model: tool to understand infectious disease (ID) workings

reasons for modeling

• explore course of ID

• mechanisms that influence ID spread

• understand transmission and control

• predict impact of intervention

SIR model: Susceptible --> Infected --> Recovered

• assumes: not birth/death, lifelong immunity, no age, no seasons

  variations exist i.e. SI, SIS, SIRS

infection peak: people infected equals people recovered

R0: number of secondary infections per primary case

• in fully susceptible population

• R0 > 1 --> outbreak

• R0 < 1 --> no outbreak

R0 = secondary cases ÷ primary cases

• calculate secondary cases (cases caused by another infection)

• calculate primary cases (cases who caused/could have caused other infections)

R0 = β x c x D

β = p(transmission) per contact

• physical distancing, hand hygiene

c = number of contacts per time (unit)

• working from home, stopping large events

D = duration of infectiousness per time (unit)

• testing cases, quarantining contacts

no units: shows amount of people infected

doubling time: time required for number of new cases to double

• generation time: average time between two consecutive "generations" of infections

• number of new infectious caused by a case

minimum effect size an intervention must achieve to have a meaningful impact on the outcome

(1 - p) x R0 ≤ 1 OR

1 - (1 ÷ R0) ≤ p

proportion of individual in a population who need to be immune to an ID to prevent sustained transmission of the ID in the population

p represents the critical vaccination level

SIR model can be specified using R0

• I = number of infected people

• N = total population size

• δ or recovery rate = 1 ÷ D

Rt: number of secondary infections per primary case when not all population is susceptible due to

• immunity post recovery

• immunity from control measures

Rt < 1 --> effective control

during an epidemic, Rt declines over time

• Rt < 1 --> pre-peak, expanding cases

• Rt = 0 --> peak, every recovery = an infection

• Rt > 1 --> post-peak, declining cases

R0: basic reproductive number

• emergence of infections

• epidemic prevention

Rt: effective reproductive number

• insights into current state of epidemic

• effect of control measures

difficult/impossible to eliminate ID

1. asymptomatic transmission

2. high global mobility

3. non-human reservoirs

economic and societal damage

1. interventions

societal response

1. uptake of interventions

ethical dilemmas

1. costs of protecting

2. who to protect

prioritizing relevant health intervention from importance

• different diseases and injuries -->

• but limited resources

incidence. prevalence, mortality --> not enough for prioritization

DALY: disability adjusted life year (since 1993)

• allows measuring of relative importance

• morbidity and mortality in one index

• time (metric), healthy time (unit)

• 0 (full health) --> 1 (death)

• summarizes info from population health into a single number

  variations exist i.e. QALY, HALE

DALY = YLL (life lost) + YLD (life lost to disability)

YLL = Σ (d x e)

• d = sum of all fatal cases

• e = remaining life expectancy (at age of death)

YLD = Σ(n x t x dw)

• n = number of cases

• t = duration of disease

• dw = disability weight

GBD to find what prevent long, healthy life

• global, national, regional levels

GBD: measure of disability and disease from different causes

• annual updates from 1990-2019

• measures

  • death, incidence, prevalence

  • YLL, YLD, DALY

classification of injuries

• force of nature, war, legal intervention

• unintentional

  • traffic

  • non-traffic

• intentional

  • self harm

  • interpersonal violence

burden of injury shows how to assess proper measures

important cause for morbidity and mortality

• patterns across cause, age, sex, region, time

• most common for NL: falls (usually elderly)

risk factor: attribute or exposure associated with disease occurrence

• modifiable factors: diet, health behavior, smoking

• non-modifiable factors: genetics

causation: needs epidemiological evidence, not just observation

• how do we determine causes of a disease?

• how important is a particular cause for a disease

• what study designs are needed to understand this?

requirements for causal inference

• temporality - exposure before disease

• plausibility - biological plausibility

• consistency - replication

• strength of association - large association is better

• dose-response relationship - more exposure causes more disease

• reversibility - removing risk factor reduces risk

elements for explanatory research question

• what do we want to know from the causal relationship?

• what is the exposure

• what is the outcome of interest

• during what time period?

finds incidence

C1 (people without disease) --> C2 (did people develop disease?)

analytical study to identify potential causes

• risk factor: risk is higher for those exposed

• protective factor: risk is higher for those non-exposed

risk difference = I(exposed) - I(non exposed)

relative risk/risk ratio = I(exposed) ÷ I(non exposed)

hazard ratio: is your risk of dying within a time higher if you are exposed

pros

• temporality

• see dynamic pattern of exposure and disease outcome

cons

• time consuming

• expensive

• needs large population

finds prevalence (in exposed or unexposed)

prevalence = P(exposed) ÷ P(unexposed)

pros

• quick

• finds association

• fixed and nonmodifiable factors

cons

• temporality

finds odds (ratio)

select cases and controls --> ask about past exposure

odds: chance of outcome occurring vs. outcome not occurring

• odd of case being exposed or not exposed

• odds of control being exposed or not exposed

• odds ratio = O(exposed) ÷ O(nonexposed)

pros

• quick

• rare diseases

• long induction period

cons

• recall bias

• selection bias

• does not show incidence or prevalence

finds correlation

data based on existing population level data

• unit of analysis: groups

pros

• easy

• used in large populations

• non-individual variables (air pollution, weather)

cons

• ecological fallacy: inferences made about individuals from population data

• confounding variables

• weak consistency/replication

how much can a risk factor be attributed to a disease?

attributable fraction (AF): proportion of rate of disease in those exposed (which is attributable to the exposure itself)

• same formula as risk difference

AF = (incidence in exposed - incidence in nonexposed) ÷ incidence in exposed

population attributable fraction (PAF): proportion of rate of disease among the whole population (which is attributable to the exposure)

PAF = (incidence in total pop. - incidence in nonexposed) ÷ incidence in total pop.

RR (relative risk) = (a ÷ (a + b)) ÷ (c ÷ (c + d))

• cohort studies

• cross sectional studies (prevalence ratio)

RD (risk difference) = (a ÷ (a + b)) - (c ÷ (c + d))

• cohort studies

OR (odds ratio) = (a x d) ÷ (b x c)

• cohort studies

• cross sectional studies

• case control studies

AF = (incidence in exposed - incidence in nonexposed) ÷ incidence in exposed

• patient level i.e. physician

PAF = (incidence in total pop. - incidence in nonexposed) ÷ incidence in total pop.

• population level i.e. public and global health

population strategy

• risk factors with modest increased risk, common in population

• more health benefits to focus on population instead of those at risk

individual strategy

• if high risk can be easy identified, an individual based strategy can be useful

prevention plan cycles: for the development, selection or adaptation of an intervention to prevent or control health problems

prevention: actions that prevent disease occurrence or limit the negative consequences of a disease

control: actions to prevent disease transmission using

1. mathematical definition

2. the epidemiological triad of an infectious disease

strategy

1. health promotion (not disease-specific) i.e. healthy lifestyles, healthy education

2. health protection (not disease-specific) i.e. protection from risk, policies, intervention

3. disease prevention (disease-specific) i.e. suicide prevention day

population approach depends on population attributable fraction

• high risk approach--> protect risk groups

• general approach --> protect whole population

• opportunistic approach --> reach out to individuals to find risks

• collective approach --> reach out to a collected to find risks

stage of disease

• primordial/precautionary: no disease or direct risk factor

  • target: system

  • modifiable risk factor

• primary: normal, no disease

  • target: risk factor

  • modifiable risk factor

• secondary/pre-clinical phase: biological onset --> first symptoms

  • target: disease

  • for this, a long preclinical phase to allow detection and treatment

• tertiary/clinical phase: after symptoms

  • target: disease consequences

  • if disease cannot be cured anymore

healthcare

• national government, municipalities covers:

  • universal healthcare (everyone)

  • selective healthcare (at risk groups i.e. inactive)

• health insurance (basic, supplement) covers:

  • indicated (people with early symptoms i.e. pre-diabetes)

  • healthcare related (chronically ill i.e. diabetes)

• NL: Public Health Act, Health Insurance Act are government based

  • population approach

  • focus on primary prevention

• Environmental Act: promotion of healthy environments

  i.e. nature, water, housing, infrastructure, pH of water

assess burden

• measure the health status of a population: what is the effect of a disease on society?

  African American women (AAW) are more likely to die of breast cancer than caucasian women, but they have a lower morbidity.

identify causes

• develop an aim: what are the causes and reasons of the disease?

  AAW are less likely to participate in adequate screening and have more no shows. they also have underlying barriers: fear (of partner abandonment, outcome), belief (faith in god), inability (competing demands).

measure effectiveness (see next)

determine efficiency

• prevention plans vs. change in health status

• results achieved vs. resources expended

• based on existing data vs. based on new data

• check confounding variables

  Are there existing programs to affect this aim? i.e. phone calls for reminder for screening, questions about important barriers.

implement intervention

• use of intervention on a large scale

• targets for health promotion

• need personnel, equipment

monitor and feedback

• monitor diagnosis levels, mortality, change in burden

screening: strategy used in a population to identify an unrecognized disease in individuals without symptoms

• pre-symptomatic, asymptomatic, unrecognizable symptomatic disease

aim: improve life expectancy/quality of life in population

• early detection, diagnosis, treatment

screening has high specificity (limits harms on health)

• population level

• asymptomatic

clinical care has high sensitivity (detects a disease)

• individual level

• symptomatic

Wilson and Jungner, 1968 --> updated in 2008

evidence for effectiveness

• lead time bias: screening might give give the diagnosis earlier but does not change treatment and QOL/death

• length time bias: disease with long latency has more chance of being screened

  i.e. cancers that grow slow will be detected better

• proof for effectiveness: randomized controlled trials

  • outcome needs to be disease specific mortality

weighing benefits vs harms

• physical, psychological, social harm

cost effectiveness

• scarce resources: we need efficient interventions for health maximization

• costs

  • inviting individuals

  • screening hosts

  • maintenance of programs

  • long term care costs for positive cases (TP and FP)

• use of QALY's or the cost of QALY (total screening cost ÷ QALYs gained)

2 types

1. organized/population/mass screening

   • high investments (inviting, registration, quality control)

   • high quality (high coverage, less FP, use of guidelines)

2. opportunistic screening

   • case finding

screening test: diagnostic test applied in screening context

• 100% sensitivity and specificity --> perfect

• useless if p(positive test) is same for cases and non-cases

• requirements: safe, affordable, acceptable

sensitivity: how good is a test at identifying people with the disease correctly?

• finds TP

• ability of test to detect disease

• independent of disease prevalence

specificity: how good is a test at identifying people without disease correctly?

• finds TN

• ability of test to distinguish non-disease people

• more important for rare diseases

• independent of disease prevalence

positive predictive value: probability a positive test result is correct

• dependent on disease prevalence

negative predictive value: probability a negative test result is correct

• dependent on disease prevalence

informativeness of study

• prior/posterior knowledge (consistency) + new knowledge (accuracy)

• to find new level of knowledge

hypothesis free studies: low prior knowledge, low p-values

i.e. genome wide association studies, proteomics

equipose: probability a new treatment is more effective than control should be 50%

• otherwise it is unethical

• not everything can have a clinical trial

accuracy or validity: correctness of the study

• more important than precision

precision: reproducibility of study

• consistency, reliability

external validity

• generalizability (are these results applicable in other populations?)

internal validity

• selection bias: sample is not representative of whole population

• information bias: measurement error

  • non differential (random): error in exposure, outcome, confounder

  • differential (nonrandom): error in outcome determinant

  • differential (misclassification): recall bias, non-blinding

  • change in effect (dilution, overestimation) --> use blinding

• confounding bias: shared cause between exposure and outcome

selection bias --> use of clinical criteria and randomization

information bias --> blinding (single blind, double blind, triple blind)

confounding bias --> randomization, stratification, multivariable modeling