This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on http://www.jmir.org/, as well as this copyright and license information must be included.
There has been relatively little research on the role of web-based support for self-care in the management of minor, acute symptoms, in contrast to the wealth of recent research into Internet interventions to support self-management of long-term conditions.
This study was designed as an evaluation of the usage and effects of the “Internet Doctor” website providing tailored advice on self-management of minor respiratory symptoms (eg, cough, sore throat, fever, runny nose), in preparation for a definitive trial of clinical effectiveness. The first aim was to evaluate the effects of using the Internet Doctor webpages on patient enablement and use of health services, to test whether the tailored, theory-based advice provided by the Internet Doctor was superior to providing a static webpage providing the best existing patient information (the control condition). The second aim was to gain an understanding of the processes that might mediate any change in intentions to consult the doctor, by comparing changes in relevant beliefs and illness perceptions in the intervention and control groups, and by analyzing usage of the Internet Doctor webpages and predictors of intention change.
Participants (N = 714) completed baseline measures of beliefs about their symptoms and self-care online, and were then automatically randomized to the Internet Doctor or control group. These measures were completed again by 332 participants after 48 hours. Four weeks later, 214 participants completed measures of enablement and health service use.
The Internet Doctor resulted in higher levels of satisfaction than the control information (mean 6.58 and 5.86, respectively;
Our findings provide initial evidence that tailored web-based advice could help patients self-manage minor symptoms to a greater extent. These findings constitute a sound foundation and rationale for future research. In particular, our study provides the evidence required to justify carrying out much larger trials in representative population samples comparing tailored web-based advice with routine care, to obtain a definitive evaluation of the impact on self-management and health service use.
There has been relatively little research on the role that web-based support for self-management might play in the management of minor, acute symptoms, in contrast to the wealth of recent research into Internet interventions to support self-management of long-term conditions. It is well known that patients already self-care for the vast majority of minor symptoms, making their own decisions about whether and how to manage symptoms themselves (eg, using over-the-counter remedies) or whether to seek medical advice [
There are compelling reasons for finding ways to use the Internet to support patients to self-manage minor symptoms. Most people say that they would find it convenient and empowering to be given enough information to be able to self-manage without seeing their doctor [
Prior to the advent of mass Internet access, patient education about self-management of minor symptoms was attempted by means of booklets and other media with some degree of success [
Previous studies of providing information on self-management of symptoms have been largely pragmatic, focusing simply on whether providing educational materials leads to better outcomes than routine care. For example, an observational study of providing a student population with online digital triage advice on whether they needed to seek medical care for minor symptoms was able to demonstrate satisfactory uptake and excellent concordance between the online advice and clinical diagnoses [
According to the Social Cognitive Theory, performance of any behavior is typically predicted by confidence that one can carry out the behavior successfully (self-efficacy) and beliefs about the likely consequences (“outcome expectancies”) [
This study forms part of a program of research into how theory and evidence can be used to design an intervention that will help patients to self-manage minor respiratory symptoms without seeking medical help. In accordance with best practice in the development of complex interventions [
This study was designed as an exploratory or phase-2 RCT [
The study was approved by the ethics committee of the School of Psychology, University of Southampton. Participants were recruited between October 2009 and March 2010 (the UK winter respiratory infection season) by advertisements providing the website uniform resource locator for the intervention and inviting adults with cold or flu symptoms to try out the website. We specifically targeted university students, as our own qualitative research [
Participants were sent an automatic email invitation to complete the intermediate follow-up 48 hours after accessing the intervention, and an invitation to complete the final follow-up after 4 weeks. An incentive (being entered into a prize draw for £100) was offered for completion of the follow-up measures, and nonrespondents received up to two additional reminders to complete the follow-up.
The intervention was a fully automated digital triage system that provided tailored computer-generated advice. Participants were presented with a homepage (
Homepage of the Internet Doctor website
The Diagnostic pages first asked a series of questions about the participant’s symptoms; participants completed these pages for one symptom at a time, and could choose from cough, sore throat, fever, and runny/stuffy nose. Then a complex algorithm provided appropriate tailored advice on whether they needed to contact health services for that symptom (see
Varieties of advice provided by the Internet Doctora
Advice type | Symptom reports prompting this advice |
Contact NHS Directb immediately and then your doctor (gives list of symptoms reported that led to this advice) | Symptoms indicating a serious, acute condition (eg, meningitis or septicemia) |
You should contact NHS Direct (gives list of symptoms reported that led to this advice) | Symptoms lasting and/or moderately severe (eg, fever ≥38.5o for ≥3 days, cough for ≥4 weeks, breathing getting worse) OR less severe symptoms together with other risk factors (eg, older age, chronic conditions, immune system suppression) |
Your symptoms are not a sign of serious illness and you do not need to contact the doctor at present (gives reassuring explanation of symptoms and advises to reconsult website if symptoms persist or worsen) | Symptoms acute and not severe or worsening |
a Screenshots illustrating each advice type are given in
b NHS Direct is a national telecare triage system providing 24-hour telephone support. We advised contacting NHS Direct in the first instance, as this service offers instant personal triage regarding appropriate next steps (eg, call ambulance, see doctor next day, etc).
The intervention was created by the research team using the LifeGuide software [
Final and intermediate outcome measures
Time point/target construct | Scale/itema | alphab | |
|
|||
Enablement |
|
||
Health Services Use | Three items asking whether since using the website the respondent had contacted (1) their general practitioner (or other practice staff), (2) NHS Direct or the National Pandemic Flu Servicec, or (3) any other health care services (eg, accident and emergency) | ||
|
|||
Satisfaction | Three items assessing satisfaction with and trust in the website advice (see |
.89 | |
|
|||
Intentions |
|
.97 | |
I plan to go to see a doctor for my symptoms | |||
I intend to go to a doctor for my symptoms | |||
Self-efficacy |
|
.94 | |
I know what to do about my symptoms | |||
I can care for my symptoms myself | |||
I can cope with my symptoms without going to a doctor | |||
Outcome expectancies |
|
.92 | |
I will get better more quickly if I go to see a doctor | |||
Seeing a doctor will help me to recover | |||
My illness may get worse if I do not see a doctor | |||
I could become very ill if I do not see a doctor | |||
Illness perceptions |
|
.95 | |
|
.91 |
a Full wording of items is provided for measures newly constructed for this study.
b Cronbach alpha coefficient is provided for scales newly constructed for this study.
c Data were collected during a period in which government advice was to contact the National Pandemic Flu Service for flu symptoms.
We assessed the primary outcomes at final (4-week) follow-up by two measures. The Patient Enablement Instrument [
Numbers of participants completing study measures and phases varied, and so precise numbers are given for each analysis. Since many variables were somewhat skewed toward low concern about symptoms, we used conservative nonparametric tests to compare groups on the final outcome variables. We used a 2-tailed Kruskal-Wallis test for between-group comparisons in patient enablement scores, and a 2-tailed chi square test to compare numbers contacting health services.
Parametric analyses were used for the secondary analyses, as there are no satisfactory nonparametric tests for time-by-group interactions and analysis of variance (ANOVA) is robust to violations of the assumption that variables are normally distributed, unless sample sizes are small [
Hierarchical linear regression was employed to examine predictors of change in intentions, pooling data across both groups. To identify bivariate predictors of change in intentions, separate regressions of each baseline and website usage predictor were carried out with intentions at follow-up as the dependent variable, controlling for baseline intentions. We then carried out a multiple regression to determine whether (1) psychological variables predicted change in intentions after controlling for relevant demographic variables, and (2) use of the Diagnostic Webpages predicted change in intentions after controlling for relevant demographic and psychological variables (ie, those with a significant bivariate relationship to change in intentions). For this regression, after controlling for baseline intentions in step 1, age was entered in step 2 (dichotomized into aged under or over 25 because of a marked skew). In step 3 we entered consultation necessity beliefs (since theory predicts these should be directly related to intentions) and in step 4 we entered illness perceptions (poor understanding of illness and emotional reactions). Finally, in step 5 we entered use of the Diagnostic Webpages. We inspected the residuals from the final regression equation to confirm that they were normally distributed.
Baseline measures were completed by 714 people; 368 (51.5%) were randomized to the Internet Doctor website and the remainder to the static website control (see
Flow of participants through the trial. aPercentages refer to proportion of group completing follow-up
Of the 214 people who completed the measures of the target outcomes at the final (4-week) follow-up, 95 (44.4%) had been assigned to the Internet Doctor group. The median patient enablement score was significantly greater in the Internet Doctor group than in the control group (median score of 3 vs 2, with an interquartile range of 0 to 5 for the whole sample,
Of the 332 (46.5%) people who completed the intermediate outcomes at first follow-up, 167 (50.3%) were in the Internet Doctor group. The Internet Doctor group rated the website more positively than the control group did on all satisfaction measures (see
Satisfaction with web-delivered advice in the Internet Doctor and control groups (n = 332)
Scale/item | Mean (SD) for each group |
|
|
Internet Doctor | Control | ||
Total scale (summed items divided by 3) | 6.58 (1.96) | 5.86 (2.27) | .002 |
The website gave me all the advice I needed | 6.40 (2.05) | 5.63 (2.51) | .002 |
The website was helpful to me | 6.41 (2.17) | 5.72 (2.51) | .007 |
I felt I could trust the website | 6.91 (2.21) | 6.25 (2.54) | .01 |
Intentions to consult the doctor declined between baseline and the intermediate (48-hour) follow-up; although the decline was greater in the Internet Doctor group this difference did not reach significance (see
Intentions and attitudes at baseline and intermediate follow-up (n = 332)
Internet Doctor group means (SD) | Control group means (SD) |
|
|
|
|||
Scale | Baseline | Follow-up | Baseline | Follow-up | Time | Group | Interaction |
Intention to consult doctor | 2.00 (2.57) | 1.66 (2.32) | 1.88 (2.57) | 1.82 (2.45) | .03 | .93 | .11 |
Consultation necessity beliefs | 2.54 (2.25) | 2.29 (2.37) | 2.38 (2.23) | 2.03 (2.37) | .01 | .62 | .61 |
Confidence to self-care | 7.75 (2.00) | 7.69 (2.08) | 7.78 (1.97) | 7.80 (2.01) | .84 | .73 | .62 |
Poor understanding of illness | 1.86 (2.13) | 1.65 (1.92) | 1.64 (2.05) | 1.70 (2.07) | .29 | .70 | .05 |
Emotional reactions to illness | 2.36 (2.14) | 2.03 (2.21) | 2.40 (2.42) | 2.17 (2.30) | <.001 | .70 | .53 |
a Significance of main effect for time, ie, change from baseline to follow-up
b Significance of main effect for between-group difference
c Significance of interaction between time and group effects, ie, group difference in change from baseline
The mean duration of website usage in the Internet Doctor group was 454 seconds (around 8 minutes), with a range from 24 seconds to over 52 minutes. Of the 368 people randomized to the website, 280 (76.1%) looked through the pages. Just over half (196; 53.3%) entered the Diagnostic section, a similar proportion (203; 55.2%) looked at the Treatment section, and over a quarter (104; 28.3%) looked at the Common Questions. Examination of the numbers of participants using each individual webpage revealed very diffuse usage, with virtually every page being used by at least some participants. Advice was provided for 146 symptoms, comprising runny nose in 57 (39.0%) cases, cough in 50 (34.2%) cases, sore throat in 29 (19.9%) cases, and fever in 10 (6.8%) cases. In 30.8% (45) of these cases the advice given was to contact health services.
Twenty-one people advised to contact health services completed the intermediate follow-up. There was no difference in satisfaction levels between those who were and those who were not advised to contact health services (mean 6.79, SD 2.03 and mean 6.18, SD 2.15, respectively;
Regression analysis (see
Baseline and website usage predictors of intentions to consult the doctor at intermediate follow-up, controlling for intentions at baseline
Variables | Bivariate regressionsa | Hierarchical regressionb | ||
Beta |
|
Final beta |
|
|
Baseline intentions | .78 | <.001 | .617 | <.001 |
Age less than 25 | .10 | .003 | .07 | .04 |
Gender | .01 | .79 | - | - |
Consultation necessity beliefs | .19 | <.001 | .13 | .01 |
Confidence to self-care | -.04 | .30 | - | - |
Poor understanding of illness | .11 | .004 | .05 | .18 |
Emotional reactions to illness | .15 | <.001 | .11 | .01 |
Diagnostic section used | .09 | .007 | .08 | .02 |
Treatment section used | .06 | .07 | - | - |
a Intention to consult the doctor entered in step 1, then contribution of each variable examined independently.
b Intention to consult the doctor entered in step 1, then variables entered in order shown, omitting those with nonsignificant bivariate relationships to intention change (see Method for details and rationale). Beta weights shown are for the last step of the equation.
The findings from this study suggest that tailored website advice may prove superior to simply providing written information about self-care. The Internet Doctor advice was rated as more helpful and trustworthy than the control information and resulted in higher levels of patient enablement a month later. Understanding of illness improved in the 48 hours following use of the Internet Doctor webpages, whereas there was no improvement in understanding of illness in the control group.
The shift toward weaker intentions to consult the doctor after using the website was more marked for the student-aged participants, consistent with our expectation that providing advice on self-care (in both groups) would have more influence on the intentions of those with less experience of independent self-care. As expected, reduction in intentions to consult was also predicted by prior beliefs that consultation was necessary to achieve recovery, poor understanding of illness, and greater emotional reactions to illness. This finding confirms that providing advice is likely to have most influence on the consultation rates of those who are most puzzled and distressed by their symptoms, and concerned that they may not recover without medical help. This profile matches that of patients who are more likely to consult [
Use of the Diagnostic section of the Internet Doctor website predicted a reduction in the strength of intentions to consult, whereas use of the Treatment section did not. This finding is not entirely surprising, since only the Diagnostic section provided specific advice about whether medical help was necessary. However, an unexpected finding was that confidence to self-care was unrelated to change in intentions to consult. Since confidence that one can carry out a behavior successfully (self-efficacy) is usually a strong predictor of behavior [
Only a minority of people were advised to contact health services, a much smaller proportion than in previous studies of triage for minor symptoms [
This study had a number of strengths as a direct test of the effects of tailored advice in the context of self-management of minor symptoms: in particular, a direct comparison with the best existing nontailored patient information, and detailed analysis of reliable, theory-based measures of relevant beliefs and perceptions. However, the findings cannot be considered definitive. The sample size was too small to reliably detect group differences in consultation rates, and reported consultation rates were not objectively verified. Future research should evaluate the effects on recorded consultations in a much larger sample, following all the usual conventions for a full trial.
While our study design provided a strong test of the efficacy of tailoring information, it did not permit evaluation of the effectiveness of the website for reducing consultation rates, since the control group was given nontailored advice that was previously shown to be effective in reducing consultation rates. It is encouraging that a reduction in intentions to consult, consultation necessity beliefs, and emotional reactions was seen in both groups after using the website. However, a further trial is needed, including comparison with a control group who are not given access to any triage advice, in order to determine to what extent reductions in consultations intentions are due to receiving web-based advice.
There was substantial dropout before follow-up, which is a common problem in internet studies with volunteer samples [
Our findings provide initial evidence that tailored web-based advice could help patients self-manage minor symptoms to a greater extent. Effect sizes on consultation rates were modest, which is consistent with previous research suggesting that often information may be obtained from the Internet in order to supplement rather than replace consultations with doctors [
This study was supported by grant number RES-149-25-1069 from the UK Economic and Social Research Council, and was carried out by the Southampton node of the Digital Social Research Programme. The website was developed by Jonathon Hare and Adrian Osmond, under supervision by Gary Wills and Mark Weal. We would like to thank Michael Moore for assisting with development of the medical advice.
None declared
Illustrative screen shots of the Internet Doctor
analysis of variance
randomized controlled trial