Medications and cognitive risk in Aboriginal primary care: a cross‐sectional study

Aboriginal and Torres Strait Islander people are ageing with high rates of comorbidity, yet little is known about suboptimal prescribing in this population.


Introduction
Aboriginal and Torres Strait Islander people experience disproportionately high rates of multimorbidity with chronic diseases such as cardiovascular disease, chronic kidney disease and dementia emerging at a younger age social determinants of health and the negative effects of racism and intergenerational trauma stemming from colonisation are thought to contribute to many of these inequitable health outcomes. 3o address this disparity, Aboriginal primary healthcare services are delivered at more than 300 locally operated Aboriginal Community Controlled Health Services (ACCHSs) across the country.These services aim to improve health outcomes by delivering culturally appropriate and holistic community and health services. 4A key function of these organisations is to ensure safe and appropriate medication use, as there are multiple challenges associated with maintaining safe prescribing in Aboriginal primary care.These challenges include ubiquitous issues such as suboptimal prescribing and poor communication between care team members and patients, in addition to specific cultural and patientrelated factors. 5otentially suboptimal prescribing for older people may involve polypharmacy (the concurrent use of ≥5 medications), the use of potentially inappropriate medications (PIMs) and underprescribing of indicated medications. 6Polypharmacy is associated with falls, functional disability, frailty and adverse outcomes for older men and women. 7,8Polypharmacy affects over a third of older people aged ≥70 years in the general Australian population, 9 and half of older communitydwelling Aboriginal and Torres Strait Islander adults. 10,113][14] Medications commonly considered to be PIMs include those with strong anticholinergic effects, such as tricyclic antidepressants and incontinence medications, that are associated with cognitive decline and functional impairment over time. 15The effects of less potent anticholinergics are thought to be cumulative, 16 and measures such as the anticholinergic cognitive burden (ACB) scale estimate this burden.Important limitations of standardised medication assessments in differentiating between appropriate and inappropriate prescribing in diverse clinical contexts are highlighted elsewhere. 17uboptimal prescribing, including polypharmacy, and ACB among Aboriginal and Torres Strait Islander peoples were investigated in two recent studies. 10,11In remote communities, one in five people were using PIMs, 10 while approximately half in regional/urban communities were taking anticholinergic medications. 11However, these studies were undertaken in only two regions with vast geographical separation, and the situation in other areas of Australia is unknown.Further work is required to clarify the prevalence of polypharmacy and suboptimal prescribing among Aboriginal and Torres Strait Islander peoples with respect to cognitive impairment and dementia risk.
This study aimed to define the prevalence of polypharmacy, PIMs and ACB among Aboriginal and Torres Strait Islander people aged ≥50 years attending ACCHSs in urban, rural and remote areas in Victoria and Western Australia.Associations between polypharmacy, PIMs, ACB and health characteristics, including multimorbidity and cognitive impairment, were assessed.

Study design
The Let's CHAT (Community Health Approaches To) Dementia in Aboriginal and Torres Strait Islander Communities project involves the co-design and implementation of a culturally responsive best practice model of dementia care in Aboriginal primary care. 18This study is a nested cross-sectional study utilising Let's CHAT Dementia baseline and additional medication data collected for a subset of ACCHSs (Fig. 1).
Baseline audits of electronic medical records at coresearching ACCHSs were completed for eligible patients.This captured information about patients' dementia risk profile, assessments and investigations relating to cognition (including whether the patients themselves, their family members and/or practice staff had raised concerns about their memory or thinking) and new diagnoses of dementia and cognitive impairment over 6 months from 1 September 2018 or 1 October 2018, until 28 February 2019 or 31 March 2019, dependent on the ACCHS.Additional audits, conducted retrospectively by author MH, captured prescription medications and comorbidities documented over the same period.MH undertook this work as part of her Doctor of Medicine/Master of Public Health studies.Medication data were matched to baseline data using patients' unique identification codes.The Charlson Comorbidity Index was generated for patients to describe comorbidities in relation to life expectancy, excluding a single criterion about HIV status because that information was unavailable. 19All data were deidentified and stored on a secure digital platform, REDCap.
Three ACCHSs, two from urban and rural Victoria and one from Western Australia (remote), were invited to participate in medication audits based on geographical setting, use of an electronic medical record (EMR) system, capacity to have audits performed remotely during COVID-19 public health restrictions and a strong relationship with the project.Records met eligibility criteria if patients were active (≥3 visits within 2 years), documented as Aboriginal or Torres Strait Islander, aged ≥50 years and without terminal illness at baseline (Fig. 1).We wanted to estimate the prevalence of polypharmacy (our primary factor of interest) with a margin of error of no more than 5%, with a 95% confidence interval (CI).Based on an assumed prevalence of 53% given previous work, 10 this required a sample size of 382 people.The baseline dataset included around 150 patients from each health service, which comfortably met the requirement.Audit lists for the baseline dataset were generated by systematically sampling patients from an alphabetical list, whereby the sample interval was defined by dividing the size of the population (N) by the size of the sample (n).For example, if N = 600 in one health service, the sample interval was 4, and every fourth person on the alphabetical list was audited.A total of 455 records were identified.

Medications and potentially suboptimal prescribing
The EMR interfaces used by the practices were MMEx, Communicare and Best Practice.Medication data were retrieved from the medications section of each EMR, which were maintained by practice staff, primarily including general practitioners (GPs) practice nurses and Aboriginal and /or Torres Strait Islander Health Workers and Health Practitioners (AHW/Ps).The patient name, dose, frequency, mode of administration and date of prescription were recorded for each medication prescribed.Medications were then coded according to the Anatomical Therapeutic Chemical (ATC) classification system. 20edications containing ≥2 pharmacological agents were recorded using a single ATC combination code.Medications prescribed for a period shorter than 4 weeks or pro re nata (as needed) were categorised as 'PRN', as distinguished from 'regular' medications taken over a longer period.PRN medications were included since some anticholinergic medications (such as opiates and sedatives) are often taken on a PRN basis.
Criteria to detect polypharmacy and PIMs and the ACB scale were used to define the prevalence of potentially suboptimal prescribing.Polypharmacy was defined as ≥5 PRN or regular medications prescribed within the audit period.The 2019 Beers Criteria were used to define PIMs, including drug-disease and drugdrug interactions. 12As a result of data limitations, criteria dependent on diagnostic markers such as renal function, and some disease states or syndromes were excluded.Taking a conservative approach, criteria Medication risk: Aboriginal primary care defining medications to be used with caution were also excluded.Criteria included in this study are listed in Table 3.
][23][24] Each medication received a score according to its anticholinergic effect, as per the ACB scale: medications with affinity for muscarinic receptors or weak anticholinergic effects were assigned a score of 1, and those with marked anticholinergic effects were assigned a score of 2 or 3 (Table S1).Medications with no anticholinergic effects were scored 0. Each patient was assigned an overall ACB score by summing the scores.

Statistical analysis
Data were analysed using the Stata statistical package, version 16 (StataCorp, College Station, TX, USA).After excluding patients who did not meet inclusion criteria, 420 records were retained for analysis (Fig. 1).We used the Kruskal-Wallis equality-of-populations rank test and Pearson's chi-squared test to assess differences between patients' demographic characteristics, cognitive risk factors and potentially suboptimal prescribing across ACCHSs.Binary logistic regression was used to explore factors associated with polypharmacy, one or more PIMs, having ACB score ≥ 1 and having ACB score ≥ 3.All variables which were significant in univariable analyses were entered into multivariable models, after which we removed non-significant covariates in a manual, backwards manner.We retained non-significant covariates in models where clinical judgement dictated this was appropriate, unless model fit was substantially worsened.We used Hosmer and Lemeshow's goodness-of-fit test and the Akaike and Bayesian information criterion to ensure appropriate model fit.Because data for cognition were missing for a small proportion of participants (2.9%), we handled the issue of missing data by

Results
The median age of patients was 58 years (interquartile range (IQR): 54-64), and the majority identified as Aboriginal (96.7%) and were female (53.3%).Across ACCHSs, patients had similar age, sex and attendance status.The prevalence of several documented chronic conditions and cognitive risk factors varied across health services (Table 1).
The majority of patients were prescribed at least one medication (70.0%), with 43.3% (95% CI = 38.7-48.1%)exposed to polypharmacy.The prevalence of polypharmacy differed significantly between locations, ranging from 32.1% at the rural ACCHS to 54.9% at the remote ACCHS.Heart disease, atrial fibrillation, hypertension, diabetes and depression were associated with greater odds of polypharmacy (Table 2).
Overall, 17.6% of patients (95% CI = 14.3-21.6%)were prescribed one or more PIMs, with lower odds for rural and remote ACCHSs compared with the urban site (Table 1).In adjusted models, the relative odds of PIMs were greater for females, those with kidney disease, urinary incontinence, depression, current heavy alcohol use and concerns for cognition (Table 2).Table 3 shows the most frequently used PIMs were psychotropics, including benzodiazepines (BZDs) (6.2%), tricyclic antidepressants (TCAs) and paroxetine or sedating antidepressants (4.5%).
Medication risk: Aboriginal primary care A small number of participants (n = 5) were taking a total of five over-the-counter and/or complementary medicine products that could not be classified.
In adjusted models, the relative odds of having ACB score ≥ 1 were lower in rural and remote ACCHSs compared to urban (Table 4).Chronic conditions remained significantly associated with higher odds of ACB score ≥ 1 after adjustment, and age was not significant.In adjusted models, significant ACB (ACB score ≥ 3) was associated with kidney disease, depression and attending an urban ACCHS (Table 4).
Figure S1 illustrates the frequency of patients with one or more medications, polypharmacy, PIMs and/or anticholinergic medications occurring concurrently.Of those prescribed one or more medications, 56 (19.0%) had combined polypharmacy, PIMS and anticholinergic medications.Of those with polypharmacy, 129 (70.9%) had PIMs or anticholinergic medications concurrently.

Discussion
This study provides important insight into the appropriateness of medication prescribing for older Aboriginal and Torres Strait Islander peoples.Across three ACCHSs, 43% of patients experienced polypharmacy.Other measures of potentially suboptimal prescribing were common; almost one in five had PIMs, and over one third were prescribed anticholinergic medications.
Overall, polypharmacy prevalence in this study was higher than that in the general older (70+ years) Australian population, 36% in 2017. 9This is consistent with high levels of multimorbidity, including cardiovascular disease and diabetes occurring at younger ages where multiple regular medications are recommended. 2,25Our findings concur with previous studies reporting PIMs for 20% of Aboriginal and Torres Strait Islander people in remote communities and regular anticholinergic medications among 47% of urban/regional Aboriginal communities.Significant ACB (score ≥ 3) was found in 12% in these Methadone 2 (0.5) 2 (0.5) 0 (0) Risperidone 2 (0.5) 2 (0.5) 0 (0) Alprazolam 1 (0.2) Carbamazepine 2 (0.5) 2 (0.5) 0 (0) Ranitidine 2 (0.5) 2 (0.5) 0 (0) Hyoscine hydrobromide (scopolamine) communities and 14% in the general older Australian population. 10,11,26olypharmacy and potentially inappropriate prescribing frequently accompanied one another in this study.Of those with polypharmacy, 71% had concurrent PIMs or anticholinergic medications.In many cases, polypharmacy in the context of multimorbidity may be appropriate to prevent cerebrovascular disease and other chronic conditions. 27However, the results of this study suggest that problematic polypharmacy is relatively prevalent in this population and could be a target for future interventions to improve prescribing and health outcomes.This is all the more compelling considering that PIMs dependent on diagnostic criteria were excluded in this study, indicating that rates are potentially higher than reported.
In this study, those with PIMs and anticholinergic medications were more likely to have concerns about their memory or thinking.Multiple conditions cause problems with memory and thinking, including dementia, depression, delirium and medication side effects, among others.For those with concerns about memory and thinking, medication review and rational deprescribing of PIMs and anticholinergic medications are important to reduce the potential effects of medication on cognition, and an area for further research. 28Larger studies have reported a positive association between PIMs or ACB and cognitive impairment. 16,29,30ur results confirm previous research reporting a strong association between ACB and depression among Aboriginal and Torres Strait Islander peoples and in the general population. 11,16Frequently detected anticholinergics in this study included amitriptyline, suggesting that tricyclic antidepressants may contribute to ACB for some older Aboriginal and Torres Strait Islander people. 11Depressive symptoms also remain highest among those treated, suggesting a need for more culturally safe psychological services such as healing services for Elders.First-line therapy for depression involves nonpharmacological intervention, often accompanied by antidepressants prescribed over a limited time frame. 31rescription of antidepressants over longer periods is relatively common in Australian general practice, which may highlight an opportunity for re-evaluation of treatment or deprescribing of antidepressants for some. 32Tricyclic antidepressants are more widely used among older groups compared to younger, 33 and this may be because of long-term use over many years and, potentially, reluctance to change to newer antidepressant agents as these become available.Prescription of tricyclic antidepressants, particularly at low dose, may also reflect off-label use such as for chronic pain, insomnia or urinary problems. 34her frequently detected PIMs involve the prescription of benzodiazepines, sometimes concurrently with opioid agents.6][37] However, high use of benzodiazepines, particularly among older and Aboriginal and Torres Strait Islander groups, suggests there is still some way to go to optimise prescribing for these groups. 38,39n this study, benzodiazepines were prescribed for 6% of older patients in ACCHSs, some of which may be prescribed inappropriately.Improving access to culturally appropriate psychological services and educational interventions to optimise prescribing of anticholinergic antidepressants and benzodiazepines 38 are likely to benefit older ACCHS patients.
Health system factors may account for observed differences between ACCHSs in this study.Availability of on-site specialist services, screening programs and documentation practices at individual ACCHSs differ, and these factors likely affect documentation of chronic conditions.Such factors may account for the high remote ACCHS polypharmacy prevalence (55%), which was similar to that found in our previous study of older Aboriginal people in remote Kimberley communities who access several different mainstream and Aboriginal health services (53%). 10Further, public funding for most medications dispensed at remote ACCHSs (under Section 100, National Health Act 1953) differs from arrangements at urban and rural ACCHSs (under Closing the Gap co-payment program); however, it is not clear whether this has influence over prescription practices in ACCHSs. 40,41urther research is required to understand differences with respect to health service and patient-related factors regarding suboptimal prescribing across urban, rural and remote areas and to understand adverse medication outcomes attributed to suboptimal prescribing.Trials of primary care interventions to address prescribing quality, such as on-site pharmacist review and culturally appropriate education for healthcare staff, may be beneficial. 42s a result of its cross-sectional nature, this study cannot draw conclusions about causal relationships.There may also be a bias towards people with more medical comorbidity in this study, owing to the exclusion of patients who had attended their health service less than three times over a 2-year period.Confounding because of medication indication or other factors may have influenced associations detected.Clinical information about specific medication indication, disease severity and geriatric syndromes could not be captured.Exact duration of medication therapy could not be determined, potentially leading to overestimation of potentially suboptimal prescribing.On the other hand, there were likely patients with undocumented cognitive impairment in the sample, which may have led to an underestimation of potentially suboptimal prescribing.Medications prescribed elsewhere (at another primary care service, specialist doctor or hospital), over-the-counter and traditional medicines were not accounted for, which could also have led to an underestimation of polypharmacy, PIM use and ACB.Potentially inappropriate omission of medications based on relevant clinical information was not reported, though this often contributes to suboptimal prescribing. 6,10dditionally, this is a study of prescribed medication rather than 'as taken' medication and concordance may be suboptimal.
For some patients, instances of potentially suboptimal prescribing identified may be appropriate.This study could not account for various medication-related factors that influence prescribing practices, such as patient preference or clinical judgement.PIM criteria and the ACB scale applied in this study were developed and validated for adults aged ≥65 years, 12,21 whereas this study included people aged ≥50 years.Although Aboriginal and Torres Strait Islander people aged ≥50 years are frequently considered to be of older age, these tools have not been validated in this younger population, and use may not be appropriate for all individuals.Although we tried to control for relevant factors in our regression models, some residual confounding may remain.This study and prevalence of potentially suboptimal prescribing should be interpreted with caution and with consideration of a patient's individual clinical circumstances and personal preferences regarding their healthcare.
A key strength of this study is that prescription medications and cognitive impairment risk factors were recorded directly from electronic patient records, which were maintained by practice staff, including GPs, nurses and Aboriginal Health Practitioners.Both regular and asneeded medications were captured to comprehensively determine PIMs and ACB, although inclusion of the latter might lead to an overestimation of potentially suboptimal prescribing.2][23][24] The sample size allowed for the assessment of important characteristics associated with potentially suboptimal prescribing.

Conclusion
In summary, polypharmacy affected almost half of older patients at urban, rural and remote ACCHSs in this study.Prevalence of PIMs and anticholinergic medications among patients was similar to estimates reported for Aboriginal and Torres Strait Islander communities elsewhere in Australia.Strong associations between polypharmacy, PIMs, ACBs and cognitive risk factors such as depression suggest patients with these conditions would benefit from regular medication review with the objective of minimising medication-related harm.The annual health check for Aboriginal and Torres Strait Islander people is an opportunity to review medications for all older people.Interventions to optimise prescribing practices in ACCHSs should also target those with risk factors for cognitive impairment and those expressing concerns about their memory or thinking.

Figure 1
Figure 1 Summary of co-researching Aboriginal Community Controlled Health Services (ACCHSs) and patient records included in the study.

Table 1
Characteristics of Aboriginal and Torres Strait Islander patients across three ACCHSs <0.001 †Concerns about memory or thinking problems.ACB, anticholinergic cognitive burden; ACCHS, Aboriginal Community Controlled Health Service; IQR, interquartile range; PIM, potentially inappropriate medication.P values are for Person's chi-square or Kruskal-Wallis equality-of-populations rank test.Bold indicates statistical significance (P < 0.05).

Table 2
Univariable and multivariable analysis of factors associated with polypharmacy and potentially inappropriate medications (PIMs) 0.130 ‡ †This column includes ORs only for age and other variables with statistically significant associations in a final multivariable model.Other variables with statistically significant associations in univariable analysis did not reach statistical significance in the initial multivariable modelling and were excluded from the final multivariable model.‡Variablenot considered in multivariable modelling as did not reach statistical significance in univariable analysis.§Concernsabout memory or thinking problems.CI, confidence interval; NS, not significant; OR, odds ratio.Bold = statistically significant result (P < 0.05).

Table 3
Potentially inappropriate medications criteria, with number and proportion detected in patients ACE, angiotensinogen-converting enzyme; CNS, central nervous system; COX, cyclooxygenase; NSAID, non-steroidal anti-inflammatory drug.Bold = most frequently detected PIMs criteria.

Table 4
Univariable and multivariable regression analysis of factors associated with ACB score NA ‡ †This column includes ORs only for age and other variables with statistically significant associations in a final multivariable model.Other variables with significant associations in univariable analysis did not reach statistical significance in the initial multivariable modelling and were excluded from the final multivariable model.‡Variablenot considered in multivariable modelling as it did not reach statistical significance in univariable analysis.§Concernsabout memory or thinking problems.

Table 5
Level 1 to Level 3 anticholinergic medications detected in patients