Central Death Rates And Exposure

Why Exposure Matters

A death count alone does not tell us how frequently deaths occurred. Forty deaths over 2,000 person-years of observation represent a different mortality experience from 40 deaths over 20,000 person-years. A central death rate accounts for both the number of deaths and the amount of time people were exposed to the possibility of dying.

In a life table, the same idea connects the number dying between ages \(x\) and \(x+1\) to the total person-years lived in that interval. The key is to use the correct denominator: exposure includes time contributed by those who survive the whole interval and by those who die during it.

Definition: The central death rate for an age interval is the number of deaths in that interval divided by the total person-years of exposure in that interval. For an observed population, write \(m_x=D_x/\mathcal{E}_x\), where \(D_x\) is the death count and \(\mathcal{E}_x\) is the total exposure. In a life table, the corresponding rate is \(m_x=d_x/L_x\), where \(d_x=l_x-l_{x+1}\) and \(L_x\) is the life-table person-years lived between ages \(x\) and \(x+1\).

What Counts as Person-Years of Exposure?

One person observed for one full year contributes one person-year. Two people each observed for half a year also contribute one person-year in total. More generally, add the time each person is under observation and at risk during the stated interval. A person who dies during the interval contributes exposure up to the time of death, not a full year automatically; a person who survives to the end contributes the time observed in the interval.

For an actual population, \(\mathcal{E}_x\) is assembled from the observed time at risk for the people in the age band. For a life-table cohort, \(l_x\) people are alive at exact age \(x\), \(d_x\) die before age \(x+1\), and \(L_x\) is their combined person-years lived in that one-year interval. The life-table identity \(d_x=l_x-l_{x+1}\) gives the death count; \(L_x\) supplies the exposure denominator.

If the survivor count within the interval is represented by \(l_{x+t}\), the cohort exposure can also be described as the area under that survivor-count curve:

L_x=\int_0^1 l_{x+t}\,dt

This expression emphasizes that exposure accumulates continuously over the year: at each time \(t\), the people still alive contribute to the total. It does not by itself prescribe how to interpolate a life table between integer ages. A particular interpolation assumption is needed if exposure is to be inferred from integer-age entries alone.

Calculating and Interpreting the Rate

The central rate has units of deaths per person-year. It is often reported per 1,000 person-years for readability: multiply the rate per person-year by 1,000. Keep the time unit attached to the exposure. For example, deaths per person-month and deaths per person-year are not the same rate.

The definition is a ratio, not a probability. Its denominator is total time at risk, rather than simply the number alive at the start of the interval. In a life table, one begins with \(l_x\) and \(d_x\), then uses \(L_x\) for the exposure:

m_x=\frac{d_x}{L_x} =\frac{l_x-l_{x+1}}{L_x}

The formula follows directly from the meaning of its parts: \(d_x\) counts deaths in the interval, while \(L_x\) counts the corresponding person-years lived. Dividing deaths by person-years gives a rate. Replacing \(L_x\) with \(l_x\) would instead divide the deaths by the number alive at the start, changing the denominator and the measure.

Worked Example: Summing Individual Exposure

A hypothetical study follows 10 people who are all age 60 at the start of a one-year interval. Six survive to the end and each contributes one year of observation. The other four die after 0.25, 0.75, 0.50, and 0.50 years, respectively. Calculate the central death rate per person-year and per 1,000 person-years. Use the recorded times as exact; no parametric mortality distribution is assumed.

Method. Add each person’s actual time at risk, then divide the death count by this total exposure. This avoids assigning a full year of exposure to every person who dies during the interval.

The six survivors contribute \(6(1)=6\) person-years. The four people who die contribute \(0.25+0.75+0.50+0.50=2.00\) person-years. Therefore,

\mathcal{E}_{60}=6+2=8.00\text{ person-years}

There are \(D_{60}=4\) deaths. Thus the central death rate is

m_{60}=\frac{4}{8.00}=0.50 \text{ deaths per person-year}

Per 1,000 person-years, this is \(0.50(1{,}000)=500\) deaths. Check: the four deaths contribute two person-years, and the six survivors contribute six, for eight person-years total; \(4/8=0.50\). The answer is a rate, not a statement that an individual has a 50% probability of dying during the year.

Life-Table Entries and Central Rates

For a life table, the number dying in the age interval is obtained from the survivor column:

d_x=l_x-l_{x+1}

Once exposure \(L_x\) is available, substitute this death count into \(m_x=d_x/L_x\). The rate is tied to the specific age interval and the exposure measure used. Two intervals can have the same \(d_x\) but different \(m_x\) if their exposures differ.

The table’s \(L_x\) should not be assumed to equal \(l_x\). The latter counts lives at the beginning of the interval; the former counts person-years accumulated throughout it. If \(L_x\) is not supplied, it must be obtained under an explicitly stated model or exposure convention. Integer-age survivor and death counts alone do not specify exactly when within the year the deaths occur, so they do not determine the exact exposure without additional information.

Worked Example: Computing a Life-Table Central Rate

A hypothetical life table has \(l_{65}=1{,}200\), \(l_{66}=1{,}140\), and tabulated exposure \(L_{65}=1{,}170\) person-years. Find \(d_{65}\), the central death rate per person-year, and the rate per 1,000 person-years. Treat the tabulated exposure as given; no interpolation or parametric mortality distribution is being assumed.

Method. First use the life-table survivor identity to obtain deaths, then divide by the stated person-years. This avoids confusing the initial number of lives with exposure.

The death count is

d_{65}=l_{65}-l_{66} =1{,}200-1{,}140 =60

Therefore,

m_{65}=\frac{d_{65}}{L_{65}} =\frac{60}{1{,}170} =0.051282\ldots

Per 1,000 person-years, the rate is \(0.051282\ldots(1{,}000)=51.282\ldots\), or 51.28 deaths per 1,000 person-years. To four decimal places, the rate per person-year is \(0.0513\). Check: \(1{,}170\) person-years is the exposure denominator, and multiplying \(0.051282\ldots\) by \(1{,}170\) recovers 60 deaths.

Worked Example: Comparing Intervals With Equal Death Counts

Two consecutive age intervals in a hypothetical life table have the following entries: at age 50, \(l_{50}=5{,}000\), \(l_{51}=4{,}900\), and \(L_{50}=4{,}950\); at age 51, \(l_{51}=4{,}900\), \(l_{52}=4{,}800\), and \(L_{51}=4{,}870\). Compare their death counts and central rates. Use the tabulated exposures directly, with no assumed parametric distribution.

Method. Calculate each \(d_x\) from adjacent survivor entries and divide by its own \(L_x\). The common error to avoid is using one interval’s exposure for both rates.

At age 50,

d_{50}=5{,}000-4{,}900=100, \qquad m_{50}=\frac{100}{4{,}950}=0.020202\ldots

At age 51,

d_{51}=4{,}900-4{,}800=100, \qquad m_{51}=\frac{100}{4{,}870}=0.020534\ldots

Both intervals have 100 deaths, but the age-51 central rate is higher because its exposure is lower. To four decimal places, the rates are \(0.0202\) and \(0.0205\) deaths per person-year, respectively. Check: \(0.020202\ldots(4{,}950)=100\), and \(0.020534\ldots(4{,}870)=100\), as required.

A Common Denominator Trap

The quantity \(d_x/l_x\) divides deaths by the number alive at the start of the interval. It is not the central death rate, whose denominator is \(L_x\). For the age-65 data above, using \(d_{65}/l_{65}\) gives \(60/1{,}200=0.05\), while the central rate is \(60/1{,}170=0.051282\ldots\). The numerical difference may look modest, but the two calculations use different denominators and answer different questions.

A useful exam check is to write the units before calculating. A death count divided by lives at the start is deaths per starting life. A death count divided by person-years is deaths per person-year. If an answer is requested as a central rate and the denominator in the calculation is \(l_x\), stop and check whether exposure \(L_x\) was provided or must be found.

Also check that deaths and exposure refer to the same age interval and population. A numerator from one age band divided by exposure from another does not form a meaningful central rate. If the exposure period is given in months, convert it to person-years when the requested rate is annual; for example, 24 person-months equal 2 person-years.

Key takeaway: A central death rate is deaths divided by person-years of exposure. In a life table, find \(d_x=l_x-l_{x+1}\), use the matching exposure \(L_x\), and calculate \(m_x=d_x/L_x\). Do not substitute \(l_x\) for \(L_x\).

Check Your Understanding

For each hypothetical situation, identify the relevant exposure and calculate or interpret the requested central death rate.

In a one-year age interval, 20 people survive the full year and contribute 20 person-years. Three others die after 0.2, 0.4, and 0.8 years. Find total exposure and the central rate per person-year. (Answer: \(21.4\) person-years; \(3/21.4=0.1402\) deaths per person-year, to four decimals.)
A life table has \(l_{70}=800\), \(l_{71}=760\), and \(L_{70}=780\) person-years. Find \(d_{70}\) and \(m_{70}\). (Answer: \(d_{70}=40\); \(m_{70}=40/780=0.0513\) deaths per person-year, to four decimals.)
A hypothetical age band records 18 deaths over 900 person-years. Express its central death rate per person-year and per 1,000 person-years. (Answer: \(18/900=0.0200\) deaths per person-year, or 20 deaths per 1,000 person-years.)
Two intervals each have 30 deaths. Their exposures are 1,500 and 1,200 person-years. Calculate both rates and state which is higher. (Answer: \(0.0200\) and \(0.0250\) deaths per person-year; the interval with 1,200 person-years has the higher rate.)
A life-table calculation uses \(l_x=2{,}400\), \(l_{x+1}=2{,}280\), and \(L_x=2{,}340\) person-years. A candidate divides 120 by 2,400 and labels the result the central death rate. Identify the error and give the central rate per 1,000 person-years. (Answer: the candidate used \(l_x\) instead of exposure \(L_x\); \(d_x=120\), so \(m_x=120/2{,}340=0.05128\ldots\), or 51.28 deaths per 1,000 person-years.)

What You'll Learn