Question 1: how do people use different social media sites?

One potential reason for the mixed findings across studies of social media sites is that people visit various sites for different reasons. For instance, a Yelp reviewer visits the website specifically to post a review. A visitor to a nursing home or hospital Facebook page might be gathering information, ‘Like’-ing something or reading a post. Thus, the mixed evidence regarding associations with online ratings and outcomes could result in part from the different uses of the websites involved. For instance, associations between online review scores and patient experience and outcome measures have more consistently emerged from analyses of online review sites, such as Yelp or the National Health Services patient review site, for which the only reasons for visiting the sites are to post or find reviews. It may be that people visiting websites for other, more positive, reasons, who happen to then leave a review, are more likely to leave positive reviews. This may lead to a more limited association with outcomes, due to an overall more positive set of reviews. The findings of Gaudet Hefele and colleagues support this hypothesis, as the Facebook scores were an average of 4.5, whereas in prior work on Yelp reviews of hospitals the average star score was 3.3.1

In another recent study in BMJ Quality & Safety, Griffiths and Leaver14 found that aggregated scores across multiple different consumer-generated sources (NHS Choices, Patient Opinion, Facebook and Twitter) predicted in-person Care Quality Commission inspection outcomes for hospitals with exceptionally high accuracy.14 This study demonstrates that aggregating patient experience ratings offers a useful approach to help guide patients in choosing providers and even potentially regulators and payors. However, as with any composite approach, we need to understand the contribution of the individual components. Better data on website users’ intentions will allow more nuanced interpretations of ratings from each website and may help us understand which sites contain more valid data on healthcare quality.

Question 2: what are the strengths and weaknesses of different types of reviewers?

Another important area for exploration is the distinct perspectives that different reviewer types may bring. For instance, some evidence suggests that friends or family members more often comment about patient safety than patients.15 This finding, also presented in greater narrative detail in Atul Gawande’s book ‘Being Mortal’,¹⁸ makes intuitive sense since family members are often more awake and alert during a hospitalisation and so potentially more likely to be detail-oriented about safety issues. They are also the ones who leave the bedside to go home and worry about loved ones remaining in the hospital. Online review websites could consider gathering data on reviewers’ relationships to the patient, in order to do exploratory analyses regarding the associations with outcomes for different reviewer types.

Public reporting websites in the USA such as Hospital Compare and Physician Compare could also consider incorporating these findings into their approach to measuring patient experience. Their current approach primarily seeks patient feedback only. Although the nursing home data presented by Gaudet Hefele et al 9 come from caregivers, neither the domains for nursing home patient experience surveys nor for HCAHPS and CAHPS include safety. Given the recent findings that patients and family members report more medical errors, verified by physician review, compared with healthcare provider-initiated reports (5% of the verified patient-reported errors),16 this represents a missed opportunity to better understand and address the persistent problem of patient safety.

It is also likely that patients with more experience (eg, a patient with congestive heart failure and multiple hospital admissions) have different types of observations than patients with less experience17 (eg, a previously well patient admitted for acute appendicitis or first-time heart attack). The differences in observations may manifest in two ways. Experienced patients and their caregivers, with a larger ‘dataset’ of observations, as well as intimate knowledge of their own disease and situation, may more easily distinguish events that deviate from a pattern. This may make them more likely, for instance, to identify a medication error as well as potentially more likely to note it in a review since they are confidently familiar with their standard medication dosing. In addition, experienced patients and their caregivers, having seen multiple aspects of the healthcare delivery system and often having sought care at more than one system, may more easily identify systems issues. For instance, in a local family advisory council meeting, a parent compared two hospitals where her child had been admitted, remarking that one had better safety equipment to decrease the risk of falls for her child, and that it would make her feel safer if both hospitals had the better equipment.

The implications of potential differences in online reviewers by level of experience are two-fold. Online narratives from more experienced patients may provide system-level observations and savvy advice for otherwise potentially less experienced online consumers. In addition, this heterogeneity of perspectives may partly explain the mixed results regarding the association of online reviews and provider quality. If the mix of experience varies across websites (eg, Facebook has fewer experienced patients leaving reviews compared with Yelp), the websites with more experienced reviewers may have an increased number of more savvy observations that inform and drive the relationship between outcomes and review scores. Whether such a difference in observations between experienced and inexperienced patients exists represents a testable hypothesis for future research. If confirmed, then websites may want to consider amplifying the voices of experienced patients, or more proactively eliciting reviews from them.

In summary, despite mixed evidence as to the validity of online reviews, their influence remains and will likely continue to grow. Thus, we need to better understand how websites with online reviews differ from one another and the strengths and weaknesses of different types of reviewers. Ideally such research would occur in partnership with online review website sponsors. Currently, the vast majority of providers probably ignore online reviews. This option will probably become less tenable in the coming years. Even if it were, though, we will better serve patients if we learn to harness the potential power of these increasingly available sources of patient experience data to improve healthcare delivery systems.