“A great surgeon performs operations for stone by a single method; later he makes a statistical summary of deaths and recoveries, and he concludes from these statistics that the mortality law for this operation is two out of five. Well, I say that this ratio means literally nothing scientifically and gives us no certainty in performing the next operation; for we do not know whether the next case will be among the recoveries or the deaths. What really should be done, instead of gathering facts empirically, is to study them more accurately, each in its special determinism…. to discover in them the cause of mortal accidents so as to master the cause and avoid the accidents.”

Source: 1880s, Incidents and Anecdotes of the Civil War (1885), p. 214

Variants: One death is a tragedy. A million deaths is just a statistic.
A single death is a tragedy; a million deaths is a statistic.
When one dies, it is a tragedy. When a million die, it is a statistic.
In Портрет тирана (1981) (Portrait of a Tyrant), Soviet historian Anton Antonov-Ovseyenko attributes the following version to Stalin: "When one man dies it's a tragedy. When thousands die it's statistics." This is the alleged response of Stalin during the 1943 Tehran conference when Churchill objected to an early opening of a second front in France.<!-- The book appears to have a footnote sourceing the claim, but I couldn't access it. Could someone please try to scare up a paper copy and have a look at footnote 188? -->
In her review "Mustering Most Memorable Quips" of Konstantin Dushenko's 1997 Dictionary of Modern Quotations (Словарь современных цитат: 4300 ходячих цитат и выражений ХХ века, их источники, авторы, датировка), Julia Solovyova states: "Russian historians have no record of the lines, 'Death of one man is a tragedy. Death of a million is a statistic,' commonly attributed by English-language dictionaries to Josef Stalin."
This quotation may originate from "Französischer Witz" (1925) by Kurt Tucholsky: "Darauf sagt ein Diplomat vom Quai d'Orsay: «Der Krieg? Ich kann das nicht so schrecklich finden! Der Tod eines Menschen: das ist eine Katastrophe. Hunderttausend Tote: das ist eine Statistik!»" ("To which a Quai d'Orsay diplomat replies: «The war? I can't find it so terrible! The death of one man: that is a catastrophe. One hundred thousand deaths: that is a statistic!»")
Another possible source or intermediary may be the concluding words of chapter 8 of the 1956 novel The Black Obelisk by Erich Maria Remarque: "Aber das ist wohl so, weil ein einzelner immer der Tod ist — und zwei Millionen immer nur eine Statistik." ("But probably the reason is that one dead man is death—and two million are only a statistic." 1958 Crest Book reprint)
Mary Soames (daughter of Churchill) claims to have overheard Stalin deliver a variant of the quote in immediate postwar Berlin (Remembrance Sunday Andrew Marr interview BBC 2011) https://www.youtube.com/watch?v=9hP2tpw9XEw
See also Jean Rostand, Thoughts of a Biologist, 1939: "Kill one man, and you are a murderer. Kill millions of men, and you are a conqueror. Kill them all, and you are a god."
In an interview given for the 1983 three-part documentary Der Prozeß by Norddeutscher Rundfunk on the Third Majdanek trial, Simon Wiesenthal attributes the quote to the unpublished auto-biography of Adolf Eichmann. According to Wiesenthal, Eichmann had been asked by another member of the Reich Main Security Office during WWII what they should answer would they be questioned after the war about the millions of dead Jews they were responsible for, to which Eichmann according to his own testimony had replied with the quote.
Misattributed
Variant: The death of one man is a tragedy, the death of millions is a statistic.

C. West Churchman, "Operations research as a profession" (1970); cited in Arjang A. Assad, Saul I. Gass (2011) Profiles in Operations Research: Pioneers and Innovators. p. 181
1960s - 1970s

Ch 1 : Production
Elements of Political Economy (1821)

As quoted in Management and the Computer of the Future (1962) by Sloan School of Management, p. 273
Context: We must include in any language with which we hope to describe complex data-processing situations the capability for describing data. We must also include a mechanism for determining the priorities to be applied to the data. These priorities are not fixed and are indicated in many cases by the data.
Thus we must have a language and a structure that will take care of the data descriptions and priorities, as well as the operations we wish to perform. If we think seriously about these problems, we find that we cannot work with procedures alone, since they are sequential. We need to define the problem instead of the procedures. The Language Structures Group of the Codasyl Committee has been studying the structure of languages that can be used to describe data-processing problems. The Group started out by trying to design a language for stating procedures, but soon discovered that what was really required was a description of the data and a statement of the relationships between the data sets. The Group has since begun writing an algebra of processes, the background for a theory of data processing.
Clearly, we must break away from the sequential and not limit the computers. We must state definitions and provide for priorities and descriptions of data. We must state relationships, not procedures.

On losing one of his wingmen during an unauthorized dive-bombing maneuver used against some authorized targets on a beach in Vietnam, as quoted in Afterburner : Naval Aviators and the Vietnam War (2004) by John Darrell Sherwood, Ch. 20 : Leighton Warren Smith and the Fall of Thanh Hoa, p. 276
Context: We went out on a very cloudy day, in an area we probably should not have been operating in, and the two of us did a maneuver that was not authorized, and I made it and he didn't. … We pulled up over some overcast to dive bomb the target. I had practice with this maneuver in the Med, but Bill was less familiar with it. I made it, but Bill just disappeared. I went back and looked for hims. I saw no race of any wreckage. I tried to tank and go back but was ordered to return to the ship.

"On Statistics"
The Silence of the Sea (1940)

Aber das ist wohl so, weil ein einzelner immer der Tod ist — und zwei Millionen immer nur eine Statistik.
Der schwarze Obelisk (1956)
A variant of this quote "One death is a tragedy. A million deaths is just a statistic." has also been attributed to Joseph Stalin, but no source for this has been found. This version appeared in the English press not later than 1958. (Ремарк, Эрих Мария // Словарь современных цитат / составитель К. В. Душенко — Москва: изд-во «Эксмо», 2006)

Göran Asplund (1975), Strategy formulation: an intervention study of a complex group decision process. p. 22