The study was conducted in 2 regions with established demographic surveillance systems (DSSs): the Kilifi Health and Demographic Surveillance System (Kilifi HDSS) [11], which is located on the Kenyan coast, and the Kenya Medical Research Institute (KEMRI) and the Centers for Disease Control and Prevention (CDC) Health and Demographic Surveillance System (Siaya HDSS) [12] in rural western Kenya near Lake Victoria. Hospitalization, diarrhea, and rotavirus counts for children under 5 years of age were obtained from 2 government hospitals: Kilifi County Hospital located in the Kilifi HDSS and Siaya County Referral Hospital located in the Siaya HDSS.

Rotavirus vaccine was introduced into the childhood vaccination program in both Kilifi and Siaya counties in July 2014. Children who were aged 6 weeks or younger or born after the start of the rotavirus immunization program were eligible to receive 1 dose at 6 weeks and 1 dose at 10 weeks, although vaccine stock-outs in November and December 2014 kept some eligible children from receiving vaccine. There was no catch-up campaign in either Kilifi or Siaya.

In the Kilifi HDSS, vaccinations have been recorded comprehensively since 2008 through a network of approximately 30 health facilities (both private and public) (Supplementary Figure 1), as part of the Kilifi Vaccine-Monitoring Study (KiVMS) [13]. In the Siaya HDSS, over 20 health facilities administer vaccines to children (Supplementary Figure 2); since 2007, immunization data have been collected for children aged less than 5 years during household data collection rounds, which occur 2–3 times per year [12]. In addition, the vaccination status of children enrolled in the rotavirus surveillance at Siaya County Referral Hospital is recorded at the time of hospital admission. In both sites, card-confirmed vaccine information was captured by the field staff.

In 2009, to prepare for vaccine introduction [12, 14], rotavirus surveillance was intensified at Siaya County Referral Hospital among children aged less than 5 years who presented with 3 or more loose stools within a 24-hour period and 1 or more vomiting episodes. In Kilifi, rotavirus surveillance was initiated in late 2009 in inpatient children aged less than 13 years with 3 or more loose stools within a 24-hour period. Parents of eligible children were briefed on the potential risks and benefits of the study before being asked to give consent by the study team. Stool samples were collected from eligible children whose parents provided voluntary informed consent and tested for rotavirus antigen by enzyme-linked immunosorbent assay, as described previously [14, 15].

Surveillance was interrupted by 12 strikes, including 7 before and 5 after rotavirus vaccine introduction, by different cadres of healthcare workers as indicated in Supplementary Table 1.

We used a controlled interrupted time-series analysis in which rotavirus-negative diarrhea hospitalizations acted as the control series to compare monthly rotavirus-associated hospitalizations (RVHs) before and after rotavirus vaccine introduction in children younger than 5 years of age. Specifically, RVHs from the period 54 months prior to (January 2010 to June 2014) were compared with hospitalizations from the period 36 months after (July 2014 to June 2017) vaccine introduction at each facility. For a month in which less than 100% of diarrhea cases were tested for rotavirus, the number of rotavirus-positive cases was scaled by multiplying the total number of diarrhea cases by the proportion who were rotavirus positive among those tested in that month.

A negative binomial regression was used to estimate the change in rotavirus hospitalizations in the first year (July 2014–June 2015), second year (July 2015–June 2016), and third year (July 2016–June 2017) post–vaccine introduction. The model included rotavirus-negative diarrhea (the control series) as an offset and calendar month and health worker industrial actions as covariates. Lag-1 Newey-West standard errors were used to adjust for autocorrelation [16]. A similar analysis was conducted for all-cause diarrhea using non–diarrhea hospitalizations as the control series.

For the analysis of rotavirus-positive diarrhea, a sensitivity analysis was conducted using a synthetic control [17–19] in the regression model instead of rotavirus-negative cases. The synthetic control was generated by weighting 3 conditions (rotavirus-negative diarrhea, pneumonia, malaria)—using the SYNTH [20] command in Stata—so that their sum closely resembled the prevaccine rotavirus-positive time series.

All statistical analyses were conducted using Stata13.1 software (Stata Corporation). All datasets, Stata files, and related documentation are available on an online data repository (https://doi.org/10.7910/DVN/PH4COG) [21].

Ethical approval to conduct this study was granted by the KEMRI Scientific and Ethics Review Unit (SERU; SSC 3049) and the CDC's institutional review board (protocol 6968).

During the study period, there were 1513 and 1652 diarrhea hospitalizations in Kilifi County Hospital and Siaya County Referral Hospital, respectively. In Kilifi, 1142 (75.5%) of the diarrhea cases were tested, and among those tested, 315 (27.6%) were positive for rotavirus. In Siaya, 877 (53.1%) were tested and 197 (22.5%) were positive for rotavirus.

The proportion of children living in the study areas who received rotavirus vaccine increased with time. Between 2014 (year of introduction) and 2017, the proportion of those aged less than 1 year with at least 1 dose increased from 31% to 73% in Kilifi and from 16% to 75% in Siaya; the proportion who were fully vaccinated (2 doses) in 2017 was 65% in Kilifi and 62% in Siaya. The highest coverage among infants was observed in 2016. Among children aged 12–23 months, the proportion with at least 1 dose increased from 0% to 86% in Kilifi and from 0% to 94% in Siaya; coverage with 2 doses in 2017 was 84% and 92%, respectively (Figure 1). The lowest proportions of vaccinated children were observed in 2014 across both sites due to vaccine roll-out in midyear and the unexpected stock-out. Among vaccinated children, the median age at receipt of the first dose was 6 weeks (interquartile range [IQR], 6–7 weeks) in Kilifi and 6 weeks (IQR, 6–7 weeks) in Siaya; median age at receipt of the second dose was 10 weeks (IQR, 10–13 weeks) in Kilifi and 10 weeks (IQR, 10–12 weeks) in Siaya.

Across the 2 DSSs, overall diarrheal illness and rotavirus-specific illness displayed seasonal patterns, which were dampened after the introduction of the vaccine (Figure 2). Siaya generally recorded higher diarrhea and rotavirus-associated diarrhea incidences compared with Kilifi in the prevaccination era (Tables 1 and 2). In Kilifi, the highest numbers of rotavirus cases were recorded from April/May to September/November of the same year (Figure 2A). In Siaya, the RVH season began earlier in the year, usually in January /February (Figure 2B). At both sites, there was a declining trend in annual incidence of diarrhea and RVHs that began prior to the rotavirus immunization program and continued following the introduction of vaccine (2015–2017). During the years affected by industrial actions of health workers (Supplementary Table 1) lower rotavirus incidence rates were recorded at both sites (Table 2).

In Kilifi, we observed reduced rates of RVHs among Kilifi HDSS resident children younger than 5 years throughout the postvaccination period (Table 3 and Figure 3). When compared with rates from the prevaccine era, reductions were observed during the first year (57%; 95% confidence interval [CI], 8.0–80.0%), the second year (80%; 95% CI, 46–93%), and the third year (76%; 95% CI, 56–87%) (Supplementary Figure 3a). Similarly, in Siaya, there were reductions in the first year (59%; 95% CI, 20–79%), second year (82%; 95% CI, 61–92%), and the third year (81%; 95% CI, 7–96%) post–vaccine introduction (Supplementary Figure 3b).

When examining trends in all-cause diarrhea using non–diarrhea hospitalizations as the control series, significant reductions were observed at both sites in post–vaccine introduction years (Table 3), although by a smaller magnitude compared with reductions in RVHs. In Kilifi, the impact of the vaccine on all-cause diarrhea increased from 41% (95% CI, 16–59%) in the first year to 48% (95% CI, 40–55%) and 46% (95% CI, 34–55%) in the second and third year, respectively (Supplementary Figure 3e). In Siaya, there was a 41% (95% CI, 27–52%) reduction in all-cause diarrhea in the first year followed by 60% (95% CI, 43–72%) in the second year and 62% (95% CI, 38–77%) in the third year (Supplementary Figure 3f).

The associated weights for each component of the synthetic control are shown in Supplementary Table 2 and the results of vaccine impact analysis in Supplementary Table 3 and Supplementary Figure 3. Vaccine impact estimates from this analysis are similar to those presented in Table 3, except that the estimate for the last year in Kilifi was slightly lower than the corresponding estimate in Table 3. In Kilifi, the impact of the vaccine was 67% (95% CI, 27–85%) in the first year, 86% (95% CI, 64–94%) in the second year, and 69% (95% CI, 45–83%) in the third year. Similarly, in Siaya, vaccine impact increased from 68% (95% CI, 33–84%) in the first year to 89% (95% CI, 68–96%) in the second year and 82% (95% CI, 8–96%) in the third year.