This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Public Health and Surveillance, is properly cited. The complete bibliographic information, a link to the original publication on http://publichealth.jmir.org, as well as this copyright and license information must be included.
Rapid reporting of human infections with novel influenza A viruses accelerates detection of viruses with pandemic potential and implementation of an effective public health response. After detection of human infections with influenza A (H3N2) variant (H3N2v) viruses associated with agricultural fairs during August 2016, the Michigan Department of Health and Human Services worked with the US Centers for Disease Control and Prevention (CDC) to identify infections with variant influenza viruses using a text-based illness monitoring system.
To enhance detection of influenza infections using text-based monitoring and evaluate the feasibility and acceptability of the system for use in future outbreaks of novel influenza viruses.
During an outbreak of H3N2v virus infections among agricultural fair attendees, we deployed a text-illness monitoring (TIM) system to conduct active illness surveillance among households of youth who exhibited swine at fairs. We selected all fairs with suspected H3N2v virus infections. For fairs without suspected infections, we selected only those fairs that met predefined criteria. Eligible respondents were identified and recruited through email outreach and/or on-site meetings at fairs. During the fairs and for 10 days after selected fairs, enrolled households received daily, automated text-messages inquiring about illness; reports of illness were investigated by local health departments. To understand the feasibility and acceptability of the system, we monitored enrollment and trends in participation and distributed a Web-based survey to households of exhibitors from five fairs.
Among an estimated 500 households with a member who exhibited swine at one of nine selected fairs, representatives of 87 (17.4%) households were enrolled, representing 392 household members. Among fairs that were ongoing when the TIM system was deployed, the number of respondents peaked at 54 on the third day of the fair and then steadily declined throughout the rest of the monitoring period; 19 out of 87 household representatives (22%) responded through the end of the 10-day monitoring period. We detected 2 H3N2v virus infections using the TIM system, which represents 17% (2/12) of all H3N2v virus infections detected during this outbreak in Michigan. Of the 70 survey respondents, 16 (23%) had participated in the TIM system. A total of 73% (11/15) participated because it was recommended by fair coordinators and 80% (12/15) said they would participate again.
Using a text-message system, we monitored for illness among a large number of individuals and households and detected H3N2v virus infections through active surveillance. Text-based illness monitoring systems are useful for detecting novel influenza virus infections when active monitoring is necessary. Participant retention and testing of persons reporting illness are critical elements for system improvement.
Novel influenza A viruses are different from currently circulating human influenza A (H1 and H3) viruses and have the potential to cause a pandemic if viruses gain the capacity to infect and transmit efficiently from person to person and cause clinical illness in humans [
To assist US states with monitoring individuals at potential risk of a novel influenza virus infection, CDC, in coordination with the National Association of County and City Health Officials (NACCHO) and Compliant Campaign, developed a text-message illness monitoring (TIM) platform. Although several state health departments had pilot-tested the TIM system, none had previously used it among members of the public during an outbreak. On August 4, 2016, public health officials in Michigan, USA, notified CDC of a laboratory-confirmed influenza A (H3N2) variant (H3N2v) virus infection in a 9-year-old child who had been exhibiting swine at an agricultural fair in Michigan [
On August 9, 2016, the Michigan Department of Health and Human Services (MDHHS) asked CDC to pilot-test the TIM system to monitor families of individuals exhibiting swine at county fairs and evaluate the functionality during an outbreak. The objectives of the pilot were to determine the ability of the TIM system to enhance detections of H3N2v virus infections and to evaluate the feasibility and acceptability of the TIM system for use in future outbreak investigations of novel influenza viruses.
The details of the H3N2v outbreak have been described previously (see
County agricultural fairs that had reported suspected H3N2v infections in the last 10 days during the 2016 H3N2v outbreak.
County fairs, to ensure illness follow-up and testing could be coordinated by a single county health department, that started during our enrollment period—August 15 to September 9, 2016—and had large (>50) numbers of swine in exhibit. Large numbers of swine increase the likelihood of swine influenza outbreaks; we chose a number of more than 50 to reach the upper 50th percentile of swine per fair as determined by Bowman and colleagues [
We identified three fairs with suspected H3N2v virus infections and approached fair organizers about participating in the TIM pilot. Of those, organizers of two fairs agreed to participate (see
As studies have shown that young age and direct contact with swine are risk factors for variant influenza virus infection [
We assigned each fair a unique texting code word. Parents of swine exhibitors were asked to text in the unique code word using their mobile phone, which would automatically enroll them in the system (see
“Yes” responses and any response other than “No” (eg, “OK”) generated an immediate email alert to designated local health department staff to contact the respondent and determine if additional follow-up was necessary. Health department staff were asked to refer anyone for respiratory testing who reported any symptoms associated with variant influenza virus infection, including mild symptoms such as runny nose and less common symptoms such as vomiting, and who had swine exposure or contact with a person with a confirmed H3N2v virus infection during the week prior to illness onset [
We monitored retention of respondents throughout the monitoring period and assessed the impact of reminder texts on daily participation. Respondents could opt out at any time by texting “STOP.” Estimates of user retention were based on formal withdrawals and respondents who ceased responding. To understand the impact of the type of texting language on respondent retention, we randomly assigned fairs to one of two groups where attendees would receive texts with either more formal or less formal language (see
Fair selection. TIM: text-message illness monitoring.
Texting language flow (formal).
Texting language flow (informal). TIM: text-message illness monitoring.
We developed a Web survey to evaluate the use and acceptability of the TIM system using Epi Info software, version 7.2.0.1 (Centers for Disease Control and Prevention). The youth program club coordinators distributed a link to the survey by email to all swine exhibitor families after the conclusion of the fair season. Questions included reasons for participating or not participating in the TIM system and likelihood of participating again if given the opportunity.
We conducted five unstructured interviews with health officials at participating health departments. These were conducted to determine how much time was spent following up on alerts generated by the system and to determine how many respondents with reports of illness sought care and were tested for influenza.
This investigation was determined to be part of a public health response; in accordance with federal human subjects’ protection regulations, it was not considered human subjects research. Participation in the survey was voluntary and anonymous.
Out of an estimated 500 households contacted at the nine selected fairs, 87 (17.4%) households enrolled, reporting for 392 household members (see
Recruitment methods and enrollment by fair.
Fair | Primary recruitment method | Recruitment occurred after fair conclusion | H3N2va infections associated with fair | Swine exhibitors (N=1052), n (%) | Estimated householdsb (N=528), n (%) | Peak household enrollment, n/N (%) | Total under TIMc surveillance (N=392), n (%) |
Fair A | Meeting | No | No | 230 (21.9) | 115 (21.8) | 21/115 (18.3) | 92 (23.5) |
Fair B | Yes | Yes | 200 (19.0) | 100 (18.9) | 11/100 (11.0) | 54 (13.8) | |
Fair C | Local health department visit | No | No | 125 (11.9) | 63 (11.9) | 3/63 (5.0) | 13 (3.3) |
Fair D | Meeting | No | No | 122 (11.6) | 61 (11.6) | 7/61 (11.4) | 31 (7.9) |
Fair E | Yes | No | 120 (11.4) | 60 (11.4) | 8/60 (13.3) | 40 (1.0) | |
Fair F | Local health department visit | No | No | 75 (7.1) | 38 (7.2) | 1/38 (2.6) | 4 (1.0) |
Fair G | Meeting | No | No | 70 (6.7) | 35 (6.6) | 30/35 (85.7) | 132 (33.7) |
Fair H | Flyers | No | No | 55 (5.2) | 28 (5.3) | 1/28 (3.6) | 4 (1.0) |
Fair I | Flyers | No | No | 55 (5.2) | 28 (5.3) | 5/28 (17.9) | 22 (5.6) |
Total | N/Ad | N/A | N/A | 1052 (100) | 528 (100) | 87/528 (16.5) | 392 (100) |
aInfluenza A (H3N2) variant.
bDerived by dividing the number of reported swine exhibitors by 2, based on experts’ estimates of the maximum likely number of swine exhibitors per household.
cTIM: text-message illness monitoring.
dN/A: not applicable.
The number of household respondents for ongoing fairs peaked at 54 respondents on the third day of the fair; after that, the number steadily declined throughout the rest of the monitoring period (see
Of the 392 persons who were actively monitored, illness was reported for 22 (5.6%) through the TIM system. Of those 22, 9 (41%) sought care; 5 out of the 9 that sought care (56%) had a specimen tested and 2 out of those 5 (40%) tested positive for H3N2v virus, representing 17% (2/12) of all H3N2v virus infections detected in Michigan during this outbreak [
Among five fairs that distributed the survey to swine exhibitor families through an estimated 500 email addresses, 70 households (14.0%) who exhibited swine responded. Respondents from 16 out of 70 (23%) households reported participating in TIM. Among the 15 respondents who answered the question regarding motivation for participation, 11 (73%) participated because it was recommended by the fair and 12 (80%) indicated that they would participate again. Among 50 respondents who reported not participating in TIM and responded to the question, 34 (68%) said it was because they had not heard about it, 7 (14%) said they did not understand the necessity, 2 (4%) said they did not have a phone to use, and 7 (14%) said they did not know why they chose not to participate. None of the respondents reported the cost of text messages as a barrier to enrollment.
Health department staff reported spending between 3 and 60 minutes per day following up on alerts generated through the system. Health systems interacting with respondents receiving the informal texting language indicated that they spent substantially more time (ie, up to 60 minutes) following up on alerts than those interacting with respondents receiving the formal texting language (ie, 3-5 minutes), likely due to a substantial number of false alerts from respondents using colloquial language such as “Aok. All fine here,” rather than using a system-recognized response of “Yes” or “No.”
Respondent retention throughout monitoring period: total and by texting language type.
During an outbreak of variant influenza virus associated with swine at agricultural fairs, we successfully deployed a text-based system for illness monitoring among potentially exposed people over a 4-week period. Two H3N2v virus infections were detected among 392 individuals monitored for illness during this outbreak [
Investigation of novel influenza A virus infection is important for early detection, treatment, and prevention of the spread of influenza viruses with pandemic potential, but passive surveillance methods may miss many infections and underestimate the burden of infection associated with exposure. Multiplier models based on an H3N2v virus outbreak in 2011 estimated that for every pediatric H3N2v virus infection detected and reported to CDC, there were approximately 200 infections in the community; for every one adult H3N2v virus infection, there were approximately 255 infections. Reasons for the gap between reported and actual infections were related to the low likelihood that someone who was ill would seek health care and be tested for a variant influenza virus [
Innovative methods, such as participatory, syndromic surveillance systems for influenza, are increasingly being explored as a means to better capture the burden of influenza at the community level [
Systems like TIM could be an important tool in future outbreak investigations. They provide a quick, scalable, and cost-effective method to actively monitor a select group of at-risk people, thereby filling a gap in current surveillance capabilities. Once adopted by a health department, the TIM system can monitor at-risk individuals in an outbreak investigation and be integrated into the outbreak response within hours. As they are intended for detection of illness among a select group of individuals, they complement current influenza surveillance systems that are better adapted for capturing population-level trends and impact [
While the TIM system was well-accepted by both respondents and health departments, we identified three areas where improvements were needed. The first was with enrollment, which varied by fair and ranged from 3% to 86%. The vast majority of households that did not enroll had not heard about the system and those that did enroll did so primarily because it was recommended by the agricultural club coordinator. This highlights the importance of effective communication by people who can reach, and are trusted by, the intended target community. Future deployments of systems like TIM should focus on identifying trusted groups and individuals early and working with them to encourage enrollment. Holding meetings to describe the system and encouraging attendants to enroll during the meeting resulted in higher enrollment, perhaps due to the convenience and personal nature of the recommendation. While strategies to increase enrollment will differ by setting, using personal communication to provide simultaneous rationale and enrollment information appeared to be an effective strategy that could be adapted and applied in most settings.
The second identified area for improvement was testing of symptomatic exposed persons; in our study, only about 20% of ill individuals were tested. Appropriate testing is the only way to confirm infection with a novel influenza virus, as the symptoms are similar to those caused by other respiratory viruses, including seasonal influenza. There are many potential barriers to care-seeking, including lack of motivation to seek influenza testing, the availability of testing at local health care centers or health departments, and the cost associated with visiting a health care provider, especially among people without insurance or with high out-of-pocket health costs. There are several ways to address these barriers and the mechanism depends on the resources and capabilities of the state and local health care systems. In future situations, health officials could consider offering testing for novel influenza virus infections free of charge at health departments or sending clinical staff to collect specimens from ill persons in their homes as a way to address some of these barriers. The latter is a strategy that has been shown to be effective in New York City for longitudinal surveillance of influenza-like illness [
The third issue identified was with respondent retention. Monitoring for a novel influenza virus requires up to 10 days of monitoring after the last day of exposure, which in the case of a 10-day county fair may necessitate as many as 20 days of monitoring. Maintaining respondents throughout the monitoring period was a challenge in this pilot and only 30% remained enrolled through the end of the recommended monitoring period. We deployed two types of texting language to see if the formality of the language could increase retention. We found that respondents receiving the informal language texts stopped responding sooner than those receiving the formal texting language and fewer remained enrolled through the end of the monitoring period. Respondents of the informal texting language also responded to texts in a conversational tone, often with abbreviations or slang terms, which generated alerts through the system that required unnecessary follow-up by health department staff. This resulted in some health department staff reporting that they spent as much as 60 minutes per day following up on alerts versus 3-5 minutes for those interacting with respondents receiving the formal texting language. We recommend that future deployments of TIM-like platforms should use the formal texting language and consider the optimal length of monitoring that weighs the incubation of variant and avian influenza against the likelihood of attrition among the respondents.
This investigation has certain limitations. First, we conducted it in the middle of an outbreak response and relied on health departments and club coordinators to recruit participants in the middle of a busy fair season. In some situations, recruitment occurred solely through email or delivering flyers, which may have limited the number of people who were aware of the system and its importance. Second, this investigation involved adult participants who responded for the entire household. Other studies also have used this method with success [
In summary, we successfully piloted a new text-based monitoring tool for detecting variant influenza virus infections and identified two H3N2v virus infections during an outbreak in Michigan. This pilot demonstrates how text messaging can complement traditional surveillance methods during an outbreak. Future activations of the system should work to improve systematic testing of exposed persons who develop symptoms and should continue exploring methods to improve participation and retention of respondents.
Epidemic curve of influenza A (H3N2) variant virus infections by date of symptom onset and detection method—Michigan, 2016.
Centers for Disease Control and Prevention
influenza A (H3N2) variant
Michigan Department of Health and Human Services
National Association of County and City Health Officials
real-time reverse-transcription polymerase chain reaction
short message service
text-message illness monitoring
We gratefully acknowledge the many people who assisted with the outbreak response and the evaluation of this monitoring tool, including James Averill, LaShondra Berman, Sally Bidol, Lenee Blanton, Andrew S Bowman, Lori Cargil, Beth Ferry, Pat Fralick, Colleen Harns, Britney James, Yungo Jang, Kathy Kacynski, Lisa Koonin, Leah Lavanway, Debra Laws, Leslie Lee, Dian Liepe, Stephen Lindstrom, Amanda Lubit, Mark McCorkle, Michelle McDonald, Guy Miller, Anita Patel, Tracy Payne, Trevor Quinlan, Erin Radke, Carrie Reed, Ruby Rodgers, Mary-Grace Stobierski, Arlee Sutton, Susan Trock, Renee Tschrihart, Kristy VanSickle, Colleen Wallace, and David Wentworth. The findings and conclusions in this report are those of the authors and do not necessarily represent the official views of CDC. Use of trade names and commercial sources is for identification only and does not imply endorsement by CDC, the Public Health Service, or the US Department of Health and Human Services.
None declared.