Study setting, period and design
A cross sectional study was conducted at the University of Gondar Comprehensive Specialized Hospital (UoGCSH), Felege-Hiwot Comprehensive Specialized Hospital (FHCSH), and Tibebe-Ghion Comprehensive Specialized Hospital (TGCSH) from June 1 to August 30 2022. UoGCSH is located in Gondar a city 730 kilometers Northwest from Addis Ababa (the capital city of Ethiopia). UoGCSH has served a total of 12,000 patients with psychiatric disorders annually. The psychiatry unit of UoGCSH contains two beds in the emergency room, nineteen beds for inpatient care, and four outpatient departments. Whereas FHCSH and TGCSH are located in Bahir Dar city, 521 kilometers away from Addis Ababa. Felege-hiwot comprehensive specialized hospital provides service for 19,200 patients with psychiatric disorders every year and it has seven inpatient beds and six outpatient sections. Tibebe-gihwon comprehensive specialized hospital has one emergency and two inpatient rooms with 6 beds.
Population Inclusion And Exclusion Criteria
All adult patients with schizophrenia having a regular follow up at the outpatient departments’ psychiatric clinic in the comprehensive specialized hospitals of northwest Ethiopia were the source population. All adult patients with schizophrenia having a regular follow up at the outpatient departments of UoGCSH, FHCSH and TGCSH during the study period were the study populations. Patients with the age of 18 years and above, taking antipsychotic medication, who had an insight to respond oral questions (satisfy the requirement in the insight assessment tool (get 3 out of 3), and patients who had one or more previous visits were included under the study. Patients who had incomplete medical record were exclude from the study.
Sample Size Determination
The sample size was calculated for APP using a single population proportion formula as follows:\(n = Z 2 P (1-p)/w2\)
where n is the desired sample size for a population of > 10,000, Z is the typical normal distribution set at 1.96 (which corresponds to 95% CI), the p-value signifies that positive prevalence was utilized in calculating the optimal sample size, and W is the degree of accuracy 0.05 required (a marginal error is 0.05). From the previous study done on the prevalence of APP, it was estimated to be 28.2% (5). Then computing for n = 1.962*0.282(1–0.282)/0.052; n = 308; by adding 10% non-respondents the estimated sample size was 339. However, for the factors associated with APP, taking AOR from previous literature (5), it was calculated as duration of treatment, 5–10 years; outcome exposed to unexposed (18.68%); computing with AOR gave 429.5. For duration of treatment > 10 years, outcome exposed to unexposed (18.68%); computing with AOR, gave 284. Finally, for the number of admissions, outcome exposed to unexposed (22.41%); computing with AOR, gave 126. So from the above sample size calculation, the largest calculated (429.5 ≈ 430) sample size was used.
Sampling Technique And Procedure
The comprehensive specialized hospitals in the northwestern Ethiopia were selected by a random lottery method. The total number of schizophrenia patients on follow up within 3 months were taken from the patients’ registration document to allocate samples proportionally within study areas. After proportional allocation, a systematic random sampling technique was used to select the study participants. The sampling fraction (k) was calculated by dividing the total number of schizophrenic patients in the study area by the total sample size; (2625/430 gives 6.1 ≈ 6). The starting point was selected randomly from 1 to 6. Then, participants were interviewed, and concurrently, relevant data was reviewed from medical charts for every sixth patient until the requirement for a sample was fulfilled. A unique patient identification card number was utilized as a questionnaire code in order to prevent the inclusion of the same patient in the study more than once. (Fig. 1: Proportional allocation of sample size of schizophrenia out patients (n = 422)
Study Variables
Antipsychotic polypharmacy was the dependent variable. The independent variables were the sociodemographic characteristics of the participants like marital status, residence, educational status, occupation and monthly income), and patient related conditions include number of admission, presence of comorbidity, duration of illness, duration of treatment and substance use
Operational Definitions
Antipsychotic polypharmacy refers to the co-prescription of more than one antipsychotic drug for a particular patient for at list 1 month and above (5, 11).
Substance use (current users) refers using at least one of a specific substance (alcohol, Khat or cigarettes) for nonmedical purposes within the last 3 months according to the Alcohol, Smoking, and Substance Involvement Screening Tool (ASSIST) (12).
Data Collection Instrument, Procedures, Quality Control
A structured questionnaire was adopted from a previous literature (5) with some modifications for the context of the study area and socio-demographic characteristics of study participants was used. It was translated to the local Amharic language, and then back translated to the English version to check consistency. Translation was not required for variables obtained from medical records of the patients. The data collected by patient interview includes sociodemographic characteristics and substance use. The patients’ medical charts was used to fill in clinically related variables like duration of illness, duration of treatment, number of admission, presence of comorbidity, type of antipsychotics, and presence of antipsychotics polypharmacy. The data collection tool had three parts. The first part contains socio-demographic characteristics of the study participants, such as sex, age, marital status, residence, religion, educational level, occupation, and income level. The second section consisted of clinical and medication related characteristics like duration of illness, presence of comorbidity, patient’s medication record, duration of treatment, and number of admissions. The third section consisted of the current substance use assessment tool. ASSIST was used to briefly screen patients' use of psychoactive substances. The tool was developed and validated by WHO (12).
Data was collected by face-to-face interview using a structured and pretested questionnaire. Data was collected by three bachelor psychiatrists in UoGCSH with supervision by one master’s degree in psychiatry, three bachelor psychiatrists in FHCSH with supervision by one bachelor psychiatrist, and two bachelor psychiatrists in TGCSH with supervision by one bachelor psychiatrist. The supervisor distributed all the necessary items for data collectors on each data collection day and was tasked with checking the filled questionnaire for completeness and solving reported problems timely during the data collection period. In all study areas, the principal investigator facilitated all the necessary materials.
To assure the quality of the data, one day training was given by the principal investigator at each study area for data collectors and supervisors. A pretest was conducted on 22 of the 5% of schizophrenia patients at Dessie Comprehensive Specialized Hospital's outpatient department to identify potential problems with the data collection tool and check the consistency of the questionnaire. Some modifications, such as correction of typing errors and the rearranging of questionnaires were made. The internal consistency of ASSIST was assessed and the Cronbach’s alpha was 0.76.
Data Entry And Analysis
Collected data was cleaned, coded, and entered into Epi Data 4.6.0 and analyzed using Statistical Package for Social Studies (SPSS) version 24. In descriptive analysis, the mean with standard deviation (SD), frequency, and percentages were used to check the distribution of the data. Bivariable and multivariable binary logistic regression analysis were employed to identify factors associated with APP. The odds ratio (OR) with a 95% confidence interval was computed for each variable for the corresponding p-value to see the strength of association. A P-value of < 0.05 was used as the cut-off for the significance of the association between the dependent and the independent variables. The model fitness was tested, and the Hosmer and Lemeshow test result was 0.797. Multicollinearity was checked, and the maximum Variance Inflation Factor (VIF) reported was less than 5, which was within the acceptable level