Facemasks and Hand Hygiene to Prevent Influenza Transmission in Households

Interpandemic human influenza virus infects millions of people every year. Some infections are mild, but others—especially in young or elderly persons—can result in more severe illness requiring hospitalization. Influenza is associated with hundreds of thousands of deaths worldwide annually (1, 2). The 2009 swine-origin influenza A (H1N1) pandemic highlighted the importance of identifying public health measures to mitigate influenza virus transmission.

Many countries would use nonpharmaceutical interventions, including facemasks, improved hand hygiene, cough etiquette, isolation of sick and quarantine of exposed individuals, social distancing measures, and travel restrictions, as their primary means to mitigate an influenza pandemic, particularly at its beginning (3–10). However, data are scarce on the effectiveness of simple personal protective measures, such as facemasks and hand hygiene, against pandemic or interpandemic influenza and on the modes of influenza virus transmission among people (5, 11). After a pilot study in 2007 (12), we conducted a prospective cluster randomized trial to test whether improved hand hygiene or surgical facemasks reduce the transmission of interpandemic influenza in households. We used a cluster design with randomization to interventions at the household level to avoid difficulties in blinding and potential contamination of interventions.

Methods

Design

From 45 outpatient clinics in the private and public sectors across Hong Kong, we enrolled persons who reported at least 2 symptoms of acute respiratory illness (temperature ≥37.8 °C, cough, headache, sore throat, or myalgia); had symptom onset within 48 hours; and lived in a household with at least 2 other people, none of whom had reported acute respiratory illness in the preceding 14 days. After participants gave informed consent, they provided nasal and throat swab specimens, which were combined and tested with the QuickVue Influenza A+B rapid diagnostic test (Quidel, San Diego, California). Participants with a positive rapid test result and their household contacts were randomly assigned to 1 of 3 study groups: control (lifestyle measures), control plus enhanced hand hygiene only, and control plus facemasks and enhanced hand hygiene. Table 1 provides detailed descriptions of the interventions. Data on clinical signs and symptoms were collected for all participants. An additional nasal and throat swab specimen was collected for laboratory confirmation of influenza virus infection by reverse-transcription polymerase chain reaction (RT-PCR).

Table 1. Study Interventions

Randomization lists were prepared by a biostatistician. The households of eligible study index patients were allocated to 3 groups in a 1:1:1 ratio under a block randomization structure with randomly permuted block sizes of 18, 24, and 30 by using a random-number generator (R software, R Development Core Team, Vienna, Austria). Interventions were assigned to households by the study manager on the basis of the randomization sequence. The allocation to specific intervention groups was concealed to recruiting physicians and clinics throughout the study. Participants and people who administered the interventions were not blinded to the interventions, but participants were not informed of the specific nature of the interventions applied to other participating households.

After randomization, a home visit was scheduled within 2 days (ideally within 12 hours) to implement the intervention and to collect informed consent, baseline demographic data, and nasal and throat swab specimens from all household members 2 years of age or older. During the home visit, index patients and household contacts were instructed in the proper use of a tympanic thermometer. During the 6 days after the initial home visit, all household contacts were asked to keep daily symptom diaries. Further home visits were scheduled around 3 and 6 days after the baseline household visit to monitor adherence to interventions and to collect further nasal and throat swab specimens from all household members regardless of illness. During the final home visit, study nurses collected and reviewed symptom diaries, and they evaluated adherence to interventions by interview and by counting the number of surgical masks remaining and weighing the amount of soap and alcohol left in bottles and dispensers. Households were reimbursed for their participation with a supermarket coupon worth approximately U.S. $25.

All participants 18 years or older gave written informed consent. Proxy written consent from parents or legal guardians was obtained for persons 17 years or younger, with additional written assent from those 8 to 17 years of age. The study protocol was approved by the institutional review board of The University of Hong Kong and the Hospital Authority Hong Kong West Cluster.

Outcome Measures

The primary outcome measure was the secondary attack ratio at the individual level: the proportion of household contacts infected with influenza virus. We evaluated the secondary attack ratio by using a laboratory definition (a household contact with a nasal and throat swab specimen positive for influenza by RT-PCR) as the primary analysis and 2 clinical definitions of influenza based on self-reported data from the symptom diaries as secondary analyses (12). The first definition of clinical influenza was at least 2 of the following signs and symptoms: temperature 37.8 °C or greater, cough, headache, sore throat, and myalgia (13); the second was temperature 37.8 °C or greater plus cough or sore throat (14). An additional secondary outcome measure was the secondary attack ratio at the household (cluster) level: the proportion of households with 1 or more secondary case.

Laboratory Methods

Specimens collected from index patients at recruitment were stored in a refrigerator at 2 to 8 °C. Specimens collected during home visits were stored in an ice chest with at least 2 ice packs immediately after collection. Before the end of the day of a home visit, study nurses obtained samples to the nearest collection point for storage in a refrigerator at 2 to 8 °C. Samples stored at 2 to 8 °C in ice chests were delivered to the central testing laboratory at Queen Mary Hospital by courier. Samples were eluted and cryopreserved at −70 °C immediately after receipt. All specimens were tested by RT-PCR for influenza A and B viruses using standard methods (15–17). The Appendix provides additional details of the laboratory procedures that we used.

Statistical Analysis

On the basis of data collected in our pilot study (12) and other studies with similar design (18, 19), we assumed that 10% to 15% of household contacts in the control group would develop RT-PCR–confirmed influenza, with an average household size of 3.8 and an intracluster correlation coefficient of 0.29. Specifying 80% power and a significance level of 5%, we aimed to follow 300 households in each intervention group to allow us to detect differences in secondary attack ratios of 35% to 45%, depending on the actual secondary attack ratios in the control group (15% or 10%, respectively). Recruiting 100 or 200 households to each group would allow 80% power to detect 55% to 70% and 45% to 55% differences in secondary attack ratios, assuming a secondary attack ratio of 10% to 15% in the control group.

To evaluate and compare secondary attack ratios by intervention group, we estimated 95% CIs by using a cluster bootstrap technique with 1000 resamples (20) and chi-square tests and multivariable logistic regression models adjusting for potential within-household correlation (21, 22). We estimated the intracluster correlation coefficient from the mean squared errors in the secondary attack ratio between and within households (21). For the multivariable logistic regression models, we used forced-entry methods to include plausible confounders, including the intervention allocated, the age and sex of the household contacts and their corresponding index patients, vaccination status of the household contacts, and antiviral use in corresponding index patients, whereas missing data on the exact age of 14 household contacts were imputed by comparison with their relationship with the index patient or occupation. Participants were analyzed in the group to which they were randomly assigned, regardless of adherence to the intervention or use of hand washing or facemasks in groups not assigned that intervention.

Our protocol specified that households with more than 1 member with RT-PCR–confirmed influenza virus infection at baseline (co–index patients) or index patients in whom influenza virus infection could not be confirmed by RT-PCR would be excluded from analyses. We excluded from analyses participants who dropped out before receiving the intervention and the few participants who dropped out after the intervention but before data on the primary outcome measure were collected (23). In sensitivity analyses, we analyzed all households in which the intervention was applied, using multiple imputation for unobserved outcomes (24) and including an additional explanatory variable for households with more than 1 index patient. Statistical analyses were conducted in R, version 2.7.1 (R Development Core Team).

Role of the Funding Source

The study was funded by the Centers for Disease Control and Prevention; the Research Fund for the Control of Infectious Disease, Food and Health Bureau, Government of the Hong Kong SAR; and the Area of Excellence Scheme of the Hong Kong University Grants Committee. The sponsors had no role in data collection and analysis, or the decision to publish, but the Centers for Disease Control and Prevention were involved in study design and preparation of the manuscript.

Results

We recruited 2750 potential index patients from 2 January through 30 September 2008; recruitment increased during periods of peak influenza activity in February and March and July and August (Appendix Figure 1).

• Download figure
• Download PowerPoint

The Figure shows the study flow. Of the 2750 potential index patients, 407 (14.8%) had influenza A or B virus infection according to the rapid test; these persons and their households were randomly allocated. In an unintentional deviation from that protocol, 49 of the 407 persons had a household contact with influenza symptoms at recruitment (a potential co–index patient). We also randomly assigned 6 of 407 persons who had symptoms for slightly more than 48 hours.

• Download figure
• Download PowerPoint

After random assignment, 76 (19%) of the households declined home visits or could not be contacted after numerous repeated attempts. We implemented the interventions in 331 households. After initial home visits, 9 households declined further participation and were excluded from analyses. Thus, 322 (97%) households completed follow-up. Influenza could not be confirmed by RT-PCR in 16 of 322 index patients in these households at baseline, and those 16 households were excluded (Figure 1). A further 47 households were excluded because 1 or more contacts had RT-PCR–confirmed influenza virus infection at baseline. Three household contacts declined to participate and were excluded from analyses. We evaluated and compared secondary attack ratios in the remaining 259 (64%) households, which included 794 household contacts. One hundred sixty (62%) index patients had influenza A virus infection, and 99 (38%) had influenza B virus infection.

Participants

Table 2 shows the characteristics of all randomly assigned index patients and of the index patients and household members that were retained in the main analysis. In general, the groups were similar. Around two thirds of index patients were children.

Table 2. Participant Characteristics

The median household size was 4 persons (interquartile range, 3 to 5 persons). A median of 1 child (interquartile range, 1 to 2 children) lived in the analyzed households in each intervention group. The median size of a household's apartment was 700 square feet (interquartile range, 581 to 1000 square feet), and the mean residential density index, defined as the number of household members divided by the household size, was 0.6 (SD, 0.3) persons per 100 square feet; this did not differ substantially or significantly between intervention groups.

Most of the initial home visits were completed within 12 hours of recruitment (Appendix Table 1). The interval between symptom onset and intervention did not significantly differ between the intervention groups (data not shown).

Appendix Table 1. Interval Between Symptom Onset in Index Patients, Random Assignment, and Application of the Intervention

Influenza Transmission

Overall, 60 (8%) household contacts in 49 (19%) households developed RT-PCR–confirmed influenza virus infection during the follow-up period, including 7 households with 2 secondary cases and 2 households with 3 secondary cases; 150 (19%) and 41 (5%) contacts met the 2 definitions of clinical influenza. There were no significant differences between intervention groups in contact infections when any of the influenza definitions were used (Table 3). Among 597 household contacts of 188 index patients who were children 15 years or younger, there were 54 (9%) secondary cases (17 siblings [secondary attack ratio, 15%], 26 parents (8%), 10 live-in domestic helpers (9%), and 1 aunt (2%). Among 197 household contacts of 71 adult index patients, there were 6 (3%) secondary cases (2 children [secondary attack ratio, 4%]) and 1 spouse [4%]). Secondary attack ratios did not significantly differ at the household level (24% in the control group, 14% in the hand hygiene group, and 18% in the facemask plus hand hygiene group; P = 0.37).

Table 3. Secondary Attack Ratios of RT-PCR–Confirmed Influenza Virus Infection and Clinical Influenza

Table 4 shows the adjusted odds ratios of RT-PCR–confirmed influenza virus infection or clinical influenza in household contacts by intervention group, allowing for within-household correlation. The risk of RT-PCR–confirmed influenza virus infection did not differ significantly between intervention groups, but it was significantly higher for children 6 to 15 years of age, and there was a nonsignificant higher risk for influenza virus infection for contacts in households in which the index patient was a child.

Table 4. Risk for Influenza Virus Infection in Included Households

In a subgroup analysis planned before study implementation (12), we found a significant reduction in RT-PCR–confirmed influenza virus infections in the household contacts in 154 households in which the intervention was applied within 36 hours of symptom onset in the index patient (Table 3). The significant difference between the treatment groups was also observed for the first definition of clinical influenza and seemed to be attributable to fewer infections in the facemask plus hand hygiene group (adjusted odds ratio, 0.33 [95% CI 0.13 to 0.87]) (Table 5). No significant difference was found between the facemask plus hand hygiene group and the hand hygiene group in RT-PCR–confirmed influenza virus infections in household contacts (odds ratio, 0.72 [CI, 0.21 to 2.48]). In an exploratory analysis, we found a borderline nonsignificant difference between intervention groups in RT-PCR–confirmed influenza virus infections among household contacts in which the intervention was applied within 48 hours of symptom onset in the index patient (Appendix Table 2).

Table 5. Risk for Influenza Virus Infection When the Intervention Was Applied Within 36 Hours of Symptom Onset in the Index Patient

Appendix Table 2. Secondary Attack Ratios for RT-PCR–Confirmed and Clinical Influenza When the Intervention Was Applied Within 48 Hours of Symptom Onset in the Index Patient

Consistent results were found in separate analyses of household contacts of index patients with influenza A or B virus infection (Appendix Tables 3 and 4). The reductions were not statistically significant in the smaller number of household contacts of index patients with influenza B virus infection.

Appendix Table 3. Secondary Attack Ratios for RT-PCR–Confirmed and Clinical Influenza A Virus Infection

Appendix Table 4. Secondary Attack Ratios for RT-PCR–Confirmed and Clinical Influenza B Virus Infection

In sensitivity analyses, we compared secondary attack ratios by using combinations of RT-PCR or clinical influenza outcomes in household contacts, by intervention group (Appendix Tables 5, 6, and 7). When the intervention was applied within 36 hours of symptom onset of the index patient, we found significant differences between groups in influenza infections that were both RT-PCR confirmed and met the first clinical definition. We also found significant differences in influenza virus infections that were either RT-PCR confirmed or met the first clinical definition, or both. In additional sensitivity analyses on all 331 households in which the intervention was applied, results were similar to the main findings (Appendix Tables 8 and 9).

Appendix Table 5. Secondary Attack Ratios for RT-PCR–Confirmed and Clinical Influenza Virus Infection When Composite Definitions Are Used

Appendix Table 6. Risk for Influenza Virus Infection in the Overall Sample, Using a Composite Definition of Infection

Appendix Table 7. Risk for Influenza Virus Infection When the Intervention Was Applied Within 36 Hours of Symptom Onset in the Index Patient, Using a Composite Definition of Infection

Appendix Table 8. Risk for Influenza Virus Infection in All Households That Received the Intervention

Appendix Table 9. Risk for Influenza Virus Infection When the Intervention Was Applied Within 36 Hours of Symptom Onset in the Index Patient

Adherence

At the final home visit, the intervention groups reported higher adherence to the interventions than the control group. Self-reported data were consistent with measurements of the amount of soap, alcohol hand rub, and facemasks used (Table 6). As part of their symptom diaries, participants in the intervention groups reported daily adherence to the respective interventions; improved hand hygiene was maintained throughout follow-up and was similar among index patients and contacts (Appendix Figure 2). Adherence to the hand hygiene intervention was slightly higher in the hand hygiene group than the face-mask plus hand hygiene group (Appendix Table 10). Index patients reported greater use of facemasks than household contacts, particularly during the first few days of follow-up (Appendix Figure 2). Adherence was similar in the subgroup of households in which the intervention was applied within 36 hours of symptom onset in the index patient (Appendix Table 10).

Table 6. Summary Measures of Adherence to Interventions During the 7-Day Follow-up Period

• Download figure
• Download PowerPoint

Appendix Table 10. Summary Measures of Adherence to Interventions During the 7-Day Follow-up Period in Households in Which the Intervention Was Applied Within 36 Hours of Symptom Onset in the Index Patient

Discussion

We report the largest study to date of the efficacy of facemasks and hand hygiene to prevent influenza virus transmission in households. Overall, the interventions did not lead to statistically significant reductions in household transmission, although we did observe statistically significant reductions where interventions were applied early after symptom onset in the index patient. The strengths of our study include laboratory confirmation of secondary influenza virus infections and the community setting with outpatient-based recruitment, which allows broad generalizability.

Our study design resulted in delays between symptom onset in the index patient and application of the interventions; thus, although adherence was incomplete, we have probably underestimated the true effectiveness of these simple interventions. Our results suggest that substantial clinically significant reductions in household infections could result if the interventions are applied soon after symptom onset (Tables 3 and 5), which is plausible given that infectiousness may be highest soon after symptom onset (25–27). Although our results suggest a benefit of hand hygiene and facemasks in combination if applied early, our study cannot precisely distinguish the relative contributions of the 2 interventions. A recent prospective trial of 143 households reported a protective effect of facemasks against a clinical outcome measure in the per-protocol (as-treated) analysis, although no evidence of efficacy was found by intention-to-treat analysis or in laboratory-confirmed respiratory virus infections (28).

In addition to statistically significant differences between the intervention groups in the primary outcome measure of RT-PCR–confirmed infections, we observed statistically significant differences between groups when we used the first definition of clinical influenza but not the second definition (Table 3). Symptom-based outcomes can lack specificity for influenza virus infections (12, 29), and the interventions in our study aimed to reduce influenza virus transmission within households and may not have been effective in preventing other respiratory infections outside the home. Another possible explanation is that our study lacked statistical power to identify differences in the second clinical definition, with few patients meeting the stricter criteria of fever plus cough or sore throat.

As in our pilot study (12), adherence to the interventions varied. We observed contamination between groups, because both interventions were practiced to some degree in the control group. Only half of the index patients in the facemask plus hand hygiene group reported regular use of a surgical mask during follow-up. Facemask adherence among household contacts was lower. Adherence to the hand hygiene intervention seemed low compared with rates recommended in health care settings but was similar to rates in previous community studies (30–32). In addition, effects in our study may tend toward a lower bound on the effects that might be observed in a pandemic with heightened public awareness (28). It is important to find ways of improving adherence for future studies.

Limitations of our study design include the potential bias from recruiting symptomatic persons, which has 3 effects. First, by using a point-of-care rapid test to detect influenza virus infection, we might preferentially have included index patients with higher viral shedding (33). However, statistical power would generally be increased if index patients were more infectious, because we might observe more household transmission; the limitation thus relates more to generalizability. Second, our study design resulted in an unavoidable delay between onset of symptoms in the index patient and the application of interventions; this may have led to underestimation of their true effects, as suggested by our statistically significant finding of reduced infection when interventions were implemented within 36 hours. Our household sample may have been biased toward including household contacts with preexisting immunity, because in households in which all contacts were susceptible, there might have been more possibility of secondary cases being observed before the index patient presented to a primary care provider (12). Our primary outcome measure is based on laboratory confirmation of influenza by RT-PCR, with specimens collected from home visits at 3-day intervals, and some infections may have been missed if peak viral shedding in the respiratory tract occurred between home visits. We may have missed secondary infections that occurred 7 days or more after illness onset in the index patient. In addition, collection of poor-quality specimens or degeneration during transport or freezing could have reduced RT-PCR sensitivity. Finally, we did not evaluate other facemasks or respirators, such as P2 or N95 masks; these might be more effective than surgical facemasks, although fit testing is usually required and adherence could be difficult to maintain (28).

Several issues should be considered when planning further studies of nonpharmaceutical interventions. We recruited index patients from outpatient clinics, and recruitment was therefore driven by influenza incidence (Appendix Figure 1). This could be problematic in temperate locations with shorter and more intense influenza seasons, where delays between recruitment and intervention may dilute effects. An alternative approach would be to recruit a cohort of uninfected households before an influenza season. However, a much larger sample would be needed, given the low attack rate of influenza. Studies over multiple influenza seasons are useful to allow for variability in incidence rates from year to year. It is challenging to obtain longitudinal laboratory specimens from participants with repeated home visits, but relying on clinical symptoms to guide testing may not yield results specific for influenza. Paired serology could be compared to determine influenza infections during follow-up; this was not feasible in our study.

In conclusion, our results suggest that hand hygiene and facemasks can reduce influenza virus transmission if implemented early after symptom onset in an index patient. During a pandemic, resources may not be available to isolate all infected individuals, and home isolation of some patients may be required. Our results directly inform the personal protective measures that should be taken in such a scenario and support the use of these nonpharmaceutical interventions in public health control measures against interpandemic influenza in annual epidemics.