PS2-21: Extracting Signs and Symptoms from Colonoscopy Reports Using NLP
Clinical Medicine & Research
Background/Aims: Adherence to certain effective and widely prescribed cardiovascular (CVD) medications is poor. The purpose of this analysis is to demonstrate the effective implementation of a cost-effective and easily disseminable population-level HIT adherence intervention. Methods: PATIENT is a pragmatic trial designed to improve adherence to antihypertensives and statins in high-risk patients in three Kaiser Permanente (KP) regions. The 1-year intervention (Dec. 2011-2012) enrolled 21,752
... ) enrolled 21,752 adults, aged 40-80, with diabetes or CVD and at least 1 dispensing of a target medication in the past year (based on pharmacy dispensing data). Participants were randomized to one of 3 arms: usual care (UC), automated telephone reminders (IVR), or an enhanced IVR arm (IVR+) that added mailed educational materials, reminder mailings and live outreach calls to IVR calls. Results: Through the first 10 months of intervention we attempted 42,010 automated reminder calls (˜ 2.9 per intervention participant). Of these, 56% (23,596) of calls successfully reached the target participant and an additional 29% (12,271) resulted in messages left. These figures were higher for calls made to those nearly due for a refill (65% and 26% of 20,182) than for calls made to those overdue (48% and 32% of 21,828). Of calls where we reached the participant, 30% resulted in a transfer to the regional pharmacy automated refill line and an additional 2.5% resulted in transfers to a live pharmacist. Among the 7,258 IVR+ patients, 3,748 medication reminder letters were mailed, 2,146 live reminder calls were made, and we have sent 36,823 educational mailings (˜4.9 per participant). Key implementation challenges that emerged related to 1) complexities in coding the intervention call flag algorithm from the EMR and 2) customization of the intervention to fit regional work-flow and support program sustainability. Close partnering with pharmacy staff was crucial in overcoming early hurdles. Outcome data (not yet available) will be presented at the meeting. Conclusions: We successfully implemented a large, multi-modal, HIT, medication adherence intervention at 3 KP regions. A high proportion of members were reached and actively participated in the intervention. Key challenges/solutions are detailed in a manual of operations to facilitate the implementation of similar interventions in other settings.