Infection dynamics for fluctuating infection or removal rates regarding the number of infected and susceptible individuals Authors: Seong Jun Park, Seong-Geun Jeong, M.Y. Choi Published: 2026 ABSTRACT In general, the rates of infection and removal (whether through recovery or death) are nonlinear functions of the number of infected and susceptible individuals. One of the simplest models for the spread of infectious diseases is the SIR model, which categorizes individuals as susceptible, infectious, recovered or deceased. In this model, the infection rate, governing the transition from susceptible to infected individuals, is given by a linear function of both susceptible and infected populations. Similarly, the removal rate, representing the transition from infected to removed individuals, is a linear function of the number of infected individuals. While nonlinear infection and removal rates have been extensively studied in deterministic epidemiological models, analytic results for stochastic dynamics with general nonlinear rates remain limited. This work presents an analytic expression for the number of infected individuals considering nonlinear infection and removal rates. In particular, we examine how the number of infected individuals varies as cases emerge and obtain the expression accounting for the number of infected individuals at each moment. This work paves the way for new quantitative approaches to understanding infection dynamics. Introduction Since the beginning of human history, infectious diseases caused by pathogens have posed a persistent threat. Viruses, which exist at the boundary between living and non-living entities, are microscopic infectious agents that replicate only within host organisms. Approximately 270 virus species are estimated to infect humans, causing over 200 diseases (Evans, 1976; Flint et al., 2020; Forni et al., 2022; L. Liu, 2014). Furthermore, illnesses caused by viruses are responsible for millions of mortalities on an annual basis (Carroll et al., 2018; Evans, 1976). The proliferation of new virus species, driven by technological advancements, has led to a steady rise in the number of identified viruses (Harko et al., 2014; Hethcote, 2000; Kermack & Mckendrick, 1927). The susceptibility of humans to viral infections is multifaceted, influenced by factors such as high population density, deforestation, and transportation infrastructure. The analysis of interhost infection dynamics aims to determine the number of infected individuals. This analysis is instrumental in predicting herd immunity and formulating effective quarantine measures, such as social distancing. A substantial body of research has been dedicated to the development and application of mathematical models in epidemiology. Mathematical modeling of infectious disease spread involves categorization of the population, with individuals capable of moving between these categories. One of the most straightforward compartment models is the susceptible, infectious, and recovered (SIR) model, which includes the numbers of susceptible, infectious, and recovered or deceased individuals (Harko et al., 2014; Hethcote, 2000; Kermack & Mckendrick, 1927; Kröger & Schlickeiser, 2020). In the SIR model, the total population, the sum of the three compartments, is fixed. Numerous studies have been conducted on modifications and extensions of the SIR model, such as models where there is no immunity after recovery (SIS model) (Hethcote, 1989; Kuhl, 2021), where immunity lasts only for a short period (SIRS) (Brauer & Castillo-Chavez, 2013; Cai et al., 2015; Kermack & McKendrick, 1933), where there is a latent period of the disease when the person is not infectious (SEIS and SEIR; E: exposed) (Anderson & May, 1991; Bianchi et al., 2019; Brauer & Castillo-Chavez, 2013; Kermack & Mckendrick, 1927), where infants may be born with immunity (MSIR; M: maternal immunity) (Brauer & Castillo-Chavez, 2013; Hethcote, 2000), and where the immunity wanes over time after infection (SEIRS) (Bjørnstad et al., 2020; Brauer & Castillo-Chavez, 2013; Hethcote, 2000). Many compartment models have been developed in epidemiology to estimate the total number of infected and recovered individuals (Allen et al., 2008; Gao et al., 2007; Kastalskiy et al., 2021; Keeling & Rohani, 2008; Paxson & Shen, 2022). Viral infection and removal (recovery or death) rates are linear functions of the number of compartments (S or I) in the aforementioned models of disease spread. However, infection, recovery, and mortality rates may not be linear functions of the number of susceptible or removed individuals because each time a virus infection occurs, the rates vary with many factors, such as individual character, social environment, infection route (e.g., contact or droplet), and intrinsic properties of the virus. Several studies have focused on viral infection dynamics involving specific nonlinear infection and removal rates (Brauer & Castillo-Chavez, 2013; Calvo-Monge et al., 2023; Hethcote, 2000; Rohith & Devika, 2020). Nonlinear infection (incidence) rates have been widely studied in epidemiological models since the early work of (Capasso & Serio, 1978) and subsequent developments (Hethcote, 2000; W. Liu et al., 1987; Xiao & Ruan, 2007). These studies introduced various functional forms, including saturated and behavioral incidence rates, to capture realistic contact processes. However, most of these studies are formulated within deterministic compartment models and focus on specific functional forms of the infection and removal rates. Stochastic epidemic models have also been extensively investigated (Allen, 2008; Bailey, 1975; Nåsell, 2002), primarily focusing on Markovian dynamics with linear or specific rate structures. Most analytic results in these frameworks rely on linear or otherwise restricted rate functions, which limit their applicability to systems with more complex, interaction-driven dynamics. Analytic results for stochastic infection dynamics with arbitrary nonlinear infection and removal rates remain limited. In this work, we address this gap by developing a general analytic framework for stochastic infection dynamics with arbitrary nonlinear infection and removal rates. Our approach is based on a birth-death process, where infection and removal events are respectively modeled as birth and death. We derive explicit expressions for the time-dependent probability distribution of the number of infected individuals, as well as for key summary quantities such as the mean population size. Park and Choi (Park & Choi, 2023) studied a birth-death process in which birth and death rates depend on the population size of a product or customer undergoing the birth-death process. Their proposed framework derives product number counting statistics to extend the framework presented in (Park & Choi, 2023) to the disease spread model. This paper discusses the number of compartments (susceptible, infected, recovered or deceased individuals) when the infection and removal (recovery or death) rates are nonlinear functions of the number of infected and susceptible individuals. This work derives analytic expressions for the numbers of susceptible, infected, and recovered or deceased individuals when the infection and removal rates are nonlinear functions of the numbers of infected and removed (recovered or deceased) cases and that the sum of the three compartments (susceptible, infected, and removed individuals) is fixed. In real-world disease spread, the number of infected individuals increases due to susceptible individuals becoming infected and the arrival of externally infected individuals. Therefore, we provide an analytic expression for the number of infected and removed cases, accounting for the arrival of infected individuals. The computational simulations prove that the analytical results are correct. We use the term “nonlinear infection or removal rates” for nonlinear functions of the number of susceptible and infected individuals. The main contributions of this study are analytical expressions for the number of individuals in three compartments (susceptible, infected, and removed) based on infection and removal rates as arbitrary functions of the number of susceptible or infected individuals. This paper is organized as follows: Section 2 explains nonlinear infection and removal rates, providing analytical expressions for the number of individuals in the susceptible, infected, and removed compartments. This work assumes that the total population remains constant, that removed individuals (recovered or deceased) cannot be reinfected, and that there is no influx of infected individuals. Next, Section 3 describes the effect of the arrival of infected individuals on the numbers of susceptible, infected, and removed individuals, presenting the analytical expression of the populations of the three compartments. Finally, Section 4 summarizes and discusses the work. Infection dynamics without the arrival of infected individuals We suggest a new approach to analyzing and predicting the number of susceptible, infected, and removed individuals under nonlinear infection and removal rates. From an infected individual's viewpoint, infectious disease spread is a birth-death process because the infection corresponds to the birth of an infected case, whereas removal corresponds to its death. The infection rate is a function of the number of susceptible and infected individuals, and the removal rate is a function of the number of infected individuals. Hence, this work considers a birth-death process where the birth and death rates depend on the number of subjects undergoing the birth-death process and the number of subjects interacting with the subjects of interest. This work obtains the analytic expression of the mean number of subjects undergoing a birth-death process whose birth and death rates are arbitrary functions of the number of subjects involved. The mean number of subjects undergoing the birth-death process is the number of infected individuals because the mean E(X) is the value of b that minimizes E((X−b)2), where E is the mathematical expectation, X denotes a random variable, and b is not a function of X (Hogg et al., 2013). Next, this work examines the birth-death process described by A→B→X, where B undergoes birth and death. The birth rate (the transition rate from A to B) depends on A and B; the death rate (the transition rate from B to X) is a function of B alone. This work assumes that the total population (the sum of populations A, B, and X) is a constant, C. Initially, there are C – 1 members in population A, one member in population B, and no member in population X. This birth-death process is similar to the spread of an infectious disease: Compartments A, B, and X correspond to the sets of susceptible, infected, and removed individuals, respectively. In general, the infection rate depends on the number of susceptible and infected individuals, and the removal rate depends on the number of infected individuals. When studying the time evolution of birth-death processes, the most common objective is determining the probability Pn(t) of n members in compartment B at time t. Since B corresponds to infected individuals, Pn(t) is the probability that there are n infected individuals at time t. When the numbers of susceptible and infected individuals are s and n, respectively, the time until the following infection occurs is an exponential random variable with rate λn,s(≥0). During an infection, the number of infected cases increases by one (changing from n to n + 1), and the number of susceptible individuals decreases by one (from s to s – 1). When the infection count exceeds one (n≥1), the time until the subsequent removal is an exponential random variable with rate (μn≥0). At death or recovery, the number of infected cases decreases by one while the number of removed individuals increases by one. In realistic interhost infections, the infection and removal rates may depend not only on the numbers of infected and susceptible individuals but also on time. We rewrite the rates as λn,s(t) and μn(t). In this work, we focus on infection and removal rates that depend only on the number of infected and susceptible individuals. Extensions to time- and compartment-size- dependent rates are left for future work. We now obtain the probability Pn(t) as a function of the infection rate λn,s and the removal rate μn, based on a disease spread model that satisfies the following three assumptions. First, the total population (the sum of the susceptible, infected, and removed populations) is constant C. Second, removed individuals cannot be reinfected. Third, at t = 0, the number of susceptible individuals is C – 1, with one index case and no individuals removed (recovered or deceased). To our knowledge, no analytic expression of Pn(t) exists for infection and removal rates given by nonlinear functions of the numbers in compartments in this infectious disease spread model. To derive Pn(t), we introduce the probability density function of time to complete m infections with n infected individuals at the instant of the mth infection, denoted by fn,m(t). Under these assumptions, Pn(t) is written as follows (see the Supplementary Material):(1){P0(t)=∫0t(μ1e−(μ1+λ1,S0)τ+∑m=1S0∑k=2m+1∫0τfk,m(τ1)vk,m,S0(τ−τ1)dτ1)dτPn(t)=∑m=nS0fn+1,m(t)λn,S0−(m−1)n=1,2,···,C−1PC(t)=∫0tfC,S0(τ)e−μC(t−τ)dτ,where fk,m(t) (m=1,2,···,S0;k=2,3,···,m+1) satisfies(2)fk,m(t)=∫0tfk−1,m−1(τ)λk−1,S0−(m−1)e−(λk−1,S0−(m−1)+μk−1)(t−τ)dτ+∑j=0m−k{(∏l=0jμk+j−l)∫0tfk+j,m−1(τ1)∫0t−τ1gk,j,S0,m(τ2)λk−1,S0−(m−1)e−(λk−1,S0−(m−1)+μk−1)(t−τ1−τ2)dτ2dτ1}with gk,j,S0,m(t) representing the inverse Laplace transform of gˆk,j,S0,m(s)=∫0∞gk,j,S0,m(t)e−stdt=∏l=0j(s+μk+j−l+λk+j−l,S0−(m−1))−1, and vk,m,S0(t) denotes the inverse Laplace transform of vˆk,m,S0(s)=∫0∞vk,m,S0(t)e−stdt=∏l=1kμls+μl+λl,S0−m. The function fk,m(t) can be interpreted as the probability density that the m-th infection occurs at time t, with k infected individuals present at that moment. Thus, it encodes the temporal sequence of infection events. The integral structure of fk,m(t) reflects a summation over all possible sequences of infection and removal events: to have k infected individuals at time t when the m-th infection occurs, the first term in Eq. (2) represents the case where there are k-1 infected individuals at the time τ of the (m-1)-th infection, and one additional infection occurs without any removal during the interval t−τ. The second term accounts for cases where there are k + j infected individuals at the (m-1)-th infection at time τ1, followed by j+1 removal events during the interval τ2, and then one infection occurs without removal during t−τ1−τ2. Equation (2) presents a hierarchical structure, where fk,m(t) is determined by f1,m−1(t),f2,m−1(t),···, and fk−1,m−1(t). Thus, we can evaluate fk,m(t) only if f2,1(t) is known. Under the initial condition (the third assumption), we have f2,1(t)=λ1,S0e−(λ1,S0+μ1)t because the first infection occurs at a rate of λ1,S0. Moreover, the index case recovery or death must not occur before the first infection. We can see from Eq. (2) that fk,m(t) is always positive, since λn,s and μn are positive. From Eq. (1), the positivity of fk,m(t), λn,s and μn ensures that Pn(t) is also positive. We now examine the physical interpretation of Eq. (1). First, we consider its first equation: the first integral represents the case where the index case is removed before any infection occurs, while the second integral accounts for the removal of all infected individuals after at least one infection has occurred. Next, we turn to the second equation, where the term fn+1,m(t)λn,S0−(m−1) represents the contribution to Pn(t) for n=1,2,···,C−1. Finally, we consider the last equation. It shows that, to have C infected individuals at time t, the system already C infected individuals at an earlier time, with the required number (S0) of infection events, and no removal occurring during the intervening period. Equation (1), which describes how the number of infected individuals varies in time, manifests that Pn(t) satisfies the normalization condition: ∑n=0CPn(t)=1. Differentiating Eq. (1), we obtain the foundational equation, i.e., the differential-difference equation governing the time evolution of a birth-death process in this infectious disease spread model:(3){∂P0(t)∂t=μ1P1(t)∂P1(t)∂t=−∑m=1S0f2,m(t)−μ1P1(t)+μ2P2(t)∂Pn(t)∂t=−∑m=nS0fn+1,m(t)−μnPn(t)+∑m=n−1S0fn,m(t)+μn+1Pn+1(t)n=2,3,···,C−1∂PC(t)∂t=fC,S0(t)−μC∫0tfC,S0(τ)e−μC(t−τ)dτ. The structure of Eq. (3) resembles differential–difference equations arising in stochastic epidemic models and branching processes (Allen, 2008; Harris, 1963). However, classical branching processes typically assume independent individuals and linear transition rates. In contrast, the present formulation incorporates nonlinear infection and removal rates that depend on both susceptible and infected populations, as well as a finite population constraint. Therefore, our approach can be viewed as a generalization of branching process models to interacting systems with arbitrary rate functions. Next, we derive the numbers of susceptible, infected, and removed individuals using Eq. (1), where the infected population is a random variable. The mean number of infections obtains the form(4)⟨n(t)⟩=∑n=0CnPn(t)=∑n=1S0∑m=nS0nfn+1,m(t)λm,S0−(m−1)+C∫0tfC,S0(τ)e−μC(t−τ)dτ,where fn,m(t) is given by Eq. (2). The first term of the right-hand side of eq. (4) is derived from ∑n=0C−1nPn(t) and the last term is equivalent to CPC(t). Given a birth-death process, removed (recovered or deceased) individuals undergo only one birth. With the removal (recovery or death) rate μn, the recovery count and death toll r(t) thus reads (Park & Choi, 2022; Ross, 2019):(5)⟨r(t)⟩=∫0t⟨μn(τ)⟩dτ,where ⟨μn(t)⟩=∑n=0CμnPn(t) is the mean removal rate. In terms of removed individuals, this infection model is identical to pure birth process. Thus, the number of removed individuals is the integral of the mean removal rate with respect time. Then the number s(t) of susceptible individuals in the total population C is given by(6)⟨s(t)⟩=C−⟨n(t)⟩−⟨r(t)⟩.In Eqs. (4), (5), (6), the number in each compartment must be computed even if the infection and removal rates are arbitrary functions of infected and susceptible populations. The positivity of Pn(t) makes it clear that the numbers of infected, susceptible, and removed individuals remain positive for all time t. Fig. 1 presents the number of susceptible, infected, and removed individuals for several rates of infection and removal, demonstrating that Eqs. (4), (5), (6) agree with the simulation results.Fig. 1Infected (I), removed (R), and susceptible (S) populations with nonlinear infection and removal rates without the arrival of infected cases. Circles and lines plot data from simulations and theoretical results from Eqs. (4), (5), (6), respectively, for rates (a) λn,s=0.07n1.2s0.6 and μn=0.2n1.3; (b) λn,s=0.2(n+s)0.8 and μn=0.1n1.2. Infected (I), removed (R), and susceptible (S) populations with nonlinear infection and removal rates without the arrival of infected cases. Circles and lines plot data from simulations and theoretical results from Eqs. (4), (5), (6), respectively, for rates (a) λn,s=0.07n1.2s0.6 and μn=0.2n1.3; (b) λn,s=0.2(n+s)0.8 and μn=0.1n1.2. The compartment sizes are obtained through a sequence of steps. The derivation of the analytic expressions for the susceptible, infected, and removed populations is summarized schematically in Fig. 2.Fig. 2Schematic overview of the analytic framework for infection dynamics without external arrivals. This scheme illustrates the derivation from nonlinear infection and removal rates to the number of susceptible, infected, and removed individuals. The infection rate λn,s increases the number of infected individuals from n to n+1 while decreasing the number of susceptible individuals from s to s-1. The removal rate μn decreases the number of infected individuals by one and increases the number of removed (recovered or deceased) individuals by one. The function fk,m(t) (Eq. (2)) is the probability density for the time at which m infections have occurred, with k infected individuals present at that moment. Equation (1) gives Pn(t), the probability of having n infected individuals at time t. Equations (4), (5), (6) provide the numbers of infected, removed, and susceptible individuals, respectively. Schematic overview of the analytic framework for infection dynamics without external arrivals. This scheme illustrates the derivation from nonlinear infection and removal rates to the number of susceptible, infected, and removed individuals. The infection rate λn,s increases the number of infected individuals from n to n+1 while decreasing the number of susceptible individuals from s to s-1. The removal rate μn decreases the number of infected individuals by one and increases the number of removed (recovered or deceased) individuals by one. The function fk,m(t) (Eq. (2)) is the probability density for the time at which m infections have occurred, with k infected individuals present at that moment. Equation (1) gives Pn(t), the probability of having n infected individuals at time t. Equations (4), (5), (6) provide the numbers of infected, removed, and susceptible individuals, respectively. It is of interest to compare Eq. (3) with the number of each compartment in the SIR model. The SIR model is described by the following system of ordinary differential equations:(7){dS(t)dt=−βI(t)S(t)dI(t)dt=βI(t)S(t)−γI(t)dR(t)dt=γI(t)with positive constants β and γ, where S(t), I(t), and R(t) denote the number of susceptible, infected, and removed (recovered or deceased) individuals, respectively, at time t. Multiplying both sides of Eq. (3) by n and summing over n (=0,1,2,···,C) yield the rate of change in the number of infected individuals, which differs from dI(t)/dt given in Eq. (7). Accordingly, Eqs. (4), (5), (6) do not agree with the solutions of the set of ordinary differential equations describing the SIR model. Fig. 3 presents the discrepancy between the results from Eqs. (4), (5), (6) and the solutions of Eq. (7).Fig. 3Infected (I), removed (R), and susceptible (S) populations resulting from Eqs. (4), (5), (6) (solid lines) and from the SIR model described by Eq. (7) (broken lines). Rates are given by (a) (λn,s,μn)=(0.1ns,0.12n) in Eqs. (4), (5), (6) and (β,γ)=(0.1,0.12) in Eq. (7); (b) (λn,s,μn)=(0.05ns,0.2n) in Eqs. (4), (5), (6) and (β,γ)=(0.05,0.2) in Eq. (7). Infected (I), removed (R), and susceptible (S) populations resulting from Eqs. (4), (5), (6) (solid lines) and from the SIR model described by Eq. (7) (broken lines). Rates are given by (a) (λn,s,μn)=(0.1ns,0.12n) in Eqs. (4), (5), (6) and (β,γ)=(0.1,0.12) in Eq. (7); (b) (λn,s,μn)=(0.05ns,0.2n) in Eqs. (4), (5), (6) and (β,γ)=(0.05,0.2) in Eq. (7). To illustrate the practical impact of nonlinear infection rates, we include a representative comparison between a nonlinear rate and its corresponding linear approximation. Specifically, we consider a nonlinear infection rate of the form (λn,s=βns1+αn) and compare it with the standard linear rate (λn,s=βns) in Fig. 4, while modeling the removal rate as μn=γn. The nonlinear infection rate (λn,s=βns1+αn) captures saturation effects due to limited contact opportunities or behavioral changes at high infection levels. The results show that the nonlinear rate leads to slower growth and a reduced peak in the number of infected individuals compared to the linear approximation. This difference arises from the effective suppression of infection at higher population levels, which cannot be captured by linear models. This example demonstrates that allowing nonlinear infection rates can significantly alter the temporal dynamics of the system.Fig. 4Comparison of infection dynamics under a nonlinear infection rate (λn,s=βns1+αn) and its linear approximation (λn,s=βns). The removal rate is μn=γn. The parameter values are β=0.05, α=1, and γ=0.2. The solid and dotted lines show compartment sizes under the linear and nonlinear infection rates, respectively, computed from Eqs. (4), (5), (6). (a) the number of infected individuals; (b) the number of removed individuals; (c) the number of susceptible individuals. Comparison of infection dynamics under a nonlinear infection rate (λn,s=βns1+αn) and its linear approximation (λn,s=βns). The removal rate is μn=γn. The parameter values are β=0.05, α=1, and γ=0.2. The solid and dotted lines show compartment sizes under the linear and nonlinear infection rates, respectively, computed from Eqs. (4), (5), (6). (a) the number of infected individuals; (b) the number of removed individuals; (c) the number of susceptible individuals. Infection dynamics with the arrival of infected individuals In many real-world situations where an infectious disease is spreading, the number of infected individuals may increase due to the arrival of infected individuals. Therefore, the increase in the number of infected cases can be divided into two types: infections started by the index case and the influx of infected cases. This work addresses these two types of increases in infected cases. The infectious disease spread model studied is the same as in Section 2, except for the arrival of infected individuals. This work focuses on the current number of infected individuals. Suppose the number of infections introduced is v, occurring at a1,a2,···,av. Starting with one infected individual and S0 susceptible individuals, the total population C reaches S0+v+1 after time av when the last introduced case occurs. As the infected population increases due to infections and arrivals, instead of fn,m(t), we introduce fk,m,r(t), the probability density of the time to complete m infections and r arrivals with k infected individuals at this time, where k runs from 1 to m + r+1 (k=1,···,m+r+1), m from 0 to S0 (m=0,1,···,S0), and r from 0 to ν (r=0,1,···,v). Under the initial condition Pn(0)=δn1, we write fk,m,r(t) as follows:(8)fk,m,r(t)=(∫0tfk−1,m−1,r(τ1)λk−1,S0−(m−1)e−(λk−1,S0−(m−1)+μk−1)(t−τ1)dτ1+∑q=0m+r−k∫0tfk+q,m−1,r∫0t−τ1gk,q,S0,m(τ2)λk−1,S0−(m−1)e−(λk−1,S0−(m−1)+μk−1)(t−τ1−τ2)dτ2dτ1)θ(ar+1−t)+(∫0arfk−1,m,r−1(τ1)e−(λk−1,S0−m+μk−1)(ar−τ1)dτ1+∑q=0m+r−k∫0arfk+q,m,r−1(τ1)∫0ar−τ1gk,q,S0,m+1(τ2)e−(λk−1,S0−m+μk−1)(ar−τ1−τ2)dτ2dτ1)δ(t−ar),where gk,q,S0,m(t) denotes the inverse Laplace transform of gˆk,q,S0,m(s)=∫0∞gk,q,S0,m(t)e−stdt=∏l=0q(s+μk+q−l+λk+q−l,S0−(m−1))−1, δ(x) represents the Dirac delta function, and θ(x) denotes the Heaviside step function. The function fk,m,r(t) can be interpreted as the probability density that the m-th infection and r arrivals occur at time t, with k infected individuals present at that moment. It thus describes the temporal sequence of infection events. Its integral structure of fk,m,r(t) represents a sum over all possible sequences of infection, arrival and removal events. To have k infected individuals at time t when the m-th infection and the r-th arrival occur, two cases are possible. (i) there are k-1 infected individuals just before time t, following the (m-1)-th infection and the r-th arrival occur, and one infection occurs during the remaining time; or (ii) there are k-1 infected individuals after the m-th infection and the (r-1)-th arrival, and one arrival occurs during the remaining time. The first and second cases correspond to the first and second terms in Eq. (8), respectively. Equation (8) can only be evaluated if f1,0,1(t), f2,0,1(t), and f2,1,0(t) are known. In the case of m = 0 and r = 1, the index case is removed to have one infection count at time t, and the infection must not occur until time a1. This case is expressed as f1,0,1(t)=∫0a1μ1e−(μ1+λ1,S0)τdτδ(t−a1). If the index case survives and becomes infected during the time a1, the number of infected individuals is two at time a1. Then, we have f2,0,1(t)=e−(μ1+λ1,S0)a1δ(t−a1). The case of m = 1 and r = 0 represents one infection, with no infected individuals arriving. There are two cases immediately after the first infection: namely, f2,1,0(t)=λ1,S0e−(λ1,S0+μ1)tθ(a1−t). For a detailed derivation of Eq. (8), see the Supplementary Material. We can see from Eq. (8) that fk,m,r(t) is always positive, since λn,s and μn are positive. The next step is to obtain Pn(t). Equation (8), which accounts for the infection count at the time of infection and arrival, leads Eq. (9) to obtain the form(9){P0t=∫0tμ1e−μ1+λ1,S0τθa1−t+∑m=0S0∑r=0m+r≥1v∑k=1m+r+1∫0τfk,m,rτ1vk,m,S0τ−τ1dτ1θar+1−tdτPnt=12∑m=0S0∑r=0m+r≥nvxn+1,m,rtλn,S0−m−1+yn+1,m,rtθar−tδt−arn=1,2,···,S0+vPS0+v+1t=∫0tfS0+v+1,S0+1,vτe−μS0+v+1t−τdτwith xk,m,r(t)=(∫0tfk−1,m−1,r(τ1)λk−1,S0−(m−1)e−(λk−1,S0−(m−1)+μk−1)(t−τ1)dτ1+∑q=0m+r−k∫0tfk+q,m−1,r∫0t−τ1gk,q,S0,m(τ2)λk−1,S0−(m−1)e−(λk−1,S0−(m−1)+μk−1)(t−τ1−τ2)dτ2dτ1)θ(ar+1−t), yk,m,r(t)=(∫0arfk−1,m,r−1(τ1)e−(λk−1,S0−m+μk−1)(ar−τ1)dτ1+∑q=0m+r−k∫0arfk+q,m,r−1(τ1)∫0ar−τ1gk,q,S0,m+1(τ2)e−(λk−1,S0−m+μk−1)(ar−τ1−τ2)dτ2dτ1)δ(t−ar), and vk,m,S0(t) the same as that given below Eq. (2). From Eq. (9), the positivity of fk,m,r(t), λn,s and μn ensures that Pn(t)(n=0,1,2,···,S0+v+1) is also positive. We now consider its physical interpretation. In the first equation, the first integral represents the case where the index case is removed before any infection or arrival occurs, while the second integral accounts for the removal of all infected individuals after at least one infection or arrival has taken place. Next, in the second equation, the terms xn+1,m,r(t)λn,S0−(m−1) and yn+1,m,r(t)θ(ar−t)δ(t−ar) represent contributions in which either an infection or an arrival is the final event before the number of infected individuals reaches n. Finally, the last equation shows that, to have S0+v+1 infected individuals at time t, the system must already have S0+v+1 infected individuals at an earlier time, with the required number (S0+v) of infection events having occurred and no removal taking place during the intervening period. The differential-difference equation satisfied by Eq. (9) has a similar structure to that of Eq. (3):(10){∂P0t∂t=μ1P1t∂P1t∂t=−∑m=0S0∑r=0m+r≥1vf2,m,rt−μ1P1t+μ2P2t∂Pnt∂t=−∑m=0S0∑r=0m+r≥nvfn+1,m,rt−μnPnt+∑m=0S0∑r=0m+r≥n−1vfn,m,rt+μn+1Pn+1tn=2,3,···,S0+v∂PCvt∂t=fCv,S0,vt−μCv∫0tfCv,S0,vτe−μCvt−τdτ Equation (9) yields the number of infected individuals as follows:(11)⟨nt⟩=12∑n=1S0+v∑m=0S0∑r=0m+r≥nvnxn+1,m,rtλn,S0−m−1+yn+1,m,rtθar−tδt−ar+S0+v+1∫0tfS0+v+1,S0+1,vτe−μS0+v+1t−τdτWhen there is no external infection arrival (v = 0), Eq. (11) reduces to Eq. (4), and Eq. (9) reduces to Eq. (1). The number of removed individuals has the same structure as Eq. (5): ⟨r(t)⟩=∫0t⟨μn(τ)⟩dτ, except for ⟨μn(t)⟩=∑n=0CvμnPn(t) with Pn(t) given in Eq. (9). The total population, denoted by Cv, fluctuates over time in such a way thatCv={S0+1t1, reflecting heterogeneous contacts and deviations from homogeneous mixing. Behavioral changes in response to increasing infection levels can also be incorporated, for example with λn,s=βnse−αn, which reduce transmission as the number of infected individuals increases. We have also examined the number of individuals in each compartment (S, I, or R) upon the arrival of infected individuals. The increase in the infected population stems from two sources: infections from the index case and external infections. This work derives the expression by considering the probability density function fk,m,r(t) of the time required to complete m infections and r arrivals with k infected individuals, written as Eq. (8). This relates to the probability Pn(t) for the number n of infected individuals at time t, given in Eq. (9). Finally, Eq. (11) yields analytic expressions for the susceptible, infected, and removed populations. Fig. 5 demonstrates that Eq. (11) is indeed correct. The most important aspect of Eq. (4) through (6) and (11) is that they provide analytic expressions for each compartment in the infectious disease spread model, even though the infection and removal rates, denoted by λn,s and μn, respectively, are arbitrary functions of the number of susceptible and infected individuals. It is easy to observe interacting subjects undergoing a birth-death process in which the birth and death rates are functions of the number of interacting compartments. For example, analytic expressions [Eqs. (4), (5), (6)] apply to consecutive reactions in chemistry, interspecific interactions in ecology, and other phenomena. The framework developed in this work to derive analytic expressions can be applied to study population dynamics, first-passage time (Park & Choi, 2024; Redner, 2001), queue waiting time (Gross et al., 2011; Kleinrock & Gail, 1996), and transport dynamics (Song et al., 2019; Spitzer, 2013). This framework provides a novel approach to analyzing quantitatively birth-death processes involving interacting subjects and feedback (i.e., autoregulation). Our goal is to extend the method of deriving the analytic results to more complex and realistic birth-death processes involving finite death channels, more than three interacting subjects, and bulk births or deaths. These extensions have broad applications in diverse fields such as queueing theory, chemical kinetics, and epidemiology. Further research could clarify the time-evolution dynamics of the product number undergoing these birth-death processes. Ultimately, we could quantitatively understand and control these processes. Future work will explore a more realistic infectious disease spread model covering reinfection, emergence of infected individuals unlinked to the index case, multiple index cases at time zero, and unlimited population size. CRediT authorship contribution statement Seong Jun Park: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Software, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Data curation, Conceptualization. Seong-Geun Jeong: Formal analysis, Funding acquisition, Investigation, Validation, Visualization, Writing – review & editing. M.Y. Choi: Writing – review & editing, Visualization, Validation, Supervision, Software, Resources, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.