To the best of our knowledge, this is one of the first studies to describe the prevalence of unplanned hospitalizations caused by ADRs determined by use of a reliable and valid algorithm among older Veterans using a population-based sample. These findings are relevant because they come from a nationally representative sample, allowing for extrapolation to the entire older Veteran population in the United States (US). Most previous studies of ADRs leading to hospitalization in older adults have been limited by the use of a sample from a more select population (e.g., individual hospital or region).3,15
Moreover, these findings were drawn from VA electronic health records, which have been shown to enhance ADR detection.35
The VA’s advanced computer system has wide implications for improving the detection of drug-related problems, which is an essential step in determining future areas for intervention.
We found that 10% of unplanned hospitalizations were caused by ADRs and that over one-third of these hospitalizations were preventable. Between FY04—FY06, 2,430,186 older Veterans received care in a VHA facility, of whom 13.5% (n=328,166) were hospitalized at least once. Given this, we can extrapolate using our data that two-thirds of these hospitalizations were unplanned and that 10% of these hospitalizations were due to ADRs. Further applying our ADR-related hospitalization preventability rate of 36.8%, it is estimated that over 8,000 admissions would be preventable during this timeframe. Using the average length of stay from the Veteran population of 7.4 days and applying conservative FY04 acute inpatient admission daily costs ($1,880 per day; personal communication with VA Health Economics Resource Center), then ADRs cost the VA over $110 million from preventable hospitalizations.36
These healthcare costs may also be applicable to non-VA settings.
Previous individual studies in non-VA settings on ADR-related hospitalizations have reported higher rates up to 21%2,4,8,11
, including a subgroup analysis of 17 studies of elderly patients from a larger meta-analysis (68 observational studies total) that reported a 16.6% prevalence rate of hospitalizations caused by ADRs.3
However, our finding of 10% of hospitalizations caused by ADRs is consistent with more recently published research. A literature review examining the relationship between study factors and the prevalence of medication-related hospital admissions reported a prevalence of 10.8% for ADR-related admissions across 3 studies where patients were admitted to a ward for care of the elderly.11
Moreover, a systematic review of prospective observational studies assessing the prevalence of hospitalizations associated with ADRs reported a median ADR prevalence rate of 10.7% across 5 studies of elderly patients (age > 60 years).15
It is important to compare our finding of the association between polypharmacy and ADR-related hospitalizations with previous studies. After controlling for physical and psychiatric measures of comorbidity, polypharmacy was found to be significantly associated with ADR-related hospitalization in both bivariate and adjusted analysis. This finding is consistent with a multicenter cross-sectional study of older adults (mean age 70 years) in academic hospitals throughout Italy that assessed the prevalence of ADR-related hospital admission in older adults. In this previous study, Onder et al
found that the number of drugs was the most important risk factor associated with ADR-related hospitalization (OR 1.24, 95% CI 1.20–1.27 for each drug increase).4
However, it is essential to recognize that polypharmacy is influenced by the complexity of prescribing for older adults with multiple comorbid conditions who often have varying social situations and fluctuating goals and priorities of care.37
Because of this, clinicians, together with their patients, are often faced with making difficult trade-offs between achieving the intended benefits and avoiding the potential harms of medication use in complex older adults. Evidence of the need for such personalized clinical decision-making comes from a study that demonstrated interindividual variability in health priorities among older adults when faced with competing outcomes (i.e., cardiovascular event vs. fall injury vs. medication-related symptoms).38
Moreover, it has been shown that primary care clinicians who care for older adults with multiple medical conditions also display variability in their beliefs regarding the benefits and harms of following guideline-directed care.39
Thus, more research is needed to determine how to best equip clinicians with informed decision-making tools across conditions in older adults who often receive polypharmacy.
Several other interesting findings deserve mention. First, we only detected one case in which the use of a HEDIS high-risk drug was implicated in an ADR-related hospitalization (i.e., hyoscyamine associated with constipation).22
This low rate compares favorably to a nationally representative study of US older adults in 2004–2005 that assessed whether ED visits due to adverse drug events (ADEs) were due to potentially inappropriate medication use determined by Beers criteria.7,33
They also found low numbers of ED visits for ADEs due to potentially inappropriate medication use (i.e., 3.6% of all visits for ADEs). In addition, we found only one HEDIS drug-disease interaction (i.e., lorazepam/dementia) and three additional drug-disease interactions defined by other explicit criteria(i.e., diltiazem/heart failure; terazosin, nortriptyline/pre-syncope, postural hypotension; and hyoscyamine/constipation) that were associated with hospitalization.18
This low rate of drug-disease interactions could be due to the fact that many patients with drug-disease interactions were successfully treated in the ED setting and did not require subsequent hospital admission.
It is also important to note that the use of a few medication classes accounted for a majority of ADR-related hospitalizations. Specifically, we found that four medication classes (i.e., cardiovascular, central nervous system, antithrombotic, and endocrine) accounted for almost 80% of all the drugs implicated with ADRs (data not shown), which is consistent with prior research.25,40
In particular, among those ADR-related hospitalizations deemed to be preventable, narrow therapeutic range drugs (e.g., digoxin and theophylline) were implicated in 4 of 25 preventable admissions, and dysglycemia accounted for 5 of the 25 admissions (data not shown). Given that it is well-documented that older adults often take unnecessary medications and that narrow therapeutic range drugs may be improperly monitored, this suggests that future reductions in ADR-related hospitalizations may be possible and could reduce unnecessary health care expenditures.41–43
What is the clinical value of these research findings? This information should be useful to clinicians in that it highlights possible areas for intervention (e.g., reducing unnecessary polypharmacy) to decrease preventable ADR-related hospitalizations. This may be particularly important for patients who have been recently hospitalized and in whom the rate of taking one or more unnecessary drugs at discharge ranges from 44–59%.44,45
One approach to reduce polypharmacy in older adults that has been published recently is the Good Palliative-Geriatric Practice algorithm, which outlines a systematic approach to evaluate medication regimens with a focus on discontinuing certain drug therapies not immediately essential for life.46
While this approach has been shown to be successful in the nursing home setting as well as in community-dwelling older adults in Israel,46
future studies are needed to evaluate this algorithm in other populations of older adults such as those residing in VA nursing homes. Moreover, using a careful, stepwise approach to discontinuing medications while also incorporating the goals of care of the patient, their social situation, and the most current pharmacotherapy evidence base should aid in the successful discontinuation of potentially unnecessary medications.47
It is also important for clinicians to be aware of the clinically relevant potential for adverse drug withdrawal events (ADWEs) when discontinuing medications. While ADWEs are relatively rare, it is notable that they may present as either physiologic withdrawal reactions (e.g., beta blockers) or return of the underlying disease (e.g., worsening of hypertension upon stopping an anti-hypertensive).48
There are several potential limitations worth discussing. First, this study relied heavily on information from electronic health records to assess ADRs, and we may have underestimated problems if the information was missing, incorrectly entered, or not recorded in the chart. We did, however, use a reliable and valid ADR algorithm to assess ADR events that were related to the hospital admission.49
In addition, it is possible that unmeasured confounding could account for the association between polypharmacy and ADRs. To minimize this threat to validity, we controlled for a number of variables shown in other studies to be related to the outcome measure, including comorbidity. Moreover, the sample consisted mostly of older male Veterans living in the community and may not generalize to older females or non-Veteran populations. Indeed, this study found that older Veterans have multiple comorbid conditions and that 80% took ≥5 regularly scheduled prescription medications. However, it is important to note that our findings may be relevant to the one-third of community-dwelling older adults who reported using ≥5 regularly scheduled prescription medications in a nationally representative study.50