Crude Rates and Their Precision

What a crude death rate is made of and how far it can be trusted. Crude, age-specific and cause-specific rates, and the crude rate proved to be a population-weighted average of age-specific rates; the Poisson model for a count of deaths, the standard error of a rate, and exact gamma limits for small counts, with the Poisson-gamma identity and their coverage proved; the NCHS rule of 20 deaths; and the comparison of two rates by their difference and their ratio. Worked on 2021 deaths in Maine and Alaska and U.S. deaths by age from CDC WONDER.

In 2021 Maine recorded 1,258.4 deaths per 100,000 residents and Alaska 847.3. What is a crude death rate made of, and could the gap between these two be chance? A crude rate is an average of the death rates at each age, weighted by the population's own age structure, and the gap is not chance: the two rates differ by 28.6 standard errors. Yet the national death rates, applied to each state's ages, would by themselves give Maine a crude rate 57% above Alaska's. The comparison is statistically precise and still misleading.

This article continues Rates in Public Health, which separates rates from proportions and ratios, counts the person-time in a rate's denominator, and relates incidence to prevalence. Here the rate is a state's death rate. First we define crude and specific rates, and assemble a crude rate from the rates of the age groups within it. Then we measure how much random variation a rate based on a given number of deaths carries, with the Poisson model, a normal interval and exact limits for small counts. Finally we compare the two states. The adjustment that corrects the comparison is the subject of Direct Standardization.

The article uses three sources of data:

  • the final 2021 death counts published by the National Center for Health Statistics (NCHS) in Deaths: Final Data for 2021 (Murphy et al., 2024);
  • the Census Bureau's population estimates for July 1, 2021 by single year of age, which NCHS used as its denominators;
  • deaths of U.S. residents by age from CDC WONDER, the federal online query system for vital statistics.

Every number below is computed by a script in the site's repository. Where NCHS published the same figure, the script asserts that the two agree to the precision published.

Crude, specific and adjusted rates

Back to Maine and Alaska, whose crude death rates opened this article and the previous one. A crude rate lumps the whole population together, and what it hides is the population's make-up by age.

A crude rate is computed for a whole population: all deaths in Maine in 2021 over all of Maine's residents. A specific rate is computed for a subgroup.

  • An age-specific rate counts the deaths and the population of one age group.
  • A sex-specific rate counts those of one sex.
  • A cause-specific rate counts only deaths from one cause, with the whole population still at risk.

Rates can be specific for several characteristics at once, such as heart disease deaths among women aged 65 to 74. An adjusted (or standardized) rate is a summary constructed to remove the effect of differences in age or another variable, and is the subject of the next article.

Age-specific rates are the most informative, because age is the strongest single influence on almost every rate of death and disease. The table gives the U.S. figures for 2019, from CDC WONDER. NCHS publishes the same rates (Murphy et al., 2024, Table 6).

Age Deaths Population Rate per 100,000 Population share Share × rate
Under 1 20,921 3,783,052 553.0 0.0115 6.4
1–4 3,676 15,793,631 23.3 0.0481 1.1
5–14 5,497 40,994,163 13.4 0.1249 1.7
15–24 29,771 42,687,510 69.7 0.1300 9.1
25–34 59,178 45,940,321 128.8 0.1400 18.0
35–44 82,986 41,659,144 199.2 0.1269 25.3
45–54 160,393 40,874,902 392.4 0.1245 48.9
55–64 374,937 42,448,537 883.3 0.1293 114.2
65–74 555,559 31,483,433 1,764.6 0.0959 169.3
75–84 688,027 15,969,872 4,308.3 0.0487 209.6
85 and over 873,746 6,604,958 13,228.6 0.0201 266.2
All ages 2,854,838 328,239,523 869.7 1 869.7

The rates fall from the first year of life to a minimum at ages 5 to 14. Then they climb steadily, roughly doubling or more with each decade of age from 35 on: each rate from ages 45 to 54 up is between 1.97 and 3.07 times the rate a decade younger. The rate at 85 and over is 987 times the rate at 5 to 14. (The all-ages deaths include 147 whose age was not stated.)

The crude rate as a weighted average

The last column of the table hints at how the crude rate is built from the specific ones. Here is the idea in one sentence: the crude rate is an average of the age-specific rates, and each age group counts in proportion to its share of the population.

Let a population be divided into $k$ age groups, with $d_i$ deaths and population $n_i$ in group $i$. Write $r_i = d_i/n_i$ for the age-specific rates, $D = \sum d_i$ for the total deaths and $N = \sum n_i$ for the total population. Then

$$R_{\text{crude}} = \frac{D}{N} = \frac{\sum_i n_i r_i}{N} = \sum_i \frac{n_i}{N}\, r_i.$$

The crude rate is a weighted average of the age-specific rates. The weights are the population's own age distribution, the shares $n_i/N$. This is the grouped-data mean of Standard Scores, Chebyshev's Inequality and Grouped Data, with the age groups as the classes.

As a check, the last column of the table shows the products $(n_i/N)\,r_i$. They add up to 869.70, the crude rate, to rounding; the 147 deaths of unknown age add 0.04.

So two populations can differ in their crude rates for two separate reasons. Their age-specific rates may differ, or their weights, their age structures, may differ. The crude rate cannot tell which.

Top: U.S. death rates by age in 2019, on a log scale; the rate at 85 and over is nearly a thousand times the rate at 5 to 14. Bottom: the age structures of Maine and Alaska in 2021. Notice that Maine has more of its population in the age groups from 55 up, where the rates in the top panel are highest, and Alaska more in every group under 45.
Top: U.S. death rates by age in 2019, on a log scale; the rate at 85 and over is nearly a thousand times the rate at 5 to 14. Bottom: the age structures of Maine and Alaska in 2021. Notice that Maine has more of its population in the age groups from 55 up, where the rates in the top panel are highest, and Alaska more in every group under 45.

Maine and Alaska differ sharply in the weights. Maine had the oldest population of the 50 states and the District of Columbia in 2021, measured by the share aged 65 and over (21.7%). Half of its residents were 44 or older. Alaska ranked 48th of the 51 (13.3%), with half its residents aged 35 or younger. And 20.5% of Alaskans were under 15, against 14.9% of Maine's residents.

Example (the same rates at every age, different crude rates). If the two states had exactly the same age-specific death rates, those of the whole country in 2021 (Murphy et al., 2024, Table 2), their crude rates would still not be equal. Applying those national rates to each state's population gives crude rates of 1,257.9 per 100,000 for Maine and 799.0 for Alaska, a ratio of 1.57, produced by age structure alone.

Now compare the actual crude rates with these predictions: Maine's, 1,258.4, is almost exactly what the national rates predict for a population of its ages. Alaska's, 847.3, is 6% above its prediction. That calculation is the idea behind indirect standardization, developed with confidence intervals in Indirect Standardization and the Standardized Mortality Ratio.

The precision of a rate

Before adjusting for age, there is one more matter to settle: whether the gap between the two states, or some part of it, could be due to chance. It could not, as the comparison at the end of this article shows, but showing it needs a measure of the random variation in a rate.

A death rate computed from a complete count of deaths is not subject to sampling error: no sample was drawn. It is still subject to random variation. The number of deaths that occurred is one outcome of a process that could have produced a few more or a few fewer under the same conditions. A comparison between two places, or two years, should not read meaning into differences of that size. NCHS takes this view (Murphy et al., 2024, Technical Notes, "Random variation").

The Poisson model

The usual model treats the number of deaths $D$ as a Poisson random variable with mean $\mu = \lambda N$. Here $\lambda$ is the underlying rate and $N$ the population (or person-years), treated as a fixed constant. The distribution itself is the subject of The Poisson Distribution.

The one property we need is that a Poisson count has variance equal to its mean, which The Poisson Distribution proves. The mean is estimated by $D$ itself, so

$$\operatorname{SE}(D) = \sqrt{D}, \qquad \operatorname{SE}(R) = \frac{\sqrt{D}}{N} = \frac{R}{\sqrt{D}}, \qquad \operatorname{RSE}(R) = \frac{\operatorname{SE}(R)}{R} = \frac{1}{\sqrt{D}},$$

where $R = D/N$ is the rate and RSE is the relative standard error. The second formula divides by the constant $N$: dividing a variable by a constant divides its standard deviation by the same constant, and $\sqrt{D}/N = (D/N)/\sqrt{D}$.

The Poisson distribution is a good model when each person has a small probability of the event and people die independently of each other. Anderson and Rosenberg (1998, Technical notes) note that it gives "reasonable, conservative" variances for an open population. The binomial model, which assumes a fixed group of people each of whom either dies or does not, tends to understate the variance.

Example (precision depends on the deaths, not the population). A city of 10 million and a town of 10,000 each record 100 deaths from some cause, and neither rate is the more precise relative to its size. The relative standard error depends only on the number of deaths, not on the population: $1/\sqrt{100} = 10\%$ in both places. The city's rate is a thousand times smaller, and so is its standard error.

For Maine in 2021, $D = 17{,}269$ and $R = 1{,}258.45$ per 100,000, so

$$\operatorname{SE}(R) = \frac{1{,}258.45}{\sqrt{17{,}269}} = 9.58, \qquad \operatorname{RSE} = 0.76\%.$$

The normal 95% confidence interval is $1{,}258.45 \pm 1.96 \times 9.58 = (1{,}239.7,\ 1{,}277.2)$ per 100,000. With seventeen thousand deaths, random variation is a small fraction of the rate.

Example (a common misstatement). The same interval is often written $\text{Rate} \pm 1.96\sqrt{\text{Rate}/\text{Population}}$, to be multiplied by the base (100,000 here) at the end. It is the interval above, since $\sqrt{R/N} = \sqrt{D}/N = R/\sqrt{D}$, but only if the rate under the square root is per person. For Maine, $R = 17{,}269/1{,}372{,}247 = 0.0125845$ per person, and $\sqrt{R/N} \times 100{,}000 = 9.58$, the standard error above. The misstatement is to put the rate per 100,000 under the square root: $\sqrt{1{,}258.45/1{,}372{,}247} = 0.0303$, which is no standard error of anything. The form $R/\sqrt{D}$ has no such trap, because it gives the standard error in whatever units $R$ is in.

Exact limits for small counts

What happens when a rate rests on a handful of deaths? The normal approximation fails, as it did for small binomial samples in Confidence Intervals for a Proportion. The Poisson distribution is skewed, and the symmetric interval $D \pm 1.96\sqrt{D}$ can even extend below zero.

Exact limits for the Poisson mean are built the way the Clopper–Pearson limits are built for a proportion in Confidence Intervals for a Proportion. Here is the idea. Having observed $d$ deaths, the lower limit is the mean so small that a count of $d$ or more would have probability only 0.025, and the upper limit is the mean so large that a count of $d$ or fewer would have probability only 0.025. Means outside those limits are the ones under which the observed count would be in an extreme tail.

The formal version. For an observed count $d$, the lower limit $L$ (for $d \ge 1$) and the upper limit $U$ (for every $d \ge 0$) solve

$$P(D \ge d \mid \mu = L) = 0.025, \qquad P(D \le d \mid \mu = U) = 0.025.$$

(For $d = 0$ the lower limit is 0.) These equations have a closed form in terms of the gamma distribution. A gamma variable $G_d$ with shape $d$, a positive whole number here, and scale 1 has density $x^{d-1}e^{-x}/(d-1)!$ for $x > 0$, and the Poisson is linked to it by the identity

$$P(D \ge d \mid \mu) = P(G_d \le \mu).$$

Proof of the identity. Both sides are functions of $\mu$ that equal 0 at $\mu = 0$. The right side is $\int_0^\mu x^{d-1}e^{-x}/(d-1)!\,dx$, with derivative $\mu^{d-1}e^{-\mu}/(d-1)!$. The left side is $1 - \sum_{j=0}^{d-1} e^{-\mu}\mu^j/j!$. Differentiating each term by the product rule,

$$\frac{d}{d\mu}\left(1 - \sum_{j=0}^{d-1} \frac{e^{-\mu}\mu^j}{j!}\right) = \sum_{j=0}^{d-1} \frac{e^{-\mu}\mu^j}{j!} - \sum_{j=1}^{d-1} \frac{e^{-\mu}\mu^{j-1}}{(j-1)!}.$$

The second sum is the first without its last term, $j = d - 1$, so the difference is $e^{-\mu}\mu^{d-1}/(d-1)!$. The two sides have the same derivative and the same value at 0, so they are equal. $\square$

The identity turns both equations into gamma percentiles. The first says $P(G_d \le L) = 0.025$. The second says $1 - P(D \ge d + 1 \mid \mu = U) = 0.025$, that is, $P(G_{d+1} \le U) = 0.975$. So the 95% limits for $\mu$, given the observed count $D$, are

$$L(D) = G^{-1}_{D}(0.025), \qquad U(D) = G^{-1}_{D+1}(0.975).$$

In words, they are the 2.5th percentile of the gamma distribution with shape $D$ and the 97.5th percentile of the one with shape $D + 1$. The limits for the rate are $L(D)/N$ and $U(D)/N$.

The identity also shows that $P(D \ge d \mid \mu)$ increases with $\mu$, since the right side does. That is what guarantees the coverage. Fix the true mean $\mu$. The interval lies entirely below $\mu$, that is, $U(d) < \mu$, exactly for the counts $d$ with $P(D \le d \mid \mu) < 0.025$. The reason is that $P(D \le d \mid \mu') = 1 - P(D \ge d + 1 \mid \mu')$ decreases as $\mu'$ grows and equals 0.025 at $\mu' = U(d)$. Those counts form a lower tail $\{0, 1, \ldots, d^*\}$, whose total probability $P(D \le d^* \mid \mu)$ is below 0.025 by the definition of $d^*$. The interval lies entirely above $\mu$, $L(d) > \mu$, exactly for the counts with $P(D \ge d \mid \mu) < 0.025$, because $P(D \ge d \mid \mu')$ increases with $\mu'$ and equals 0.025 at $\mu' = L(d)$. Those counts form an upper tail, again of total probability below 0.025. (A count of 0 is never among them, since its lower limit is 0.) So the interval misses with probability below 0.05, and covers every $\mu$ with probability above 0.95.

The gamma-percentile limits are the formulas NCHS gives, in spreadsheet form, for counts below 100 (Murphy et al., 2024, Technical Notes). They are also the exact Poisson limits that Breslow and Day (1987, §2.3) tabulate. The script reproduces NCHS's table of limit factors $L(D)/D$ and $U(D)/D$ to all six published decimals at the counts it checked (1, 7, 20 and 50 deaths; Table XI).

Suppose a town of 25,000 records 7 deaths from some cause in a year, a rate of 28.0 per 100,000.

  • The normal interval is $28.0 \pm 1.96 \times 28.0/\sqrt{7} = (7.3,\ 48.7)$.
  • The exact limits for the count are 2.814 and 14.423. So the exact interval for the rate is (11.3, 57.7) per 100,000.

The exact interval is shifted upward and asymmetric about 28.0, as a skewed distribution requires.

Which interval keeps its promise of 95%? Only the exact one, as the argument above guarantees and the following computation shows. Because the count takes whole-number values, the coverage of each interval can be computed exactly for any true mean $\mu$. We add the Poisson probabilities of the counts whose intervals contain $\mu$, as the proportions unit did for the binomial.

Exact coverage of two 95% intervals for a Poisson mean, against the expected count. Notice how far the normal interval (orange) falls short for small counts: to 0.63 when one death is expected and 0.87 at five. The exact interval (blue) never falls below 0.95. Both saw-tooth because the count is discrete. Above about 100 expected deaths (dotted line), the two are close.
Exact coverage of two 95% intervals for a Poisson mean, against the expected count. Notice how far the normal interval (orange) falls short for small counts: to 0.63 when one death is expected and 0.87 at five. The exact interval (blue) never falls below 0.95. Both saw-tooth because the count is discrete. Above about 100 expected deaths (dotted line), the two are close.

For expected counts between 1 and 10, the normal interval's coverage fell as low as 0.632 and averaged 0.889 (averages are over a grid of expected counts 0.25 apart). At $\mu = 5$ it was 0.870, against 0.980 for the exact interval. Between 20 and 100 expected deaths, the normal interval's coverage averaged 0.945, with a low of 0.929. The exact interval's coverage never fell below 0.950 anywhere from 1 to 150. NCHS uses the normal approximation for counts of 100 or more and the gamma limits below that.

Too few deaths to publish

NCHS replaces a crude or age-specific death rate with an asterisk when it is based on fewer than 20 deaths. It describes that threshold as "the equivalent of an RSE of 23% or more" (Murphy et al., 2024, Technical Notes). The equivalence is the formula above: $1/\sqrt{20} = 22.4\%$ and $1/\sqrt{19} = 22.9\%$. So 19 deaths is the largest count whose relative standard error rounds to 23% or more. At 100 deaths the RSE is 10%. A rate based on a handful of deaths can move by half its value from one year to the next without any change in the underlying risk.

Comparing two rates

Now the two states can be compared formally. When both rates rest on 100 or more deaths, NCHS tests their difference with

$$z = \frac{R_1 - R_2}{\sqrt{\operatorname{SE}(R_1)^2 + \operatorname{SE}(R_2)^2}}.$$

The denominator is the standard error of the difference: the two states' counts are independent, and the variance of a difference of independent quantities is the sum of their variances. This is the same construction as the unpooled comparison of two means or two proportions (Murphy et al., 2024, Technical Notes, formula 12). For Maine and Alaska, $\operatorname{SE}(R_1) = 9.58$ and $\operatorname{SE}(R_2) = 847.3/\sqrt{6{,}208} = 10.75$, so

$$z = \frac{1{,}258.4 - 847.3}{\sqrt{9.58^2 + 10.75^2}} = 28.6.$$

The 95% interval for the difference is (382.9, 439.4) per 100,000.

The rate ratio $R_1/R_2 = 1.485$ is usually analyzed on the log scale, where its standard error has a simple approximate form. The approximation is the delta method, a first-order Taylor expansion, which Intervals and Adjustment for Relative Risks and Odds Ratios sets out in general. Here it runs as follows. The log of a rate is $\log R = \log D - \log N$, and $N$ is a constant. Near its mean $\mu$, $\log D \approx \log \mu + (D - \mu)/\mu$, so $\operatorname{Var}(\log D) \approx \operatorname{Var}(D)/\mu^2 = \mu/\mu^2 = 1/\mu$, estimated by $1/D$. The log of the ratio is $\log R_1 - \log R_2$, a difference of independent quantities, so the variances add:

$$\operatorname{SE}(\log(R_1/R_2)) = \sqrt{\frac{1}{D_1} + \frac{1}{D_2}} = \sqrt{\frac{1}{17{,}269} + \frac{1}{6{,}208}} = 0.0148,$$

and the 95% interval is $1.485 \times e^{\pm 1.96 \times 0.0148} = (1.443,\ 1.529)$.

Example (what a large $z$ does not show). A $z$ of 28.6 is far beyond any conventional threshold, and it still does not show that a Maine resident of a given age was more likely to die than an Alaskan of the same age. This is the misreading the whole unit is built to prevent. The difference is enormous by the standard of random variation: no plausible chance fluctuation produces a $z$ of 28.6. But the test answers only whether the two crude rates differ by more than chance, and they do. It cannot say whether the difference reflects health or demography, because the crude rates have different weights.

The calculation in the previous section already settled part of that. With the same national age-specific rates applied to both states, age structure alone would give Maine a crude rate 57% higher than Alaska's.

Precision and validity are different properties: a large population makes a rate precise, but only a comparison of like with like makes it mean what it appears to.

Summary

Quantity Definition
Crude rate $D/N = \sum_i (n_i/N)\, r_i$
Standard error of a rate $R/\sqrt{D}$; relative standard error $1/\sqrt{D}$
Exact 95% limits for a count $D$ $G^{-1}_{D}(0.025)$ and $G^{-1}_{D+1}(0.975)$
Two rates, 100 or more deaths each $z = (R_1 - R_2)/\sqrt{\operatorname{SE}_1^2 + \operatorname{SE}_2^2}$; rate ratio with $\operatorname{SE}(\log) = \sqrt{1/D_1 + 1/D_2}$

Worked problems

These three problems are posed on the article's own data: the U.S. deaths of 2019 and the 2021 state figures. Each exercises one of the main results, and the script named under "In Python" computes every number and asserts the main ones.

Problem 1 (a crude rate from two age groups). Split the U.S. population of 2019 at age 65. (a) Find the death rate in each part and show that the crude rate is their weighted average. (b) What would the crude rate be, with the same two rates, in a population with 21.66% aged 65 and over, like Maine's in 2021, and in one with 13.33%, like Alaska's?

Solution. (a) Adding the last three rows of the table, 2,117,332 deaths occurred among 54,058,263 people aged 65 and over, a rate of 3,916.8 per 100,000. The other eight rows give 737,359 deaths among 274,181,260 people under 65, a rate of 268.9. The share aged 65 and over is $54{,}058{,}263/328{,}239{,}523 = 0.1647$. So

$$0.1647 \times 3{,}916.8 + 0.8353 \times 268.9 = 645.1 + 224.6 = 869.7 \text{ per 100,000},$$

the crude rate. The 16.5% of the population aged 65 and over account for 74.2% of the deaths.

(b) With a share of 0.2166 aged 65 and over, the crude rate would be $0.2166 \times 3{,}916.8 + 0.7834 \times 268.9 = 1{,}059.1$. With a share of 0.1333 it would be $0.1333 \times 3{,}916.8 + 0.8667 \times 268.9 = 755.2$. The two hypothetical populations have identical death rates in both age groups, and their crude rates differ by a factor of 1.40. With the same 2019 rates in all eleven age groups, the two states' actual age structures give crude rates of 1,028.8 and 646.0, a factor of 1.59 (the section on the crude rate found 1.57 with the national rates of 2021). Two broad groups capture much of the effect of age structure. Eleven capture more, because the two states also differ in their ages within each broad group.

Problem 2 (a rate at the publication threshold). A county of 50,000 people records 20 deaths from some cause in a year. Find the rate, its relative standard error, and the normal and exact 95% intervals.

Solution. The rate is $20/50{,}000 = 40.0$ per 100,000, with relative standard error $1/\sqrt{20} = 22.4\%$ and standard error $40.0/\sqrt{20} = 8.94$.

  • The normal interval is $40.0 \pm 1.96 \times 8.94 = (22.5,\ 57.5)$ per 100,000.
  • The exact limits for the count are $G^{-1}_{20}(0.025) = 12.217$ and $G^{-1}_{21}(0.975) = 30.888$. Dividing by the population, the exact interval is (24.4, 61.8) per 100,000.

The script confirms what the limits mean: a Poisson count with mean 12.217 is 20 or more with probability 0.025, and one with mean 30.888 is 20 or fewer with probability 0.025. The exact interval sits higher than the normal one at both ends and is longer above the estimate (21.8) than below it (15.6). At this count, the smallest for which NCHS publishes a rate, the interval runs from about 0.6 to 1.5 times the rate.

Problem 3 (two crude rates and their ratio). In 2021 Florida recorded 261,369 deaths, a crude rate of 1,200.0 per 100,000, and Utah 22,569 deaths, a crude rate of 676.1 (Murphy et al., 2024, Table 17). Test the difference between the crude rates, and give the rate ratio with a 95% interval. What does the result establish?

Solution. The standard errors are $1{,}200.0/\sqrt{261{,}369} = 2.35$ for Florida and $676.1/\sqrt{22{,}569} = 4.50$ for Utah. The difference is $1{,}200.0 - 676.1 = 523.9$, with standard error $\sqrt{2.35^2 + 4.50^2} = 5.08$, so

$$z = \frac{523.9}{5.08} = 103.2,$$

and the 95% interval for the difference is $523.9 \pm 1.96 \times 5.08 = (514.0,\ 533.8)$ per 100,000. The rate ratio is $1{,}200.0/676.1 = 1.775$, with $\operatorname{SE}(\log) = \sqrt{1/261{,}369 + 1/22{,}569} = 0.00694$ and 95% interval $1.775 \times e^{\pm 1.96 \times 0.00694} = (1.751,\ 1.799)$.

The result establishes that the two crude rates differ by far more than random variation can explain, and nothing more. Florida had the second-highest share of residents aged 65 and over in 2021 (21.1%) and Utah the lowest (11.7%). The age-adjusted rates of the next article, 832.9 and 815.8, differ by a factor of only 1.021.

What we have learned

A crude rate is a weighted average of age-specific rates, with the population's own age structure as the weights. Its precision depends only on the number of deaths: a relative standard error of $1/\sqrt{D}$, and exact gamma limits when the count is small.

Maine's crude death rate is far above Alaska's, by a margin no chance fluctuation could produce. Yet the national rates, applied to each state's ages, predict an even larger gap. The remedy is to compare the states' rates at the same ages. Direct standardization applies both states' age-specific rates to a single standard population; indirect standardization applies one set of standard rates to both states and compares their observed deaths with the deaths expected. Direct Standardization develops the first method and Indirect Standardization and the Standardized Mortality Ratio the second. They establish that Alaska's mortality was the higher once the states are compared at the same ages, and that the size of such a comparison depends on whose ages are chosen as the standard. So can its direction, when the age-specific rates being compared are not all higher in one population.

In Python

from math import sqrt, exp, log
from scipy import stats

deaths, pop = 17_269, 1_372_247              # Maine, 2021
rate = deaths / pop * 1e5                     # 1,258.4 per 100,000
se = rate / sqrt(deaths)                      # 9.58
print(rate - 1.96 * se, rate + 1.96 * se)     # normal 95% interval

d, n = 7, 25_000                              # a small count
lo = stats.gamma.ppf(0.025, d)                # 2.814
hi = stats.gamma.ppf(0.975, d + 1)            # 14.423
print(lo / n * 1e5, hi / n * 1e5)             # exact interval: 11.3, 57.7

d1, r1, d2, r2 = 17_269, 1258.4, 6_208, 847.3  # Maine and Alaska
se_log = sqrt(1 / d1 + 1 / d2)
print(r1 / r2, r1 / r2 * exp(-1.96 * se_log), r1 / r2 * exp(1.96 * se_log))

Every number in this article is computed by tools/experiments/applied-statistics/rates-in-public-health.py in the site's repository, the script it shares with Rates in Public Health. The script downloads the Census estimates and queries CDC WONDER, checks both against the published NCHS figures, and runs its simulations with fixed seeds.

References

  • Anderson, R. N., and Rosenberg, H. M. (1998). Age standardization of death rates: Implementation of the year 2000 standard. National Vital Statistics Reports, 47(3). National Center for Health Statistics. record · 4 pages
  • Breslow, N. E., and Day, N. E. (1987). Statistical Methods in Cancer Research, Volume II: The Design and Analysis of Cohort Studies, Chapter 2 (Rates and rate standardization). IARC Scientific Publications No. 82. International Agency for Research on Cancer. record · 5 pages
  • Centers for Disease Control and Prevention, National Center for Health Statistics. Underlying Cause of Death 1999–2020, CDC WONDER online database, from the Multiple Cause of Death Files compiled through the Vital Statistics Cooperative Program. https://wonder.cdc.gov/ucd-icd10.html (queried November 2026). record · 3 pages
  • Murphy, S. L., Kochanek, K. D., Xu, J., and Arias, E. (2024). Deaths: Final data for 2021. National Vital Statistics Reports, 73(8). National Center for Health Statistics. Tables 2, 6, 17, IX and XI and the Technical Notes. record · 6 pages
  • Pagano, M., Gauvreau, K., and Mattie, H. (2022). Principles of Biostatistics (3rd ed.). CRC Press. The course text. record · 48 pages
  • U.S. Census Bureau, Population Division (2022). Annual State Resident Population Estimates for 6 Race Groups (5 Race Alone Groups and Two or More Races) by Age, Sex, and Hispanic Origin: April 1, 2020 to July 1, 2021 (Vintage 2021; released June 2022), file sc-est2021-alldata6. https://www2.census.gov/programs-surveys/popest/datasets/2020-2021/state/asrh/ record · 5 pages