Disease transmission and mass gatherings: a case study on meningococcal infection during Hajj

Background Mass gatherings can not only trigger major outbreaks on-site but also facilitate global spread of infectious pathogens. Hajj is one of the largest mass gathering events worldwide where over two million pilgrims from all over the world gather annually creating intense congestion. Methods We developed a meta-population model to represent the transmission dynamics of Neisseria meningitidis and the impact of Hajj pilgrimage on the risk of invasive meningococcal disease (IMD) for pilgrims population, local population at the Hajj site and country of origin of Hajj pilgrims. This model was calibrated using data on IMD over 17 years (1995–2011) and further used to simulate potential changes in vaccine policy and endemic conditions. Results The effect of increased density of contacts during Hajj was estimated to generate a 78-fold increase in disease transmission that impacts not only pilgrims but also the local population. Quadrivalent ACWY vaccination was found to be very effective in reducing the risk of outbreak during Hajj. Hajj has more limited impact on IMD transmission and exportation in the pilgrim countries of origin, although not negligible given the size of the population considered. Conclusion The analysis performed highlighted the amplifying effect of mass gathering on N. meningitidis transmission and confirm vaccination as a very effective preventive measure to mitigate outbreak risks. Supplementary Information The online version contains supplementary material available at 10.1186/s12879-022-07234-4.

where D X c,i and I X c,i correspond respectively to the demographic process and infection process for compartment X i.e. transition rates between compartments at a given time t due to birth, death or ageing (demographic) or change in disease status (infection).
In the absence of vaccination, the demographic process is identical whatever the disease status and is given for every age group except the first one by Θ i is the transition rate from age group i to age group i+1, defined as follows: where g stands for the growth rate of the population, µ the death rate and l i for the length of age class i. In the first age group, except for the compartment corresponding to disease status at birth, the demographic process simplifies to D X c,1 = −(µ i + Θ i )X c,i . (S1.4) We consider that, in the absence of vaccination, all newborns are susceptible (corresponding to compartment S c,1 ). Therefore where B c is the number of births at date t defined as a fraction of the population size: where P c,i is the population size in age group i and f c the birth rate in cluster c. At demographic equilibrium, the following equations hold (S1.7)

S1.2. Infection process in the absence of vaccination
In the absence of vaccination, the population in a given age group and cluster can be divided in 5 compartments : Susceptible(S), Short-term carrier(C), Asymptomatic carrier(A), Diseased(D), and Recovered(R).The corresponding set of equations is given by : (S1.8) The force of infection in the absence of vaccination and if transmission only occurs locally in cluster c is given by where P c,j = S c,j + C c,j + A c,j + D c,j + R c,j is the total population in age group j.
By normalizing the force of infection with the size of the population, we considered here a density-dependent transmission [2]. Infectivity is considered to be identical for short-term and asymptomatic carriage but lower for diseased individuals ( < 1). This formulation of the force of infection enables considering age-specific transmission rate (corresponding to the matrix of β ij ). However, for sake of simplicity, we assume throughout the analysis presented here homogeneous mixing (∀i, j β ij = β ji = β c ).

S1.3. Model with vaccination
If one excepts Hajj-related vaccination, vaccination occurs when people move from one age group to another. This impact the ageing process, since in case of vaccination, the compartment before vaccination (X c,i−1 ) differs from the compartment after vaccination (X c,i ).
Vaccination is also accounted for in the model through the addition of 6 vaccinerelated compartments: Vaccinated and protected against infection (V), Vaccinated but susceptible to meningococcal infection (SV), Vaccinated and short-term carrier (CV), Vaccinated and asymptomatic carrier(AV), Vaccinated and Diseased (DV), and Vaccinated and immune following infection (RV). The infection process is given by (S1.10) The force of infection becomes: (S1.11)

S1.4. Hajj-related transmission
Hajj-related transmission is first represented by dividing the population of each cluster at the start of each Hajj period between pilgrims and non-pilgrims. The proportion of pilgrims is both cluster and age-dependent. For instance, the proportion of individuals in disease status X in cluster c and age group i transferred at the start of a given Hajj period (t = t hs ) to the pilgrim-specific compartments associated to cluster c is given by: We also account for Hajj vaccination that occurs in the model at the time pilgrims are transferred to the pilgrim-specific states. The vaccination process is otherwise similar to the one described previously. At the end of a given Hajj period (t = t he ), subjects in pilgrim-specific compartments are reintegrated in the non-pilgrim equivalent compartment: The demographic process is unchanged for pilgrim-specific compartments. Pilgrims can change age group or die during the Hajj. We however restricted births to non-pilgrims compartments.
The infection process for pilgrims is also identical to the one defined in S1.10. The only difference relates to the increased contact rates between pilgrims (Hajj density effect) that led to the specific force of infection for pilgrims during Hajj (t hs < t < t he ): (S1.14) β H corresponds here to the specific Hajj density effect defined as a multiplicative factor. In addition, we also consider year-to-year variation in the level of meningitis transmission which corresponds to β y in (S1.14) We also account for an increased risk of carriage acquisition for the local population of the Hajj site linked to contacts with pilgrims (β L ). If we consider that the first cluster considered in the model corresponds to the population of the Hajj site (c=1). The force of infection during Hajj for the local population is given by The model was coded in C# and the results were generated and analyzed through a web-based simulation platform called DENMOD.

S2.1. Demography
Population is derived from WHO 2017 population assessment [3] for clusters outside KSA (high, medium and low transmission). We estimated the average annual growth rate per cluster, based on 2000-2019 demographics evolution per country [4]. Demographic data for Mecca and KSA area are based on official KSA statistics [5].
For Mecca and KSA outside Mecca, the whole populations are modeled. For the three other clusters, we only model one tenth of their real population. As the number of Hajj pilgrims compared to the size of these three clusters' population is very small, this choice does not impact the dynamic of the meningitis's spread due to the Hajj and the relative proportions between the three clusters are respected.
Age groups are the same for all clusters. There are twelve age groups: an age group from birth to one year old, an age group from 1 to 5 years old, then 5 age groups of 5 years until 30 years old, then 3 age groups of 10 years until 60 years old and finally 2 age groups of 20 years covering people until 100 years old. Ageing of the population is based on life tables provided by WHO [6]. They are specific to a country, so the larges country in the cluster was selected to represent the ageing of the population, cf Table S3 i.e. the one with largest impact on the representativeness of the cluster. Source :KSA General Authority for Statistics [7]. Asian Non-Arab corresponds to Asian countries except Turkey, Malaysia, the Philippines, Indonesia, Russia, and China

S2.2. Pilgrimage
Pilgrimage is modeled as a twenty days event, occurring every 354 days. 20 days cover both the period of the religious events -5 days -and the progressive arrival and departure of pilgrims. Real start dates of the pilgrimage were taken as initial dates of the 20-days period for calibration and validation years (1995-2012) and are automatically computed in simulations. As the time between two Hajj is shorter than a Gregorian year, it occurs that 2 Hajj happen the same Gregorian calendar year, like in 2006. For easiness of reading, all results are based on Gregorian calendar years.
During pilgrimage, the part of the population of each cluster that represent the pilgrims is modeled as a dedicated cluster. The transition mechanism of pilgrims from local population to their dedicated cluster, mimicking the travel from their residency country to the Hajj site, is described in equations (S1.12) and (S1.13). The annual number of pilgrims is taken from KSA Hajj statistics [7] and is used to calculate the proportion (h c,i ) of pilgrims. Obviously, the fractions of the pilgrims in the clusters are small as shown in Table S3, especially in non-KSA ones. Usually, two to three millions of pilgrims attend the Hajj each year. The typical number of pilgrims and their country of origin can be seen in Table S4. Age distribution of the pilgrims differs from the age distribution of non-pilgrims but remains stable over time [8] and is described in Table S4. Source : Annual number of pilgrims [7], Age distribution [8,9] S2.3. Disease When infected, an individual may progress rapidly to the disease state with symptoms within a week or may carry the infection in an asymptomatic state for several months [10]. From the 2000-2001 Hajj outbreak we noticed a very short period of carriage [11,12]. The IMD cases were reported in a week or two following the Hajj. From this study we consider the short term carriage to be a week (i.e. q = 1). We estimated the long-term (asymptomatic) carriage to 1 year, taking an average value from the 9 to 15 months period evoked in Bennet et al. review [13], consistent with 11.1 up to 14.5 months of carriage estimated by De Wals et al. in [14]. The infectious period of IMD cases is about a week which yields α D = 52 [10].
The waning rate of recovery-induced-immunity was set to 12 years [15], which yields δ = 0.0839/year. No reliable data are found for the relative transmission scaling ( ). We assume = 0.5. All these parameters are summarized in Table S5.

S2.4. Vaccine
In 2002, upon available, the polysaccharide quadrivalent ACWY vaccines were made mandatory for all the pilgrims and the residents of Mecca and Medina who are older than two years [21]. KSA also introduced routine vaccines for the toddlers at the age of 9 months [18], still accurate. We model that as routine vaccine for the Mecca and KSA clusters at the age of one year for all simulated years. One is a proxy for 9 months due to age structure of the model and the fact it is the moment where vaccination scheme ends and is efficient. Coverage rate was set to 96%, estimated from analysis of other diseases coverage rate, always very high in KSA [19,20]. Pilgrims from all the five clusters are considered vaccinated according to the Hajj guidelines and regulation [22]. Since there is no consistent and routine vaccines throughout the non-KSA clusters, and that all pilgrims are vaccinated and we study here the impact mainly on pilgrims and on mass gathering event local site, we do not include routine vaccines in those clusters.
Our model is calibrated over the 1995-2001 period. We do not consider any vaccination during this period as vaccination program for pilgrims was only bivalent A/C, that did not cover all circulating strains as outbreaks in 2000 and 2001 demonstrate.
Model is calibrated on the number of cases both in pilgrims and in Mecca and KSA local clusters.
The quadrivalent vaccine is used worldwide from 2002 for the local population and pilgrims against the meningococcal infection which is found to be 93% effective [16]. This allows us to estimate the vaccine efficacy ϑ to be 0.93. Other studies considered 0.6 − 0.9 for ϑ [15]. Following Karachaliou et al. [15], the waning rate of vaccine-induced-immunity was set to θ = 0.1/year.

S2.5. Hajj outbreaks, IMD cases, and annual incidence
During the past decades several large meningitis outbreaks occurred in the Hajj along with some sporadic cases. Yezli, et al. reported an annual IMD cases in Mecca and KSA for the period of 1987-2015 [23]. They did not distinguish the cases in terms of pilgrims and non-pilgrims. Using the proportion of Hajj participants and estimation reported by Memish et al. [21], we approximate the pilgrims and non-pilgrims cases. Based on these studies, we also estimate the annual number of cases (pilgrims and non-pilgrims) in KSA and Mecca clusters for the calibration period as shown in Table  S6. From these observed annual IMD cases we estimate the annual incidence rates (number of cases per 100000 population) for the pilgrims and residents of KSA and Mecca clusters over the period of 1995-2011, shown in Figure S2.  2  17  1996  26  1  10  1997  72  9  27  1998  27  5  10  1999  14  0  6  2000  175  51  112  2001  159  31  126  2002  7  3  45  2003  5  2  37  2004  1  1  8  2005  3  0  15  2006  3  0  19  2007  2  0  11  2008  1  0  6  2009  1  0  5  2010  0  0  3  2011  1  0  5 Source : Derived from data reported in Memish et al. [21] and Yezli et al. [23]  Source : Derived from data reported in Memish et al. [21] and Yezli et al. [23]. Note that the incidence is calculated on annual basis and do not account for differences in actual duration of exposure between pilgrims (Hajj period) and local population (year-round)