Study design and setting
An institution-based cross-sectional study was conducted in Dire Dawa City Administration in Ethiopia from 10 June to 20 July 2020. Dire Dawa is located at 515 Kilometres, East of Addis Ababa, Capital of Ethiopia. Based on the 2007 Central Statistical Agency population census, the total population of Dire Dawa Administration was projected to be 342, 827 in 2019 [20]. According to the Dire Dawa Adminstration Health Bureau annual report of 2019, there were two hospitals, 15 health centers, and 34 health posts with a total of 1080 health workers [21].
Population and sampling
All health professionals in public health facilities in Dire Dawa were the source population. Health professionals in randomly selected public health facilities in Dire Dawa were the study population. Critically sick health professionals and those who were on leave of any type or long-term training or education or field activities during the data collection period were excluded from the study.
The sample size was calculated with Epi Info version 7.1 considering the assumption for single (HMIS utilization) and double (for factors) population proportion formulas. Accordingly, maximum sample size of 379 was obtained using a single population proportion formula considering 78.5% proportion of good routine health information utilization [16], confidence level of 95%, margin of error of 5%, design effect of 2, non-response of 10%, and a total population of 1080.
We applied a two-stage stratified sampling technique to select the study participants. First, public health facilities were stratified into hospitals and health centers, and then, one out of two hospitals and six out of 15 health centers were randomly selected. Then, the sample size was proportionally allocated to each selected facility based on the actual numbers of registered health professionals serving in the selected health facility during the last six months before the interview. We prepared a separate sampling frame for each health facility using their actual numbers of (permanently) hired health professionals in 2020 and recruited participants using a simple random sampling technique.
Data collection tools and measurements
Pretested-structured questionnaire adapted from PRISM framework assessment tool version 3.1 [7] and published literature [13, 14, 16, 19, 22-28] were used to collect data from participants through a self-administered interview conducted over a month. The questionnaire contains information on sociodemographic characteristics of the participants, technical factors, organizational factors, behavioral factors, and the level of utilization of routine health information (Supplementary Table 1).
Before starting analysis, internal consistencies of items were checked for each domain scale of dependent and independent variables using reliability analysis (Cronbach α). Accordingly, we checked for the internal consistency of each domain of PRISM framework tool for assessing organizational, technical and behavioral determinants of RHIS utilization and computed summary statistics, mean±SD, minimum and maximum scores, and standard error. We observed high internal consistency across all domains with the minimum in attitude toward RHIS items (Cronbach’s α=0.72) and the maximum in self-efficacy items (Cronbach’s α=0.97) (Supplementary Table 2).
Utilization of routine health information system (RHIS): The use of information for improving effectiveness and efficiency of healthcare services through better management at all levels [7]. Utilization of RHIS was measured using 13 five-points Likert Scale items each rated from '1' (strongly disagree) to '5' (strongly agree) and then, composite index score was computed from 13 items and the level of RHIS utilization was 'good' when the participants scored above the mean and 'poor' unless otherwise [16, 17, 24].
Organizational support: It was measured using 11 five-points Likert Scale items each rated from '1' (strongly disagree) to '5' (strongly agree) and then, composite index score was computed from 11 items, and the level of perceived organizational support was 'good' when the participants scored "quartile 3 and above", 'fair' when scored between "quartile 3 and quartile 1" (excluding both) and 'poor' unless otherwise [7, 16].
Training: Considered 'yes' when the participants received a short-term education or orientation on health management information system and its related reporting tools and procedures according to the national health information system training manual on how to use data elements, indicators, and their definitions in the last two years and 'no' unless otherwise.
Training adequacy: It was measured using seven five-points Likert Scale items each rated from '1' (strongly disagree) to '5' (strongly agree) and then, composite index score was computed from seven items and the training adequacy was 'good' when the participant scored "quartile 3 and above", 'fair' when scored between "quartile 3 and quartile 1" (excluding both) and 'poor' otherwise [7, 16, 17].
Supervision: Considered 'yes' when the participant had taken health information system-specific or related supportive follow up aimed at enabling providers to perform RHIS properly through providing on-job training and technical support in the last six months and 'no' otherwise [7, 25].
Supervision quality: It was measured using five dichotomous items each coded '1' when the participants responded "right answer" and '0' when responded "the wrong answer" and then, composite index score was computed from five items and the level of supervision quality was 'good' when the participant scored "quartile 3 and above", 'fair' when scored between "quartile 3 and quartile 1" (excluding both) and 'poor' unless otherwise [7, 17].
Technical support: It was measured using seven five-points Likert Scale items each rated from '1' (strongly disagree) to '5' (strongly agree) and composite index score was computed from seven items, and the technical support was 'good' when the participant scored "quartile 3 and above", 'fair' when scored between "quartile 3 and quartile 1" (excluding both) and 'poor' otherwise [7, 17].
The perceived complexity of RHIS formats: It was measured using four dichotomous (yes/no) questions each coded '1' when the participant responded 'yes' (i.e. complex) and '0' when responded 'no' (i.e. not complex) and then, composite index score was computed from four items and level of the perceived complexity of RHIS formats was "not complex" when the participant scored "quartile 1 and below", "fairly complex" when scored between "quartile 1 and quartile 3" (excluding both) and 'complex' when scored between "quartile 3 and above " [7, 17].
Self-efficacy: It was measured using seven 10-points Likert Scale items each rated from '0' (no competency) to '10' (best competency) and composite index score was computed from seven items and the self-efficacy was 'good' when the participant scored "quartile 3 and above", 'fair' when scored between "quartile 3 and quartile 1" (excluding both) and 'poor' unless otherwise [7, 17].
Attitude toward RHIS: It was measured using six five-point Likert Scale items each rated from '1' (strongly disagree) to '5' (strongly agree) and composite index score was computed from six items, and health professional’s attitude, was 'good' when participant scored "quartile 3 and above", 'fair' when between "quartile 3 and quartile 1" (excluding both) and 'poor' unless otherwise [7, 14, 16].
Decision-making autonomy: It was measured using six five-points Likert Scale items each rated from '1' (strongly disagree) to '5' (strongly agree) and composite index score was computed from six items and decision-making autonomy was ‘good’ when the participants scored "quartile 3 and above", ‘fair’ when scored between "quartile 3 and quartile 1" (excluding both) and ‘poor’ unless otherwise [7, 16, 17, 24].
Data quality control
To maintain the data quality, standard questionnaires adapted from validated scales and published literature were contextualized to the study purpose and setting. Six trained nurses collected the data, and two experts holding BSc degree in Public Health was supervised the overall data collection process with the principal investigator. We also pretested a questionnaire on 19 health professionals (5% of the total sample) to check its validity in Melka Jabdu Health Center (a non-selected health facility in Dire Dawa, Ethiopia). Epi-Data was used for data entry to minimize the potential errors that could occur during entry. During data collection, strict supervision of data collectors and validation of collected data was carried out by supervisors and investigators.
Statistical analysis
After checking for completeness, data were entered using EpiData version 3.1 and analyzed using Stata version 16.0. Descriptive statistics such as frequency, measures of central tendency, and dispersion were used to characterize the participants. Bivariable and multivariable binary logistic regression analyses were conducted to identify factors associated with good utilization of routine health information. Independent variables with P-value <0.25 during our bivariable analysis were considered in our multivariable analysis model. The overall model adeaucy was confirmed using Hosmer and Lemeshow goodness of fit test at P-value>0.05. We rulled out and confirmed the absence of numerical errors and multicolliniearity in the model (each predictor has standardized resiedul of less than absolute value of three and cook’s distance of less than absolute value of one and their standard errors of each cofficient (β) was less than two). Adjusted Odds Ratio(AOR) with a 95% confidence interval was used to report association and significance declared at P-value<0.05.