Destination probabilities under constant competing hazards

Splits the probability of leaving a state within an interval between two absorbing destinations with constant cause-specific hazards h_1 and h_2. The total exit probability depends on the summed hazard, and each destination receives its share h_1 or h_2 of that sum.

Signature

p_1 = h_1 / (h_1 + h_2) * (1 - exp(-(h_1 + h_2) * Delta)); p_2 = h_2 / (h_1 + h_2) * (1 - exp(-(h_1 + h_2) * Delta)); p_stay = exp(-(h_1 + h_2) * Delta)
Inputs
InputsDefinitionUnit
h_1Constant cause-specific hazard of moving to destination 1events per person per unit of time
h_2Constant cause-specific hazard of moving to destination 2events per person per unit of time
DeltaLength of the model cycle, in the same time unit as the hazardstime, for example years
Output
p_1Probability of leaving the starting state for destination 1 within the intervalprobability from 0 to 1
p_2Probability of leaving the starting state for destination 2 within the intervalprobability from 0 to 1
p_stayProbability of still being in the starting state at the end of the intervalprobability from 0 to 1

Function

Rate and transition probability conversion function

Maps an event rate and a time interval to the probability that the event occurs within that interval, and back again. Rates are converted rather than probabilities divided, because a probability is tied to the length of the interval over which it was observed. The converted probability then enters a transition matrix used in the cohort update s_(t+1) = s_t P described on the Markov Model page.

Implementations

  • Excel

    Competing exit probabilities in three cells

    With named cells Hazard1, Hazard2 and CycleLength, the three formulas return p_1, p_2 and p_stay respectively.

    =Hazard1/(Hazard1+Hazard2)*(1-EXP(-(Hazard1+Hazard2)*CycleLength)); =Hazard2/(Hazard1+Hazard2)*(1-EXP(-(Hazard1+Hazard2)*CycleLength)); =EXP(-(Hazard1+Hazard2)*CycleLength)

Assumptions

  • Constant cause-specific hazards

    Both hazards are constant within the interval and their sum is greater than zero.

  • Absorbing destinations within the cycle

    A person who reaches destination 1 or 2 does not move again within the same cycle. If a further transition is possible from a destination, for example from ill to dead within one cycle, the formulas are no longer valid and the full matrix is derived from the transition rate matrix.

Worked examples

  • Two exits over a three-month cycle

    With hazards of 0.12 and 0.08 per year and a cycle of 0.25 years, the destination probabilities are about 0.02926 and 0.01951 and the probability of remaining is about 0.95123. The unrounded values sum to 1 and match the article's table.

    h_1 = 0.12; h_2 = 0.08; Delta = 0.25; p_1 = 0.02926; p_2 = 0.01951; p_stay = 0.95123

Common errors

  • Converting each competing rate with the single-event formula

    Using p equal to 1 minus exp of minus h_k times Delta for each exit ignores that people who leave by one route are no longer at risk of the other. With both hazards equal to 1 per cycle, each exit would receive about 0.632 and the probability of remaining would be about minus 0.264, which is impossible. The joint formula gives about 0.432 for each exit and about 0.135 for remaining.

Sources

  • Competing-risk rate to probability conversion

    Jones E, Epstein D, García-Mochón L. A procedure for deriving formulas to convert transition rates to probabilities for multistate Markov models. Medical Decision Making. 2017;37(7):779-789. Introduction, which shows that the simple formula is always wrong with competing risks, and the special case with two competing absorbing exits.

    View source →

Canonical Identity

Stable URI · Machine-readable · Resolvable · CC BY 4.0