{
  "id": "person-time",
  "title": "Person-Time",
  "lang": "en",
  "source_lang": "en",
  "summary": "Person-time is the total time individuals contribute while under observation and at risk, used as the denominator for rates of disease, death, or other events.",
  "category": "health",
  "subcategory": "public-health",
  "type": "concept",
  "aliases": [
    "Person time",
    "person-time at risk",
    "population-time"
  ],
  "generated_at": "2026-11-04",
  "model": "pro/gpt-6.1-sol",
  "url": "https://aiwiki.page/en/person-time/",
  "translations": {
    "zh-Hans": {
      "url": "https://aiwiki.page/zh/person-time/",
      "json": "https://aiwiki.page/zh/person-time.json"
    }
  },
  "sections": [
    {
      "heading": "",
      "text": "Person-time is the sum of the time that individuals spend under observation and at risk of a specified event. In epidemiology, it provides a denominator for measuring how frequently new events occur when people are followed for different lengths of time. Common units include person-years, person-months, and person-days. Unlike a count of participants, person-time incorporates both the number of people observed and the duration of their observation."
    },
    {
      "heading": "Definition and calculation",
      "text": "If individual \\(i\\) contributes an eligible observation time \\(t_i\\), total person-time is\n\n\\[PT=\\sum_{i=1}^{n}t_i.\\]\n\nOne person observed for one year contributes one person-year. Two people observed for six months each also contribute one person-year. Fractional contributions are added in the same way as whole years.\n\nFor example, a study of first diagnoses might record:\n\nParticipant | Time under observation and at risk | Reason observation ends\nA | 2.0 years | Study ends\nB | 1.0 year | Develops the condition\nC | 0.5 years | Lost to follow-up\nD | 1.5 years | Develops the condition\nTotal | 5.0 person-years | 2 new cases\n\nFor a first-event analysis, participants B and D stop contributing time when they develop the condition. Participant C contributes only the period during which their event status is observed.\n\nPerson-years measure accumulated observation, not elapsed calendar time or the average age of participants. A study lasting two calendar years can accumulate thousands of person-years if it follows enough people."
    },
    {
      "heading": "Use in incidence and mortality rates",
      "text": "An incidence rate, also called an incidence density, is calculated as\n\n\\[IR=\\frac{D}{PT},\\]\n\nwhere \\(D\\) is the number of new events and \\(PT\\) is the corresponding person-time at risk. It is one of the principal measures of incidence.\n\nIn the example above,\n\n\\[IR=\\frac{2}{5}=0.4\\text{ cases per person-year}.\\]\n\nThis can also be expressed as 40 cases per 100 person-years. The multiplier changes presentation, not the underlying rate.\n\nThe same structure is used for mortality rates, with deaths in the numerator and the relevant time alive and under observation in the denominator. In large populations, person-time is often approximated using population size and the length of the observation period. A midyear population multiplied by one year is a common approximation for annual population person-time.\n\nThe numerator and denominator must refer to the same population, period, event definition, and eligibility rules. Counting events occurring outside the time represented in the denominator produces an inconsistent rate."
    },
    {
      "heading": "Defining time at risk",
      "text": "Person-time is specific to the outcome and research question. Time counted for one analysis may be ineligible for another.\n\nEntry and exit\n\nObservation can begin at recruitment, birth, treatment initiation, or another defined starting point. In a cohort study, participants may enter on different dates and leave at different times.\n\nFor a first occurrence of a condition, eligible follow-up usually ends at the earliest of:\n\n- The event of interest.\n- Death.\n- Loss to follow-up.\n- The end of the study.\n- Another prespecified endpoint that ends eligibility.\n\nPeople who already have the condition generally do not contribute time to an analysis of its first onset. They may nevertheless contribute to analyses of recurrence, complications, or death.\n\nCensoring\n\nCensoring occurs when the event time is not fully observed. For example, a participant who remains event-free when a study ends contributes their observed time, even though an event might occur later.\n\nAccounting for different observation lengths does not automatically remove bias from incomplete follow-up. If people who leave observation have a different event rate from those who remain, an unadjusted rate may not represent the intended population. Statistical analyses therefore require assumptions about censoring and, where necessary, methods to address its relationship with event risk.\n\nRecurrent events\n\nFor repeatable outcomes, such as episodes of an infectious disease, an individual may contribute more than one event. The denominator must specify whether time during an episode, recovery, or temporary ineligibility is included.\n\nAn event rate that counts repeated episodes differs from a first-event rate. Repeated events within the same individual may also be correlated, which affects statistical uncertainty."
    },
    {
      "heading": "Difference from risk and prevalence",
      "text": "Person-time rates are not proportions or probabilities. A rate has units of inverse time, whereas cumulative incidence, or risk, is the probability of experiencing an event within a specified period.\n\nA rate of 40 events per 100 person-years does not mean that 40% of participants necessarily experience an event each year. That interpretation depends on the event process, duration of follow-up, and whether repeat events are counted. Rates can exceed one event per person-year; probabilities cannot exceed one.\n\nUnder a constant event hazard \\(\\lambda\\), with no competing events, the risk by time \\(t\\) is\n\n\\[R(t)=1-e^{-\\lambda t}.\\]\n\nWhen \\(\\lambda t\\) is small, this is approximately\n\n\\[R(t)\\approx\\lambda t.\\]\n\nThe constant-hazard assumption is important: an average rate over follow-up does not generally establish the risk at every time point.\n\nPrevalence is different again. It measures the proportion of a population that has a condition at a specified time or during a specified period, rather than new events per unit of time at risk."
    },
    {
      "heading": "Comparing groups",
      "text": "Person-time is particularly useful when groups have unequal follow-up or when populations change through entry and exit. Applications include observational cohorts, clinical trials, and public health surveillance.\n\nFor two groups, an incidence rate ratio is\n\n\\[IRR=\\frac{D_1/PT_1}{D_0/PT_0}.\\]\n\nThis compares event rates rather than cumulative risks. Although both are relative measures of occurrence, a rate ratio is not generally interchangeable with a risk ratio.\n\nEqual amounts of person-time do not necessarily provide equivalent information. One hundred people followed for one year and ten people followed for ten years both contribute 100 person-years, but may differ substantially in age distribution, duration of exposure, and timing of events."
    },
    {
      "heading": "Time-varying characteristics",
      "text": "An individual's follow-up can be divided into intervals according to characteristics that change over time, such as age, treatment, or exposure status. Each interval contributes person-time to the appropriate category.\n\nFor example, a participant who starts a medication halfway through follow-up may contribute time to both an unexposed and an exposed category. Correct allocation matters: classifying all earlier follow-up as exposed merely because the person later received treatment can produce immortal time bias. This occurs when time during which a person must remain alive or event-free to qualify for a category is handled incorrectly.\n\nSplitting person-time by age or calendar period also permits comparison of rates within more comparable groups."
    },
    {
      "heading": "Statistical interpretation and limitations",
      "text": "A single rate compresses an event history into one numerator and one denominator. It does not show whether events were concentrated early or late in follow-up. Survival analysis can describe how event occurrence changes over time and account explicitly for the timing of events and censoring.\n\nPerson-time rates also require attention to:\n\n- Population differences: unequal age, exposure, or disease severity can make crude comparisons misleading.\n- Event ascertainment: missed events or inaccurate dates can distort the numerator or denominator.\n- Competing events: death may prevent a nonfatal outcome; its cause-specific rate alone does not determine the probability of that outcome.\n- Statistical uncertainty: confidence intervals depend strongly on the number of observed events, not just the amount of accumulated follow-up.\n\nUnlike a quality-adjusted life-year, person-time does not weight time by health status or quality of life. It measures eligible observation time, not health benefit."
    }
  ],
  "links": [
    "epidemiology",
    "incidence",
    "mortality-rate",
    "cohort-study",
    "infectious-disease",
    "cumulative-incidence",
    "prevalence",
    "clinical-trial",
    "public-health-surveillance",
    "relative-risk",
    "immortal-time-bias",
    "quality-adjusted-life-year"
  ],
  "sources": [],
  "license": "https://creativecommons.org/licenses/by/4.0/"
}