represents a trade-off between anticipated benefits, possible harms, cost and opportunity cost. To advance this discussion of value and inform debates about price, the authors propose a taxonomy of population-level drug use. With this taxonomy, broad categories related to the most probable balance of benefit and harm to support discussions about cost have been created. The authors propose that the taxonomy be employed for broad examination of drug use at the population level. When sufficient granular data are available, the taxonomy could be applied to individual-level clinical situations. The proposed taxonomy builds on past efforts to classify drug use patterns; the current expanded effort will provide rich context for ongoing discussions of drug use and ultimately value (Greenway and Ross, 2017; Lexchin, 2017).
To date, efforts to improve drug use, maximise benefits, minimise harms and optimise allocation of limited resources, include diverse lists and guidelines for prescribers, as well as explicit coverage policies by health insurance plans. The World Health Organization (WHO, 2019) maintains the Essential Medicines List, which includes a relatively narrow selection of drugs deemed effective, safe and responsive to the most pressing healthcare needs of a population. In some countries, governmental organisations strive to identify the highest value drugs for inclusion in national formularies, a list of drugs covered by a national health plan. In the UK, for example, the National Institute for Health and Care Excellence (NICE) fills this role. NICE (2018) conducts evidence and cost effectiveness analyses, seeking stakeholder input in its decision-making process. The resulting formulary reflects the nation's best attempt to rationalise drug access policies and manage prescription spending to the greatest good of the population. In the US, health plan formularies are similarly set by payers for their enrolled populations. Examples include commercial insurance companies, the Veterans' Administration, and Medicaid programs. However, in these cases of decision making, the relative role played by cost, benefits, harm and negotiated rebates, is not generally transparent (Cole et al, 2019). In general, population or budget-based priorities guiding these diverse lists may or may not reflect the drug needs of every individual; they are intended to guide collective drug use. An appeal process often affords opportunity to apply for exceptions to insurance coverage restrictions (Centers for Medicare and Medicaid Services (CMS), no date).
Other efforts aimed at optimising drug use focus on limiting potential harm by guiding prescribers or empowering informed patient choice. For example, the Beers Criteria are employed to generate lists of medications that are ‘typically best avoided by older adults in most circumstances or under specific situations’ (American Geriatrics Society (AGS), 2019). This and similar lists aim to help prescribers identify and reduce use of potentially harmful medications in individuals at highest risk for adverse drug effects (AGS, 2019). Efforts such as the Drug Facts Box have focused on concisely communicating benefits and risks of drugs, to help consumers and doctors optimise drug use decisions for a given individual, a specific drug or drug group, or a given indication (Schwartz and Woloshin, 2013).
The authors have built on their past effort to measure, categorise and communicate drug use patterns at the population level. Their 2013 Dartmouth Atlas of Prescription Drug Use describes effective, potentially discretionary, and potentially harmful drug use and measures regional variation in the use of these categories among Medicare Part D enrollees (Munson et al, 2013). A refined version of the prior taxonomy aimed at enriching discussion of population-level prescription drug use patterns is proposed. The cost or cost-effectiveness is not generally explored but rather the paper is focused on the range of benefit-risk trade-offs that are believed to be essential to meaningful discussions of drug cost and ultimately value.
Approach
The authors began by acknowledging that a drug's benefit-risk balance depends on the situation in which it is applied. It is recognised that evidence and patient-level data are often insufficient to permit precise prediction of a drug's health impact for an individual, but evidence (and lack of evidence) can inform estimates that permit broad categorisation of drug use patterns. This proposed taxonomy of population-level drug use patterns relates to the most probable health impact based on a selective literature review, the author's previous work and insights from experts. A selective literature review was performed to capture the main resources that are commonly used to either guide prescriber's prescription decisions or assess prescribing decisions. This literature review also yielded examples of prescription drug use that represent each of the proposed taxonomy domains described in this paper.
Overall, the method builds on the 2013 Dartmouth Atlas of Prescription Drug Use (Munson et al, 2013), in which three categories of prescription drug were chosen to permit study of prescription drug use applying (and modifying as needed) the Dartmouth Atlas of Healthcare framework to classify healthcare services. Prior to the authors' 2013 publication, the Dartmouth Atlas team classified healthcare services into three categories, according to the factors that primarily drive their provision or consumption: effective care, preference-sensitive care, and supply-sensitive care.
The first of these categories is effective care, which includes services that are of proven value, have no significant tradeoffs and are generally backed by well-articulated medical theory and strong evidence of efficacy, determined by clinical trials or valid cohort studies (Dartmouth Atlas of Healthcare Project, 2007a). The analogous effective category for prescription drug is well-aligned with this definition and includes treatments that are widely viewed as effective (Munson et al, 2013).
The second category, preference-sensitive care, includes services that have significant trade-offs that affect a patient's quality and/or length of life and decisions about these services should reflect patients' well-informed personal values and preferences. (Dartmouth Atlas Project, 2007b). In the authors' taxonomy of prescribing, preference-sensitive care is categorised, such as potentially discretionary drug use; this category of prescribing includes treatments that involve a high degree of prescriber and/or patient discretion (Munson et al, 2013).
The third category of the Dartmouth Atlas of Prescription Drug Use was potentially harmful medication use. This category includes treatments with good evidence of harm in a specific population (Munson et al, 2013). In consultation with experts in the field of prescription drug use to create a more inclusive categorisation, the authors expanded the Atlas of Prescription Drug Use taxonomy to include two additional categories: wasteful drug use and lifestyle drug use, which are defined below.
The final category defined by the Dartmouth Atlas is supply-sensitive care, which includes services whose frequency is not driven by well-articulated medical theory or scientific evidence, but rather by the service supply (eg hospital beds or surgeons) (Dartmouth Atlas Project, 2007c). While it is not possible to say for certain that supply-sensitive care causes harm to patients, preliminary studies have found greater mortality rates and no increase in patient quality of life in regions with higher rates of supply-sensitive care (The Dartmouth Atlas Project, 2007c). As prescription drug supply chains in developed countries are vast, supply was not considered a key factor in prescribing patterns.
Proposed taxonomy
This paper proposes the following taxonomy of prescription drug use: effective, potentially discretionary, potentially harmful, wasteful and lifestyle. Table 1 defines each domain and provides illustrative examples.
Table 1. Proposed Prescription Drug Use Taxonomy, Domains and Examples
Domain | Definition | Examples |
---|---|---|
Effective | Supported by evidence of improved health outcomes, may inform guidelines and performance metrics |
|
Potentially discretionary | Shown effective in a relatively narrow population. A subset of observed drug use is ‘effective’, but much common use does not meet criteria for effective use |
|
Wasteful | Drug use that does not add additional health risk but is not supported by evidence as superior to lower cost equivalents |
|
Lifestyle | Drug use that is not medically necessary for physical health; the risk benefit tradeoff and thus the value are best determined by the well-informed patient who often bears the cost |
|
Source: American Geriatrics Society; 2019; O'Mahony et al, 2015
Effective drug use patterns are those supported by evidence demonstrating improved health outcomes (Munson et al, 2013). On average, this type of drug use confers benefits in excess of associated risks for patients aligned with evidence (disease, disease stage, population with demonstrated benefit) (Munson et al, 2013). Drug use in this category is generally broadly accepted by clinician groups, often included in clinical practice guidelines, and can inform quality measures (National Committee for Quality Assurance (NCQA), 2018). Effective prescription use is supported by high-quality evidence, often clinical trials assessing health outcomes or commonly-accepted surrogate endpoints (Neumann and Cohen, 2015). Examples of effective drug use include continuous beta blockers and statins for acute myocardial infarction survivors (NCQA, 2018), disease-modifying anti-rheumatic drugs for rheumatoid arthritis patients (NCQA, 2018), antihypertensives among hypertensive patients (Whelton et al, 2018), angiotensin active antihypertensives among diabetes patients with hypertension or nephropathy (NCQA, 2018), statin use among diabetics with high cardiovascular risk (NCQA, 2018), antibiotics for susceptible bacterial infections (Centres for Disease Control and Prevention, 2017), vaccines among those vulnerable to a target infection (NCQA, 2018), antiretrovirals for human immunodeficiency virus (Panel on Antiretroviral Guidelines for Adults and Adolescents, 2019), many chemotherapy regimens (American Society of Clinical Oncology, 2019) and contraception (Curtis et al, 2016).
Potentially discretionary drug use involves broad use of drugs shown to be effective in a narrowly defined set of individuals. Being outside the narrowly defined evidence base, these drug use patterns are of uncertain benefit and harm. The decision to use a drug in such cases may depend on nuanced clinical inputs, clinician practice culture and, ideally, informed patient choice (Munson et al, 2013). Such drug use may occur among patients with mild forms of the approved indications (eg antidepressants for treatment of short-term depression symptoms or dysthymia) (CMS, 2013), duration longer than approved (such as long-term use of new generation sedative hypnotics approved for short-term treatment) (McMillan et al, 2013), or conditions similar to approved indications (such as gabapentinoids for chronic musculoskeletal pain) (Pfizer, 2018). Additionally, individuals may have the condition for which the drug has been shown effective but fall outside the population in which the drug was tested (such as pediatric use of drugs tested and approved in adults). Potentially discretionary drug use has been suggested in some US use of antidepressants (CMS, 2013), antipsychotics, proton pump inhibitors (Heidelbaugh et al, 2012; CMS, 2015), opioids (Dowell et al, 2016) and sedatives (Choosing Wisely, 2017).
Potentially harmful drug use patterns are those that generally can be expected to confer more risk than benefit for a specified population. This is especially important when lower-risk alternatives exist. Such drug use may be warranted in rare cases. There are several lists of potentially inappropriate medications that should be avoided in the elderly (eg HEDIS, Beers and STOPP). The authors assign these to the potentially harmful drug use category. Additionally, Prescrire publishes a yearly list of marketed drugs that have evidence indicating more harm than benefit and should be avoided (Prescrire, 2019). Other examples of potentially harmful drug use patterns include the use of testosterone in men over the age of 50 with coronary artery disease (Morden et al, 2019), contraindicated drugs in dementia patients (AGS, 2019), thiazolidinediones in patients with heart failure (Castagno et al, 2011) or estrogen replacement without progesterone replacement in women who have not had a hysterectomy (American College of Obstetricians and Gynecologists, 2020).
Wasteful drug use is that which, on average, does not add additional health risk but is not supported by evidence as superior to lower cost equivalent alternatives. This is the only category in the proposed taxonomy that explicitly considers cost. These patterns of drug use often cost the recipient, the healthcare system, payers, and society, without additional health gain. Examples are the use of brand name drugs when lower-cost generics are available and the use of expensive drugs in patients who have not yet tried and failed less expensive alternatives (eg expensive new drugs for diabetes mellitus before a trial of metformin) (Wexler, 2019).
Lifestyle drug use is not medically necessary for physical health but generally confers minimal risk. For such drug use decisions, the benefit-risk trade-off and thus the value are best determined by the well-informed patient who often bears the cost of the drug use decision. Since much drug use in this category is not included in prescription insurance plans (or is included in a limited way), the paying customers only obtain drugs in this category if the perceived potential benefit warrants the price. Examples of this domain include botulinum toxin injections, sexual dysfunction treatments, retinoids for wrinkles, and eyelash lengtheners.
Discussion
The proposed taxonomy sorts prescription drug use patterns into broad categories, with the intention of advancing descriptions of population-level drug use and to provide context for value debates. The authors hope to support comparison of prescription drug use across regions and populations, framed by the expected balance of benefits and harm based on available evidence, ranging from effective to potentially harmful.
This proposed taxonomy has several limitations. First, the taxonomy relies on relatively non-controversial examples demonstrative of the range of drug use suggested by published work on this topic, as well as clinical and pharmacology insights. The authors recognise that for many drugs, the balance of benefit and harm will depend on diverse, individual factors including the specific product selected, the condition for which it is prescribed, age, comorbidities, genetics, and other drugs with which the prescribed product will be combined. While many of these factors are known at the time a prescription is written, filled, and then consumed, often much information needed to anticipate impact on outcome is not known because patient-level data are missing (especially pharmacogenetic data) and the patient's specific combination of condition, demographic and other drug mix has never been explicitly studied. Despite the very individual nature of drug use decisions, the authors believe there is value in measuring population-level drug use as a high-level signal of drug use intensity and probable benefit-risk balance or (in the case of potentially discretionary drugs) uncertainty. Additionally, the paper's wasteful category assumes that on average when generics are available, generics are the more cost-effective option. However, there are instances where brand use may be appropriate because of an arrangement between a payer and manufacturer that makes a brand drug cheaper than its generic alternative (Dusetzina et al, 2019), and in some drug categories generic equivalence is debated (Yamada et al, 2011; American Thyroid Association, 2004). The authors do not pretend to understand which patterns are optimal, and it is believed optimal drug use varies by population and culture as a function of illness, resources, patient preference and clinician practice culture.
The fact that much drug use falls into the potentially discretionary category because of limited evidence and relatively high use intensity (eg opioids, antidepressants, sedatives) is an important limitation. That many potentially discretionary drug use examples involve treatment of subjective symptoms (eg pain, depression, insomnia) suggests preference may and likely should strongly influence drug use, as long as the decision is well informed. This caveat applies to lifestyle drugs as well. Drugs used to treat sexual dysfunction could be classified as lifestyle or medically essential, depending on how one frames the role of sexual health in overall health and wellbeing. There is no commonly agreed upon demarcation between lifestyle and health, though we have not yet heard eyelash lengtheners described as contributing to health. Finally, the authors limited their effort to a selective review of literature focused on population-level drug use.
While the examples used draw largely from prescription drug use observed in the US, this paper believe the concepts and categories are broadly applicable to all developed nations, where prescription drug access is relatively broad and the evidence base is common to all.
This taxonomy has been designed to advance discussion and understanding of observed wide variation in population-level drug use across geographic regions and subpopulations. The aim is to support insightful descriptions of prescription use, going beyond intensity and spending, and getting closer to anticipated benefit-harm balance.
The authors hope the proposed taxonomy will support physicians, researchers and policy makers concerned with optimal drug use from the societal perspective. This taxonomy may help healthcare systems think about internal drug use patterns to better understanding the range of observed patterns across clinical groups and relative peer organisations. It is expected that others will build upon this work to support evolution and use of broadly applicable taxonomy that can inform drug value discussions including, but not limited to safety, clinical outcomes and cost.
Key Points
- This paper proposes the following taxonomy of population-level prescription drug use: effective, potentially discretionary, potentially harmful, wasteful and lifestyle
- Effective drug use patterns are those supported by evidence demonstrating improved health outcomes
- Potentially discretionary drug use involves broad use of drugs shown effective in a narrowly defined set of individual.
- Potentially harmful drug use patterns are those that generally can be expected to confer more risk than benefit for a specified population.
- Wasteful drug use is that which, on average, does not add additional health risk but is not supported by evidence as superior to lower cost equivalent alternatives
- Lifestyle drug use is not medically necessary for physical health but generally confers minimal risk.
CPD reflective questions
- How can categories of value advance our debates on drug price and total drug spending?
- What are some challenges to classifying prescription drug use patterns at the population level?
- How can each of the follow groups (policy makers, payers, physicians, researchers) use this framework to support optimal drug use from a societal perspective?