Effectiveness of Electronic Health Record-Based Interventions for Improving Follow-Up in Primary Care

The goal of this proposal is to demonstrate and test methods by which large health care systems can efficiently identify cancer patients who are more likely to experience diagnostic delays and pre-emptively rectify these delays. This study will build upon tools developed in our recent work (Aim1, prior IRB Protocol Number: H-23801) and test their effectiveness to identify patients at risk of experiencing delays in cancer diagnosis followed by an intervention that the investigators hypothesize will reduce these delays. This is Aim 2 (for which the investigators are seeking approval) is the final Aim of this proposal. Aim 1 was approved under Protocol Number: H-23801. In Aim 2 the investigators will determine the effectiveness of an IT-based intervention (consisting of data mining using triggers tested in Aim 1 followed by targeted electronic communication and surveillance techniques) to facilitate cancer diagnosis as compared with usual care (no use of trigger or electronic communication and surveillance). Hypothesis 1: The time from first appearance of a diagnostic clue to follow-up action (e.g. colonoscopy performance after a positive FOBT) will be significantly less in the intervention arm than in usual care. Hypothesis 2: The percentage of patients receiving timely follow-up care will be significantly more in the intervention arm than in usual care. To improve the generalizability of our findings to multiple ambulatory care environments, the investigators will conduct our research in two settings: an urban Veterans Affairs facility in Houston, Texas and a large primary care network in central Texas. These settings include internal medicine and family medicine, academic and nonacademic practices, and significant racial, gender, ethnic, age, urban/rural, and socioeconomic diversity. Our study addresses coordination and timeliness of care, both of which are priorities to achieve high quality care. Hypothesis 3: Overall, the trigger will achieve a positive predictive value (PPV) of at least 50% in identifying delays in care. PPV is defined as the number of charts correctly identified with a delay in diagnostic evaluation, divided by the total number of charts identified by the trigger, and was deemed to be the approximately level necessary to avoid substantial contribution to provider alert fatigue.