August 21, 2012
More on Missed Followup of Tests in Hospital
The problem of significant abnormal test results “slipping through the cracks” is a serious cause for preventable adverse patient outcomes and a leading cause of malpractice settlements. For many years we have stressed the importance of including “test results pending” in discharge summaries and in verbal handoffs as well (see our March 1, 2011 Patient Safety Tip of the Week “Tests Pending at Discharge” and the list of other columns below).
In our March 9, 2010 Patient Safety Tip of the Week “Were et al 2009) which showed that only 16% of tests with results pending actually are documented in discharge summaries.” we noted a study (
A new study (Ong 2012) further quantifies some of the risks of tests ordered late in the course of a hospitalization. They reviewed clinical pathology tests ordered on inpatients at a large teaching hospital in Australia. Overall 37% of patients had at least one test whose results were not reviewed at discharge. The percent of tests not reviewed varied dramatically by when they were ordered. While the overall rate of test results not reviewed was 3% at discharge and 1.5% two months after discharge, 47% of all missed test results occurred with tests ordered on the day of discharge. 21% of tests ordered on the day of discharge were not followed up, compared to only 1.8% ordered on other days. Moreover, 14.7% of unreviewed tests at discharge were abnormal. Of those still not reviewed by 2 months after discharge 10.8% had abnormal results.
Our March 1, 2011 Patient Safety Tip of the Week “Tests Pending at Discharge” noted a study (Callen 2011) that found lack of followup on test results ranged from 20-60% for inpatients and 1-75% for emergency room patients. And the lack of followup had clinical implications since some of the results were considered critical or urgent or otherwise actionable.
So there is little doubt that failure to follow up on test results during such transitions of care is both common and potentially dangerous.
So what are the solutions? When we actually sat down to dictate discharge summaries and tried to identify these tests that had been done but not yet resulted we were astonished at how difficult it was to identify those tests! Even as lengths of stay plummet it is almost impossible to wade through either a paper chart or an electronic medical record and easily identify what tests were ordered, then look up the results and identify those lacking results. However, you can write computer scripts in the electronic medical record that will do that for you. So at one hospital we work with the IT staff developed 2 solutions: (1) a screen that shows “tests done but not yet resulted” and (2) a screen that shows physicians “test results coming in after discharge” for recently discharged patients. Hospitalists or any physicians managing inpatients typically go to the first screen when doing a discharge summary and can then include those tests in the discharge summary to alert the person providing the next level of care to check those results. That screen is patient-specific. The other screen is available any time the physician logs on and lists all that physician’s recently discharged patients so the physician can click on a patient name and see results that have come in after discharge. Then (hopefully) they convey those results to the next caregiver in the continuum who needs those results to manage the patient on an ongoing basis.
One of the things we found after implementation was that physicians were often deluged with long lists of results, many of which were not considered particularly important. So the risk of “alert fatigue” reared its ugly head. We had to pare back the lists and prioritize the tests listed to include the more important ones. (Anecdotally, seeing the longer lists initially probably did help physicians recognize that many of the tests they ordered during a hospital stay probably were of little value and we did see a slight decline in inpatient lab tests.) Note also we had to do the same paring and prioritizing when we began tracking test results for patients seen in the ambulatory clinics.
Our system required the physician to log onto the computer system to see the “test results coming in after discharge” list. Another group (Dalal 2012) designed and implemented an automated system that notified attending physicians by email when those test results came in. They also sent a carbon copy to the primary care physicians. They, too, found they had to suppress selected tests to avoid too many email notifications. Ultimately they delivered an average of 1.6 email notifications per discharged patient. An overwhelming majority (84%) of inpatient attending physicians in that study were satisfied with the system. Note that this stands in sharp contrast to their first attempt at a computer solution (Dalal 2011) that identified numerous workflow issues and other barriers to adoption.
So we know that IT solutions can be helpful. But there remain some concerns. One concern we always have is related to the transfer of responsibility for care of the patients. We’ve noted on numerous occasions the study (Singh 2009) that demonstrated notification of more than one provider actually doubled the likelihood of no one responding. It needs to be made clear who will be responsible for following up on those pending test results. We always recommend that be discussed in a phone call at the transition of care but that is not always possible.
Secondly, the issue of attribution of a patient to a physician is harder than it sounds. Tests may originally be ordered in the ER prior to the decision to admit. In such cases the ER physician’s name may be attached to such tests. The patient then gets admitted under one attending or hospitalist but those physicians may also change during the course of a hospitalization. Yet tests they ordered may be attributed to them. The situation gets even more complicated in academic medical centers where multiple housestaff are involved in care. Ultimately you want all test results to come back to the attending of record at the time of discharge so that he/she can identify pending tests results and note them in the discharge summary. The previously noted study by Were et al. (Were et al 2009) also identified multiple changes in attendings as an issue and difficulty identifying the physician who will ultimately follow the patient after discharge.
While technological solutions are likely to be of benefit, to date there remains scant evidence that they have had a significant impact on resolution of this problem. The Callen paper notes that having an online endorsement or acknowledgement feature would at least help organizations better track how important test results are being communicated.
Suffice it to say that the ideal system for ensuring followup of all test results remains elusive.
See also our other columns on communicating significant results:
Were MC, Li X, Kesterson J, et al. Adequacy of Hospital Discharge Summaries in Documenting Tests with Pending Results and Outpatient Follow-up Providers. Journal of General Internal Medicine 2009; 24(9): 1002-1006
Ong M-S, Magrabi F, Jones G, Coiera E. Last Orders: Follow-up of Tests Ordered on the Day of Hospital Discharge (Research Letter). Arch Intern Med 2012; published online first August 13, 2012
Callen J, Georgiou A, Li J, Westbrook JI. The safety implications of missed test results for hospitalised patients: a systematic review. BMJ Qual Saf 2011; 20: 194-199 Published Online First: 7 February 2011
Dalal AK, Schnipper JL, Poon EG, et al. Design and implementation of an automated email notification system for results of tests pending at discharge. J Am Med Inform Assoc 2012; 19(4): 523-528
Dalal AK, Poon EG, Karson AS, et al. Lessons learned from implementation of a computerized application for pending tests at hospital discharge. Journal of Hospital Medicine 2011; 6(1): 16-21 (first published online November 15, 2011)
Singh H, Thomas EJ, Mani S, et al. Timely Follow-up of Abnormal Diagnostic Imaging Test Results in an Outpatient Setting. Arch Intern Med. 2009; 169(17): 1578-1586.