Randomised trial of personalised computer based information for cancer patientsBMJ 1999; 319 doi: http://dx.doi.org/10.1136/bmj.319.7219.1241 (Published 06 November 1999) Cite this as: BMJ 1999;319:1241
- Ray Jones, senior lecturer in health informatics ()a,
- Janne Pearson, research assistanta,
- Sandra McGregor, research assistanta,
- Alison J Cawsey, lecturer in computer scienceb,
- Ann Barrett, professor of radiation oncologyc,
- Neil Craig, lecturer in health economicsa,
- Jacqueline M Atkinson, senior lecturer in behavioural sciencesa,
- W Harper Gilmour, senior lecturer in medical statisticsa,
- Jim McEwen, professor of public healtha
- a Department of Public Health, University of Glasgow, Glasgow G12 8RZ
- b Department of Computer Science, Heriot Watt University, Edinburgh EH14 4AS
- c Beatson Oncology Centre Western Infirmary, Glasgow G11 6NT
- Correspondence to: R Jones
- Accepted 11 October 1999
Objective: To compare the use and effect of a computer based information system for cancer patients that is personalised using each patient's medical record with a system providing only general information and with information provided in booklets.
Design: Randomised trial with three groups. Data collected at start of radiotherapy, one week later (when information provided), three weeks later, and three months later.
Participants: 525 patients started radical radiotherapy; 438 completed follow up.
Interventions: Two groups were offered information via computer (personalised or general information, or both) with open access to computer thereafter; the third group was offered a selection of information booklets.
Outcomes: Patients' views and preferences, use of computer and information, and psychological status; doctors' perceptions; cost of interventions.
Results: More patients offered the personalised information said that they had learnt something new, thought the information was relevant, used the computer again, and showed their computer printouts to others. There were no major differences in doctors' perceptions of patients. More of the general computer group were anxious at three months. With an electronic patient record system, in the long run the personalised information system would cost no more than the general system. Full access to booklets cost twice as much as the general system.
Conclusions: Patients preferred computer systems that provided information from their medical records to systems that just provided general information. This has implications for the design and implementation of electronic patient record systems and reliance on general sources of patient information.
Most cancer patients want as much information as possible and wish to be involved in treatment decisions.1–3 Some argue that tailoring information is important to meet patients' different backgrounds.4 Computer based methods can be used to tailor information to patients,5–10 but no major randomised trials have examined the outcome of tailoring information to cancer patients. The importance of the electronic patient record has been recognised in the NHS information strategy.11 If using medical records to tailor information for patients is worth while then it has implications for the design and implementation of electronic patient records and patients' use of computer based resources such as the internet.
Our primary aim in this study was to compare patients' use and satisfaction, doctors' perceptions, and the costs of a system providing information for patients that was personalised using the medical record with more general computer based information. Subsidiary aims were to compare the effect of providing such personalised information with that of providing conventional information booklets and to assess the impact of providing information on patients' psychological status. Although many information booklets are produced, in practice these are not freely available in hospitals; therefore, computer based information may provide a cost effective alternative. Too much technical information 12 13 and, some argue, access to medical records14 may increase anxiety, whereas appropriate information may reduce it.15 We therefore measured these effects.
Participants and methods
Study population and sample
The Beatson Oncology Centre provides specialised non-surgical cancer treatment for patients throughout western Scotland. Between August 1996 and December 1997, 1261 eligible patients with breast, cervical, prostate, or laryngeal cancer were identified from radiotherapy booking sheets (fig 1) Patients receiving palliative treatment, with no knowledge of their diagnosis, with visual or mental handicap, or with severe pain or symptoms were excluded from the study. We obtained ethical approval for the study from the Western Ethics Committee.
Recruitment and randomisation
Eligible patients were sent a letter describing the study and were contacted when they attended the centre, within three days of their starting treatment At this contact, we gave further details of the study and assessed the patients' eligibility. We randomly allocated 525 patients who agreed to enter the study to one of three intervention groups (For further details of recruitment and randomisation, see appendix 1 on the BMJ's website.)
Computer based information
Two groups were offered a “computer consultation” using a touch screen computer (fig 2).
General information group—Patients were offered a system giving general information about cancer organised as a hypertext document.
Personal information group—Patients were offered a system that allowed them to see a summary of their medical record, and from there (via hypertext links) to information about all the concepts and terms mentioned in the record (such as “grade II” being linked to information about what this meant). Half of the personal group was also given access to the general system menu, so that we could measure more directly which system was preferred and used first.
Patients in both personal and general information groups were sent printouts of the information they had viewed. After the intervention, the patients had open access to the same information system via another computer sited in a waiting area.
Booklet information group
Patients were invited to look through a folder of printed booklets and to take as manyas they wished. There were folders for each type of cancer containing 47, 32, 34 and 30 booklets for breast, cervical, prostate, and laryngeal cancers respectively.
(For further details of computer systems and booklets, see appendix 2 on the BMJ's website.)
At recruitment or shortly after, patients were asked about the information they had already been given, what newspapers they read, their use of technology, and what information sources they used.16 They completed a hospital anxiety and depression scale17 and mental adjustment to cancer questionnaire.18
The intervention took place up to a week after recruitment. The computer systems automatically recorded the time patients spent using the computer and the choices they made, both at the time of their computer consultation and afterwards. For the booklet information group, we recorded the number and type of booklets chosen. Patients' views on the information were obtained from a home questionnaire after the intervention.
At their consultation with a doctor, three weeks after intervention, the patients completed a second hospital anxiety and depression scale and mental adjustment to cancer questionnaire. The doctors assessed the patient's participation, anxiety, knowledge, and time spent in consultation and compared these results with those of the “average” patient with that type of cancer.
Three months after the intervention, patients completed a third hospital anxiety and depression scale and mental adjustment to cancer questionnaire, and they were asked about their information preferences and their use of the printed material at home.
We assessed differences in patients' views, doctors' assessments, use of information, and cost of intervention using cross tabulations and χ2 tests. We assessed differences between intervention groups in scores on anxiety and depression scales and mental adjustment questionnaires using cross tabulations and χ2 tests, Student's t tests, and Mann-Whitney U tests.
In a subsidiary analysis using multiple logistic regression analysis, we examined differences in patients' anxiety and depression scores by their cancer type, time since diagnosis, age, sex, deprivation category,10 newspapers read, and use of information in the intervention.
We calculated current time costs in maintaining both computer information systems, using the cost of a research assistant's salary. We recorded the costs of the booklets taken by each patient. Capital computer costs were written off over four years, with maintenance charges of 5% for years 2-4. Costs incurred by patients were not included as the interventions took place during visits for treatment. We modelled four different scenarios and compared their four year cost profiles using a 6% discount rate.
Patients completing the study
Of the 715 patients invited to participate, we recruited 525; 190 (27%) refused to take part (fig 1) Probability of refusal increased with age. Of the 525 patients recruited, 467 continued to the three month follow up and 438 of these returned questionnaires. The 87 who did not complete follow up were more likely to be in the general computer information group (23% v 13%, χ2=8.1 (1 df), P=0.004), to not have breast cancer (20% v 14%, χ2=3.83 (1 df), P=0.05), live in poorer areas (deprivation categories 4-7, 20% v 11%, χ2=8.3 (1 df), P=0.004), and to have had a diagnosis of cancer for more than a year (40% v 15%, χ2=12.9 (1 df), P<0.001). There was no difference by age, sex, or newspapers read. A further 47 patients returned an incomplete hospital anxiety and depression scale at three weeks or three months and were not included in the analyses of anxiety and depression scales and mental adjustment questionnaires.
Use of computer
The average time spent using the computer at intervention was 12 minutes (range 1-44). Of those patients in the personal information group who were offered both personal and general information systems, two thirds (57/88) chose the personal information first. Twenty nine per cent of the patients used the computer again. Patients in the personal information group were more likely than those in the general information group to use the computer between the three week and three month follow ups (20/169 v 4/155, χ2=12 (2 df), P=0.002).
Patients' views and preferences
Patients given personal information were more likely to have a high satisfaction score, calculated from seven attributes (table 1), than were those given general computer information (mean difference 12%, 95% confidence interval 0.7% to 23.9%). More patients given personal information thought that the information was relevant and that they had learnt something new.
The patients who received booklet information were more likely to feel overwhelmed with information than were those given computer information, while patients in the computer information groups were more likely to think that the information provided was limited.
At three months' follow up, although 80% of patients expressed a preference for 10 minutes with a specialist nurse or radiographer to computer or booklet information, 20% preferred unlimited time with a computer, and those in the computer groups were more likely to do so.
Doctors thought more patients (35%) in the general computer information group were above average in knowledge compared with both the personal information group (25%) and booklet information group (20%) (P=0.01). They perceived no other difference.
Use of printed material at home
More of the patients offered booklet information (83%) used the material at home compared with those offered personal information (70%) or general information (57%) (χ2=22.4 (2 df), P<0.001). More patients in the personal information group showed the computer printouts to family or friends (36%) compared with those with general information (22%) or with booklets (21%) (χ2=6.7 (2 df), P=0.035).
In the absence of an electronic patient record, the personalised computer information system requires manual extraction of data from patients' case records and would currently cost over nine times the cost of the general information system. With the introduction of an electronic patient record, however, it would cost the same as the general system over time as fewer additions would be required.
Despite our buying the booklets at a discount, the average cost of booklets taken was over £7 per patient. A general computer information system would cost 40% of the costs offull access to booklets; even in the first year it would cost less. (For further details of costing, see appendix 3 on the BMJ's website.)
There were no significant changes in the patients' depression scores or mental adjustment scores between the start of treatment and follow up at three weeks and three months. However, 327 (84%) of the patients showed improvement in anxiety scores, of whom 255 (65%) improved in the first three weeks. At three months, 37% of patients in the general computer information group were still anxious compared with only 19% in the personal information group (mean difference 18%, 95% confidence interval 3.7% to 26.5%) (table 2). Exploration of other predictors by multiple logistic regression showed that type of cancer, age, sex, and type of newspaper read were all predictors of anxiety, but type of intervention remained a significant predictor with more patients in the general computer information group being anxious. (See appendix 4 on the BMJ's website for details of the intention to treat analysis).
Recruitment and follow up
The validity of our results is affected by patients' declining to take part in the study, many of whom may not have wanted more information. More patients in the general computer group failed to complete the three month follow up It may be that use of the personal rather than general computer information helped to retain the patients' interest, but the evidence is inconclusive.
Personalised v general computer systems
The NHS information strategy11 emphasises the importance of electronic patient record systems for clinical information but suggests the internet for patient information. Although most patients in this study preferred more time with a professional to computer or booklet information, one in five did not, and for all patients a computer could provide complementary information. However, the patients preferred the personalised information system to the general one and were more likely to use both computer and printout.
There was little difference in doctors' perceptions of the intervention groups. All the doctors at the Beatson Oncology Centre were willing collaborators in this study, and few clinicians now object to controlled patient access to medical records. Other reasons for combining patient information with electronic patient records include patients' audit of records14 and easier maintenance of integrated health service systems.19 Routine use by patients should be built into electronic patient records as they are implemented over the next few years.
Computers systems v booklets
Written information is important. 20 21 More of the patients offered booklet information used the material at home than did those given computer printouts, but the patients given personal computer information were more likely to use printouts than those given general computer information and were most likely to show their information to others. The printed booklets were more attractive than computer printouts, although these could be improved. Printed booklets are expensive, and tailored computer printouts could be produced more cheaply.
More of the patients offered booklets felt overwhelmed by the amount of information available, possibly because of the large number of booklets from which to choose. Providing a more restricted set of booklets, or nurse guidance in their selection, might have produced different results.
It is unclear why more patients in the computer groups thought that the information given was limited. The information presented was more complex than in other local systems.22–24 However, the patients received printouts only of the material they inspected during the computer session, so they did not have additional information to work through at home, as the booklet group did. Furthermore, the patients spent a relatively short time using the computer and may not have realised the depth of information available.
Improvement in anxiety
We did not directly try to reduce anxiety but measured it in case giving patients access to their medical records increased anxiety.14 Our results suggest the converse. There was some evidence that the patients given general computer information were more anxious than other patients at three months. Although we did not collect information on physical health to which anxiety might be related 25 26 and doubts have been expressed about the use of the hospital anxiety and depression scale among cancer patients,27 we think that the observed difference in anxiety at three months is most likely explained by the interventions. As more patients given personal information used their printouts with their family, we could hypothesise that this contributed to the difference.
What is already known on this topic
Various studies have examined different ways of “personalising” computer based information for patients
There has been no randomised trial testing the assumption that personalisation using the medical record is worth while
What this paper adds
This randomised trial showed that cancer patients thought a system giving them information based on their medical record was better than one giving only general information
Patients were more likely to use the personal system again and to show the printouts from that system to their family
There were no major differences in doctors' perceptions of the patients, but patients using the general information system seemed more anxious at three months' follow up
The study has implications for the design and implementation of electronic patient record systems and of patient information systems
Others have found that general information given after a consultation can inhibit patients' recall of the consultation28 and that patients used audiotaped consultations to inform their family of their situation.29 Whether providing patients with information helps to reduce their anxiety may depend on their coping style30–32 and warrants further study. Finding information on the internet can be difficult, and more thought is needed about its role as a primary information source for patients.
(See appendix 5 on the BMJ's website for a discussion of the feasibility of our computer system and of alternative computer based forms of patient information.)
This study strongly suggests that patient information should be linked to electronic patient records Patient information booklets are expensive, and computer based information could prove cheaper. However, further study is needed of how information, particularly from general sources, affects anxiety.
We thank the consultants (particularly Nick Reed and Tim Habeshaw, who were directors of the Beatson Oncology Centre during the study), other medical staff, cancer nurse specialists, radiographers, medical records staff, and other staff at the Beatson Oncology Centre for their collaboration with this project; the patients who took part in the study; Lynn Naven, who worked as a locum for SMcG during three months' sick leave; Ed Carter; Ross Morton and Keith Murray, who helped with various aspects of computing; Charles Gillis and Cathy Meredith, who advised on research design; Sally Tweddle, who made available unpublished protocols and papers; and colleagues in the University of Glasgow and the Beatson Oncology Centre who commented on the manuscript.
Contributors: RJ had the original idea for the study, designed the study, was the main grant holder, supervised the research assistants, analysed the data, and wrote the paper. JP carried out the pilot study, developed the computer system, contributed to the design, was a research assistant with day to day responsibility for data collection and running the study, undertook preliminary analysis, and contributed to final analysis and editing of paper. SMcG contributed to the design of the study, was a research assistant with day to day responsibility for data collection and running the study, undertook some preliminary analysis, and edited the paper. AJC discussed the original idea, contributed to the design and development of the computer systems and design of the study, discussed the analysis of data, edited the paper, and was a grant holder. AB discussed the original idea, set up opportunities for the pilot study and main study, contributed to the design of the computer system and study, edited the paper, and was a grant holder. NC advised on the design of the study and analysis of the data and edited the paper. JMA advised on the design of the study and choice of psychological measures, edited the paper, and was a grant holder. WHG advised on design of the study and analysis of the data and edited the paper. JMcE discussed the original idea, set up opportunities for the pilot study and main study, edited the paper, and was a grant holder. RJ and AB are guarantors for the study.
Funding Scottish Office Home and Health Department grant number K/OPR/2/2/D248. The views expressed in this paper are those of the authors and do not represent the views of the Scottish Office.
Competing interests None declared.
website extra Further details of the trial and its implications appear on the BMJ's website www.bmj.com