http://orcid.org/0000-0002-8604-6258
Figures
Abstract
Background
Adverse drug reactions (ADRs) are important causes of morbidity and mortality in the healthcare system; however, there are no studies reporting on the magnitude and risk factors associated with ADR-related hospitalisation in Ethiopia.
Objectives
To characterise the reaction types and the drugs implicated in admission to Jimma University Specialized Hospital, Southwest Ethiopia, and to identify risk factors associated with ADR-related hospitalisation.
Methods
A prospective cross-sectional study was conducted from May 2015 to August 2016 among consenting patients aged ≥18 years consecutively admitted to medical wards taking at least one medication prior to admission. ADR-related hospitalisations were determined through expert review of medical records, laboratory tests, patient interviews and physical observation. ADR causality was assessed by the Naranjo algorithm followed by consensus review with internal medicine specialist. ADR preventability was assessed using Schumock and Thornton’s criteria. Only definite and probable ADRs that provoked hospitalisation were considered. Binary logistic regression was used to identify independent predictors of ADR-related hospitalisation.
Results
Of 1,001 patients, 103 (10.3%) had ADR-related admissions. Common ADRs responsible for hospitalisation were hepatotoxicity (35, 29.4%) and acute kidney injury (27, 22.7%). The drug classes most frequently implicated were antitubercular agents (45, 25.0%) followed by antivirals (22, 12.2%) and diuretics (19, 10.6%). Independent predictors of ADR-related hospitalisation were body mass index (BMI) <18.5 kg/m2 (adjusted odd ratio [AOR] = 1.69; 95% confidence interval [CI] = 1.10–2.62; p = 0.047), pre-existing renal disease (AOR = 2.84; 95%CI = 1.38–5.85, p = 0.004), pre-existing liver disease (AOR = 2.61; 95%CI = 1.38–4.96; p = 0.003), number of comorbidities ≥4 (AOR = 2.09; 95%CI = 1.27–3.44; p = 0.004), number of drugs ≥6 (AOR = 2.02; 95%CI = 1.26–3.25; p = 0.004) and history of previous ADRs (AOR = 24.27; 95%CI = 11.29–52.17; p<0.001). Most ADRs (106, 89.1%) were preventable.
Conclusions
ADRs were a common cause of hospitalisation. The majority of ADRs were preventable, highlighting the need for monitoring and review of patients with lower BMI, ADR history, renal and liver diseases, multiple comorbidities and medications. ADR predictors should be integrated into clinical pathways and pharmacovigilance systems.
Citation: Angamo MT, Curtain CM, Chalmers L, Yilma D, Bereznicki L (2017) Predictors of adverse drug reaction-related hospitalisation in Southwest Ethiopia: A prospective cross-sectional study. PLoS ONE 12(10): e0186631. https://doi.org/10.1371/journal.pone.0186631
Editor: Sreeram V. Ramagopalan, University of Oxford, UNITED KINGDOM
Received: July 4, 2017; Accepted: October 4, 2017; Published: October 16, 2017
Copyright: © 2017 Angamo et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: The authors received no specific funding for this work.
Competing interests: The authors have declared that no competing interests exist.
Introduction
Adverse drug reactions (ADRs) are one of the leading causes of morbidity and mortality in the healthcare system [1]. Globally, studies have reported that the overall prevalence of ADR-related hospitalisation varies from 0.2% [2] to 54.5% [3]. A recent review of 43 observational studies identified a comparable prevalence of ADR-related hospitalisation in developed and developing countries [1]. However, most of the studies included in this review were conducted in developed countries, where the disease characteristics and prevalence, access to medicines, drug use patterns and management systems differ markedly from those in developing countries [4]. With respect to potential risk factors for ADRs, developing countries differ from developed countries in several important areas. These include greater proportions of patients taking antituberculosis (anti-TB) and antiretroviral therapy (ART) [5], a high prevalence of anaemia and malnutrition [6], widespread use of traditional remedies [7], a higher incidence of concomitant anti-TB drugs and ART with overlapping adverse effects [6], and increasing rates of concomitant infectious and non-communicable diseases demanding multiple medications with potential interactions [8–12]. There are also higher frequencies of genetic variants conferring increased risk for ADRs for commonly used drugs treating cancer, human immunodeficiency virus (HIV)/acquired immunodeficiency syndrome (AIDS) and tuberculosis (TB) in persons of African descent [13].
In Ethiopia, a developing country, the risk of ADR-related hospitalisation is a health concern due to an increasing number of patients eligible for anti-TB and ART, due to decentralisation and scale-up of the HIV/AIDS care programme [14], and concomitant drug management of HIV/TB co-infected patients [14]. According to a 2014 World Health Organization (WHO) report, the prevalence of all forms of TB was 211 per 100,000 of the population, of whom 13% of patients were HIV co-infected [15]. The prevalence of HIV among the adult population in 2015 was estimated to be 1.0% with national ART coverage of 52.0% [16]. There is growing attention to chronic diseases management with multiple medications [11, 17]. The use of new and complex therapies for chronic diseases has increased due to the establishment of community care programmes [18] and strengthening of health systems for chronic care at the level of local health centres, primary, general, and specialised hospitals [18]. The provision of primary health care services by health extension workers (health professionals working at the lowest health care level targeting preventive and curative health services at the household level) [19] and the expansion of community pharmacies and drug stores at the community level has remarkably increased medication access and use in the community, although medication regulation is poor [20]. There are also higher rates of drug-related problems [21] and irrational use of medicines [22] among patients with chronic illnesses in ambulatory care clinics that could lead to drug-related harm.
Studies have identified several factors contributing to ADR-related hospitalisations including older age [23, 24], female gender [25, 26], increased number of co-morbidities [27, 28], increased number of medications [29, 30], renal diseases [31, 32], liver diseases [33], heart failure [28], higher Charlson Comorbidity Index (CCI) [34], presence of chronic illnesses [35, 36], and history of previous ADRs [37, 38]. HIV/AIDS patients taking ART have been identified as a risk factor for ADR-related hospitalisations only in developing countries [8, 39]. Drugs commonly reported as contributing to ADRs include anticoagulants [24, 40], non-steroidal anti-inflammatory drugs (NSAIDs) [41], and angiotensin-converting enzyme inhibitors [41]. However, studies focussing on risk factors for ADR-related hospitalisation in developing countries are very limited in number [1]. Therefore, identification and reporting of factors contributing to ADR-related hospitalisation for community-based patients is crucial to develop preventive strategies to decrease the burden in the developing world [9, 10].
To our knowledge, there are no studies reporting on the prevalence and risk factors associated with ADR-related hospitalisation in Ethiopian patients. Thus, the main aim of this study was to characterise the reaction types and the drugs implicated in admission to Jimma University Specialized Hospital (JUSH), Southwest Ethiopia, and to identify risk factors associated with ADR-related hospitalisation.
Methods
Study setting, design and population
This study was conducted at the JUSH, which is the major public hospital in southwest Ethiopia with a catchment population of about 15 million people [42].
A prospective cross-sectional study was conducted from May 2015 to August 2016. Consenting patients with complete medical records aged ≥18 years and taking at least one medication prior to admission to medical wards were included in the study. Patients were excluded if they were unwilling to participate, unable to be interviewed due to health or other reasons, not taking at least one medication prior to admission and had incomplete medical and medication records.
Data collection
One of the authors (MTA) interviewed consecutively admitted patients as soon as practical for socio-demographic information, social drug use, medical history, drug allergies, use of over-the-counter and herbal medicines (S1 Table). Patients’ medical records were reviewed for admission diagnosis, ADR history, and clinical data within 48 hours of admission. Medication exposure in the month preceding hospitalisation was obtained through review of patients’ medical records and /or interview with the patient or family members. Common laboratory tests were evaluated within 48 hours of admission for each patient case including: renal function (serum creatinine mg/dL, blood urea nitrogen mg/dL and estimated glomerular filtration rate (eGFR) mL/min/1.73m2), liver function (alanine transaminase IU/L, aspartate transaminase IU/L, alkaline phosphatase IU/L and total bilirubin mg/dL) and complete blood count (white blood cell count cells/mm3, red blood cell count cells/mm3, haemoglobin g/dL, haematocrit %, and platelet count cells/mm3). Parameters describing nutritional and metabolic status (serum albumin g/dL, total triglyceride mg/dL, low-density lipoprotein mg/dL, high-density lipoprotein mg/dL, glucose mg/dL), serum electrolytes (serum calcium mmol/L, potassium mmol/L and sodium mmol/L) and coagulation status (prothrombin time) were recorded. Vital statistics such as body temperature, blood pressure, respiratory rate and pulse rate were recorded at admission. Other diagnostic data, such as echocardiography, ultrasound, electrocardiogram and viral markers and urinalysis results, were also evaluated.
Definitions of terms and variables used
The WHO definition of an ADR was used in this study: “any response to a drug which is noxious and unintended, and which occurs at doses normally used in man for prophylaxis, diagnosis, or therapy of disease, or for the modification of physiological function” [43]. This definition excludes treatment failure, drug abuse, intentional drug overdose and accidental or self-poisoning. Alcohol consumption was recorded as number of standard drinks per day. Khat chewing, which was of interest as it contributes to elevated blood pressure [44], was defined as regular chewer if a patient chewed khat at least four times per week. Drugs were classified using the WHO Anatomical Therapeutic Chemical Classification (ATC) System [45] and diagnoses were coded according to International Classification of Primary Care 2nd edition [46]. Calculation of the number of medications was based on the number of active ingredients in single and combination products [47]. A cut off point for polypharmacy was ≥6 medications [48]. A cut off point for comorbidities was ≥4 diseases based on a similar study [39]. Adult nutritional status was assessed using the body mass index (BMI) [49] and classified as <18.5 kg/m2 and ≥ 18.5 kg/m2 for analysis purpose. Patients were considered to have pre-existing chronic kidney disease if the eGFR was <60 mL/minute/1.73m2 and they had documented abnormal renal ultrasound (abnormal renal echogenicity or kidney size or presence of cysts) for at least 3 months prior to admission [50]. Drug-induced acute kidney injury was suspected among patients with baseline renal insufficiency (eGFR < 60 mL/minute/1.73m2), volume depletion and multiple exposures to nephrotoxic agents prior to admission, as long as other potential causes were excluded. Chronic liver disease was considered pre-existing if liver diseases (such as cirrhosis, chronic viral hepatitis) or liver dysfunction or liver injury were documented by the treating physician prior to admission [33]. Drug-induced hepatotoxicity was suspected when aspartate aminotransferase (AST) or alanine aminotransferase (ALT) levels were 3 or more times the upper normal limit (UNL) or total bilirubin was 2 or more times UNL [51], as long as other potential causes were excluded. The response to rechallenge or re-exposure to some drugs, such as anti-TB drugs, were implemented based on the WHO guidelines for the treatment of TB [52]. HIV/AIDS and TB patients were those patients diagnosed with HIV/AIDS and TB, respectively, and taking ART and anti-TB drugs prior to the current admission. All comorbidities were defined as present if documented in the medical records. A patient who had studied at least to a primary school level and was able to read and write in the local language(s) was considered as educated, whereas other patients were considered uneducated. According to the Ethiopian context, cities and small towns were considered as urban areas while rural villages or other similar clusters were considered as rural areas.
Identifying ADRs as a cause for hospitalisation
The primary researcher (MTA) evaluated all patients admitted to medical wards during the study period to assess if the admission had been caused by an ADR. The identification of whether one or more drugs led to the hospitalisation was based on a review of medical records, evaluation of laboratory tests, interview with patients or family members about medication usage and physical observation. An ADR was suspected if there was a relationship between the time of drug administration and the onset and course of the adverse reaction, while excluding other potential causes. The known adverse reaction profile of each drug was evaluated based on Ethiopian National Drug Formulary (2014), British National Formulary [53] and Up-To-Date 19.3 [54]. Confirmation of the causal relationship of an ADR to the suspected medication was performed using the Naranjo ADR assessment scale [55]. Applying the Naranjo algorithm, ADRs were classified as definite (9–12 points), probable (5–8 points), possible (1–4 points), or doubtful (0 points). The senior supervising internal medicine specialist (DY) independently and blindly reviewed all cases of suspected ADRs and cases without suspected ADRs for the presence of ADRs and to confirm the Naranjo rating using a similar approach to other studies [56, 57]. The primary researcher (MTA) and the senior supervising internal medicine specialist met to reach a consensus decision on the presence of an ADR-related admission and excluded possible and doubtful cases. When consensus was not reached on the causality assessment, an additional clinical pharmacy specialist’s opinion was sought for majority decision. Only definite and probable ADRs that provoked hospitalisation were considered. ADRs were assessed for preventability using Schumock and Thornton criteria [58] through the same approach. ADRs were classified as type A (dose dependent, augmented pharmacological and predictable reactions) and type B (bizarre, dose independent and non-predictable reactions) according to the Rawlins and Thompson classification method [59]. ADRs observed during the hospital stay were excluded.
Data analysis and interpretation
Data was recorded into an Access database (2016, Microsoft, Redmond, Washington) and analysed using the IBM Statistical Package for the Social Science (IBM SPSS version 23.0 Inc., Chicago, Illinois). Frequency of adverse reaction types and drugs implicated were determined. The chi-square test or Fisher’s exact test were used to compare categorical data between ADR- and non-ADR patients. For non-normally distributed variables, comparisons were undertaken with Mann-Whitney tests and the results were presented as medians and interquartile ranges (IQR). For normally distributed continuous variables, comparisons were performed using Student’s t-tests with results presented as means and standard deviations (SD).
Independent variables were assessed for multicollinearity and association to rule out correlation between two or more independent variables using variance inflation factor. Independent variables with p<0.25 in univariate analyses were entered into a multivariable binary logistic regression model to determine independent predictors of ADR-related hospitalisation. The performance of the ADR-related admission risk prediction model was assessed using the area under the receiver operator curve (AUROC), which assesses the ability of independent risk score to predict ADR-related admissions. The ROC curve was prepared using R (2015, R Foundation for Statistical Computing, Vienna) [60]. A p value of <0.05 was considered statistically significant in all analyses.
Ethics
The study was approved by the Tasmanian Health and Medical Human Research Ethics Committee (reference number H0014718) and the Jimma University Institutional Review Board (reference number RPGC/58/2015). We received permission from Jimma University Specialised Hospital management to conduct the research. Written informed consent was obtained from all individual participants included in the study.
Results
General information
A total of 3,224 patients were screened; 2,223 patients were excluded due to age <18 years (255), not taking at least one medication prior to admission (576), incomplete medical and medication records (70), unwillingness to participate (463) and inability to be interviewed as a result of health or other reasons (859). Of the 1,001 patients who met the inclusion criteria, 103 (10.3%) were deemed to have experienced an ADR-related admission (Fig 1).
One hundred and nineteen ADRs were identified among the 103 patients, equating to 1.2 ADRs per patient. Of these, 26.1% were definite and 73.9% were probable. Most ADRs (106, 89.1%) were considered preventable. Ninety-nine (83.2%) of the ADRs were considered pharmacologically predictable (type-A reactions) (Table 1).
ADR characteristics and drugs implicated in ADRs
The most common ADRs responsible for hospital admissions were hepatotoxicity (35, 29.4%) followed by acute kidney injury (severely declined glomerular filtration rate and hypovolemia) (27, 22.7%), skin reactions (8, 6.7%), hypokalaemia (7, 5.9%) and gastrointestinal bleeding or gastritis (7, 5.9%). Commonly implicated drugs in hepatotoxicity were isoniazid (21, 11.7%) followed by pyrazinamide (18, 10.0%), efavirenz (5, 2.8%) and atorvastatin (5, 2.8%). Commonly implicated drugs in acute kidney injury were tenofovir (9, 5.0%), furosemide (7, 3.9%), and enalapril (7, 3.9%). The drugs most commonly suspected of causing skin reactions were sulfamethoxazole (2, 1.1%), trimethoprim (2, 1.1%) and rifampicin (2, 1.1%). The drug most commonly suspected of causing hypokalaemia was furosemide (6, 3.3%). The drugs most commonly suspected of causing gastrointestinal bleeding were warfarin (3, 1.7%), heparin (3, 1.7%) and diclofenac (3, 1.7%). In general, anti-TB agents constituted the major source of ADRs (45, 25.0%) followed by antivirals (22, 12.2%) and diuretics (19, 10.6%). Table 2 provides a summary of ADRs and causative agents.
Drugs used
Out of the 4,018 drugs used by the 1,001 patients, 562 (14.0%) were used by patients who experienced ADR-related admissions. One-hundred and eighty of the 562 drugs were implicated in the 119 ADRs. Anti-infectives constituted the major proportion (52.1% versus 40.5%, p = 0.001) followed by cardiovascular system agents (18.9% versus 24.9%, p = 0.013) in both the ADR and non-ADR groups, respectively (Table 3).
Patient characteristics in univariate analysis
The main socio-demographic difference between the ADR-and non-ADR groups was a difference in urban versus rural residence (P = 0.004), where urban residents were more likely to be admitted with ADRs than rural residents. The mean body mass index (kg/m2) was lower in the ADR group, 19.1±2.8 versus 20.1±2.9, p = 0.004. There was no difference between the ADR and non-ADR groups for age, gender, educational status, smoking, alcohol use, khat chewing and herbal use. Patients with a previous ADR history (p<0.001), pre-existing renal diseases (p = 0.003), pre-existing liver diseases (p<0.001), TB taking anti-TB drugs (p<0.001) and HIV/AIDS taking ART (p<0.001) were more likely to be admitted with ADRs than those without these risk factors. Similarly, ADR admission was more likely in patients who had been hospitalised in the preceding 3 months (p = 0.030). Patients in the ADR group had a more comorbidities (4.0±1.4 versus 3.2±1.2, p<0.001) and used more drugs than the non-ADR group (5.5±2.7 versus 3.9±2.1, p<0.001) (Table 4).
Factors associated with ADR-related admissions
Variables included in the multivariable binary logistic regression model were those with p<0.25 in the univariate analyses. There was no multicollinearity identified in the included variables. Six risk factors associated with ADR-related hospitalisation were identified in the multivariable binary logistic regression model: BMI < 18.5kg/m2, pre-existing renal diseases, pre-existing liver diseases, number of diagnoses ≥4, number of medications ≥6 and previous ADR history (Table 5). The ADR-related admission risk prediction model was found to have an AUROC of 79.0% (95% CI 73.9%-84.1%) (S1 Fig). The model had a sensitivity of 59.2% and specificity of 86.6%, suggesting that it can moderately rule-in patients at risk of ADRs and strongly rule-out those patients not at risk of ADRs.
Discussion
Identification and reporting of predictors of ADR-related hospitalisation for community-based patients is crucial to develop preventive strategies and responsible care of patients. Medical practitioners may lack awareness of factors predicting ADRs leading to hospitalisation [61, 62]. To overcome this, studies, mainly from developed countries [28, 37], have identified several predictors of ADRs causing hospitalisation. However, Ethiopia is a developing country with different healthcare issues. These issues include a lesser ability to provide healthcare [19], a rising proportion of colliding epidemics of infectious and non-communicable diseases demanding multiple medications with potential of interactions [11, 63], a greater prevalence of HIV/AIDS and TB co-infection [64] with overlapping adverse effects of their medications [6], and a less health-literate population [14]. Additionally, malnutrition and anaemia are more common than in developed countries [65, 66]. Our study, the first in Ethiopia, identified risk factors associated with ADRs, and characterised the reaction types and drugs implicated in admission to the JUSH in Southwest Ethiopia.
We found that 10.3% of admissions were related to ADRs, a prevalence comparable to other studies from South Africa [39] and Argentina [67]. Similarly, a review of studies with a similar design to ours found comparable rates of ADRs in both developed and developing countries [1]. More than half of the ADR-related admissions were due to hepatotoxicity (mainly due to isoniazid and pyrazinamide) and acute kidney injury (mainly due to tenofovir, enalapril and furosemide). Very few studies, only Patel et al. [68] from India and Mouton et al. [39] from South Africa, have reported comparable proportions of anti-TB-induced hepatotoxicity and tenofovir-induced renal impairment. The higher incidence of hepatotoxicity in the current study could be due to concomitant anti-TB, ART and other medications with overlapping hepatotoxic effects. In addition, there was a substantial number of patients with malnutrition in the current study, and potentially the slow acetylation status of some Ethiopian patients with isoniazid [69] could have exacerbated the hepatotoxic reactions. ADRs commonly responsible for hospitalisation in previous studies included gastrointestinal bleeding [25, 31], electrolyte and metabolic disturbances [8, 28, 36] and cardiovascular disorders [28, 30]. Similar ADRs were reported in the current study, but at lower rates probably due to differences in population, diseases characteristics and drug therapy used.
Most of the predictors of ADRs, such as number of comorbidities [27, 28], number of drugs [28, 29, 36], pre-existing renal failure [31, 70], pre-existing liver diseases [37] and history of previous ADRs [37, 38], were in line with reported findings from both developed and developing countries. HIV/AIDS patients taking ART was identified as an independent predictor that raised the risk of experiencing ADRs in South African studies [8, 39] in 2005 and 2013, respectively. However, this variable was not included in the predictive model in the current study. This is possibly due to the intrinsic relationships between HIV/AIDS and malnutrition, multiple comorbidities and polypharmacy, although multicollinearity was not detected. In addition, the strength of the effect of HIV/AIDS patients taking ART as an independent variable is likely less than that of the other variables possibly because HIV/AIDS patients taking ART would have shown crossover-interaction with ADR-related hospitalisation. Additionally, compared to earlier South African studies, the burden of HIV/AIDS is now markedly reduced due to improvements in health care and greater availability of more tolerable ART. The two South African studies were conducted in communities with a high prevalence of HIV/AIDS, when more toxic antiretroviral drugs were in use. Finally, pharmacogenomic variations and other clinical characteristics of patients may have contributed to this difference [69].
Patients with pre-existing renal diseases were more likely to be hospitalised with ADRs than patients with normal renal function, as has been identified in other studies [71, 72]. About one-third of the patients presented to JUSH with an eGFR <60 mL/min/1.73m2 (Table 4), of whom, 18.5% had pre-existing renal diseases. Nearly one-third of patients were underweight (Table 4) and patients with lower BMI were at increased risk of developing ADRs in the current study (Table 5), which is in line with another study [73]. The combined effect of reduced muscle mass (malnutrition) with chronic illnesses may have contributed to a depressed eGFR despite normal serum creatinine levels, and concealed renal insufficiency may impact on the clearance of hydrophilic drugs [74]. In addition, patients in the current study might unintentionally be overdosed based on their weight and/or renal function (using eGFR) that may have led them to develop ADRs. It is therefore important to evaluate renal function to assess the potential therapeutic benefit against the risk of ADRs before initiating renally cleared drugs.
Similar to previous studies [33, 75], patients with pre-existing liver diseases were more likely to develop ADRs. Patients with liver diseases may have multiple comorbidities that require complex medical regimens [76]. Pharmacokinetic and pharmacodynamic changes such as a decreased drug elimination or increased toxic metabolites, alteration in drug distribution or protein binding provide opportunity for adverse reactions [33]. In addition, some drugs (such as anti-TB and ART) commonly used in the sample population are more likely to be associated with hepatotoxicity, in contrast to drugs used in developed countries.
We found that ADR patients were prescribed a higher number of medications compared to non-ADR patients. That means, increasing medical complexity, both number of co-morbidities and number of medications, were associated with an increased risk for ADR-related hospitalisations. This is clearly described in the literature [28, 30, 37].
Patients with a previous ADR history were more likely to be admitted with ADRs which is consistent with other studies [37, 38]. This might be explained by immunological reactions tending to become worse on repeated exposure due to immunologic memory or cross-reaction to alternative drugs, or because no alternative drugs are available more toxic alternatives must be used. Most of the ADRs observed were preventable. Most of the ADRs in the current study occurred due to lack of close review of patients’ previous clinical and medication-related progress. ADR prevention could have been improved with better knowledge of patients’ medical and medication history and associated risk factors.
The ADR-related hospitalisation risk prediction model demonstrated a capacity of 79.0% to discriminate patients who are at risk of ADR-related hospitalisation and those patients who are not. The sensitivity and specificity of the ADR-risk prediction model was 59.2% and 86.6%, respectively, which suggests that it can moderately rule-in patients at risk of ADRs and strongly rule-out those patients not at risk of ADRs. To our knowledge, there is no previously developed similar ADR-risk prediction model in similar study populations aged ≥18 years, especially in low and middle-income countries. Previous models developed by Zopf et al. [37] and Parameswaran et al. [57] showed comparable results (predictive abilities of 80.0% and 70.0%, respectively), but these models were developed in different populations with different clinical characteristics and drug therapy. Our model was further augmented by an ADR preventability assessment using Schumock and Thornton’s preventability assessment criteria, in which the majority of the ADRs were preventable provided these risk factors were reviewed and monitored closely. Although our ADR risk prediction model was not validated in other populations, the majority of the variables identified as independent predictors of ADRs have been described in previous studies [37, 57], suggesting that it is a useful model to assist healthcare practitioners to moderately identify patients at risk for ADRs for implementation of intervention strategies.
Limitations and strengths
Due to the cross-sectional nature of this study, we were only able to consider the incidence of ADRs at a time, so there is a need for future longitudinal studies to consider the incidence of initial and repeat events, and interventional strategies to identify root causes and reduce ADR-related burdens. The self-reported responses for previous ADR history, for instance, may be limited by recall bias and could have influenced the identified predictors of ADRs, especially in patients taking multiple medications for chronic illnesses. Some questions in the Naranjo causality assessment tool were designed for controlled clinical trials; they were not feasible nor ethical for clinical practice, such as observing effects on giving placebo. Commonly used over-the-counter medicines, contraceptives, topical agents, and herbal remedies were not typically recorded in drug histories, which may have resulted in underestimation of the rate of ADR-related admissions. The results of this study should be extrapolated to other countries with caution, as the study findings depend on the patient characteristics, disease distribution, healthcare infrastructure, detection methods and definitions of ADRs adopted.
The strength of our study is the prospective identification of ADRs immediately upon admission, allowing for accurate evaluation of the clinical presentation and laboratory parameters. We used a three-step process in ADR assessment, evaluating individual patient’s clinical and laboratory parameters, causality assessment using the Naranjo algorithm followed by consensus review with a senior supervising internist, to help mitigate the subjectivity associated with interpretation of some ADRs. Our study is the first to identify independent predictors of ADR-related hospitalisation in Ethiopian patients. Our yearlong sampling avoided the bias associated with anti-infective use pattern due to seasonal variation. Our study was conducted in a teaching and referral hospital serving a population of 15 million; therefore, our results allow extrapolation to other settings in the southwest of Ethiopia.
Conclusions
ADRs were a common cause of hospitalisation in adults admitted to medical wards of the JUSH. The majority of ADRs were preventable, highlighting the need for close monitoring and review of patients with lower BMI, previous ADR history, pre-existing renal and liver diseases, multiple comorbidities and medications. The ADR-related hospitalisation risk prediction model demonstrated some ability to identify patients at higher risk for ADRs, and clearly identify patients at lower risk of ADRs. ADR predictors should be integrated into clinical pathways and pharmacovigilance systems. However, validation and refinement of the model is necessary prior to its implementation in routine clinical practice. The prevention of incident ADR may be of paramount importance, as previous ADR was a strong predictor of subsequent events in this patient population. Assessment of ADR causality and effective use of a pharmacovigilance system to monitor drug response in patients should be considered at ambulatory care units of all health care levels to minimise the burden of admissions related to ADRs by targeting the occurrence of preventable reactions.
Supporting information
S1 Fig. Area under the receiver operator curve (AUROC) showing ADR-risk prediction capacity of the model.
https://doi.org/10.1371/journal.pone.0186631.s001
(TIF)
Acknowledgments
The authors gratefully acknowledge the study participants for their time during the interview and medical record evaluation. We also would like to thank Mr Tsegaye Melaku for his support in data collection and participation when third party opinion was sought. Sincere thanks to the clinical and administrative staff of Jimma University Specialised Hospital for their assistance and cooperation in this study.
References
- 1. Angamo MT, Chalmers L, Curtain CM, Bereznicki LR. Adverse-drug-reaction-related hospitalisations in developed and developing countries: A review of prevalence and contributing factors. Drug Saf. 2016;39 (9):847–57. pmid:27449638
- 2. Sonal Sekhar M, Adheena Mary C, Anju PG, Hamsa NA. Study on drug related hospital admissions in a tertiary care hospital in South India. Saudi Pharmaceutical Journal: SPJ: The Official Publication of the Saudi Pharmaceutical Society. 2011;19 (4):273–8. pmid:23960769
- 3. Varallo FR, Lima MFR, Galduroz JCF, Mastroianni PC. Adverse drug reaction as cause of hospital admission of elderly people: A pilot study. Latin American Journal of Pharmacy. 2011;30 (2):347–53.
- 4. Lopez AD, Mathers CD, Ezzati M, Jamison DT, Murray CJ. Global and regional burden of disease and risk factors, 2001: Systematic analysis of population health data. Lancet. 2006;367 (9524):1747–57. pmid:16731270
- 5.
United Nations Acquired Immunodeficiency Syndrome (UNAIDS) global update 2016. http://www.who.int/hiv/mediacentre/news/global-aids-update-2016-news/en/. Accessed on 24 June 2017.
- 6. Subbaraman R, Chaguturu SK, Mayer KH, Flanigan TP, Kumarasamy N. Adverse effects of highly active antiretroviral therapy in developing countries. Clinical Infectious Diseases: An Official Publication of the Infectious Diseases Society of America. 2007;45 (8):1093–101. pmid:17879931
- 7. Isah AO, Pal SN, Olsson S, Dodoo A, Bencheikh RS. Specific features of medicines safety and pharmacovigilance in Africa. Therapeutic Advances in Drug Safety. 2012;3 (1):25–34. pmid:25083223
- 8. Mehta U, Durrheim DN, Blockman M, Kredo T, Gounden R, Barnes KI. Adverse drug reactions in adult medical inpatients in a South African hospital serving a community with a high HIV/AIDS prevalence: Prospective observational study. British Journal of Clinical Pharmacology. 2008;65 (3):396–406. pmid:18070223
- 9. Dimie O, Reginald OO, Haruna MM, Mukhtar A, Aliyu B, Abdulaziz H, et al. Morbidity and mortality patterns of hospitalised adult HIV/AIDS patients in the era of highly active antiretroviral therapy: A 4-year retrospective review from Zaria, Northern Nigeria. AIDS Research and Treatment. 2012;2012:1–11. pmid:23019521
- 10. Henry NL, Marie-Solange D, Simeon-Pierre C, Elvis T, Gloria A, Henry A, et al. Adverse drug reactions of highly active antiretroviral therapy (HAART) in HIV infected patients at the General Hospital, Douala, Cameroon: A cross sectional study. The Pan African Medical Journal. 2012;12 (87):1–7.
- 11. Van Olmen J, Schellevis F, Van Damme W, Kegels G, Rasschaert F. Management of chronic diseases in sub-Saharan Africa: Cross-fertilisation between HIV/AIDS and diabetes care. J Trop Med. 2012;2012:349312. pmid:23209477
- 12. Misganaw A, Mariam DH, Ali A, Araya T. Epidemiology of major non-communicable diseases in Ethiopia: A systematic review. Journal of Health, Population, and Nutrition. 2014;32 (1):1–13. pmid:24847587
- 13. Aminkeng F, Ross CJ, Rassekh SR, Brunham LR, Sistonen J, Dube MP, et al. Higher frequency of genetic variants conferring increased risk for ADRs for commonly used drugs treating cancer, AIDS and tuberculosis in persons of African descent. The Pharmacogenomics Journal. 2014;14 (2):160–70. pmid:23588107
- 14.
Ethiopian Federal Ministry of Health HIV/AIDS Prevention and Control Office Country progress report in HIV/AIDS response. Ethiopian Federal Ministry of Health HIV/AIDS Prevention and Control Office, Addis Ababa; 2014. http://www.unaids.org/sites/default/files/country/documents/ETH_narrative_report_2014.pdf.
- 15.
World Health Organisation. Tuberculosis progress in Ethiopia. http://www.afro.who.int/en/ethiopia/country-programmes/topics/4481-tuberculosis.html. Accessed on 24 June 2017.
- 16. Wang H, Wolock TM, Carter A, Nguyen G, Kyu HH, Gakidou E, et al. Estimates of global, regional, and national incidence, prevalence, and mortality of HIV, 1980–2015: The Global Burden of Disease Study 2015. The lancet HIV. 2016;3 (8):e361–87. pmid:27470028
- 17.
World Health Organization. Noncommunicable Diseases (NCD) Country Profiles 2014. http://www.afro.who.int/en/ethiopia/country-programmes/topics/4591-ethiopia-non-communicable-diseases-ncd.html. Accessed on 10 Feb 2017.
- 18. Mamo Y, Seid E, Adams S, Gardiner A, Parry E. A primary healthcare approach to the management of chronic disease in Ethiopia: An example for other countries. Clinical Medicine (London, England). 2007;7 (3):228–31.
- 19.
Netsanet WW, Gandham NVR. Health Extension Program in Ethiopia. Universal health coverage studies series (UNICO) No.10. The World Bank, Washington DC; 2013. http://documents.worldbank.org/curated/en/356621468032070256/Ethiopia-The-health-extension-program-in-Ethiopia.
- 20. Gebretekle GB, Serbessa MK. Exploration of over the counter sales of antibiotics in community pharmacies of Addis Ababa, Ethiopia: Pharmacy professionals' perspective. Antimicrobial Resistance and Infection Control. 2016;5:2. pmid:26835006
- 21. Tegegne GT, Gaddisa T, Kefale B, Tesfaye G, Likisa J, Albachew M, et al. Drug therapy problem and contributing factors among ambulatory hypertensive patients in Ambo General Hospital, West Shoa,Ethiopia. Global Journal of Medical Research. 2015;15 (4):20–6.
- 22. Oumer S. Irrational use of medications among elderly patients in an Ethiopian referral hospital. African Journal of Pharmacy and Pharmacology. 2017;11 (15):191–4.
- 23. Pedros C, Quintana B, Rebolledo M, Porta N, Vallano A, Arnau JM. Prevalence, risk factors and main features of adverse drug reactions leading to hospital admission. European Journal of Clinical Pharmacology. 2014;70 (3):361–7. pmid:24362489
- 24. van der Hooft CS, Dieleman JP, Siemes C, Aarnoudse AJL, Verhamme K, Stricker BH, et al. Adverse drug reaction-related hospitalisations: A population-based cohort study. Pharmacoepidemiology and Drug Safety. 2008;17 (4):365–71. pmid:18302300
- 25. Ruiter R, Visser LE, Rodenburg EM, Trifiro G, Ziere G, Stricker BH. Adverse drug reaction-related hospitalizations in persons aged 55 years and over: A population-based study in the Netherlands. Drugs & Aging. 2012;29 (3):225–32. pmid:22372725
- 26. Zopf Y, Rabe C, Neubert A, Gassmann KG, Rascher W, Hahn EG, et al. Women encounter ADRs more often than do men. European Journal of Clinical Pharmacology. 2008;64 (10):999–1004. pmid:18604529
- 27. Hopf Y, Watson M, Williams D. Adverse-drug-reaction related admissions to a hospital in Scotland. Pharm World Sci. 2008;30 (6):854–62. pmid:18654836
- 28. Onder G, Pedone C, Landi F, Cesari M, Della Vedova C, Bernabei R, et al. Adverse drug reactions as cause of hospital admissions: Results from the Italian group of pharmacoepidemiology in the elderly (GIFA). Journal of the American Geriatrics Society. 2002;50 (12):1962–8. pmid:12473007
- 29. Mannesse CK, Derkx FHM, de Ridder MAJ, t Veld A, van der Cammen TJM. Contribution of adverse drug reactions to hospital admission of older patients. Age and Ageing. 2000;29 (1):35–9. pmid:10690693
- 30. Patel H, Bell D, Molokhia M, Srishanmuganathan J, Patel M, Car J, et al. Trends in hospital admissions for adverse drug reactions in England: Analysis of national hospital episode statistics 1998–2005. BMC Clinical Pharmacology. 2007;7:9. pmid:17894876
- 31. Sánchez Muñoz-Torrero JF, Barquilla P, Velasco R, Fernández Capitan MDC, Pacheco N, Vicente L, et al. Adverse drug reactions in internal medicine units and associated risk factors. European Journal of Clinical Pharmacology. 2010;66 (12):1257–64. pmid:20689943
- 32. Leendertse AJ, Egberts AC, Stoker LJ, van den Bemt PM. Frequency of and risk factors for preventable medication-related hospital admissions in the Netherlands. Archives of Internal Medicine. 2008;168 (17):1890–6. pmid:18809816
- 33. Verbeeck RK. Pharmacokinetics and dosage adjustment in patients with hepatic dysfunction. European Journal of Clinical Pharmacology. 2008;64 (12):1147–61. pmid:18762933
- 34. Chen YC, Fan JS, Hsu TF, Chen MH, Huang HH, Cheng KW, et al. Detection of patients presenting with adverse drug events in the emergency department. Internal Medicine Journal. 2012;42 (6):651–7. pmid:22188441
- 35. Zhang M, Holman CDJ, Price SD, Sanfilippo FM, Preen DB, Bulsara . Comorbidity and repeat admission to hospital for adverse drug reactions in older adults: Retrospective cohort study. British Medical Journal. 2009;338. pmid:19129307
- 36. Alexopoulou A, Dourakis SP, Mantzoukis D, Pitsariotis T, Kandyli A, Deutsch M, et al. Adverse drug reactions as a cause of hospital admissions: A 6-month experience in a single center in Greece. European Journal of Internal Medicine. 2008;19 (7):505–10. pmid:19013378
- 37. Zopf Y, Rabe C, Neubert A, Hahn EG, Dormann H. Risk factors associated with adverse drug reactions following hospital admission: A prospective analysis of 907 patients in two German university hospitals. Drug Saf. 2008;31 (9):789–98. pmid:18707193
- 38. McDonnell PJ, Jacobs MR. Hospital admissions resulting from preventable adverse drug reactions. Ann Pharmacother. 2002;36 (9):1331–6. pmid:12196047
- 39. Mouton JP, Njuguna C, Kramer N, Stewart A, Mehta U, Blockman M, et al. Adverse drug reactions causing admission to medical wards: A cross-sectional survey at 4 hospitals in South Africa. Medicine. 2016;95 (19):e3437. pmid:27175644
- 40. Olivier P, Bertrand L, Tubery M, Lauque D, Montastruc JL, Lapeyre-Mestre M. Hospitalizations because of adverse drug reactions in elderly patients admitted through the emergency department: A prospective survey. Drugs & Aging. 2009;26 (6):475–82. pmid:19591522
- 41. Tipping B, Kalula S, Badri M. The burden and risk factors for adverse drug events in older patients: A prospective cross-sectional study. South African Medical Journal = Suid-Afrikaanse Tydskrif Vir Geneeskunde. 2006;96 (12):1255–9. pmid:17252155
- 42.
Jimma University Specialised Hospital. https://www.ju.edu.et/jimma-university-specialized-hospital-jush. Accessed on 28 December 2016.
- 43.
World Health Organisation. Safety of Medicine: A guide to detecting and reporting adverse drug reaction. World Health Organisation, Geneva; 2002. http://apps.who.int/medicinedocs/en/d/Jh2992e/.
- 44. Getahun W, Gedif T, Tesfaye F. Regular Khat (Catha edulis) chewing is associated with elevated diastolic blood pressure among adults in Butajira, Ethiopia: A comparative study. BMC Public Health. 2010;10:390. pmid:20594361
- 45. Pahor M, Chrischilles EA, Guralnik JM, Brown SL, Wallace RB, Carbonin P. Drug data coding and analysis in epidemiologic studies. European Journal of Epidemiology. 1994;10 (4):405–11. pmid:7843344
- 46.
Classification Committee of the World Organization of Family Doctors (WICC) 1998. ICPC-2: International Classification of Primary Care. 2nd edn. Oxford, Oxford University Press.
- 47. Wawruch M, Zikavska M, Wsolova L, Kuzelova M, Kahayova K, Strateny K, et al. Adverse drug reactions related to hospital admission in Slovak elderly patients. Archives of Gerontology and Geriatrics. 2009;48 (2):186–90. pmid:18313773
- 48. Bushardt RL, Massey EB, Simpson TW, Ariail JC, Simpson KN. Polypharmacy: Misleading, but manageable. Clinical Interventions in Aging. 2008;3 (2):383–9. pmid:18686760
- 49.
Department of Health and Human Services. National Heart, Lung, and Blood Institute. www.nhLbi.nih.gov/health/public/heart/obesity. Accessed on 20 June 2017.
- 50. Levey AS, Coresh J. Chronic kidney disease. Lancet. 2012;379 (9811):165–80. pmid:21840587
- 51. Reuben A. Hy's law. Hepatology. 2004;39 (2):574–8. pmid:14768020
- 52.
World Health Organisation. Treatment of Tuberculosis: Guidelines for national programmes. 4th edition. World Health Organisation; 2010. http://www.who.int/tb/publications/tb_treatmentguidelines/en/.
- 53.
British National Formulary 67 (BNF). Joint formulary committee, Pharmaceutical Press; 2014.
- 54.
Up-To-Date 19.3. http://freemedicalresources.blogspot.com.au/2012/08/use-uptodate-database-offline-with-your.html.
- 55. Naranjo CA, Busto U, Sellers EM, Sandor P, Ruiz I, Roberts EA, et al. A method for estimating the probability of adverse drug reactions. Clinical Pharmacology and Therapeutics. 1981;30 (2):239–45. pmid:7249508
- 56. Gurwitz JH, Field TS, Avorn J, McCormick D, Jain S, Eckler M, et al. Incidence and preventability of adverse drug events in nursing homes. The American Journal of Medicine. 2000;109 (2):87–94. pmid:10967148
- 57. Parameswaran Nair N, Chalmers L, Connolly M, Bereznicki BJ, Peterson GM, Curtain C, et al. Prediction of hospitalization due to adverse drug reactions in elderly community-dwelling patients (The PADR-EC Score). PLoS One. 2016;11 (10):e0165757. pmid:27798708
- 58. Schumock GT, Thornton JP. Focusing on the preventability of adverse drug reactions. Hospital Pharmacy. 1992;27 (6):538. pmid:10118597
- 59.
Rawlins MD, Thompson JW. Pathogenesis of adverse drug reactions. In: Davies DM, ed. Textbook of adverse drug reactions, 2nd edn. Oxford, Oxford University Press;1981.pp.11.
- 60. Robin X, Turck N, Hainard A, Tiberti N, Lisacek F, Sanchez JC, et al. pROC: An open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinformatics. 2011;12:77. pmid:21414208
- 61. Aziz Z, Siang TC, Badarudin NS. Reporting of adverse drug reactions: Predictors of under-reporting in Malaysia. Pharmacoepidemiology and Drug Safety. 2007;16 (2):223–8. pmid:16947117
- 62. Rehan HS, Vasudev K, Tripathi CD. Adverse drug reaction monitoring: Knowledge, attitude and practices of medical students and prescribers. The National Medical Journal of India. 2002;15 (1):24–6. pmid:11855588
- 63. Belachew T. Are we ready for the rising silent epidemic of metabolic syndrome and chronic non-communicable disease in Ethiopia? Ethiopian Journal of Health Sciences. 2015;25(1):1–2. pmid:25733778
- 64.
Acquired immunodeficiency syndrome (AIDS) resource centre. http://www.etharc.org/index.php/resources/healthstat/hivaids-estimates-and-projections-in-ethiopia-2011-2016. Accessed on 09 January 2017
- 65. Gedle D, Gelaw B, Muluye D, Mesele M. Prevalence of malnutrition and its associated factors among adult people living with HIV/AIDS receiving anti-retroviral therapy at Butajira Hospital, Southern Ethiopia. BMC Nutrition. 2015;1 (1):5.
- 66. Dargie B, Tesfaye G, Worku A. Prevalence and associated factors of undernutrition among adult tuberculosis patients in some selected public health facilities of Addis Ababa, Ethiopia: A cross-sectional study. BMC Nutrition. 2016;2v(1):7.
- 67. Alvarez PA, Bril F, Castro V, Meiville I, Gonzalez CD, Centurion IG, et al. Adverse drug reactions as a reason for admission to an internal medicine ward in Argentina. The International Journal of Risk & Safety in Medicine. 2013;25 (3):185–92. pmid:24047689
- 68. Patel KJ, Kedia MS, Bajpai D, Mehta SS, Kshirsagar NA, Gogtay NJ. Evaluation of the prevalence and economic burden of adverse drug reactions presenting to the medical emergency department of a tertiary referral centre: A prospective study. BMC Clinical Pharmacology. 2007;7:8. pmid:17662147
- 69. Yimer G, Ueda N, Habtewold A, Amogne W, Suda A, Riedel KD, et al. Pharmacogenetic & pharmacokinetic biomarker for efavirenz based ARV and rifampicin based anti-TB drug induced liver injury in TB-HIV infected patients. PLoS One. 2011;6 (12):e27810. pmid:22162992
- 70. Sharif-Askari FS, Syed Sulaiman SA, Saheb Sharif-Askari N, Al Sayed Hussain A. Development of an adverse drug reaction risk assessment score among hospitalized patients with chronic kidney disease. PLoS One. 2014;9 (4):e95991. pmid:24755778
- 71. Caamano F, Pedone C, Zuccala G, Carbonin P. Socio-demographic factors related to the prevalence of adverse drug reaction at hospital admission in an elderly population. Archives of Gerontology and Geriatrics. 2005;40 (1):45–52. pmid:15531022
- 72. Hoigne R, Sollberger J, Zoppi M, Muller U, Hess T, Fritschy D, et al. Significance of age, sex, kidney function, atopy and number of prescriptions for the occurrence of adverse drug reactions, studied by multivariate statistical methods. Results from the comprehensive hospital drug monitoring berne (CHDMB). Schweizerische Medizinische Wochenschrift. 1984;114 (49):1854–7. pmid:6515386
- 73. Hassen AA, Belachew T, Yami A, Ayen WY. Anti-tuberculosis drug induced hepatotoxicity among TB/HIV co-infected patients at jimma university hospital, ethiopia: Nested case-control study. Plos One. 2013;8 (5):e64622. pmid:23696901
- 74. Corsonello A, Onder G, Bustacchini S, Provinciali M, Garasto S, Gareri P, et al. Estimating renal function to reduce the risk of adverse drug reactions. Drug Saf. 2012;35 Suppl 1:47–54. pmid:23446785
- 75. Zeeh J, Platt D. The aging liver: Structural and functional changes and their consequences for drug treatment in old age. Gerontology. 2002;48 (3):121–7 pmid:11961363
- 76. Manley HJ, McClaran ML, Overbay DK, Wright MA, Reid GM, Bender WL, et al. Factors associated with medication-related problems in ambulatory hemodialysis patients. American Journal of Kidney Diseases: The Official Journal of the National Kidney Foundation. 2003;41(2):386–93. pmid:12552501