Patient choice, Internet based information sources, and perceptions of health care: Evidence from Sweden using survey data from 2010 and 2013

Sampling

The VB-survey has a cross sectional study design and the study population consists of individuals 18 years and older at the time of the survey (around 949,000 and 976,000 people in Skåne in 2010 and 2013, respectively). Data are collected during two separate rounds, one in the spring and one in the fall, through phone interviews where the interviewer reads the questions to the respondents and registers the responses. The sampling frame of the survey is a database that contains the telephone numbers (cell and fixed-line) of all people living in the country (except unlisted numbers) [26]. The VB-survey aims to interview a representative sample of around 0.5 % of the adult population by making a random draw from the sampling frame. This translates into a minimum of 1,000 people from a medium-sized county, i.e. the survey terminates when the target number of participants have been reached. In 2010, 1,000 persons were interviewed in Skåne and in 2013 6,000 people were interviewed. The survey response rate was similar in both years at 40.5 % and 45.9 %, respectively; i.e. slightly more than twice the number of persons were contacted for interview than the final number of participants (the interviewers make a maximum of five contact attempts and no rewards are given for participation). The response rate of the Skåne VB-survey is similar to that in other counties and to that of other comparable household surveys [26].

Descriptive statistics

Table 1 shows the frequencies and relative share of the respective response categories for all analytical variables used in the study.

Around two-thirds of the sample in both years report that they have high confidence in their provider. A fairly stable majority also report that they do think that they have access to the care that they need. Most people are in relatively good health and have good general knowledge about their health. The largest age category in the sample is Old (over 60 years) and the majority of respondents have primary and secondary (below university) education.

Prevalence of IBIS use and user characteristics

The actual use of the Internet to obtain health care information is relatively low in this sample of the general adult population of Skåne (Table 1). In 2010, around 5.8 % of the full sample stated that they had used an Internet based information source for this type of information; see Table 1 for details. Three years later this share had increased to around 6.4 %, an approximate 10 % increase over this period.

In 2010, self-assessed health, chronic ill health, and self-reported health knowledge do not appear to matter significantly for whether a person used the Internet or not to obtain health information. However, with the exception of self-assessed health (SAH), these factors did appear to matter in 2013. For the other factors there were no large differences in these frequencies between the two survey years.

The use of IBIS and patient satisfaction

The effect of IBIS appear to be somewhat larger in 2013 than in 2010 for both outcomes and the models explain a slightly larger share of the variation in perceived access to care for both years.

With respect to the other factors, age would seem a relatively important aspect in explaining the observed outcomes: Young and Middle age both have lower odds of stating they have adequate access to care compared to the Old group (Odds ratios 0.608 [95 % CI: 0.374–0.988] and 0.511 [0.341–0.765], respectively).

For those assessing their own health as Good, the odds ratio of 1.788 [95 % CI: 1.499–2.131] in 2013 showed that they had higher odds of perceiving that access to care was good than those assessing their health as Not good. For all other independent variables entered in the model, no statistically significant relationships were established by the models.

In addition to the specifications applied above, the models were also tested using alternative definitions of some of the variables. For example, the IBIS variable was also defined to include reported knowledge of the existence of the Internet based information source, www.1177.se. While this resulted in slightly different absolute values of the beta coefficient to those in Table 3, the overall results were robust to these alternative specifications. The next section includes further sensitivity analyses.

Methodological discussion

One limitation of the study is the nature of the data. While the samples are of acceptable size, the cross-sectional design effectively prohibits a causal analysis of the relationships. The implication of this is that it is not possible to say, from the results of this study alone, that using the Internet to obtain health information leads to less satisfaction with health care, as it could also be the case that for some reason, people who are less content with health services tend to use this source of information to a larger extent than do more generally satisfied persons [27]. One such reason could be an inherent attitude toward both the Internet as such and toward medical care [28]. Failure to include an indicator of such an attitude would render the IBIS variable endogenous in the models and its estimated coefficients would be biased and possibly inconsistent. The analytical solution to the issue of possible unobserved heterogeneity and associated endogeneity of, in this case, the IBIS variable is to apply instrumental variable (IV) estimation techniques [29]. Specifically, this would involve replacing the IBIS variable with an indicator that is associated with IBIS use, but not with individuals’ attitudes toward the Internet and health care (now present in the epsilon error term in the models). As no such information is available in the current sample data the interpretation of the results needs to be made with caution, especially the absolute sizes of the parameter coefficients. Relatedly, the deletion of all observations with incomplete data in the estimation models may also lead to some informational loss.

A second methodological concern related to the data is the way in which the variables are measured. For example, access to care and confidence in providers are measured from specific questions on these issues in the VB-survey. While this provides a transparent and straightforward way of measuring these outcomes it may limit the measures’ comparability with those of other studies that use different ways of measuring access and patient confidence [30, 31]. In addition, the re-categorization of some of the dichotomous variables used in the analyses facilitates their interpretation, but may also reduce the power of the estimates.

A final data issue relates to the sampling of participants. While the response rate is comparable with that of other similar surveys, the VB-survey aims to include a certain number of participants (1,000 and 6,000, respectively). Although this approach ensures a sufficient number of individuals in the final sample, it may also subject the survey to the element of self-selection given that participation is voluntary. Those who decline to participate may do so for certain, non-random reasons, such as having worse health or being significantly younger than those who do participate. For example, at the national level, in the 2013 survey, a total of 15,489 contacted persons declined to participate. Other reasons for non-response included Wrong telephone number (3,257 persons), Deceased (100), and Not eligible for other reasons (1,202) [26].

Discussion of main findings

Notwithstanding these limitations in the data, the study’s findings are of relevance to both the scientific evidence base on the role of the Internet in individuals’ health care seeking behaviors and to policy development. The main finding of the study is that the use of Internet based information sources to obtain relevant information about choosing a service provider or learning of the quality of different providers reduces the odds of reporting general satisfaction with the health services. While patient perceptions of health care is likely to depend on many other things than the particular source of information, the effect of Internet use on various outcomes, including these ones, is important, particularly in light of the raised expectations among policy makers that this source of information is having, also in the area of health care [32, 33]. Depending on the role that the act of seeking information plays in individuals’ overall health care consumption, this may or may not be seen as counter-intuitive. If the gathering of information is seen as an integral part of the health care consumption experience, it may well be that overall dissatisfaction with health services also includes unhappiness with the information about health care as such information has been found to be both incomplete and inaccurate [15, 34, 35].

While it may be that people are displeased with the Internet as the source of information as such, rather than the actual content, it has been noted that the information available to compare caregivers – frequently various types of quality indicators and performance data – is not always what the public is looking for [11]. Furthermore, recent analyses from Sweden suggest that the demand for various types of provider and health care information varies considerably by gender, education, and health status [17, 36]. For example, those with higher education report being more interested in factors such as provider competences and ownership of clinic compared with those with lower education who say they seek information on the clinical quality of the providers [36].

Regardless of the mechanisms, the findings speak to a general need to consider carefully how health authorities and others inform their citizens about the choice of health care provider and their relative quality. In a recent review of patients’ attitudes toward internet based health information sources in the Netherlands, Hendrikx et al. [37] noted the importance of designing such sources in a clear and useful manner. In addition, international analyses stress the importance of the possibility of being able to make an informed provider choice and its effect on satisfaction with health care [11, 38, 39]. Similarly, from the perspective of patient choice policies, Curtice and Heath [21] report that people in the UK generally do want to be able to choose their provider, but also that those who are more in favor of choice are also less satisfied with the health services in the UK. These and similar findings point at the need for an improved understanding of the complexities involved in consuming health care.

One possibly important, although seemingly overlooked, factor in understanding the impact of health information and its source, on the one hand, and outcomes, such as satisfaction and confidence in provider, on the other, is the way that being more informed about one’s illness and different treatment alternatives changes the expectations that one has on treatment experiences [40, 41]. Although it is beyond the reach of this study to assess the effect of expectations, much evidence has been generated about the important role that individuals’ expectations play in understanding patient satisfaction [42, 43]. An important hypothesis worthy of further investigation is the role that pre-consultation information plays in modifying the subsequent treatment experience. Recent studies have suggested that this involves understanding the complex interplay between information, its source, and actual patient-provider interaction [44, 45].

Use of IBIS

The finding that relatively few people appear to use the Internet as a source of health care information is an important one and is in line with those of other studies [17, 22, 46]. In the current choice-based reforms across European and other health systems, much of the information needed to make an informed choice across provider options is Internet-based. In addition, in the case of Sweden, the way individuals choose a provider differs from one county to another.⁴ For example, in the county of Skåne, people are asked to make the provider choice using two alternative methods, both of which are online-based and one which requires filling out a form (for signing and submission as hard-copy) available on the Skåne region web-page.⁵ While it may seem rational to develop web-based systems for citizens to make their provider choices, it also risks excluding those with no or reduced access to the Internet, limited skills in using such systems, and those who refrain from accessing the web-pages for other reasons.

A general observation is that as the ability to choose health care provider increases, the demand for relevant information on critical aspects of such a choice also increases. However, Glenngård et al. [46] note that making a choice does not necessarily mean switching provider, but choosing to stay with the current one. If such a “passive choice” is the result of resorting to the default option, this could decrease the perceived need for and use of information, regardless of the source. However, if this is the case, the purpose of the reform is somewhat defeated as the key underlying assumption on which the choice-reforms are based is that well-informed consumers of health care move to those providers that are of higher quality and leave those of lower quality [17]. This and other studies point to the need for further investigations in these and related matters of availability of information, patients’ actual choices, and market changes [47].

Several previous studies, both Swedish and international ones, have noted the relatively low levels of Internet use for health care information. Dixon et al. [48] show in their report on patient choice in the UK that only eight percent of patients asked had used the internet to find health information and only four percent used the government funded site “NHS choices” (at www.nhs.uk), created specifically to aid in the choice making process in that system. Furthermore, in the literature surrounding health status and information seeking behavior, it is noted that having a health problem can interfere with the information seeking process because of the energy it takes [49]. Searching for and interpreting information is such a laborious task that to willingly make the effort, there has to be a certain urgency to the situation [11, 50] and this urgency is often not associated with a planned choice of primary care provider. More generally, studies have found a certain level of “choice fatigue”, as choosing one’s PHC-provider is only one of many choices citizens are expected to make across multiple social and economic domains [36].

Profile of users

While the finding that the use of the Internet to seek health care information is low in general is relevant in and of itself, actual users of this source of information also vary by demographic indicators and by socioeconomic status. This finding has also been reported elsewhere and is relevant to health authorities when designing and disseminating information to citizens about their rights and options to choose provider. The existing evidence suggests that the demand for and use of information sources is largely dependent on an individual’s socioeconomic status, age and health status [51]. The socioeconomic status often refers to length of education and the consensus is that having more education has a positive influence on the information seeking behavior [51, 52]. In addition, more educated persons are usually better at interpreting the available information, all of which would suggest that better educated people obtain an advantage over less educated when it comes to making an informed choice of health care.

It is also stated that the elderly are less prone to use the internet as a source for their health information needs [46, 53]. Age was a factor that provided a paradoxical relationship to information use. In terms of use of sources like the Internet, elderly were described in the literature to be less likely to use them than the younger generation. Lack of knowledge of how to work a computer and the internet as well as distrust towards information found on the Internet may create a barrier for use [17, 51]. The elderly are often described as more concerned with their health and provider options than are the young and, furthermore, the elderly are more positive towards receiving health care information. Gatto and Tak (ibid.) describe how Internet use among the older generations may change as they gradually bring computer skills from their work into retirement with them, providing a possible explanation to the increase in older IBIS users. If this trend continues, the overall number of IBIS users is likely to increase over the years as also the older generation becomes more skilled in computer related tasks.

Health status has also been proposed to influence the act of seeking information [49]. However, the way a person’s health affects her demand for health information most likely differs whether it is a more urgent episode of disease or if the person suffers from a chronic illness. It might be postulated that an urgent need for medical care would prompt an interest in learning of treatment options and possible providers and the Internet might be one source of such information. A person suffering from a chronic illness may, on the other hand, be more interested in learning of alternative treatments from other sources, such as personal communication with a health care staff or other persons with similar experiences.

One additional aspect that appears to be mostly overlooked in the data is the issue of availability of health care information in minority languages. In Sweden the information about choosing a provider is also available in about ten languages other than Swedish.⁶ However, this and other studies from Sweden [17] have not been able to shed light on the ethnic profile of Internet users. Given the relatively high shares of immigrants in many European countries, this would seem to be an issue of relevance, including from an equity perspective [54].