We analyzed a variety of epidemiological datasets to quantify the exposure patterns and transmission characteristics of MERS and SARS by disease generation and in different settings, including aggregated case counts, individual-level case data, and detailed transmission trees, as detailed below.
Individual-level case data to quantify exposure patterns
Middle East respiratory syndrome
We analyzed a publicly available line list of MERS cases reported between March 2013 and May 2015 to the Kingdom of Saudi Arabia (KSA) Ministry of Health . For each case, we obtained the date of reporting, healthcare worker status, and whether the infection had been linked to healthcare facilities. Healthcare facilities in KSA report MERS cases to the Ministry of Health through an electronic case-reporting system once all appropriate testing is complete . Case confirmation is based on laboratory diagnosis through detection of viral nucleic acid or serology, regardless of the presence of clinical signs and symptoms .
Severe respiratory syndrome
Total SARS case counts for Canada, China, Hong Kong, Singapore, and Vietnam including cases among healthcare workers were obtained from the World Health Organization (WHO) website for the outbreaks in 2003 . A probable case of SARS was defined as radiographic evidence of pneumonia or respiratory distress syndrome on a chest X-ray, positivity for SARS virus infection by one or more laboratory assays, or autopsy findings consistent with the pathology of respiratory distress syndrome . A confirmed case was defined based on a positive laboratory test combined with clinical evidence compatible with SARS.
Transmission trees to quantify transmission characteristics by generation time
We obtained and analyzed detailed transmission trees for hospital clusters of MERS and SARS. Transmission trees provide information on the epidemiological links between successive cases and allow for quantification of the reproduction number, R, which is a key parameter in outbreak investigations. In general, R quantifies the transmission potential of an infectious pathogen, which informs the likelihood of large-scale outbreaks [31, 32]. Estimates of R > 1 indicate the potential for an infectious pathogen to generate a major outbreak while R < 1 indicates that transmission of a given pathogen cannot be sustained in the population. Here, we note Rg as the reproduction number at disease generation g, where g ≥ 0. If g = 0 then R0 denotes the index reproduction number, or the number of secondary cases ascribed to the index case in a given outbreak. Overall, the reproduction number is a function of several inter-related factors, including the epidemiology of the disease, local cultural factors, and environmental conditions. Moreover, the reproduction number is affected by population behavior changes and control interventions occurring over the course of an epidemic.
We defined hospital clusters as outbreaks that started in a healthcare setting with a hospitalized patient (the index case) and ended when the chain of transmission subsided and no further infections were linked to the healthcare setting. We searched for past MERS and SARS outbreaks with information on types of exposure. Specifically, we searched PubMed for articles on SARS published after 1 January 2003 with the search “(SARS AND hospital) OR (SARS AND healthcare)” and articles on MERS published after January 2012 with the search “(MERS AND hospital) OR (MERS AND healthcare).” We also screened relevant articles cited within selected articles.
Overall, we found information on types of exposure for nine hospital outbreaks (three MERS and six SARS), while detailed transmission trees were available for four of these (two SARS and two MERS clusters) [11, 13, 14, 16, 17, 33, 34]; an ill-defined SARS transmission tree from Taiwan had to be discarded.
Each case within the hospital cluster was classified according to occupational and social exposure, including healthcare workers, patients, family members and visitors, and non-clinical staff. Healthcare workers were defined as personnel responsible for the direct care of patients and included physicians, nurses, laboratory technicians, and emergency medical personnel. Hospital personnel who do not directly work with patients were categorized as non-clinical hospital staff, a group that included janitors, clerks, ambulance drivers, and firefighters.
Transmission tree of MERS hospital cluster in South Korea, 2015
We constructed the transmission tree of the South Korean outbreak, comprising 186 cases with the last case reported on 5 July 2015. For this purpose, we employed publicly available detailed case data from the WHO, the Korean Centers for Disease Control, and the Ministry of Health & Welfare of South Korea [13, 14, 33, 34]. The index patient developed symptoms on 11 May 2015 but was not diagnosed with MERS until 20 May 2015. This nosocomial outbreak involved 15 healthcare settings. While exposure information was available for all cases, only 168 (91 %) out of 185 transmission links tied to healthcare settings were ascertained through outbreak investigations.
Transmission tree of MERS hospital cluster in Al-Hasa, Saudi Arabia, 2013
We obtained a transmission tree for a nosocomial MERS outbreak comprising 25 cases that occurred between 1 April 2013 and 23 May 2013 in Al-Hasa, Saudi Arabia . This nosocomial outbreak involved four healthcare facilities . Transmission links were inferred for all secondary cases comprising this outbreak except for one case.
Transmission tree of SARS hospital cluster in Singapore, 2003
We obtained the transmission tree for a nosocomial SARS outbreak comprising 188 cases in Singapore between 25 February 2003 and 11 May 2003 . The index patient was a local resident who developed symptoms in late February 2003 and was subsequently admitted to a hospital after returning from a holiday in Hong Kong. This nosocomial outbreak involved three major hospitals in Singapore .
Transmission tree of SARS hospital cluster in Toronto, Canada 2003
Detailed information was available on a nosocomial SARS outbreak in Toronto, Canada , resulting in 90 cases between 23 February 2003 and 15 April 2003 . The index patient was a traveler returning from Hong Kong on 23 February 2003. This nosocomial outbreak developed in a single 249-bed secondary case community hospital . Infection control precautions were implemented throughout the hospital including the closing of the hospital to admissions, closing of the outpatient clinics, and quarantine orders to discharged patients .
MERS case data from Saudi Arabia were publicly available from the KSA Ministry of Health . Similarly the dataset of MERS cases in South Korea was publicly available from the WHO, the Korean Centers for Disease Control, and the Ministry of Health & Welfare of South Korea [13, 14, 33, 34]. All of the data were de-identified. These openly available datasets were generated as part of emerging outbreak investigations and were, therefore, deemed exempt from institutional review board assessment.
We tabulated the pathogen-specific frequency of healthcare and familial exposure using individual-level data and aggregated case counts. To evaluate differences in exposure, we used chi-square and Fisher’s exact tests of independence in a cross-tabulation of exposure category and outbreak. Transmission trees allowed for tabulation of exposure frequency and number of secondary infections by pathogen, individual outbreak (n = 4), and disease generation—the time interval elapsed between successive generations of cases. Next, to quantify and compare the transmission potential and extent of transmission heterogeneity for MERS and SARS nosocomial outbreaks, we used an approach recently developed to characterize the distribution of cluster sizes for subcritical pathogens . A hallmark of high transmission heterogeneity is a preponderance of very small and very large clusters (the latter being associated with super-spreading events), together with a low frequency of intermediate-size clusters. Based on this approach, we fit a negative binomial to the distribution of secondary cases obtained from the MERS and SARS transmission trees and estimated the reproduction number R and dispersion parameter k (with lower values indicating higher heterogeneity) . Armed with these estimates, we simulated the expected distribution of future outbreaks that may occur in South Korea or elsewhere in terms of final size, peak size, and outbreak duration. We used branching process models  to simulate 5,000 MERS-like and 5,000 SARS-like outbreaks based on the distribution of secondary cases inferred from empirical transmission trees. Each simulated outbreak was initiated with a single infectious individual.