Factors Associated with the Detection of Inappropriate Prescriptions in Older People: A Prospective Cohort

(1) Background: Ageing is associated with complex and dynamic changes leading to multimorbidity and, therefore, polypharmacy. The main objectives were to study an older community-dwelling cohort, to detect inappropriate prescriptions (IP) applying the Patient-Centred Prescription model, and to evaluate the most associated factors. (2) Methods: This was a prospective, descriptive, and observational study conducted from June 2019 to October 2020 on patients ≥ 65 years with multimorbidity who lived in the community. Demographic, clinical and pharmacological data were assessed. Variables assessed were: degree of frailty, using the Frail-VIG index; therapeutical complexity and anticholinergic and sedative burden; and the number of chronic drugs to determine polypharmacy or excessive polypharmacy. Finally, a medication review was carried out through the application of the Patient-Centred Prescription model. We used univariate and multivariate regression to identify the factors associated with IP. (3) Results: We recruited 428 patients (66.6% women; mean age 85.5, SD 7.67). A total of 50.9% of them lived in a nursing home; the mean Barthel Index was 49.93 (SD 32.14), and 73.8% of patients suffered some degree of cognitive impairment. The prevalence of frailty was 92.5%. Up to 90% of patients had at least one IP. An increase in IP prevalence was detected when the Frail-VIG index increased (p < 0.05). With the multivariate model, the relationship of polypharmacy with IP detection stands out above all. (4) Conclusions: 90% of patients presented one IP or more, and this situation can be detected through the PCP model. Factors with higher association with IP were frailty and polypharmacy.


Introduction
High-income countries face significant population ageing [1,2], which is associated with complex and dynamic changes that lead to the appearance of one or more chronic diseases, giving rise to multimorbidity [3]. Older patients with multimorbidity often meet frailty criteria [3].
Frailty is defined as an increased vulnerability to stressors resulting from a decrease in the physiological reserves of different systems [4]. It has been determined by identifying a critical number of impairments in physical strength, physical activity, nutrition, and mobility [4]. It is known that frailty is associated with a higher need for healthcare resources and

Data Collected
Personal data: Age and gender. Functional data: Dependence or independence for medication management and the Barthel Index (BI) to assess basic activities of daily living were graded [21].
Medical data: we collected morbidities (from the diagnostic clusters within the Johns Hopkins University ACG system) [22] and adjusted-age Charlson Index [23]; dementia diagnosis, as stated in patients' medical records, and the degree of deterioration established following the GDS (Global Deterioration Scale) [24]; blood pressure available in the last year; and geriatric syndromes.
Analytical data: Full blood count, sodium, potassium, urea, and glycosylated haemoglobin (HbA1c) were collected if available during the last year.
Pharmacological data: Number of chronic medicines prescribed for at least six months before the Medication Review (MR). It was determined if the patient had moderate polypharmacy (between 5 and 9 medications) or excessive polypharmacy (10 or more medications) [7]. Type of medication (qualitative classification) was recorded by ATC (Anatomical Therapeutic Chemical) system. Detection of therapeutical complexity through the MRCI [25] and DBI [26].
Patients in end-of-life (EOL) were identified according to the NECPAL CCOMS-ICO© tool criteria [29]. These patients are considered to be in the last months or the year of their life. The identification of EOL was based on: (a) the previous identification by the primary care team, (b) advanced disease criteria [29], or (c) Frail-VIG index >0.50.
Main therapeutic goal: According to the patients' baseline situation, an individualized therapeutic goal was established: (i) survival when the patient's baseline was optimal; (ii) functionality in patients in an intermediate situation; and (iii) symptomatic control in patients with a very vulnerable established baseline situation (patients in EOL situation were included).

Medication Review
Each patient's pharmatherapeutic plan was reviewed through the application of the PCP model [19]. This model was a process with four systematic stages and a multidisciplinary team carried it out made up of the patient's primary care physician and nurse, with a consulting team (a geriatrician and a clinical pharmacist). The model focused all therapeutic decisions on the individualized global assessment of each patient: comprehensive geriatric assessment (CGA), the frailty index calculation (Frail-VIG index) [30], and the resulting individual therapeutic goal (prolonging survival maintaining functionality or prioritizing symptomatic control) [31]. The decisions were taken together with the patient or with their main caregiver in case of incapacity ( Figure 1). and the resulting individual therapeutic goal (prolonging survival maintaining functionality or prioritizing symptomatic control) [31]. The decisions were taken together with the patient or with their main caregiver in case of incapacity ( Figure 1).

Inappropriate Prescription (IP)
With the MR, different criteria were used to determine IP; for example, in patients at EOL, Type 2 Diabetes Mellitus, hypertension and cardiovascular therapy, dyslipidaemia, mental health and dementia, pain, and osteoporosis.
Patients at EOL (NECPAL CCOMS-ICO© tool criteria [29]): according to STOPPFrail criteria, medications aimed at prolonging survival and those for primary prevention were assessed for potential discontinuation. Medications for secondary prevention were individualised based on patient goals [31,32].
Type 2 Diabetes Mellitus (T2DM): Two main proposals were used to individualise hypoglycaemic treatment: (a) therapeutic intensity criteria, following the American Diabetes Association (ADA) recommendations [33][34][35]; (b) type of medication: sulphonylureas (SU) were considered inappropriate because of the high risk of hypoglycaemia [34,36]; metformin was considered inappropriate if there were non-adjusted doses in cases of renal failure [34]; glifozins (SGLT2 inhibitors) were considered inappropriate when it was prescribed in patients without heart failure and chronic renal failure (glomerular filtration rate (GFR) < 45 mL/min) [34,37]; and short-acting insulin or mixtures were also considered inappropriate, except when it could be justified [34]. Table 1 describes the therapeutic goals in T2DM according to the patient profile.

Inappropriate Prescription (IP)
With the MR, different criteria were used to determine IP; for example, in patients at EOL, Type 2 Diabetes Mellitus, hypertension and cardiovascular therapy, dyslipidaemia, mental health and dementia, pain, and osteoporosis.
Patients at EOL (NECPAL CCOMS-ICO© tool criteria [29]): according to STOPPFrail criteria, medications aimed at prolonging survival and those for primary prevention were assessed for potential discontinuation. Medications for secondary prevention were individualised based on patient goals [31,32].
Type 2 Diabetes Mellitus (T2DM): Two main proposals were used to individualise hypoglycaemic treatment: (a) therapeutic intensity criteria, following the American Diabetes Association (ADA) recommendations [33][34][35]; (b) type of medication: sulphonylureas (SU) were considered inappropriate because of the high risk of hypoglycaemia [34,36]; metformin was considered inappropriate if there were non-adjusted doses in cases of renal failure [34]; glifozins (SGLT2 inhibitors) were considered inappropriate when it was prescribed in patients without heart failure and chronic renal failure (glomerular filtration rate (GFR) < 45 mL/min) [34,37]; and short-acting insulin or mixtures were also considered inappropriate, except when it could be justified [34]. Table 1 describes the therapeutic goals in T2DM according to the patient profile. Hypertension (HT) and Cardiovascular Therapy: There is currently evidence suggesting less intensive monitoring in people with multimorbidity, particularly in cases of dementia or limited life expectancy [38]. Globally, blood pressure under 140/90 mmHg has been associated with a higher risk of falls and mortality [39][40][41]. We considered an antihypertensive medication as an IP in EOL patients when the patient's mean systolic blood pressure has been lower than 130 mmHg over the last year [31].
Dyslipidaemia: Statins are not recommended in EOL patients [32], regardless of the indication, particularly in primary prevention cases. In secondary prevention, we can individualise decision-making based on each patient's associated risks and benefits [31]. We considered a lipid-lowering drug as an IP when prescribed to a patient with a total cholesterol level under 150 mg/dL, because it is a malnutrition marker [41].
Mental Health and Dementia: The European Association of Palliative Care's recommendations were used to make decisions; they propose a different therapeutic main goal in patients with dementia according to the stage of their pathology, based on evidence and consensus among experts [42]. We considered chronic antipsychotic drugs as an IP when prescribed to patients without behavioural disorders over the last 3-6 months or when prescribed to treat insomnia, as there is no evidence to support this indication [31,42,43].
Pain: Following Beers/STOPP criteria, the following proposals were made [36,[44][45][46]: (a) Tricyclic antidepressants were considered an IP because of their anticholinergic effects; (b) non-steroidal anti-inflammatory drugs (NSAIDs) were considered inappropriate when they were not prescribed at the lowest dose or for the shortest time possible, because of their high risk of ADEs; (c) weak opioids such as tramadol and codeine were registered as IP unless prescribed at low doses, due to the risk of ADEs; (d) major opioids, such as morphine and oxycodone, were considered IP if they were not associated with a laxative; and (e) meperidine was considered an IP due to its anticholinergic potential.
Osteoporosis: We considered calcium supplements (except in cases of symptomatic hypocalcaemia), vitamin D, or anti-resorption drugs as inappropriate in EOL patients [32].
Other groups: Based on the PCP model, the medications that could not be considered a correct indication or the most optimised posology were recorded as IP.

Sample Size
IP prevalence in the frail older population was estimated at 71% to calculate sample size [47]. With a 95% confidence level and 5% accuracy, a minimum of 352 patients should be included.

Statistical Methods
IBM SPSS Statistics v27.0 statistical software was used to perform statistical analysis. The results for categorical variables were described as absolute and relative frequencies.
Outcomes for continuous variables were expressed by means and standard deviations (SD). The statistical tests used to evaluate the relationship between two qualitative variables were the Chi-square test (or Fisher's exact test in 2 × 2 tables where the expected frequencies were <5). The Student's t-test was used to analyse the relationship between quantitative and qualitative variables. To identify the factors associated with IP, we used univariate and multivariate logistic regression. Statistical significance was established when the value of p was under 0.05.

Subject Baseline Data
A total of 428 patients were enrolled (66.6% women). The mean age was 85.5 years (SD 7.67). Almost half of them lived in a nursing home (50.9%). Globally, they had moderate dependence for basic daily activities, with a mean Barthel Index of 49.93 (SD 32.14), a prevalence of frailty of 92.5%, with 73.8% of patients suffering some degree of cognitive impairment. Table 2 outlines the COP-cohort's baseline demographic, clinical, functional, and cognitive data, and Table 3 lists the baseline pharmacological data. Globally, a particularly high prevalence of IP was detected. Up to 90.0% of the patients had at least one IP.  Moreover, an increase in the prevalence of IP was detected when the Frail-VIG index increased (p < 0.05) (Figure 2).  Moreover, an increase in the prevalence of IP was detected when the Frail-VIG index increased (p < 0.05) (Figure 2 Table 4 shows the descriptive analysis of the baseline situation and the number of IPs. The clinical variables most associated with presenting at least one IP were BI and the number of morbidities. BI had a mean of 60.7 in patients without IP and 48.7 in patients with at least one IP (p = 0.020). Regarding the morbidities, 6.9% of patients without IP presented five or more morbidities, and the percentage was 93.1% when they had at least one IP (p = 0.011).  Regarding morbidities, 51.1% of patients without IP had HT and 69.6% of patients with at least one IP had HT (p = 0.003); 4.7% of patients without IP had T2DM and 21.8% of them had at least one IP had T2DM (p = 0.008). Concerning FI, there was also an association with IP. Patients with one or more IPs presented a higher mean of FI than those without IP (0.39 (SD 0.1) vs. 0.34 (SD 0.1) (p = 0.023)). Figures 3 and 4 show the prevalence of each polypharmacy degree and the prevalence of MRCI degree considering the number of IPs. In both, all comparisons showed statistically significant differences (p < 0.001). Remarkably, patients with no IPs presented the lowest polypharmacy and MRCI rates. Moreover, on the contrary, those patients with more IPs had increased polypharmacy and MRCI rates (p < 0.001). Figure 5 shows the prevalence of the DBI degree considering the number of IPs, and differences were statistically significant in all groups, except when IPs were 0 versus ≥1.  Figures 3 and 4 show the prevalence of each polypharmacy degree and the prevalence of MRCI degree considering the number of IPs. In both, all comparisons showed statistically significant differences (p < 0.001). Remarkably, patients with no IPs presented the lowest polypharmacy and MRCI rates. Moreover, on the contrary, those patients with more IPs had increased polypharmacy and MRCI rates (p < 0.001). Figure 5 shows the prevalence of the DBI degree considering the number of IPs, and differences were statistically significant in all groups, except when IPs were 0 versus ≥1.    Figures 3 and 4 show the prevalence of each polypharmacy degree and the prevalence of MRCI degree considering the number of IPs. In both, all comparisons showed statistically significant differences (p < 0.001). Remarkably, patients with no IPs presented the lowest polypharmacy and MRCI rates. Moreover, on the contrary, those patients with more IPs had increased polypharmacy and MRCI rates (p < 0.001). Figure 5 shows the prevalence of the DBI degree considering the number of IPs, and differences were statistically significant in all groups, except when IPs were 0 versus ≥1.      Table 5 outlines the types of IP analysed according to the Anatomical, Therapeutic, Chemical (ATC) classification. It is essential to highlight that the groups most frequently prescribed inappropriately were ATC A (alimentary tract and metabolism), B (blood and blood-forming organs), C (cardiovascular system), and N (nervous system), with a percentage of 24.72%, 9.29%, 30.85%, and 24.62%, respectively. Table 5. Inappropriate prescriptions identified considering the ATC (Anatomical, Therapeutic, and Chemical) classification system.  Table 6 shows the univariate analysis, which highlights the relationship of the detection of IP with the following variables: T2DM, number of morbidities (especially if they were ≥5), the Frail-VIG index (severe frailty), polypharmacy (both moderate as well as excessive), therapeutic complexity (high complexity), and DBI (high DBI).

Univariate Analysis and Multivariate Analysis
With the multivariate model, the relationship of polypharmacy with IP detection stands out above all, both moderate and excessive. The frailty index was a significant predictor factor in the univariate model for those with a high IPs presence (≥2 or, or ≥3), but this did not remain significant in the multivariate analysis. 48 Table 5 outlines the types of IP analysed according to the Anatomical, Therapeutic, Chemical (ATC) classification. It is essential to highlight that the groups most frequently prescribed inappropriately were ATC A (alimentary tract and metabolism), B (blood and blood-forming organs), C (cardiovascular system), and N (nervous system), with a percentage of 24.72%, 9.29%, 30.85%, and 24.62%, respectively. Table 5. Inappropriate prescriptions identified considering the ATC (Anatomical, Therapeutic, and Chemical) classification system.  Table 6 shows the univariate analysis, which highlights the relationship of the detection of IP with the following variables: T2DM, number of morbidities (especially if they were ≥5), the Frail-VIG index (severe frailty), polypharmacy (both moderate as well as excessive), therapeutic complexity (high complexity), and DBI (high DBI).

ATC Group Total
With the multivariate model, the relationship of polypharmacy with IP detection stands out above all, both moderate and excessive. The frailty index was a significant predictor factor in the univariate model for those with a high IPs presence (≥2 or, or ≥3), but this did not remain significant in the multivariate analysis.

Discussion
In this study describing a sample of older patients recruited at the community level, we detected a high prevalence of functional and cognitive impairment and frailty in a specific health region with semi-urban characteristics. Similarly, high rates of moderate and excessive polypharmacy, therapeutic complexity, and anticholinergic and sedative burden with the pharmacological data were observed. These results are higher than might be expected in a standard cohort with patients aged 65 or older [14,48]. This fact could be explained by the inclusion criteria that selected patients according to one objective criterion (presenting multimorbidity), as well as one subjective criterion (patients with multimorbidity whose primary care physician identified prescription management difficulties).
The application of the PCP model detected a prevalence of IPs of up to 90%. This result is a much higher proportion of IPs than those detected in other studies using explicit criteria (Beers and STOPP-START criteria) [49]. This data is probably due to two main reasons: (i) inclusion criteria with patients with multimorbidity [50,51] and (ii) the PCP model allows the optimisation of individualised medication, thus resulting in a more thorough analysis of the prescription.
Thus, PCP should generally be considered an advanced MR (based on medication history, patient information, and clinical information) that optimises the prescription process [52].
Regarding ATC groups, we found that four of the 13 groups included in this classification accounted for almost 90% of IPs (alimentary tract and metabolism, blood and blood-forming organs, and cardiovascular and nervous system). Once again, this shows that IP is usually concentrated in a small number of pharmacological groups [53,54]. According to the multivariate analysis, it is remarkable that polypharmacy is the variable most commonly associated with the IP presence.
Notably, the positive relationship between frailty and IP was detected in this descriptive study. Furthermore, in the univariate model, a relationship between the FI and IP was observed. Nevertheless, in the multivariate model, this relationship disappeared. This fact could be due to FI being the result of summarising all the other data analysed. Indeed, this study could help open a path towards new studies to investigate the relationship between frailty and IP.
The current study has some limitations, such as the lack of a larger sample of non-frail patients, which would allow us to conduct a more accurate statistical analysis.
As a future goal, it would be interesting to assess clinical and pharmacological outcomes after applying different proposals to individualise the therapeutic approach through a longitudinal follow-up study.

Conclusions
The application of the PCP model in older adults with multimorbidity enabled to identify up to 90% of them presenting at least one IP. Frailty had a positive association with IP detection, and polypharmacy was the most involved factor in IP detection.
However, more studies should be performed with frail and non-frail patients to validate the potential of this tool. Funding: This study was granted a 'PERIS'-research program, funded by the Health Department of the Catalan Government (file numbers SLT008/18/00139 and SLT008/18/00152). The funding organization had no involvement in the study design, data collection, analysis and interpretation, writing of the manuscript and submission for publication. No other funding sources were used to assist in the conduct of this study or preparation of this manuscript.

Informed Consent Statement:
We obtained verbal informed consent from patients or their main caregivers. Afterwards, we included the patient's verbal informed consent in their electronic health record.

Data Availability Statement:
The datasets generated during and/or analysed during the current study are available from the corresponding author on reasonable request.