A retrospective evaluation of the association between potentially inappropriate prescribing and adverse drug reactions among hospitalized elderly Nigerians

Background Potentially inappropriate prescribing is associated with adverse clinical outcomes in the elderly. The Beers Criteria were developed to improve the quality of medication prescribing and safety to the elderly. However, the nexus between the Criteria’s potentially inappropriate medicines (PIMs) and adverse drug reactions (ADRs) remains controversial in clinical practice. This study aimed to evaluate the association between the 2015 American Geriatrics Society-Beers’ (AGS-Beers) PIMs and ADRs among elderly inpatients. Methods A cross-sectional retrospective study was carried out among elderly patients aged ≥ 60 years that were hospitalized between 1st January and 31st December 2016 at the internal medicine wards of a Nigerian University teaching hospital. Eligible elders that were hospitalized or died within 24 hours were excluded from the study. The medical records of eligible participants were randomly selected and information including patients’ socio-demographics, medication and medical histories, and medicines utilization during hospitalization were extracted from the records. Two clinical pharmacists evaluated the medical charts for PIMs using the 2015 AGS-Beers Criteria and ADRs using the Naranjo algorithm. Bivariate analyses and subsequently, a binary logistic regression, were carried out to determine the association between independent variables and ADRs with P<0.05 being considered significant. Results A total of 268 participants mean age, 70.53 (8.22) years were evaluated. According to the AGS-Beers Criteria, 32.1% (86/268) received at least one PIM, 13.8% (37/268) experienced 43 ADRs during hospitalization of which diclofenac was the most implicated medication for it (8/43, 18.6%). The AGS-Beers’ PIMs were not associated with ADRs in the bivariate analysis (p=0.24). The diagnoses of musculoskeletal disorders (OR=7.38, 95%CI=2.68-20.34, p<0.001) and the use of anticholinergic medications (OR=4.02,95%CI=1.04-15.58, p=0.04) were significantly associated with ADRs in the logistic regression. Conclusion The

elderly inpatients experienced ADRs which were associated with diagnoses of musculoskeletal disorders and anticholinergic medications but not the 2015 AGS-Beers' PIM in the Nigerian teaching hospital. Keywords: adverse drug reaction, Beers criteria, elderly inpatient, potential inappropriate prescribing Running title: Potential inappropriate prescribing and adverse drug reactions in elderly Background Population ageing demands cost-effective healthcare practices that will assure the safety of the elderly and maximize their well-being [1]. Adverse Drug Reactions (ADRs) present a safety concern in the elderly due to its associations with increased morbidity, mortality and healthcare costs [2]. Literature suggests that at least 10% of elderly experience ADRs either as inpatients or as outpatients [3]. Simple interventions that can reduce ADRs in this group of population is therefore of utmost importance.
Although many factors including age, comorbidity, polypharmacy and a diagnosis of heart failure have been associated with ADRs in the elderly, the prescribing of potentially inappropriate medicines (PIMs) appears to be consistently reported in the literature [2,[4][5][6][7]. The PIMs are medications with higher risks of causing harms or less therapeutic effects among the elderly in the presence of safer and more effective alternatives [8]. The Beers Criteria and the Screening Tool of Older People's Prescriptions and Screening Tool to Alert to Right Treatment (STOPP/START Criteria) are amongst the explicit criteria developed to improve the safety of medication in the elderly by providing lists of PIMs to guide physicians' prescribing [8][9][10].
The Beers Criteria which were initially developed in the 1990s has been revised as more evidence unfold aimed at mitigating preventable ADRs and improving clinical outcomes among the elderly [8]. However, there is no conclusive evidence to show that the Criteria have achieved this goal in clinical practice because reports of associations between the earlier itinerary of the Criteria and ADRs are inconsistent and sometimes contradictory. A few studies among elderly Europeans, specifically in France and Italy that evaluated the association between the explicit criteria PIM list and ADRs found no association [11,12] while studies among the Americans and Brazilians contradicted this negative finding [13,14]. Few studies that compared the association between the Beers Criteria, STOPP/START criteria and ADRs found the latter PIMs to be better associated with ADRs among the elderly than the former [5,15]. Although, most of these previous studies applied the earlier versions of the Beers Criteria which have many shortcomings, the controversy remains even with the major review the Criteria have undergone since its adoption by the American Geriatrics Society in 2012. This is because a few studies that applied the 2012 American Geriatrics-Beers (AGS-Beers) associated its PIM list with ADRs and identified it as a predictor of mortality [16,17], while many other studies that used the same version of the Criteria found no associations among the study populations [18][19][20].
The 2015 AGS-Beers Criteria have incorporated new classifications including the PIMs to be avoided due to non-anti-infective drug-drug interactions (DDIs), the drug-disease or drug-syndrome interactions, or to be avoided or have their dosages reduced due to level of kidney functions in the elderly, in addition to the traditional classification of "independent of diagnosis" [8]. Medications without sufficient evidence to be regarded as PIMs were recommended for cautious use in the Criteria. There is, however, a dearth of information concerning the association between this latest itinerary of the AGS-Beers Criteria and ADRs among the elderly. In Nigeria, studies have applied the Beers Criteria to screen for PIMs without exploring the clinical outcomes associated with the prescription of such medications [21][22][23]. This study aimed to evaluate the prevalence of ADRs and to determine the association between the 2015 AGS-Beers' PIMs and ADRs among hospitalized elderly in Nigeria.

Study design and setting.
This study was a cross-sectional retrospective evaluation of ADRs among hospitalized elderly, using a medical chart review at the internal medicine wards of a public University teaching hospital, in Southwest, Nigeria. The hospital which is 205-bedded serves as a referral centre for many healthcare facilities in the region and had a number of geriatricians consulting in the medical wards at the time of this study.

Study population/Eligibility Criteria
Patients aged ≥60 years that were hospitalized in the internal medicine wards of the hospital between 1 st January and 31 st December 2016 were eligible for inclusion in the study. In Nigeria, adults aged ≥60 years are recognized as elderly in line with the United Nations cut-off and this was consistent with studies in many low and middle-income countries [4,17,19,24]. Eligible participants with incomplete socio-demographic information, discharged or transferred to another level of care, died within 24 hours of hospitalization, readmitted within the study period or on chemotherapeutic agents were excluded from this study. The data collection was carried out between 13 th April and 30 th July 2017. Figure 1 depicts the selection process of the participants. The study participants were identified from the hospital admission records. The eligible participants' medical record numbers were listed, and the included records selected using a simple non-blinded randomization technique with the aid of computer-generated random numbers. Included records with missing data were deleted before analysis.

Sample size calculation
The sample size was calculated based on the population of 846 elderly patients that were hospitalized at the internal medicine wards within the study period using the hospital admission records. Using this total population with a margin error of 5%, a power of 95% and 50% response distribution yielding a minimum sample size of 265, with additional 10% being added for attrition resulting in 292 participants for the maximum sample size as computed using a formula previously described [25].

Data collection
With the help of a trained research assistant, the selected records were retrieved and information including patients' socio-demographics, medical and medication histories, medicine utilization during hospitalization and duration of hospital stay were extracted using the researchers' designed checklist. The complexity of the patients' health conditions was determined using the Charlson's comorbidity index [26]. The patients' specific laboratory data including serum creatinine (mg/dL) or estimated glomerular filtration rate (eGFR) mL/min) calculated using the Cockcroft-Gault formula were documented [27]. Platelet counts (cells/mm 3 ), level of potassium and sodium electrolytes (mmol/L) at hospitalization and during hospital stay (where available) were also extracted.
Physicians' clinical judgment of ADR occurrence documented in the charts and the clinical decisions that were taken to mitigate the ADRs including the substitution of medication and reduction in doses were extracted from the review sheet.

Measurement of ADRs
A multifaceted approach including a detailed review of nursing and physicians' charts, laboratory data, and other clinical parameters was conducted. Attention was paid to patients' verbalized complaints documented in the charts. Radio-imaging data, such as ultrasound, echocardiography, and electrocardiogram parameters were however not considered for review because they were rarely ordered by the physicians. Potential ADRs were identified using the Institute for Healthcare Improvement (IHI) global trigger tool which has been found useful in retrospective evaluations of ADRs among inpatients in many healthcare settings [28].
Two clinical pharmacists independently assessed the ADRs using the Naranjo algorithm which was preferred to other tools for its high specificity and ease of use [29]. A further analysis was performed for only "definite", "probable" and "possible" ADR categories. The assessors also agreed on the causality of the ADRs. Where there were disagreements between the assessors, it was referred to a physician and consensus were reached. This study evaluated the PIMs for both predictable and unpredictable ADRs but determined the association for the predictable only.

Assessment of PIMs
The appropriateness of medication use during hospitalization was assessed using the 2015 AGS-Beers Criteria. The PIMs of "independent of diagnosis", to be avoided due to drugdisease or drug-syndrome interactions to non-anti-infective DDIs and those to be avoided or have their doses reduced due to the level of kidney functions in elderly were evaluated.
The strong anticholinergics listed in the Criteria were evaluated. The prevalence of medications that increase falls risks in the elderly according to the AGS clinical practice guidelines on fall prevention was also assessed [30].

Data management and analysis
The data were fed into the Statistical Packages for the Social Sciences software (SPSS version 25) and were primarily analyzed using descriptive statistics. Participants with missing data were deleted before the analysis. Student's independent t-test was used to compare normally distributed continuous variables and the results were presented in means and standard deviations. Bivariate analyses and subsequently, a binary logistic regression, were carried out to determine the associations between categorical variables and ADRs. Independent variables reported being associated with ADRs in previous studies including gender, age, comorbidity index, PIMs, diagnoses of heart failure, chronic kidney diseases and musculoskeletal disorders, use of anticholinergics and Falls associated medications and duration of hospitalization [2,[4][5][6][7] were included in the model.

Results
Of the 292 included participants 91.8% (268/292) were eventually analyzed, the remainder 8.2% (24/292) with missing data were deleted before analysis. Table 1 presents the baseline socio-demographic characteristics of the study participants.

Prevalence of PIMs among the study population
Almost half of the PIMs (39/86, 45.3%) was known to increase fall risks in the elderly (Supplement Table1). The PIMs were significantly associated with gender (p=0.049).   The prevalence of ADRs in this study (13.8%) was similar to 13.2% reported among the elderly inpatients in England [33] and consistent with the global average in this population [3]. It was however, lower than the 25.0% and 21.1% reported among the Italians and Brazilians respectively [4,7]. The variation in the definitions of ADRs, differences in cut-off age for defining the elderly and the methods of identification of the ADRs in the studies could partly account for the variations in the findings of these studies.

Adverse drug reactions among the study population
In this study, cardiovascular medications and non-steroidal anti-inflammatory drugs (NSAIDs) were implicated in many of the ADRs among the study participants consistent with similar studies in England and Brazil [4,33]. In this study, diclofenac caused gastrointestinal (GIT) bleeding and acute kidney injury among the participants. Diclofenac  [36]. In this study, the diagnoses of chronic kidney disease and heart failure were not associated with ADRs in contrast with other studies [4,6]. Although the duration of hospitalization was associated with ADRs in the bivariate analysis, it was no longer significant in the regression model. This result probably indicates that with good clinical practice, the duration of inpatient care may not be a determinant for ADR occurrence in the elderly. There was a negative association between age and ADRs in this study in contradiction of many studies [3,4,33], probably due to the high proportion of participants in the lower age range which was a reflection of the young old population in Nigeria.
A few studies have applied the Beers Criteria to evaluate the quality of medicine prescribing among elderly Nigerians [21-23], however, to the best of these authors' knowledge, this was the first study to assess associations between the Beers PIMs and ADRs in this population. This study will, therefore, add to the existing information about the potential adverse effects of medicines among the elderly in Nigeria.

Limitations of the study
This study has some limitations including the retrospective nature of the study.
Information such as laboratory data and medical history were missing in some patients' records. This limited the full application of the Criteria to all patients. The possibility exists that the PIMs may be higher than reported in this study. It is equally possible that the ADRs may have been underestimated since retrospective studies are known to detect lower ADRs compared to prospective evaluations [4]. Although adjustments were made for some cofounders in this study, it is possible that other cofounders existed that were not evaluated. The radio-imaging data such as ECG were not reviewed for ADRs, and this study was carried out in a teaching hospital which may limit generalizations to other healthcare facilities in the country.     *significant in bivariate analysis **significant in the logistic regression at p<0.05 Figure 1 Participant selection

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to