Running decision support logic retrospectively to determine guideline adherence: a case study with diabetes

We describe the application of our previously developed analytical infrastructure called RetroGuide to conduct studies on retrospective data using a clinician friendly flowchart paradigm. Introduction Requiring additional clinician?s input in a new decision support system (DSS) is often a major implementation obstacle. Another limitation is the process of fine tuning the exact logic of the new DSS, which is often done in the production environment after piloting an initial version and gathering clinicians? feedback. Our approach was to utilize only currently available Electronic Health Record data (EHR) and beta test the logic and impact of a potential decision support module using only retrospective data. The main objective of this study is to demonstrate the use of a suite of medical informatics tools called RetroGuide [1] for such an effort. As a case study, the problem of blood pressure control in diabetic patients and adherence to the JNC7 hypertension guideline is presented. Methods Intermountain Healthcare?s Enterprise Data Warehouse (EDW) was used as the source of coded lifetime EHRs. RetroGuide was used to create graphical scenarios representing a temporal sequence of events of interest associated with analytical questions. RetroGuide uses a workflow engine to execute the modeled scenario on real retrospective data and produces a series of reports about the scenario execution. Type II diabetic patients with at least 2 available blood pressure measurements and some additional criteria were included in the study (194 subjects total). Results A simplified version of the executable flowchart is shown in Figure 1. The scenario logic included some temporal restrictions on the considered blood pressure values as shown in Figure 1. The variables inside brackets show that the engine can remember the timestamp or value associated with a previously identified event. According to the JNC7 guideline the systolic blood pressure value should always be maintained less than 140 mmHg. RetroGuide output showed that 15.9% of study patients had both values over 140 (31/194; 95% CI 10.4 - 21.4). We investigated 27 additional related clinical questions in a more complex scenario. Discussion: In contrast with a comparable study also investigating blood pressure control in diabetics [2] our methodology enables easy integration of additional temporal restrictions on the considered blood pressure values or other events of interest. Additional advantages of RetroGuide when compared to traditional SQL-based database tools are: (a) a user-friendly flowchart model as a shared logic formalism between the DSS developer and clinicians; and (b) support for extensive ?drill-down? capability into available EHR data via a hierarchy of customizable reports. We have also initiated a controlled study to evaluate the suitability of RetroGuide?s methodology to lower the skill barrier for ?champion? clinicians to run analyses on data in large EDWs. The value of our graphical tool will increase as more detail is stored in EHR, longer time-span is available and its temporal aspects are better.