Under and overtaking medications.

Two prospective cohort studies reported under and overtaking medications were common [18, 20]. In one study, 17.4% of participants reported at least one incident of medication non-adherence during the 12-month prospective follow up period [20]. These comprised one incident (14.7%), two incidents (2.4%) and three incidents (0.3%) [20]. The other study, examined all domains (excluding abstract reasoning) in participants with heart failure who screened positive for CI [18]. Through pill counts of all prescribed medication, this found that MCI and severe CI persons were 30% and 27% non-adherent (pills were not taken, over taken or a combination of both) respectively within a 30 day period[18].

Omission of a single dose.

Omission of a single dose was specified in seven studies [14, 18, 21, 25, 26, 32, 34]. When reported, adherence frequencies fell to 42%-71%. The majority used pill counts to determine medication adherence (n = 4) whilst the remaining used self-reports (n = 1), survey (n = 1) and a medication-monitoring device (n = 1). Most studies did not specify cognitive domains and instead used a global description of CI (n = 4) [25, 26, 32, 34].

Three studies specified cognitive domains including, attentional [18], speed of information processing [18], visuospatial and constructive skills [18], new learning and memory [14, 18, 21], receptive and expressive language [18], praxis [18] and executive functioning [14, 18, 21] (Table 3).

Deviation from prescribed time/dose intervals.

Deviation from a prescribed time and deviation from dose intervals was documented in one case control study using a global description of CI [27]. The control (non-CI) group performed better in tasks of timing and dosing compared to participants with AD. Participants with AD also over-estimated their ability to time and dose medications correctly (92% self-predicted versus 78% actual, respectively) whereas controls were accurate at both tasks.

Attentional and speed of information processing.

Attention [18, 20, 30] and speed of information processing [18] were not statistically significant predictors of medication adherence and management. These were prospective [18, 20] and retrospective cohort studies [30] using the DRS [20, 30] and the Saint Louis University Mental Status (SLUMS) examination to ascertain cognitive impairment [18].

Visuospatial and constructional skills and praxis.

Visuospatial [18, 20, 30] or praxis [18] test scores were not statistically associated with medication adherence however, one study found participants often knew medications by colour and shape rather than name or indication [18].

New learning and memory and executive functioning in medication adherence.

Most studies focused on memory (n = 5) and executive function (n = 5) [14, 18, 20, 21, 30] when assessing medication adherence.

Three studies reported better memory as a significant predictor of adherence [18, 20, 30] while two studies did not [14, 21]. CI participants were used in two of the studies reporting significance [18, 20]. The remaining studies did not focus on CI populations. Two studies used the same methods to ascertain memory scores [20, 30] finding better performance on the DRS memory subscale a predictor of better medication adherence [20] and poor performance a predictor of missed medical appointments [30]. Furthermore, CI was the sole factor associated with poor medication adherence.

CI participants who completed further neuropsychological testing in one study demonstrated deficits in immediate memory and delayed verbal memory [18]. Intact executive functioning was protective for adherence in two studies [14, 21] and was determined through different forms of testing: Wisconsin Card Sorting Test (WCST); WMS III letter- number sequence, mental control and digit span backward [14] and Trailmaking B Test [21]. These recruited participants showing signs of CI (e.g. MMSE scores <25) [21] or deficits in domains indicative of impairment [14]. The remaining studies, using CI population samples, reported poorer executive performance increased the likelihood of non-adherence (n = 1) [20] or was not statistically significant (n = 2) [18, 30]. Interestingly, two used the DRS initiation/preservation subscale [20, 30].

Abstract reasoning.

One prospective cohort study reported better performance on the DRS conceptualization subscale increased the likelihood of non-adherence [20].

Receptive and expressive language.

There was not a significant association between language domain tested and medication adherence in one prospective cohort study [18].

Individual factors.

Eleven studies documented individual factors associated with medication adherence and non-adherence in cognitively intact and cognitively impaired individuals [14, 18, 20, 21, 26, 27, 30, 31, 33–35] (Table 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Significant Factors for Medication Adherence and Non-Adherence in Cognitively Intact and Cognitively Impaired Individuals.

https://doi.org/10.1371/journal.pone.0170651.t004

Individual hazardous factors reported: African American and Hispanic ethnicity [31]; forgetting [21, 26]; lower scores in cognitive domain test indicative of deficits [14, 20, 21, 30]; dementia/CI [18, 27, 33, 35]; depression [30]; inadequate self-care confidence [33]; lower level of education, concern about taking prescribed drugs and intentional noncompliance [35].

Of note, global dementia/CI scores using MMSE [14, 26] were not associated with adherence in two prospective cohort studies and when reported, where total MMSE score ranged from 14–23 (M = 20.72) [26]. Presence of dementia was associated with not knowing the name and purpose of medications, having others assisting and preparing medications and an adult child assisting with medications [26].

Dyad/Carer.

The likelihood of participants with CI having a caregiver to assist with medications was reported in six studies [25–29, 34]. Presence of caregiver established self-administration dependence and low capacity of self-medication [26, 28] as well as improved adherence [27, 34]. Informants were able to accurately predict the care recipient’s adherence rates and performance on medication management tasks in one case-control study [27].

Participants with CI were more likely to have someone prepare medication [26, 27, 29]. Hazardous factors affecting medication adherence for CI persons included caregiver distress, not living with a caregiver [34], absence of assistance (e.g. reminders, pill box check and setting-up lists) [27] and spouses as primary caregivers who were also cognitively impaired [25].

Medication.

Three studies reported medication factors in relation to non-adherence [20, 21, 35]. Taking fewer drugs was associated with improved adherence in one study using self-reports [21] and CI persons taking four or more daily medications had a 2.5 fold increase in non-adherence compared to those taking less than four medications according to independent rater-reviews [20]. Conversely, a cross-sectional study reported a non significant association between medication adherence and: the number of drugs; or frequency of drug administration; or with/without one dose package; or use of medication calendar or written drug information using pill counts [35].

Medication aids.

A memory assistive device was used more often by CI participants (32/35) than non-CI participants (132/143) [26]. The lack of such device was associated with non-adherence [35]. Better scores on memory subscales were also associated with participants setting up their own medication schedules [21]. Additionally, environmental cues associated with the repetitive task of taking medication were reported to have contributed to increased adherence [14].

Healthcare system factors.

Health system factors were reported in three studies [21, 30, 35]. There was a negative correlation between physician rating and patient’s ratings of medication treatment adherence [30]. Furthermore, poor client-physician relationships was an independent predictor of poor adherence (defined as <80%) [35] whilst prescription insurance was positively associated with medication adherence on univariate analysis only [21].

Cognitive domains.

Specific cognitive domains receiving the most attention were memory and executive functioning. Studies with CI populations found intact memory was a significant predictor of medication adherence [18, 20]. Any associations with executive functioning remain unclear due to discordant study results. Persons with CI may not be able to understand, retain or follow instructions, implying that interventions focusing on traditional models of patient education may fall short in this population [18].

The impacts of a better level of executive performance in persons with CI on medication adherence were discordant, one reported improvement [20] and another did not [18].

In contrast, studies with populations considered cognitively intact reported better executive function to be a significant factor for adherence [14, 21]. These discordant results may be due to differing methods utilized to test this domain. Executive functions correlate with instrumental activities of daily living requiring goal directed activities [37] suggesting executive abilities involving mental flexibility, including implementing, planning and maintaining intentions, may be important for medication adherence [20].

Interestingly, better performance on the subscale of abstract reasoning increased the likelihood of medication non-adherence in a sample of CI older adults. This may be explained by intentional non-adherence due to individuals’ concerns about the medication [35] as this requires the ability to abstract and form fundamental connections between medication and possible side effects–the ability measured by the selected subscale [20]. There was not a significant association between language domain tested and medication adherence [18]. However, subtle impairments in verbal memory amongst cognitively intact women taking oral anti-estrogen therapy was a potential predictor of non-adherence [38]. Such findings have implications for clinical practice. For example, interventions to improve adherence could potentially benefit from providing written instructions and resources[38], not just verbal and the use of assistive technologies. Furthermore, it has implications for the use of brief screening tools to more efficiently identify at-risk patients for closer monitoring and the development of assessment tools to inform targeted adherence interventions[38].

These findings need replicating, though physicians should be aware of the error in assuming impairment in all areas of cognitive functioning will increase non-adherence. Health care providers need to explore CI patient’s understanding and concerns regarding medications. This may reduce a patient’s intentional non-adherence. Education about indication, medication and side effects, along with considering patient tolerance of side effects may improve intentional non-adherence [20]. Other specific cognitive domains received much less attention, with inconsistent or non-significant findings of those examined [18, 20, 30].

It is surprising that impairment in specific domains was not as useful in understanding and preventing medication non-adherence in this population [18, 20, 30]. Further research is needed to understand this complex relationship [15, 39] and elucidate if different patterns of suboptimal adherence may exist depending on the combinations of neurocognitive impairment [39].

Risk factors.

Most frequently reported risk factors for medication non-adherence were CI (or suspected CI) and absence of a caregiver or spouse living with the patient. The impact of impairment on adherence begun with MCI and did not worsen for participants with severe CI. This is important for clinical practice as the presence of MCI is easily missed [40]. Although it is not feasible to screen all older peoples for CI [41], clinicians should be aware of the relationship of these risk factors to non-adherence. Methods of ascertaining dementia or CI varied amongst the studies making it difficult to compare results. The assistance of a caregiver may potentially indicate advanced dementia/CI and may provide an explanation why a relationship between cognitive functioning and medication adherence is not reported [26]. Research should further explore these risk factors in order to reliably ascertain adherence.

Studies that did not find a relationship between dementia or CI and adherence most frequently used self-reports or interviews to ascertain adherence, both tending to lead to under-estimation of medication non-adherence [9]. These findings may be biased because of inherent methodological limitations as patient self-reports and interviews of CI persons are poor measures of treatment adherence [42]. Future studies should consider the use of other methods to identify adherence/non-adherence in an older population, including personalized medical records and direct pharmacist questioning, which have been suggested as optimal measurement tools in previous studies [43].

A significant number of people with dementia have a comorbid health condition, which may have serious implications for the way specialist services are delivered to people with dementia [44]. Regimen complexity [21]and the number of prescribed medications [20] were cited as risk factors for medication non-adherence. Physicians should be aware of the relationship between medication adherence and comorbidity, a factor likely to increase and complicate medication intake. Patients with multiple chronic disease may vary their opinions about health outcomes, such as, longer survival, prevention of disease-specific events, physical and cognitive function and tolerable risk of adverse drug reaction[45]. The difficulty for clinical practise is to rationally prescribe medications for older adults with multiple chronic conditions and reduced life expectancy whilst also analysing: the likelihood of benefit and goals of care and satisfying the basic principles of optimal medication use. Therefore, future research should focus on a subpopulation of persons with dementia or CI with co-morbid diseases [46].

Interventions.

Unfortunately, interventions were investigated by only two studies [29, 32] which had small sample sizes.

A recurring theme is the importance of caregivers for the success of interventions [29]. This, however, reinforces the dependency of older people with dementia and is inconsistent with the philosophy of promoting self-determination and independence critical to a person’s quality of life [47]. Individual strategies (e.g. medication regimens or setting up schedules) used in two studies [21, 26], led to recruitment of willing participants with caregivers, hence may not represent the general older population.

Several interventions (such as medication and disease education, medication reviews and packaging/dispensing of medications) for the general older adult population have been documented. Despite moderate success, few studies have attempted to translate these findings to older adults with dementia or CI. This study updates and extends the knowledge of a previously published systematic review [9]. This study included a more comprehensive search to retrieve a broader scope of articles and focuses on the impact of specific cognitive domains on medication adherence.

Research into the effect of a memory-prompting device designed for participants with HIV-associated memory impairment demonstrated improvements in adherence to highly active-antiretroviral therapy (HAART) for these participants but not for memory-intact participants [48]. These results suggest different approaches are necessary for each population.

Medical consequences.

Medication non-adherence, particularly overdose, may result in toxicity due to altered pharmacodynamics in the older population [49]. Well-known consequences: poor disease control, increased hospitalizations, disability and early death [20] were rarely examined in the appraised studies. Research in this field tends to exclude persons with dementia/CI, limiting the ability to extrapolate results to the growing cognitively impaired population [8, 9].