Healthcare Quarterly

Healthcare Quarterly 20(3) October 2017 : 47-51.doi:10.12927/hcq.2017.25288
Digital Technologies in Healthcare

Monitoring Receptivity to Online Health Messages by Tracking Daily Web Traffic Engagement Patterns: A Review of More than 13 Million US Web Exposures over 1,235 Days

Neil Seeman and Bob Seeman

Abstract

Reaching the recipient of online health messages is necessary to Web-based health promotion applications. To measure willingness to adhere to a health-related Web message, we explored the frequency with which more than 13 million Web users ignored or opted to receive a random inbound message. The findings suggest declining curiosity among Web users about online messages, and that certain days may be more propitious than others for communicating with users. This approach can be modified to gather more granular insights into how messages, including timing and design features, can be tailored to promote improved public health messaging.

Introduction and Context

The US Internet-using public, which is the population parameter for this study and analysis, sends more than 3.5 million texts each minute. In that same minute, YouTube™ users share roughly 400 hours of new video, users watch 7 million videos and, at this pace, mobile connections alone in the USA will generate more than 18 million megabytes of data every minute within five years (Domo 2016).

How can preventive health and health promotion messages resonate in an increasingly connected world where Web users enjoy so many data feeds, such a variety of digital information, and where users receive these data continually?

An effective and bespoke preventive health, health promotion, drug monitoring or drug adherence message distributed to a Web user through an app or other online modality requires capturing a Web user's increasingly fractured attention online. This means that the sender of the health-related information, among other things, requires advance insights into the ideal timing for message delivery, into the speed and mode of delivery (e.g., video, social media or text) and into the frequency with which the message is best delivered.

A torrent of text messages, push alerts, social media messages and other noisy notifications can undermine the objective of the message; the more intrusive the messaging, the more inured the user becomes to the content and the more likely the user is to disable the notifications or delete the app. Enabling easier reach to the recipient, therefore, is a necessary precondition to the ultimate behaviour modification or monitoring goals that the health application seeks.

Objective

To measure natural rates of online human "interruptibility," or willingness to adhere to a health-related Web message, we explored the frequency with which more than 13.1 million US Web users, from December 11, 2013 to May 3, 2017, either ignored or opted in to a random inbound message – either advising about a health concern or asking about socioeconomic status – preceded by a question about age and gender. We were interested in the changing frequency of opt-ins to the survey, and, especially to the message. Frequency alteration is a proxy for the Web-using public's changing receptivity to online healthcare messaging.

Methodology

For this study of online health message engagement trends, we deployed Random Domain Intercept Technology™ (RDIT™), by which Web users encounter random opt-in inbound messages, surveys or videos after they manually type in a lapsed or dormant Website destination (e.g., phonyurl.com) into the Web address (URL) bar (Seeman 2011; Seeman et al. 2010, 2016). The speed with which potential message respondents land on an RDIT-controlled website – that is, one that is not a "pop-up" or a "pop-under" but a real registered non-trademarked website, such as www.posttcards.com – uniquely enables data scientists to collect behavioural response data from infrequent survey or message test respondents whose reactions and response times are rapid and intuitive under this method. This distinguishes the resulting data from all other online survey or message testing data, which are almost universally collected online from habitual, panel-based respondents who are self-selected members of a "panel." Members of a panel are persons pre-recruited via social media or recruited on gaming sites and elsewhere on the Web enticed by cash or rewards, and who, on average, complete 40 online surveys per month for various research organizations; they answer paid surveys for 29 minutes daily (Kelly and Rico 2015).

Unlike other digital survey or message testing methodologies that collect personally identifiable information to better provide incentives and thereby improve opt-in rates, the method described here recruits only random anonymous respondents who provide no personally identifiable information. The messages or digital campaigns are hosted on real, registered, non-trademarked websites that contain no malware and respondents are not contacted in any way. Because message test links are not e-mailed, as is the case with most panel survey solution companies, it is not possible to hack into respondent e-mails.

Because of this total anonymity, responses are not influenced by social desirability bias as they are conducted via telephone, face-to-face and online panel surveys (Fisher 1993).

Governments cannot shut down ephemeral, scattered and changing websites controlled by the global RDIT router system, as they can do with Facebook™ or Twitter™ or other popular website destinations, and RDIT is not susceptible to the increasing prevalence of ad block technologies.

Given this unique approach to data collection from randomly exposed Web users (Seeman et al. 2010; Seeman and Seeman 2010), the statistical likelihood of any Web user in the US encountering the opt-in question is the same, irrespective of receiving device, browser or Operating System (Seeman 2015; Seeman et al. 2016, 2017). The opt-in question format remained the same for the duration of the study – i.e., it first asked the anonymous user to click on his or her age and gender. Further, the layout, the visual design of the opt-in question, the sampling frame (the proportion of US Web users capable of being interrupted), and the latency (i.e., the speed with which any user was presented with the message) remained stable over the 1,235 days. No personally identifiable information was collected. All Internet Protocol (IP) addresses were scrubbed and transformed automatically into unique identifiers. IP geo-location software (MaxMind™) and longitude and latitude geographic region validation (at a city- and region-level) were conducted using the Haversine formula (Seeman and Ing 2013) to ensure that the Web users randomly intercepted were located in the US at the time of interception and were not using a proxy server that could potentially log them as being outside the country. Using the Haversine formula, the distance between the target city centre and an observation (i.e., an exposed individual to the message) can be calculated from their latitudes and longitudes. Latitude and longitude data are readily available from the aforementioned geo-location database provider.

Findings

Of the 13,138,782 Web users randomly interrupted from December 11, 2013 to May 3, 2017, a daily average of 4.69% opted to complete either a brief survey or to see a message online. Major findings were: (a) a general downward opt-in percent rate over the 1,235 days, suggesting a declining curiosity among Web users about all online messages no matter what the modality (e.g., video treatment exposure, visual imagery or text); (b) consistent inter-daily variability in opt-in trends, suggesting that certain time periods may be more propitious for communicating with Web users and, (c) a regular seven-day cycle (i.e., there is a spike and a dip in the actual trend every week) within the data period that suggests a pattern of recurring frequencies, wherein the weekly spike typically takes place on either a Saturday or a Sunday in the US, with the dips typically taking place at the mid-point of the week.

More generally, the results of this study suggest that this novel Web traffic analysis is capable of being expanded to cover any region of the world, and can be restricted to specific device types or Operating Systems. In this way, it would be possible to gather more granular insights into how specific messages, including their timing and design features, might be tailored to promote prevention in healthcare or public health messaging.

Figure 1 illustrates the changing percentage of opt-in over the 1,235 days, with a low of 2.37% on Day 931 (July 1, 2016) and a high of 9.43% on Day 456 (March 12, 2015). We then replicated the 1,235-day time-series signal with sinusoidal functions – using the Fourier Transform function used to detect patterns in signal processing – and we re-analyzed the same time-series to detect what, if any, frequencies were dominant within the period. We observed that there is a regular seven-day cycle – i.e., there is a spike and a dip in the actual trend every week within the data that suggests a pattern of recurring frequencies. The weekly spike generally takes place on either a Saturday or a Sunday in the data, with the dips typically taking place at the mid-point of the week. The weekly trend does not explain the aforementioned outlier dates, which are more likely explained by irregularities in Internet usage (e.g., regional outages, in the case of an outlier dip, or a major breaking news event, in the event of an outlier spike).


Click to Enlarge
 

This general weekly cycle trend is illustrated in Figure 2, where f = 1/T (T = time). Figure 2 shows this weekly periodicity extracted from the 1,235-day time series, segmented into seven-day cycles, which, in turn, validates the more general observation of this paper that there are potentially predictable times when a Web user is more or less receptive to adhering to an online health message.


Click to Enlarge
 

Discussion

The approach used in this study measures Web user interruptibility patterns without the need for conventional focus groups, or for biosensor data or for any mechanical attachment to a human. The approach can measure receptivity to health messages globally and provides stable baseline data that can thereafter assess changes over time. Thus, message receptivity can be quantified based on time of day, day of week, day of month, day of year or season. These measurements can be positioned to country, city, region, self-reported age, sex, ethnicity, chronic health condition, to the stage of a diagnostic spectrum (e.g., pre-diagnosis, at-risk, recovery), device type or to Operating System.

Our approach described in this paper is not invasive, does not require predictive algorithms derived from small samples and continually updates, without storing personally identifiable information. It informs about the receptivity of different segments of the population (e.g., self-reported age and gender) according to device used, time of day, week, city or sub-city region. A consent-based "cookie" placed on the exposure Websites (which was not done in this study) that "follows" the interrupted user across different websites could theoretically enable further, in depth reporting about the habits and characteristics of the Web user and obtain even more granular segmentations. For example, the following questions could be addressed: What is the health message pattern receptivity of Web users who do, versus those who do not, visit health-related educational websites on a frequent basis? Is the receptivity of Web users who predominantly browse the Web for work-related purposes qualitatively different from that of respondents who predominantly use it for social connectedness?

There have been attempts to deliver health promotion messages over the Web, some reportedly successful whereas others less so (Webb et al. 2010). The relative success of public health interventions has been variously attributed. Some stress the salience of the message and emphasize design features such as simplicity (Peters et al. 2015) or length (shorter is better) (van Genugten et al. 2016). Special techniques such as gamification (goal-setting, self-monitoring, incentives, rewards) have been advocated as benchmarks of efficacy (Edwards et al. 2016).

Using the right combination of behaviour change techniques is reportedly crucial (Dusseldorp et al. 2014). The mode of delivery (whether by video, live chat, testimonials, games, newsletters) has been evaluated (Webb et al. 2010). Issues associated with targeting specific populations have also been described (McCabe et al. 2010). Some interventions have been reported as exerting large positive effects on behaviour, notably those that have targeted physical activity (Carr et al. 2008; Kim and Kang 2006). Still others have had negative effects (Frosch et al. 2003; Forbes et al. 2017) attributed by researchers to the targeted population's lack of motivation for change. Some interventions have been linked to positive but modest change (Schippers et al. 2017). Whatever the approach, however, one element is common to all online interventions: a requirement that the intended recipient of the message is actually reached, and is reached repeatedly (Schäfer et al. 2012).

Very few studies have considered the important issue of when individuals are at their most receptive points to messaging in the context of Web-based health promotion or prevention. We found one study identifying the best time to offer a free travel card to encourage people to abandon cars and use public transportation (Thøgersen 2012). We also found references to when it is best to keep silent during bereavement counselling (Back et al. 2009) and to the critical importance of timing when offering psychological interpretations during psychotherapy sessions. In psychotherapy, it is well established that an interpretation delivered at the wrong time falls on deaf ears (Holmes 1997; Leitner 2001; Waldron et al. 2004). The issue of appropriate timing has also been recognized in public health messaging about vaccination (de Montigny et al. 2013) and nutrition (Rowe and Alexander 2011).

Conclusion

Our conclusion, extrapolated from the data revealed in this paper, in the clinical healthcare context and the limited literature on interruptibility and receptiveness to online messages, is that the health-related Web communications that work best are those that are delivered when consumers are most prepared to receive them. We therefore have introduced a novel way to gather intelligence on this time-identification challenge at a global, voluminous scale, and with granularity, in the interests of public health promotion and personalization. In the interests of personalizing healthcare, it is important to discover when a person is sufficiently interested (i.e., in the right mood and state of readiness) to receive reminders, instructions, recommendations and warnings (Knox and Cooper 2011).

Author disclosure

The authors are directors of RIWI Corp. (CSE: RIW), which owns the patent for Random Domain Intercept Technology™, which was used to collect the data in this study. Neil Seeman, the chairman and chief executive officer, is the holder of certain non-voting preferred shares of RIWI Hold Inc., which owns equity in RIWI Corp. Neither Neil Seeman nor Bob Seeman exercise control over the securities held by RIWI Hold Inc. This study was conducted in the public interest, with all funding and associated labour costs borne by RIWI Corp. RIWI Corp. is a global survey, message testing and predictive analytics firm. RIWI does not provide any mobile health services or products to any of its current, past or intended customers.

About the Author(s)

Neil Seeman, JD, MPH, is chairman and chief executive officer of RIWI Corp., a senior fellow of Massey College and adjunct lecturer in the Institute of Health Policy, Management and Evaluation, University of Toronto.

Bob Seeman, BASc, JD, MBA, is chief executive officer of Clera Inc.

References

Back, A.L., S.M. Bauer-Wu, C.H. Rushton and J. Halifax. 2009. "Compassionate Silence in the Patient–Clinician Encounter: A Contemplative Approach." Journal of Palliative Medicine 12(12): 1113–17. doi:10.1089/jpm.2009.0175.

Carr, L.J., B.R. Todd, C. Dorozynski, J.F. Broomfield, M.L. Smith and D.T. Smith. 2008. "Internet-Delivered Behavior Change Program Increases Physical Activity and Improves Cardiometabolic Disease Risk Factors in Sedentary Adults: Results of a Randomized Controlled Trial." Preventive Medicine 46(5): 431–38. doi:10.1016/j.ypmed. 2007.12.005.S0091-7435(07)00503-8.

de Montigny, L., K. Charland, A. Verma, J.S. Brownstein, P. Le Guerrier and D.L. Buckeridge. 2013. "Predictors of the Timing of Vaccination Uptake in the 2009 Influenza Pandemic (H1N1) in Montreal." American Journal of Preventive Medicine 45(5): 622–28.

Dusseldorp, E., L. van Genugten, S. van Buuren, M.W., Verheijden and P. van Empelen. 2014. "Combinations of Techniques that Effectively Change Health Behavior: Evidence from Meta-CART Analysis." Health Psychology 33(12): 1530–40. doi:10.1037/hea0000018.

Edwards, E.A., J. Lumsden, C. Rivas, L. Steed, L.A. Edwards, A. Thiyagarajan et al. 2016. "Gamification for Health Promotion: Systematic Review of Behaviour Change Techniques in Smartphone Apps." BMJ Open 6(10): e012447. doi:10.1136/bmjopen-2016-012447.

Fisher, R.J. 1993. "Social Desirability Bias and the Validity of Indirect Questioning." Journal of Consumer Research 20(2): 303–15. doi:10.1086/209351.

Forbes, C.F., C.M. Blanchard, W.K. Mummery and K.S. Courneya. 2017. "A Pilot Study on the Motivational Effects of an Internet-Delivered Physical Activity Behaviour Change Programme in Nova Scotian Cancer Survivors." Psychology and Health 32(2): 234–52. doi:10.1080/08870446.2016.1260725.

Frosch, D.L., R.M. Kaplan and V.J. Felitti. 2003."A Randomized Controlled Trial Comparing Internet and Video to Facilitate Patient Education for Men Considering the Prostate Specific Antigen Test." Journal of General Internal Medicine 18(10): 781–87. doi:10.1046/j.15251497.2003.20911.x.

Holmes, J. 1997. "'Too Early, Too Late': Endings in Psychotherapy – An Attachment Perspective." British Journal of Psychotherapy 14(2): 159–71.

James, J.; Domo. 2016. Big Data Never Sleeps 4.0. Domo Inc. Retrieved May 9, 2017. <https://www.domo.com/blog/data-never-sleeps-4-0/>.

Kelly, F. and M. Rico. 2015. "Quick Quality Quantitative: IIeX Conference." Presentation by F. Kelly, Atlanta, GA, Insights and Information Exchange, June 30, 2015. <http://insightinnovation.org/wp-content/uploads/2015/06/PDF/quality.pdf>.

Kim, C.-J. and D.-H. Kang. 2006. "Utility of a Web-based Intervention for Individuals with Type 2 Diabetes: The Impact on Physical Activity Levels and Glycemic Control." Computers, Informatics, Nursing 24(6): 337–45. doi:10.1097/00024665-200611000-00008.00024665-200611000-00008.

Knox, R. and M. Cooper. 2011. "A State of Readiness: An Exploration of the Client's Role in Meeting at Relational Depth." Journal of Humanistic Psychology 51(1): 61–81.

Leitner, L.M. 2001. "Experiential Personal Construct Therapeutic Artistry: The Therapy Relationship and the Timing of Interventions." Humanistic Psychologist 29(1-3): 98–113. doi:10.1080/08873267.2001.9977009.

McCabe, M.P., L.A. Ricciardelli and G. Karantzas. 2010. "Impact of a Healthy Body Image Program among Adolescent Boys on Body Image, Negative Affect, and Body Change Strategies." Body Image 7(2): 117–23. doi:10.1016/j.bodyim.2009.10.007.

Peters, G.-J. Y., M. de Bruin and R. Crutzen. 2015. "Everything Should Be as Simple as Possible, but No Simpler: Towards a Protocol for Accumulating Evidence regarding the Active Content of Health Behaviour Change Interventions." Health Psychology Review 9(1): 1–14. doi:10.1080/17437199.2013.848409.

Rowe, S. and N. Alexander. 2011. "The 7 Cardinal Sins in Nutrition Communication." Nutrition Today 46(6): 276–80. doi:10.1097/NT.0b031e31823acfb4.

Schippers, M., P.C.G. Adam, D.J. Smolenski, H.T.H. Wong and J.B.F. de Wit. 2017. "A Meta-Analysis of Overall Effects of Weight Loss Interventions Delivered via Mobile Phones and Effect Size Differences according to Delivery Mode, Personal Contact, and Intervention Intensity and Duration." Obesity Reviews 18(4): 450–59.

Schäfer, M., M. Jaeger-Erben and S. Bamberg. 2012. "Life Events as Windows of Opportunity for Changing towards Sustainable Consumption." Journal of Consumer Policy 35(1): 65–84.

Seeman, N. 2011. "Method of Obtaining a Representative Online Polling Sample." US Patent 8069078 B2, Published: November 29, 2011.

Seeman, N. 2015. "Use Data to Challenge Mental-Health Stigma." Nature 528(7582): 309. doi:10.1038/528309a.

Seeman, N., S.G. Fogler and M.V. Seeman. 2016. "Mental Health Promotion through Collection of Global Opinion Data." Journal of Preventive Medicine and Care 1(1): 23–36.

Seeman, N. and A. Ing. 2013. 2013 GRIT Consumer Participation in Research Report: A Study of Survey-Takers in 200+ Countries and Regions around the World. New York AMA Communication Services and Greenbook. Retrieved April 20, 2017. <https://riwi.com/wp-content/uploads/2015/06/GRITCPR2013_SeemanIng.pdf>.

Seeman, N., A. Ing and C. Rizo. 2010. "Assessing and Responding in Real Time to Online Anti-Vaccine Sentiment during a Flu Pandemic." Healthcare Quarterly 13(Special Issue): 8–15.

Seeman, N., D.K. Reilly and S. Fogler. 2017. "Suicide Risk Factors in U.S. College Students: Perceptions Differ in Men and Women." Suicidology Online 8: 24–30.

Seeman, N. and M.V. Seeman. 2010. "Autism and the Measles, Mumps, and Rubella Vaccine: Need to Communicate a Health Study Retraction to Patients." Journal of Participatory Medicine 2. Retrieved May 9, 2017. <http://www.jopm.org/evidence/research/2010/12/17/autism-and-the-measles-mumps-and-rubella-vaccine-need-to-communicate-a-health-study-retraction-to-patients/>.

Seeman, N., S. Tang, A.D. Brown and A. Ing. 2016. "World Survey of Mental Illness Stigma." Journal of Affective Disorders 190: 115–21. doi:10.1016/j.jad.2015.10.011.

Thøgersen, J. 2012. "The Importance of Timing for Breaking Commuters' Car Driving Habits." Collegium 12: 130–40.

van Genugten, L., E. Dusseldorp, T.L. Webb and P. van Empelen. 2016. "Which Combinations of Techniques and Modes of Delivery in Internet-Based Interventions Effectively Change Health Behavior? A Meta-Analysis." Journal of Medial Internet Research 18(6): e155. doi:10.2196/jmir.4218.

Waldron, S., R. Scharf, J. Crouse, S.K. Firestein, A. Burton and D. Hurst. 2004. "Saying the Right Thing at the Right Time: A View Through the Lens of the Analytic Process Scales (APS)." Psychoanalytic Quarterly LXXIII(4): 1079–125. doi:10.1002/j.2167-4086.2004.tb00193.x.

Webb, T.L., J. Joseph, L. Yardley and S. Michie. 2010. "Using the Internet to Promote Health Behavior Change: A Systematic Review and Meta-Analysis of the Impact of Theoretical Basis, Use of Behavior Change Techniques, and Mode of Delivery on Efficacy." Behavior Journal of Medical Internet Research 12(1): e4. doi:10.2196/jmir.1376.

Comments

Be the first to comment on this!

Note: Please enter a display name. Your email address will not be publically displayed