Morbidity burden, seasonality and factors associated with the human respiratory syncytial virus, human parainfluenza virus, and human adenovirus infections in Kenya.

Background: Human respiratory syncytial viruses (HRSV), human parainfluenza viruses (HPIV), and human adenoviruses (HAdVs) cause a substantial morbidity burden globally. Objective: We sought to estimate morbidity burden, assess seasonality, and determine factors associated with these respiratory viruses in Kenya.
Methods: The data were obtained from Kenyan sites included in the Köppen-Geiger climate classification system. We defined the proportion of morbidity burden by descriptive analysis and visualized time-series data for January 2007-December 2013. Logistic regression was used to identify factors associated with infection outcomes.
Results: The morbidity burden for HRSV was 3.1%, HPIV 5.3% and HAdVs 3.3%. Infants were more likely to be infected than other age groups. HRSV exhibited seasonality with high occurrence in January-March (odds ratio[OR] = 2.73) and April-June (OR = 3.01). Hot land surface temperature (≥40 °C) was associated with HRSV infections (OR = 2.75), as was warmer air temperature (19-22.9 °C) (OR = 1.68), compared with land surface temperature (<30) and cooler air temperature (<19 °C) respectively. Moderate rainfall (150-200 mm) areas had greater odds of HRSV infection (OR = 1.32) than low rainfall (<150 mm).
Conclusion: HRSV, HPIV and HAdVs contributed to morbidity burden, and infants were significantly affected. HRSV had a clear seasonal pattern and were associated with climate parameters, unlike HPIV and HAdVs.

Background

The morbidity burden of influenza-like illness (ILI) is attributed to several viral agents (World Health Organization et al., 2015). The prevalence of ILI varies depending on the type of respiratory virus, population demographics, geography, season and other factors (Zhang et al., 2015). The burden of influenza as a cause of ILI in Kenya has been well studied (Umuhoza et al., 2020), unlike other respiratory viruses. One important cause of ILI is the human respiratory syncytial virus (HRSV), which causes a substantial burden of acute respiratory diseases in children (<5 years) and the elderly (>65 years) (Falsey, 2005). HRSV is widely distributed across the world. In a systematic review in 2015, China reported an 18.7% HRSV prevalence (Zhang et al., 2015). Many other countries have also documented the occurrence of HRSV infections (Salimi et al., 2016). In Africa, a systematic review and meta-analysis of 2017 revealed that HRSV had a prevalence of 14.6% (Kenmoe et al., 2018).

HRSV has 2 main types, A and B, which occur either independently or co-circulate with a slight predominance of type A. However, clinically, infection with either of the 2 types does not present differently (Hirsh et al., 2014). Other respiratory viruses that cause ILI include human parainfluenza viruses (HPIV) and human adenoviruses (HAdVs) (Lim et al., 2017). HPIV has 4 major types, namely HPIV-1, HPIV-2, HPIV-3 and HPIV-4 (Henrickson, 2003), with a predominant occurrence of types HPIV-3, HPIV-1 and HPIV-2. Although the HPIV-4 type is less common, there is evidence that it can cause both the mild and severe respiratory illnesses described for the HPIV 1-3 types (Vachon et al., 2006; Villaran et al., 2014). HAdVs constitute 7 known species that vary from HAdV-A–G and cause a range of syndromes, including pneumonia.

Amongst the HAdVs, Species B (serotypes 3 and 7), C (serotypes 1, 2, and 5) and E (serotype 4) viruses cause ILI (Schmttz et al., 1983). HAdV-B3 occurs most frequently amongst HAdVs serotypes.

Understanding the seasonality of acute viral respiratory tract illnesses is important for timely preventive and treatment interventions (Janet et al., 2018). In temperate climates, most acute viral respiratory illnesses show distinct seasonality, with peaks during the winter months (Paynter, 2015). However, this seasonal variation is less evident in the tropics due to diminished fluctuations in meteorological parameters (Li et al., 2019). HRSV seasonality varies significantly within geographical regions and across time (Mullins et al., 2003). HRSV infections are mostly reported during winter in temperate regions and are also associated with the rainy season in tropical countries (van der Sande et al., 2004). Nevertheless, epidemics of HRSV are also observed during the dry season in areas located south of the equator (Shek and Lee, 2003). The epidemic duration of HRSV is approximately 5 months in both temperate and tropical regions (Sundell et al., 2016).

The epidemics of HPIV last for slightly longer (6 months) and occur in the spring or early summer months (Li et al., 2019). Differences in seasonality are observed for HPIV types in the temperate zones. HPIV-3 epidemic peaks occur in spring (Fry et al., 2006), whereas HPIV-1 and HPIV-2 circulation occur periodically in early winter in the temperate regions (Murphy et al., 1980). The reported prevalence of HPIV-4 is low; hence the seasonality of HPIV-4 is not well defined. In contrast, HPIVs occur throughout the year in the tropics. It has been described in autumn mainly for the HPIV-1 subtype; HPIV-3 and HPIV-2 peaked in spring.

The seasonality of the HAdVs is different; the infections occur throughout the year, with peaks observed in late winter, spring or early summer in temperate climates (Chen et al., 2016). Similarly, in tropical climates, HAdVs infections were detected throughout the year with peaks of different amplitudes, without a clear seasonality (Faden et al., 2011).

Factors influencing the occurrence of HRSV, HPIV and HAdVs are similar. The 3 viruses are transmitted directly through human contact or indirectly through fomites (Kutter et al., 2018). The common demographic factors associated with the 3 viruses have been reported in various studies and include age, sex and preexisting health conditions (Simoes, 2003). Additionally, host attributes and behaviors, comprising household overcrowding, daycare attendance, birth during the seasonal peak of infection, lower parental education level, inadequate hygiene, and lower breastfeeding rates, have been suggested to affect the distribution of these viruses (Sommer et al., 2011). Furthermore, environmental factors such as humidity, temperature, precipitation and airflows have been suggested to influence occurrence substantially (Pica and Bouvier, 2012). This study sought to estimate morbidity burden, assess seasonality, and determine factors associated with HRSV, HPIV and HAdVs infections in Kenya between 2007 and 2013.

Methodology

Study regions

Kenya is the 47th largest country in the world with a population of 48 million and a landmass of 580 367 km² characterized by variable geographical features and diverse climate (Kenya Population 2019). The climate varies from warm to cool across the different geographic regions of the country. Figure 1 shows the current Kenyan Köppen-Geiger climate classification map at a 1-km resolution (Beck et al., 2018), combined with the Kenya regional boundaries and ILI surveillance sites generated by qGIS 3.8.1-Zanzibar. The western region of the country has an equatorial tropical climate and some temperate savanna areas. This region experiences rainfall throughout the year, with the heaviest rain in April and a temperature range of 14-36 °C. The Rift Valley region has different climates ranging from the hot desert or arid region in the north to the tropical savanna and cooler temperate areas in the south. The average monthly rainfall ranges from 20 mm in July to 200 mm in April, and temperate varies from 20 °C to 40 °C. In the northeast part of the country, the climate is characterized by the warmest desert and arid areas that become cooler toward the center, followed by the oscillation of arid and tropical savanna climes in the coastal region. The average temperature range on the coast is 22-30 °C, with annual rainfall varying between 20 mm to 300 mm (Ayugi et al., 2016). ILI surveillance sites presented in this study are geographically distributed in each regional climate, except in the warm desert, characterized by low population density, which was not accessible due to security concerns. Eight surveillance sites were used to represent the target population in these different climatic zones.

Figure 1

Influenza and other respiratory viruses surveillance sites distribution map, Kenya

Source of the data

In this study, the primary dataset was participants presenting with ILI, recruited by the Kenya Medical Research Institute's (KEMRI) influenza and other respiratory virus surveillance program from 2007 to 2013. Participants were enrolled based on the World Health Organization case definition for ILI (World Health Organization 2015). Briefly, any individual presenting at the selected outpatient departments with (1) fever >38° C (oral or equivalent), (2) cough/sore throat, and (3) onset of illness within the previous 10 days was eligible for participation. The nasopharyngeal specimens and demographic data from consenting participants were submitted to the surveillance program that had received prior approval from the KEMRI Science and Ethics Research Unit under protocol numbers SSC#981.

The surveillance program assigned each hospital site a health care provider who collected patient specimens and demographic data on a weekly basis. The residence of each patient, including the village and the estate, was recorded along with other demographic information (age, gender and occupation). The collected nasopharyngeal specimens were tested for influenza and other respiratory viruses, including HRSV, HPIV and HAdVs, using a series of assays. These assays included polymerase chain reaction, viral culture in the appropriate cell lines, and immunofluorescence using appropriate virus-specific antisera. Data management and retrieval of the ILI dataset was performed using a program-specific Microsoft Access database.

Supplementary datasets comprised spatial coordinates of each surveillance site, monthly mean land temperature, air temperature, rainfall and regional Köppen-Geiger climate classification (1980-2016) (Beck et al., 2018). The spatial coordinates of each surveillance site were sourced via Earth Pro (7.3 Google LLC). Monthly mean land temperature data was derived from measurements of the Earth Observing System Moderate Resolution Imaging Spectroradiometer instrument aboard the Terra (EOS AM-1) spacecraft (Wan et al., 2002). The rainfall data were obtained from the African Rainfall Climatology dataset with data derived from satellite cold clouds.(Novella and Thiaw, 2013). Air temperature data were produced from several sources, including but not limited to the Global Historical Climatology Network Monthly (Willmott and Matsuura, 2001).

Data analysis

The ILI dataset was reviewed to ascertain participants’ characteristics and error checking completed. We performed a descriptive analysis to determine the morbidity burden of infection caused by HRSV, HPIV and HAdVs. In this analysis, the morbidity burden was defined and expressed as a proportion of laboratory-confirmed positive cases for HRSV, HPIV and HAdVs per the total number of ILI cases recorded by the surveillance program (World Health Organization et al., 2015). In addition, we obtained estimates for each specific virus by participants’ demographic characteristics and by the surveillance sites over the study period.

The seasonal pattern of HRSV, HPIV and HAdVs infections were recorded as the monthly total number of cases for the study period of January 2007 to December 2013. To increase the signal to noise ratio, we visualized by line plot the total number of each specific respiratory virus per annual quartile. We applied Fourier transformation to the time series data to describe monthly trends and showed if any steady seasonal patterns of HRSV, HPIV and HAdVs were present. Initially, a fast Fourier transformation was performed to identify the magnitude and phases of each 12-month cycle component. Also, inverse discrete Fourier transformation was computed to describe seasonal patterns in the data. To increase the frequency of resolution, we padded the de-trended data with zeros (Bloomfield, 2004). This analysis was performed in Microsoft Excel with NumXL software add-in, while all other analyses were completed in STATA®13 (STATA Corporation, College Station, TX, USA).

We performed a comparative analysis for HRSV, HPIV and HAdVs-specific outcomes and measured the differences in demographic variable categories with 95% CIs and chi-square test. The logistic regression was applied to measure the association of demographic (age, gender, occupation, origin, sick contact, school attendance, location and year), climate (temperature, rainfall and humidity), seasons (quarter-1, quarter-2, quarter-3 and quarter-4), clinical (online supplemental list) variables and the outcome of interest (presence or absence of HRSV, HPIV and HAdVs) by bivariate and multivariate models. We then reported the probability value (P-value) and odds ratios (OR) for each specific virus outcome. In the multivariate model, a P-value of <0.05 was considered statistically significant after adjusting for predictor variables to account for confounding factors. We assessed the two-way interaction by adding a hashtag symbol (#) in the model where necessary. Finally, the best-fitted model was selected by the lower Bayesian Information Criterion (Hosmer and Lemeshow, 2000).

Results

The ILI surveillance program recruited 17 261 participants from 8 surveillance sites from January 2007 to December 2013. The data in this study were limited to unsubtyped HRSV and HAdV, and the 3 subtypes of HPIV 1-3 (HPIV-4 was not tested for). As shown in Table 1, the majority of study participants came from Kisii (17%), New Nyanza (17%) and Mbagathi (16%). A lower number of study participants came from Port Reitz (12%), Kericho (11%) and Alupe (11%). The other 2 surveillance sites of Malindi and Isiolo each registered a lower number (10%) of study participants.

Table 1

Demographic characteristics of study participants by respiratory viruses.

Variable/ Outcome		HRSV			HPIV			HAdVs
Overall	Total Population	Positive	Negative	Chi-square	Positive	Negative	Chi-square	Positive	Negative	Chi-square
	N (%)	n (%)	n (%)	P-value	n (%)	n (%)	P-value	n (%)	n (%)	P-value
	17261	539 (3)	16722 (97)		922(5)	16339(95)		581(3)	16680(97)
Gender				0.429			0.471			0.106
Male	8998(52)	290(3)	8708(97)		470(5)	8528(95)		322(4)	8676(96)
Female	8263(48)	249(3)	8014(97)		452(5)	7811(94)		259(3)	8004(97)
Age				<0.001			<0.001			<0.001
≤1year	9650(56)	351(4)	9299(96)		557(6)	9093(94)		391(4)	9259(96)
2 to 4 years	6139(35)	166(3)	5973(97)		327(5)	5812(95)		175(3)	5964(97)
5 to ≤ 18years	1163(7)	20(2)	1143(98)		32(3)	1131(97)		14(1)	1149(99)
19-49 years	300(2)	2(0.6)	298(99)		6(2)	294(98)		1(0.3)	299(99)
50+ years	9(0.05)	0(0)	9(100)		0(0)	9(100)		0(0)	9(100)
Occupation				0.023			0.001			0.012
Children	16482(95)	528(3)	15954(97)		905(5)	15577(95)		567(3)	15915(97)
Students	420(2)	9(2)	411(98)		11(3)	409(97)		13(3)	407(97)
Others	355(2)	2(0.5)	353(99)		6(2)	349(98)		1(0.2)	354(99)
Origin				0.335			0.458			0.241
Urban	16932(98)	532(3)	16400(7)		902(5)	16030(95)		574(3)	16358(97)
Rural	319(2)	7(2)	312(98)		20(6)	299(94)		7(2)	312(98)
Sick contact				0.213			0.362			0.368
Yes	6001(35)	174(3)	5827(97)		308(5)	5693(95)		192(3)	5809(97)
No	11245(65)	365(3)	10880(97)		614(5)	10631(95)		389(3)	10856(97)
attend school				<0.001			<0.001			<0.001
Yes	3191(18)	60(2)	3131(98)		123(4)	3068(96)		60(2)	3131(98)
No	14052(81)	479(3)	13573(97)		798(6)	13254(94)		521(4)	13531(96)
Location				<0.001			0.002			<0.001
Alupe	1884(11)	30(2)	1854(98)		75(4)	1809(96)		30(2)	1854(98)
Isiolo	1347(8)	44(3)	1303(97)		79(6)	1268(94)		33(2)	1314(98)
Kericho	1996(11)	64(3)	1932(97)		127(6)	1869(94)		65(3)	1931(97)
Kisii	2950(17)	128(4)	2822(96)		177(6)	2773(94)		94(3)	2856(97)
Malindi	1327(7)	62(5)	1265(95)		80(6)	1247(94)		68(5)	1259(95)
Mbagathi	2810(16)	46(2)	2764(98)		125(4)	2685(96)		117(4)	2693(96)
New Nyanza	2938(17)	79(3)	2859(97)		142(5)	2796(95)		108(4)	2830(96)
Port Reitz	2009(12)	86(4)	1923(96)		117(6)	1892(94)		66(3)	1943(97)
Year				<0.001			<0.001			<0.001
2007	2925(17)	22(1)	2903(99)		82(3)	2843(97)		133(5)	2792(95)
2008	3052(18)	103(4)	2949(96)		185(6)	2867(94)		101(3)	2951(97)
2009	3806(22)	86(2)	3720(98)		123(3)	3683(97)		131(3)	3675(97)
2010	3027(17)	116(4)	2911(96)		188(6)	2839(94)		102(3)	2925(97)
2011	2289(13)	140(6)	2149(94)		138(8)	2151(94)		44(2)	2245(98)
2012	1338(8)	53(4)	1285(96)		110(8)	1228(92)		52(4)	1286(96)
2013	824(5)	19(2)	805(98)		96(12)	728(88)		18(2)	806(98)

The overall morbidity burden of ILI caused by HRSV was 3.1% (2.8%–3.3%; 95% CI). HPIV and HAdVs represented 5.3% (5.0%–5.6%; 95% CI) and 3.3% (3.1%–3.6%, 95% CI), respectively. Amongst the HPIV types, HPIV-3 was the most dominant (38.6%), followed by HPIV-1 (34.1%) and HPIV-2 (10.4%). The proportion of multiple HPIVs coinfections was 16.8%.

The proportions of ILI caused by HRSV, HPIV and HAdVs varied by demographics. HRSV infection differed significantly across the age categories (P<0.001), as did HPIV and HAdVs. Further, differences were observed in the proportions of HRSV (P = 0.023), HPIV (P = 0.001), and HAdVs (P = 0.012) among occupation categories. There was a significant difference in the proportion of participants who attended school (P<0.001) for HRSV, HPIV and HAdVs infections compared with non-school attending participants. The proportion of disease caused by the 3 respiratory viruses varied significantly across surveillance sites and throughout the years of the surveillance period. HRSV, HPIV and HAdVs occurred at all surveillance sites with varying distribution of infections. HRSV infections were most commonly found in Malindi (5%), Port-Reitz (4%) and Kisii (4%). Other surveillance sites had lower proportions of HRSV infection. The proportions of HPIV infections were the highest in Malindi, Port-Reitz, Kisii, Kericho and Isiolo. The proportion of HAdVs infections were highest in Malindi (5%), New-Nyanza (4%) and Mbagathi (4%), compared with Port-Reitz, Kisii and Kericho.

The ILI participants presented with various clinical symptoms (S1 figure 1). Fever (100%), cough (98%), runny nose (87%) and nasal stuffiness (53%) were the most common symptoms. Many clinical characteristics, including malaise, vomiting, fatigue, difficulty in breathing, diarrhea, headache, sore throat, retro-orbital pain, sputum production, abdominal pain, joint pain, muscle aches, bleeding, and neurological signs, were reported in <35% of participants. These symptoms varied by HRSV, HPIV and HAdVs infection type. After adjusting for age, none of the clinical characteristics, except for fever and cough, were significantly associated with a specific respiratory virus outcome among the participants.

HRSV, HPIV and HAdVs circulated throughout the 7 years of surveillance (2007-2013), roughly following a quarterly pattern (S2 Figure 2). There were 3 major spikes of HRSV activity in the years 2008, 2010 and 2011. HRSV had a defined seasonal peak appearing around April–May every year of the surveillance period (Figure 2). Fourier series analysis revealed no clear pattern in the seasonal trends for HPIV and HAdVs cases. HPIV was the most prevalent respiratory virus, with spikes occurring irregularly every year during the surveillance period. In contrast, major erratic spikes were observed for HAdVs in 2007, followed by 2009; the number of cases then decreased progressively toward 2013.

Figure 2

The monthly trend of HRSV, HPIV and HAdVs infections during the study period

*Monthly trend of Influenza-like illness (ILI) by HRSV, HPIV and HAdVs, Kenya (2007-2013). Human respiratory syncytial virus (HRSV), human parainfluenza virus (HPIV), and human adenoviruses (HAdVs). The inverse discrete Fourier transform (IDFT) represents the monthly periodicity trend for every 12 months cycle. Blue dots denote the observed seasonal peak of HRSV around April–May.

Demographic factors, season, and climate had significant associations with the occurrence of HRSV, HPIV and HAdVs (Table 2). Infants had higher proportions of all 3 viruses than other age groups. In addition, HRSV was 2.73 times more likely to occur in Jan–Mar (Q1) and 3.01 times in Apr-Jun (Q2) than in Oct-Dec (Q4). Hot land surface temperature (≥40 °C) favored HRSV infections more than cooler land surface temperature (<30 °C) (OR: 2.75). HRSV infections also had higher odds (OR: 1.68) of occurring in warmer (19-22.9 °C) than cooler air temperatures (<19 °C), and during moderate (150-200 mm) compared with low rainfall (<150 mm) (OR = 1.32). There were no associations between season and climatic conditions for either HPIV or HAdVs infections. Higher land surface temperature (≥40 °C) as opposed to cooler land surface temperature (<30 °C) was associated with HAdVs infections (OR = 2.25).

Table 2

Factors associated with the respiratory syncytial virus, human parainfluenza, and adenoviruses.

Variable/Outcome	HRSV				HPIV				HAdVs
	Crude OR*	P-Value	Adjusted OR*	P-Value	Crude OR*	P-Value	Adjusted OR*	P-Value	Crude OR*	P-Value	Adjusted OR*	P-Value
	(95% CI)		(95%CI)		(95% CI)		(95%CI)		(95% CI)		(95%CI)
Age		<0.001				<0.001				<0.001
≤1year	Ref		Ref		Ref		Ref		Ref		Ref
2 to 4 year	0.73(0.61-.088)	0.001	0.75(0.62-0.91)	0.003	0.91(0.79- 1.05)	0.235	0.90(0.78-1.03)	0.153	0.69(0.57-0.83)	<0.001	0.69(0.58-0.83)	<0.001
5 to ≤ 18year	0.46(0.29-0.73)	0.001	0.50(0.31-0.79)	0.003	0.46(0.32-0.66)	<0.001	0.45(0.31-0.66)	<0.001	0.28(0.16-0.49)	<0.001	0.29(0.17-0.49)	<0.001
19-49 year	0.17(0.04-0.71)	0.015	0.19(0.04-0.77)	0.02	0.33(0.14-0.75)	0.008	0.33(0.14-0.76)	0.009	0.07(0.01-0.56)	0.011	0.08(0.01-0.59)	0.013
50+ year	NA
Quartile		<0.001				<0.001				0.0082
Oct-Dec(Q4)	Ref		Ref		Ref		Ref		Ref		Ref
Jan-March(Q1)	3.10(2.32-4.14)		2.73(2.00-3.73)	<0.001	0.77(0.63-0.93)		0.83(0.68-1.02)	0.091	0.69(0.54-0.88)		0.62(0.47-0.80)	<0.001
Apr-Jun(Q2)	2.98(2.23-3.97)		3.01(2.23-4.07)	<0.001	1.07(0.90-1.28)		1.07(0.90-1.27)	0.432	0.74(0.58-0.93)		0.81(0.63-1.04)	0.101
Jul-Sept(Q3)	0.94(0.66-1.35)		1.05(0.73-1.52)	0.768	0.69(0.56-0.84)		0.71(0.58-0.87)	0.001	0.90(0.71- 1.13)		0.87(0.69-1.11)	0.28
LTM LST*		<0.001				0.0048				0.0032
Cooler(<30°C)	Ref		Ref		Ref		Ref		Ref		Ref
Warmer(30-39.9°C)	1.05(0.87-1.25)		0.93(0.75-1.15)	0.529	0.85(0.74-0.98)		0.90(0.78-1.05)	0.199	0.96(0.80-1.14)		1.02(0.85-1.23)	0.775
Hot(≥40°C)	3.42(2.44- 4.78)		2.75(1.79-4.23)	<0.001	0.51(0.29-0.87)		0.56(0.32-0.97)	0.041	2.00(1.36- 2.95)		2.25(1.48-3.43)	<0.001
MM AT*		0.0001				0.5487				0.7885
Cooler(<19°C)	Ref		Ref		Ref				Ref
Warmer(19-22.9°C)	2.23(1.38-3.60)		1.68(1.03-2.76)	0.037	1.15(0.88-1.51)				1.06(0.76-1.47)
Hot(≥23°C)	2.59(1.58-4.23)		1.61(0.95-2.73)	0.072	1.11(0.83-1.49)				0.99(0.70- 1.41)
LTM rainfall*		<0.001				0.3126				0.0337
Low (<150mm)	Ref		Ref		Ref				Ref		Ref
Moderate(150-200mm)	1.62(1.30-2.01)		1.32(1.05-1.66)	0.016	1.08(0.89-1.30)				0.86(0.67-1.11)		0.93(0.72-1.21)	0.627
Heavy(>200mm)	1.67(1.25- 2.22)		1.04(0.75-1.44)	0.781	1.19(0.93-1.52)				0.63(0.43-0.93)		0.66(0.43- 1.00)	0.055

* LTM LST: Long term mean land surface temperature; * MMAT: monthly mean air temperature; * LTM rainfall: long term mean rainfall; *Q1: quarter one; *Q2: quarter two; *Q3: quarter three; *Q4: quarter four.

Discussion

In this study, we identified that HRSV, HPIV and HAdVs contributed substantially to the morbidity burden of ILI at all hospitals in the influenza and other respiratory viruses surveillance program network across Kenya. The record of multiple infections of HPIV subtypes suggested a co-circulation and/or co-morbidity with other respiratory viruses. However, co-infections of the 3 viruses were outside the scope of this study. HRSV exhibited seasonality, occurring more frequently in January through June, whereas HPIV and HAdVs exhibited no seasonality. Hot land surface temperature, warmer air temperature, and increased monthly rainfall were also associated with HRSV infections; no associations were seen between temperature or rainfall with either HPIV or HAdVs.

In a recent systematic review for the East African region, we found that the overall proportions of HRSV, HPIV and HAdVs reported among ILI cases were 10%, 9% and 14%, respectively (publication in press). Thus, the proportions reported in the current study were much lower than those in the systematic review. These differences may result from various factors, including the study design, sampling technique, type of specimen analyzed, populations studied, or severity of illness. The proportion of infections for HRSV, HPIV and HAdVs in infants was higher than other age groups at 4%, 6% and 4%, respectively. However, this was not surprising because it is known that infants are naïve to those infections and easily acquire these respiratory viruses within households (Munywoki et al., 2014), leading to several episodes of repeat infections.

The proportion of HRSV, HPIV and HAdVs infections was greater in sites in the coastal tropical savanna (Malindi and Port-Reitz) and the western tropical forest (Kisii and Kericho) climatic regions. Indeed, climatic factors are an important driver for viral respiratory infection dynamics (Weber et al., 2001). The increasing number of HRSV cases from January to June with a peak around April–May coincided with the rainy season in Kenya, and moderate rainfall (150-200 mm) showed a positive effect on the occurrence of HRSV. This observation concurs with various studies in the tropics that recognized similar trends and further associated viral respiratory infections with rainfall (Murray et al., 2012).

Other climate parameters that were suitable for HRSV circulation included warm air temperature (19-22.9 °C) and hot land surface temperature (≥40 °C). Temperature is a known metrological predictor of respiratory syncytial virus infections; several studies have reported an association of monthly average temperature with respiratory syncytial virus Infections in the tropics (Rodriguez-Martinez et al., 2015). In contrast, HPIV and HAdVs exhibited no seasonality for either respiratory virus. This finding supports previous studies, which have also shown clear seasonal patterns for HRSV, but not for HPIV and HAdVs (Li et al., 2019).

This study had several limitations. Whereas influenza as a major cause of ILI burden was not included here because it has been reported previously (Umuhoza et al., 2020), the morbidity burden of other viruses causing ILI was not examined. Further, possible co-infections between HRSV, HPIV or HAdVs and other viruses were not examined. In addition, this study likely overestimated the prevalence of the 3 respiratory viruses in the underlying population at risk since all participants had symptoms of ILI at the time of enrollment. As 91% of study participants were aged <5 years, the associations seen in this study may have had insufficient power to assess associations in other age groups. Finally, other unmeasured environmental or temporal confounding factors may have influenced the associations seen in this study. Notwithstanding these weaknesses, this study has considerable strengths, including the large sample size, long study period, broad geographic region covering the entire country, reliable data source from a robust surveillance network, and application of a pre-defined protocol and stratified analysis.

Conclusion

Our findings indicate that HRSV, HPIV and HAdVs cause a substantial proportion of ILI morbidity burden in Kenya. The age category of infants had a higher proportion of these 3 respiratory viruses. HRSV has a steady seasonal pattern, with cases increasing early in the year and peaking around April–May coincidental with the rainy season. Hot land surface temperature, warmer air temperature, and moderate rainfall were associated with increased HRSV infection. In contrast, HPIV and HAdVs were not seasonal and were not associated with climate parameters. mmc1.docx

Conflicts of interest

The authors declare that they have no financial or personal relationships which may have an inappropriate influence on conducting this study.