What is the evidence that giving healthcare professionals better tools makes them
perform the complicated tasks they do better? A systematic review [1] of computer
based clinical decision support systems (CDSS) shows two things. It demonstrates that
there are many studies in a wide variety of different clinical areas. It shows that
they work.
Search
Studies using CDSS in a clinical setting by a healthcare practitioner and assessing
the effects in a prospective setting with a concurrent control were sought. Five
databases, reference lists and conference proceedings were searched.
Results
Sixty-eight studies were found (Table 1). Of 15 studies of drug dosing systems, 60%
found benefit. Of 19 studies on preventive care systems, 74% found benefit. Of 26
studies in other clinical areas, 73% found benefit. Only one of five diagnostic
decision support systems found benefit; that used a system to identify postoperative
patients at risk of respiratory complications for physiotherapy.
|
Table 1: Healthcare professional performance in studies of
CDSS
|
|
Study area
|
Studies showing benefit/total
|
| Drug dosing systems |
9/15 |
| Diagnostic aids |
1/5 |
| Preventive care systems |
14/19 |
| Other medical care |
19/26 |
| All CDSS studies |
43/65 |
Comment
The systematic review is a terrific bit of work. But the question still hanging
is whether computer systems can contribute significantly to reduce adverse drug
events in hospitals. Two US examples show exactly what can be achieved [2,3].
Boston
In the Brigham and Women's Hospital, which is a 726-bed tertiary referral
centre, the use of a physician computer order-entry (POE) system was evaluated,
in which doctors wrote all drug orders online. The study had a baseline period
during which an audit of medication errors was examined, followed by
implementation of the POE system and re-audit. Incidents were identified by three
mechanisms: nurses and pharmacists reported incidents, an investigator visited
wards twice daily to solicit information, and patient charts were examined daily
by an investigator.
The main outcome was the number of nonintercepted serious medication errors.
These were either an error preventable by systems currently in use, or had the
potential for harm but did not result in injury.
Results
Use of the POE system prevented more than half of the serious medication errors.
There were just under 11 of these per 1000 patient days at baseline, and under 5
per 1000 patient days during use of the POE system. Potential errors which had
not been intercepted fell most, by 84%. Preventable errors fell by 17%.
The authors concluded that their system could be extended to different drug
types, like sedatives, which actually rose, which had not been included in their
original system, and by extending the system in other ways. They also show that
the cost of running a POE system for their large, complicated, hospital, would be
of the same order as money saved directly. When other costs, like extra work
caused by serious drug errors, or malpractice litigation, were included, it could
save $5-10 million a year. The system could both save money and improve quality
of care.
Phoenix
The Good Samaritan Regional Medical Centre in Phoenix is a 650-bed referral
centre. It has an integrated hospital information system. A multidiciplinary team
of professionals met and identified 37 drug or class-specific areas where adverse
drug events might be expected. The system was modified so that if circumstances
arose where an adverse drug event might occur (digoxin toxicity was one example),
then a pharmacist or radiologist was alerted. If necessary, the physician
attending the patient was contacted.
Results
Over six months there were 9306 non-obstetric admissions. There were 1116 alerts
(Table 2). Physicians needed to be contacted 794 times, and 596 times the event
had not been recognised. The average time taken for each contact was 15 minutes.
|
|
Table 2: Adverse drug event alerts
|
|
Event
|
Number
|
per 1000 admissions
|
Percent
|
| Nonobstetric admissions |
9306 |
|
|
| ADE alerts |
1116 |
120 |
100 |
| Evaluator needed to alert physician |
794 |
85 |
71 |
| True potential adverse drug events |
596 |
64 |
53 |
| Physicians unaware of potential for harm |
265 |
28 |
24 |
| Changes in treatment |
265 |
28 |
24 |
The rates of clinically unrecognised events varied for different clinical
circumstances. For instance, more than half of the potential problems for renal
toxicity with the use of radiocontrast media had been previously recognised, but
it was felt that potential benefit outweighed potential harm.
Using some literature data on costs, the authors calculated that the potential
saving to their 650-bed hospital was some $3 million a year, and could be more if
the system were extended to other areas.
Comment
These are two different types of interventions. One depends on putting systems
in place to stop mistakes happening. The other depends on real-time interventions
to stop mistakes when they happen. Both had a major effect in stopping medication
errors in large, complex institutions. Both would improve patient care. Both
would reduce costs.
Concentrating on stopping bad things happening is what quality control is all
about. These are two excellent examples of how to do it. Moreover, all three
papers have extensive referencing of a wealth of literature in this area, and are
worth reading for that alone.
Reference:
- DL Hunt, RB Haynes, SE Hanna, K Smith. Effects of computer-based clinical
decision support systems on physician performance and patient outcomes. JAMA
1998 280: 1339-1346.
- DW Bates et al. Effect of computerized physician order entry and a team
intervention on prevention of serious medical errors. JAMA 1998 280:
1311-1316.
- RA Raschke et al. A computer alert system to prevent injury from adverse
drug events. Development and evaluation in a community teaching hospital. JAMA
1998 280: 1317-1320.
|
previous
or
next
story in this issue
