Courses


INFS 892: Health Informatics Research
Literature Review #3: Mobile Computing in Health Care Settings
Cory Allen Heidelberger
April 25, 2011

Introduction

Only a few years ago, discussion of meaningful use of electronic health records on mobile devices would likely have conjured images of rolling workstations and conventional laptop computers. In the last few years, the form factor for common mobile computing devices has folded down significantly, as touchscreen-enabled smartphones and tablets have experienced faster adoption that most previous technological innovations. Making meaningful use of EHR systems still requires systemwide investment in technology and training, but it can now take advantage of personal mobile computing devices widely adopted by health care professionals for their personal use. “Besides the obvious benefits of always-on, ubiquitous connectivity, it leverages something most doctors and patients already own — a mobile phone or, increasingly, a tablet. This is why we have seen so many healthcare solutions coming out that incorporate wireless” (Lewis, 2011b). This multiply advantageous form factor may not “crush” laptops in the healthcare space, as one champion proclaimed one month after the iPad hit the market (Merrill, 2010a), but new mobile computing tools are enjoying remarkably fast adoption in the healthcare field.

With the meaningful use verification period for the CMS EHR Incentive Program opened Monday (“‘Meaningful Use’ Reporting Period Starts for Electronic Health Records Program,” 2011), it is thus important and timely to see if adoption of mobile computing devices by health care professionals may help health care incorporate health IT in their regular service provision. This paper reviews current rates of adoption, uses, and motivating factors of mobile computing technology by health care practitioners.

Earlier Mobile Computing Devices

Mobile computing preceded smartphones and tablets in the form of personal digital assistants. 26% of U.S. physicians used PDAs in 2001 (Lu et al., 2003). However, at that time, PDA users in health care settings faced more barriers. Many more institutions lacked integrated information systems that could provide PDA users access to patient databases. Screen size and resolution, poor interfaces, short battery life, and hardware fragility hindered usability (Lu et al., 2003). Studies expected PDA use to rise rapidly (Lu, Xiao, Sears, & Jacko, 2005), but the above barriers prevented the devices from gaining dominance in the health care industry.

Laptop computers with tablet capability brought some expanded interface capacity. However, these devices posed challenges the same problem as PDAs with poor battery life, with the greater weight and bulk compounding usability problems. Older tablets also generally ran operating systems designed for use with keyboard and mouse, making operation in tablet mode unwieldy (Robertson, Miles, & Bloor, 2010).

Current Mobile Adoption

Now mobile computing tools appear to be rising to meet the hopes expressed for mobile computing years ago. Current literature echoes past promises of halcyon days for mobile computing—smartphone technology, for instance, is called a transformative, paradigm-changing technology (Bottles, 2011). However, current adoption and use rates appear to give ground for such statements. Smartphones are leading the way in modern penetration of mobile computing technology in the health care field. Knowledge Networks survey finds 64% of doctors have a smartphone. 27% of primary care providers and specialists have a tablet, an adoption rate five times the rate for the general population. That statistic is all the more remarkable given that health care practitioners generally lag other industries and the general population in information technology adoption. Manhattan Research mid-2010 found 72% of U.S. physicians have a smartphone or PDA; Chilmark Research finds 22% of physicians with iPad by end of 2010 (Dolan, 2011). A 2010 PricewaterhouseCooper survey found 63% of physicians are already using personal devices for mobile health solutions (Lewis, 2010b).

At the institutional level, in fall 2010, one IT enterprise supplier found that healthcare institutions constituted 10% of iPad-deploying customers, the third-largest segment among its clients, just behind the technology sector at 11% and well behind the leading financial services sector at 43% (Merrill, 2010b). An October 2010 HIMSS survey found 70% of responding members planning to deploy iPad and other Apple mobile devices in their health care facilities by the end of 2011 (Lewis, 2010a). Interestingly, 30% of physicians in a survey last year reported that, benefits notwithstanding, their hospital or practice leaders would not support the use of mobile devices (Lewis, 2010b).

Operating systems adopted show some significant differences between health care practitioners and the general consumer population. Nielsen numbers on the broader consumer market show devices running the Android operating system take up 48% of the market. iOS holds 31%, while Blackberry OS claims 18% (Zeman, 2011). A Bulletin Healthcare analysis (Hirsch, 2011) of its health professional subscribers found a much different breakdown of OS choice. It found 93% of its subscribers who use mobile devices use Apple-based machines. iPhone users decreased from an 86% share of the mobile audience in this group in June 2010 to 79% in February 2011. iPad users rose over the same period from 8% to 14%. Android doubled its share of users to 6%. This study found the three in ten subscribers used mobile devices to read Bulletin healthcare’s daily email briefings, showing that physicians use mobile devices for basic information seeking and ongoing professional learning.

The Bulletin Healthcare survey also found notable differences in mobile device adoption rates among different specialties:

Specialty

Mobile Adoption

Physician Assistants

41%

Emergency Room Physicians

40%

Cardiologists

33%

Urologists

31%

Nephrologists

31%

Dermatologists

30%

Gastroenterologists

30%

Psychiatrists

28%

Optometrists

28%

Radiologists

24%

Rheumatologists

22%

Endocrinologists

21%

Oncologists

20%

Clinical Pathologists

16%

Table: Mobile device adoption rates among different health specializations (Hirsch, 2011)

These different rates appear to be higher among more mobile practitioners like P.A.’s and ER physicians and lower among more lab-based analysts like radiologists and pathologists who may travel less in their facility in the course of their normal workflow.

Tools Available

Mobile computing funtionality ranges from access to integrated EHR solutions to standalone applications for communication, reference, and administration. DrChrono introduced its iPad EMR the same month the iPad launched; by September, ClearPractice followed with the Nimble EMR application (Merrill, 2010a). The MyChart Apple application lets patients access medical history, appointment times, and lab results. The University of Iowa Hospitals and Clinics are testing Canto (for iPad) and Haiku (for iPhone) for likely implementation in October with an upgrade of the Epic EHR system (Bennett, 2011).

The Emergency Medical Spanish Guide was developed by an EMT whose experience with Spanish-speaking patients led him to create an application that could provide and even pronounce simple yes/no questions in Spanish to bridge language barriers in the ER (Usatine, 2010). That application is now paying for the former EMT’s medical school tuition. A similar application exists to serve French-speaking patients (Castro, 2010). InVivoLink this month released OrthoPod, an iPad app to help orthopedic surgeons access their personal implant registry and see patterns in their procedures and patient data (press release, 2011).

Reference apps like Epocrates and WebMD are most popular mobile apps; apps from pharma manufacturers get little use (Dolan, 2011). As of April, 2010, over 125,000 doctors were using Epocrates on iPhone and iPod touch devices; one study claimed that “60 percent of Epocrates users avoided three or more medical errors a month” (Bottles, 2011). The American Medical Association released its first mobile application this month, a free application through iTunes to allow doctors to look up, track, and organize Current Procedural Terminology billing codes (Gillette, 2011). This release for Apple devices acknowledges the dominance of Apple products in this sector.

Articles on mobile applications provide multiple examples of physicians educating patients with mobile images from Netter’s Atlas of Human Anatomy (Davis, 2011; Porter, 2011). The FDA this February approved Mobile MIM, a mobile radiology application that allows clinicians to share and make diagnoses based on CT, MRI, and PET scans (Lewis, 2011b).

The AirStrip Cardiology app (iPhone, iPad; planned for Android) gets data from GE Healthcare’s Muse Cardiology Information System’s cloud database of current electrocardiograms. Through the AirStrip system, doctors can access ECGs immediately anywhere instead of relying on potentially distorted faxes or PDFs. On mobile screens, physicians can zoom in on small differences, as small as a half-millimeter, that make make a big diagnostic difference but that might be lost in the resolution of a typical fax (Horowitz, 2011). This application complements other apps planned by GE Healthcare to expand mobile EHR access (Lewis, 2011b). AT&T last month announced similar expansion of its cloud-based Healthcare Community Online to include a mobility interface for smartphones (Lewis, 2011b).

Cloud computing and virtualization broaden mobile health IT from the realm of specific mobile apps to full mobile access to an institution’s integrated EHR system. Dell this month released a mobile clinical computing program that allows hospitals using the Meditech health Care Information System to virtualize their system and provide secure access to desktop applications via any number of devies, including tablets (Lewis, 2011c). Virtualized clients also mitigate concerns about information security on personal devices (Lewis, 2010a).

Mobile hardware is improving the capabilities available to doctors. This year’s iPad 2 processes images nine times faster than the original device in 2010, allowing faster processing of more detailed medical images. The dual cameras of the tablet also make it easier for rural doctors to share photos of wounds and conference with remote colleagues (Lewis, 2011b). These capabilities can facilitate communication among radiologists and clincians and expedite diagnosis and treatment (Choudhri & Radvany, 2010). Radiologists who are accustomed to more sednetary workflow in a dedicated imaging lab with specialized equipment may find value in accommodating mobile platforms to support better interaction with referring physicians in patient management teams (Shih, Lakhani, & Nagy, 2010).

Motivations for Mobile Adoption

The largest motivators for this apparent rush to mobile deployment include point-of-care applications, clinical decision support, medical image viewing applications, and general administration. Another survey finds that physicians agree the greatest benefit of mobile health will be the ability to make decisions faster by accessing more accurate data in real time (Lewis, 2010b). The perception overlaps with the perception among more than 60% of physicians that the best benefit of EHR systems is real-time patient information access (staff, 2011).

A Knowledge Networks marketing survey finds physicians relying on mobile units for checking email, researching medications and conditions, and taking surveys. Physicians still prefer in-person visits with drug company representatives over e-marketing—markedly more so among physicans over 55, but still with a strong majority among those under 40 (Dolan, 2011). Nonetheless, Epocrates has just launched a mobile drug-sampling service, offering their 315,000 physician members the ability to order custom samples via mobile devices (Millard, 2011).

Christ Community Health Services in Memphis chose iPads as its main hardware platform for its coming switch to electronic health recorsd. The hospital chose from among numerous technological options based on tests in clinical settings and criteria including battery life, ease of use, and portability. As a result, CCHS Memphis is now preparing to implement 38 iPads in five health care centers for use in all charting. CCHS physician John David Williamson said “that intimate, non-disruptive doctor-patient interaction is the ‘definite advantage of the iPad. Rather than having my back turned to a patient while on a computer, I can actually continue to have a normal interaction with the patient…’” (Maki, 2011). Similar advantages have driven an Ottawa hospital to order 1800 iPads to be in the hands of its physicians, residents, and pharmacists by July (CBC, 2011).

Methodist Le Bonheur Hospital in Memphis started with in-house iPad pilot project and is now testing use of a virtual private network to give authorized iPad users outside the hospital access to medical records (Maki, 2011). Such a move is a logical extension of the mobility allowed by such devices, not just from room to room within the facility, but beyond the walls of the facility. MLBH user cites portability, ease of use, and quick access (no long boot-up, no long log-in and log-out from room to room) as reasons for adoption. Hospitals that have already implemented EHR may also have invested in desktop computers in each room, lessening the need for mobile solutions (Bennett, 2011; Maki, 2011). One official at Stern Cardiovascular Center in Memphis suggests putting desktop on cart to wheel from room to room (Maki, 2011), although physicians who have tried newer handheld mobile solutions might balk at the idea that the traditional computer on wheels (the COW, as it is abbreviated with just a touch of mockery) can provide comparable utility.

Dr. John Halamka, chief information officer of Beth Israel Deaconess Hospital in Boston, puts mobile medical device criteria in concrete terms that argue strongly in favor of iPads as a perfect fit for doctors: “The secret for the ideal clinical device is it has to weigh a pound, it has to last 10 hours, because that’s [a doctor’s] shift, you have to be able to disinfect it so there’s no risk of contamination, and you have to be able to drop it 5 feet onto carpet without damage” (Davis, 2011). Size, weight, and battery life are cited by other users and observers as reasons newer mobile computing devices are gaining acceptance among health care practitioners (Merrill, 2010b). Doctors at Beth Israel Deaconess also say ability to share information with patients is key. Consider the difference between visiting with a doctor who is ensconced a few feet away at a computer terminal, data upon which the patient cannot see, and a doctor bedside, entering data on a screen the patient can see and even touch and manipulate.  Mobile devices create the possibility for more natual bedside communication and interaction (Bennett, 2011) between health care providers and their patients.

Mobile applications provide another opportunity to reduce the separation between health care provider and patient. When tools are downloadable to common platforms like mobile phones and tablets, doctors and patients may be able to use these devices together on their own devices with minimal expense. A psychiatriast may be able to provide information to a visiting patient from an application like the new PTSD Coach, just released this month by the Department of Veterans Affairs and Department of Defense to help veterans with post-traumatic stress disorder (Montalbano, 2011). If the patient has a compatible smartphone or tablet, the psychiatrist can just as easily “prescribe” that the patient install the application on his or her own mobile device and use the application outside the office to manage his or her own symptoms. Unlike a large integrated EHR system accessible only to physicians at hospital or clinic workstations, mobile applications open more possibilities for patient use and interaction with their health care providers and data. Common mobile devices also offer physicians channels by which to provide remote healthcare, an option a majority of consumers and physicians find acceptable (Lewis, 2010b). Such remote monitoring could mark an electronic return to the practice of “house calls” (Lewis, 2010b).

Integrating mobile devices into regular care may also promote patient confidence in the quality of their care. A Sage Healthcare Insights study released this month finds 81% of patients “have a positive perception of documenting patient care electronically” (staff, 2011), a level of positive perception greater than that found among physicans (62%). Patients report greater confidence in physicians who use EHRs and icnreasingly expect physicians to give them access to electronic records and tools (staff, 2011). These findings stand in contrast to concerns expressed that patients may object to their physician answering questions by seeking answers from a machine instead of consulting their own professional knowledge; besides, the push for evidence-based medicine will demand increased use of technology in health care (Shaw, 2011). Also driving expectations of greater health care technology use may be the againg and retirement of baby boomers. This demographic has egnerally lead other groups in technology spending and adoption (Bottles, 2011); as they become senior citizens and increasingly need health care services, they will bring their preferences for technology to that industry. Physicians can respond to such perceptions and expectations by working with EHR data via mobile devices in sight of patients and even showing them the information they are accessing and entering on those devices.

Personal choices and innovativeness appear to be playing an important role in mobile device adoption. At one Memphis hospital, for example, almost all iPads in use are owned by individual physicians (Davis, 2011). Even if their institutions lag in promotion and integration of health information technology, individual practitioners are finding the new generation of smartphones and tablets sufficiently useful to incorporate into their own workflow, even if just for reference and communication. The new mobile technologies are also sufficiently accessible that health care practitioners have designed many of the simple apps available for varios devices.

Conclusion

Past health information research contains various predictions of great technological advances in mobile computing that did not catch on. However, the swift penetration of smartphones and tablets in the heallth care industry, coming on the heels of increased institutional and government investment in EHR, suggest that mobile computing has reached a new level of integrative utility for health care providers.

Works Cited

  1. Bennett, M. (2011, April 13). Doctors use iPads, iPhones for medical records. The Daily Iowan. Retrieved April 19, 2011, from http://www.dailyiowan.com/2011/04/13/Metro/22798.html
  2. Bottles, K. (2011). Physician Executives Should Not Ignore How Smartphones Will Transform Healthcare. The Health Care Blog. Retrieved April 22, 2011, from http://thehealthcareblog.com/blog/2011/01/31/physician-executives-should-not-ignore-how-smartphones-will-transform-healthcare/
  3. Castro, H. (2010). Doc APProvED: A Day in the Life of an iPad. Emergency Medicine News, 32(6).
  4. CBC. (2011, April 20). 1,800 iPads ordered by Ottawa Hospital. CBC News. Retrieved April 22, 2011, from http://www.cbc.ca/news/canada/ottawa/story/2011/04/20/ottawa-ipads-hospital374.html
  5. Choudhri, A. F., & Radvany, M. G. (2010). Initial Experience with a Handheld Device Digital Imaging and Communications in Medicine Viewer: OsiriX Mobile on the iPhone. Journal of Digital Imaging, 24(2), 184-189. doi:10.1007/s10278-010-9312-7
  6. Davis, M. (2011, April 4). For some doctors, the iPad is claiming a key spot next to the stethoscope. Boston Globe. Retrieved from http://articles.boston.com/2011-04-04/business/29380774_1_ipad-tablet-computers-ulcer
  7. Dolan, B. (2011, March 31). Survey: 27 percent of US doctors have tablets. mobihealthnews. Retrieved April 21, 2011, from http://mobihealthnews.com/10627/survey-27-percent-of-us-doctors-have-tablets/
  8. Gillette, B. (2011, April 7). AMA introduces CPT coding app. Modern Medicine. ArticleStandard, . Retrieved April 21, 2011, from http://www.modernmedicine.com/modernmedicine/Cosmetic+Surgery/AMA-introduces-CPT-coding-app/ArticleStandard/Article/detail/715114?contextCategoryId=40174
  9. Hirsch, R. (2011, April 14). Physician Mobile Use Grows 45%; Apple® Dominates Android™ and Blackberry® | Business Wire. Business Wire. Retrieved April 19, 2011, from http://www.businesswire.com/news/home/20110414005326/en/Physician-Mobile-Grows-45-Apple%C2%AE-Dominates-Android%E2%84%A2
  10. Horowitz, B. T. (2011, April 7). AirStrip, GE Launch Cardiology App for iPhone, iPad. eWeek: Health Care IT. Retrieved April 21, 2011, from http://www.eweek.com/c/a/Health-Care-IT/AirStrip-GE-Launch-Cardiology-App-for-iPhone-iPad-590966/
  11. Lewis, N. (2010b, September 10). Mobile Devices To Transform Healthcare. InformationWeek: Healthcare. Retrieved April 21, 2011, from http://www.informationweek.com/news/healthcare/mobile-wireless/227400122
  12. Lewis, N. (2010a, December 10). Healthcare iPad Deployment To Approach 70% In 2011. InformationWeek: Healthcare. Retrieved April 21, 2011, from http://www.informationweek.com/news/healthcare/EMR/showArticle.jhtml?articleID=228800929
  13. Lewis, N. (2011b, March 7). iPad 2 Highlights Mobile Healthcare Advancements. InformationWeek: Healthcare. Retrieved April 21, 2011, from http://www.informationweek.com/news/healthcare/mobile-wireless/229300499
  14. Lewis, N. (2011c, April 12). Dell Launches Meditech Desktop Virtualization. InformationWeek: Healthcare. Retrieved April 21, 2011, from http://www.informationweek.com/news/healthcare/mobile-wireless/229401432
  15. Lu, Y. C., Lee, J. J. K., Xiao, Y., Sears, A., Jacko, J. A., & Charters, K. (2003). Why Don’t Physicians Use Their Personal Digital Assistants? AMIA Annual Symposium Proceedings (Vol. 2003, p. 405).
  16. Lu, Y. C., Xiao, Y., Sears, A., & Jacko, J. A. (2005). A review and a framework of handheld computer adoption in healthcare. International Journal of Medical Informatics, 74(5), 409–422.
  17. Maki, A. (2011, April 14). Medical Breakthrough: Health Care Providers Hope iPad Can Enhance Their Services. Memphis Daily News. Retrieved from http://www.memphisdailynews.com/editorial/Article.aspx?id=57807
  18. “Meaningful Use” Reporting Period Starts for Electronic Health Records Program. (2011, April 18). Government Technology. Retrieved April 21, 2011, from http://www.govtech.com/health/Meaningful-Use-Electronic-Health-Records-Program.html?elq=94f4d7ef7d0f4fbca95f1f700fb73c98
  19. Merrill, M. (2010a, December 1). iPad helps docs go paperless. Healthcare IT News. Retrieved April 21, 2011, from http://www.healthcareitnews.com/news/ipad-helps-docs-go-paperless
  20. Merrill, M. (2010b, November 15). Healthcare sector among top adopters of iPad. Healthcare IT News. Retrieved April 22, 2011, from http://www.healthcareitnews.com/news/healthcare-sector-among-top-adopters-ipad
  21. Millard, M. (2011, April 15). Vendor Notebook: M*Modal and WebChartMD partner on speech IT. Health. Retrieved April 21, 2011, from http://www.healthcareitnews.com/news/vendor-notebook-mmodal-and-webchartmd-partner-speech-it-0
  22. Montalbano, E. (2011, April 19). PTSD iPhone App Launches For Veterans. InformationWeek: Government. Retrieved April 21, 2011, from http://www.informationweek.com/news/government/mobile/229401873
  23. Porter, S. (2011, April 1). Health IT Can Make Good Family Medicine Practices Even Better. AAFP News Now | American Academy of Family Physicians. Retrieved April 21, 2011, from http://www.aafp.org/online/en/home/publications/news/news-now/ehr/20110401fpehrstories.html
  24. press release. (2011, April 5). InVivoLink Releases iPad App for Analyzing Orthopedic Practice Patterns. EMR Daily News. Retrieved April 21, 2011, from http://emrdailynews.com/2011/04/05/invivolink-releases-ipad-app-for-analyzing-orthopedic-practice-patterns/
  25. Robertson, I., Miles, E., & Bloor, J. (2010). The iPad and medicine.
  26. Shaw, G. (2011, April 19). Do Decision Support Tools Make Docs Look Dumb? HealthLeaders Media. Retrieved April 22, 2011, from http://www.healthleadersmedia.com/print/TEC-265124/Do-Decision-Support-Tools-Make-Docs-Look-Dumb
  27. Shih, G., Lakhani, P., & Nagy, P. (2010). Is Android or iPhone the Platform for Innovation in Imaging Informatics, 23(1), 2-7. doi:10.1007/s10278-009-9242-4
  28. staff. (2011, April 21). Study: Patients believe EMRs bring accuracy to their records. Healthcare IT News. Retrieved April 21, 2011, from http://www.healthcareitnews.com/news/study-patients-believe-emrs-bring-accuracy-their-records
  29. Usatine, R. (2010). Smartphones in Medicine 2010. The Teaching Physician, 9(4), 3-4. Retrieved from http://www.oucom.ohiou.edu/FD/Teaching%20Physician/1010TeachPhy.PDF
  30. Zeman, E. (2011, April 14). Doctors Favor iPhone, iPad Over Android. InformationWeek: Healthcare. Retrieved April 19, 2011, from http://www.informationweek.com/news/healthcare/mobile-wireless/229401603

 

I’m speaking today on my effort to create an instrument for coding narrative content in online discussions. Following are sources cited in the presentation.

ReCal2: Reliability for Coders: a handy online calculator for intercoder/interrater reliability for nominal data. Calculates Percent Agreement, Scott’s Pi, Cohen’s Kappa, and Krippendorff’s Alpha (nominal). Site also offers tools for more than two raters coding ordinal, interval, and ratio data.

Altman, D. G. (1991). Practical Statistics for Medical Research. Chapman & Hall/CRC.

Greene, K., & Brinn, L. S. (2003). Messages Influencing College Women’s Tanning Bed Use: Statistical versus Narrative Evidence Format and a Self-Assessment to Increase Perceived Susceptibility. Journal of Health Communication, 8(5), 443–461.

Greenhalgh, T., & Hurwitz, B. (1999). Narrative based medicine: Why study narrative? BMJ, 318(7175), 48-50. Retrieved from http://www.bmj.com

Jones, D., Turner, M., Singleton, C., & Ramsay, J. (2009). A study analysing inconsistent responses from people with multiple sclerosis in a recent national audit. Disability and Rehabilitation, 31(25), 2064-2072.

Klenke, K. (2008). Qualitative Research in the Study of Leadership. Emerald Group.

Landis, J. R., & Koch, G. G. (1977). The measurement of observer agreement for categorical data. Biometrics, 33(1), 159.

Lombard, M., Snyder-Duch, J., & Campanella Bracken, C. (2010, June 1). Intercoder Reliability. Matthew Lombard. Retrieved 02:40:32, from http://astro.temple.edu/~lombard/reliability/

Osborne, J. W. (2008). Best Practices in Quantitative Methods. SAGE.

Papacharissi, Z. (2007). Audiences as Media Producers: Content Analysis of 260 Blogs. In M. Tremayne (Ed.), Blogging, Citizenship, and the Future of Media. New York: Routledge.

Polkinghorne, D. (1988). Narrative Knowing and the Human Sciences. Albany, NY: State University of New York.

Sarbin, T. R. (1986). Narrative Psychology: The Storied Nature of Human Conduct. Praeger.

Winterbottom, A., Bekker, H., Conner, M., & Mooney, A. (2008). Does narrative information bias individual’s decision making? A systematic review. Social Science & Medicine, 67(12), 2079.

INFS 892: Health Informatics Research

Literature Review #2: Health Care Professionals’ Use of Online Social Networks

Cory Allen Heidelberger

March 25, 2011

 

In less than a decade, online social networks have become a prominent and powerful extension of social interactions throughout our culture. Through tools like blogs, Facebook, and Twitter, we keep in touch with friends, make new friends, do business, and gain new knowledge.

Health care professionals do not exist in isolation from online social networks. These new online tools offer the same social, informational, and professional benefits to doctors, nurses, and pharmacists as they do to executives, students, merchants, and retirees. Since so much of health care relies on effective and professional communication among health care professionals and between them and their patients, a communication tool like online social networks has significant potential to affect the effective delivery of health care. Preliminary to studying how the use of online social networks may affect the practice and outcomes of health care, it is important to understand how and how much health care professionals are adopting and using online social networks to support their professional practice. Given the sensitive, high-stakes nature of health care, it is also important to understand the ethical dimensions of online social network use among health care professionals.

 

Online Social Networks: Defining the Field

The “backbone” of online social networks is the connected collection of user profiles (Boyd & Ellison, 2008). Such profiles usually include user names, photos, and basic demographic information. Online social networks generally include information about connections between members. That information may reside in explicitly declared links, like unidirectional “Like” or “Follower” links or bidirectional “Friend” links (Boyd & Ellison, 2008). Health care providers may interact with patients and with each other on any established online social networking site. Health care providers may turn to health social networks specializing in health issues. They may focus their online activity even more narrowly in online social networks like Sermo.com, membership to which is open exclusively to physicians.

Inquiry into health care provider use of online social networks need not be confined to platforms like Facebook or Sermo. While such platforms dedicated to social networking create an easy point of entry, standardized format, and accessible audience for most casual Internet users, some health care professionals may create their own online social networks through their own blogs, wikis, and other social media. For example, several nurses may create independent blogs but over time find each other’s websites. They may include each other on their blogrolls, comment on each other’s blogs, and write blog posts responding to items they read on each other’s blogs. This nurse “blogosphere” may exist on several different platforms (Blogger, Tumblr, WordPress, etc.), have no formal leadership or organization, follow no business model, and offer no standardized format or features, but it would still constitute an online social network. Its “backbone” of user profiles may be much less well defined than the profile/friend structure of platforms like Facebook, but a blogosphere still offers a clearly definable network of social interactions supported by online technology.

 

Professional–Patient Use

Patients are turning increasingly toward the Internet for health information (Fox & Jones, 2009). No longer the “sole custodian[s] of medical data” (Eysenbach, 2008), health care professionals are increasingly “one of many input sources” (Swan, 2009). Some health care providers are responding to this shift in patient demand: as of January 2011, 906 U.S. hospitals (less than 20% of all U.S. hospitals) had established 3,087 social networking sites (Bennett, 2011). 170 Canadian hospitals (12% of total) have been found to be using social media (Fuller, 2011). Social media adoption rates among European hospitals vary, with around 45% of Norwegian and Swedish hospitals using LinkedIn, but 22% hospital adoption of Facebook in Norway compared to 0% in Sweden. The percentage of German hospitals using online social networks is in the single digits, while adoption in the United Kingdom ranges from 16% for Facebook, 21% for Twitter, and 41% for LinkedIn (Engelen, 2011). Australia’s hospitals are lagging 12 to 18 months behind the U.S. in social media adoption (Cadogan, 2011).

Medical schools are also catching up with adoption of the Web and social networking sites. As of March 31, 2010, 100% of U.S. medical schools had websites. 95% of U.S. medical schools had some sort of Facebook presence: a quarter had official school pages, over 70% had student groups, and more than half had alumni groups on the social networking site. Just over 10% had Twitter accounts (Kind, Genrich, Sodhi, & K. C. Chretien, 2010).

Creating a social media presence does not mean hospitals and other institutions are using them effectively to promote interaction with patients. A marketing study of 120 American hospitals selected at random found all had Facebook pages, but less than 40% posted content to those pages daily, 25% posted twice a week, 25% posted once a month, and 5% had posted nothing (Dolan, 2011). Any discussion of health care providers’ use of online social networks requires remaining mindful that effective use of online social networks requires being social—i.e., being present, producing content, and interacting, not just creating a static electronic brochure.

Patients are seeking information to supplement, not replace, the advice of health care professionals. Overwhelming majorities say professional sources are more helpful in providing accurate medical diagnoses and information about prescription drugs; strong majorities also favor professional sources for information about alternative treatments and recommendations for doctors, specialists, or medical facilities (Fox, 2011). Smaller majorities prefer non-professional sources for emotional support and quick remedies to everyday health issues (Fox, 2011). This split suggests that patients may be receptive to informational support from health care professionals in online social networks but that professionals may want to extend their professional emotional reserve to the online realm and leave laypeople the room they value to provide each other emotional support.

Health care professionals have been able to enter patient health social networks to recruit participants for medical trials. As part of its multi-platform social media strategy during the 2009 H1N1 outbreak the CDC monitored and responded to social network conversations to provide the public with accurate disease and treatment information (Keckley & Hoffman, 2010).

While engaging patients in the online forums they are adopting has the capacity to build effective provider-patient relationships, the health care industry lags in adoption of social media in part due to lack of a clear business model. Online social network activities require time and effort; compensating the physician and the facility for such engagement, from information systems development and maintenance to the actual medical information shared by practitioners, is complicated. Charging patients by the Tweet is problematic in a realm where users are accustomed to free-flowing, unmetered exchanges. Advertising is restricted by professional guidelines and regulations (Keckley & Hoffman, 2010), thus hindering another possible revenue source to make social media efforts pay for themselves.

Health care professionals may find support for an online social-networking business case in the marketing potential of such online tools. Online social networks offer health care professionals the ability to disseminate information quickly, broadly, and at almost no cost. They allow providers to cheaply advertise health-related seminars and community activities. From a pure marketing standpoint, using online social networks to interact with patients sends a message to return and potential “customers” that the providers and their hospital or clinic are cutting edge businesses (Tariman, 2010). Engaging in social media may also be “essential” for institutions and practitioners to combat misinformation that patients and others will spread via those same channels (Pho, 2011a). High-quality physician blogs like KevinMD humanize the healthcare industry as a whole, giving physicians’ perspectives and offering popularly accessible explanations of medical decisions (Bhargava, 2009).

 

Professional–Professional Use

Health care professionals can also use online social networking to obtain information and other support for themselves. Professionals are using these online resources, especially younger professionals (Guseh, R. W. Brendel, & D. H. Brendel, 2009). Roughly one in six U.S. physicians have created accounts on Sermo.com (Bureau of Labor Statistics, 2009; “Introduction | Sermo.com,” 2011). Ozmosis and SocialMD offer similar “walled (and safe) communities for physicians to share opinions and interact in a guarded environment” (Bhargava, 2009). 65% of nurses say they plan to use online social networks for professional purposes (Keckley & Hoffman, 2010).

Health care professionals, like professionals in other fields, have found blogs useful as public document repositories, discussion space, and opportunities to expand professional networks and knowledge base (Thielst, 2007). Such blogs become part of the literature and public face of the profession, informing and reflecting on the medical community as a whole (Lagu, Kaufman, Asch, & Armstrong, 2008). Less public, members-only social networks for physicians may support more valuable sharing of specific medical knowledge and support. Professional online social networks Sermo, Ozmosis, and radRounds allow members to share cases for community discussion and collaboration (Keckley & Hoffman, 2010).

 

Ethical Issues

Doctor-patient interaction online remains relatively rare. Only 5% of adults report receiving information, care, or support from health professionals online (Fox, 2011), a number no higher than the number of adults who reported exchanging e-mails with their doctors in 2008 (Cohen, 2009). Such interaction is stymied not by an absence of health care professionals in online social networks but by ethical concerns. In 2009, 60% of U.S. physicians said they were already using online social networks or were interested in doing so (Darves, 2010). A survey of medical residents and fellows at one French facility in fall 2009 found 73% of respondents had Facebook profiles, with 99% of those including the user’s real name, 97% including birthdates, and 91% including a personal photo. 85% of responding medical professionals said they would automatically decline “friend” requests from patients, and 76% expressed concern that a patient discovering a physician’s Facebook account and gaining access to the content would affect the doctor-patient relationship (Moubarak, Guiot, Y. Benhamou, A. Benhamou, & Hariri, 2011). Another study found more than 80% of University of Florida medical students and residents included personally identifiable information in their Facebook accounts, and only 33% of those users imposed privacy protections on that information (Thompson et al., 2008).

A fundamental tension exists between establishing appropriate boundaries (Luo, 2009) and promoting education and empowerment, a problem addressed by developing “a more sophisticated awareness of privacy and engagement within online communities” (Lewis, Goldman, Bennett, Shine Dyer, & Kolmes, 2011). That understanding of engagement may require simply applying the common sense of face-to-face workplace communication: treat the online social network as a public space at the hospital, and make the publicity explicit to patients who may share that space to give them a sense of what personal matters they should address offline (Giurleo, 2011a; Sydney, 2007). Such public prudence may not differ significantly from the professional ethics doctors have wrestled with in social situations for generations; however, the stakes of maintaining that professionalism are arguably higher in the online realm, where indiscretions can cause damage much more quickly across a much larger social network (Jain, 2009). Ethical professional use of social media also requires constant compassion, with a concerted awareness that the avatars and text with which professionals interact are still real people (Giurleo, 2011b), an awareness that may too easily be lost in online realms that convey less social presence.

Ethical demands may differ among different health care fields. For example, psychotherapists use transference, in which patients experience the psychotherapist in ways similar to their connections with people from their past, to help patients work through their problems. Psychotherapists avoid self-disclosure and maintain professional boundaries to avoid hindering that process. Self-disclosure via social networks may directly impact treatment (Luo, 2009). Because of the nature of their work, psychiatrists who engage in public activities on blogs or Facebook may draw unwelcome attention from emotionally unstable or dependent individuals (Perez-Garcia, 1998). Direct communication and the face-to-face process of narrating their own stories are part of treatment; accessing information about therapists online may short-circuit those processes (Yan, 2009).

On the other hand, psychiatrists may find uniquely valuable information about their patients’ thoughts, emotions, and relationships by using Web searches and social networking sites to incorporate Internet habits into their history-taking (Perez-Garcia, 2010). Health care providers may be able to use online information to verify patient information, especially in mental health situations where patients may be prone to falsehood (Luo, 2009). Acting on false information supplied by patients may lead health care providers to deliver incorrect or harmful treatments; however, online information may be just as prone to inaccuracy and requires active efforts at verification (Hughes, 2009). The APA Ethics Committee has ruled that using the Internet to gather information about a patient is ethical “only in the interests of promoting the patient’s care and well-being and never to satisfy the curiosity or other needs of the psychiatrist,” but another expert contends that Googling patients without their knowledge, even in the interest of providing care, violates patient autonomy and dignity (Yan, 2009).

The problem of unintended disclosure on online social networks may affect patients as well as providers. For example, a physician discovered via Facebook photos that a patient who had denied smoking was indeed a smoker (Guseh et al., 2009). Such unintended disclosure may provide the physician information that may affect recommended treatment; however, if included on medical records, that unintended disclosure could also cause the patient to face higher medical insurance premiums (Chin, 2010).

To avoid ethical pitfalls and harm to patient care, some professionals may also adopt the position that as social and recreational spaces, popular online social networks like Facebook are as inappropriate a space for professional–patient interaction as the local bar; such professionals may thus declare online social networks totally off limits (Darves, 2010; Tariman, 2010). Others recommend very cautious guidelines for online social network engagement, with a first principle of “friending” patients being don’t (Guseh et al., 2009). However, one may question whether health care providers restricting their online social network content to purely professional material will miss out on the social utility of such we tools and whether limiting disclosure on Facebook and personal blogs will make any meaningful contribution to professional privacy when vast amounts of information about health care professionals is already available on other sites outside of their control (Holm, 2009). While certain one-to-one interactions like “friending” on Facebook may complicate professional detachment, forthright engagement with the general public in health care provider blogs and other social networking tools may help put a human face on the industry and provide consumers with a better understanding of health care (Bhargava, 2009). Medicine and law are still catching up with technology, so to avoid running afoul of HIPAA and other rules, practitioners generally avoid blogging about patients, even though discussions of certain challenging cases could be greatly informative for the general public (Darves, 2010). Despite ethical complications—or perhaps because of them, practitioner blogs may be the most logical venue for discussion of ethical and practical guidelines for engaging patients and fellow professionals in social media settings to improve health care delivery. One might even argue that practitioners have a professional, ethical obligation to use the blogs and other social networking tools by which misinformation might spread to combat that misinformation by helping patients find reputable health data (Pho, 2011b).

Awareness of privacy issues online is not universal. A 2006 sampling of medical blog content found 33% providing first and last name of authors and 16% providing sufficient identifying information (Lagu et al., 2008). 16% included positive comments about patients; 18% included negative comments about patients. Blogging allows some popular health care professionals to disseminate good health information to the masses (Darves, 2010), but that same channel can carry incorrect and harmful information just as quickly. To avoid harm and personal liability, health care professionals engaging in blogging appear to be developing voluntary “self-regulation regarding patient privacy, transparency, anonymity, and patient respect” (Kruglyak, 2006; Lagu et al., 2008).

Such self-regulation appears to rise with experience and training: multiple investigations find younger medical students and recent medical school graduates frequently exhibiting unprofessional behavior in online social networks, although concerns in this area seem to arise as much from injudicious posting of their own personal information and evidence of behavior outside healthcare settings that might impinge on their and their schools’ or employers’ reputations as from actual improper healthcare practice or improper direct interaction with patients or other professionals (Cain, Scott, & Akers, 2009; K. C. Chretien, Greysen, J.-P. Chretien, & Kind, 2009; MacDonald, Sohn, & Ellis, 2010; Thompson et al., 2008). Institutions like Harvard Medical School and Drexel University College of Medicine already caution students about the potential unintended professional consequences of injudicious personal disclosures on social networking sites (Jain, 2009). However, as of March 31, 2010, while nearly 97% of U.S. medical schools had posted student guidelines on publicly available websites, only 10% had published conduct policies specific to social media (Kind et al., 2010). Given that prohibiting online social network use is unlikely to stop the widespread adoption and use of these tools by health care professionals, it seems more fruitful to follow to route recommended for using e-mail in health care: proactively defining boundaries, improving user knowledge, and developing practical guidelines centered around patient privacy and trust (Chin, 2010). A similar route in online social networking—developing training in privacy, identity protection, and e-professionalism (Mattingly, Cain, & Fink, 2010; Thompson et al., 2008)—seems a more mature route (van den Broek, 2010) that will trains medical students to develop the professionalism necessary to navigate difficult situations rather than simply avoiding them. Such an approach that addresses the challenges of professionalism online would then allow practitioners, professional organizations, and health care businesses to harness online social networks for advancement of the profession

 

Directions

Adoption of online social networks is unlikely to subside, especially as mobile tools accelerate the blurring of boundaries between online and offline social networks. That adoption process may be slower among health care professionals in their work than among other users in other professions, due to the sensitive and literally life-or-death nature of health care and the professional and ethical considerations that arise therefrom. The health care profession is moving more cautiously into this realm of electronic communication just as it has moved more cautiously from paper to electronic medical records. An inappropriate use of new tools in health care could cause enormous harm. But just as with electronic medical records, the appropriate use of online social networks to communicate with patients and fellow professionals could greatly improve the delivery of health care.

Understanding the current state of online social network use by health care professionals, we can proceed to investigating those potential improvements. Some evidence already exists that patients can find online interaction with doctors satisfactory (Cohen, 2009). Research should further investigate the capacity of online social networks to improve patient perceptions of health care and well-being. Similarly, it will be valuable to determine the satisfaction health care providers obtain in using such online tools, as well as potential professional and organizational benefits such as better workflow, cost savings, acquisition of expertise, and development of and engagement with professional organizations.

While more complicated to quantify, research should also investigate whether health care professionals’ engagement results in better health outcomes. Does advice given online affect the likelihood of patients adopting and sticking with prescribed health behaviors? Does online engagement increase patients’ likelihood to consult with physicians on subsequent health issues? Could online interaction lead to a reduction in face-to-face visits that might in turn lead to health care providers missing certain health indicators that would be obvious in a physical meeting? Answers to all of these question will be of keen interest to providers and patients alike.

Parallel to this course of health investigation should run a line of ethical investigation. As we investigate the impacts of online social networks on provider-patient relationships, we should engage providers and patients in conversations about their expectations of privacy and professionalism in the online realm. These conversations will help shape guidelines to maintain quality and propriety in the increasingly virtual doctor’s office. Such discussions and investigations of current use will also inform the necessary legal scholarship that will develop around online social networks so health care providers may better understand their liability for online communication. Answering these ethical questions alongside the practical questions of health outcomes and provider and patient satisfaction will support increasing appropriate use of online social networks in health care delivery.

 

Works Cited

Bennett, E. (2011). Hospital Social Network List. Found in Cache. Retrieved March 25, 2011, from http://ebennett.org/hsnl/

Bhargava, R. (2009, September 22). How Doctors Are Using Social Media. Ogilvy Public Relations Worldwide: Fresh Influence. Retrieved April 6, 2011, from http://blog.ogilvypr.com/2009/09/how-doctors-are-using-social-media/

Boyd, D. M., & Ellison, N. B. (2008). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13(1), 210–230.

van den Broek, W. (2010, August 9). Facebook and physician professionalism. KevinMD.com. Retrieved April 6, 2011, from http://www.kevinmd.com/blog/2010/08/facebook-physician-professionalism.html

Bureau of Labor Statistics. (2009). Physicians and Surgeons. Occupational Outlook Handbook, 2010-11 Edition. Retrieved March 25, 2011, from http://www.bls.gov/oco/ocos074.htm

Cadogan, M. (2011). Australian Hospital Social Network Directory. Life in the Fastlane. Retrieved March 25, 2011, from http://lifeinthefastlane.com/resources/aussie-hospital-social-network-list/

Cain, J., Scott, D. R., & Akers, P. (2009). Pharmacy students’ Facebook activity and opinions regarding accountability and e-professionalism. American Journal of Pharmaceutical Education, 73(6), 104. Retrieved from http://www.ncbi.nlm.nih.gov/pubmed/19885073

Chin, J. J. (2010). Medical Professionalism in the Internet Age. Annals of the Academy of Medicine, Singapore, 39(5), 345–347.

Chretien, K. C., Greysen, S. R., Chretien, J.-P., & Kind, T. (2009). Online posting of unprofessional content by medical students. JAMA: The Journal of the American Medical Association, 302(12), 1309-1315. doi:10.1001/jama.2009.1387

Cohen, E. (2009, September 3). Should you “friend” your doctor on Facebook? CNN.com: Health. Retrieved March 24, 2011, from http://www.cnn.com/2009/HEALTH/09/03/friending.your.doctor/index.html

Darves, B. (2010, March). Social Media and Physicians. NEJM Career Center. Retrieved May 17, 2010, from http://www.nejmjobs.org/career-resources/social-media-and-physicians.aspx

Dolan, P. L. (2011, March 23). Conversation Lags on Hospital Facebook Pages. American Medical Association: American Medical News. Retrieved April 6, 2011, from http://www.ama-assn.org/amednews/2011/03/21/bise0323.htm

Engelen, L. (2011). European Hospitals. HospitalSEU: UMC St. Radboud REShape & Innovation Center. Retrieved March 25, 2011, from http://hospitalseu.wordpress.com/

Eysenbach, G. (2008). Medicine 2.0: Social Networking, Collaboration, Participation, Apomediation, and Openness. Journal of Medical Internet Research, 10(3), e22. doi:10.2196/jmir.1030

Fox, S. (2011). Peer-to-peer Healthcare. Pew Internet & American Life Project. Pew Research Center’s Internet & American Life Project. Retrieved from http://www.pewinternet.org/Reports/2011/P2PHealthcare.aspx

Fox, S., & Jones, S. (2009). The Social Life of Health Information. Pew Research Center’s Internet & American Life Project. Retrieved from http://www.pewinternet.org/Reports/2009/8-The-Social-Life-of-Health-Information.aspx

Fuller, A. (2011, March 6). Hospital Social Networking List. SMiCH+. Retrieved March 25, 2011, from http://www.smich.ca/?page_id=12

Giurleo, S. (2011a, March). Why do we make social difficult? (or we’re not in middle school anymore). DrSusanGiurleo.com. Retrieved March 25, 2011, from http://drsusangiurleo.com/why-do-we-make-social-difficult/

Giurleo, S. (2011b, March 1). The Most Important Rule of Health Care Social Media. DrSusanGiurleo.com. Retrieved March 25, 2011, from http://drsusangiurleo.com/the-most-important-rule-of-health-care-social-media/

Guseh, J. S., Brendel, R. W., & Brendel, D. H. (2009). Medical professionalism in the age of online social networking. Journal of Medical Ethics, 35(9), 584 -586. doi:10.1136/jme.2009.029231

Holm, S. (2009). The concise argument. Journal of Medical Ethics, 35(9), 525. doi:10.1136/jme.2009.032375

Hughes, L. (2009). Is it ethical to “google” patients. Psychiatr News, 44(9), 11.

Introduction | Sermo.com. (2011). Sermo.com. Retrieved March 25, 2011, from http://sermo.com/about/introduction

Jain, S. H. (2009). Practicing medicine in the age of Facebook. New England Journal of Medicine, 361(7), 649–651.

Keckley, P. H., & Hoffman, M. (2010). Social Networks in Health Care: Communication, Collaboration and Insights. Deloitte. Retrieved from https://www.deloitte.com/view/en_US/us/industries/US-federal-government/center-for-health-solutions/disruptive-innovations/2fbc755f3c1b9210VgnVCM100000ba42f00aRCRD.htm

Kind, T., Genrich, G., Sodhi, A., & Chretien, K. C. (2010). Social media policies at US medical schools, 15(10). doi:10.3402/meo.v15i0.5324

Kruglyak, D. M. (2006, October 30). HealthTrain, the Open Healthcare Manifesto. Trusted.MD. Retrieved March 25, 2011, from http://trusted.md/manifesto#axzz1Hd88DXoU

Lagu, T., Kaufman, E. J., Asch, D. A., & Armstrong, K. (2008). Content of Weblogs Written by Health Professionals. Journal of General Internal Medicine, 23(10), 1642-1646. doi:10.1007/s11606-008-0726-6

Lewis, D., Goldman, D., Bennett, E., Shine Dyer, J., & Kolmes, K. (2011, March 13). Patients/Caregivers on Facebook: Establishing Boundaries Without Barriers. Presented at the SXSW: South by Southwest, Austin, Texas. Retrieved from http://schedule.sxsw.com/events/event_IAP5652

Luo, J. S. (2009). The facebook phenomenon: boundaries and controversies. Primary Psychiatry, 16(11), 19–21.

MacDonald, J., Sohn, S., & Ellis, P. (2010). Privacy, professionalism and Facebook: a dilemma for young doctors. Medical Education, 44(8), 805-813. doi:10.1111/j.1365-2923.2010.03720.x

Mattingly, T. J., Cain, J., & Fink, J. L. (2010). Pharmacists on Facebook: online social networking and the profession. Journal of the American Pharmacists Association: JAPhA, 50(3), 424-427. doi:10.1331/JAPhA.2010.08174

Moubarak, G., Guiot, A., Benhamou, Y., Benhamou, A., & Hariri, S. (2011). Facebook activity of residents and fellows and its impact on the doctor-patient relationship. Journal of Medical Ethics, 37(2), 101.

Perez-Garcia, G. J. (2010). Challenges Faced by Psychiatrists in the Internet Age. Psychiatric Times, 27(5).

Pho, K. (2011a, March 23). Hospitals need an engaged social media presence on Facebook and Twitter. KevinMD.com. Retrieved April 6, 2011, from http://www.kevinmd.com/blog/2011/03/hospitals-engaged-social-media-presence-facebook-twitter.html

Pho, K. (2011b, April 1). Patients using Facebook for health information. KevinMD.com. Retrieved April 6, 2011, from http://www.kevinmd.com/blog/2011/04/patients-facebook-health-information.html

Swan, M. (2009). Emerging Patient-Driven Health Care Models: An Examination of Health Social Networks, Consumer Personalized Medicine and Quantified Self-Tracking. International Journal of Environmental Research and Public Health, 6(2), 492-525. doi:10.3390/ijerph6020492

Sydney. (2007, March 17). Blogging Doctors — Nyet. Medpundit. Retrieved March 25, 2011, from http://medpundit.blogspot.com/2007/03/blogging-doctors-nyet-are-doctors-who.html

Tariman, J. D. (2010). Where to Draw the Line: Professional Boundaries in Social Networking. ONS connect, 25(2), 10-13.

Thielst, C. B. (2007). Weblogs: a communication tool. Journal of healthcare management/American College of Healthcare Executives, 52(5), 287.

Thompson, L. A., Dawson, K., Ferdig, R., Black, E. W., Boyer, J., Coutts, J., & Black, N. P. (2008). The intersection of online social networking with medical professionalism. Journal of General Internal Medicine, 23(7), 954-957. doi:10.1007/s11606-008-0538-8

Yan, J. (2009). Psychiatrists must beware the perils of cyberspace. Psychiatric News, 44(14), 9.

Text of presentation on prior research, delivered in INFS 892, 2011.03.03

Health information technology (HIT) research frequently cites privacy as a concern of health care providers and patients. Privacy is recognized internationally as critical factor in HIT adoption (Anderson, Forgner, Johns, & Reinhardt, 2006). Technological, social, organizational, and legal changes since the 1960s have contributed to steadily growing privacy concerns (Westin, 2003). However, the new technology of online social networks as applied to health—health social networks—poses new conditions and benefits that may change previous privacy concerns.

Health Social Networks

The “backbone” of online social networks is the connected collection of user profiles (Boyd & Ellison, 2008). Such profiles usually include user names, photos, and basic demographic information. On health social networks, these profiles may include medical information that would generally be kept private in traditional interactions with health professionals but which HSN users share to help each other find users with information and experience relevant to their specific health concerns. Online social networks generally include information about connections between members. That information may reside in explicitly declared links, like unidirectional “Like” or “Follower” links or bidirectional “Friend” links (Boyd & Ellison, 2008).

A health social network is an online social network formed around shared interest in a specific health condition like obesity or cancer, a specific area of health care like children’s medicine or hospice, or health information in general. The key characteristic of an online health social network is interaction focused on sharing knowledge and providing emotional support for individuals dealing with health-related questions and problems affecting themselves or other people they care about. Swan (2009) defines a health social network as “a website where consumers may be able to find health resources at a number of different levels.” Swan’s definition includes the word consumer, an increasingly inapt term (Adams, 2010a) (Heidelberger, 2009) given that an important characteristic of health social networks and any other type of online social network is that users are not merely consuming content but producing and organizing much of the content available in the form of, for example, posts, comments, discussions, and ratings. While Internet use is often characterized by “lurking,” the act of reading online discussions without making one’s presence known through participation, some evidence shows that health-related online groups show a markedly above-average level of user participation (Nonnecke & Preece, 2000). This increased participation and the blurring of producers and consumers into conducers affects how individuals configure their understandings of health (Adams, 2010b).

The value of any online social network lies in the sharing of information. In health social networks, one of the vital “different levels” (Swan, 2009) at which users seek information is peer information, advice from users just like them that relates specifically to their experience. Lurkers can find certain information relevant to their health situations without ever making their presence known, but to make the information on a health social network fully relevant to one’s specific health situation, a user generally must choose to engage the community in a conversation, share some personal detail, and surrender some level of privacy.

Privacy

Privacy is subject to different interpretations in different situations (Allen, 1988). Privacy is not simply the desire to withhold information; more comprehensively, privacy refers to the desire to control disclosure of information (Lanier & Saini, 2008). Privacy means controlling what personal information is revealed to others, when that information is revealed, and how it is used (Westin, 1967, 2003). The view of privacy as a matter of control enjoys broad acceptance (Elgesem, 1996; Fried, 1984; Lessig, 2002). However, some scholars argue that, while control plays an important part in managing privacy, having control does not equate with having privacy (Dinev & Hart, 2004; Laufer & Wolfe, 1977; Xu, 2007). Control may merely be a means to the privacy ends of restricted access and protection from intrusion (Tavani & Moor, 2001).

Westin (1967) identifies four states of privacy:

  1. solitude: freedom from observation by others
  2. intimacy: small-group seclusion in which members can achieve close, relaxed, frank relationship
  3. anonymity: freedom from identification and surveillance in public places and for public acts
  4. reserve: desire to limit disclosure to others (as summarized in Margulis, 2003).

Westin (1967) also identifies four main functions of privacy:

  1. personal autonomy
  2. emotional release
  3. self-evaluation
  4. limited and protected communication

Some scholars narrow the concept of privacy to informational privacy to focus on the ability of individuals to limit the access others have to their information (Alpert, 2003). In developing the “Concern for Information Privacy” construct, (Smith, Milberg, & Burke, 1996) identify four main components: collection, errors, unauthorized access, and secondary use. (Stewart & Segars, 2002) validated and extended the CFIP instrument. CFIP focuses on organizational handling and uses of information. However, the Internet creates important changes in how individuals interact with organizations and with each other with respect to personal information. The Internet gives consumers more access to and control over the data they share in the marketplace (Malhotra, Kim, & J. Agarwal, 2004). Grounded in social contract theory, the construct of Internet users’ information privacy concerns (IUIPC) focuses on consumers’ engagement with e-commerce and consists of three main components:

  1. collection: equitable exchange of information according to mutually accepted rules
  2. control: freedom to voice an opinion or exit the contractual agreement
  3. awareness of privacy practices: understanding of the rules and practices established by the firm and its agreement with consumers (Malhotra et al., 2004).

Subsequent research has used social contract theory similarly to understand patient interaction with medical websites (Gaurav Bansal, Zahedi, & Gefen, 2010).

The National Consumer Health Privacy Survey 2005 (Bishop, Holmes, & Kelley, 2005) provides these key findings about user interaction

  1. Consumers are concerned about health information privacy.
  2. Consumers are unaware of their privacy rights.
  3. A small fraction (13%) of consumers engage in specific privacy-protection behavior. Most cited behaviors have to do with keeping insurers from finding out information that might affect payment or premiums.
  4. Consumers are willing to trade privacy for benefits. Note that at this time, more consumers identified paper records as secure than identified electronic records as secure.

Privacy in Health Social Networks

More recent research suggests that privacy concerns are affected by health status: individuals in poor health are less likely to share information with consumer health websites (G. Bansal & Davenport, 2010). Connected research attempts to link privacy concerns with other personal dispositions: emotional instability may work through perceived sensitivity of health information to relate to heightened privacy concerns, but extroversion, agreeableness, conscientiousness, and intellect show little sign of significant connection to privacy concerns (Gaurav Bansal et al., 2010). Notice that the preceding studies, like many discussions of information privacy, address privacy as a consumer issue. Westin (2003) broadens the discussion of privacy to look at individual privacy concerns in the context of their relationships as citizens to government and as employees to employers in addition to the commercial relationship of consumers to business. The IUIPC focuses on the social-contract relationship between individuals and the firm as each entity surrenders different rights and makes different promises for different benefits (Malhotra et al., 2004). These conceptualizations fit with the theme of much literature on health information privacy as it considers the impacts of how health care providers may use (or misuse) electronic health records to process genetic information and other sensitive patient data (e.g., Alpert, 2003).

However, these conceptualizations assume a hierarchy that does not necessarily manifest itself in an online social network. Certainly online social networks may involve citizen–government, employee–employer, and consumer–business interactions. While there may well be consumer–business interactions in health social networks in the form of network members seeking information from staff of the online service (e.g., formal forum moderators or coaches, designated medical experts) or advertisements soliciting business, online social network interactions will generally be dominated by a form of interaction not addressed by most discussions of privacy concerns: non-hierarchical citizen–citizen interactions, users communicating with users as equals seeking the same balance between privacy rights and informational benefits.

One of the above studies acknowledges that health social networks may play by different rules from the typical consumer medical websites. In finding that extroversion shows no significant relationship with health information sensitivity, (Gaurav Bansal et al., 2010) note the the consumer-type website they used in their investigation provides a “leaner” online setting where extroverts “apparently are less forthcoming.” They acknowledge that personality traits may have different influence in different communication settings and recommend further investigation along those lines. Such investigation makes sense for health social networks, as they would provide a richer communication setting where, instead of simply making a commercial transaction for a product, users are engaging directly and often by name with each other. This richer channel can carry more social presence, which could well have more influence over privacy perceptions and behaviors than a “lean” commercial transaction. (Gaurav Bansal et al., 2010) recognize that trust in a website (a factor with a strong positive relationship to intent to disclose health information) could be enhanced for some “customers” by social interaction, the heart and soul of health social networks. Evidence that consumer-oriented research may not capture the full privacy picture in health social networks comes from recent findings that individuals with chronic conditions are more likely to go online to find others with similar health concerns than individuals who do not have chronic conditions (Fox, 2011).

Unique characteristics of online social networks pose challenges to normal privacy expectations and behaviors. The social Web is inherently leaky: they increase the chances for individuals to lose control of information they share (Solove, 2007). Online social networks constitute a “mediated public” where information becomes persistent, searchable, replicable, and open to invisible audiences (boyd, 2007). Information that would be transient in a private face-to-face conversation persists online in entries in the Google cache and copies stored on other computers. Information that would be lost in noise and fallible memory can be retraced and recalled almost instantly by search engines. Information shared online can be replicated in numerous other contexts that may have no direct relation to the original intent that motivated the sharing (such as a discussion on a weight-loss surgery forum about constipation finding its way into a doctoral dissertation on the use of narrative on health social networks). And information directed at one specific person or group in an online forum may be viewed by and provoke reactions from numerous individuals whom the original speaker does not and may not ever know. All of these factors can make social interactions online “eternal,” transcending the moment and context in which they happen and escaping the control of the people originating them (Albrechtslund, 2008). This “eternal” nature is seen in health social network profiles whose stories and data continue to inform their communities even after the originators of those profiles die (Goetz, 2008).

However, concerns about privacy in online social networks often operate from an assumption of vertical hierarchy underpinning the traditional view of surveillance (Albrechtslund, 2008). Concerns about surveillance from some “Big Brother” lead to assumptions that sacrifices of privacy in online social networks stem either from cost-benefit analysis or from ignorance of the dangers. Albrechtslund posits that online social networks support “participatory surveillance” in which individuals empower themselves and each other by sharing information rather than trading it. This participatory paradigm improves our understanding of privacy dynamics online as we witness a shift from the anonymity and pseudonymity that dominated Internet discourse in the first decade of the Web to the increased voluntary posting of personal information on social networks and integration of online and offline activities (Adams, 2010b), a shift enhanced by the increasing availability and popularity of Web 2.0 technology like blogs, Facebook, and Twitter.

Privacy is generally seen as a positive construct, and research focuses on how to protect or preserve it (Angst & R. Agarwal, 2009). One study views sharing health information online as a “disutility” (G. Bansal & Davenport, 2010). Arguably, though, from a health perspective, privacy could be seen as a negative factor standing in the way of building and sharing knowledge and support. In a discussion of blogging in a health social network context, (Adams, 2010b) suggests there is a need for further investigation of whether engaging in online documentation and communication about personal health may motivate individuals to stick with routines in pursuit of their health goals. This suggestion hearkens to the concept of participatory surveillance: health social network users may benefit from being able to “keep an eye on” each other and know that others are similarly keeping an eye on them. Increased information sharing also creates a richer database of disease treatment and patient experience, an advantage that PatientsLikeMe explicitly embraces in an “Openness Philosophy” that challenges the primacy of privacy:

“Currently, most health-care data is inaccessible due to privacy regulations or proprietary tactics,” it declares. “As a result, research is slowed, and the development of breakthrough treatments takes decades. . . . When you and thousands like you share your data, you open up the health-care system. . . . We believe that the Internet can democratize patient data and accelerate research like never before” (Goetz, 2008).

Some users of health social networks report finding informational support among peers that they may not find among their regular physicians:

Thank you for the replies. Of course I will talk to my doctor about this but wanted to come on here because it is more beneficial to me to find out how it affects a person that has had WLS [weight loss surgery]. Doctors are not always familiar with how our new systems work (ObesityHelp.com user, 2009).

User behavior in health social networks suggests different attitudes toward privacy than we might expect in public face-to-face interactions. In ObesityHelp.com, for instance, plenty of interaction takes place privately, in member-only chat rooms and via other protected channels. However, thousands of users post medical information in publicly accessible profiles, blogs, and forum posts. Users identifiable by username, location, and photo openly discuss personal details like weight loss and gain, prescriptions, constipation, sexual dysfunction, and surgical complications that would be considered impolite if not imprudent to share in the physical company of strangers and which would certainly violate HIPAA if released by those users’ health care providers.

It is possible, of course, that many users do not grasp the privacy implications of using health social networks. A study of health social networks for diabetics found varying capabilities for users to control privacy settings and share information; only eight of the ten studied networks offer accessible privacy policies, and most of them are hard to read (Weitzman, Cole, Kaci, & Mandl, 2011). Providing users with more control over publication may actually “induce them to reveal more sensitive information,” a result which raises concerns that users may conflate increased control over publication with increased control over access and usage of their personal information  (Brandimarte, Acquisti, & Loewenstein, 2010). Access and usage pose the real risk of intrusion and harm, but as potential actions of others, they are less salient than the satisfaction users may derive from the immediate exercise of control over their own information publication.

Concern for information privacy has been shown to correlate negatively with likelihood of adopting electronic health records (Angst & R. Agarwal, 2009). However, our own research at DSU on adoption of electronic health records in South Dakota suggests that privacy concerns are not nearly as prominent in the minds of practitioners as other barriers to ER adoption. And whatever privacy concerns may keep some people from engaging with health information technology, there are thousands of users who are willing to share personal stories and health information on health social networks like PatientsLikeMe (Goetz, 2008).

Privacy concerns in health social networks may be mitigated by the sense of community. Consider this observation from multiple sclerosis patient and PatientsLikeMe user Laurie Fournier:

Pretty much everybody I know over 45 has some kind of medical condition. Some people have had cataracts surgery, or they have high blood pressure, or high cholesterol or diabetes. Everyone has something. And if everyone has something, that really levels the playing field (Goetz, 2008).

The sense of community in health social networks may be even stronger among users with rare diseases. Individuals with rare diseases—i.e., diseases affecting no more than 20,000 people—are “power users” of Internet health resources (Shute, 2011). The observation that such users are “ready and willing to share with each other so that other people can benefit from their experiences” (Shute, 2011) suggests the need to look beyond a purely selfish privacy calculus: users may surrender privacy not to gain emotional or informational support for themselves but to provide such support to other members of their online community.

Privacy remains a valid concern for all individuals. Privacy sets boundaries, and people need physical and psychological boundaries to define their sense of identity. The increasing integration of online social networks into daily activities, especially in the area of patient use of health social networks, will not end the value of privacy. However, health social networks emphasize the difference in privacy concerns with hierarchical institutions and with peers who share health interests and goals.

Works Cited

Adams, S. A. (2010a). Revisiting the online health information reliability debate in the wake of “web 2.0”: An inter-disciplinary literature and website review. International journal of medical informatics, 79(6), 391–400.

Adams, S. A. (2010b). Blog-based applications and health information: Two case studies that illustrate important questions for Consumer Health Informatics (CHI) research. International Journal of Medical Informatics, 79(6), e89-e96. doi:10.1016/j.ijmedinf.2008.06.009

Albrechtslund, A. (2008). Online social networking as participatory surveillance. First Monday (Chicago), 13(3).

Allen, A. L. (1988). Uneasy access: Privacy for women in a free society. Rowman & Littlefield Pub Inc.

Alpert, S. A. (2003). Protecting medical privacy: challenges in the age of genetic information. Journal of Social Issues, 59(2), 301–322.

Anderson, G. F., Forgner, B. K., Johns, R. A., & Reinhardt, U. E. (2006). Health Care Spending and Use of Information Technology in OECD Countries. Health Affairs, 25(3), 819-831. Retrieved from http://www.ezproxy.dsu.edu:2108/pqdweb?index=17&did=1038492821&SrchMode=3&sid=1&Fmt=4&VInst=PROD&VType=PQD&RQT=309&VName=PQD&TS=1249163046&clientId=18865&aid=1

Angst, C. M., & Agarwal, R. (2009). Adoption of Electronic Health Records in the Presence of Privacy Concerns: The Elaboration Likelihood Model and Individual Persuasion. MIS Quarterly, 33(2), 339-370. doi:Article

Bansal, G., & Davenport, R. (2010). Moderating Role of Perceived Health Status on Privacy Concern Factors and Intentions to Transact with High versus Low Trustworthy Health Websites. MWAIS 2010 Proceedings, 7.

Bansal, G., Zahedi, F. “., & Gefen, D. (2010). The impact of personal dispositions on information sensitivity, privacy concern and trust in disclosing health information online. Decision Support Systems, 49(2), 138-150. doi:10.1016/j.dss.2010.01.010

Bishop, L., Holmes, B. J., & Kelley, C. M. (2005). National consumer health privacy survey 2005. Oakland: California Healthcare Foundation, 1–5.

Boyd, D. M., & Ellison, N. B. (2008). Social network sites: Definition, history, and scholarship. Journal of Computer-Mediated Communication, 13(1), 210–230.

boyd, D. (2007, May). Social Network Sites: Public, Private, or What? The Knowledge Tree.

Brandimarte, L., Acquisti, A., & Loewenstein, G. (2010). Misplaced Confidences: Privacy and the Control Paradox. Presented at the Ninth Workshop on the Economics of Information Security (WEIS 2010), Cambridge, MA.

Dinev, T., & Hart, P. (2004). Internet privacy concerns and their antecedents-measurement validity and a regression model. Behaviour & Information Technology, 23(6), 413–422.

Elgesem, D. (1996). Privacy, respect for persons, and risk. Philosophical perspectives on computer-mediated communication, 45–66.

Fox, S. (2011). Peer-to-peer Healthcare. Pew Internet & American Life Project. Pew Research Center’s Internet & American Life Project. Retrieved from http://www.pewinternet.org/Reports/2011/P2PHealthcare.aspx

Fried, C. (1984). Privacy. In F. D. Schoeman (Ed.), Philosophical dimensions of privacy: An anthology. Cambridge Univ Pr.

Goetz, T. (2008, March 23). Practicing Patients. The New York Times. Retrieved from http://www.nytimes.com/2008/03/23/magazine/23patients-t.html

Heidelberger, C. A. (2009). Citizens, Not Consumers. In V. Weerakkody, M. Janssen, & Y. K. Dwivedi (Eds.), Handbook of Research on ICT-Enabled Transformational Government: A Global Perspective (pp. 51-71). Hershey, PA: IGI Global.

Lanier, C. D., & Saini, A. (2008). Understanding consumer privacy: A review and future directions. Academy of Marketing Science Review, 12(2), 1–48.

Laufer, R. S., & Wolfe, M. (1977). Privacy as a concept and a social issue: A multidimensional developmental theory. Journal of Social Issues, 33(3), 22–42.

Lessig, L. (2002). Privacy as property. Social Research: An International Quarterly, 69(1), 247–269.

Malhotra, N. K., Kim, S. S., & Agarwal, J. (2004). Internet Users’ Information Privacy Concerns(IUIPC): The Construct, the Scale, and a Causal Model. Information Systems Research, 15(4), 336–355.

Margulis, S. T. (2003). On the Status and Contribution of Westin’s and Altman’s Theories of Privacy. Journal of Social Issues, 59(2), 411–429.

Nonnecke, B., & Preece, J. (2000). Lurker demographics: counting the silent. In Proceedings of the SIGCHI conference on Human factors in computing systems (pp. 73-80). The Hague, The Netherlands: ACM. doi:10.1145/332040.332409

Shute, N. (2011, March 1). People Coping With Rare Disease Are Internet Power Users. NPR: Shots. Retrieved March 3, 2011, from http://www.npr.org/blogs/health/2011/02/28/134140813/people-coping-with-rare-disease-are-internet-power-users?ps=sh_sthdl

Smith, H. J., Milberg, S. J., & Burke, S. J. (1996). Information privacy: measuring individuals’ concerns about organizational practices. MIS quarterly, 20(2), 167–196.

Solove, D. J. (2007). The future of reputation: gossip, rumor, and privacy on the Internet. Yale Univ Pr.

Stewart, K. A., & Segars, A. H. (2002). An empirical examination of the concern for information privacy instrument. Information Systems Research, 13(1), 36–49.

Swan, M. (2009). Emerging Patient-Driven Health Care Models: An Examination of Health Social Networks, Consumer Personalized Medicine and Quantified Self-Tracking. International Journal of Environmental Research and Public Health, 6(2), 492-525. doi:10.3390/ijerph6020492

Tavani, H. T., & Moor, J. H. (2001). Privacy protection, control of information, and privacy-enhancing technologies. ACM SIGCAS Computers and Society, 31(1), 6–11.

Weitzman, E. R., Cole, E., Kaci, L., & Mandl, K. D. (2011). Social but safe? Quality and safety of diabetes-related online social networks. Journal of the American Medical Informatics Association. Retrieved from http://jamia.bmj.com/content/early/2011/01/24/jamia.2010.009712.abstract

Westin, A. F. (1967). Privacy and freedom (Vol. 97). London.

Westin, A. F. (2003). Social and political dimensions of privacy. Journal of Social Issues, 59(2), 431–453.

Xu, H. (2007). The effects of self-construal and perceived control on privacy concerns. In Proceedings of the 28th Annual International Conference on Information Systems (ICIS 2007). Montréal, Québec, Canada.

 

I keep forgetting what I planned to write about. That’s what the Internet is for: to remember stuff so I don’t have to. Here are the three topics I’ll tackle in our health informatics research class:

  1. Patient privacy concerns in health social networks (with potential for research topic on ObesityHelp.com user-generated content)
  2. Mobile/ubiquitous computing in health care settings (remember, this can address use by practitioners and/or patients!)
  3. Health care provider use of online social networks

 

I’m working on my CET 751 paper/presentation on solid-state drives. Here are some sources:

  1. Samsung marketing video that made me go “Holy crap! I want SSD!” (stay tuned for the trampoline!).
    1. Samsung SSD three-story drop test (2008)
    2. Intel SSD laptop roof drop (2010)
    3. This guy showed his Macbook SSD opening 50 apps. He got over three million views.
  2. Wikipedia on solid-state drives (no, really: Wikipedia is pretty good, especially on tech stuff) and, for comparison, on hard drives.
  3. Lucas Mearian (2010), “Why Aren’t SSDs Getting Cheaper?” NetworkWorld.com.
  4. Logan G. Harbaugh (2010), “Storage Smackdown: Hard Drives vs. SSDs,” NetworkWorld.com.
  5. Mark Kyrnin (2007… I think!), “SSD — Solid-State Drives: A Hard Drive Alternative Based on Flash Memory,” About.com.
  6. Elexis Marie (no date given), “How Do Solid-State Hard Drives Work?” eHow.com
  7. Jeff Atwood (2009), “The State of Solid-State Drives,” Coding Horror.
  8. Linus Torvald (yes, the Linux inventor — 2008), “…so I got one of those new Intel SSDs,” Linus’ Blog.
  9. Leo Noteboom (2009), “Can a USB Thumb Drive ‘Wear Out’?Ask-Leo.com.
  10. no author cited, “OLPC XO Laptop: First Look,” ConsumerReports.org

The term solid-state has evolved just a little over the past few years. In older usage, every computer in this room is “solid-state”: our machines all run on semiconductors, not vacuum tubes. Currently, though, solid-state is used primarily to distinguish memory devices with no moving parts from conventional spinning disk drives.

The standard hard drive has two motors. One motor spins one or more magnetic platters, the disks on which data is stored. That data storage layer is a magnetic coating 10 to 20 nanometers thick — several thousand of these layers stacked atop one another would equal the thickness of one sheet of paper. Another motor moves an actuator arm that sweeps the read/write heads across the platters. These motors do some serious work: the platters spin at 4200 to 15000 rpm, while the tip of the actuator arm experiences accelerations as high as 550 g’s. These high-speed movements in tight tolerances leave a hard drive susceptible to crashes caused by sudden jolts, contaminants, temperature, and even changes in air pressure (operating a typical hard drive more than 10,000 feet above sea level increases the risk of a crash).

A solid-state drive looks much like a hard drive, since it is designed to swap in for traditional hard drives (Kyrnin, 2007). They come in the same chassis as 2.5-inch or 3.5-inch hard drives, ready to plug in to the same ATA or SATA interface.

The big difference is that only movement in a solid-state drive is the movement of electrons. There are no spinning platters or swinging arms. Instead of storing data in alternating magnetic fields, the solid-state drive moves electrons in and out of tiny semiconductor transistors. A transistor that is full of electrons and cannot accept electrical flow reads as a binary 0; a transistor that accepts flow is a binary 1.

The absence of moving parts presents several advantages:

  1. Reliability: Fewer moving parts mean fewer things that can go wrong. Hard drives require gentle treatment, lest a sudden jolt send heads grinding into platters. Mobile devices may require accelerometers to sense sudden movement and stop hard drive operations before a shock causes a crash. While users should not get too reckless with solid-state drives, sudden movement does not affect the reading of transistors. Solid-state drives can withstand various effects that would cripple hard drives, which makes them superior memory solutions for mobile devices.
  2. Performance: Hard drives must spin the platters to the proper position for the heads to read or write the data. That means hard disk reads and writes include some spin-up and seek time during which data isn’t moving. Solid-state drives experience no such delays. Recognizing these time savings, programmer Jeff Atwood characterizes solid-state drives as “the most cost effective performance increase you can buy.” Linus Torvald offered the following positive assessment of Intel’s SSD in October 2008:

    The whole thing just rocks. Everything performs well. You can put that disk in a machine, and suddenly you almost don’t even need to care whether things were in your page cache or not. Firefox starts up pretty much as snappily in the cold-cache case as it does hot-cache. You can do package installation and big untars, and you don’t even notice it, because your desktop doesn’t get laggy or anything.

    By the numbers, a solid-state drive can perform 16,000 Input/Output operations per second (IOPS). A top of the line 15,000-rpm enterprise-grade Fibre Channel hard drive will perform 200 IOPS (Mearian, 2010). The solid-state drive is 80 times faster. Think of it this way: suppose it takes me 15 minutes to deliver this presentation. Replace me with a solid-state drive, and you could get the message in 11 seconds. Currently, consumer-grade SSDs offer a much smaller advantage, “only” twice as fast.

  3. Power usage: Motors eat electricity. Practically, energy usage may not hinder a desktop computer’s performance, but mobile device us”ers demand every power savings they can get to extend battery life. Unlike toning down screen brightness, a solid-state drive saves power while improving performance, a nice bonus for mobile device users.
  4. Noise: what noise? With no moving parts, no mechanical energy is lost to movement turning into sound waves. Of course, if you like that little grindy-grindy sound when you put your computer through the paces, then solid-state drive silence will bug you the way the Prius bugs some people as it starts noiselessly.

No device is perfect. One weakness of the solid-state drive is the vulnerability of transistors to wearing out. They can take only so many fillings and emptying of electrons — in this case, only so many read and writes. You can use your thumb drive as a sort of solid-state hard drive, but you would wear it out the same way. Estimates for flash memory lifetime range from 10,ooo to 100,000 read/writes. For copying and backing up files, that lifetime isn’t bad. But if you start using your flash drive for various applications, especially database applications, reads and writes can add up quickly (Noteboom, 2009). Solid-state drives can have compensating mechanisms, but one bad bit can be enough to render the entire drive unreadable (Noteboom, 2009).

To extend the life of a solid-state drive, engineers use wear leveling, a technique that distributes read/write activity roughly equally around the disk. If a frequently accessed file were kept in a static location, the transistors in that “sector” of the solid-state drive would wear out sooner than other transistors that were lucky enough to be assigned some archied file. Wear-leveling moves files around and even breaks them up from read to read and rwite to write, spreading out the abuse to all portions of the drive roughly equally.

Wear leveling creates some security issues. If a user loads, encrypts, and saves a file, the solid-state drive does not overwrite the original file. The drive instead creates the desired encrypted copy in a new location, leaving the original unencrypted and possibly accessible. Users also can not overwite files directly, as the solid-state drive will distribute new files to locations that require wear-leveling rather overwriting already existing files in more often-used areas. Solid-state drives can get around these problems, but the solutions require some processing overhead.

Solid-state drives also cost more than hard drives. One analyst predicts that for manufacturers, solid-state drive costs will drop from $1.90 per gigabyte today to $1.70/GB next year, while online consumers will likely see prices stay steady at $3.00 to $3.30 per gigabyte (Mearian, 2010). Prices did drop 60% in both 2007 and 2008, but manufacturers were making no money. When the recession hit, manufacturers could not afford to take that hit any more. They slowed production, tightening supply even as demand increased, which flattened out the price curve and drove some price increases (Mearian, 2010).

Mearian (2010) finds 1-terabyte hard drives available for $90, compared to Intel SSD’s for $400. Mearian (2010) cites one expert from Gartner who says solid-state drives will never match hard drives on price per gigabyte. Mearian (2010) suggests that if that cost difference persists, a viable option may be hybrid systems with a lower capacity SSD boot drive combined with a big external drive for multimedia storage.

Cost likely means solid-state drives will not fully replace hard drives. However, solid-state drives can  meet important needs in a number of realms. For some mobile users, the increased reliability and lower power consumption may be worth the cost overhead. Military and aerospace users are big purchasers of solid-state systems to provide reliable computing power in high-performance situations. Even the One Laptop Per Child project incorporated a solid-state drive in its XO laptop at the end of 2007. The XO had a 1GB SSD with 800 MB available. OLPC was trying to provide a cheap yet rugged laptop for use in isolated and low-income locations around the world. In OLPC’s case, the advantage of a machine that would withstand rough conditions with little to no opportunity for tech support or replacement outweighed their very strong determination to hold down up-front costs.

The same cost-performance tension OLPC faced also applies to our own educational environment. Distributing mobile computers to hundreds of K-12 students is no task for the timid. Computers in the hands of kids will get shaken and stirred. Solid-state drives offer one more avenue to provide students with technology that will withstand their abuse, but we still have to decide whether we can afford to spend the extra dollars for that advantage.

————————————————-

Bonus link: Kate Bush, “Rubber Band Girl“: not related, but fun to study by.

Object-Oriented Databases

Traditional database models:

  1. hierarchical
  2. network
  3. relational (since 1970, commercially since 1982, works!)

Object-Oriented data models started mid-1990s

  1. we wanted more complex apps (storing more than numbers and text)
  2. we wanted more data modeling features
  3. increased OOPL use sparked OO database desire

Commercial OODb products came up in 1990s, didn’t really hit mainstream: relational model works well enough, very entrenched

OOPLs: Simula, Smalltalk, C++, Java

Experimental OODb systems: Postgres, IRIS Orion, OPen-OODB

Commercial OODB products: Onotos, Gemstone,…

OODB about maintaining direct correspondence between real-world objects and database objects so objects do not lose their integrity and identity and can easily be identified and operated upon.

Object has state (value: what the object knows) and behavior (operations: what the object knows how to do)

OO captures complex structures in simpler abstract containers or names or nouns.

OODB has objects of arbitrary complexity; traditional database can scatter data about a gievn object across numerous records/tables or records.

OOPL defines instance variables (data members in C++) which hold the values that define the object in a given state.

Two parts of operation:

  1. Signature/Interface of operation specifies operation name and arguments (or parameters)
  2. Method/Body specifies implementation of operation — that’s the instructions.

Polymorphism here refers to the ability of an operation to be applied to different types of objects; a.k.a. operator overloading

Every object has a unique object identifier, the OID. This is tricky if you have a big system with multiple machines/users creating multiple objects simultaneously. However you do it, OID should not change (like in real world: individuals don’t change identity! You are always Bob!).

Every object is a triple: (OID, type, value)

Object types:

  1. atom: single value
  2. set: several values
  3. tuple: like line in database
  4. list: ordered set
  5. bag: collection of objects
  6. set: bag with no duplicates

OODB market growing very slowly; OO ideas being used in many apps but not going to full-tilt OODB.

Ultimately, making the computer’s representation of reality match ours does not matter: databases are a matter of efficient and effective storage. If the computer thinks in terms of relational tables and I think in terms of objects, it doesn’t matter, as long as I get the data I want.

Chapter 22: Object-Relational and Extended Relational DBS

SQL3 added object-relational capacity

We usually don’t see the O in the object-relational database systems; vendors just slap on some object capabilities and don’t sweat calling it a whole different system.

The big push for object capabilities is new media, using audio, video, images, BLOBs. But these big files didn’t really drive the development/adoption effort. RDB folks were already finding ways to deal with those big chunks of data. ODB thinking came about because folks were using OO thinking for other problems (programming, which itself went OO because developers were using OO thinking in planning/design!).

—————————-

Last word: Database administration is a service industry! We and the stuff we are in charge of have no intrinsic value! We are valuable only to the extent we help out clients figure out what they need and optimize what they do. We serve our clients.

Disk Storage

Primary file organization: how files are physically stored on disk; where the read/write head has to go to read the file. You want to mess with this as little as possible! Find your data a good home and leave it there. There’s only one way to do this per file: you will store/arrange your records by one particular field (ID, SSN, LName, etc.)

Secondary file organization: auxiliary means of accessing data; indices! You can do all of these you want! Index by City, index by SSN… you basically list the records and their physical locations in order of whatever field suits your needs.

Primary file organization (Chapter 13!)

Things to remember when working with disks :

  1. Disks offer random access, read by block (LBA: Logical Block Address — usu. starts at 0, goes to n-1, where n = # blocks on disk)
  2. We still use tape backup because tape is the cheapest storage per bit!
  3. reading by bits or records is inefficient! We read by blocks!
  4. seek time: time to move head to block
  5. latency (a.k.a. rotation delay): time to spin disk to right positiontion under head
  6. block transfer time: time to move block from disk to head to memory (or back, for write) — usually takes the most time
  7. placing blocks contiguously reduces seek time and latency, but that’s beyond scope of this course, so we will concentrate on reducing the number of block reads and writes
  8. Double buffering allows us to read Block 1, let CPU start beating on it while we load Block 2 into the other buffer. Then the moment CPU gets done beating on #1, it turns to the other buffer and beats on #2, no waiting! This speeds up the total process by allowing processing to happen simultaneously with some data transfer
  9. Non-spanning: records can’t be broken up across blocks. This wastes some block space
  10. Blocking factor: how many records we can fit per block
  11. Fixed-length records waste space, but they are much more efficient to process and are what we will usually use.
  12. Read takes one block transfer; write usually requires two transfers, unless we luck out and the block we want is sitting in memory. Systems may allow pegging: we designate certain blocks as worth keeping around in the buffer.
  13. Choose your primary file organization to make most efficient the operations most important to you (search, insert, update, delete)
    1. Heap: unordered! INSERT by APPEND, records ordered by time created.
      1. Inserting takes 1-2 block reads (find last block, see if there’s room, write here or create new block to write).
      2. Searching is a pain, requires b/2 block reads on average; requires reading all b blocks on search of non-existent file.
      3. DELETE: must first search (avg. b/2). Then must decide what to do with the empty space! Instead of physically deleting the record and either Swiss-cheesing the disk or having to haul records in to backfill, you can just flag the record with a deletion bit. Odds are we’ll use downtime to compress the disk to fill the holes (garbage collection).
    2. Physical sort: ordered on some oft-used field (like EmpID, SSN, ZIP…)
      1. INSERT: pain! You’ve got to move everyone over to make room for new records… or you create overflow file in heap, deal with searching the whole overflow until we have time to do garbage collection during downtime.
      2. search: easier on the sort field (so you use this method when you do a lot of searches on that particular field)
      3. DELETE: again, sped up. Leave flag on DELETEd record. Note that DELETE could allow us to stick in INSERTs if there’s room… but making the computer consider that possibility usually eats more time than just sending every INSERT to the overflow.
      4. Note that physical sort allows binary search! Go to the middle, check whether we need to go below or above, keep halving until we hit target! Binary search will take average log2(b) block reads.
  14. Hashing: way of mapping things on disk, reducing the range of numbers; mod of primary key often used;  good academic topic that we’re not spending a lot of time on!

Secondary File Structures (Indexing! Chapter 14!)

  1. Making index will save time in most cases
  2. Primary Index: built on primary (unique) key
  3. Clustering index: orders in chunks (all the Bobs, all the Marys, etc.)
  4. Primary Index is another table, a separate file, with two fields: the primary (candidate) key and the block number in which the record with that primary key value is stored. Primary key may be the ordering key for the primary storage; doing a binary search on this index is still shorter than doing a binary search on the disk itself (fewer blocks to church through).
  5. Primary index allows binary search! Zoom! Searches now take log2(#blocks in Primary Index) + one more block read from disk.
  6. Remember, the primary index may still be huge (imagine Blogger’s primary index on usernames), but it won’t be as huge (use as many blocks) as the actual database
  7. Dense index: every record in database gets entry in index.
  8. Sparse index: not dense!
  9. A primary index can be sparse: since it uses the primary key, it can simply list the first or last key value (the anchor value) of each block on disk. This is good!
  10. Secondary index uses some key other than the ordering field of the primary file organization. Since the value of the key in this index has nothing to do with the block number, we need to have an index entry for every record.
  11. If search matters, create an index!
  12. You probably still have to read the index from disk. If you can keep the index in memory for lots of opeations, you really speed things up.
  13. Binary search tree: every child to left smaller, every child to right larger; efficiency comes from minimizing the depth of the roots (number of levels necessary = ROUNDUP((log2(n)+1)) where n = # records)
  14. B-tree: index scheme based on binary search tree: idea is that since we live in the world of disks and have to read entire blocks, we might as well pack as much into the block as we can
  15. B+-tree: internal nodes do not store pointer to the data block, which means there’s room for more index values and thus larger fan-out; only the leaf nodes have records paired with block pointers, so every search must go to the bottom of the tree. But the B+-tree’s bigger fan-out gives an advantage!
  16. So, B-tree, you probably need to go to the bottom leaf, but you might get lucky. B+-tree always goes to the bottom leaf. (Somewhere there’s a forumla for this!)
  17. B+-tree also has the advantage that you can sequentially read the bottom row and get every record! B-tree, you’ve got to jump all over to get that complete list.

Next Time: Optimization!

Chapter 6: Relational Algebra (yum!)

plan: expect midterm posted Monday or Tuesday; due Monday following spring break. Open note, open Net….

SQL is based on relational calculus, which asks for results but doesn’t specify an order of operations. Relational algebra is the basic set of operations for the relational model, more prescriptive, spelling out the steps. We’re looking at RA here because that’s what query optimizers (Chapter 15) use to perform the transformations necessary.

All objects in RA are relations: no duplicate tuples (records)! SQL doesn’t mind duplicates, eliminates them with UNIQUE or DISTINCT. Sorting to removing duplicates is expensive! In the old days, we didn’t want to incur that expense, so the guys who came up with SQL let duplicates slide.

sigma σ = SELECT

  • subscript like WHERE clause
  • horizontal partition: selects rows
  • results always have fewer or same number of tuples as original R
  • commutative

PROJECT = pi π

  • vertical partition: selects columns
  • note! this is like a straight SQL SELECT, without WHERE
  • results must be set of tuples, no duplicates
  • results always have fewer or same number of tuples as original R
  • not commutative

You can write RA ops in a big nested expression, or you can step it and name the intermediate results

RENAME = rho ρ

  • renames table and/or columns

UNION ( ∪ )

  • the two relations we unite must be “type compatible”: same number of attributes, and each pair of corresponding attributes must be type compatible (have same or compatible domains: strings, CHARS, integers, etc…. it’s the datatype that matters, not the field name: default is to take on the field name of the first table)
  • Note that usually what you’ll do is PROJECT two disparate tables, then UNION the results, since we don’t leave a bunch of type-compatible tables sitting around in the database

Type compatibility is a requirement for all three set operations, UNION, INTERSECTION ( ∩ ), and DIFFERENCE ( – )

UNION and INTERSECTION are commutative

CARTESIAN/CROSS PRODUCT ( × )

  • has lots of tuples: multiply tuple count from each relation
  • add attribute count from each relation
  • usually not a meaningful relation in itself!

THETA JOIN, EQUIJOIN, NATURAL JOIN (NJ removes duplicate attributes)

SQL.SELECT = PROJECT
SQL.FROM = JOIN
SQL.WHERE = SELECT

We used to use rule-based query optimization, rules of thumb; now we use statistics-based query optimization: we have stats on the tables that inform our choices (see Chaper 15).

Relational databases took off because we can mathematically prove that stuff will work, using relational algebra! It is not the most efficient model (JOINs take time!), but that’s outweighed by the formal definitions that make stuff work.

Chapter 17 (Elmasri & Navathe): Transaction Processing

Assume transactions are correct: we aren’t worrying in this class about bad programming.

Remember that CPUs sit idle during I/O (secondary process, not central process). We are greedy: we want to use those extra CPU cycles. We thus interleave transactions, letting T2 get some CPU time while T1 dinks around elsewhere. So remember: we caused the problems we are trying to solve. :-)

Classic transaction concurrency problems:

  1. Lost Update: T2 writes over results of T1
  2. Temporary Update (a.k.a. Dirty Read): T1 reads X, modifies i; T2 reads X; T1 has error and aborts. T2 operates on dirty data, a value that shouldn’t have existed and got undone by T1’s abort
  3. Incorrect Summary (accountants hate this): I seek aggregate data, down a column/field, for each record. I count 100,000 records. I get to record 50,000, but while I’m still counting, someone updates record #1-100. Or while I’m counting #1-100, someone changes #50,000-51,000. My data is meaningless, unpinnable in time (well, no moer problematic than the U.S. Census
  4. Unrepeatable Read: T1 reads X, gets value, continues, then reads X again, gets different value due to operation of T2. Note that having to read X again probably happens because of memory constraints: you read it, then had to dump buffer, have to go back and get data.

Remember, you don’t get to control interleaving order. The system fits concurrent transactions wherever it can.

Corruption of data due to concurrency is huge problem. When your business relies on data (everybody’s business relies on data), you can’t let this corruption happen! If your database has bad data, it will cause mistakes. And if your users see they’re getting bad data, they will stop using the database. Then you have two problems! If you give your people a data tool that they have to double-check, you’ve given them crap. You and your database are the double-check!

Transaction states:

  1. begin
  2. active
  3. partially committed
  4. committed
  5. terminate

Note that a transaction can abort to the fail state at the active and partially committed state. We create a system log on non-volatile memory (i.e., pull the plug, that info persists) to recover database after transactions fail. System log records transaction starts, reads, writes, commits, and aborts.

Commit point: all operations executed successfully and effect of all operations recorded in system log.

Remember: transactions usually write to buffer, rarely to disk. When transaction commits, its effects may still be only in buffer, not on disk! System may crash before we write the buffer to disk. That’s why the log matters! The log records that info so we can REDO after a crash and fix things with the least possible re-input of data from users.

ACID Properties of Transactions

  1. Atomicity: all or nothing — a transaction happens completely or not at all. If it aborts, you better have a routine that undoes what it managed to do before abort.
  2. Consistency Preservation: transaction takes database from one consistent state to another. (Business rules, program logic… programmers and end users handle this!)
  3. Isolation: transaction acts as if it is the only thing happening, independent of other transactions.
  4. Durability: if transaction commits, it should stick, surviving any failure.

The system (i.e., what we design) has to handle #1, 2, and 4.

Four levels of isolation:

  • Level 0: doesn’t allow dirty reads
  • Level 1: no lost updates
  • Level 2: no dirty reads, no lost updates (level 0 + level 1)
  • Level 3: no dirty reads, no lost updates, no unrepeatable reads (0 + 1 + 2)

Conflicting operations must (1) be in different transactions, (2) access the same data item, and (3) involve at least one write.

Levels of Recoverability (exam: be able to identify schedules by these levels!):

  1. Recoverable: If T2 reads from data written by T1, T1 must commit before T2 commits. If T1 is committed, it never gets rolled back. But note that if T1 aborts, T2 must abort… cascading rollback! Lots of work!
  2. Avoid Cascading Rollback: T2 only reads from T1 if T1 has committed. Only read from committed transactions!
  3. Strict: T2 cannot read or write X until the last transaction that read X has committed or aborted.

Strict is easiest: all you have to do to recover is UNDO. Your log records old value and new value for each write. you then work backward through the log and replace new value with old value.

Transactions done in serial schedule (no interleaving) are by definition correct.

Serializability: equivalence to a serial schedule of the same transactions. We never serialize schedules; we need to know, though, that we could if we (and our CPU) had all the time in the world. Equivalent means more than just result equivalent (p. 628), since you can get the same results from two different schedules given particular data. View equivalence is stronger, says every read operation reads the same value across schedules. We use conflict equivalence, which says conflicting operations happen in the same order.

Remember: there’s overhead for serializability. It takes away interleaving options and slows us down. But it also keeps our data clean. Doing things right takes time!

Chapter 18: Concurrency Control Techniques

Chapter 17 gives the mathematical basis. Chapter 18 tells us how to get stuff done. (Mathematicians can prove the transition.)

Read locks are sharable; write locks are exclusive.

2 Phase Locking:

  • Basic 2PL: growing phase (more locks), the shrinking phase (fewer locks). Once you give up a lock, you can’t ask for any more (it’s all downhill from here). Enforce this rule, and your schedule is serializable and strict recoverable. Unfortunately, gradual growing phases allow deadlocks: T1 locks Y, T2 locks X, T1 asks for X, T2 asks for Y — neither can proceed.
  • Conservative 2PL: growing phase vertical — grab all locks at once! shrinking phase gradual. No deadlocks, but starvation possible: you have to grab all locks at once, and there’s much more chance that at least one of those locks is already taken.
  • Strict 2PL: gradual growing phase, then gradual shrinking phase for read locks but vertical shrinking phase for write locks (drop all write locks at end). Deadlocks back, but strict recoverable, better than basic.
  • Rigorous 2PL: gradual growing phase, then drop all locks at end: guarantees strict recoverable.

Timestamp ordering: bigger numbers mean younger (20100127 > 20091225). Timestamp ordering gets rid of locks.

Chapter 19: Database Recovery Techniques

When you write to the log, it records the old value and the new value. This allows idempotent UNDO: you can do it over and over and getting the same result as if you had done it just once. If the log recorded an operation (“add 5, multiply by 104%”) instead of the old value, your UNDO process could really goof things up! Record the effect, not the cause.

Checkpoint: suspend operations, flush buffers, add checkpoint. This gives us a chance to guarantee that a certain string of data really has been written to the hard drive. When the system crashes, we need recover and UNDO/REDO only as far back as the checkpoint. We need not REDO any transaction that committed pre-checkpoint, because we know its commit made it to the disk. A transaction that committed post-checkpoint definitely wrote to the buffer, but it may or may not have reached the disk, so we should REDO. A transaction that starts and writes some data before the checkpoint does indeed hit the disk! The transaction may continue and hit commit after the checkpoint, but the chekpoint flushes the buffer, including writes from partially complete transactions, so we should only need to REDO the writes between the checkpoint and the commit.

Immediate Update: each operation writes straight to the buffer. This makes more UNDO!

Deferred Update: we collect writes, really write them all at once at the end of the transaction. This is nice for recovery: if a crash hits before a transaction commits, that transaction hasn’t had a chance to change the database. No UNDO or REDO necessary.

Deferred Update with Forced Write: Commit doesn’t happen until we write to disk. Do this, and you UNDO/REDO no committed transactions.

Always record your write of X to disk before you write to disk! Otherwise, you might lose your history of what you did! Checkpoints work backwards: flush buffer, then record checkpoint upon success. If system crashes between flush and checkpoint, you just go back to the previous checkpoint.

Note: MySQL doesn’t have concurrency control. Not everyone needs concurrency control.

Next Page »