Introduction

Globally, the population is ageing. The number and proportion of oldest old people, who are defined as individuals aged 85 years and over, are increasing more rapidly than those in other age groups [1]. Oldest old people comprise about 2% of the population in the United States, while in 2030, this number is projected to increase to 2.5% and 3.7% in 2040 [2]. Oldest old patients often present with several comorbidities that require the use of multiple medications.

Inappropriate medication use involves prescribing potentially inappropriate medications (PIMs) and potential prescribing omissions (PPOs). In relation to PIMs, the risks of using particular medications outweigh their benefits, especially if there are safer and more effective options. With respect to PPOs, these are omitted medications that have a therapeutic use in the absence of contraindications [3, 4]. Inappropriate prescribing can cause an increase in adverse drug events (ADEs), which are situations leading to patient harm with respect to an increase in morbidity, mortality and health care costs [5, 6]. With an ageing population, inappropriate prescribing is an international health care problem [7].

Screening tools for inappropriate prescribing have been formulated to assist clinicians in providing efficient and appropriate prescribing practices, as well as in reducing the prevalence of preventable ADEs [4, 8]. The STOPP (Screening Tool of Older Persons’ potentially inappropriate Prescriptions) and START (Screening Tool to Alert doctors to the Right Treatment) criteria were developed in Ireland in 2003 to identify PIMs and PPOs, respectively. The STOPP and START were later validated by a Delphi consensus in 2006 [9]. Version Two of the criteria was released in 2014, comprising 80 STOPP and 32 START criteria [10]. The STOPP and START criteria have been applied in various studies and have been shown to be effective screening tools [11, 12]. Despite their potential benefits and utility, little work has been undertaken using these criteria amongst hospitalised oldest old patients.

The aims of this paper are: to examine the prevalence of inappropriate medication use in oldest old patients across transitions of care using the STOPP and START criteria; to determine which medication categories have a high prevalence of inappropriate medication use in oldest old patients; and to examine challenges associated with medication prescribing in oldest old patients from the perspectives of geriatricians.

Methods

Study design, setting and participants

A mixed-methods, sequential explanatory design was used, involving a retrospective clinical audit of electronic medical records and semi-structured interviews. The study was undertaken at an Australian public, teaching hospital with two sites, comprising 570 beds. This hospital served over 550,000 people living in the northern and western regions of the state of Victoria. At this hospital, doctors were taught about the importance of deprescribing in older patients, and also about the process of doing therapeutic reconciliation.

For the audit, the patient sample was randomly selected by an independent researcher who extracted data such as age, gender, admission dates, and primary diagnostic codes. The inclusion criteria were patients aged 85 years and over, who presented to the emergency department (ED) of the study hospital, and were admitted to the hospital between January 1st 2016 and December 31st 2016 with at least 1 medication on presentation. The exclusion criteria comprised patients who were admitted to the intensive care unit, died before discharge, were palliated prior to discharge, were discharged to other hospitals or private rehabilitation following ED presentation, had incomplete or unobtainable records, had duplicate admission dates in the electronic system, declined medications on ED presentation or were a tourist and therefore medication prescriptions were not reflective of the local situation. There were 1458 potentially eligible patients admitted to the emergency department across the 1-year audit period. Using a random numbers table, medical records were randomly sampled to determine the eligibility for inclusion and till the required sample was obtained.

Using purposive sampling by considering gender and years of working experience, interviews were conducted with geriatricians who worked in the geriatric evaluation and management or general medical units of the hospital.

Procedure

Demographic and medication information were obtained from the ED notes, admission notes, discharge summaries relating to patients’ discharge from acute and subacute settings, medication charts and progress notes. Pathology results and observation charts were also perused to obtain details of adverse clinical outcomes experienced by older patients. Data were obtained for four different time points during oldest old patients’ hospital stay. These time points were as follows: time point 1: medications on presentation to ED; time point 2: medications on admission to acute care; time point 3: medications on discharge from acute care to subacute care; and time point 4: medications on discharge from acute care directly to home/residential care or on discharge from subacute care to home/residential care. It is important to note that not all patients were discharged from acute care to subacute care.

Data were documented in an Excel spreadsheet (version 2010), which was directly imported into IBM SPSS (version 25). Medications prescribed were noted for each patient at each of the four time points. The STOPP/START criteria (version 2) were applied and the PIMs and PPOs were documented for each time point. For patients who had an identified PIM, vital sign observations, physical examinations, pathology tests and imaging results were examined to identify adverse clinical outcomes as possible clinical manifestations occurring during the patients’ stay. The Adverse Clinical Outcomes Tool was used to record potential adverse clinical outcomes which were developed from the STOPP criteria. Thus, in identifying adverse clinical outcomes, these related to medications that were prescribed and not withdrawn.

Data were collected on patient age, gender, use of walking aids, presence of documented allergy, presence of documented dementia, and presence of documented faecal or urinary incontinence. The Charlson comorbidity index (CCI) was manually calculated using an online calculator [13]. Transition points were defined as the number of changes in the location of a patient during hospitalisation. Change in discharge destination was calculated and defined as a change in the place of dwelling due to higher care requirements, such as moving from home to a residential care facility. Communication barriers were identified and defined by the presence of confusion from delirium or dementia, aphasia, dysphasia, dysarthria, hearing impairment, or an inability to communicate in English without an interpreter. The total length of hospital stay was calculated as the length of stay in acute care and subacute care. Functional Status was defined as patients requiring assistance with activities of daily living (ADL) in each of the domains (personal, domestic and community). Personal ADLs involved showering, dressing, toileting, eating, and grooming. Domestic ADLs involved completing household tasks like cooking and cleaning. Community ADLs comprised driving and shopping. If patients required assistance in any of these areas, they were allocated a score of 1 and a score of 0 if they were independent. The presence of a falls history was defined as having clear documentation of being a recurrent faller or having been referred to the falls clinic. Polypharmacy and excessive polypharmacy were defined arbitrarily as ≥ 5 and ≥ 10 medications.

In organising structured interviews with geriatricians, the questions posed included the barriers to safe prescribing in older patients, and challenges they experienced in ceasing or commencing medications.

Data analysis

Descriptive statistical analysis was performed, with categorical variables analysed using summary counts and percentages. For continuous variables with skewed distributions, medians and inter-quartile ranges were calculated. The following explanatory variables were examined at the univariate level to determine their effects on PIMs or PPOs: age, use of aids, documented allergy, documented dementia, history of falls, incontinence, personal ADLs, domestic ADLs, community ADLs, communication barrier, comorbid conditions, number of medications prescribed on admission, total length of stay, and age. Univariate associations with p values of ≤ 0.25 were included in the Poisson regression modelling. Poisson regression modelling was performed using the number of counts of PIMs or PPOs as the dependent (outcome) variable and explanatory variables. The level of significance utilised was alpha = 0.05.

Using a prevalence rate for PIMs of 54% found in the previous Australian study conducted by Manias et al. [14], and based on a desired 95% confidence interval width of 0.125, the sample size was calculated to be 245 patients. Therefore, data were required to be collected on about 245 patient medical records.

Audio-recorded interviews of geriatricians were transcribed verbatim. Thematic analysis was undertaken, which involved each author independently reading and rereading the data transcripts many times, identifying initial themes and subthemes, and summarising and synthesising the data.

Results

Demographic characteristics

Data on 249 randomly selected medical records were collected (Fig. 1). The demographic characteristics are summarised in Table 1. The median age of patients was 88.5 years, with the oldest patient being 103 years old. The most common diagnosis of admitted patients to acute care was congestive cardiac failure, followed by sepsis, pelvic or femur fracture and cerebral infarction. There were 90 (36.7%) patients who had acute kidney injury on admission.

Fig. 1
figure 1

Flow chart for access of sample

Table 1 Demographic characteristics (N = 249)

On presentation to ED, a total of 2,425 medications were prescribed to the entire study population with a median number of 10 medications (range 1–20). On admission to the acute unit, a total of 2254 medications were prescribed with a median number of 9 medications (range 2–10). On admission to the subacute unit, a total of 1,062 medications were prescribed to 108 patients admitted to the subacute unit with a median number of 10 medications (range 3–21). On discharge, a total of 2,391 medications were prescribed with a median of 9 medications (range 1–21). Polypharmacy increased slightly from 90.2% on presentation to ED, to 95.1% at discharge while excessive polypharmacy marginally dropped from 50.8% on presentation to ED, to 48.8% at discharge. All patients had changes to their medications during hospitalisation.

Characteristics of potentially inappropriate medications

Characteristics of potentially inappropriate medications are summarised in Table 2. There were 476 occurrences in the whole sample. The total numbers of PIMs at the four different time points were 195, 123, 51 and 107, respectively. The prevalence of having at least 1 PIM at the different time points was 51.0%, 37.3%, 40.4% and 36.9%, respectively. The most common PIMs were the use of proton pump inhibitors for uncomplicated peptic ulcer disease, followed by use of benzodiazepines, and medications prescribed without an evidence-based clinical indication. The common medications that were prescribed without any clear clinical indication were aspirin, frusemide, spironolactone and amitriptyline.

Table 2 Frequency of incidents across four snapshots of time for the Screening Tool of Older Persons’ potentially inappropriate Prescriptions (STOPP)

Characteristics of potential prescribing omissions

For the purpose of PPO characteristics, results reported here exclude the prevalence of vaccination omission as information about vaccinations tended not to be collected by the hospital. For information about PPOs relating to vaccination—omission of documentation about having a pneumococcal vaccine at least once after the age of 65, and omission of a seasonal trivalent influenza vaccine annually—refer to results in the supplementary material. The total numbers of PPOs at the four different time points were 158, 153, 66 and 125 respectively. There were 502 occurrences in the whole sample. The prevalence of having at least 1 PPO at the four different time points was 44.6%, 43.8%, 41.8% and 36.9%, respectively. The most common PPOs were omission of Vitamin D supplements in housebound patients or patients experiencing falls, followed by the omission of angiotensin converting enzyme (ACE) inhibitors in patients with systolic heart failure or ischaemic heart disease. The next most common PPOs involved the omission of beta-blockers with ischaemic heart disease (Table 3).

Table 3 Frequency of incidents across four snapshots of time for the Screening Tool of Alert doctors to the Right Treatment (START)

Adverse clinical outcomes

The total number of adverse clinical outcomes across the sample at the 4 different time points was 71, 42, 13 and 19 respectively. The most common adverse clinical outcomes were associated with prolonged use of a proton pump inhibitor, benzodiazepine, neuroleptic medication and non-steroidal anti-inflammatory drugs. As time progressed during the patients’ hospitalisation, the prevalence of adverse clinical outcomes decreased from admission to discharge (Table 4).

Table 4 Frequency of incidents across four snapshots of time for adverse clinical outcomes

Poisson regression results

Poisson regression analyses were undertaken to predict the incident count for PIMs and PPOs in relation to explanatory variables (Table 5). Use of mobility aids and a higher number of medications on admission were associated with a higher incident count for PIMs. A higher comorbidity number and a higher length of hospital stay were associated with an increased incident count for PPOs. A higher number of medications on admission was associated with a reduced incidence rate of PPOs.

Table 5 Poisson regression analysis for predicting the count of potentially inappropriate medications (PIMs) and potential prescribing omissions (PPOs)

Qualitative interview results

Interviews were undertaken with five female and four male geriatricians. Identified themes related to challenges in prescribing medications, challenges confronting geriatricians, and challenges faced by patients (see supplementary material).

Challenges in prescribing medications

Geriatricians referred to challenges in prescribing medications. Time constraints created difficulties in seeking the relevant medication history from specialists and general practitioners (GPs). Oldest old patients came to hospital from residential aged care where there were many GPs managing them rather than a regular GP, leading to loss of information regarding reasons for prescribing.

Since oldest old patients were often excluded in research studies, geriatricians believed a lack of evidence-based medication management existed in this population. Geriatricians had to extrapolate results from studies in younger patients. A dearth of data existed on the safety profile and efficacy of commonly used medications.

Difficulties arose from patients’ inability to afford certain medications. While geriatricians attempted to prescribe medications according to evidence, patients sometimes faced problems with accessing beneficial medications from a lack of government subsidisation.

The nature of adverse effects of certain medications, such as anticoagulants and insulin, meant that geriatricians had to carefully evaluate before commencing these medications. This caution sometimes led to delays in starting potentially beneficial medications.

Challenges in prescribing confronting geriatricians

Geriatricians referred to the differential knowledge about the oldest old sometimes contributed to inappropriate prescribing. While aged care trainee doctors usually had good understanding of safe prescribing, geriatricians perceived that doctors from other specialities and junior doctors may have had gaps in prescribing knowledge. These knowledge gaps could have led to potentially toxic levels of medications and patients taking inappropriate medications over the long-term.

Patients were managed by multiple specialists who sometimes considered their area of expertise rather than patients as a whole. A lack of coordinated care existed between specialists, resulting in patients being prescribed many medications with an increased risk of drug interactions. Patients were also reluctant to discontinue medications prescribed by specialists.

Geriatricians stated that while disease management guidelines were useful, strict observation of guidelines could lead to polypharmacy. Guidelines needed to be tailored to the oldest old, especially since guidelines were mostly based on younger populations. Strict adherence to guidelines could also lead to patients missing out on beneficial medications.

Geriatricians had difficulties in completing timely medication reconciliation on hospital admission. Inadequate information was sometimes available about the type of medications, dosages and indications.

All geriatricians believed that patients needed to be actively involved in decisions and goals of care. Active involvement meant that patients’ wishes could be considered in consultations.

As the hospital used a paper-based approach in information management, geriatricians stated there was a potential loss of information regarding medications. This situation increased the risk for medication errors and miscommunication, especially with patient movements between clinical settings.

Geriatricians believed that their medical colleagues were at times over-cautious about prescribing. These colleagues occasionally overestimated the risk of adverse effects, leading to patients missing out on potentially beneficial medications.

Challenges faced by patients

Geriatricians believed that patients were susceptible to adverse effects. Challenges existed in diversity in patterns and severity of patient illness.

Patients were reluctant to cease medications, especially those prescribed by specialists or general practitioners in private practice. This reluctance related to doctors previously informing patients that some medications had to be continued for life. Convincing patients otherwise, was challenging. Families of patients with dementia were also opposed to medication changes because of patients’ worsening cognition.

Geriatricians referred to multiple comorbidities leading to polypharmacy. Other challenges related to patient problems with medication adherence. Reasons for non-adherence related to adverse effects experienced by certain medications, and patients’ inability to swallow.

Discussion

The study provided comprehensive information about inappropriate prescribing in oldest old patients. The prevalence of having at least one PIM varied between 36.9% and 51.0% while the prevalence of having at least one PPO varied between 36.9% and 44.6% during patients’ hospitalisation. Use of mobility aids and an increasing number of medications on admission were associated with a higher incident count for PIMs. A higher comorbidity number and a longer length of hospital stay were associated with an increased incident count for PPOs. A higher number of medications on admission was associated with a reduced incidence rate of PPOs. Interviews with geriatricians identified many complex challenges in prescribing medications for the oldest old, which related to medication characteristics, difficulties confronting geriatricians, and concerns faced by patients.

Marked variability exists in determining PIMs in oldest old patients. Of the small number of studies undertaken in this population group, in community-dwelling people over 3-year cycles (2003, 2007 and 2011), Ble et al. [15] found between 34.9% and 41.1% of patients aged 85 years and over had at least one PIM using Beers (2012 version), while Wauters et al. [16] found 56.1% of community-dwelling people aged 80 years and over had at least one PIM using STOPP (version 2). Using a translated version of the Beers criteria (2012), Lai et al. [17] showed, of patients aged 80 years and over admitted to hospital, 27.1% had at least one PIM while San Jose et al. [18] found 63.3% of patients aged 85 years and over admitted to hospital had at least one PIM using the STOPP (version 2). The baseline level of at least one PIM (51.0%) was therefore similar to those of previous studies using the STOPP (version 2). In the current study, the prevalence was shown to decrease to 36.9% during hospitalisation up to hospital discharge. Previous studies have been cross-sectional in nature while the current study has been able to demonstrate downward trends in PIMs during patients’ hospitalisation. In geriatricians’ interviews, while there was recognition of the challenges involved in polypharmacy, they were cognisant of reducing inappropriate medication prescribing throughout patients’ hospitalisation.

The most common PIMs were the use of PPIs and benzodiazepines. These findings are consistent with previous studies [19,20,21,22,23,24,25,26]. PPIs have been shown to be one of the most commonly prescribed medications in older people [27]. The standard General Practice management for patients with PPIs includes conducting frequent medication optimisation to reduce the dosage or cease PPIs if asymptomatic [27]. Despite these guidelines, active de-prescribing of PPIs is not necessarily practised [28, 29]. The complications associated with long-term use of PPIs include Clostridium difficile infections, hypomagnesaemia, renal disease, dementia and pneumonia [30].

Benzodiazepine use was also shown to be a major type of PIM identified. In the study cohort, benzodiazepine use was found to possibly contribute to the presence of adverse clinical outcomes, such as documented prolonged sedation, confusion, impaired balance, falls, dyspnoea, confusion, and reduced oxygen saturation. Although benzodiazepine use appeared to impact on adverse clinical outcomes, the prescription of this medication group decreased during hospitalisation, which appeared to also be associated with a reduction in related adverse clinical outcomes. Past research has shown that oldest old patients who consumed benzodiazepines in hospital were more likely to have cognitive and psychomotor impairment compared with patients aged younger than 80 years [17]. Benzodiazepine use, especially for those prone to falls, was shown to be the most common cause of PIMs in the oldest old patients admitted to hospital [18].

The prevalence of having at least one PPO showed a downward trend for the four time points. If the omission of influenza and pneumococcal infections is included, the prevalence of at least one PPO ranged between 98.4% and 99.1%. There is a need for strategies and guidelines to improve the prescription of appropriate medications for the oldest old. In using the START (version 2), Wauters et al. [16] found a higher prevalence of 67.0% of at least one PPO in community-dwelling people aged 80 years and over, while San José et al. [18] found 53.6% at least one PPO using START (version 1). There are considerable differences between START version 1 and version 2. One of the most obvious is the presence of the two vaccinations relating to influenza and pneumococcal infections.

The most common PPO was vitamin D supplements, which is consistent with the findings of the few international studies in the oldest old population [19,20,21]. Studies have shown that vitamin D supplements reduce the risk of fractures and falls [31, 32]. Given the high incidence of falls-related ED presentations and the subsequent morbidity and mortality in this group, more vigilance is needed to ensure appropriate prescription of vitamin D supplements. Thereafter, common PPOs were ACE inhibitors in patients with ischaemic heart disease or systolic heart failure and beta-blockers in patients with ischaemic heart disease, respectively, which have previously been identified as common PPOs in the oldest old [19,20,21]. One study reported that age > 75 years was an independent risk factor for the under-prescription of ACE inhibitors and beta-blockers [33]. Given that congestive cardiac failure was the most common primary diagnosis in this study sample, more steps need to be taken to prescribe medications such as ACE inhibitors and beta-blockers that have a proven mortality benefit [33].

Poisson regression results demonstrated that an increase in comorbidities was associated with an increased incident count in PPOs and with no effect on PIMs. The current study involved calculation of comorbidity using the Charlson index, which may not be an accurate portrayal of how multimorbidity affects the oldest old. Development of a specific multimorbidity measurement for the oldest old would provide more appropriate measurement, which can be subsequently used to determine its associations with inappropriate prescribing.

Use of mobility aids was associated with an increased incident count of PIMs. Mobility aids comprised four-wheel frames, two-wheel frames, single pronged sticks and crutches. Use of mobility aids was significantly linked with a history of falls. Many of these patients had also used long-term benzodiazepines, which increase the prevalence of confusion and dizziness.

There was a complex picture with the number of medications prescribed on admission, with a significant positive association involving the incident counts of PIMs and a significant negative association involving the incident counts of PPOs. Many studies have previously reported the positive association between the number of medications prescribed and prevalence of PIMs [14, 21, 22, 24, 34]. With an increase in the number of medications, there could be a potential increase in the medication complexity regimen, leading to greater likelihood of adverse events and difficulties in medication reconciliation [35]. Greater attention on medications known to improve therapeutic benefits in the oldest old while at the same time being cognisant of the adverse effects that are likely with other medications, would enable a more balanced approach to prescribing.

As the length of hospital stay increased, patients also experienced an increased prevalence for PPOs. As indicated by the geriatrician interviews, there was a focus on reducing polypharmacy as much as possible during the patients’ stay. This situation was evident by the number of PIMs that progressively reduced during patients’ hospitalisation. However, at the same time, there was little focus on attempting to commence appropriate medications that are warranted.

There were limitations relating to this study. For the STOPP criteria, patients were assumed to have been on a PPI for > 8 weeks. If possible, the long-term prescription of the medication was verified by checking discharge summaries from other admissions or the medication list from the GP, to reduce the potential overestimation in the prevalence of PPI prescription. If no information was provided about the patients’ vaccination status, it was assumed that they did not have pneumococcal or influenza vaccination. Full adherence of patients with their medications was assumed. This was a single-centre study conducted in a tertiary teaching hospital with two sites, and the results cannot be generalised to other healthcare settings. The START and STOPP criteria do not capture the nuances of prescribing for patients on a case-by-case basis. While they are useful guides to safe prescribing, they are not substitutes for clinical reasoning.

This study demonstrated that oldest old patients are discharged with high rates of PIMs and PPOs. Greater attention should be placed on actively de-prescribing medications that are not beneficial while simultaneously commencing medications that would be advantageous. Audits of a larger scale are needed to identify barriers to safe prescribing and implement steps towards improving prescribing practices for this vulnerable population.