Essentials of Aggregate System Dynamics Infectious Disease Models

Essentials of Aggregate System Dynamics Infectious Disease Models Nathaniel Osgood CMPT 394 February 5, 2013

Comments on Mathematics & Dynamic Modeling Many accomplished & well-published dynamic modelers have very limited mathematical background Can investigate pressing & important issues Software tools are making this easier over time Can gain extra insight/flexibility if willing to push to learn some of the associated mathematics Achieving highest skill levels in dynamic modeling do require mathematical facility and sophistication To do sophisticated work, often those lacking this background or inclination collaborate with someone with background

Applied Math & Dynamic Modeling Although you may not use it, the dynamic modeling presented rests on the tremendous deep & rich foundation of applied mathematics Linear algebra Calculus (Differentia/Integral, Uni& Multivariate) Differential equations Numerical analysis (including numerical integration, parameter estimation) Control theory Real & complex analysis For the mathematically inclined, the tools of these areas of applied math are available

Models in Mathematical Epidemiology of Infectious Disease Long & influential modeling tradition, beginning with Ross & Kermack-McKendrick (1920s) Models formulated for diverse situations (Cf. Anderson & May) Latent & incubation period/diversity in contact rates/heterogeneity/preferential mixing/vaccinated groups/zoonoses/variations in clinical manifestations/network structure Important tradition of closed-form analysis

Mathematical Models of Infectious Disease Link Together Diverse Factors Typical Factors Included Infection Mixing & Transmission Development & loss of immunity both individual and collective Natural history (often multistage progression ) Recovery Birth & Migration Aging & Mortality Intervention impact Sometimes Included Preferential mixing Variability in contacts Strain competition & crossimmunity Quality of life change Health services interaction Local perception Changes in behavior, attitude Immune response

Emergent Characteristics of Infectious Diseases Models Instability Nonlinearity Tipping points Oscillations Multiple fixed points/equilibria Endemic equilibrium Disease free equilibrium

Jan. 1942 Mar. 1942 May 1942 July 1942 Sept. 1942 Nov. 1942 Jan. 1943 Mar. 1943 May 1943 July 1943 Sept. 1943 Nov. 1943 Jan. 1944 Mar. 1944 May 1944 July 1944 Sept. 1944 Nov. 1944 Jan. 1945 Mar. 1945 May 1945 July 1945 Sept. 1945 Nov. 1945 Instability Slight perturbation (e.g. arrival of infectious person on a plane) can cause big change in results Contrast with goal seeking behaviour 600 500 400 300 200 100 0 Chickenpox

Oscillations & Delays The oscillations reflect negative feedback loops with delays These delays reflect stock and flow considerations and specific thresholds dictating whether net flow is positive or negative Stock & Flow: Stock continues to deplete as long as outflow exceeds inflow, rise as inflow>outflow The stock may stay reasonably high long after inflow is low! Key threshold R*: When # of individuals being infected by a single infective = 1 This is the threshold at which outflows=inflows

Jan. 1942 Mar. 1942 May 1942 July 1942 Sept. 1942 Nov. 1942 Jan. 1943 Mar. 1943 May 1943 July 1943 Sept. 1943 Nov. 1943 Jan. 1944 Mar. 1944 May 1944 July 1944 Sept. 1944 Nov. 1944 Jan. 1945 Mar. 1945 May 1945 July 1945 Sept. 1945 Nov. 1945 1200 1000 800 Measles Childhood Diseases in Saskatchewan Instability 600 400 200 0 Jan. 1942 Mar. 1942 May 1942 July 1942 Sept. 1942 Nov. 1942 Jan. 1943 Mar. 1943 May 1943 July 1943 Sept. 1943 Nov. 1943 Jan. 1944 Mar. 1944 May 1944 July 1944 Sept. 1944 Nov. 1944 Jan. 1945 Mar. 1945 May 1945 July 1945 Sept. 1945 Nov. 1945 600 500 400 300 200 100 0 Chickenpox

Jan. 1921 Nov. 1921 Sept. 1922 July 1923 May 1924 Mar. 1925 Jan. 1926 Nov. 1926 Sept. 1927 July 1928 May 1929 Mar. 1930 Jan. 1931 Nov. 1931 Sept. 1932 July 1933 May 1934 Mar. 1935 Jan. 1936 Nov. 1936 Sept. 1937 July 1938 May 1939 Mar. 1940 Jan. 1941 Nov. 1941 Sept. 1942 July 1943 May 1944 Mar. 1945 Jan. 1946 Nov. 1946 Sept. 1947 July 1948 May 1949 Mar. 1950 Jan. 1951 Nov. 1951 Sept. 1952 July 1953 May 1954 Dynamic Complexity: Tipping Points 700 600 Chickenpox in SK 500 400 300 200 100 0

Jan. 1921 Aug. 1921 Mar. 1922 Oct. 1922 May 1923 Dec. 1923 July 1924 Feb. 1925 Sept. 1925 Apr. 1926 Nov. 1926 June 1927 Jan. 1928 Aug. 1928 Mar. 1929 Oct. 1929 May 1930 Dec. 1930 July 1931 Feb. 1932 Sept. 1932 Apr. 1933 Nov. 1933 June 1934 Jan. 1935 Aug. 1935 Mar. 1936 Oct. 1936 May 1937 Dec. 1937 July 1938 Feb. 1939 Sept. 1939 Apr. 1940 Nov. 1940 June 1941 Jan. 1942 Aug. 1942 Mar. 1943 Oct. 1943 May 1944 Dec. 1944 July 1945 Feb. 1946 Sept. 1946 Apr. 1947 Nov. 1947 June 1948 Jan. 1949 Aug. 1949 Mar. 1950 Oct. 1950 May 1951 Dec. 1951 July 1952 Feb. 1953 Sept. 1953 Apr. 1954 Nov. 1954 June 1955 Jan. 1956 Aug. 1956 Dynamic Complexity: Tipping Points 3500 Measles (Red) & Mumps (Green) in SK 3000 2500 2000 1500 1000 500 0

Slides Adapted from External Source Redacted from Public PDF for Copyright Reasons

Nonlinearity (in state variables) Effect of multiple policies non-additive Doubling investment does not yield doubling of results Leads to Multiple basins of tracking (equilibrium)

Multiple Equilibria & Tipping Points Separate basins of attraction have qualitatively different behaviour Oscillations Endemic equilibrium Disease-free equilibrium

Equilibria Disease free No infectives in population Entire population is susceptible Endemic Steady-state equilibrium produced by spread of illness Assumption is often that children get exposed when young

Slides Adapted from External Source Redacted from Public PDF for Copyright Reasons

TB In SK

Jan. 1921 Nov. 1921 Sept. 1922 July 1923 May 1924 Mar. 1925 Jan. 1926 Nov. 1926 Sept. 1927 July 1928 May 1929 Mar. 1930 Jan. 1931 Nov. 1931 Sept. 1932 July 1933 May 1934 Mar. 1935 Jan. 1936 Nov. 1936 Sept. 1937 July 1938 May 1939 Mar. 1940 Jan. 1941 Nov. 1941 Sept. 1942 July 1943 May 1944 Mar. 1945 Jan. 1946 Nov. 1946 Sept. 1947 July 1948 May 1949 Mar. 1950 Jan. 1951 Nov. 1951 Sept. 1952 July 1953 May 1954 Mar. 1955 Jan. 1956 Nov. 1956 Sept. 1957 July 1958 May 1959 Mar. 1960 Jan. 1961 Nov. 1961 Dynamic Complexity: Tipping Points 700 600 Chickenpox in SK 500 400 300 200 100 0

Example: STIs

R 0 < 1 : 200 HC Workers, I 0 =1425

R 0 < 1 : 200 HC Workers, I 0 =1400

R 0 < 1 : 200 HC Workers, I 0 =1425

Kendrick-McKermack Model Partitioning the population into 3 broad categories: Susceptible (S) Infectious (I) Removed (R)

Contacts per Susceptible Per Contact Risk of Infection Force of Infection <Fractional Prevalence> Mean Time with Disease Births Susceptible Incidence Infectives Recovery Recovered <Mortality Rate> <Mortality Rate> Initial Population Population Initial Fraction Vaccinated Fractional Prevalence

Shorthand for Key Quantities for Infectious Disease Models: Stocks I (or Y): Total number of infectives in population This could be just one stock, or the sum of many stocks in the model (e.g. the sum of separate stocks for asymptomatic infectives and symptomatic infectives) N: Total size of population This will typically be the sum of all the stocks of people S (or X): Number of susceptible individuals

Basic Model Structure Population Size Immigration Rate Contacts per Susceptible Per Contact Risk of Infection Fractional Prevalence Average Duration of Infectiousness Immigration Susceptible Incidence Infective Recovery Recovered

Mathematical Notation Immigration M Rate Per Contact Risk of Contacts per Infection Fractional Susceptible c Prevalence Population Size N Mean Time with Disease Immigration of Susceptibles Susceptible Incidence Absolute Prevalence S I R Recovery Recovered

Underlying Equations I S M c S N I I c S N I R I

Our model : Set c=10 (people/month) =0.04 (4% chance of transmission per S-I contact) μ=10 Birth and death rate=0 Initial infectives=1, other 999 susceptible

Key Quantities for Infectious Disease Models: Parameters Contacts per susceptible per unit time: c e.g. 20 contacts per month This is the number of contacts a given susceptible will have with anyone Per-infective-with-susceptible-contact transmission probability: This is the per-contact likelihood that the pathogen will be transmitted from an infective to a susceptible with whom they come into a single contact.

Intuition Behind Common Terms I/N: The Fraction of population members (or, by assumption, contacts!) that are infective Important: Simplest models assume that this is also the fraction of a given susceptible s contacts that are infective! Many sophisticated models relax this assumption c(i/n): Average number of infectives that come into contact with a susceptible in a given unit time c(i/n) : Force of infection : (Approx.) likelihood a given susceptible will be infected per unit time The idea is that if a given susceptible comes into contact with c(i/n) infectives per unit time, and if each such contact gives likelihood of transmission of infection, then that susceptible has roughly a total likelihood of c(i/n) of getting infected per unit time (e.g. month)

Key Term: Flow Rate of New Infections This is the key form of the equation in many infectious disease models Total # of susceptibles infected per unit time # of Susceptibles * Likelihood a given susceptible will be infected per unit time = S*( Force of Infection ) =S(c(I/N) ) Note that this is a term that multiplies both S and I! This non-linear term is much different than the purely linear terms on which we have previously focused Likelihood is actually a likelihood density (e.g. can be >1 indicating that mean time to infection is <1)

Another Useful View of this Flow Recall: Total # of susceptibles infected per unit time = # of Susceptibles * Likelihood a given susceptible will be infected per unit time = S*( Force of Infection ) = S(c(I/N) ) The above can also be phrased as the following: S(c(I/N) )=I(c(S/N) )=I(c*f* )= # of Infectives * Mean # susceptibles infected per unit time by each infective This implies that as # of susceptibles falls=># of susceptibles surrounding each infective falls=>the rate of new infections falls ( Less fuel for the fire leads to a reduced burning rate)

A Critical Throttle on Infection Spread: Fraction Susceptible (f) The fraction susceptible (here, S/N) is a key quantity limiting the spread of infection in a population Recognizing its importance, we give this name f to the fraction of the population that issusceptible

New Recoveries Associated Feedbacks Contacts between Susceptibles and Infectives Susceptibles Infectives New Infections - -

Example Dynamics of SIR Model (No Births or Deaths) 2,000 people 600 people 10,000 people SIR Example 1,500 people 450 people 9,500 people 1,000 people 300 people 9,000 people 500 people 150 people 8,500 people 0 people 0 people 8,000 people 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 Time (days) Susceptible Population S : SIR example Infectious Population I : SIR example Recovered Population R : SIR example people people people

Explaining the Stock & Flow Dynamics: Infectives & Susceptibles Initially Each infective infects c(s/n) c people on average for each time unit the maximum possible rate The rate of recoveries is 0 In short term # Infectives grows (quickly)=> rate of infection rises quickly (Positive feedback!) Susceptibles starts to decline, but still high enough that each infective is surrounded overwhelmingly by susceptibles, so efficient at transmitting Over time, more infectives, and fewer Susceptibles Fewer S around each I =>Rate of infections per I declines Many infectives start recovering => slower rise to I Tipping point : # of infectives plateaus Aggregate Level: Rate of infections = Rate of recoveries Individual Level: Each infective infects exactly one replacement before recovering In longer term, declining # of infectives & susceptibles => Lower & lower rate of new infections! Change in I dominated by recoveries => goal seeking to 0 (negative feedback!)

Key Points Minimum value of stock of infectives occurs at different time than minimum of incidence! # of Infectives continues to decrease even after incidence is rising, as long as the rate of recoveries is higher than rate of infection Maximum value of stock of infectives occurs at different time than maximum of incidence! Maximum of incidence depends on both susceptible count and force of infection Stock of infectives will keep rising as long as incidence is greater than the recovery rate

Case 1: Outbreak 2,000 people 600 people 10,000 people 1,500 people 450 people 9,500 people Contacts between Susceptibles and Infectives Susceptibles Infectives - New Infections - SIR Example 1,000 people 300 people 9,000 people 500 people 150 people 8,500 people 0 people 0 people 8,000 people Contacts between Susceptibles and Infectives Susceptibles Infectives New Recoveries New Infections - - New Recoveries Contacts between Susceptibles and Infectives Susceptibles Infectives New Recoveries 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 Time (days) New Infections - - Susceptible Population S : SIR example Infectious Population I : SIR example Recovered Population R : SIR example people people people

Shifting Feedback Dominance 2,000 people 600 people 10,000 people 1,500 people 450 people 9,500 people Contacts between Susceptibles and Infectives Susceptibles Infectives - New Infections - SIR Example 1,000 people 300 people 9,000 people 500 people 150 people 8,500 people 0 people 0 people 8,000 people Contacts between Susceptibles and Infectives Susceptibles Infectives New Recoveries New Infections - - New Recoveries Contacts between Susceptibles and Infectives Susceptibles Infectives New Recoveries 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200 Time (days) New Infections - - Susceptible Population S : SIR example Infectious Population I : SIR example Recovered Population R : SIR example people people people