Has general practitioner computing made a difference to patient care? A systematic review of published reportsBMJ 1995; 311 doi: http://dx.doi.org/10.1136/bmj.311.7009.848 (Published 30 September 1995) Cite this as: BMJ 1995;311:848
- Correspondence to: Dr Sullivan.
- Accepted 30 August 1995
Objective: To review findings from studies of the influence of desktop computers on primary care consultations.
Design: Systematic review of world reports from 1984 to 1994.
Setting: The computerised catalogues of Medline, BIDS, and GPlit were searched, as well as conference proceedings, books, bibliographies, and references in books and journal articles.
Subjects: 30 papers met the inclusion criteria and were included for detailed review.
Interventions: A validated scheme for assessing methodological adequacy was used to score each paper.
Main outcome measures: Papers were rated on sample formation, baseline differences, unit of allocation, outcome measures, and follow up. Differences in outcomes were also recorded.
Results: Four of the six papers dealing with the consultation process showed that consultations took longer. Doctor initiated and “medical” content of consultations increased at the expense of a reduction in patient initiated and “social” content. Each of the 21 studies which looked at clinician performance showed an improvement when a computer was used (from 8% to 50%, with better results for single preventive measures). Only one of the three studies looking at patient outcomes showed an improvement (diastolic blood pressure control 5 mm Hg better after one year, with fewer doctor-patient consultations).
Conclusions: Using a computer in the consultation may help improve clinician performance but may increase the length of the consultation. More studies are needed to assess the effects on patient outcomes of using a computer in consultations.
Uing a computer during a consultation lengthens the consultation time by 48-90 seconds
Use of a computer during consultations improves immunisation rates by 8-18% and other preventive tasks by up to 50%
Using a computer during consultations does not seem to have appreciable impact on patient satisfaction
More work is needed to assess the effects of primary care computing on patient outcomes
Almost 90% of general practices in Britain are now computerised, many using computers to carry out clerical tasks and repeat prescribing. In addition, 55% of general practitioners use desktop computers to access clinical data during consultations.1 Indeed, as general practitioner computing is “an integral part of the NHS IT [information technology] strategy,”2 the government currently spends around £47m on primary care computing each year (NHS Management Executives, personal communications). In the United States computing is seen as “an essential technology for health care.”3
Computers can help with the care of individual patients through clinical decision support and with the care of groups of patients through strategic decision support. This help can operate at several levels--via access to scientific publications,4 provision of guidelines and protocols,5 prompting for missing information,6 and structured knowledge based systems.7
It is often considered axiomatic that the more structured the information system the better the care. However, computers are less likely to be of value in the loosely ordered world of general practice, where people present with a wide variety of undifferentiated problems. This makes evaluating the impact of computers in consultations even more important for primary care, in which intuitive responses may be as valid as more structured management.8
Another issue is that in order to improve outcome, possibly computer systems should develop a specifically clinical orientation allowing them to store and generate primarily medical data. However, “the design of many existing electronic medical records derives, implicitly or explicitly, from support for the use of aggregated data for research, audit, finance or planning,”9 producing a tool possibly structured more for information management than for clinical management.
This review concentrates on the use of computers by clinicians in terms of the effects on consultations, and on patient care in particular, rather than any advantages for administration or research, though both may influence patient care indirectly. The concern that “expensive computing systems are developed and installed in health care institutions without sufficient informed clinical improvement”10 exists in all aspects of medicine. This review, however, is concerned exclusively with primary care.
A SYSTEMATIC REVIEW OF PUBLISHED WORK 1984-94
A worldwide review of published work was conducted and prospective studies selected if (a) they concerned doctors or nurses in a primary care setting and (b) they described any computing system designed for use by a doctor, either in routine clinical practice or for a specific research project. The aspects examined effects on the consultation process, on doctors' task performance, and on patient outcomes.
We searched the computerised databases of Medline, BIDS (which accesses the science, social science, and arts and humanities citation indexes), and GPlit (the primary care subset of the biomedical databases) by using “computers in medicine,” “primary care,” “family practice,” and “medical informatics” as the search terms. We also reviewed books,11 12 13 bibliographies,14 15 16 17 18 19 and conference proceedings of related topics20 as well as citations in these books and articles and references provided by colleagues. We excluded studies on aspects such as attitudes, accuracy, and completeness of data; comparisons with consultant letters; and epidemiological studies.
CRITERIA FOR EVALUATING STUDIES
Johnson et al reviewed the impact of clinical decision support systems on clinician performance and patient outcomes and proposed a scheme for assessing the methodological adequacy of studies on the impact of computers.15 In their system random allocation to study groups is rated more highly, as it reduces bias. Baseline differences between groups should be eliminated or adjusted for, and the unit of allocation to groups should be the practice in order to minimise the Hawthorne effect (the beneficial effect of participation in research). Measures of outcome should be objective and follow up should be as complete as possible (table I).
Thirty evaluations of computers in primary care met the criteria for review. Six examined the effects on the consultation process21 22 23 24 25 26 (see table II), 21 evaluated effects on clinicians' performance of tasks21 22 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 (see table III), and only three measured the impact on patient outcome47 48 49 50 (see table IV). Two studies examined more than one topic (consultation process and doctors' performance)21 22 and are included in both relevant tables (II and III). By means of the system described, each paper was reviewed and scored by each of us separately. Twenty two of the studies were scored identically. Differences in scores for the others were discussed and an agreed score reached. All 30 studies, including those with low scores, were incorporated to emphasise how little rigorous evaluation of computers in primary care has actually been carried out.
EFFECTS ON CONSULTATION PROCESS
Studies of the effects of computers on the consultation process were concerned mainly with the length of the consultations and the activities included (table II). The studies contained only few doctors (range one to six), indicating the difficulty of assessing the content of consultations. Three studies showed that consultations were 48-54 seconds longer when a computer was used.21 22 23 This difference was mainly due to tasks involving the computer. Doctor initiated and “medical” content of the consultations increased at the expense of a reduction in patient initiated and “social” content. Only one study tried to observe the longer term impact of introducing a computer to consultations.25 After 30 months it found that consultations were on average 90 seconds longer--10 minutes as compared with 8 1/2 minutes for controls. Only one study found no change in the content of the consultations and used a subjective measure to detect differences in the “standard of care attained.”26
CLINICIAN PERFORMANCE STUDIES
Studies of the effects of computers on clinician performance were the most numerous and were concerned with preventive care, clinical tasks, screening, and repeat prescribing. Many used a more robust methodology, including patient follow up. The emphasis was on immunisation and other preventive tasks (14 studies) and on prescribing (four), fewer studies being concerned with the management of disease (diabetes, one study; hypertension, one study). Only one study examined the performance of doctors in recording presenting symptoms and in generating problem lists.46 Most of the improvements were in the positive direction (table III). Immunisation rates improved by 8-18%27 28 29 and other preventive tasks performed improved by up to 50%.21 25 30 31 32 33 34 35 36 37 38 The biggest improvements were noted when single rather than repeated measurements were performed.
Results were better when studies concerned more deprived patient populations,35 emphasising the potential for opportunistic case finding to reverse the “inverse care law”39 when supported by an adequate information infrastructure. Consultation based prompting could work only for attenders. Letters or telephone contacts, usually by a nurse,28 34 35 were more effective strategies for those who rarely visited. Tierney et al used a randomised block design and showed that clinical decision support was marginally better than strategic decision support.31 They also showed that there was no additive effect when both were employed.
Early studies of prescribing confirmed the anticipated time savings for doctors and receptionists, which probably persuaded most practices to buy computers in the first place.40 Further studies showed that more generic prescribing is encouraged as electronic formularies are adopted, which partly explains the 13-30% reduction in prescribing costs reported.41 42 43 However, the few practices concerned in prescribing studies makes their generalisability less certain.
Both studies examining process measures of chronic disease management suggested that improvements are encouraged.44 45 The study by Brownbridge et al on hypertension44 also examined a paper protocol, so that the effects of computerisation by itself were difficult to disentangle. The remaining study46 indicated that a computer can encourage more complete data capture of aspects of consultations which doctors consider important. However, doctors are more resistant to recording data which they consider less essential.46
PATIENT OUTCOME STUDIES
Only three studies could be classified as examining patient outcome (table IV). The first concerned doctors completing an encounter form for each patient seen. The data were entered in a remote clinical program and feedback was returned to the study doctors. Though the critique provided was remote from the consultation which produced the data, the doctors nevertheless changed their management of patients in the light of suggestions made (one in three consultations). An average reduction of 5 mm Hg in diastolic blood pressure was recorded for moderately hypertensive patients despite four visits a year fewer than controls.47 It is reasonable to expect that “patient specific reminders at the time of consultation” would be even more successful.48 The remaining two studies, which used patient satisfaction as an outcome measure, failed to detect any appreciable change.49 50
Despite the major cost of computers to the health service, systematic review of published work yielded only 30 papers evaluating their effects on the consultation process, clinician performance, and patient outcomes. Most of these studies indicated a neutral or positive effect when a computer was used. However, every study of the introduction of computing evaluates more than simply a new information system. The reason for any observed effect can be hard to disentangle from the effects of other changes that may occur at the same time--for example, greater teamwork, redefining working relations, and consultation with outside resources and training.
Many of the papers highlighted the clinical aspects of decision support in a research setting. However, in many cases it is the more strategic approach which is measured by outcome studies. Their emphasis on the more easily quantified aspects of performance reflects the current reality of available technology. Current systems are poorly placed to support the provision of feedback for strategy planning. Only 24% can audit the clinical content of a patient review and only 52% can audit prescribing activity.51 Seventy nine per cent of systems cannot perform any statistical analysis and 76% have no graphical ability. The capacity to export data to third party software is also often limited.52
The studies reviewed provide evidence that using a computer in a consultation may lengthen that consultation by as much as 90 seconds. Patient initiated and social content may be reduced, though this may be offset by increased clinical performance on the part of the physician. The focus has been on the usefulness of computers in highly structured tasks. These may be laid open to computerisation, but surely the rich interaction of the consultation cannot be replaced by a computer. As Bleich et al noted, “Any doctor who could be replaced by a computer deserves to be.”53
Most of the studies reviewed assessed the effects of computers on the clinician, but future research should centre on outcomes of care for patients. This is problematical because of the wide ranging tasks in primary care consultations, including prevention, current health problems, and public health issues. It is also difficult to determine the link between the process of care and outcomes for patients when clinical problems are diffuse.
The way forward would be randomised controlled trials to evaluate the benefits and drawbacks of existing computer use in consultations for clinicians, support staff, and patients. However, over half of the general practitioners in Britain already have desktop computers and the remainder could be described as laggards who may be reluctant to follow suit. Therefore, we need to look at new methods of evaluating these major changes in “the essential unit of medical practice”54 such as quasiexperimental and pragmatic trials. The introduction of new aspects of information technology (such as Medline access, Cochrane databases,55 and decision support systems) should also be examined.
It is five years since Mike Pringle challenged the suppliers of general practitioner computing systems to work with clinicians to improve the quality of patient care.56 Only by clearly documenting the successes, failures, and lessons learnt will computers enable general practitioners “cum technologica caritas.”
We thank Sue Ross and Professor John Howie for their comments, Bill Dodd (NHS Management Executive, Leeds) for his cooperation, and Michele Beaumont for secretarial work.
Funding EM is funded by the Clinical Resource and Audit Group of the Scottish Office Home and Health Department.
Conflict of interest None.