Participatory Surveillance of Hypoglycemia and Harms in an Online Social Network | Clinical Pharmacy and Pharmacology | JAMA Internal Medicine | JAMA Network
[Skip to Navigation]
Figure. Collecting and returning hypoglycemia reports. A, Screenshot of the survey in TuAnalyze. Initial aggregate responses were blogged about on TuDiabetes. B, Screenshot of the research blog on TuDiabetes. The blog and survey were not co-displayed. Reprinted with permission from Diabetes Hand Foundation.

Figure. Collecting and returning hypoglycemia reports. A, Screenshot of the survey in TuAnalyze. Initial aggregate responses were blogged about on TuDiabetes. B, Screenshot of the research blog on TuDiabetes. The blog and survey were not co-displayed. Reprinted with permission from Diabetes Hand Foundation.

Table 1. Demographic and Disease Correlates for the Total Sample and for Respondents Reporting Recent and Severe Hypoglycemia
Table 1. Demographic and Disease Correlates for the Total Sample and for Respondents Reporting Recent and Severe Hypoglycemia
Table 2. Risks for Harms Among Participants Reporting Frequent Recent Hypoglycemia and Severe Hypoglycemia
Table 2. Risks for Harms Among Participants Reporting Frequent Recent Hypoglycemia and Severe Hypoglycemia
Original Investigation
Mar 11, 2013

Participatory Surveillance of Hypoglycemia and Harms in an Online Social Network

Author Affiliations

Author Affiliations: Children's Hospital Informatics Program at the Harvard-MIT Division of Health Sciences and Technology (Drs Weitzman and Mandl and Ms Kelemen), Divisions of Adolescent Medicine (Dr Weitzman) and Emergency Medicine (Dr Mandl), Diabetes Program (Dr Quinn), and Manton Center for Orphan Disease Research (Dr Mandl), Boston Children's Hospital, Boston, Massachusetts; Departments of Pediatrics (Drs Weitzman, Quinn, and Mandl) and Population Medicine (Dr Eggleston), Harvard Medical School, Boston; and Harvard Pilgrim Health Care Institute and Division of Endocrinology, Diabetes, and Hypertension, Brigham and Women's Hospital, Boston (Dr Eggleston).

JAMA Intern Med. 2013;173(5):345-351. doi:10.1001/jamainternmed.2013.2512

Importance Surveillance systems for elucidating the burden of hypoglycemia are limited.

Objective To quantify experiences of hypoglycemia and related harms, members of an international online diabetes social network with insulin-dependent diabetes mellitus were polled through a software application (“app”). Aggregate results were returned to participants through network channels.

Design The study period was from March 2011 through April 2012, during which time retrospective reports about experiences with hypoglycemia and related harms were collected from participants using the app.

Setting The study was undertaken within the international online diabetes social network.

Participants Eligibility criteria included TuDiabetes membership, age 13 years or older, a self-reported diagnosis of diabetes mellitus, ability to read and write English, and Internet access. Of 2827 app users, 687 (response rate, 24.3%) opted in to the volunteer sample.

Main Outcome Measures Primary outcomes included the following: frequency of “going low” (having a low glucose value in the past 2 weeks) and episodes of severe hypoglycemia (in the past 12 months), and, for respondents reporting recent and/or severe hypoglycemia, lifetime experience of vehicle crashes or severe medical injury, daily debilitating worry, and withdrawal from driving, exercise, sex, and going outside of the home to avoid hypoglycemia and consequences. Secondary outcomes included measures of research engagement.

Results Of 613 respondents (24.3% of app users), 49.1% reported more than 4 episodes of “going low” in the past 2 weeks and 29.2% reported 1 or more severe low in the past year; 16.6% reported both more than 4 recent low episodes and 1 or more severe event in the past year. Harms were common, including daily debilitating worry (45.8%), vehicle crash or injury (15.0%), and withdrawal from exercise, driving, leaving home, and having sex (54.0%, 37.4%, 24.8%, and 22.7%, respectively). Of all respondents, 54.2% reported multiple harms, the risks for which were highest (73.7%) among respondents with a past-year severe event (odds ratio, 2.39; 95% CI, 1.60-3.58; P < .001 controlling for frequent recent low episodes and demographic and disease factors). Engagement was high, with 96.6% of the sample permitting recontact for research and 31.7% posting personal study data on their app profile page; 40.5% of 2825 unique page views of research-related materials published on the community site involved views of returned research results.

Conclusions and Relevance Participatory surveillance of hypoglycemia in an online diabetes social network enables characterization of patient-centered harms in a community sample and bidirectional communication with affected persons, augmenting traditional surveillance.

Hypoglycemia is a serious sequela of diabetes treatment not consistently tracked by formal health surveillance.1 Associated harms span vehicle crashes (car, truck, bicycle, or other) and injuries,2 missed work,3 debilitating worry,4 medical complications, and death.5 Insulin is the second most commonly prescribed drug associated with serious adverse drug reports made to the US Food and Drug Administration (FDA).6 Insulin-associated hypoglycemia accounts for one-third of all US emergency department visits related to medication use by older adults.7 Despite insulin's widespread use and the magnitude of hypoglycemia-related harms, monitoring of hypoglycemia has been limited to emergency department visit surveillance8-10 or adverse event reporting in clinical trials.11,12 Reports about events below the threshold for an emergency department visit and reports from community-based patient samples are lacking.

Pogach and Aron13 call for improved surveillance of hypoglycemia by the Centers for Disease Control and Prevention (CDC) and the FDA. Current CDC surveillance assets such as the Behavioral Risk Factor Surveillance System and the National Health and Nutrition Examination Study do not acquire information specific enough to advance this goal nor use suitably targeted sampling frames; moreover, traditional use of landline-based telephone systems to contact potential study subjects to participate in health interviews now misses important populations.14,15 Voluntary drug problem reporting to the FDA by pharmacists, physicians, and consumers provides some insight into the occurrence of hypoglycemia but is greatly underused.16,17 Reports triggered by scanning electronic health records may augment monitoring but lack specificity.18 Furthermore, standard surveillance approaches are directed toward the patient only and lack channels for patients' feedback and opportunities to collaborate with them to elucidate and quantify patient-centered outcomes.

Dynamic engagement of online communities may augment standard surveillance.19-21 Patients with chronic illnesses increasingly gather in online communities (eg,,, Large “self-organizing” cohorts based in online networks comprise reservoirs of patients uniquely positioned to describe their health care, treatments, adverse effects, symptoms, and behaviors, and surveillance is undertaken with their collaboration.22-25 Patients are highly willing to share electronic data for research and clinical care.26-29 “Crowdsourced” intelligence produces results comparable, in some circumstances, to randomized trials.30 While participatory surveillance of distributed online populations cannot provide generalizable, denominator-based rates, it can characterize risk within online communities using a bidirectional communication channel that enables repolling, feedback, and even rapid-cycle intervention.

We engaged an online diabetes social network in participatory surveillance to measure hypoglycemia-related patient-centered outcomes. Building on prior work engaging this community,31 we used a closed-loop approach of polling and feeding back results through network channels. We quantified both recent and past-year severe hypoglycemia, testing for elevated risks for harms. We sought to understand harms among persons reporting severe hypoglycemia in the past year, a group that may be identified in standard reporting systems as “at risk” for problems. We also assessed harms among persons reporting frequent recent, but not necessarily severe, hypoglycemia, who may not be identified as at risk by traditional surveillance but whose experience of harms may be substantial. Secondary outcomes included measures of community engagement with the research.


We collected structured self-reports of hypoglycemia and related harms using a software application (“app”) called TuAnalyze, which supports privacy-preserving participatory surveillance by members of the international online diabetes social network.22,31 TuDiabetes has more than 25 000 members; 77% reside in the United States, 62% are female, and 72% have either type 1 diabetes mellitus (T1D) or latent autoimmune diabetes in adults (LADA). The median member age is 41 years. TuAnalyze is freely available to TuDiabetes members.22 The app reinforces participation through contextualized feedback of maps and graphs of user-reported glycosylated hemoglobin (HbA1c) values. Summary data from member surveys are returned to the community through a site research blog (the main dissemination vehicle) and through newsletters and postings on the site home page. Consent is obtained within the app for use of member data. Study activities were approved by the Boston Children's Hospital institutional review board. Neither TuDiabetes nor study investigators have any commercial interest in TuAnalyze.

Sample eligibility and participation

Eligibility criteria include TuDiabetes membership, age 13 years or older, a self-reported diagnosis of diabetes mellitus, ability to read and write English, and Internet access. Persons using the app as a proxy for another person, such as family members (n = 74), were excluded from analyses. During the study period of March 2011 through April 2012, 687 of 2827 app users (response rate, 24.3%) took 2 complementary surveys about hypoglycemia and diabetes care using the app (Figure, A). Compared with all TuDiabetes members, TuAnalyze survey respondents were more likely to reside in the United States (86% vs 77% [odds ratio, 1.85; 95% CI, 1.48-2.30]; P < .001), have T1D/LADA (80% vs 72% [odds ratio, 1.56; 95% CI, 1.28-1.90]; P < .001), and were older on average (44 years vs 42 years; P < .001). There was no difference in the sex makeup of survey and source populations. Of the 687 respondents, 613 (89%) reported using insulin and comprise the final sample. Respondents who entered survey data before and after feedback of summary survey data on the site research blog (47% and 53%, respectively) did not differ in age, sex, or type.

Sources of data

Information about recent and severe hypoglycemia, associated harms, disease, and demographic and health care characteristics were collected from respondents using the app. Data about community engagement with the research included the following: (1) total unique page views for the TuAnalyze research blog used to report summary statistics about the frequency of recent hypoglycemia, severe hypoglycemia, and related harms generated from TuAnalyze surveys (tallied using Google analytics for March 2011 through March 2012); (2) the proportion of users permitting recontact through the app for ongoing research; and (3) the proportion of users opting to display on their social network profile page their most recent HbA1c value entered through the app. An export of current TuDiabetes member profile information was obtained from site administrators for analyses of response bias; privacy protections within the app precluded identification and exclusion of TuAnalyze survey respondents from the member list for comparisons of survey and source populations.

Frequency of Recent and Severe Hypoglycemia

Survey respondents reported “recent lows” (low glucose values) as the number of times they had “gone low” in the past 2 weeks. No specific definition for going low was offered in this context, since the goal was to elicit patient-defined episodes of hypoglycemia. Severe hypoglycemia was assessed separately, as the number of occasions in the past 12 months when survey respondents experienced a low blood glucose level resulting in their becoming unconscious, seizing, or requiring glucagon, medical treatment, and/or help from another person to treat the hypoglycemia. This definition is consistent with other published reports of conditions for serious or severe hypoglycemia.32 Because of the nonnormal distribution of these 2 measures, we dichotomized them at the median value into groupings of more than 4 or fewer for “recent lows” and 1 or more or none for “severe lows.”

Hypoglycemia Unawareness

Respondents were asked to report on hypoglycemia unawareness, defined in the survey as the absence of the typical early signs and symptoms of low blood glucose level. Respondents who reported being unsure about their hypoglycemia unawareness (5% of the sample) were classified as hypoaware for analyses. In addition, participants reported on the blood glucose value range at which they typically first start to feel their blood glucose level is “low.” These values were reported in 10-unit increments, from “below 40 mg/dL” to “100 mg/dL or higher” (to calculate as millimoles per liter, multiply by 0.0555).

Disease and Demographic Characteristics

TuAnalyze users indicated their diabetes type as T1D/LADA, type 2 diabetes mellitus (T2D), prediabetes, or gestational diabetes. No insulin-using respondents reported gestational or prediabetes, and therefore none were included in the final sample for these analyses. We dichotomized diabetes type into T1D/LADA and T2D, based on underlying disease mechanism. Self-reported values of HbA1c were collected through the app as described elsewhere31; consistent with American Diabetes Association (ADA) standards,33 respondents with a most recent HbA1c value of 7% or higher were classified as “above target” compared with those with values lower than 7% (7.5% was used for respondents younger than 19 years). Respondents identifying solely as white and non-Hispanic were classified as white. Age was dichotomized at the median of 44 years and diabetes duration at the median of 17 years.

Adherence With Recommended Care

Adherence with clinical practice recommendations was defined as receiving all five 2012 ADA clinical practice recommendations: any history of a pneumonia vaccination, and past-year history of an influenza vaccination, dilated eye examination, foot examination, and lipid profile.33 Self-care measures included typical frequency of self-monitoring of blood glucose (SMBG), dichotomized at the ADA minimum recommendation of 5 or more times per day, and 4 or more visits with a diabetes clinician in the past year.

Harms Attributed to Hypoglycemia

Participants reported on the following: experiencing a vehicle crash or severe injury in their lifetime, worry affecting their daily life, and withdrawal from typical daily functioning in the past 6 months in 4 areas (driving, exercise, sex, and leaving the home, to avoid hypoglycemia and possible adverse consequences).

Data analyses

We used χ2 tests and the 2-sample t test for analyses of response bias. Demographic correlates and harms associated with hypoglycemia were identified using logistic regression. Odds for harms were calculated among respondents reporting frequent recent and severe lows, using multivariate logistic regression. Using a propensity score approach,34 we predicted each respondent's probability of experiencing frequent recent or past-year severe lows, controlling for effects of significant bivariate correlates, including sex, race (white or other), age (continuous), disease duration (continuous), diabetes type (T1D/LADA or T2D), hypoglycemia unawareness, SMBG (≥5 times per day, or fewer), pump and/or continuous glucose monitor (CGM) use. Adjusted odds of harms related to recent lows were estimated controlling for the experience of severe lows, and vice versa, to characterize the effects on outcomes of each category of low. Analyses were conducted in SAS version 9.3 statistical software (SAS Institute Inc).

Demographic characteristics of the sample

Of 613 respondents, the majority were white, located in the United States, and female. Most (88.3%) reported having T1D/LADA (Table 1). The age range of the sample was from 14 to 81 years (mean, 43 years; median, 44 years). Reported disease duration ranged from 0 to 67 years (mean, 19 years; median, 17 years). In their application settings, 96.6% of respondents permitted recontact by the research team and 31.7% chose to have their HbA1c value entered in the application displayed on their TuDiabetes profile page.

Experience and correlates of hypoglycemia in the sample

Reported number of recent lows ranged from 0 to 75, with a median of 4 and a mean of 6.2. More than half (53.2%) of those with T1D/LADA and 18.2% of those with T2D reported experiencing more than 4 lows in the past 2 weeks. Only 9% of the sample reported no lows in the past 2 weeks. Severe lows in the past 12 months were less common, with 70.8% of the sample reporting none and more than one-quarter (29.2%) reporting at least 1—including 30.3% of those with T1D/LADA and 20.8% of those with T2D. Of the total sample, 38.3% reported experiencing 4 or fewer recent lows and no past-year severe lows. One-third of the sample (32.5%) reported more than 4 recent lows and no past-year severe lows; 12.6% reported 4 or fewer recent lows and 1 or more past-year severe low; and 16.6% reported more than 4 recent lows and at least 1 past-year severe event. The median blood glucose value at which respondents reported typically experiencing the first symptoms of hypoglycemia was 60 to 69 mg/dL, reported by almost one-third of respondents (32.3%). Nearly three-quarters of participants (73.2%) reported first feeling low at values between 50 and 79 mg/dL.

In bivariate analyses, experiencing more than 4 recent lows was associated with female sex, younger age, T1D/LADA, insulin pump and/or CGM use, longer disease duration, an HbA1c value under 7%, high daily SMBG, and dissatisfaction with diabetes health care; experiencing severe hypoglycemia in the past year was associated with female sex, CGM use, longer disease length, an HbA1c value higher than 7%, hypoglycemia unawareness, and cost barriers to care (Table 1). A small percentage of respondents (9.6%) did not report an HbA1c value through the application; compared with respondents who had entered an HbA1c value, this group was less likely to report high daily SMBG and more likely to report cost barriers to care, but did not differ on other measures.

Hypoglycemia-related harms

Harms from hypoglycemia were common (Table 2). The most frequently reported harm in the total sample was avoiding exercise, reported by 54.0% of respondents, followed by daily debilitating worry about hypoglycemia, which was reported by 45.8%. These harms were also the most common among those reporting more than 4 recent lows or a past-year severe low. Fifteen percent of respondents reported having an accident or injury as a result of low blood glucose—including 19.7% of those who reported frequent recent hypoglycemia and more then one-quarter (25.7%) of those reporting a past-year severe low. More than half (54.2%) of all respondents reported experiencing more than 1 of the assessed harms. Compared with respondents who did not report frequent recent lows, respondents reporting this were more likely to regularly limit their driving and sexual activity to avoid experiencing hypoglycemia and its consequences. Participants reporting at least 1 severe low in the past year were more likely than those who did not to report an accident or injury as a result of hypoglycemia, daily debilitating worry about hypoglycemia, and 3 of 4 assessed withdrawal behaviors. Experience of multiple harms was common among those experiencing frequent recent lows and severe lows (63.0% and 73.4%, respectively). Risks for multiple harms were highest among respondents with a past-year severe event, in analyses controlling for demographic, disease, and health care factors and experience of frequent recent hypoglycemia.


Almost all users (96.6%) opted to permit research recontact. Approximately one-third of users (31.7%) opted to display their HbA1c value on their app profile page where persons authorized by them could see it. Of 2825 page views of study materials posted in the TuAnalyze research blog, 40.5% involved a review of interim research results (Figure, B). Engagement measures did not differ by experience of recent or severe hypoglycemia.


Using participatory surveillance, we quantified recent and severe hypoglycemia and patient-centered harms among members of an online international diabetes social network. Though our sample is not denominator based and cannot be used to estimate general population rates, reported hypoglycemia levels were similar to those found among carefully selected clinical cohorts: 36.7% of patients with T1D reported experiencing severe hypoglycemia in the past year in a large multicenter British-Danish clinical cohort35; 30.3% of T1D respondents in our sample reported this. In the Veterans Affairs Diabetes Trial of patients with T2D,36 17.6% of participants reported symptomatic (not necessarily severe) hypoglycemia; among 119 T2D TuAnalyze respondents using insulin or an oral agent, 22 (18.5%) reported severe hypoglycemia in the past year. Correlates were consistent with published reports and hypoglycemia was strongly associated with “hypoglycemia unawareness”37,38 and use of a CGM. The latter association likely reflects device use to control an established hypoglycemia risk.39-42 There was no clear relationship between hypoglycemia and report of receiving helpful information about hypoglycemia from one's physician or the social network, self-care patterns, health care interactions, or adherence with clinical practice recommendations. This may reflect limited influence of these factors on hypoglycemia occurrence or be an artifact of the self-selecting nature of the study sample.

Harms were especially evident among respondents reporting a severe event in the past year, among whom risks were elevated for 5 of 6 assessed harms. Daily debilitating worry and withdrawing from normal activities to avoid or mitigate the effects of hypoglycemia were common and may be driving quality of life down and productivity losses up among persons with diabetes—outcomes reported in large sample clinical and prospective cohort studies.4,43 A less common, but extremely serious, harm involves experiencing an accident or injury related to hypoglycemia.

Engagement was high—virtually all respondents opted in to research recontact and one-third shared personal research data with others via their TuDiabetes profile, using standard tools. Many respondents and network members reviewed research results, a proof of concept for establishing bidirectional communication with research participants through an online disease-centered social network.31 Bidirectional communication can include feedback of results, as implemented here, as well as future interventions to ameliorate harms including efforts informed by participatory surveillance and potentially delivered through the social networking medium and engagement of patients in using the application clinically for panel management. Our approach leverages communication channels available within many online diabetes social networks19,44 but privacy controls available within TuAnalyze specifically.31

Patient engagement in disease-focused social networking is strong and growing,45 reflecting both the appeal of this organic and grassroots phenomenon and patients' need for community,46 information, and support.47-49 Harnessing this engagement may enable population health monitoring and clinical care redesign28 through enabling collection of patient-reported data. Use of the inherently open and dynamic system to engage an online community sample in quantifying patient-centered measures may introduce bias related to the inherently opt-in model. Outcome estimates are not generalizable to nonnetwork populations nor network members who opt not to use TuAnalyze.31 Reported levels of type, device use, glycemic control, SMBG, and insurance among the study sample may differ from population-based comparator groups. This may be acceptable when the goal is rapid as opposed to nationally representative health surveillance and preservation of a feedback channel. Self-reports present a validity challenge, as in traditional monitoring. Standards for benchmarking engagement are lacking, as are measures of impact of returning research results on engagement and outcomes. Midstudy feedback to participants may influence subsequent reporting.

In conclusion, quantification of experiences of hypoglycemia and harms in an online diabetes social network enables characterization of serious patient-centered adverse outcomes using an approach that complements traditional surveillance and provides a communication channel back to affected participants. Establishing participatory surveillance and bidirectional communication among a motivated sample using appropriate safeguards for privacy, safety, and autonomy establishes a new toolkit for outcomes research. This may be a boon for monitoring and ameliorating diabetes.

Back to top
Article Information

Correspondence: Elissa R. Weitzman, ScD, MSc, Children's Hospital Informatics Program, One Autumn Street, Room 541, Boston, MA 02215 (

Accepted for Publication: August 28, 2012.

Published Online: February 11, 2013. doi:10.1001/jamainternmed.2013.2512

Author Contributions: Dr Weitzman and Ms Kelemen had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Weitzman, Quinn, and Mandl. Acquisition of data: Weitzman, Kelemen, and Mandl. Analysis and interpretation of data: Weitzman, Kelemen, Quinn, Eggleston, and Mandl. Drafting of the manuscript: Weitzman, Kelemen, Quinn, and Mandl. Critical revision of the manuscript for important intellectual content: Weitzman, Quinn, Eggleston, and Mandl. Statistical analysis: Weitzman, Kelemen, and Mandl. Obtained funding: Weitzman, Eggleston, and Mandl. Administrative, technical, and material support: Weitzman and Mandl. Study supervision: Weitzman and Mandl. Assistance with survey development: Eggleston.

Conflict of Interest Disclosures: None reported.

Funding/Support: This work was supported by grant PO1HK000088-01 from the Centers for Disease Control and Prevention; grants 5R01LM007677 and G08LM009778 from the National Library of Medicine; and grant 1U54RR025224-01 from the National Center for Research Resources, National Institutes of Health.

Role of the Sponsor: The study sponsors had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation, review, or approval of the manuscript.

Cryer PE, Davis SN, Shamoon H. Hypoglycemia in diabetes.  Diabetes Care. 2003;26(6):1902-191212766131PubMedGoogle ScholarCrossref
Cox DJ, Penberthy JK, Zrebiec J,  et al.  Diabetes and driving mishaps: frequency and correlations from a multinational survey.  Diabetes Care. 2003;26(8):2329-233412882857PubMedGoogle ScholarCrossref
Rhoads GG, Orsini LS, Crown W, Wang S, Getahun D, Zhang Q. Contribution of hypoglycemia to medical care expenditures and short-term disability in employees with diabetes.  J Occup Environ Med. 2005;47(5):447-45215891522PubMedGoogle ScholarCrossref
Marrett E, Radican L, Davies MJ, Zhang Q. Assessment of severity and frequency of self-reported hypoglycemia on quality of life in patients with type 2 diabetes treated with oral antihyperglycemic agents: a survey study.  BMC Res Notes. 2011;4:25121777428PubMedGoogle ScholarCrossref
Gerstein HC, Miller ME, Genuth S,  et al; ACCORD Study Group.  Long-term effects of intensive glucose lowering on cardiovascular outcomes.  N Engl J Med. 2011;364(9):818-82821366473PubMedGoogle ScholarCrossref
Moore TJ, Cohen MR, Furberg CD. Serious adverse drug events reported to the Food and Drug Administration, 1998-2005.  Arch Intern Med. 2007;167(16):1752-175917846394PubMedGoogle ScholarCrossref
Budnitz DS, Lovegrove MC, Shehab N, Richards CL. Emergency hospitalizations for adverse drug events in older Americans.  N Engl J Med. 2011;365(21):2002-201222111719PubMedGoogle ScholarCrossref
Simeone JC, Quilliam BJ. Predictors of emergency department and outpatient visits for hypoglycemia in type 2 diabetes: an analysis of a large US administrative claims database.  Ann Pharmacother. 2012;46(2):157-16822234990PubMedGoogle ScholarCrossref
Levine BS, Anderson BJ, Butler DA, Antisdel JE, Brackett J, Laffel LM. Predictors of glycemic control and short-term adverse outcomes in youth with type 1 diabetes.  J Pediatr. 2001;139(2):197-20311487743PubMedGoogle ScholarCrossref
Booth GL, Hux JE, Fang J, Chan BT. Time trends and geographic disparities in acute complications of diabetes in Ontario, Canada.  Diabetes Care. 2005;28(5):1045-105015855565PubMedGoogle ScholarCrossref
Garber A, Marre M, Blonde L,  et al.  Influence of initial hyperglycaemia, weight and age on the blood glucose lowering efficacy and incidence of hypoglycaemic symptoms with a single-tablet metformin-glibenclamide therapy (Glucovance) in type 2 diabetes.  Diabetes Obes Metab. 2003;5(3):171-17912681024PubMedGoogle ScholarCrossref
Holstein A, Plaschke A, Egberts EH. Lower incidence of severe hypoglycaemia in patients with type 2 diabetes treated with glimepiride versus glibenclamide.  Diabetes Metab Res Rev. 2001;17(6):467-47311757083PubMedGoogle ScholarCrossref
Pogach L, Aron D. Balancing hypoglycemia and glycemic control: a public health approach for insulin safety.  JAMA. 2010;303(20):2076-207720501929PubMedGoogle ScholarCrossref
Saaddine JB, Engelgau MM, Beckles GL, Gregg EW, Thompson TJ, Narayan KM. A diabetes report card for the United States: quality of care in the 1990s.  Ann Intern Med. 2002;136(8):565-57411955024PubMedGoogle Scholar
Li C, Balluz LS, Okoro CA,  et al; Centers for Disease Control and Prevention (CDC).  Surveillance of certain health behaviors and conditions among states and selected local areas—Behavioral Risk Factor Surveillance System, United States, 2009.  MMWR Surveill Summ. 2011;60(9):1-25021849967PubMedGoogle Scholar
Budnitz DS, Layde PM. Outpatient drug safety: new steps in an old direction.  Pharmacoepidemiol Drug Saf. 2007;16(2):160-16516634121PubMedGoogle ScholarCrossref
Wysowski DK, Swartz L. Adverse drug event surveillance and drug withdrawals in the United States, 1969-2002: the importance of reporting suspected reactions.  Arch Intern Med. 2005;165(12):1363-136915983284PubMedGoogle ScholarCrossref
Brajovic S, Piazza-Hepp T, Swartz L, Dal Pan G. Quality assessment of spontaneous triggered adverse event reports received by the Food and Drug Administration.  Pharmacoepidemiol Drug Saf. 2012;21(6):565-57222359404PubMedGoogle ScholarCrossref
Weitzman ER, Cole E, Kaci L, Mandl KD. Social but safe? quality and safety of diabetes-related online social networks.  J Am Med Inform Assoc. 2011;18(3):292-29721262920PubMedGoogle ScholarCrossref
Fox S, Purcell K. Chronic Disease and the Internet. Washington, DC: Pew Research Center; 2010
Cobert B, Silvey J. The Internet and drug safety: what are the implications for pharmacovigilance?  Drug Saf. 1999;20(2):95-10710082068PubMedGoogle ScholarCrossref
Weitzman ER, Kelemen S, Mandl KD. Surveillance of an online social network to assess population-level diabetes health status and healthcare quality.  Online J Publ Health Inform. 2011;3(3):1-7Google Scholar
Hand E. Citizen science: people power.  Nature. 2010;466(7307):685-68720686547PubMedGoogle ScholarCrossref
van Noort SP, Muehlen M, Rebelo de Andrade H, Koppeschaar C, Lima Lourenço JM, Gomes MG. Gripenet: an internet-based system to monitor influenza-like illness uniformly across Europe.  Euro Surveill. 2007;12(7):E5-E617991409PubMedGoogle Scholar
Brownstein JS, Freifeld CC, Madoff LC. Digital disease detection—harnessing the Web for public health surveillance.  N Engl J Med. 2009;360(21):2153-2155, 215719423867PubMedGoogle ScholarCrossref
Weitzman ER, Kaci L, Mandl KD. Sharing medical data for health research: the early personal health record experience.  J Med Internet Res. 2010;12(2):e1420501431PubMedGoogle ScholarCrossref
Zulman DM, Nazi KM, Turvey CL, Wagner TH, Woods SS, An LC. Patient interest in sharing personal health record information: a web-based survey.  Ann Intern Med. 2011;155(12):805-81022184687PubMedGoogle Scholar
Kaye J, Curren L, Anderson N,  et al.  From patients to partners: participant-centric initiatives in biomedical research.  Nat Rev Genet. 2012;13(5):371-37622473380PubMedGoogle ScholarCrossref
Weitzman ER, Kelemen S, Kaci L, Mandl KD. Willingness to share personal health record data for care improvement and public health: a survey of experienced personal health record users.  BMC Med Inform Decis Mak. 2012;12(1):3922616619PubMedGoogle ScholarCrossref
Frost J, Okun S, Vaughan T, Heywood J, Wicks P. Patient-reported outcomes as a source of evidence in off-label prescribing: analysis of data from PatientsLikeMe.  J Med Internet Res. 2011;13(1):e621252034PubMedGoogle ScholarCrossref
Weitzman ER, Adida B, Kelemen S, Mandl KD. Sharing data for public health research by members of an international online diabetes social network.  PLoS One. 2011;6(4):e1925621556358PubMedGoogle ScholarCrossref
Bonds DE, Kurashige EM, Bergenstal R,  et al; ACCORD Study Group.  Severe hypoglycemia monitoring and risk management procedures in the Action to Control Cardiovascular Risk in Diabetes (ACCORD) trial.  Am J Cardiol. 2007;99(12A):80i-89i17599428PubMedGoogle ScholarCrossref
American Diabetes Association.  Standards of medical care in diabetes—2012.  Diabetes Care. 2012;35:(suppl 1)  S11-S6322187469PubMedGoogle ScholarCrossref
D’Agostino RB Jr. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group.  Stat Med. 1998;17(19):2265-22819802183PubMedGoogle ScholarCrossref
Pedersen-Bjergaard U, Pramming S, Heller SR,  et al.  Severe hypoglycaemia in 1076 adult patients with type 1 diabetes: influence of risk markers and selection.  Diabetes Metab Res Rev. 2004;20(6):479-48615386817PubMedGoogle ScholarCrossref
Duckworth W, Abraira C, Moritz T,  et al; VADT Investigators.  Glucose control and vascular complications in veterans with type 2 diabetes.  N Engl J Med. 2009;360(2):129-13919092145PubMedGoogle ScholarCrossref
Bolli GB. How to ameliorate the problem of hypoglycemia in intensive as well as nonintensive treatment of type 1 diabetes.  Diabetes Care. 1999;22:(suppl 2)  B43-B5210097899PubMedGoogle Scholar
Kushner P. Minimizing the risk of hypoglycemia in patients with type 2 diabetes mellitus.  Diabetes Metab Syndr Obes. 2010;3:49-5321437076PubMedGoogle ScholarCrossref
Cemeroglu AP, Stone R, Kleis L, Racine MS, Postellon DC, Wood MA. Use of a real-time continuous glucose monitoring system in children and young adults on insulin pump therapy: patients' and caregivers' perception of benefit.  Pediatr Diabetes. 2010;11(3):182-18719958460PubMedGoogle ScholarCrossref
Chase HP, Kim LM, Owen SL,  et al.  Continuous subcutaneous glucose monitoring in children with type 1 diabetes.  Pediatrics. 2001;107(2):222-22611158450PubMedGoogle ScholarCrossref
Fiallo-Scharer R, Cheng J, Beck RW,  et al; Juvenile Diabetes Research Foundation Continuous Glucose Monitoring Study Group.  Factors predictive of severe hypoglycemia in type 1 diabetes: analysis from the Juvenile Diabetes Research Foundation continuous glucose monitoring randomized control trial dataset.  Diabetes Care. 2011;34(3):586-59021266651PubMedGoogle ScholarCrossref
Kashmer L, Clarke W, Gurka M, Elchuri S, Nyer M, Gonder-Frederick L. Predictors of parental interest in continuous glucose monitoring for children with type 1 diabetes.  Diabetes Technol Ther. 2009;11(6):373-37819459766PubMedGoogle ScholarCrossref
Nordfeldt S, Ludvigsson J. Adverse events in intensively treated children and adolescents with type 1 diabetes.  Acta Paediatr. 1999;88(11):1184-119310591418PubMedGoogle ScholarCrossref
Greene JA, Choudhry NK, Kilabuk E, Shrank WH. Online social networking by patients with diabetes: a qualitative evaluation of communication with Facebook.  J Gen Intern Med. 2011;26(3):287-29220945113PubMedGoogle ScholarCrossref
Fox S. Peer-to-Peer Healthcare. Washington, DC: Pew Research Center; 2011
Frost JH, Massagli MP. Social uses of personal health information within PatientsLikeMe, an online patient community: what can happen when patients have access to one another's data.  J Med Internet Res. 2008;10(3):e1518504244PubMedGoogle ScholarCrossref
Seeman N. Web 2.0 and chronic illness: new horizons, new opportunities.  Healthc Q. 2008;11(1):104-11018326388PubMedGoogle Scholar
Chou WY, Hunt YM, Beckjord EB, Moser RP, Hesse BW. Social media use in the United States: implications for health communication.  J Med Internet Res. 2009;11(4):e4819945947PubMedGoogle ScholarCrossref
Idriss SZ, Kvedar JC, Watson AJ. The role of online support communities: benefits of expanded social networks to patients with psoriasis.  Arch Dermatol. 2009;145(1):46-5119153342PubMedGoogle ScholarCrossref