Helping Researchers Improve Health Outcomes: D4I’s How-To Videos on Advanced Data Analysis As part of its mission to improve global health outcomes, Data for Impact (D4I) has created two video tutorials to strengthen individuals’ capacity to analyze routine data for program evaluations. Routine data can include patient data collected in hospital charts, community information systems, and disease-specific registers and health service specific registers at health facilities (e.g., family planning services, antenatal care services, sick childcare services). These videos are designed to increase viewers’ skills and confidence in using two popular methods for analyzing routine data: Interrupted Time Series Analysis (ITS) and Logistic Regression (LR). These methods were selected based on the results of a recent D4I activity that focused on creating practical guidance for using routine data in evaluation. The activity included a literature review identifying these methods as the most commonly used. The tutorials are aimed at practitioners who have some statistical knowledge and experience using Stata, including researchers, students, and local evaluators. The videos’ brevity is noteworthy, as “many people incorrectly believe they need to take long courses to be able to use many statistical techniques,” says D4I’s Eva Silvestre, Partner Team Lead at Tulane University. After participating in an internal meeting wherein presenters discussed Stata commands used in their analysis, Silvestre thought, “that is really useful,” and got to work spearheading the development of these videos. Silvestre and her team created the tutorials alongside a professor of statistics in Ghana to help ensure accuracy and usefulness. Working with anonymous data modified from actual D4I projects, learners get the chance to see real data in action. In the videos, a facilitator poses example research questions (e.g., “What was the impact of the policy on the utilization of antenatal care visits on pregnant women?”) then presents possible answers, including step-by-step instructions on how to analyze and interpret the data. Results interpretation is critical, helping connect the analysis to program and policy improvement and ultimately, health outcomes. For both tutorials, Silvestre notes, “we were sure to add research questions that could be answered with routine data.” Sample data sets, .do files, and background materials are provided for each method. The three-part ITS video helps viewers understand how their time-dependent data should be structured, how to run commands, and how to interpret the analysis results. ITS can be used when a researcher needs to analyze social statistics like maternal and neonatal outcomes, and malaria incidence rates, among other time-dependent outcomes. The LR video focuses on one of the most common methods for statistical analysis of categorical data. This specialized form of regression can be used with a categorical outcome variable. Viewers who work in healthcare and the social sciences will likely be familiar with this technique as it can be used to predict the probability of disease outcomes. Check out the tutorials here, and let us know how they impact your team’s work.
