Table 3.

Tools used in pharmacovigilance studies of influenza vaccine.

Country	Tool/initiative	Characteristics (including strengths and limitations)	Example of applications
Australia	AusVaxSafety	An automated active near real-time vaccine safety surveillance system. Uses an opt-out monitoring platform integrated with immunization provider software to issue automated surveys to vaccine recipients or caregivers via text messaging. Capable of independently reporting brand-specific data using participant-reported outcomes.	Used to evaluate safety of quadrivalent inactivated influenza vaccine brands for 2017.41 Data collected from 73,892 subjects (71.8% of all survey recipients); comparable safety outcomes across brands was demonstrated.41
Australia	Text messaging	Text messaging was introduced in the active surveillance FAST-Mum programme in pregnant women in 2013 as an alternative to telephone interview	In one survey, more women responded to text messaging (90%) than to telephone interview (67%).42 Timeliness of data collection was improved with text messaging. Despite the higher response rate with text messaging, women surveyed by this method were less likely to report an AEFI than women surveyed by telephone, with the greatest discrepancies being for injection site reactions and unsolicited events.42
Australia	Vaxtracker	A web-based active surveillance system in children; automates contact with parents by email or text message to answer 11 symptom questions if the child experienced any kind of reaction after immunization. All serious events are followed up by telephone.	In a study during the 2013 influenza season, of 477 children recruited to use Vaxtracker, 61% of parents completed the survey after the first vaccine dose.43 Completion rates were highest when participants provided both email and mobile phone contact details (74%) compared with email (58%) or mobile phone alone (60%). After the first vaccine dose, 8% of respondents reported a local reaction and 3% reported fever. The system allowed rapid analysis of AEFI by health authorities.43
Canada	Mobile phone app	Used in the CANVAS network to complete post-vaccination surveys or report AEFI as they occur rather than waiting until the survey is distributed.39	Evaluated in a proof of concept study.44 A total of 76 people participated, of whom 48 (63%) downloaded the app. Of these 48, 79% completed the day 8 survey, 56% completed the day 30 survey, and 6% completed a joint day 8/30 survey. Eleven participants reported an AEFI, including one spontaneous report that was also reported using the day 8 survey.44 Twenty-one participants completed a usability survey, of whom 86% agreed or strongly agreed that they preferred an app over a web-based system.
US	Text-mining	Used in VAERS for automated classification of reports	In one study, medical officers evaluated 6034 VAERS reports for H1N1 vaccine to determine whether the reports met the Brighton Collaboration case definition for anaphylaxis Text-mining techniques extracted three feature sets, i.e. important key words, low-level patterns and high-level patterns. The authors concluded that this approach could be applied effectively to VAERS data, potentially reducing staff workload and providing more timely information.45
Multinational	Time-to-onset methodology	Used in a GSK database (spontaneous reports of EFI with rotavirus and influenza vaccines). Adverse events were identified as safety signals if their time-to-event distribution was significantly different from the distribution of other events with the same vaccine or from the distribution of the same event for other vaccines.46 The product label was used as a proxy to evaluate a realistic threshold for safety signals.	For the influenza vaccine, 36 safety signals were identified (based on Medical Dictionary for Regulatory Activities preferred terms), of which 11 appeared in the product label (i.e. a true positive signal). This compared with four preferred terms identified using disproportionality analysis, none of which were in the product label. It appeared that the time-to-event method had a higher sensitivity compared with the standard disproportionality method (14.5% versus 0%), but slightly lower specificity (98.0% versus 99.7%).46

AEFI: adverse event following immunization; CANVAS: Canadian National Vaccine Safety; FAST-Mum: Follow-up and Active Surveillance of Trivalent influenza vaccine in Mums; VAERS: Vaccine Adverse Event Reporting System