Roger E. Thomas Comments

Ho et al. (p. 1776) have made a commendable effort to document and help remedy low influenza immunization coverage among a large sample of individuals aged 65 years or older attending health clinics in Singapore. While stimulating interest among researchers and health practitioners about interventions to prevent influenza in highly vulnerable populations in and beyond Singapore, this article should draw readers’ attention to some methodological issues as well; I attempt to address those problems here, leaving aside the critical issues of optimal influenza vaccine matching,^1,2 immunization timing, and prevention of transmission among travelers.^3–5

I focus on central methodological questions that need to be addressed in further research. Given that the authors reported a 17% influenza vaccination rate for Singapore and very low rates in both their intervention (5.9%) and wait-list control (4.8%) groups, how can influenza immunization uptake in countries with very low coverage rates be measured and increased?

The Ho et al. article illustrates the need for sharpening research methods to best inform public health policy and practice, as follows. First, choose tested interventions and improve them. Ho et al. briefly noted a 2018 Cochrane systematic review and meta-analysis of 61 randomized controlled trials (n = 1 854 698) that focused on influenza vaccination interventions targeting community residents aged 60 years or older.⁶ However, they limited their intervention to displays of posters and flyers and patients bringing posters or flyers to physicians while neither explaining why they chose this intervention nor citing supportive research findings. In a study in China, the key factors influencing vaccination were health care workers’ recommendations, previous influenza vaccinations, public perceptions of the vaccine being safe and effective, and public perceptions of the severity of influenza effects.⁷ The Singapore study could have contributed more to interventions in the Asia–Pacific region if it had built on and improved previous research by broadening the scope of the intervention.

Second, choose a strong design (a randomized controlled trial). The authors did not randomize with a control group but, rather, used a wait-list control design. Randomized controlled trials are designed to control for known as well as unknown confounders. The authors also argued post hoc that there were no important community interventions during their study; however, this information was unknown to them before the study. The group receiving the intervention first thus became an informed group and remained so when it became the wait-list control.

Third, ascertain baseline vaccination rates. Although vaccination in previous years is a key vaccination predictor, the authors did not ascertain baseline vaccination rates from records of clinics or other health care institutions (as a result of the absence of a comprehensive national adult vaccination database). General practice patients are stable attenders, and clinic records could be checked. The experimental and wait-list control groups’ vaccination rates could have been entirely due to their being accustomed to undergoing past vaccinations.

Fourth, ascertain the completeness of delivery of the intervention. For instance, staff motivation and time are among the factors determining whether interventions are delivered completely. The authors reported that they were unable to measure clinics’ true compliance with the intervention owing to limitations in staff capacity for collecting these data.

Fifth, measure and analyze effects of attrition. It is important to identify all eligible participants and the characteristics of those accepting and declining participation. The analysis should be an intention-to-treat analysis. Differential attrition of patients after assignment to the intervention and wait-list control groups should be analyzed.

Finally, identify early large clinic differences in vaccination rates and seek explanations. In Figure 1 of Ho et al., vaccination rates vary from 1% to 21%, and it is impossible to identify the rates that would have been obtained if the intervention had been completely delivered in a randomized controlled trial and in a patient-centered manner.

There is a need for major interventions to enhance uptake of influenza immunization in the aging population at the national level.^6,7 Future studies supporting these efforts should be designed to anticipate and solve the problems described here.