In this study of 808 elderly patients, contrary to our hypothesis, we observed few positive associations between individual geriatric conditions or cumulative geriatric burden and risk of ADRs. In contrast, ADRs were observed less often in patients with at least one dependency in ADLs than in those who were fully independent in their ADLs. Results were generally similar when our analyses were restricted to preventable and serious ADRs, although a history of recent falls was positively associated with future development of the latter outcome.
Few studies have closely evaluated the association between geriatric conditions and risk of ADRs, but other research in outpatient and inpatient settings is generally consistent with our findings. In a small methodologically limited study of hospitalized older adults, von Renteln-Kruse and colleagues described higher rates of ADRs in patients with urinary incontinence and poor nutrition (although only bivariate analyses were presented) (15
). In contrast, four higher quality studies identified no association between ADR risk and degree of mobility impairment (4
) or independence in ADLs (1
). Results of studies that evaluated cognitive impairment as a risk factor for ADRs have produced seemingly contradictory results, with two studies finding that worse cognitive function was associated with lower risk of ADRs (30
) and three others finding no effect or associations in the oppose direction (1
). These contradictory results may in part be explained by the complex interactions observed in these studies, with the association between cognitive impairment and ADR risk varying substantially by the type of medications used and recent introduction of new medications (30
). Thus, although subtlety is required in interpreting these findings, the overall gist of the existing literature demonstrates little evidence of a clear and convincingly positive impact of geriatric conditions on ADR risk.
One potential explanation for our results—and those of other studies evaluating risk factors for ADRs—is that older adults with greater degrees of geriatric burden might be less likely to report ADRs. If present, this bias would negatively confound the observed association between geriatric conditions and ADR risk.
For example, in explaining their finding of a negative association between cognitive impairment and ADR risk, Onder and colleagues postulated that ADRs may be more difficult to detect in cognitively impaired adults due to underreporting of symptoms in this group, reduced physician attention to patients with dementia, and given the frequently heavy burden of comorbidity in patients with dementia, greater difficulty distinguishing an ADR from an underlying disease process (31
We are unaware of any prior research that has directly evaluated differences in patient’s self-report or physician’s vigilance toward identifying ADRs in patients with versus without various geriatric conditions. Although we could not measure this directly, we did find that ADRs in elders with greater levels of geriatric burden were more likely to be accompanied by objective evidence. The relative paucity of ADRs without objective evidence in this group suggests that elders with higher geriatric burden may underreport subjective symptoms or otherwise present greater challenges in diagnosing ADRs that lack objective evidence to confirm the diagnosis. However, several features of our study reduce the risk that such potential bias in reporting had a substantial impact on our results. First, we found no association between geriatric burden and ADR rates after excluding ADRs that lacked objective confirmatory evidence. In addition, only 7% of ADRs were identified exclusively by self-report, with the remaining 93% identified wholly or in part through chart review. Although chart evidence of ADRs in part relies on the patient reporting their symptoms to clinicians, in other instances, clinical signs, laboratory tests, and directed physician questioning can be used to detect ADRs without patient prompting (33
). Finally, findings from the parent trial that outpatient GEM reduces the rate of serious ADRs compared with usual care demonstrates that differences in ADR risk between groups of patients can be identified using the methodologies employed in this study (18
Many physicians are reluctant to prescribe medications to such people out of fear that these patients are at disproportionately high risk of developing adverse events. Such instincts are appropriate because vulnerable elders often suffer from substantial burden of comorbid illnesses and are commonly prescribed multiple medications, which both can increase risk of ADRs (2
). However, our results suggest that specific geriatric conditions are not themselves positively associated with ADR risk. Thus, if a patient's comorbid illnesses and medication burden do not contraindicate adding additional medications, the presence of the geriatric conditions we studied should not necessarily dissuade the provider from prescribing potentially beneficial therapy.
We have no clear explanation for the finding that ADL dependency is associated with lower risk of ADRs. This observation may be spurious given its borderline significance and the relatively large number of predictors we studied, which increases the risk of false-positive results. In addition, the vulnerable nature of participants in the study may create a floor effect, whereby even those without ADL dependency had other characteristics that put them at risk. We are thus reluctant to conclude that ADL dependency is truly protective against ADRs, but we feel more confident in concluding that ADL dependency does not appear to substantially increase risk of these events.
Several limitations of our study merit discussion. First, as noted above, we cannot rule out the possibility that ADRs were detected differently in patients with greater or lesser degrees of geriatric burden, potentially biasing our results. Second, all participants had a minimum degree of vulnerability, operationalized by the study designers to encompass a wide range of potential problems ranging from specific geriatric syndromes to recent hospitalization or stroke (19
). Thus, our results should not be construed as comparing healthy versus vulnerable elders but as comparing the association between ADR risk and geriatric conditions among patients who all had a baseline degree of vulnerability. It is possible that more pronounced associations would be observed in comparing healthy versus vulnerable elders. Third, our major predictors of interest were largely assessed by self-report, which may differ from physician diagnoses of these features. However, most of these features are typically diagnosed by physicians based on patient reports (eg, incontinence, falls). Fourth, our measure of cumulative geriatric burden has not been independently validated, and the scoring system (1 point for each problem) fails to account for the severity of each condition or its expected contribution to ADR risk. However, a similar accumulation-of-deficits approach (albeit more comprehensive) has been validated extensively by Rockwood and colleagues and may represent a fair approximation of the degree of geriatric burden that a clinician may perceive in daily office-based encounters with patients (27
). Finally, because study patients were being discharged from the hospital at study baseline, it is uncertain how our results generalize to ambulatory participants without recent hospitalization.
In summary, we observed no positive association between a variety of geriatric conditions and risk of developing an ADR. Patients with such features often have a substantial degree of comorbid illnesses and multiple medication use, and it is prudent to exercise special caution in prescribing to such patients. Nevertheless, our findings suggest that prescribers need not be overly timid in prescribing medications that are truly appropriate to vulnerable older patients on account of geriatric conditions they might have.