The household plays an important role in influenza transmission due to the high frequency and intensity of contact. It has been estimated that in the US, approximately 30% of transmission occurs in the household.¹ The household also provides a useful setting to track influenza infections among close contacts of cases because the number of individuals exposed to the index case can be defined, exposure is likely similar, and identification of household contacts is logistically feasible.

Prior transmission studies with secondary cases of laboratory-confirmed seasonal and pandemic influenza have shown secondary infection risk (SIR) among household contacts to range from 4 to 10% when based on laboratory-confirmed illness,^2–4 and from 13 to 30% when estimated based on symptoms.^1,3,5–8 Thus far, the mean estimated serial interval found in household transmission studies on influenza has been 2–4 days.^2,9–13 To date there have been few reports on household transmission of seasonal influenza in middle-income countries such as South Africa.^14–18 South Africa has a temperate climate and influenza is highly seasonal circulating in the Southern hemisphere winter between May and September each year.¹⁹ The estimated annual incidence of influenza-associated hospitalization for lower respiratory tract infection was approximately 50 per 100,000 population during non-pandemic years (2010–2011), and data from national surveillance indicate that 9% of patients hospitalized for acute lower respiratory tract infections had respiratory specimens that tested positive for influenza.²⁰ However, there are few data describing the household transmission of influenza virus in South Africa.¹⁴

We aimed to determine the SIR, serial interval, and associated risk factors for influenza transmission in household contacts of individuals infected with seasonal influenza in communities in South Africa.

We performed a case-ascertained household transmission study during May–October 2013.

Thirty index cases were enrolled and 110 eligible household contacts were identified (Fig. 1). Seven (23%) index cases were ≥5 years of age and had 20 household contacts; 23 (77%) index cases were >5 years of age and had 90 household contacts. We had complete laboratory data for all 30 index cases, and completed questionnaires for 29 index cases. Of the 110 eligible household contacts, 107 (97%) were enrolled. Three household contacts declined due to lack of availability for follow-up visits, however consented to completing symptom and contact diaries. Baseline questionnaires were completed for 106 (99%) of 107 enrolled household contacts and risk factor questionnaires were completed by 87 (81%) of 107 enrolled household contacts.

Households had a median of 5 (range 3–7) members including the index case, 5 total rooms (range 1–9), and 3 (range 1–7) rooms for sleeping (Table 1). Sixty-six percent (19/29) of index cases were female with a median age of 16 years (IQR 6–37) and 30% (9/29) were confirmed to be HIV-infected by laboratory testing (Table 2). Sixty percent (64/106) of household contacts were female with a median age of 23 years (IQR 12–43); 11% (12/106) were HIV-infected (3 tested, 8 from patient-held medical records, 1 self-report, 50 declined response) (Table 2). Among index cases, 3% (1/29) reported receiving one dose of influenza vaccine in the previous 12 months (which included the current influenza season). Amongst household contacts, 5% (5/106) reported having received one dose of the influenza vaccine in the previous year (including the current influenza season). At the beginning of the study, 27 (25%) household contacts reported sharing a bed with the index case, and 30 (28%) reported sharing a cup or plate with them. Of those who completed the risk factor questionnaires at the end of the study, most (62/82; 76%) reported that they had never avoided contact with the index patient and 43 (52%) had shared a bed with the index case during the study period.

Symptom diaries were completed by 82/110 (75%) eligible household contacts, and contact diaries by 74/110 (67%) eligible household contacts. Among responders, 60 reported new onset ARI during the study, yielding a SIR for ARI of 55% (CI 45–64%) (60/110) for all ages, assuming non-responders did not have symptoms. The median age of symptomatic household contacts was 19 years (IQR 11–39) and 38 (63%) were female. There was no significant difference in SIR for ARI in household contacts by age group of the index case (index cases ≥5 years SIR 43% (9/21) vs index cases >5 years SIR 55% (51/92) p = 0.34).

Of the 60 (55%) household contacts reporting ARI, 16 (27%) completed additional risk factor questions assessing change in behavior after symptom onset. Fourteen (88%) reported that they never avoided household contacts and 11 (73%) never slept in a separate bed from other household contacts after symptom onset. In addition, 14 (88%) never avoided school, work, or social gatherings to protect others from catching their illness. Twelve (75%) reported always or often covering their mouths when they coughed/sneezed.

Of the 107 enrolled household contacts, 24 (22%) were positive for influenza by rRT-PCR during the study period. In three cases, the subtype of the household contact did not match that of the index case, and those were excluded from the estimation of SIR as it was likely that this infection was due to transmission from the community and not the index case. There were no co-primary cases (household contacts reporting symptom onset on the same day as the index case). Eligible contacts who were not enrolled were assumed to be negative for influenza rather than excluded to ensure SIR would not be falsely elevated. There were 31 (29%) participants missing one laboratory sample, 7 (7%) missing two laboratory samples, and 1 (1%) missing three laboratory samples. In total, 21 of 110 household contacts (including asymptomatic cases) were positive for the same influenza subtype as the index case by rRT-PCR in 16 households and were included in the analysis, yielding a SIR for laboratory-confirmed influenza of 19% (CI 12–27%). Sensitivity analysis excluding the three eligible household contacts that were not enrolled from the analysis resulted in an SIR of 20% (CI 12–27%). Five households had two secondary cases, and 11 households had one secondary case. The median age of household contacts with laboratory-confirmed influenza was 18 years (IQR 3–51), and 11 (52%) were female. When stratified by age, the SIR for laboratory-confirmed influenza in household contacts with index cases ≥5 years old was 30%, and with index cases >5 years old was 17% (p = 0.17).

Of the 21 household contacts with laboratory-confirmed influenza, 18 were infected with influenza A. Of these, 9 (50%) were H1N1 subtype, 7 (39%) were H3N2 subtype, 2 were unable to be subtyped. The remaining 3 (14%) were influenza B. There was no significant difference in SIR by subtype (17% (H1N1), 16% (H3N2), and 21% (influenza B), p = 0.89).

Of the 21 laboratory-confirmed index cases, data on index case and household contact symptom onset were available for 8 index case-household contact pairs. The mean serial interval for transmission of laboratory-confirmed influenza was 2.1 days (standard deviation = 0.35, range 2–3 days; Fig. 2). Of the 60 cases with ARI, data on index and household contact symptom onset were available for 44 index case-household contact pairs. The mean serial interval for ARI was 2 days (standard deviation = 2.3, range 1–11 days; Fig. 2).

Fifteen (71%) of the 21 household contacts with laboratory-confirmed influenza completed symptom diaries. Of these 15, 3 (20%) were asymptomatic. The most common symptoms reported by household contacts with laboratory-confirmed influenza were cough (14/15, 93%), nasal discharge (13/15, 87%), and fever (10/15, 67%). Overall, of the 60 who reported symptoms (including laboratory-confirmed influenza and ARI), the most common symptoms were cough (51/60, 85%), nasal discharge (48/60, 80%), and myalgia (38/60, 63%). Other symptoms reported included headache, sore throat, and diarrhea. Diarrhea was the least commonly reported symptom amongst both those with laboratory-confirmed influenza (3/21, 14%) and those with ARI (18/60, 30%).

Of the 21 household contacts with laboratory-confirmed influenza, 13 (62%) completed contact diaries. Table 3 shows the amount of time spent and activities done with the index case for laboratory-confirmed influenza cases and household contacts without influenza during the first 3 days of enrollment. On Day 3 of the study, more household contacts with confirmed influenza reported spending most of the day in the home with the index case than well household contacts (p = 0.01). There were no other significant differences between household contacts with or without laboratory-confirmed influenza on time spent or activities done with the index case (Table 3).

We found an SIR of 19% for laboratory-confirmed seasonal influenza in South African households. This is higher than in prior studies in other countries, which range from 4 to 10%.^2–4 This difference in infection risk could be due to many factors that have been shown to be risk factors for increased influenza transmission, such as younger age of index case, larger number of household members, and crowded sleeping conditions, which may occur more frequently in low and middle-income settings.^9,15,25

Our SIR for ARI (55%) was much higher than what we found for laboratory-confirmed influenza (19%), but similar to what was seen when comparing studies that report SIR based on ARI alone^1,3,5–8 versus those which are laboratory-confirmed. ^2,3 However in ARI, unlike with laboratory-confirmed cases, we were unable to exclude cases that may not have been related to the index cases. SIR for ARI of up to 38% and for laboratory-confirmed influenza up to 13% have been reported in prior studies.^1,3,6–8 Our findings are consistent with the range of results reported by a study done on pandemic influenza in South Africa.¹⁴ In that study, the authors were unable to perform confirmatory testing on all cases, but estimated a SIR of pandemic influenza ranging from 10% (laboratory-confirmed cases) to 17% (including symptomatic cases who did not have confirmatory testing), and a serial interval of transmission of 2.3 days.¹⁴

In contrast a previous study in Hong Kong¹⁰ showed that 80% (12/15) of laboratory-confirmed contacts who completed symptom diaries reported symptoms. This may indicate that asymptomatic carriers may not be a driver of transmission in these communities. This indicates that laboratory-confirmation, while difficult, is important in giving a more accurate picture when calculating SIR. As there are typically many respiratory viruses circulating during the winter season,²³ it is not surprising that there is a high rate of ARI in individuals testing influenza negative. Similar to other studies^2,4,9,26, our study reported a higher SIR in laboratory-confirmed cases when the index case was ≥5 years old, although this was not statistically significant, likely due to low numbers of enrolled cases.

A study in Mongolia, which similar to South Africa is a low to middle-income country, found a much lower SIR of 6%; however, Mongolia is a sparsely populated country that may have less household crowding.²⁶ In our study 15 (50%) households reported having 2 or more people share a room for sleeping, which has been found to be a risk factor for secondary influenza transmission.²⁵ A study done in a similar setting in Kenya found a SIR of 8%,¹⁸ similar to prior studies but less than what was found in our study; however, this study was unable to account for asymptomatic cases or provide laboratory-confirmation. The Kenya study also indicated that household contacts of HIV-infected index cases may be at higher risk of developing secondary ILI, which may be relevant in the communities in which our study was conducted due to the high HIV prevalence.²⁷ Our study also showed time spent in the home with the index case on Day 3 of the study was a risk factor for influenza infection. This indicates that increased exposure to the index case through contact is a risk factor, however our sample size was too small to demonstrate this consistently.

Our study has limitations that warrant discussion. The case-ascertainment study design has inherent bias as only patients coming to the clinic to seek care within the first 3 days of illness could be included.² These patients could have more severe or infectious illness that may have falsely elevated the SIR in this population.² Screening cases for inclusion with RIDT may have led us to fail to identify some potentially eligible index cases because of low sensitivity of these tests.²⁸ Also, although HIV infection status of household contacts has been suggested as a risk factor for influenza transmission, we were not able to examine this due to small sample size and missing data on HIV infection status for approximately half of the household contacts. Our analysis also assumed similar exposure for all household contacts, which may not have been the case. While data was collected on interactions and activities of household contacts with the index cases, small sample size and missing data did not allow us to perform an adjusted analysis to look for differences in exposure and risks. The assumption was also made that secondary cases with the same type of influenza as the index case were due to household transmission and not from the community, however this is supported by prior studies showing that risk of infection is much higher in the household.²⁹ As study teams only interacted with the families every 3 days, if a form was not filled out daily, there may have been recall bias when reporting symptoms and interactions with the index case. Strengths of our study include that our study is based on laboratory-confirmed findings in addition to symptom report, close follow-up of contacts, and testing of asymptomatic individuals.

This is the first household transmission study on seasonal influenza in South Africa. SIR for laboratory-confirmed influenza and ARI were higher than what has been reported in previous studies and studies from other countries,^2–4 but are consistent with the study done in South Africa on pandemic influenza.¹⁴ There is a large burden of respiratory symptoms in this community, yet household contacts did not report changing behaviors to prevent transmission, influenza vaccination rates were low, and risk factors of crowded sleeping conditions and spending time in the home with the index case were reported. Education on transmission-reducing behaviors could guide public health measures to improve influenza transmission control in households in similar settings. This study provides local data on influenza transmission that can be used to understand and predict influenza transmission through modeling in similar middle-income community settings.