Adverse drug reactions and its associated factors among geriatric hospitalized patients at selected comprehensive specialized hospitals of the Amhara Region, Ethiopia: a multicenter prospective cohort study | BMC Geriatrics

Study design, and period

The study design was a multicenter prospective cohort study of geriatric patients admitted to the medical wards of four selected hospitals in the Amhara regional state. The study period was from May 2023 to August 2023. The study participant was followed until they were discharged from the hospital to provide appropriate therapies for ADR alleviation and to monitor them during that time.

Study setting

This research was carried out in public health institutions located in the Amhara regional state of Ethiopia. The Amhara Region is situated in the northwestern and central parts of Ethiopia, positioned between latitudes 9° and 23° 45’N and longitudes 36° and 40° 30’E. Elevation varies from 700 m in the eastern regions to over 4620 m in the northwestern area. The total area is 170,000 km², organized into 11 administrative zones and 105 Woredas. This region contains eight comprehensive specialized hospitals. Felege Hiwot, University of Gonder, Debere Tabor, and Dessie Comprehensive Specialized Hospitals were randomly selected as data sources for the study.

Inclusion and exclusion criteria

Patients admitted to the selected hospitals during the study period who were 65 years of age or older, geriatric patients who stay in the hospital for a minimum of 24 h or longer, providing enough time to monitor and identify ADRs, patients in the hospital have had at least one prescription filled out. This makes sure that the study’s emphasis is on patients who are at risk of ADRs as a result of medication and who give their written, informed consent to take part in the research were included in the study. Unless a caregiver or legal representative is available to assist, patients with severe cognitive impairment or dementia are unable to provide informed permission or participate consistently in follow-up, and patients who do not take medicine throughout their hospital stay and who are admitted for reasons unrelated to medication (such as surgery or trauma) were not included.

Sample size determination and sampling technique

The sample size was calculated based on a single proportional formula n =\(\:\frac{\:\left({\text{Z}{\alpha\:}/2)}^{2}\text{P}\right(1-\text{P})}{{d}^{2}}\); Z = 1.96, the proportion of ADE occurrence (P) = 0.5 and marginal error (d) = 5%, then the sample size is equal to 384. Including a 10% contingency for patients who declined to participate in the study and non-respondents led to a final sample size of 422. We allocated a portion of the total sample size to each comprehensive hospital based on the patients admitted within the previous three months. Proportional allocation of samples to the total population of each hospital was applied using the formula as follows: n = nf×N/ni

Where n = required sample size for each hospital, nf = patients admitted in the previous three months at each hospital, N = sample size calculated from a single proportional formula and ni = total number of geriatric patients from four selected hospitals who were admitted in the previous three months. The source population and the samples were N = 809 and n = 422, respectively.

Study participants at FHCSH = 230*422/809 = 120 (17 were excluded during follow-up).

Study participants at UGCSH = 201*422/809 = 105 (12 were excluded during follow-up).

Study participants at DTCSH = 167*422/809 = 87 (9 were excluded during follow-up).

Study participants at DCSH = 211*422/809 = 110 (11 were excluded during follow-up).

Finally, 373 patients were used for analysis after 49 patients were eliminated because they were unable to complete their follow-up (Fig. 1). We applied a consecutive sampling technique among all patients who met the inclusion criteria.

Flow chart of enrolments and exclusion among geriatric patients from May 2023 to August 2023

Study variable

The incidence of adverse drug reactions is the dependent variable for this study. The study examined factors such as age, gender, education level, marital status, occupation, body mass index (BMI), GFR (Modification of Diet in Renal Disease [23], drug source, social drug use (such as alcohol, khat, and smoking), drug class, number of prescribed drugs, type of medical condition, number of comorbidities, and length of hospitalization to determine the incidence of adverse drug reactions.

Data collection instrument, procedures, and quality assurance

The data abstraction tool was developed by reviewing the literature for important variables [24,25,26,27,28,29]. An interdisciplinary team validates and examines the assessment tools to ensure the accuracy and dependability of the results. The assessment tools were assessed by one epidemiologist and two MSc. in clinical pharmacy. We first created the data collection sheet in English, translated it into the local language of Amharic, and then translated it back into English to maintain consistency. We used 10% of randomly chosen sample patients in a pretest to evaluate the tool’s quality before actual data gathering. Before it was put to use for data collection, a few adjustments were performed.

Not only does the questionnaire cover sociodemographic information such as age, gender, residence, marital status, educational attainment, occupation, alcohol use, and cigarette smoking, but it also covers clinical and related factors such as past medical history and current diagnosis, comorbidities, complications, hospitalization history within the last six months, and ADRs history. Along with the Schumock and Thornton scale, Hartwig’s Severity Assessment Scale Naranjo Causation Scale, and the intervention tool were used.

Prior to starting work, data collectors were trained on the goals of the study, the data collection checklist, and how to recognize and record adverse drug reactions to ensure data accuracy. Under the supervision of four senior clinical pharmacists (MSc. in clinical pharmacy), four clinical pharmacists with bachelor’s degrees in pharmacy were to collect the data. Data was gathered through patient interviews, in-person observations, and a review of their lab, prescription, and medical records. Every day, the completed data collection forms were reviewed and monitored. Doctors and nurses were informed of any changes to medication management.

Outcome measures and ADRs detection

The main finding of this study is the prevalence of ADRs. The secondary outcome focuses on ADRs’ severity, preventability, and outcome, as well as their causal link to medications. Medical and pharmaceutical records, laboratory inspections, patient interviews, and direct observation were used to identify ADRs. Standard instruments have been used to evaluate the study’s ADRs’ severity, causative connection, and preventability.

To assess the causality of ADRs, the modified Naranjo Causation Scale was employed. The alternatives for ten of the questions on this tool are “yes,” “no,” and “don’t know.” Four different point values (-1, 0, + 1, or + 2) are assigned to each response. Response was considered after each participant’s points were added up to a total score that might be anywhere between − 4 and + 13. A score of nine or more was considered “definite,” five to eight was considered “possible,” one to four was considered “possible,” and zero or less was considered “doubtful” [25]. Using Hartwig’s Severity Assessment Scale, which comprises seven questions, the severity of the present ADRs was evaluated and categorized as mild (levels 1 and 2), moderate (levels 3 and 4), and severe (levels 5 and above) [26]. The precise criteria developed by Schumock and Tornton were applied to determine the preventability of the ADRs.

Schumock and Tornton’s criteria were employed to determine the preventability of the ADRs. The tool comprises three sections: non-preventable (part III), possibly preventable (part II), and preventable (part I). There are five questions categorized as preventable, four as possibly preventable, and one as non-preventable. Each response was divided into yes and no categories. When answering one or more of the par I questions are true, and then adverse medication occurrences were definitely preventable. The assessors will move on to part II if every response is negative. When answering one or more of the part II questions is yes, then the adverse medication event is likely avoidable. Part III moved forward if all of the responses were negative which is not preventable [24]. Reports indicate the organ most commonly affected by adverse drug reactions and the drugs most frequently associated with these reactions. The incidence of adverse drug reation per 100 admissions was calculated by dividing the total number of ADRs identified by the total number of admissions and then multiplying by 100.

Pharmacists play a part in the management of patients in geriatrics to maximize pharmacotherapy. They are responsible for making sure that the medications they give patients are appropriate and safe for usage. After examining various literatures [30,31,32], we created an intervention tool and tailored it to our research objectives and study subjects.

Data processing and analysis procedures

Every day, the data’s completeness was verified. All patient data that had been gathered was entered into Epi Data version 7.1 and exported to STATA version 14.1 for cleaning and analysis respectively. Categorical data were represented by frequency and percentages, while continuous variables were represented by the mean (standard deviation) or median (interquartile range, IQR), depending on the kind of distribution. After an evaluation of the Hosmer-Leme show goodness-of-fit test, a logistic regression model was used. Using bivariate logistic regression analysis, potential variables for multivariate logistic regression analysis were identified. Thus, covariates having a p-value < 0.25 were added to a multivariate logistic regression in order to identify statistically significant predictors of ADR frequency. A P-value of less than 0.050 was then used to assess statistical significance.