For a variety of reasons, substantial percentages of at-risk patients don't continue to take medications prescribed to help them managed their chronic conditions. Ideally, the VA's electronic health care records could be used to efficiently identify those that stop taking medications for reasons that could be addressed by health care providers.
Develop and evaluate a computer program to screen VA electronic health care records and identify patients as they become past due for resupply of a prescription medication. Specific aims were to estimate the percentage of true and false positive screens, to determine the patients' reasons for not continuing to get the medication from the VA, and to estimate the screening program's yield of patients that potentially could benefit from efforts to continue use of the medication.
The screening program was evaluated using a cohort of all veterans who were currently getting prescription medications known as statins from the Minneapolis VA Health Care System. Statins are commonly prescribed to reduce a patient's high cholesterol, hence the risk of having a heart attack or stroke. Prescription files (VISTA) were screened biweekly starting in February 2010 with 9 months of prior prescription records. Eligible individuals had received at least 2 separate 30-day statin supplies with no record of death in the VISTA patient or patient treatment files. A 'past due date' was calculated as the date of release of the last VA statin supply plus the days of supply dispensed, any carry over from the previous supplies, any inpatient days and a 120-day grace period. If a patient was past due when the screening program was run using local health care records, similar records maintained at other VA health care systems were searched. Addresses were checked to exclude residents of long-term care facilities. Surveys were mailed to the first 1000 that became past due to ask whether they had stopped taking the statin they had received from the VA, and, if they did, why? After a 9-month lag, administrative records at the Austin Information Technology Center (outpatient pharmacy and visits, inpatient, enrollment) were searched to determine true and false positive screens defined as those that didn't or did get another VA statin supply before their calculated past due date. Several baseline (year prior to getting their last statin supply) variables were extracted including patient characteristics, the number and types of prescriptions each patient received, and recorded diagnoses. Monthly enrollment files were searched for deaths among the positive screens.
Within 5 months, 1000 patients amounting to 4.6% of the 21,935 in the statin cohort being monitored became past due. Subsequent examination of statin supplies provided by the VA indicated 824 (3.8%) were true positives (positive predictive value 824/1000; 82%). Follow-up of the 824 true positives indicated 11% were deceased (n=95) or nursing home residents (n=17). Excluding the deceased, nursing home residents and undeliverable addresses (n=21), the follow-up survey response rate was 786/867 (91%). Of the 640 true positives, 30% (n=302) reported reasons for ceasing to get statins from the VA that most likely would not be amenable to reinstating use of the VA statin including having a non-VA supply (n=178), physician orders, side effects or use alternative treatments. An additional 22% of the true positives received another statin supply from the VA within 6 months after their past due date. For undetermined reasons, the screening program missed statin supplies dispensed to 176 patients in February 2010, i.e. 18% of the 1000 that screened positive were false positives. Thus, the overall estimated yield of potential candidates for efforts to reinstate statin use was only 18% of the 1000 positive screens or 0.8% of the screened cohort.
Allocation of limited health care resources to this type of mass screening program would have to be justified by the magnitude of improvement in patient outcomes that would depend on the yield as well as the reasons patients discontinue the medication and whether they could be satisfactorily addressed by health care providers. This prospective electronic screening of the VA health care records quickly identified a large number of ostensible statin dropouts that would require follow-up. However, most did not appear to be good candidates for remedial action. Unwarranted follow-up could be reduced by more up to date records of deaths and transitions to long-term care facilities. A more efficient screening program would search for and exclude those with orders to discontinue the medication, documented side effects, initiation of alternative treatments, and laboratory results consistent with a side effect or adequate cholesterol control despite discontinuation of the statin provided by the VA. A more sophisticated screening program may require electronic searches of provider notes. A more efficient screening program would need to recognize medication supplies obtained outside the VA. Perhaps the process of medication reconciliation could provide readily accessible information on discontinued medications including the reason. The positive screens reported a variety of potentially addressable reasons for stopping their statin such as not being convinced the medication was worthwhile and wanting to lower their out-of-pocket costs for medicines. However, these reasons may not be documented in health care records and require patient follow-up. The method used in this evaluation to get a good response rate to the follow-up patient survey that included a $5 payment, sending and tracking as many as three mailings per patient most likely would not be practical for routine screening of a large medication cohort. .
- Rector TS, Nugent S, Spoont M, Noorbaloochi S, Bloomfield HE. Screening electronic veterans' health records for medication discontinuation. The American journal of managed care. 2012 Jul 1; 18(7):352-8.
- Rector TS. Screening Electronic Prescription Records for Dropouts. [Cyberseminar]. 2011 Nov 15.