Exploring instruments used to evaluate potentially inappropriate medication use in hospitalised elderly patients in Kosovo

Abstract

Objectives

A number of instruments are used to identify potentially inappropriate medications (PIMs) in the elderly. In this study we identify PIMs in elderly patients and aim to compare three different instruments used to assess PIMs.

Methods

In this prospective cohort study, we compared medications of elderly patients against three commonly used instruments: Beers’ list, PRISCUS and STOPP/START, at the point of hospital admission and discharge in the nephrology clinic of Kosovo’s largest hospital. Readmission risk was evaluated using the LACE Index and correlations with the number of PIMs and PIMs criteria were analysed.

Results

Of 184 patients admitted to the nephrology clinic, 84 met study inclusion criteria. Patients had a median of three drugs at admission and four at discharge. Hospital readmission risk was high with median LACE Index being 11 (63% of patients). A higher number of PIMs was associated at the point of discharge compared with admission for all three tools (Beers’ list: 29% vs 38 %, P=0.04; STOPP/STRART: 20% vs 23%, P<0.001; PRISCUS list: 12% vs 21%, P<0.001). The number of drugs at admission predicted the number of PIMs at discharge only when using Beers' criteria (P=0.006). At discharge, each increase in medication was associated with an increase in PIMs based on Beers’ [0.134; (P=0.007)] and STOPP/START criteria [0.130; (P=0.005)]. Nitrofurantoin was the main PIM identified with Beers’ and PRISCUS list in comparison to proton- pump-inhibitors being the most prevalent agents identified with STOPP/START criteria.

Conclusions

There are differences when using Beers’ criteria, STOPP/START criteria and PRISCUS list during identification of PIMs in elderly patients with high readmission risk. These differences should be considered when identifying PIMs in hospital settings.

Introduction

Drug-related problems as a result of inappropriate prescribing are common in the elderly population.^1–3 Numerous studies found these drug-related problems to be based on altered pharmacokinetics and pharmacodynamics^{4 5} and to be associated with a higher risk for adverse drug events (ADE), an increase in hospitalisation rates,^6–8 morbidity and mortality.⁹ In light of these findings, one key strategy to mitigate the susceptibility of this population group towards high drug-related risks has been to reduce the use of potentially inappropriate medication (PIM) with the help of explicit criteria such as Beers’ criteria, Screening Tool of Older People’s potentially inappropriate Prescriptions (STOPP) and Screening Tool to Alert to Right Treatment (START) criteria, and the PRISCUS list. Beers’ criteria, which provide a list of criteria to identify PIMs, were first developed by the American Geriatric Society (AGS) in 1991 for use in nursing home patients,¹⁰ and then were updated and expanded in 1997,¹¹ 2003,¹² 2012¹³ and 2015¹⁴ to focus on all patients over 65 years in ambulatory, acute and institutionalised settings. The STOPP/START criteria were developed by a group of experts in geriatric pharmacotherapy in 2008¹⁵ and were updated in 2015¹⁶ to include 81 STOPP and 34 START recommendations organised by physiological organ systems. The STOPP/START criteria were also designed to address the duality associated with the identification of inappropriate medication use by providing a list of PIMs in STOPP criteria (therefore suitable for screening medications during admission) and afterwards addressing potential prescribing omissions with START.¹⁶ Lastly, the PRISCUS list (from project ‘priscus’ derived from latin for ‘old and venerable’)¹⁷ was published in 2010, as a result of the German Health Ministry’s drug safety initiative 2008/2009 to develop a list applicable to the German healthcare system.¹⁷ In this regard, it should also be emphasised that differences in approved drugs and therefore applicability in Europe, were also a reason for the development of STOPP/START criteria, which were referred to as ‘European Beers’ criteria’.¹⁸ PIM detection using different STOPP/START and Beers’ criteria has been previously compared.^{18 19} However, we identified only one study comparing all three i.e: Beers’ with STOPP and PRISCUS, which found that Beers’ criteria were not applicable to the German healthcare system.²⁰ In addition to above, these tools were not previously studied in the Kosovo setting, therefore additionally prompting the need to compare and explore their applicability in this setting.

Aims of the study

This study aimed to compare the number of detected inappropriately prescribed drugs by the three mentioned different PIM-lists, identify inappropriate key medications and medication groups, and to assess the prevalence of PIMs among elderly patients in nephrology. We also aimed to explore potential predictors associated with PIM use.

Ethics approval

This study complies with ethical principles for medical research as outlined by the Declaration of Helsinki and was approved by the Ethics and Professional Committee of the University Clinical Centre of Kosovo (No.1198).

Method

This prospective cohort study was carried out in the Clinic of Nephrology of the University Clinical Centre of Kosovo. Data was collected between August and October 2016 from all patients who were admitted to the Clinic of Nephrology during that period, and patients were recruited consecutively as they came in. The inclusion criteria was restricted to patients 65 years or older, who were hospitalised in the department and had at least one medication at the point of admission or at discharge and gave informed consent. Patients seen in the ambulatory setting and not admitted as inpatients were excluded. On admission, data on patient's age, gender, diagnosis, reason for admission, and medication used prior to admission and at discharge was collected by medical residents. Patient information was de-identified and securely stored by the investigators.

To predict the readmission risk among patients included in the study, the LACE Index, a validated risk assessment tool, was used.^21–23 This tool generates a composite risk score based on four parameters: length of stay, acuity of admission, co-morbidities and Emergency Department visits. A LACE score of ≥10 indicates a high readmission risk. The patient’s medication list was compared against three main criteria Beers’ (2015 update), PRISCUS and STOPP criteria to identify PIMs prescribed and used in elderly patients in the Clinic of Nephrology. Similar to the determination of LACE score, initially PIMs were identified by a pharmacist and then reviewed and confirmed by another consultant pharmacist specialised in medication therapy review and a nephrologist working at the nephrology clinic of the hospital. All pharmacists involved in the identification and subsequent review of PIMs (n=5) had working knowledge of the use of the three tools. The number of PIMs was calculated at admission and discharge as well as the frequency of PIMs within different drug/system categories outlined by each of the criteria. The prevalence of PIMs at admission and discharge was determined for each criterion by dividing the total number of patients with at least one PIM with the total number of patients.²⁴ The correlation between the number of PIMs in each criterion and the LACE Index score was analysed. Data collected was analysed using SPSS statistical analysis software (SPSS 21.0 for Windows, IBM corp., Armonk, NY, USA). Descriptive analysis was completed to identify differences in baseline characteristics. Chi-Square testing was done to assess for association between categorical data as well as linear regression modelling to estimate the relationship among one dependent variable and one or more predictor. Statistical significance was defined as two-tailed ≤to 0.05.

Results

There was a total of 84 patients who met the inclusion criteria and were included in the study out of all 184 patients admitted to the Clinic of Nephrology between August and October 2016. Patient baseline characteristics such as age, gender and number of drugs at admission and discharge are outlined in table 1. Patients had a median of three drugs at admission and four drugs at discharge. Patients with one to four drugs at admission and discharge made 59.5% and 65.4% of the sample, respectively. When evaluating the total LACE Index score, the median score was 11% and 63% of patients were identified to have a total score of 10 or more, which predicts a high risk for hospital readmission.

Table 1.

Patientdemographics

Patient characteristics
Gender	N (%)			Number of drugs
Female	46 (54.8)				Admission N (%)	Discharge N (%)
Male	38 (45.2)			0* drugs	10 (11.9)	1 (1.2)
Age	Years			1–4 drugs	50 (59.5)	55 (65.4)
Median	71			5–8 drugs	23 (27.4)	25 (29.8)
Range	65–91			>8 drugs	1 (1.2)	3 (3.6)
LACE Index score results
LACE parameters	Median	Min.	Max.	Length of stay in the hospital	N (%)	Total LACE Index score	N (%)
Length of stay (days)	1	1	7	1 day	70 (83.3)	0–9	31 (37)
Acuity of admission	3	0	3	3 days	3 (3.6)	ten or greater	53 (63)
Co-morbidities	5	0	5	4–6 days	4 (4.8)
ED visits	2	0	4	7–13 days	1 (1.2)
Total Lace Index	11	2	19	14 or more days	6 (7.1)

*10 patients had 0 mediations at admission point but nine of them were discharged with medication.

Number and prevalence of PIMs at admission and discharge

Among the 84 patients evaluated, the total number of PIMs identified using Beers’, STOPP and PRISCUS criteria were 24, 17 and 10 respectively. PIM prevalence at admission according to Beers’, STOPP and PRISCUS criteria were 29%, 20% and 12%, respectively. PIM prevalence at discharge according to Beers’, STOPP/START and PRISCUS, was 38%, 23% and 21%, respectively. There was a statistically significant difference between the use of these tools in the identification of PIMs (see table 2).

Table 2.

Number and prevalence of PIMs at admission and discharge

Number and prevalence of PIMs at admission and discharge		Beers' criteria	STOPP/START criteria	PRISCUS list
					Admission	Minimum PIM number	0	0	0
Maximum PIM number	3	3	2
Sum PIM number	29	22	12
PIM prevalence	24 (29%)	17 (20%)	10 (12%)
P-value	0.000	0.000	0.000
Discharge	Minimum PIM number	0	0	0
	Maximum PIM number	3	3	2
	Sum PIM number	40	25	20
	PIM prevalence	32 (38%)	19 (23%)	18 (21%)
	P-value	0.04	0.000	0.000
Association between length of stay score and PIM at discharge
PIM at discharge	Length of stay score	Scores 1–3	Scores 4–7	P-value
Beers' criteria	0-1PIMs	69	8	0.02
	2–3 PIMs	4	3
STOPP criteria	0-1PIMs	69	10	0.64
	2–3 PIMs	4	1
PRISCUS list	0-1PIMs	71	11	0.58
	2–3 PIMs	2	0
Correlation between the number of medications at admission and discharge with LACE Index scores
	Number of medications at admission Correlation coefficient (P-value)		Number of medications at discharge Correlation coefficient (P-value)
Length of stay score	- 0.040 (0.715)		0.108 (0.328)
Acuity of admission score	- 0.033 (0.766)		0.027 (0.808)
Co-morbidities score	0.226 (0.038)		0.383 (0.000)
Emergency department visit score	0.036 (0.743)		0.153 (0.165)
Total LACE Index score	0.064 (0.561)		0.274 (0.012)

Percentage of PIMs based on categories

The top three PIMs based on the Beers’ list were: nitrofurantoin (21.4%), drugs not dose adjusted or avoided in reduced renal function (16.7%) and antidiabetic agents – sliding scale insulin and sulfonylureas (14.3%). Based on the STOPP list, the top three PIMs were: proton pump inhibitors (20%), sulfonylureas (16%) and potential drug-drug interaction (beta blocker concurrently used with verapamil or diltiazem) (10%). Lastly, based on the PRISCUS list, the top three PIMs were: nitrofurantoin (45%), anti-hypertensive agents (ie, non-sustained-release nifedipine and alpha-blockers) (10%) and analgesics (ie, NSAIDs) (10%). The percentage breakdown of PIMs for each list is shown in table 3.

Table 3.

Number of PIMs based on categories

	N (%)
Beers’ criteria—PIM categories
Anti-infective: nitrofurantoin	9 (21.4)
Not dose adjusted or avoided in reduced kidney function: spironolactone, ranitidine	7 (16.7)
Endocrine: sliding scale insulin, sulphonylureas	6 (14.3)
GI: proton-pump inhibitors	5 (11.9)
Cardiovascular: nifedipine IR, digoxin, peripheral alpha blockers	4 (9.5)
Anti-cholinergic: anti-histamines, anti-spasmodics	3 (7.1)
Corticosteroids (long duration)	3 (7.1)
CNS: benzodiazepines	2 (4.8)
Analgesics: NSAIDs, meperidine	2 (4.8)
Drug-drug interaction: peripheral alpha-1 blocker + loop diuretic	1 (2.4)
STOPP/START criteria—PIM categories
GI: proton-pump inhibitors	5 (20)
Beta blocker + verapamil or diltiazem	4 (16)
Endocrine: sulphonylureas	4 (16)
Aldosterone antagonist + K conserving agents without potassium monitoring	3 (12)
Anti-cholinergics	3 (12)
Benzodiazepines	2 (8)
Analgesics: NSAIDs	2 (8)
Peripheral alpha-1 blocker + loop diuretic	1 (4)
Corticosteroids (long duration)	1 (4)
PRISCUS list—PIM categories
Anti-infective: nitrofurantoin	9 (45)
Anti-dementia, vasodilator, circulation-promoting agents (ie, pentoxifylline)	4 (20)
Analgesics, anti-inflammatory	2 (10)
Sedatives and hypnotics	2 (10)
Anti-hypertensive: alpha-blockers, non-sustained-release nifedipine	2 (10)
Antiarrhythmics	1 (5)

Predictors of the total number of medications and PIMs at admission and discharge

The total number of drugs at admission was found to be positively correlated to the total number of drugs at discharge (P=0.0001). Additionally, an increased number of drugs at admission was correlated to an increase in the number of PIMs at admission, based on Beers’ criteria only (P=0.006) (table 4). While there was no correlation of the total number of drugs at admission with the number of PIMs at discharge (P>0.05), the total number of drugs at discharge was associated with a statistically significance difference in the number of PIMs at discharge based on Beers’ and STOPP criteria, but not the PRISCUS list. At discharge, each increase in medication was associated with 0.134 (P=0.007) and 0.130 (P=0.005) increase in PIMs based on Beers’ and STOPP criteria respectively. Lastly, for all three criteria, the number of PIMs at admission was also positively associated with the number of PIMs at discharge. Age and gender did not make a statistical difference in the number of PIMs at either admission or discharge. Detailed results of regression models are presented in table 4.

Table 4.

Regression coefficients with associated P-values

Independent variable	Dependent variable	Regression coefficient	P-value	95% CI
Number of total medications at admission	Beers’: PIM at admission	0.081	0.006	0.024 to 0.139
	Beers’: PIM at discharge	0.043	0.205	−0.024 to 0.110
	PRISCUS: PIM at admission	0.011	0.579	−0.029 to 0.052
	PRISCUS: PIM at discharge	0.012	0.623	−0.036 to 0.059
	STOPP: PIM at admission	0.051	0.074	−0.005 to 0.107
	STOPP: PIM at discharge	0.053	0.078	−0.006 to 0.113
	Number of total medications at discharge	0.817	0.000	0.671 to 0.964
Number of total medications at discharge	Beers’: PIM at discharge	0.134	0.007	0.038 to 0.232
	PRISCUS: PIM at discharge	0.024	0.524	−0.050 to 0.098
	STOPP: PIM at discharge	0.130	0.005	0.041 to 0.219
Beers’: PIM at admission	Beers’: PIM at discharge	0.942	0.000	0.808 to 1.076
PRISCUS: PIM at admission	PRISCUS: PIM at discharge	0.920	0.000	0.765 to 1.075
STOPP: PIM at admission	STOPP: PIM at discharge	0.971	0.000	0.881 to 1.065

Correlation of LACE Index scores with total number of medications, and PIMs

Length of stay, acuity of admission and Emergency Department visit scores did not make a statistically significant difference (P>0.05) in the total number of medications at either admission or discharge. However, the co-morbidity score was positively associated (P=0.0001) with the number of medications at both admission and discharge, as well as total LACE Index with number of medications at discharge (P=0.012) (table 2). Further, acuity of admission, co-morbidity, Emergency Department visit and total Lace Index scores did not make a statistically significant difference to the total number of PIMs at admission or discharge. In contrast, when comparing the length of stay score and PIMs at discharge among patients based on Beers’ list, there was a statistically significant association (P=0.02). However, this association was not significant for PIMs identified at discharge when using PRISCUS and STOPP criteria (table 2).

Discussion

Our study found that the total number of drugs as well as the number and prevalence of PIMs, consistently increased from the time of admission to the time of discharge, regardless of the criteria used to identify the PIMs. PIM prevalence in our study was highest with Beers’ criteria, second with STOPP/START criteria, and lowest with the PRISCUS list and within the reported range of 14%–66% of previous studies.^{18 25 26} This is also in line with previously reported suggestions that a combination of tools may lead to a higher number of PIMs identified.¹⁸ The prevalence of PIMs found by the three PIM lists in our study with Beers’>STOPP>PRISCUS is in line with many other studies (table 5). The study of Siebert et al ²⁰ reported poor concordance between international lists and the drugs available in Germany and hence might be regarded as an exception.

Table 5.

Comparison of PIM prevalence obtained by different PIM lists, studies and settings

Study	Setting	Highest prevalence of PIM (%)	Second highest prevalence of PIM (%)	Third highest prevalence of PIM (%)
Our study	hospital, nephrology, admission	Beers' (29)	STOPP (20)	PRISCUS (12)
Our study	hospital, nephrology, discharge	Beers' (38)	STOPP (23)	PRISCUS (21)
Siebert et al20	hospital, admission	STOPP (48)	PRISCUS (35)	Beers' (31)
Wickop et al28	hospital, admission	STOPP (79)	FORTA (61)	PRISCUS (36)
Tosato et al18	hospital, during hospitalisation	Beers' (58)	STOPP (50)
Gallagher et al19	hospital, admission	Beers' (25)	STOPP (12)
Oliveira et al29	primary care	Beers' (52)	STOPP (34)
Grace et al30	nursing home	Beers' (89)	STOPP (85)

Apart from clinical relevance, the Beers’ list seems to be appropriate to identify most PIMs, for use in the elderly in Kosovo and potentially other Balkan countries with similar healthcare systems and drug inventories. However, when clinicians use any of the three criteria explored in this study, they should consider potential differences in individual-approved medications for countries where the screening criteria are applied. Additionally, clinicians both locally and internationally, should have in mind that these tools need to always be evaluated in the clinical context given that the percentage of PIM detection may not be associated with clinical relevance.

The leading categories of PIMs in our study varied among all three lists. Nitrofurantoin use was the leading PIM for Beers’ and PRISCUS lists, while for STOPP criteria the proton pumps inhibitor class was the leading PIM. Beers’ and STOPP criteria were concordant on the inappropriate use of sulfonylureas. Some of the leading PIMs such as nitrofurantoin and NSAIDs have been associated with increased risk for unplanned hospitalisation according to Price et al.⁸ These discordances suggest that there are intrinsic differences among the three lists in regard to which drugs should be considered potentially inappropriate in the geriatric population and therefore be of a clinical concern. In light of these findings, more research is required to identify differences and gaps between these three lists and other criteria currently being used for the management of pharmacotherapy in the elderly in order to develop more inclusive and standardised guidelines. It might also be interesting to study the clinical and practical relevance of the discrepancies in a larger multi-centric study across various countries, which would also address potential differences in approved drugs.

When assessing the relationship between LACE Index scores and the total number of drugs at admission and discharge, co-morbidity scores were positively associated with the total number of drugs at both admission and discharge, which reinforces the well-reported link of multiple comorbidity and polypharmacy in the geriatric population. Moreover, an increase in total medications at discharge was associated with an increase in the total LACE Index score, which is indicative of the higher risk of hospital readmission. This finding is important given the high percentage (63%) of patients in our sample had a high LACE Index score (10 or greater).

The regression modelling identified few predictors for the total number of medications and PIMs at admission and discharge when controlling for other independent variables. An increase in the total number of medications at admission was associated with an increase in the total number of medications at discharge and the number of PIMs at admission (based on Beers’ criteria only) but not at discharge. Meanwhile, as the total number of medications increased at discharge, the number of PIMs at discharge increased concurrently (based on Beers’ and STOPP lists only). Lastly, an increase in the number of PIMs at admission consistently predicted an increase in PIMs at discharge regardless of the criteria used to identify the PIMs, which is similar to what Osei et al²⁴ reported. A systematic review by Hill-Taylor et al²⁷ found polypharmacy as well as age (75 years or older) and female gender as predictors of PIM in a higher number of studies. Our findings suggest that polymedication at admission and discharge can be an indicator to address PIMs. Furthermore, our data shows that resolving PIMs during hospitalisation is not a matter of course but needs to be resolved proactively.

Strengths and limitations

To our knowledge, this is the only conclusive study which compares the prevalence of PIM use in the geriatric population using three international criteria (Beers’, STOPP and PRISCUS). Using the regression models, we highlighted the relationship between different variables and identified key predictors that drove the number of total medications and the number of PIMs at admission and discharge. Additionally, by using the LACE Index tool, we were able to assess the impact of medication use during hospitalisation on a long-term outcome—risk of hospital readmission. This is especially important given that ultimately one of the aims of reducing PIM use among the elderly is to reduce hospital readmission. However, this study has several limitations. One limitation of the study is the small sample size which was contributed to by the study’s time limitation and the use of only one department. Moreover, collecting the data from one department potentially introduced selection bias and therefore the results of the study should be interpreted with caution. Even though the PIM lists are validated, it should be kept in mind that a comparison of the number of detected PIMs does not rate clinical relevance.

Conclusions

The present study suggests that the prescribing of PIMs in the elderly patients in Kosovo is common and that PIM prevalence is similar to the prevalence reported in previous studies internationally. Hospitalisation appears to be a particularly high-risk period for prescribing of PIMs. Our study identified the total number of medications at admission and discharge are important predictors for total number of PIMs, therefore clinicians should carefully screen for unnecessary medications and PIMs. Furthermore, the variation in PIM number and categories identified using criteria such as Beers’, STOPP/START and PRISCUS reflects the lack of standardised consensus on PIMs and therefore can lead to confusion among healthcare professionals. Thus, although these criteria are evidence-based tools essential in the clinical decision-making process, they should not supersede clinical judgement and patient values and needs. Further research in this field is needed and could include the relevance of the found PIMs in clinical practice as well as an analysis of the other instruments used to identify PIMs.

What this paper adds.

What is already known on this subject

• PIM prevalence analysed with Beers’ criteria, STOPP/START criteria and PRISCUS is highest with Beers’ criteria and is similar to other reported studies.

• Polypharmacy at hospital admission and discharge is associated with higher comorbidity scores in the elderly.

What this study adds

• Explores application of Beers’ criteria, STOPP/START criteria and PRISCUS in the Kosovo setting.

• Identifies differences between Beers’ criteria, STOPP/START criteria and PRISCUS in relation to leading drug categories identified by each instrument;.

• Identifies that number of drugs at hospital admission point predict the number of PIMs at the hospital discharge point only when using Beers’ criteria.

Data availability statement

Data are available upon reasonable request. All data relevant to the study are included in the article or uploaded as supplementary information.

Ethics statements

Patient consent for publication

Not required.

Ethics approval

This study complies with ethical principles for medical research as outlined by the Declaration of Helsinki and was approved by the Ethics and Professional Committee of the University Clinical Center of Kosovo (No.1198).