In this session, we will discuss the definition and some uses of incubation period. The incubation period of an infectious disease is defined as the time from infection to symptoms onset. The incubation period may be substantially different among people infected with the same pathogen due to several factors. One of the most important factors is the dose and route of infection. For example, higher infection dose of salmonella is associated with faster symptoms onset and, therefore, a shorter incubation period. Biological factors of the infected host are also important. Some people may be more susceptible to faster disease progression and severe complications because of their age, genetics or competence of their immune systems. For example, the incubation period of AIDS tends to be longer in HIV-infected individuals who are younger. Pharmacologic factors are also important. For example, between 1975 and 1985, the incubation period of AIDS among HIV-infected individuals was substantially lengthened by the advent and timely use of effective antiretroviral therapies. The frequency distributions of incubation period for many diseases are typically right-skewed, which means that they have long right tails, which, in turn, means that the incubation period is much longer than average for a significant proportion of infected people. To summarize the statistical variations of incubation period, we often describe its empirical frequency distribution with commonly-used probability distributions, for example, lognormal distribution, which means that the frequency distribution looks like a normal distribution if the incubation period is plotted on a log-scale along the horizontal axis. Let us now look at some real examples. This is the frequency distribution of the incubation period of 168 SARS patients in Hong Kong in 2003. The best-fit lognormal distribution for these data has a median of 4.4 days and a 90th percentile of around 10 days, which means that 90% of SARS patients had incubation period shorter than 10 days. Obviously, the incubation periods of different infectious diseases can be very different. The incubation period of human cases of avian influenza H5N1 and H7N9 has a median of around four days. In comparison, the incubation period of seasonal influenza is shorter with a median of around one day, while the incubation period of measles is longer with a median of around 12 days. Knowledge about the incubation period of an infectious disease is important for clinical management. For example, shorter incubation period is associated with high risk of severe complications for tetanus. Similarly, delaying the onset of symptoms using antiretroviral therapies can reduce the risk of opportunistic infections and lengthen the survival time for AIDS patients. Knowledge about the incubation period is also important for public health control. For example, to control the spread of a communicable disease, especially for emerging infectious diseases such as SARS, health policymakers may decide to quarantine suspected cases and contacts of confirmed cases. Quarantine means isolating suspected cases and contacts from the general community to prevent disease transmission in case they have been infected but do not yet have the clinical or virological evidence of infection. The incubation period is often used as the basis for determining the duration of quarantine. For example, by setting the quarantine duration to be the 90th percentile of the incubation period distribution, we can expect that at least 90% of the infected people who are quarantined will show symptoms during quarantine and, therefore, be identified and followed up. During the SARS epidemic in Hong Kong in 2003, the quarantine duration for suspected cases and contacts of confirmed cases was 10 days, which corresponded to the 90th percentile of the best estimate of incubation period distribution at that time. An accurate estimate of the incubation period distribution is also crucial for interpreting disease surveillance data and understanding epidemic dynamics. During the early decades of the HIV epidemic, disease surveillance of HIV was mostly based on AIDS diagnoses because there was no HIV screening. During those years, the incubation period of AIDS had a median of around nine years and could range from three to more than 12 years, so the observed AIDS cases represented only a small proportion of the total number of HIV infections in the population. As a hypothetical example, suppose the incubation period distribution of AIDS looks like this, which ranges from 3 to 15 years with a median of 9 years, and suppose the annual number of new HIV infections from Year 1 to 20 looks like this. Those infected in Year 1 would develop AIDS in Year 4 to 16 according to the incubation period distribution. Similarly, those infected in Year 2 would develop AIDS in Year 5 to 17, and so on. In any given Year t, the cumulative number of AIDS cases is equal to the number of HIV infections in Year 1 times the proportion of infections whose incubation period is shorter than t minus 1 years, plus the number of HIV infections in Year 2 times the probability that the incubation period is shorter than t minus 2 years, and so on. For example, the cumulative number of AIDS cases in Year 10, which is six years after the first diagnosis of AIDS, represents only 20% of the total number of HIV infections in this example. In short, the cumulative number of AIDS cases in Year t is equal to the sum of the number of HIV infections in all years before t, weighted by the probability that the incubation period is shorter than the time elapsed between the year of infection and Year t. This method of inferring HIV incidence from AIDS epi-curves and incubation period is called backcalculation. Robust backcalculation of HIV incidence from AIDS epi-curves obviously requires an accurate estimate of the incubation period distribution. To summarize, in this session, we have defined incubation period and discussed some practical uses of knowledge about incubation period.