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.
The Internet is one of the primary sources for health information. However, in research, the effects of Internet use on the perception of one’s own health have not received much attention so far.
This study tested how Internet use for acquiring health information and severity of illness influence patients with a chronic disease with regard to the perception of their own health. Negative psychological states are known to lead to preferential processing of positive information. In particular, the self-directed nature of Internet use provides room for such biases. Therefore, we predicted that patients experiencing negative health states more frequently, due to more frequent episodes of a chronic illness, will gain a more positive perception of their health if they use the Internet frequently to gain health information, but not if they use the Internet rarely. This effect was not expected for other sources of information.
A longitudinal questionnaire study with two measurement points—with a 7-month time lag—tested the hypothesis in a sample of patients with chronic inflammatory bowel disease (n=208). This study assessed patients’ frequency of Internet use, their participation in online social support groups, their use of other sources of health information, and several indicators of the participants’ perceptions of their own health. A structure equation model (SEM) was used to test the predictions separately for Internet searches and other sources of information.
Data analysis supported the prediction; the interaction between frequency of health-related information searches and frequency of episodes at the first measurement point (T1) was related to participants’ positive perceptions of their own health at the second measurement point (T2) (B=.10, SE=.04,
The results of this study suggest that patients might process information from the Internet selectively, in an unbalanced, biased fashion, with the formation of a self-serving (ie, positive) perception of own health. At the same time, this bias contributes to the ability of patients to cope psychologically with their disease.
The Internet provides lay people with access to health information that was in earlier days available only to physicians and other health care professionals. Already in 2001, about 40% of Internet users searched online for health information [
The risks and benefits associated with the availability of health information on the Internet do not result only from the content of this information, but also from how patients process it (see Kalichman et al [
Health information is processed in a biased fashion. Positive or self-serving information (eg, suggesting own good health state) is, for instance, more easily accepted and less avoided than negative self-relevant information (eg, suggesting own illness) [
There are, however, reasons to assume that the extent to which the positive, or self-serving, bias occurs depends crucially on individuals’ health states and their psychological states resulting from them [
We believe that during Internet searches, a positive bias in information processing is very likely to occur because searching and surfing the Internet is completely self-directed (ie, not guided by external restrictions). Also, it can be done via multiple paths due to the hypertext structure and the virtually unlimited amount of information available [
Indeed, experimental research [
What are the long-term consequences of such a positive bias during Internet searches? As mentioned above, research on the immediate outcomes of Internet searches has shown that negative states lead to better memory for positive information and more positive attitudes toward the target around which the search centers [
A longitudinal study with two measurement points and a time lag of 7 months was conducted with patients who suffer from chronic inflammatory bowel disease to test this hypothesis. We focused on this chronic disease as it is characterized by infrequently occurring acute episodes of illness, which substantially restrict patients in their ability to cope with everyday demands; the episodes are, thus, likely to elicit negative psychological states. Moreover, patients with chronic diseases rely particularly on the Internet as an informational source because their illnesses strongly restrict them in their daily and social activities, which is why they are often unable to leave the house [
We used three indicators for the positive perception of own health that captured the immediate outcome of the positive bias—the perception of the risks resulting from the disease—as well as how it affects participants personally—health-related stress and health-related self-esteem. As information about their health and their disease is self-relevant to patients, we expected that our hypothesis would apply to all three indicators of perception of own health.
Participants were recruited and data were collected via the Internet in order to gain a sample of patients who were experienced in using the Internet for health-related purposes. To be able to test whether the predicted frequency of Internet searches by frequency of episodes interaction occurs for all sources of information or just for health-related Internet searches, we assessed online social support group participation and the consultation of offline sources of health-related information (ie, interactive and noninteractive sources). No frequency of episodes by source of information interaction was expected for sources other than Internet searches because these other sources do not provide the degree of self-directedness required to provide room for counter-regulation and the preferential processing of positive cues.
Patients with chronic inflammatory bowel disease participated in an open, online questionnaire study with two measurement points that were 7 months apart. Participants were recruited via the German Association for Crohn’s Disease and Ulcerative Colitis, which has more than 20,000 members and is, thus, the biggest organization of patients with chronic inflammatory bowel disease in German-speaking countries. The association advertised the study in their members’ journal, on their website, via email, and on social networking sites.
When following the link in the advertisement, participants first had to provide informed consent online. After receiving information about the duration, the content, and the aims of the study, as well as the data storage policy, each participant had to actively check a box and, thereby, indicate informed consent. Participation was completely voluntary. To get access to the survey, participants had to enter their email address into a Web form. Email addresses were stored separately from the other data and were only accessible to the system administrator (ie, not to the researchers), who deleted them after the second wave of data collection was completed. Participants received an email with a link to the actual questionnaire.
The reported study was ethically approved by the ethics committee of the Faculty of Medicine at the University of Tübingen, Germany. As compensation, participants who completed questionnaires from both waves received a gift voucher of €10—approximately US $11—to be redeemed at an online sales platform.
All measures of the survey at the first measurement point (T1) and the second measurement point (T2) were assessed via the online survey program, Questback. Items were presented in a fixed order. First, participants answered demographic questions (ie, gender, age, and education) and general questions about their health and illness (ie, type of diagnosis, time since receipt of diagnosis, time since occurrence of symptoms, number of episodes, current acute episode, and severity of acute episode). These questions included the assessment of the key predictor, frequency of episodes. Within a larger battery of measures, participants’ health-related self-esteem, health-related stress, and health-related risk perceptions were assessed and served as the key dependent measures. Finally, participants answered questions about their Internet use—the second key predictor—the Internet services they used, which online social support group they joined, and which offline sources of information they used. The functionality and readability of the questionnaire was pretested by healthy participants. The questionnaire contained up to 14 items. Returning to already completed pages of the questionnaire was not possible. It was mandatory to complete the items assessing key concepts. Participants were contacted via email about 7 months after they had completed the questionnaire from the first wave. Those who did not reply within a week received a reminder. All measures at T2 were exactly the same as at T1.
In the next passage, questions on the use of health information sources had to be answered by checking or not checking boxes. Possible information sources were search engines, forums, encyclopedias, patient association websites, Internet portals, scientific search engines, newsletters, and social networks.
For
Participants had to be 18 years of age or older and have chronic inflammatory bowel disease. The only exclusion criterion was that the answers to the two main disease-related variables—time since receipt of diagnosis and number of episodes per year—had to be within a plausible range: < 60 years and < 20 episodes, respectively.
To test the hypothesis, a structure equation model (SEM) was applied using SPSS Amos version 22 (IBM Corp). In this model, the three manifest indicators of the positive perception of own health were summarized as a latent variable, separately for T1 and T2. Relationships between the assessed indicators and the latent variables were set equally across both time points. The model regressed the perception of own health at T2 on the same variable at T1, as well as on frequency of health-related information searches, frequency of episodes, and their interaction (see
The correlation between the measurement error of frequency of episodes at T1 and the error term for positive perception of own health at T1 was set free as it was likely that both were related. Moreover, measurement errors of health-related stress, health-related self-esteem, and health-related risk perception at T1 were allowed to covary with their respective measurement errors at T2. This was done because we assumed that components of these indicators that are not part of the latent variable of positive perception of own health might still be stable. The model did not include any further covariates. To test for the opposite causal relationship, we also computed a model in which the perception of own health at T1 predicts the interaction between frequency of episodes and health-related Internet searches at T2.
Finally, additional models were tested to find out whether Internet searches alone or also other types of information searches interact with the frequency of episodes in their impact on the perception of own health. These alternative models replaced health-related information searches on the Internet by social interactive sources, by noninteractive sources, or by participation in online social support groups. Except for this change in the predictor and its interaction with frequency of episodes, the alternative models did not exhibit any changes in comparison to the main model. We did not expect to find an interaction between the respective information sources and the frequency of episodes on positive perception of own health in these models.
Structure equation model (SEM) for the effects (unstandardized regression coefficients, B) of frequency of episodes, frequency of health-related information searches, and their interaction from the first measurement point (T1) on the latent variable, positive perception of own health, and from the second measurement point (T2), controlling for positive perception of own health from T1. NFI: normed fit index; TLI: Tucker-Lewis index; CFI: comparative fit index; RMSEA: root mean square error of approximation; ns: not significant; *P<.05; ***P<.001.
A total of 8 out of 216 participants (3.7%) who completed both questionnaires were excluded from the analyses reported below based on the exclusion criteria: time since receipt of the diagnosis (> 60 years) and number of episodes during the last year (> 20 episodes). Thus, 208 participants were included in the analyses: 154 women (74.0%), 54 men (26.0%), mean age 37.77 years (SD 11.51, range 18-63). Out of these participants, 0.5% (1/208) had not finished school, 8.7% (18/208) had a certificate of secondary education after 9 years of schooling (German: Hauptschulabschluss), 26.4% (55/208) had a general certificate of secondary education after 10 years of schooling (German: Mittlere Reife), 30.8% (64/208) had a high school degree (ie, a certificate of general university maturity after 12 years of schooling), 10.1% (21/208) had a polytechnic degree, and 23.6% (49/208) had a university degree. Thus, the sample was well educated.
The self-report about participants’ diseases revealed that 57.7% (120/208) of the participants had Crohn’s disease, 40.4% (84/208) had ulcerative colitis, 1.0% (2/208) had indeterminate colitis, and 1.0% (2/208) reported to have a disease different from the aforementioned ones. The participants received the diagnosis an average of 11.79 years (SD 10.01) ago and had had symptoms for an average of 13.61 years (SD 10.83).
Participants reported using the following online sources of health information: search engines (196/205, 95.6%), forums (154/205, 75.1%), encyclopedias (109/205, 53.2%), patient association websites (108/205, 52.7%), Internet portals (68/205, 33.2%), scientific search engines (eg, Google Scholar and PubMed) (53/205, 25.9%), newsletters (51/205, 24.9%), and social networks (44/205, 21.5%). Out of 208 participants, 3 (1.4%) did not answer this question. In addition, they indicated using the following offline sources of health information: doctors (187/204, 91.7%), family and friends (108/204, 52.9%), psychologists or advisory centers (52/204, 25.5%), books (131/204, 64.2%), presentations (113/204, 55.4%), newspapers and journals (90/204, 44.1%), and television or radio (67/204, 32.8%). Out of 208 participants, 4 (1.9%) did not answer this question.
The first page of the questionnaire had 319 hits. A total of 258 patients with inflammatory bowel disease completed the whole questionnaire at T1 (258/319, 80.9%) and 216 patients completed the questionnaire at T2 (216/319, 67.7%): dropout rate of 16.3% (42/258) from T1 to T2. These subsamples did not differ with respect to the following variables: gender (χ2
1=0.1,
The correlations between variables and their means and standard deviations are displayed in
Correlation of all variables.
Number | Variables | Mean | SD | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 |
1 | Health-related Internet search T1a | 4.24 | 1.46 |
|
|
|
|
|
|
|
|
|
2 | Frequency of episodes T1 | 1.95 | 2.11 | .07 |
|
|
|
|
|
|
|
|
3 | Health-related stress T1 | 2.63 | 0.95 | .19b | .22b |
|
|
|
|
|
|
|
4 | Health-related self-esteem T1 | 3.42 | 1.06 | -.16c | -.23b | -.66d |
|
|
|
|
|
|
5 | Health-related risk perception T1 | 3.30 | 0.65 | .17c | .20b | .38d | -.30d |
|
|
|
|
|
6 | Health-related Internet search T2e | 4.07 | 1.48 | .61d | .04 | -.01 | .05 | .08 |
|
|
|
|
7 | Frequency of episodes T2 | 1.64 | 1.98 | .16c | .53d | .23b | -.16c | .21b | .04 |
|
|
|
8 | Health-related stress T2 | 2.58 | 0.99 | .18b | .12 | .75d | -.55d | .26d | .08 | .25d |
|
|
9 | Health-related self-esteem T2 | 3.49 | 1.08 | -.19b | -.15c | -.62d | .72d | -.32d | -.07 | -.17c | -.73d |
|
10 | Health-related risk perception T2 | 3.17 | 0.66 | .10 | .18b | .32d | -.28d | .57d | .19b | .14c | .35d | -.40d |
aT1: first measurement point.
b
c
d
eT2: second measurement point.
We predicted that frequency of episodes and health-related information searches on the Internet interact in their impact on patients’ positive perceptions of their own health. To be more precise, we expected that in patients using the Internet frequently, but not in patients using it rarely, more frequent episodes should lead to a more positive perception of their own health. Multi-colinearity is often an issue when computing interactions between variables assessed in one source. Yet, our predictors—frequency of episodes and frequency of health-related information searches—were not correlated (r=.07, n=208,
The predicted model showed a good fit to the data (see
Frequency of health-related information searches at T1 did not predict the perception of own health at T2 (B=-.03, SE=.04,
Additional SEMs were computed to gain further insights about this effect. First, a model testing the opposite causal direction did not find evidence for any influence of perception of own health at T1 on the interaction between frequency of episodes and health-related Internet searches at T2 (B=-.03, SE=.10,
Moreover, another model tested whether the same interaction effect occurred for participation in online social support groups, instead of health-related Internet searches. This analysis served to rule out that any health-related Internet search, which might similarly occur in online social support groups, results in the predicted effects. In this model, no interaction between online social support group participation and frequency of episodes on the perception of own health was found (B=.02, SE=.05,
Finally, two additional models tested whether offline sources of health-related information, namely, social interactive and noninteractive (ie, purely informational) sources, lead to effects similar to those of health-related Internet searches. Again, no interaction occurred between the frequency of episodes and the consultation of health-related offline sources (social interactive sources: B=-.02, SE=.04,
In sum, the results support the hypothesis that health-related Internet searches for knowledge acquisition and the frequency of episodes associated with the illness have an impact on participants’ perception of own health over time; with increasing frequency of episodes, the perception of their own health became more positive when participants used the Internet for health-related information acquisition often, but not when they used it rarely. The additional analyses did not find similar effects for information acquisition from offline media or other individuals, nor from online social support groups. This indicates that the effects described here are unique for self-directed information acquisition on the Internet.
Results from the regression analysis within the structure equation model (SEM) analysis with the latent variable, positive perception of own health, from the second measurement point (T2) as criterion, and frequency of episodes and frequency of health-related information search as predictors controlling for positive perception of own health from the first measurement point (T1). ns: not significant; **
The current research aimed to test the long-term effects of patients’ Internet use on the perception of their own health. We predicted that patients experiencing episodes of illness more frequently have a more positive perception of their health when frequently using the Internet, but not when rarely using the Internet to search for health-related information. This prediction relied on research on counter-regulation [
The interaction between frequency of health-related Internet searches and frequency of episodes of their chronic disease predicted patients’ perception of their own health across a 7-month period. Additional analyses did not find evidence for similar effects for other sources (ie, external sources and social support group participation) or for the opposite causal direction. The results clearly indicated that the effects occur, in line with our prediction, only as a consequence of information-related Internet searches and not when social information sources (eg, friends or health professionals) or other offline information sources (eg, books) were resorted to. This suggests that, in line with our assumption, the self-directed nature of information acquisition on the Internet provides the contextual basis for counter-regulation and its impact on information processing and acquisition. In other words, the degrees of freedom the Internet provides during information searches allow for selective information acquisition and the formation of a self-serving (ie, positive) perception of own health when experiencing episodes of illness more frequently. This interpretation of the findings is speculative insofar as this study did not assess any indicators of this assumed process (ie, room for self-regulation leading to selective processing of information in threatened individuals). At the same time, earlier studies have provided ample evidence that self-regulated information searches on the Internet lead to counter-regulation and a preferential processing of positive information in individuals in a negative state [
To the best of our knowledge, this study was the first to study long-term effects of frequency of episodes and Internet searches to acquire health-related information. Earlier research on health within the context of the Internet mostly focused on online prevention programs [
Beyond its originality, the strengths of this study are its longitudinal design and the fact that several indicators for the dependent variable have been assessed targeting the perception of own health, as well as the perception of the illnesses patients are suffering from. Separate tests for each of the three indicators show the same results as the analysis across indicators reported above.
Only patients with specific chronic diseases were included in the study. In this study, relying on patients with chronic inflammatory bowel diseases allowed for the assessment of a more homogenous and less subjective indicator of the severity of illness, namely, the number of episodes during the last year. In this vein, focusing on one particular group of diseases allowed for the assessment of an appropriate indicator of severity of illness. Nonetheless, further research should definitely aim to replicate the current findings with other chronic diseases and beyond.
One might consider the limited internal consistency of the indicator of health-related risk perception as a weakness of this study. As no scale for risk perception in the context of chronic inflammatory bowel diseases exists in the literature, we developed our own scale that did not work out perfectly. However, given that we use a latent variable for the perception of one’s own health, the error variance included in the risk perception scale does not affect our main statistical test. In addition, risk perception loads relatively low on the latent variable; the model also holds when only the other two indicators of perception of one’s own health are used. For these reasons, we do not consider the unsatisfying internal consistency of the risk perception scale as a threat to the validity of these findings.
A final limitation of this study is that we did not assess what type of information participants studied on the Internet. This would have allowed us to find out whether they select, receive, or remember information in an unbalanced fashion. However, collecting that information and analyzing it across a 7-month period hardly seems feasible. In addition, earlier experimental research has already addressed this aspect and found that all steps of the information search process on the Internet are guided by a preference for positive information in a negative psychological state (ie, threat) [
The findings from this study have ambivalent implications for patients. On the one hand, Internet use seems to help patients with severe illnesses to develop a positive perception of their own health because the information-related Internet searches lead to a more positive picture of their own health situation—in particular, if they are severely affected by their disease (ie, frequently suffer from it). Hence, this research has added one more aspect to the list of positive implications of health information searches on the Internet, such as increased empowerment and heightened compliance among patients using the Internet [
This distorted and possibly too-positive perception, in turn, can have implications for the doctor-patient relationship and the medical treatment of those patients. Internet-informed patients might be less willing to accept their doctors' advice and claim more autonomy in health decisions, due to an increased positive perception of their own health as was found in this study. This might in turn lead to less willingness to comply. Therefore, physicians should pay particular attention to this implication in order to prevent their patients, fatally enough, from making decisions against appropriate treatments. At the same time, patients should be aware of the need to carefully conduct Internet searches and examine the integrity of an information source before relying on its information. Further research is needed to gain more insight about these implications as they rely more on speculative conclusions from this study's findings than on the findings themselves.
This study provided evidence for long-term outcomes of Internet searches for health information. The more severe the patients’ illnesses (ie, the more frequently they suffer from them), the more that frequent Internet searches lead to a positive perception of own health. Thus, the accessibility of health-related information on the Internet renders patients more informed than in the past when health-related information was accessible only to health professionals; it also opens an avenue for unbalanced information processing. This unbalanced processing might help patients cope emotionally with their situation; however, it might result in a biased perception of their own health, particularly in those who suffer severely from a disease.
comparative fit index
normed fit index
not significant
root mean square error of approximation
structure equation model
first measurement point
second measurement point
Tucker-Lewis index
The research reported here was supported by a grant from the ScienceCampus Tuebingen (TP5.1) awarded to KS.
None declared.