Introduction

In Western countries, adverse drug reactions are an important medical problem, resulting in 3–5% of all hospital admissions, accounting for 5–10% of in-hospital costs and being associated with a substantial increase in morbidity and mortality [1, 2, 3, 4, 5]. Older patients are particularly vulnerable to drug-related illnesses because they are usually on multiple drug regimens, and age is associated with changes in pharmacokinetics and pharmacodynamics [6, 7].

The use of inappropriate medications, defined as medications in which the risk outweighs the benefit, is a major factor influencing the likelihood of adverse drug events among the elderly. Since 1991, Beers and colleagues have developed a comprehensive set of explicit criteria for inappropriate medication use, with the intent of providing a useful tool for assessing the quality of prescribing in older persons [8, 9, 10]. Despite these criteria not representing substitutes for careful clinical consideration by physicians, they may be used in drug utilization review, as the basis for educational materials and in assessing the quality of prescribing. In addition, the identification of factors related to inappropriate drug consumption can be used to design and implement effective drug utilization programs aimed at influencing prescribing patterns [10, 11].

Previous studies have documented widespread inappropriate medication use in the community, nursing homes, board and care facilities, physician office practices, and homebound elderly with a prevalence ranging from 12% to 40% [8, 9, 11, 12, 13, 14, 15, 16, 17, 18]. However, the use of inappropriate medications among in-hospital patients has rarely been evaluated [19, 20]. Hospitalized older adults are usually 'frail' and present with acute diseases, which may increase their susceptibility to adverse medication effects and raise the severity of drug-related illnesses [2]. Moreover, in-hospital patients, who often have a genuine need for many drugs, are usually victims of a "prescribing cascade", which leads to an increased likelihood of receiving inappropriate drug therapy [21].

Therefore, the aims of the present study were (a) to determine the prevalence of inappropriate medication use in a hospitalized elderly population and (b) to identify predictors of this use. To accomplish this issue, we used data from the study of the Italian Group of Pharmacoepidemiology in the Elderly (GIFA), which has been designed to explore drug use and quality of in-hospital care in Italy. Nationwide, continuous data acquisition has led to the creation of a database, containing information on a large, and representative population of elderly patients admitted to acute care hospitals.

Methods

GIFA database

The GIFA is a group of investigators operating in community and university-based hospitals throughout Italy. The GIFA periodically surveys drug use, occurrence of adverse drug reactions, and quality of hospital care. The methods of the GIFA study have been described in detail elsewhere [4, 22]. Briefly, all patients admitted to 81 geriatric and internal medicine wards participating in the study were enrolled and followed until discharge. The study periods were the following: 1 May to 30 June and 1 September to 31 December 1988; 15 May to 15 June 1991; and 1 May to 30 June and 1 September to 31 October in 1993, 1995, 1997 and 1998.

For each participant, a questionnaire was completed at admission and updated daily by a study physician who received specific training. Data recorded included socio-demographic characteristics, indicators of physical function and cognitive status, clinical diagnoses on admission and at discharge, medications taken prior to admission, during hospital stay, and those prescribed at discharge.

Inappropriate medication use

To identify inappropriate medication use, we adopted the criteria developed and published by Beers in 1997, which identify patterns of medication use that unnecessarily place older persons at risk of adverse drug reactions [10]. These criteria were developed with the intent to be generalizable to any population of persons older than 65 years, regardless of their level of frailty or their place of residence. We used 25 of the 26 criteria for inappropriate medication use independent of diagnosis. These are listed in Table 1. We did not apply the criterion for cough and cold preparations containing antihistamines because our system for drug coding did not specifically identify this class of drugs. Only medications used during hospital stay were considered for the present study.

Table 1. Prevalence of inappropriate medication use based on Beers' criteria. Propoxyphene, trimethobenzamide, meperidine, doxepin were not available on the market in Italy during the surveys periods. Phenylbutazone, meprobamate, methyldopa and were not used by any patient. The criterion for cough and cold preparations containing antihistamines was not used because the system for drug coding did not specifically identify this class of drugs
Full size table

Covariates

Cognitive performance was assessed at hospital admission using the Hodkinson abbreviated mental test (AMT) [23]. This test has proven to be reliable for detecting both mild cognitive impairment and dementia in older populations, and has previously been adopted in epidemiological surveys [24, 25]. A score below 7 was used to identify patients cognitively impaired [25].

Drugs were coded according to the Anatomical Therapeutic and Chemical codes [26]. Diagnoses were coded according to the International Classification of Diseases, Ninth Edition, Clinical Modification codes [27]. Co-morbidity was quantified using the Charlson co-morbidity index by adding scores assigned to specific discharge diagnoses, as illustrated in the original publication [28]. For analytical purposes, we computed a variable for heart disease, which includes diagnoses of ischemic heart disease, atrial fibrillation and congestive heart failure. ADL disability was defined as need of assistance to perform one or more of the following tasks: eating, dressing, bathing, transferring, and toileting. Alcohol use was defined as consumption of any amount of alcohol before hospital admission. Three levels of variables for number of drugs used during hospital stay and length of hospital stay were computed based on tertiles.

Data analysis

To reduce the variability in medication use related to changes in drug availability on the market and to select a uniform sample of patients, we limited the analyses to participants aged 65 years or older hospitalized in 1997 or 1998 according to surveys of those years. Therefore, from the initial sample of 32,181 patients, we excluded those enrolled in the study between 1988 and 1995 (n=24837), and those younger than 65 years (n=1610). In order to identify independent predictors of use of any inappropriate medication during hospital stay, a multivariate analysis was performed in the resulting sample of 5734 patients. From bivariate comparisons, variables associated with the use of any inappropriate medication at a P level ≤0.10 were selected and entered into a logistic regression model adjusted for age and gender.

To compare characteristics of participants receiving and not receiving inappropriate medications, we used the Chi-square test for categorical variables. Differences between continuous variables were assessed using analysis of variance (ANOVA) comparisons for normally distributed parameters; alternatively, the Kruskal-Wallis test was adopted. A value of P below 0.05 was considered statistically significant. All analyses were performed using SPSS for Windows version 10.0.

Results

In the 1997–1998 period, a total of 5734 patients aged 65 years or older were enrolled in the study. The principal characteristics of the population are illustrated in Table 2. Mean age ±SD was 79.0±7.5 years; males and females were equally represented. During hospital stay, 837 (14.6% of the study sample) patients received one or more medications classified as inappropriate based on Beers criteria. More specifically, 697 (12.2%) patients used one inappropriate medication and 140 (2.4%) two or more. As shown in Table 1, ticlopidine (n=346; 6.0% of the study sample) was the most frequently used medication, among those in Beers' list, followed by digoxin (n=174; 3.0%), amytriptyline (n=113; 2.0%), chlordiazepoxide (n=91; 1.6%) and diazepam (n=91; 1.6%).

Table 2. Bivariate comparison of principal characteristics of the population (n=5734) according to use of inappropriate medications
Full size table

At the bivariate analyses, subjects receiving an inappropriate medication were younger, had a lower prevalence of cognitive and functional impairment and were more likely to be enrolled in a 1998 survey than other participants (Table 2). In addition, these subjects had a higher co-morbidity index, consumed a higher number of drugs during hospital stay and had a longer length of hospital stay than participants not receiving inappropriate medications (Table 2). Entering these variables in a logistic regression model, age and cognitive impairment were associated with a reduced likelihood of receiving an inappropriate medication. In contrast, Charlson co-morbidity index and overall number of medications used during hospital stay were associated with a higher use of inappropriate medications (Table 3).

Table 3. Factors associated with inappropriate medication use. Results form the multivariate analysis. All the variables were simultaneously entered in the logistic regression model. The likelihood ratio test, assessing the statistical significance of fit improvement resulting from the addition of the variables to the model, was 248,566 (P<0.001)
Full size table

Since ticlopidine was responsible for about one-third of the cases of the inappropriate medication use, we repeated the analysis excluding this drug from the list elaborated by Beers. Based on these revised criteria, 527 participants (9.2% of the study sample) received an inappropriate medication. However, the results of the multivariate analysis showed that the inverse association of inappropriate medication use with age [75–84 years vs 65–74 years, odds ratio (OR) 0.72, 95% confidence interval (CI) 0.59–0.89; ≥85 years vs 65–74 years, OR 0.58, 95% CI 0.45–0.77] and cognitive impairment (OR 0.75, 95% CI 0.59–0.95) was confirmed. In addition, the number of medications used during hospital stay (5–8 medications vs <5 medications, OR 2.42, 95% CI 1.73–3.40; ≥9 medications vs <5 medications, OR 4.97, 95% CI 3.55–6.96) still showed a significant direct relationship with the outcome, while no association was observed for the Charlson co-morbidity index (≥2 vs 0–1, OR 0.99, 95% CI 0.82–1.20).

Finally, to confirm the inverse relationship between age and inappropriate medication use, and in consideration of the fact that the two drugs most frequently responsible for inappropriate use (ticlopidine and digoxin) are usually prescribed in subjects with heart disease, we repeated the multivariate analysis in participants with and without this disease. Overall, 2754 participants (48% of the study sample) had heart disease (which includes diagnoses of ischemic heart disease or atrial fibrillation or congestive heart failure). As expected, compared with other participants, subjects with heart disease had a higher likelihood of receiving an inappropriate medication (16.9% vs 12.5%; P<0.001). However, as shown in Fig. 1, in the multivariate analyses, the inverse association of age groups with inappropriate drug use was confirmed in both groups of subjects with and without heart disease.

Fig. 1.
figure 1

Odds ratio and 95% confidence interval of inappropriate medication use by age groups, in subjects with and without heart disease. Heart disease includes diagnoses of ischemic heart disease, atrial fibrillation and congestive heart failure. Analyses are adjusted for gender, cognitive impairment, ADL disability, Charlson co-morbidity index, number of medications used during hospital stay, length of hospital stay and year of survey

Full size image

Discussion

In this study, we showed that among hospitalized older adults inappropriate drug use was common, and about 15% of participants were receiving at least one of the medications listed in the 1997 Beers criteria. In this sample, age and cognitive impairment were inversely associated with inappropriate drug use, while a direct relationship was observed for a number of drugs used during hospital stay and Charlson co-morbidity index.

Beers criteria have been criticized, since they do not identify all causes of potentially inappropriate prescribing and sometimes identify appropriate prescribing as inappropriate [29, 30]. However, these criteria represent a widely used and standardized tool for pharmacological research, despite the fact that they can not be used as a substitute for careful clinical judgement [10]. In this context, we did not apply diagnoses-related criteria for inappropriate medication use. Indeed, the high co-morbidity rate observed in this 'frail' population may have required careful adjustments of pharmacological regimens based on individual needs of each patient, leading to use of medications contraindicated for a certain disease to treat a coexisting one.

The prevalence of inappropriate medication use in this population was lower than those reported from other studies conducted in the United States [8, 9, 11, 12, 13, 14, 15, 16, 17], but in line with a previous observation in a European population [18]. This result may be explained by differences in pharmaceutical market among countries. Several of the drugs on Beers' list are not available in Italy. For example, propoxyphene, which appeared to be one of the most frequently prescribed inappropriate medications in the United States [11, 12], is not available on the market in Italy. Similarly, the fact that clopidogrel was not available on the market in Italy during the surveys periods, may have been responsible for the elevated consumption of ticlopidine as an alternative to aspirin. In this context, differences between the USA and Italy in indications to use of medications may have influenced our results. For example, Italian guidelines to anti-thrombotic treatment in the elderly, considered as appropriate the use of ticlopidine in subjects with ischemic heart disease intolerant to aspirin [31].

Finally, the characteristics of this study population may have influenced the prescribing pattern. For example, the elevated rate of heart disease observed in this sample, may explain the high consumption of ticlopidine and digoxin, which accounted for about 50% of any inappropriate medication use.

In contrast with other studies [9, 11, 14], we found that age was inversely related to the likelihood of receiving an inappropriate medication. We believe that this finding may be explained by the fact that older hospitalized adults in this population were 'frail' and at high risk of experiencing drug-related illnesses [2]. In this context, physicians may have been cautious in prescribing high-risk medications in this vulnerable population and they may have preferred to use medications with a safe profile. An alternative explanation is that older patients, compared with other participants, had a higher need of 'life-saving' medications with indubitable positive effects to treat acute severe conditions. This, by converse, may have reduced the use of other medications, which were not clearly beneficial for the patient.

Similarly, cognitive impairment may represent a marker of frailty. Therefore, the fact that, compared with other participants, cognitively impaired patients were less likely to use inappropriate medications probably reflected, among these clinically complex patients, a careful choice by physicians of medications whose risks do not outweigh benefits. Interestingly, the different use of inappropriate medications related to cognitive status may partially explain the inverse association between cognitive impairment and drug-related illnesses we have previously documented [32].

The most important factor predicting the probability of inappropriate medication use was the total number of prescription medications. The most likely hypothesis is that each prescription had a certain probability of being inappropriate, thus proportionally increasing a subject's probability of being on an inappropriate therapy with each additional medication. Similarly, the direct association between co-morbidity and inappropriate medication use may be explained by the fact that the elevated number of medications needed to treat multiple coexistent conditions increased a patient's likelihood of receiving an inappropriate prescription.

An important limitation of this study relates to generalizability of the results. Our findings, which are based on a hospitalized population cannot be extrapolated to subjects living in the community. Moreover, as we mentioned above, as consequence of differences in the pharmaceutical market among countries, the results of this study should not be compared with those of previous observations in the United States or other European countries [33]. In addition, the measurement of some predictors of inappropriate medication use may not have been accurate. For example, data on alcohol intake were self-reported and, therefore, subject to misclassification, and information on type of alcoholic beverage and duration of alcohol use were not gathered. Finally, this study is only assessing the problem of use of potentially harmful medications, not exploring the issue of underuse of beneficial therapies. Such underuse problem has been identified in the management of a broad range of chronic conditions in elderly patients, including cardiovascular disease, hypertension, osteoporosis and pain management [34, 35, 36, 37, 38] and it deserves the same level of attention and scrutiny given to problems of overuse of drug therapies.

In conclusion, the present study showed that inappropriate medication use is common among hospitalized older adults in Italy. The most important determinant of risk of receiving an inappropriate medication was the number of drugs being taken. Older age and cognitive impairment were associated with a reduced likelihood of using an inappropriate medication. We believe that the understanding of the factors associated with inappropriate prescribing may be valuable in designing and implementing effective drug utilization programs to influence prescribing patterns and they can be incorporated into medication use control programs.