Search

Much of the methodological literature on rapid qualitative analysis describes processes used by a relatively small number of researchers focusing on one study site and using rapid analysis to replace a traditional analytical approach. In this paper, we describe the experiences of a transnational research consortium integrating both rapid and traditional qualitative analysis approaches to develop social theory while also informing program design. Research was conducted by the Innovations for Choice and Autonomy (ICAN) consortium, which seeks to understand how self-injection of the contraceptive subcutaneous depot medroxyprogesterone acetate (DMPA-SC) can be implemented in a way that best meets women's needs, as defined by women themselves. Consortium members are based in Kenya, Uganda, Malawi, Nigeria, and the United States. Data for the ICAN study was collected in all four countries in sub-Saharan Africa. In order to both illuminate social phenomena across study sites and inform the program design component of the study, researchers developed tools meant to gather both in-depth information about women's contraceptive decision-making and data targeted specifically to program design during the formative qualitative phase of the study. Using these two bodies of data, researchers then simultaneously conducted both a traditional qualitative and rapid analysis to meet multiple study objectives. To complete the traditional analysis, researchers coded interview transcripts and kept analytical memos, while also drawing on data collected by tools developed for the rapid analysis. Rapid analysis consisted of simultaneously collecting data and reviewing notes developed specifically for this analysis. We conclude that integrating traditional and rapid qualitative analysis enabled us to meet the needs of a complex transnational study with the added benefit of grounding our program design work in more robust primary data than normally is available for studies using a human-centered design approach to intervention development. However, the realities of conducting a multi-faceted study across multiple countries and contexts made truly “rapid” analysis challenging.

Identifying patients with a low likelihood of paying their bill serves the needs of patients and providers alike: aligning government programs with their target beneficiaries while minimizing patient frustration and reducing waste among emergency physicians by streamlining the billing process. The goal of this study was to predict the likelihood of patients paying the balance of their emergency department visit bill within 90 days of receipt. Three machine learning methodologies were applied to predict payment: logistic regression, decision tree, and random forest. Models were trained and performance was measured using 1,055,941 patients with non-zero balances across 27 EDs from 1 August 2015 to 31 July 2017. The decision tree accurately predicted 87% of unsuccessful payments, providing significant opportunities to identify patients in need of financial assistance.