Statement of principal findings
This current study found that almost 15 % of people aged ≥65 years in the Republic of Ireland received at least one PIP, according to a subset of 26 STOPP criteria. PIP was more likely in older adults (≥75 years) when adjusting for gender and polypharmacy. The overall prevalence of potential prescribing omissions was 30 % considering the ten included START criteria. Omissions were significantly more likely in males when compared to females when adjusting for age and polypharmacy. There was a significant association between PIP/PPOs and polypharmacy when adjusting for age, gender and multimorbidity.
Results in the context of current literature
This study presents a snapshot of the prevalence of PIP in a nationally representative population of people aged ≥65 years using data from the first wave of TILDA. Previous research is limited by having focused on specific groups in particular clinical settings as well as having measured PIP using Beers’ criteria derived in the US [13–16]. A limited number of studies of PIP in the general population of older people exist. In a previous Irish study using data from the HSE-PCRS, an overall prevalence of 36 % in patients ≥70 years was reported using a subset of 30 STOPP criteria [6]. Similar results were reported in a study of the Enhanced Prescribing Database (EPD) in Northern Ireland where the overall prevalence of PIP was 34 % [5]. However, the differences between the TILDA database and these prescription reimbursement databases make comparisons more difficult. The lack of clinical information relating to diagnoses (e.g., hypertension, CVA) in the EPD and PRCS databases and the lack of information on drug duration, dose or frequency of administration of prescriptions in the TILDA dataset mean that only a limited number of the STOPP criteria applied in the three studies are similar. In the EPD and PCRS databases, the most prevalent PIP drugs were proton pump inhibitors (PPIs) at maximum therapeutic dosage for >8 weeks, followed by NSAIDs for >3 months and long-acting benzodiazepines for >1 month [5, 6]. These instances of PIP were not captured in TILDA. Similarly, the use of a NSAID with moderate-severe hypertension was the most prevalent PIP in TILDA, and this could not be assessed in the prescription databases due to a lack of clinical information. Of the 19 STOPP criteria that were the same in PCRS and TILDA, the prevalence of PIPs are broadly similar, but low in both datasets (<2.5 %).
The significant association between PIP and polypharmacy is consistent with findings reported in previous studies [5, 6, 17], even when age, gender and multimorbidity is accounted for [18]. While it is well recognised that the use of multiple medications in the older adults is associated with increased risk of adverse events such as adverse drug events in patients taking NSAIDs (e.g., gastrointestinal haemorrhage) and increased incidence of falls and cognitive impairment in those taking sedative-hypnotic drugs, these medications are routinely among the most common PIPs [3, 6, 19], as was the case in our study. We also found an association between advancing age and PIP after adjusting for polypharmacy, similar to a previous primary care based study [20]. We reported no association between PIP and gender after adjustments for age. These findings are contrary to previous studies reporting that women are more likely to receive a PIP when compared with men [6, 17, 21].
Our overall findings relating to the prevalence of PPOs are similar to previous studies that have used the START criteria in a primary care setting [20]. Other studies have reported higher prevalence rates of omissions among an older hospital inpatient population [22, 23]. However, older patients admitted to hospital are sicker and frailer than those in the community so the findings need to be considered in the context of the population of interest. The cardiovascular system accounted for the majority of PPOs in keeping with previous reports [20]. Our study demonstrated that polypharmacy is associated with the under-prescribing of indicated medicines even when adjusting for age, gender and co-morbid conditions. Similar findings have been reported in other studies, particularly those with co-morbid conditions [24, 25]. However, these findings warrant further investigation.
We also noted that over 70 % of study participants who reported moderate-severe depressive symptoms were not prescribed an antidepressant drug. Depressive symptoms were assessed in TILDA using the Center for Epidemiologic Studies Depression Scale (CES-D) [26]. We used a cut off score of ≥27 points on the CES-D scale as a proxy for diagnosis of ‘moderate-severe’ depressive symptoms. This interpretation is the most conservative estimate reported in the literature [27]; therefore, our findings need to be considered in the context of these short-comings. It may be that other management approaches were adopted in the treatment of depressive symptoms such as cognitive behavioural therapy [28].
Clinical and policy implications
The term ‘potentially’ is used as there is often limited evidence to support the inclusion of particular medications in lists of drugs to avoid in particular populations. However, their use or lack thereof may be indicated and appropriate in certain circumstances considering the clinical presentation of the patient and the balance of the risk and benefits [5]. Clinical vigilance and quality efforts should place particular focus on medication appropriateness, both with respect to under-prescribing and over-prescribing. Reducing PIP in older people will require implementation of more robust methods of medication reviews to routinely assess drug effectiveness, dosage, duration, interactions and adverse effects [29, 30].
The updated NICE guidelines (2013) on falls assessment and prevention in older people highlight the importance of a medication review in older people at risk of falls who present to healthcare professionals. There is a lack of experimental research on the use of the STOPP criteria as a tool to reduce the incidence of adverse drug events (ADEs) such as falls in older people. However, a recent randomised controlled trial demonstrated that the application of the STOPP criteria significantly improved medication appropriateness when compared with the usual pharmaceutical care in 400 hospital inpatients [3]. The study was not powered to detect a reduction in secondary outcomes such as ADEs but further research is warranted to examine the role of medicines reviews based on the STOPP criteria as an intervention tool to attenuate ADE incidence. In addition, reducing the number of drugs used by older people may serve to reduce the risk of falls and the associated direct and indirect costs [29].
Strengths and weaknesses of the study
The TILDA database contains data on morbidities (>99.5 % complete) that enable the assessment of PIP in the context of a given diagnosis (e.g., hypertension, cardiovascular conditions). These data also facilitate the assessment and measurement of errors of omission, i.e., the lack of use of appropriate drugs when needed. However, it was only possible to apply a subset of the STOPP/START criteria to the database, as described in Table 1, due to a lack of information on drug strength, dose and duration of prescriptions. This may have led to an underestimate of the true prevalence of PIP/PPO among older people in TILDA. In addition, the validity of the original criteria is diminished by only applying a subset and the comparability to other population based studies is limited due to the differences in criteria applied.
The cost of PIP was not considered in the present study, but a previous study estimated the cost of PIP to be in excess of €45 million using a subset of 30 STOPP criteria [6]. Further research is planned to link the TILDA dataset to the national prescribing database (HSE-PCRS) to facilitate the application of 19 additional STOPP criteria and to estimate the cost of PIP using these linked datasets. In addition, there is a need to compare these findings with other European population based studies. The association between explicit process measures of PIP and prescribing omissions, and health outcomes (morbidity and mortality) and healthcare utilisation also requires further investigation to facilitate the development and testing of appropriate interventions to reduce PIP and PPOs in older adults. In the Irish context, this can be completed using future waves of the TILDA data.