Objectives This study aims to summarise existing evidence on risk factors for drug-related problems (DRPs) in hospitals as well as ambulatory care or nursing homes and adds additional empirical evidence on risk factors for DRPs in non-elective hospitalised patients.
Methods A comprehensive literature review was performed to compose an overview of demographic, clinical and pharmacological risk factors associated with DRPs in different settings (ambulatory care, nursing homes and hospitals). A cross-sectional study on rehabilitation, cardiology and pulmonology wards of three hospitals in Nijmegen, the Netherlands, was performed to assess possible risk factors for DRPs in a clinical setting.
Results The comprehensive review identified 21 papers discussing risk factors for drug-related hospital admissions, use of potential inappropriate drugs, adverse drug reactions and other types of DRPs. The majority of these studies had been carried out in ambulatory care (11 papers; 52%). Polypharmacy, comorbidity and the use of specific drugs (antithrombotics, antidiabetics) were most often positively associated with the occurrence of DRPs. Our cross-sectional study demonstrated that admission to the rehabilitation ward, admission at the intensive care unit and comorbidity were associated with the occurrence of potential DRPs in a clinical setting.
Conclusions Although risk factors associated with DRPs differ greatly among published papers, comorbidity, polypharmacy and the use of specific drugs (antithrombotics, antidiabetics) were frequently associated with DRPs. Although several guidelines advise to use prespecified risk factors (like age, polypharmacy and renal impairment), one should be aware that most of these risk factors are insufficiently grounded on empirical evidence.
- Drug Utilization
- Drug-Related Side Effects and Adverse Reactions
- Medication Errors
Statistics from Altmetric.com
- Drug Utilization
- Drug-Related Side Effects and Adverse Reactions
- Medication Errors
Prescribing medication is one of the most commonly applied interventions in healthcare. In addition to the beneficial effects, medication therapy also introduces risks of medication errors, adverse drug events and other drug-related problems (DRPs).1 In order to reduce the number of DRPs and thereby aim to improve the effectiveness and safety of medication therapy, medication review is often proposed. Medication review is a structured evaluation of a patient's medicines, aimed at reaching agreement with the patient about drug therapy, optimising the impact of medicines and minimising the number of DRPs.2 The effectiveness of medication review is assessed in several randomised controlled trials, indicating that medication review reduces both the number of DRPs and the number of potentially inappropriate medicines.3–6
Conducting medication reviews without targeting patients at risk, however, is a time-consuming process and therefore expensive. Considering the expanding population of older people, medication reviews are likely to become resource intensive.2 Although the Preventing Hospital Admissions by Reviewing Medication (PHARM) study demonstrated that a pharmaceutical care process seems to reduce the number of medication-related (often costly) hospital admissions, the authors stated that a pharmaceutical care process like PHARM is unlikely to be cost saving in its present form.7 To improve the cost effectiveness of medication reviews, targeting high-risk patients is required for instance by identifying patients at high risk of having DRPs.8 Several studies tried to identify patients at risk of medication errors, DRPs and potential DRPs (pDRPs; DRPs who are not confirmed by the physician or patient) in different settings, but the results are still inconclusive and no useful identification of these patients has been described for a hospital setting. Furthermore, there is no published (systematic) review giving an overview of potential risk factors for selecting the most eligible patients for medication review. Therefore, our objective is to both summarise the literature on risk factors that are associated with patients having potential DRPs in different settings (ambulatory care, nursing homes and hospitals) and subsequently assess in an empirical study whether these risk factors are also applicable in a clinical setting.
Design: This study consists of (1) a comprehensive review of the existing literature in order to identify candidate risk factors to detect high-risk patients for DRPs and (2) subsequently a cross-sectional study to assess whether these candidate risk factors are also applicable in a clinical setting.
In order to identify published evidence on risk factors for patients at risk of having DRPs in different settings, a literature search was performed by the first author, in accordance with one of the coauthors in April 2013 in MEDLINE. The following Mesh terms were used: (“Home Care Services” OR “Ambulatory Care” OR “Hospitalization”) AND (“Adult” OR “Aged” OR “Aged, 80 and over” OR “Frail Elderly”) AND (“Comorbidity” OR “Polypharmacy” OR “Drug Utilization” OR “Drug Utilization Review” OR “Drug Prescriptions”) AND (“Medication Errors” OR “Inappropriate Prescribing” OR “Drug Interactions” OR “Drug-Related Side Effects and Adverse Reactions”). Titles were screened for relevance and abstracts of articles considered to be relevant were reviewed. Papers were included when they fulfilled the inclusion criteria: (1) the study must assess the association between risk factors/patient characteristics and patients having DRPs; (2) studies must be published in English, German or Dutch; and (3) an abstract has to be available. There were no restrictions with regard to study type. Risk factors for having DRPs extracted from the full-text articles were discussed in relation to applicable Dutch guidelines on medication review9 ,10 in a multidisciplinary panel with experts from different hospitals. This discussion resulted in a set of demographic, clinical and pharmacological factors supplemented with patient-related, drug-related and process-related risk factors based on expert opinion. The multidisciplinary panel consisted of a physician-clinical pharmacologist, hospital pharmacists, a hospital pharmacist in training (i.t.) and an outpatient pharmacist. All panel members were employed in at least one of the (three) hospitals in Nijmegen, the Netherlands; Sint Maartenskliniek (SMK), Canisius-Wilhelmina Hospital (CWZ) or Radboud University Medical Centre (RadboudUMC).
Cross-sectional study assessing risk factors for potential DRPs
Subsequently, the applicability of these risk factors in a clinical setting was determined in a cross-sectional study by classifying the presence of the predefined risk factors in hospitalised patients admitted to one of the three participating hospitals.
Design/setting of the cross-sectional study: Assessment of pDRPs took place at five non-elective wards; on respectively pulmonology and cardiology of RadboudUMC and CWZ and rehabilitation of the SMK. These wards were selected because there was a Dutch standard level of usual (pharmaceutical) care without additional pharmaceutical care interventions (medication review). Patients admitted on these wards on 24 July, 13 September, 28 and 29 November, 18 December 2013, and 8 and 27 January 2014 were included in these assessments.
Assessment: The assessment of pDRPs was conducted by a (hospital) pharmacist (i.t.) in consultation with a physician-clinical pharmacologist (i.t). Assessments were performed in a standardised way, based on a combination of explicit review criteria (Screening Tool to Alert doctors to Right Treatment/Screening Tool of Older Person's Prescriptions criteria10 ,11) and implicit medication review criteria, of which a composite medication review checklist was derived.12–17 Clinical data combined with a list of the patients' current medication were used as source of information in the assessments. Therefore, the performed assessments are technical medication reviews (reviews only based on a list of patients' medicine). Identified pDRPs were not discussed with the patient or treating physician. Only in case of severe potential DRPs, judged by the reviewing pharmacist or physician, the treating physician was contacted. Examples of these cases were no discontinuation of acetylsalicylic acid with concomitant melaena or initiation of captopril in a patient with a history of tickling cough on an ACE inhibitor.
Measurements: Potential DRPs were recorded and classified according to DOCUMENT,18 together with all demographic, clinical and pharmacological risk factors for the patients whose medication file was reviewed. The DOCUMENT system was designed for community pharmacy; however, it was applicable as well in a clinical setting. Potential DRPs were classified to categories drug selection, overdose or underdose, undertreated, monitoring and toxicity or adverse drug reaction.
Statistical analysis: For the statistical analysis of associations between possible explanatory factors and the number of patients with pDRPs, multivariate logistic regression using a forward selection method was performed. Potential risk factors were included in the logistic regression analysis if univariate analysis revealed a p value of <0.1. Univariate differences between the groups were tested with a χ2 test. p Values ≤0.05 were considered significant in the multivariate logistic regression.
Sample size calculation
There is an ongoing debate on the appropriate method of calculation of sample sizes needed to build reliable multivariate prediction/association models with a certain number of predictors with a given effect size. Multiple assumptions about prevalence of outcome and predictor variables, about intercorrelation of variables and about the nature of variables (estimations of degrees of freedom in case of continuous variables) are necessary, leading to false precision of estimates of sample size. Instead, various authors recommend to use a rule of thumb; a recommendation that is supported by simulation studies.19 A common rule of thumb is to formulate sample size requirements as events per variable (EVP), with a minimum of 10 EVP. In this project, a stepwise selection of candidate predictors will be followed, thus increasing the probability to obtain a reliable and concise prediction model. Assuming a sample size requirement of 15 EVP data, a sample of 150 patients is sufficient to build a reliable model including a maximum of 10 predictors.
The MEDLINE search revealed 328 publications. Twenty-one of these papers associated clinical/pharmacological factors with either risk of hospital admissions, use of potential inappropriate drugs, adverse drug reactions or other kinds of DRPs. These papers assessed risk factors in an ambulatory setting (11 papers, 52%), at hospital settings (8 papers, 38%) or at nursing homes (2 papers, 10%). Polypharmacy was most often associated with DRPs (16/21 papers with positive association). Factors such as comorbidity (11/18 positive associations), age (9/18 positive associations, 3/18 negative associations), gender (8/17 positive associations) and renal impairment (3/6 positive associations) were most often correlated to the occurrence of DRPs. All associations of characteristics with DRPs, statistically significant in one or more papers, are displayed in table 1, which also shows the setting it applies to (ambulatory, nursing homes or hospital).
Based on results of the comprehensive review, the expert panel defined a final set of possible risk factors for potential DRPs by combining the risk factors of the literature review and expert opinion. As several sociodemographic characteristics such as dependent living situation, poor economic status, living alone, alcohol abuse, non-adherence to drug therapy and self-medication (1) are not easily collectable in a hospital setting and (2) were not strongly associated with DRPs, these factors were not collected in this cross-sectional study. The length of hospital stay could not be identified since the patient's medical charts were assessed during hospitalisation. Depression was not defined as a separate feature, but as one of the possible comorbidities. The characteristic ‘anxiolytic drug use’ was modified by the expert panel to ‘use of benzodiazepines during daytime’ since this introduces a risk of falling. Antithrombotic drug use was further divided into characteristics ‘therapeutic use of anticoagulants’ and ‘antiplatelet use’. Comorbidity was assessed based on the medical charts. Additional to these characteristics based on the comprehensive review, the expert panel added the factors ‘liver function disorder’ (indicated as Child-Pugh class B), ‘obesity’ ( body mass index >30 or body weight male >120 kg/female >100 kg), ‘prescription of renally cleared drugs with narrow therapeutic indices’ (digoxine, lithium, sotalol, aminoglycosides, methotrexate), ‘delirium’ and ‘intensive care unit (ICU) stay during current hospitalisation’ since these factors are likely to introduce DRPs in clinical practice, based on practical experience.
Cross-sectional study assessing risk factors for potential DRPs
The baseline characteristics are displayed in table 2. A total of 131 patients were included. Within this group of patients, 83 (63%) aged 65 years or older, 43 (33%) patients had a impaired renal function (estimated glomerular filtration rate <60 mL/min/1.73 m2). In total, 100 patients (76%) were defined as polypharmacy patients (use of ≥5 drugs chronically). The population on cardiology wards is the oldest (78% aged 65 years or older), covers the highest percentage of patients with renal impairment (49%) and has the highest amount of polypharmacy patients (90%). The rehabilitation ward has the least aged population (only 40% aged 65+ years) and the highest prescription rate of benzodiazepines during daytime, and therefore, patients might be at risk of falling.
Among 131 medication assessments, 70 (53%) patients were identified having one or more potential DRPs. In total, 124 pDRPs (average 1.0 pDRPs per patient) were identified, pDRPs occurred most frequently in the rehabilitation ward (43; 1.4 average per patient), closely followed by the cardiology wards (74 pDRPs, average 1.0 per patient). Patients identified with pDRPs on cardiology and rehabilitation wards had >1 pDRP average (1.8 on cardiology vs 2.0 on rehabilitation wards). In pulmonology wards, the number of potential DRPs was the lowest (7 pDRPs, 0.2 average per patient). Patients in the pulmonology wards had no more than 1 pDRP per patient. The distribution of potential DRPs among the different populations on the three wards assessed is shown in table 3.
The majority of the potential DRPs were identified in classes U (undertreated) (35.5%), D (drug selection) (33.9%) and O (overdose or underdose) (25%) of the DOCUMENT classification system.18 Undertreatment of a condition covered 20.2% of all pDRPs. Potential DRPs in classes C (compliance) and E (education or information) could not be identified because a technical medication review without patient involvement was conducted. The distribution of pDRPs among different classes is depicted in table 4.
Using a less restrictive alpha level (p<0.1), univariate analysis showed that patients having at least one potential DRP tend to have more comorbidity, have more frequent polypharmacy, are more often obese, are more often admitted to the rehabilitation ward and are less often admitted to the lung ward. All patients admitted to the ICU had at least one potential DRP. After multivariate analysis on the possible explanatory variables, only comorbidity (OR 16.4; CI 3.7 to 73.2) and being admitted to the rehabilitation ward (OR 21.4; CI 4.2 to 108.2) were significantly associated (p<0.05) with the presence of at least one potential DRP in an individual patient.
This study tried to identify risk factors for pDRPs for patients clinically admitted on pulmonology, cardiology and rehabilitation wards. In order to assess these risk factors, first a comprehensive review of existing literature was performed, which showed that the evidence on factors to identify patients at risk for DRPs is limited and that mentioned risk factors were not unequivocal. Polypharmacy demonstrated the strongest association in literature with patients having various DRPs, followed by comorbidity, and the use of specific drugs (antithrombotics, antidiabetics), which were the only factors unambiguously found to be associated with patients having DRPs, irrespective of setting. All other patient characteristics seemed to be of limited relevance since they were not strongly associated with DRPs in a clinical setting.
Subsequently, in this cross-sectional study comorbidity, being admitted on the rehabilitation ward or a stay at ICU during current hospitalisation only were associated with the occurrence of potential DRPs in the clinical setting.
In our cross-sectional study, comorbidity was found to be significantly associated with patients having at least one pDRP, in accordance with results of the comprehensive review. Polypharmacy, however, was only possibly associated with potential DRPs in the univariate analysis (χ2 test), while this factor was not associated in the multivariate analysis. This might possibly be explained by the co-linearity of polypharmacy with comorbidity, as earlier shown in the literature and also confirmed in our analysis (p<0.001).39 The characteristics ‘admission at the rehabilitation ward’ and ‘being admitted to the ICU during current hospitalisation’ are new findings as they seemed not to be identified as risk factors for occurrence of DRPs in other (mostly non-hospital) studies.
This study has several limitations: although the sample size (n=131) was insufficient to draw firm conclusions, strong associations (and consequently risk factors with a low number to detect) applicable for the total hospital population should appear in this sample. Due to organisational limitations, the reviews were performed by different physicians together with different pharmacists, which might have introduced some variation in the assessment of pDRPs. However, all teams used the same structured protocol.
This study shows, that, besides comorbidity, the characteristics for patients at risk of (p)DRPs collected in a specific setting, mostly ambulatory care, might not be automatically extrapolated to clinical settings. For instance, 76% of hospitalised patients in this study used five or more drugs chronically, in contrast to 10% of patients in primary care.41 So distinctiveness of a prognostic factor depends strongly on the setting where it is applied. Even when differences between hospital settings, nursing homes and ambulatory settings are acknowledged, the results of this cross-sectional study cannot be extrapolated to other clinics or other types of wards because of considerable heterogeneity between wards and hospitals. Also, hospitalised patients should not be considered as a single population, but as several unique populations, depending on the types of patients and DRPs present on a ward.
In conclusion, this study could not clearly identify a set of risk factors, besides previous ICU stay, admission to the rehabilitation ward or comorbidity to identify hospitalised patients at risk of (p)DRPs. The urgent need to select patients who would benefit most from medication review at non-elective wards cannot be met and therefore, based on this study, no tool (besides selecting patients with multiple comorbidity) could be provided to deploy medication reviews with more efficiency.
Although current available guidelines on medication review already incorporated risk factors (like age and renal impairment), these risk factors should be used with caution. First, because evidence for these risk factors is often inconclusive. Our comprehensive review, for example, did not demonstrate sufficient data to prove that age is an independent risk factor for DRPs. Second, there is a significant heterogeneity between the setting in which medication reviews are performed (ambulatory, clinical settings) and even within a clinical setting risk factors varied between wards. Finally, current literature only identified risk factors associated with preventable hospital admissions related to medication, DRPs or potential inappropriate medication. For none of these factors, causality had been proven. In other words, it had not been demonstrated that intervening on these risk factors would result in decreased medication-related hospital admissions, and consequently, better quality or higher efficiency in care.
These conclusions emphasise the needs for additional research to enable healthcare providers to provide medication review to patients who would benefit most and therefore improve efficiency of medication review.
What is already known on this subject
Medication review seems to be an effective intervention to reduce drug-related problems (DRPs).
Several studies assessed possible associations between demographic and clinical risk factors and DRPs.
Medication review is time consuming.
What this study adds
A comprehensive literature review indicating that patients with comorbidity, polypharmacy and specific drugs (antidiabetics and antithrombotics) seem to benefit most of medication review.
Experimental data illustrating that comorbidity, being admitted to the rehabilitation ward, or admission to intensive care unit during current hospitalisation is associated with the occurrence of potential DRPs in a clinical setting.
The authors would like to thank Ala Keyany, pharmacist at Sint Maartenskliniek, for her contribution to the assessment of medication files and medical records.
Competing interests None declared.
Provenance and peer review Commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.