Characterising polypharmacy in the very old: Findings from the Newcastle 85+ Study

Background Polypharmacy is potentially harmful and under-researched amongst the fastest growing subpopulation, the very old (aged ≥85). We aimed to characterise polypharmacy using data from the Newcastle 85+ Study—a prospective cohort of people born in 1921 who turned 85 in 2006 (n = 845). Methods The prevalence of polypharmacy at baseline (mean age 85.5) was examined using cut-points of 0, 1, 2–4, 5–9 and ≥10 medicines—so-called ‘no polypharmacy’, ‘monotherapy’, ‘minor polypharmacy’, ‘polypharmacy’ and ‘hyperpolypharmacy.’ Cross-tabulations and upset plots identified the most frequently prescribed medicines and medication combinations within these categories. Mixed-effects models assessed whether gender and socioeconomic position were associated with prescribing changes over time (mean age 85.5–90.5). Participant characteristics were examined through descriptive statistics. Results Complex multimorbidity (44.4%, 344/775) was widespread but hyperpolypharmacy was not (16.0%, 135/845). The median medication count was six (interquartile range 4–8). Preventative medicines were common to all polypharmacy categories, and prescribing regimens were diverse. Nitrates and oral anticoagulants were more frequently prescribed for men, whereas bisphosphonates, non-opioid analgesics and antidepressants were more common in women. Cardiovascular medicines, including loop diuretics, tended to be more frequently prescribed for socioeconomically disadvantaged people (<25th centile Index of Multiple Deprivation (IMD)), despite no difference in the prevalence of cardiovascular disease (p = 0.56) and diabetes (p = 0.92) by IMD. Conclusion Considering their complex medical conditions, prescribing is relatively conservative amongst 85-year-olds living in North East England. Prescribing shows significant gender and selected socioeconomic differences. More support for managing preventative medicines, of uncertain benefit, might be helpful in this population.


Introduction
Polypharmacy is potentially harmful and under-researched amongst the fastest growing section of society, the very old (aged �85) [1]. Few studies have explored polypharmacy in this population [2][3][4][5][6][7][8][9][10], despite medication sensitivity often increasing with the late-life problems of physiological decline, multimorbidity and frailty. Even in the younger old, polypharmacy is most often measured through reductionist categorical definitions that offer no indication of the complexities or appropriateness of prescribing [11].
This study thus characterised polypharmacy in the very old using data from the Newcastle 85+ Study-examining its prevalence, most common medications and medication combinations, as well as prescribing differences across gender and socioeconomic status.

Recruitment and study protocol
Health and medication data were extracted from the Newcastle 85+ Study: a population-based longitudinal study of very old adults living in North East England who were born in 1921, aged 85 in 2006 and permanently registered with one of 53 participating general practices in Newcastle or North Tyneside [17,30,31]. When the study began (2006), the sociodemographic composition of the cohort was representative of England and Wales, but participants with end-stage terminal disease were excluded (n = 11) [17]. Data were collected in two ways: multidimensional health assessments and general practice medical records [17,[30][31][32]. Full details of the questions asked are available at: http://research.ncl.ac.uk/85plus/.

Ethics approval
The Newcastle 85+ Study was approved by the Newcastle and North Tyneside Local Research Committee One (Ref: 06/Q0905/2), and informed written consent was obtained from all participants.

Disease and disability status
Data on fifteen disease groups were analysed to provide a complete picture of multimorbidity in the baseline cohort. Some were analysed individually (e.g. hypertension) and others as composite variables (e.g. a record of any arthritic disease was taken as a diagnosis of arthritis) [16]. Full details of disease status construction and composition can be found in S2 Table. A disability score (0/1-6/7-12/13-17) was assigned from the total number of activities of daily living, instrumental activities of daily living and mobility items performed with any self-reported difficulty. The higher score, the greater the level of disability [39, 40].

Statistical methods
Socioeconomic position was measured at the small area level via the Index of Multiple Deprivation (IMD)-a weighted construct of income, employment, education, health, crime and the living environment. IMD was cut at the <25th, 25th-75th and >75th centiles in line with previous research [41,42]. Univariate ordinal logistic regression assessed whether polypharmacy differed by gender and socioeconomic position at baseline. Mixed-effects models then assessed whether these variables were associated with prescribing changes over time (mean age 85.5-90.5 years, baseline to 60 months post-baseline). Only medicines prescribed at a frequency of �40 were included, as those with very small sample sizes would have insufficient power for longitudinal analysis, and any gender or socioeconomic differences they may have would not be clinically relevant. Basic health and sociodemographic characteristics were analysed separately by sex using the chi-squared test for descriptive purposes. Gender and socioeconomic differences in disease prevalence were also examined to inform our understanding for prescribing (S3 and S4 Tables). All analyses were performed using R 3.5.0.
In the month prior to baseline health assessment, polypharmacy (49.6%, 419/845) was more common than minor polypharmacy (24.6%, 208/845) and hyperpolypharmacy (16.0%, 135/845) ( Table 2). The median medication count was six (IQR 4-8). Table 3 presents the twenty most frequently prescribed medicines amongst people with minor, polypharmacy and hyperpolypharmacy at baseline. Cardiovascular medications (including statins, aspirin and beta-blockers), non-opioid analgesics and proton-pump inhibitors were among the ten most frequently prescribed medicines in all three categories. Subjects of prescribing indicators including anxiolytics, antipsychotics and urinary antispasmodics were sparse. Antidepressant and anti-dementia medications were also uncommon.

Polypharmacy combinations
At baseline, polypharmacy regimens were highly individualised but often included cardiovascular and analgesic medications. Aspirin and statins were most commonly co-prescribed amongst people with minor polypharmacy and polypharmacy. Non-opioid analgesics, statins, aspirin and loop diuretics were most common in hyperpolypharmacy.

Gender differences in medication prescription
At baseline, and across the full polypharmacy spectrum, women were 37% (95% CI: 5%-77%) more likely to belong to the monotherapy plus all other categories of polypharmacy vs none than men. Longitudinally, gender differences in individual medications were also observed in mixed-effects models adjusted for deprivation.  Table).

Socioeconomic differences in medication prescription
At baseline, compared to the referent 25th-75th centile, there was a non-significant difference in polypharmacy between those most affluent (>75th centile IMD)  Table), despite no difference in the prevalence of cardiovascular disease (p = 0.56) and diabetes (p = 0.92) across each IMD centile (S4 Table).

Summary
Polypharmacy is common and highly individualised in the Newcastle 85+ cohort (n = 845), driven by both long-term preventative and symptomatic medicines, as well as significant gender and subtle socioeconomic differences. That said, considering the clinical complexity of the sample in terms of widespread multimorbidity and geriatric syndromes, prescribing for this population was relatively conservative.

Comparison with existing literature
Prescribing is said to increase with age [43, 44] and polypharmacy was indeed common in the Newcastle 85+ cohort. Almost half of participants took five to nine medicines (49.6%, 419) with a median of six-a slightly higher number than that of other comparable old age cohorts [45,46]. This may reflect international differences in multimorbidity and prescribing patterns [37,47]. 'Six' may seem a high number of medicines to take but given the many drivers for polypharmacy in our sample, including complex multimorbidity, chronic pain, disease-specific clinical guidelines and health service use [37, 48-51], we could argue that prescribing was relatively contained. With deprescribing in its infancy at study inception (2006)  The contribution of preventative and symptomatic treatments to polypharmacy is consistent with previous research [5,57] and the challenges of prescribing to people in advanced age. 'Preventative' statins could, for example, partly reflect the difficulty in recognising when the end-of-life is approaching, limited time-to-benefit information or the aforementioned dearth of evidence for deprescribing [58, 59], and 'symptomatic' non-opioid analgesics and laxatives, the prevalence of pain and reduced mobility in later life [40,48]. Some of the most frequently prescribed medicines are also 'high risk.' ACE inhibitors, beta-blockers and loop diuretics have all been linked to unplanned hospital admissions in younger populations, for example [60]. That said, encouraging prescribing trends were also observed. Subjects of prescribing indicators including anxiolytics, urinary antispasmodics and antipsychotics were infrequently prescribed.
Reasons for the socioeconomic differences in prescribing are less clear (S6 Table). The greater prescription of tricyclic (and related) antidepressants (OR: 2.98, 95% CI: 1.15-7.75) to those more affluent (>75th centile IMD) may reflect inequitable pain management. Topical NSAIDs (OR: 0.37, 95% CI 0.17-0.84) were also less common amongst disadvantaged people (<25 th centile IMD). Reasons for this are unproven but could include socioeconomic advantage lessening prescribers' perception of risk, or differences in symptom reporting and access to specialist care. Greater prescription of selective beta-2 agonists to the >75 th group might similarly suggest that their overuse is monitored more closely amongst those living in socioeconomic disadvantage, who may be perceived to be at greater risk of associated exacerbation or death [61]. The tendency for disadvantaged people (<25th centile IMD) to use more SSRIs (OR: 1.44, 95% CI: 0. 65-3.19) may reflect poorer social support, coping mechanisms and greater stress [62], or an increased vulnerability to prolonged or recurrent late-life depression [63]. Though no difference in the prevalence of cardiovascular disease (p = 0.56) and diabetes (p = 0.92) across each IMD centile, socioeconomically disadvantaged people also tended to be prescribed more of the cardiovascular medicines including ACE inhibitors and calcium-channel blockers, statins, beta-blockers and oral anticoagulants, whose risks can outweigh the benefits in later life [60,[64][65][66]. Thus, limited deprescribing in the very old may particularly impact deprived individuals.
Finally, polypharmacy combinations were diverse, reflecting the heterogeneity of multimorbidity [37] and propensity for other determinants of medication use,-such as renal function, frailty and medication tolerability -, to vary from person to person in very late life. Indeed, 'one size does not fit all' when it comes to polypharmacy [67].

Strengths and limitations
Our study has several strengths. We characterised polypharmacy in a cohort of 85+ year olds and examined gender and socioeconomic differences in late-life prescribing, thus adding to the scarce literature in these areas. Our pragmatic definition for polypharmacy helped to avoid information loss, and visualising medication combinations helped to communicate prescribing complexity. Finally, we analysed medication and disease data from general practice records as opposed to the less reliable method of self-report [17].
We were unable to judge whether prescribing was potentially inappropriate as we had no information on dosages, durations or current indications; lacked many drug names by chemical substance and aggregate (rather than individual level data) was used. Some findings might also reflect former prescribing practices as the study started in 2006. For example, aspirin was prescribed more often than clopidogrel [68,69]. The extent to which non-adherence influenced prescribing is also unclear. We could not measure prescribing outside of data collection points, or assess disease severity, so the mixed-effects models may not reflect the true extent to which gender and socioeconomic position influence late-life prescribing [70]. There may have also been a survivor effect longitudinally, but adjustment for mortality made no difference in a sensitivity analysis. As is customary in prescribing research, over-the-counter medicines were excluded from our polypharmacy definition to make our results directly applicable to primary care practice, but we acknowledge that this can lead to an under-estimation of medication use.

Implications for practice and conclusion
Considering their complex clinical conditions, prescribing is relatively conservative amongst 85-year olds living in North East England. That said, more support for managing preventative medicines, of uncertain benefit and harm, might be helpful in this population. Pharmacists in primary care teams may be an ideal way to provide this. In highlighting prescribing heterogeneity, our findings also underscore the need for personalised healthcare in the very old. This could help to address the significant differences in prescribing prevalence for gender, and the subtle but potentially important differences for socioeconomic position.
Supporting information S1