Departiurut of (flutnmrrrc anti labor STATISTICAL TREATMENT OF CAUSES OF DEATH COOPERATIVE WORK RELATIVE TO TREATMENT OF JOINTLY RETURNED CAUSES AND THE RE- VISION OF THE INTERNATIONAL CLASSIFICATION d BUREAU OF THE CENSUS S. N. D. NORTH DIRECTOR Qepartmrnt nf Cnntmrrn' attb iCahnr STATISTICAL TREATMENT OF CAUSES OF DEATH Cooperative work relative to treatment of jointly returned causes and the re- vision of the international classification Plan proposed by the Committee on Demography of the American Public Health Association & 4* & & BUREAU OF THE CENSUS S. N. D. NORTH DIRECTOR 8-419 V. S. No. 105 2 STATISTICAL TREATMENT OF CAUSES OF DEATH. INTRODUCTION. This pamphlet is issued in accordance with the general plan of coop- erative work between the United States Census Bureau, the Public Health and Marine-Hospital Service of the United States, committees of national associations representing the organized sanitary sendees of the country, individual sanitary workers, and the medical profession. All are alike dependent upon accurate and fully comparable mortality statistics for many important purposes, and it is necessary that the opin- ions of the users of vital statistics upon certain questions of detail be obtained as fully as possible in order that the methods of classification and presentation of data may be generally acceptable. The agencies cooperating with the Census in this movement for the improvement of vital statistics in the United States are the following: American Public Health Association-Committee on Demography and Statistics in their Sanitary Relation. W. A. King, Chairman, Chief Statistician, United States Census Bureau, Washing- ton, D. C. Cressy L. Wilbur, M. D., Chief of Division of Vital Statistics, Department of State, Lansing, Mich. J. S. Fulton, M. D., Secretary of State Board of Health, Baltimore, Md. H. M. Bracken, M. D., Secretary of State Board of Health, St. Paul, Minn. Jesus E. Monjaras, M. D., Mexico, D. F. F. L. Hoffman, Newark, N. J. Conference of State and Provincial Boards of Health of North America-Committee on Vital Statistics. Irving A. Watson, M. D., Chairman, Secretary of State Board of Health, Concord, N. H. Charles A. Ljndsley, M. D., Secretary of State Board of Health, New Haven, Conn. Peter H. Bryce, M. D., Secretary of Provincial Board of Health and Deputy Registrar-General, Toronto, Ontario. Cressy L. Wilbur, M. D., Lansing, Mich, (honorary member). American Medical Association-Committee on Public Health. Charles O. Probst, M. D., Chairman, Secretary of State Board of Health, Columbus, Ohio. Maj. W. C. Gorgas, Surgeon, U.S. Army. J. N. Hurty, M. D., Secretary of State Board of Health, Indianapolis, Ind. H. A. West, M. D., Galveston, Tex. H. W. Sanders, M. D., State Health Officer, Montgomery, Ala. Victor C. Vaughan, M. D., Ann Arbor, Mich. Public Health and Marine-Hospital Service of the United States. Walter Wyman, M. D., Surgeon-General. 3 4 At the meeting of the American Public Health Association held at New Orleans, December 8 to 12, 1902, the following resolution relating to vital statistics was passed: Resolved, That the efforts of the Committee on Demography and Statistics of this Association, in conjunction with the United States Census Office, to secure the extension of the registration area by the enactment of suitable laws and the use of a standard form of certificate of death, as shown in Census Circular No. 71, be heartily commended, and that the committee be further authorized and directed to cooperate with the Census Office, the United States Public Health and Marine-Hospital Serv- ice, and other departments of the Federal Government interested in vital statistics, and with similar committees from other associations, in the work of promoting the adoption of suitable registration laws and the extension of the registration area, the proper compilation and presentation of vital statistics by States and cities in weekly or monthly bulletins and annual reports, and also in further work relating to the extension and practical use of the International Classification of causes of death, the disposition of jointly 'returned causes, and all preliminary work relating to the next decennial revision. At the annual meeting of the American Medical Association at New Orleans, May 5 to 8, 1903, a similar resolution was adopted, providing for the fullest cooperation between the Public Health Committee of the American Medical Association, the committee of the American Public Health Association, a similar committee authorized by the conference of state and provincial boards of health of North America, at New Haven, October 28 and 29, 1902, and the Public Health and Marine- Hospital Service and the Bureau of the Census on the part of the Government, not only for the improvement of the classification, but also for the extension of adequate registration methods and the proper presentation of statistical data relating to diseases and deaths. It is thus seen that there is a widespread desire to improve the methods of collec- tion and presentation of vital statistics in this country, and it is hoped that general advantage will be taken of this opportunity to aid in the work. Special attention is called to the italicized portion of the above resolution. The classification of deaths in which a single cause is assigned has been very completely provided for by the Manual of International Classifica- tion of Causes of Death, published and distributed by the Bureau of the Census, and its acceptance by registration officials as the authority for this purpose insures uniformity in the statistics to that extent; but the classification of deaths in which two or more apparently independent causes are assigned by the physicians has not been so adequately provided for, although equally essential to complete uniformity in statistical results, and the subject as usually considered presents many difficulties. The vast number of terms, with their variations and synonyms, appear- ing in the certificates makes it practically impossible to provide an alpha- betical index which would give the proper reference in cases where two or more such terms appear in combination, even if there were no difficulty in deciding where such cases should be classified. The actual combina- 5 tions appearing even in a comparatively small number of cases are so numerous .as to preclude this method, and the possible combinations are innumerable. Even a partial index of them would require a volume the size of a dictionary, and would be too bulky and cumbersome for ready reference. On the other hand, general rules upon the subject must necessarily permit a wide range in construction and application, and result in the variations in classification which it is most desirable to avoid. The necessities of the case and the obvious objections to either of the methods referred to have led to what is thought to be a practical solu- tion of the difficulties and the adoption of a simple system for disposing of such cases by reference to the alphabetical index of single terms, as given in the manual. This system was devised by Dr. Cressy L. Wil- bur, medical referee of the Census, and is advocated by the Committee on Demography and Statistics of the American Public Health Associa- tion, of which he is a member, and the object of this publication is to explain the method involved, and to enlist the assistance of all persons interested in perfecting it and putting it into proper shape for general use. Briefly stated, the plan is to assign to each title of the classification a numerical weight or rating indicating its importance, and to each term compiled under any title a similar numerical weight, indicating the validity of that term with reference to the title under which it is com- piled. The product of the two will show the relative value of each term in the general scheme of the classification in comparison with any other term or terms, and may be printed, as a simple quantitative exponent, in the ordinary alphabetical index. Should the proposed method be found to work satisfactorily, the results will also be available for the purpose of amending the classi- fication at its next decennial revision. In the manual of the international classification referred to (which will be sent on request) will be found, beginning on page 19, last column, a detailed list of the titles composing the international classification. On pages 29 to 109, inclusive, the original forms of returns included under each title of the classification are given; no attention need be paid to the bracketed subdivisions of titles. Reference should be made also to the introductory text, pages 5 to 25, inclusive, and especially to the section on "jointly reported causes of death," pages 12 and 13. CLASSIFICATION OF "JOINTLY RETURNED" CAUSES. Jointly returned causes of death may be defined as returns of two or more causes for a single death. Sometimes the arrangement of the certificate of death contributes to such returns, as by providing for a statement of primary and secondary causes, chief or determining and consecutive or contributory causes, immediate cause, etc. When the order and duration of each cause are definitely stated upon the certificate, 6 or when the physician has plainly indicated the disease which was the primary cause of death, then the registration office has no difficulty in determining the title under which the death is to be compiled. No mat- ter how many causes may be assigned in the original certificate, the death must be compiled under a single title in all present systems of classifi- cation, and the title selected should be the one which most satisfactorily represents the cause of death for all purposes. It is evident that the order of the terms originally reported can not be accepted as the invariable order of preference, but that the terms assigned by the physician must be considered as a whole in order to determine the cause of death to be selected for compilation. The following are a few examples selected from the thousands that occur. It should be understood that the order is of no importance, since the terms may be returned in any order: Heart disease, bronchitis, and Bright's disease. Heart disease and chronic nephritis. Chronic Bright's disease and valvular heart disease. Chronic nephritis, uremia, and cardiac hypertrophy. Apoplexy, heart disease, and chronic Bright's disease. Valvular disease of heart, nephritis, and apoplexy. Whooping cough, bronchopneumonia, and scarlet fever. Diabetes mellitus and phthisis. Abscess of brain, otitis media, and scarlet fever. Chronic endocarditis and acute nephritis. Cerebral hemorrhage, Bright's disease, and arterio-sclerosis. Chronic parenchymatous nephritis, heart disease, and apoplexy. Many more examples of jointly returned causes of death might be given. In some cases there is little difficulty in choosing, while in others it is extremely perplexing to decide upon, and even more difficult to maintain a consistent course among hosts of synonymous terms of varying degrees of accuracy. The systematic treatment of jointly returned causes of death may afford a key to the criticism of the classification as a whole and in detail, forming a valuable basis for a decennial revision, which must enhance the value of the classification and thus lead to its use by an ever-increasing number of registration officers and users of mortality statistics. When a physician names two or more diseases as being more or less responsible for a single death, it is evident that the compiling office must choose between them. Excluding cases in which the natural preference may be indicated by the relative duration of the several diseases, or by the fact that one disease is plainly stated as a consecutive or contributory cause, there remain for final decision in the registration office many cases in which there exists no guide except the application of more or less definite general principles. It is very important that decisions shall be uniform, since the comparability of the statistics will be impaired by variations. 7 It is evident that if deaths were invariably reported in terms corre- sponding exactly to the titles of the classification, the only thing to be considered would be the relative importance of the titles in question. Thus, a death from typhoid fever might be considered as more impor- tant for the ends subserved by statistical statement than a death from pneumonia, because as regards the restriction and prevention of disease, typhoid fever is at present perhaps of greater consequence to sanitary administration, and therefore to sanitary workers, who are the chief users of mortality statistics. Accordingly it may be more important tq have all deaths from typhoid fever stated than all deaths from pneu- monia, it being impossible, under any method of classification now in use, to fully represent both terms. Another factor to be taken into con- sideration in fixing the relative importance of two titles is the question as to whether one disease is a complication of the other. Pneumonia may be a complication or a form of typhoid fever, but typhoid fever does not result from pneumonia. Hence typhoid fever would have a higher rank as an independent disease. But causes of death are not always reported in the exact titles used in the classification, or even in their precise synonyms. If continued fever, for example, should be returned by the physician, although it would be compiled under typhoid fever, if it were the only cause given, it is evident that there might be considerable doubt as to whether it really belonged under that title. Its validity as a representative of the title would be less than a return of typhoid fever or enteric fever. Practically, in determining the relative weight of causes of death jointly returned by physicians, two factors are considered: (a) Importance of the title under which the term returned falls. (b) Validity of term returned, with reference to title. If each of these factors be assigned a numerical rating-for example, upon a scale from o to io-then their product, ranging from o to 100, will show the relative weight of a given return in a form readily com- parable with any other. As an example of the way in which such ratings could be used, we may take a few terms under typhoid fever and pneumonia as given in the tabular list of the manual, and assign to them values which are merely examples of possible ratings: i. Typhoid fever, io (importance). This title includes: Abdominal fever, 6 (validity), typhoid, io. typhus, io. Adynamic fever, 3. Continued fever, 4. Enteric fever, 10. Typhoid fever, 10. Typhomalarial fever, 9. 8 93. Pneumonia, 8 (importance). This title includes: Acute pneumonia, io (validity). Alcoholic pneumonia, 8. Apical pneumonia, 9. Bilious pneumonia, 5. Chest, inflammation of, 4. Croupous pneumonia, 10. Lung fever, 9. Typhoid pneumonia, 5. Each of the judgments as to importance of the title and validity of the term under the title is thus capable of definite expression on a numer- ical scale. Given this expression, then, in the index of the manual the weight of each term, represented by the product of the two ratings, might be written in superior figures, as follows: Abdominal fever60 (i). typhoid100 (i). typhus100 (i). Acute pneumonia80 (93). Adynamic fevers0 (j). Alcoholic pneumonia64 (93). Apical pneumonia?2 (93). Bilious pneumonia40 (93). Chest, inflammation of32 (93). Continued fever40 (i). Croupous pneumonia80 (93). Enteric fever100 (1). Lung fever (93). Typhoid fever100 (1). Typhoid pneumonia40 (93). Typhomalarial fever90 (1). Etc. The numbers in parentheses indicate the title of the international classification to which the term belongs. When any two terms are returned together, the one having the higher exponent should be given the preference in the classification. Thus enteric fever100 would be selected instead of acute pneumonia80, while the latter would be preferred to abdominal fever60 or continued fever40. In case the exponents are the same, some rule, as of priority, should govern. But certain titles, as yellow fever, should be especially weighted with regard to sanitary importance, as by writing 10+ ; and a product containing 10+ would of course take precedence of the same product derived from 10. One practical advantage of such a system of rating terms returned as causes of death lies in the fact that these ratings, which make it possible to determine instantly the relative compiling value of any of the thousands of possible pairs of terms, might be included in the index of the manual of classification without increasing its length. When it is considered that the 179 titles of the classification alone would involve 15,931 possible pairs, and that the number of pairs derivable from the whole number of terms included in the index would rise into the millions, it will be seen that this is no small advantage. It is not only as a method of treating jointly returned causes of death, however, that a system of definite numerical ratings of titles and terms of the classification would be useful. The rating submitted by each individual is in effect a criticism of the classification, and as such can be utilized more readily than verbal criticism in the work of revising the system. Titles are first rated in the order of importance; it follows that 9 titles universally rated very low might advantageously be omitted from the classification. Terms are rated as to their validity under their respective titles; any low rating, say less than 5, vitiates the statistics of the title to which it is assigned, indicating that the term either belongs under some other title or else should be relegated to the limbo of indefi- nite returns. The rating of the validity of terms with respect to their titles will necessarily be a task of considerable magnitude, because of the large numbers involved. In order to do the work satisfactorily, some idea of the exact meaning and relative importance of the titles themselves must first be obtained, and it is the purpose of this publication to enlist assist- ance in establishing the rating of these titles. The aid of clinicians, pathologists, sanitarians, vital-statisticians, and all other persons actively interested in the use and improvement of statistics of causes of death is earnestly solicited. The information desired may be conveniently given by means of the accompanying blank, on which space is provided not only for rating the importance of each title as a constituent of mortality tables, but also for the insertion of any special definition or desirable modification of a title. Thus, under title 93, "pneumonia," it might be suggested that a statement of "lobar or croupous pneumonia" and of "pneumonia (not otherwise defined)," as in the recent English revision, would be preferable to the present practice. Incidentally, it may be noted that the lower rating of pneumonia as compared with typhoid fever in the examples previously given might depend upon the somewhat indefinite inclusion of title 93 at present. Other general criticisms of the classification are solicited, and will be carefully compared and studied in formulating suggestions for the next revision. These may relate to the general principle or arrangement of the system, the addition of new terms or groups, etc. For convenience, it will be desirable to use a separate sheet for each suggestion, and to refer to the title in question by number. In making the ratings requested, attention is especially directed to the suggestions which follow. Reference should also be made to a paper upon the subject (see appendix) read by Doctor Wilbur (a member of the Committee on Demography and Statistics of the American Public Health Association) at the annual meeting of the American Medical Association, at New Orleans, May 5 to 8, 1903. SUGGESTIONS FOR CRITICAL EXAMINATION AND RATING OF TITLES OF THE INTERNATIONAL CLASSIFICATION OF CAUSES OF DEATH. The accompanying blank form (8-307) presents the titles of the inter- national classification conveniently arranged for criticism and rating. It consists of five columns, as follows: Column 1: Title number.-Titles are numbered serially for con- venience. 10 Column 2: Title.-The duplicated titles (in italics) repeat special data included in the more general titles immediately preceding. (See manual, page 24.) Column 3: Rating.-Please mark in this column your estimate of the relative importance of each title of the classification. If a title is definite and includes deaths concerning which you consider it of the first importance to have statistical information, mark it "10," the maximum in the scale. If a title is indefinite and practically worthless from the point of view of mortality statistics, mark it "o." If a title is of about average importance and precision, mark it "5" on the same scale. Other titles should be inserted between these rates according to your own personal judgment, based on the practical uses and special interest which statistics of causes of death possess for you, individually. It is understood that individual lists may be biased in certain directions because of special uses made of the data, but in comparing the ratings made by all classes of users of mortality statistics these differences will be eliminated. It is earnestly hoped that ratings may be given for each title in column 3. Should that be impracticable, then ratings for as many titles as pos- sible will be of service. Column 4.: Definition.-Define the title as it should be understood for use in mortality tables. The present inclusion of the title can be found in the tabular list of the manual (pages 29-109). If a more precise and limited definition would be desirable, note the fact. Column 5: Remarks.-Under this head any remarks as to uses, assign- ment to other titles, frequent complications, etc., may be entered, and new titles may be suggested for the separate statement of causes included under present titles. When the space provided in columns 4 and 5 is insufficient for com- ments on any title, use a separate sheet, referring by number and name to title discussed. Merely as an example, without suggesting the rating for this specific case, the blank might be filled out somewhat as follows, for title 35, ''scrofula'': 1 No. 2 Title. 3 Rating. 4 Definition. 5 Remarks. 35 Scrofula.... 2 Tuberculosis of lymph glands. Belongs under proper subdivision of tuberculosis. Term is a relic of the past, and should not appear in modern mortality tables. APPENDIX. COOPERATIVE METHODS FOR IMPROVING THE USEFUL- NESS OF STATISTICAL CLASSIFICATIONS OF CAUSES OF DEATH.® Cressy L. Wilbur, M. D., Chief of division of vital statistics, department of state, Michigan, and medical referee of the Bureau of the Census in vital statistics. One of the most important matters involved in the administration of a registration office is the selection and proper use of a statistical classi- fication of causes of death. The mortality from various diseases must be known in order that the value of methods adopted by sanitary officers for their prevention and restriction may be appreciated, and it is impor- tant for many reasons that accurate and fully comparable statistics of causes of death be published. This presupposes the use of some classi- fication of causes of death. It may be said that the use of a formal classification is unnecessary, as the causes returned by physicians might be stated in alphabetical order. This would be possible for only the smallest areas of registra- tion, and for short periods of time. For a large city, a state, or for the country as a whole, the number of different terms returned as causes of death would be so numerous that it would not be feasible to give each one a separate line in statistical tables. Even where it has been attempted to make use of the alphabetical arrangement, numerous terms have usually been condensed under single titles, so that a method of classification was actually in use. Where this is not the case the statis- tics may be very misleading, and may give rise to annoying blunders on the part of those consulting them unless they first complete the work of proper statistical compilation that was left unfinished by the registration office. As an amusing instance, I may quote the following from a recent issue of The Journal of the American Medical Association, under the heading, "Clippings from lay exchanges": Tuberculosis of the lungs carried off twenty-five of the victims of disease last month. Seven others died of tuberculosis and four of phthisis. « Read at the fifty-fourth annual session of the American Medical Association, in the section on Hygiene and Sanitary Science, and approved for publication by the executive committee-Dr. John S. Fulton, Dr. Seneca Egbert, and Dr. H. M. Bracken. Reprinted from the Journal of the American Medical Association, August 15, 1903. 11 12 It may be well at this point to call attention to the limited sense in which the word "classification" is now employed in this connection. It does not refer to a broad and comprehensive philosophical arrange- ment of diseases, based on general principles and intended to show the natural or scientific relations of the diseases or causes of death.. If this were the object of a statistical classification of causes of death at the present time, it would be necessary to acknowledge failure from the start. No ideal classification fully exhibiting the natural relations of diseases and meeting the demands of all classes of scientific workers is yet possi- ble. The tendency is strongly in the direction of doing away with all so-called ' ' groups " or " classes ' ' of diseases and of placing stress chiefly on the definite meaning and inclusion of the individual titles. ' ' Classifi- cation," said Doctor Farr, "is another name for generalization." Our object at the present time, in statistical classification of causes of death, is to do away with generalization as far as possible. The old statistical groups of "zymotic diseases," "constitutional diseases," "diathetic diseases," and the like, have disappeared, or remain only as belated relics of a former nomenclature. Even the distinction between "general dis- eases ' ' and ' ' local diseases ' ' can not be considered of any special impor- tance to-day. The boundary is elastic, and a so-called "local disease" of to-day is placed by the advanced pathology of to-morrow among the general infections. By the adoption of a statistical classification of causes of death, there- fore, we mean nothing more than the use of a certain number of titles of diseases, each with a definite inclusion of terms as returned by physicians on certificates of death. For convenience a certain general arrange- ment is usually followed, as in the international classification, now in general use in this country. The titles of this system and the detailed list of terms included under each title are given in the Man- ual of International Classification, published last year by the United States Census Office. This classification was reported by Dr. Jacques Bertillon, as chairman of a committee of the International Statistical Institute, at the session held in Chicago in 1893. If was the result of a movement for uniformity in statistics of causes of death that had engaged the attention of registra- tion officials and users of mortality statistics for half a century. The conference of state and provincial boards of health indorsed it at Detroit in 1897, an(I the American Public Health Association recommended its general use at Ottawa in 1898, and reported a plan for its regular decen- nial revision by an international commission to meet at Paris in 1900. This recommendation was indorsed by the International Statistical Insti- tute at Christiania in 1899, and the international commission for the first revision of the classification met at Paris, August 18-21, 1900, by invi- tation of the French Government, representatives from twenty-six countries taking part in the sessions. This section of the American 13 Medical Association adopted a resolution in its favor at Atlantic City in 1900, and it has since been accepted by every registration state, most of the leading cities, the United States Census Office, United States Department of Labor, etc. It has been adopted by Canada, Mexico, and by the countries of Central and South America, and has no rival as an international classification. Its imperfections have never been ignored by the strongest advocates of its general adoption. It is faulty, as are all other systems. The reason for its use lies not in any perfection of the system, but because it was possible by adopting it to unite on a general basis of uniformity. This we have done, and we are now ready, by cooperative work, to amend and improve it, and make it, by means of the periodical revisions that all classifications require, more definite, more practical, more useful in every sense of the term to all who have occasion to employ statistics of causes of death. How shall this be done? Only by enlisting the aid of all users of mortality statistics and by ascertaining in just what respects the system requires alteration to best serve their wishes. Here it may be necessary to explain in fuller detail just what are the relations of vital-statisticians to the matter of classification of causes of death generally. The source of information in regard to causes of death is the medical profession. The individual physician who makes a statement on the medical certificate of cause of death in regard to a case that has occurred in his practice makes the first contribution to that mass of data from which come the mortality statistics. The certificate signed by the phy- sician goes into the local registration office and is there recorded. Possibly, if in a city, it may be compiled there, and the mortality statis- tics accompanying the annual report of the city board of health be derived therefrom. In our state, as in some other registration states, the certificate is next sent to the central registration office at the capital, and is there permanently preserved. It is again compiled, first, for the monthly bulletin of vital statistics, and, second, after all delayed returns and corrections have been received, for the annual registration report. Lastly, a transcript of this certificate goes to the United States Census Office at Washington, where it is. not recorded, but where the information is transferred to the cards employed in the electric tabu- lating machinery, and it is again compiled-this time for the annual report on mortality, now required by law to be prepared by the perma- nent Census Office. This is the final statistical report, but all alike-the national, state, and local compilations-agree in the fact that the origi- nal certificate of death signed by the attending physician is the source of all information. The validity of the statistics can not rise above the authority of their source, and, if the original statements of causes of death as made by the physicians are vague, indefinite, and full of error, 14 then the statistics must be defective and unsatisfactory to precisely the same degree. Compilation is that process by which statistics are derived from the mass of individual records, and, for our special purpose, may be limited to the production of the finished tables of causes of death, showing their various relations by locality, date of occurrence, age, sex, color, nativity, etc. It should be a colorless medium, by which all of the important features of the original returns may be preserved, in condensed form, and nothing should be added to or subtracted from the significance of the original statements. Here we may note two leading difficulties of compilation. These are, first, the condensation of the extensive list of terms reported by physicians to fit the comparatively short list of titles alone feasible in statistical reports; and, second, the selection of the proper title for com- pilation when two or more causes have been assigned by the physician. These difficulties increase with the mass of returns to be handled, and are more pressing in the larger offices, where the vast mass of causes and complications renders some uniform system of procedure indispensably necessary. Some idea of the large number of causes reported by physicians may be gained from examination of the Census Manual of the Inter- national Classification. Each one of these terms, with the exception of a few translated from the report of the French commission, has been actually returned on a certificate of death by some physician in the United States. And, moreover, any two or three of these terms may occur on a single certificate and in any combination. If there were only 5,000 different terms-and there are many more than that-then there would be 25,000,000 different binary combinations of those terms possible, or, disregarding the order of the terms, which is seldom important, there would be 12,500,000 combinations of different terms. Practically, such combinations are endless in variety, and yet in the process of compilation all of these 5,000 terms and 12,500,000 possible combinations must be reduced to the exact number of titles in the system selected. In the international classification there are only 179 titles in the detailed form, and no classification much exceeds this number. This process of condensation of medical terms under synonymous titles, or the selection of the true cause of death from several causes assigned by the physician, constitutes a large part of the practical work of a registration office in the treatment of causes of death. To be of value it must be under medical supervision, and yet, from the con- siderable amount of clerical labor involved, it is rare to find that a qualified medical statistician passes on each individual certificate of death. As a rule, certain principles and methods of procedure are estab- lished, and only exceptional cases are referred to expert judgment. 15 Even if competent medical statisticians gave their personal attention to the assignment of each certificate of death compiled in the various offices, there would be numerous differences of practice, and consequent variations in the resulting statistics, unless a comprehensive system of treatment were adopted in all alike. The adoption of such a system of assignment of individual single terms is the characteristic feature of the international classification, which renders it superior, for general adoption and the promotion of uni- formity, to any other system now in use. It gives the title to which each individual term is to be assigned. The question of assignment to one title or another, or the formation of new titles, is a proper subject for settlement at the regular decennial revisions, but between the revisions each term has a definite place in the classification, and all offices using the classification put it there. For the use of other systems there is no authoritative guide. A registration office in this country that undertook to employ the English registrar-general's classification might do so, so far as outward form was concerned, but there would be no assurance that the terms received from physicians' certificates were assigned as the registrar-general would assign them. Each office would be a law unto itself, just as in the old days before the adoption of uniform methods under the international classification. For the treatment of jointly returned causes of death no systematic method applicable to all classes of returns has as yet been introduced. It is usual in statistical offices to keep a book of practice showing decisions in such cases, and in time precedents may be collected showing the assignments of the most common causes when returned together. Attention was called to the importance of the subject by Wiirzburger, at the session of the International Commission at Paris, and a tabular arrangement showing the preferences to be given individual titles of the classification was appended to the report of the commission by Doctor Bertillon. This method, however, deals only with the definite titles, not with the numerous terms of varying degrees of accuracy included under the titles, and the judgment of the statistician would often be required as to whether the relative assignments of certain titles should govern the assignments of certain terms included. Many cases, more- over, including some of the most difficult, and practically all of the terms embraced under general or residual titles, are left to unguided individual decision. Consideration of the difficulties attending this class of decisions, and analysis of the process by which a final judgment was usually arrived at, have suggested another plan for the systematic treatment of jointly returned causes, and one which, in addition, seems adapted to the thor- ough revision of the international classification so that it will most fully meet the wishes of all of its users. This plan is set forth in a circular of the Committee of the American Public Health Association on Demog- raphy and Statistics in their Sanitary Relations. 16 When a death is returned from two causes, e. g., typhoid fever and valvular disease of the heart, one of these terms must be selected. The death can not be compiled under both typhoid fever and organic heart disease-at least it can not in any present system of classification. Which shall be chosen? Which title will best represent this case in the compiled statistical tables? Which will best serve the purposes of the users of the statistics? When the terms, as in the example above, are exactly the same as, or fully equivalent to, the titles of the classification, then the question resolves itself into ' ' Which title is more important?' ' Different answers may be given to such a question. The sanitarian, regarding typhoid fever as a dangerous communicable disease and interested in its preven- tion, wants to know the exact number of deaths from this disease. He cares nothing about organic heart disease, relatively, but says at once, ' ' Put it under typhoid fever. ' ' The actuary or medical director of a life insurance company, dealing with such a return, might consider the state- ment of heart disease the more important. As a matter of fact, the indi- vidual might not have succumbed from typhoid fever if the heart had not been affected, and might have lived for many years with the heart affection but for the intercurrent attack of typhoid fever. The clinician, the pathologist, the sanitarian, the actuary-each may have a different opinion. It is the duty of the vital-statistician to endeavor to satisfy the wishes of all, as far as possible, and for that purpose it is necessary to ask for the opinion of each class of users of mortality statistics. Such opinions to be available for use must be comparable, and the most obvi- ous way of attaining this object was to ask for numerical answers, as: ' ' What would you rate the importance of each title on a definite scale, say from o to io?" Then the more important title, as numerically rated for general purposes, should be preferred, no data as to relative duration, primary or secondary occurrence, etc., being available to modify such judgment. In a similar case, suppose the return to have been continued fever and valvular heart disease. Now continued fever, stated alone, is com- piled under typhoid fever, but there is some uncertainty about it. It can not be considered as fully equivalent to this title. The registrar in estimating its value as compared with the other cause assigned, neces- sarily considers this fact, and may decide that a doubtful statement of typhoid fever is less satisfactory for compilation than a positive one of organic heart disease. In other words, we consider two factors in decid- ing every case of independent joint causes, first, the relative importance of the titles to which the terms would be assigned, and second, the validity of the terms as more or less certainly belonging under the titles selected. Both of these factors are capable of numerical rating on the decimal or other scale. Having rated the importance of each title, and also the 17 validity of each term as related to its title assigned, then the product of these numbers will give a general weighting number or combining num- ber for each term returned on a certificate of death, by means of which the preference to be given in classification, i. e., assignment to titles, is definitely indicated. Provision would necessarily be made for equal rates, but a very large proportion of cases would be decided directly by the combining numbers. A very practical advantage of such a system is that the combining numbers or weights could be readily inserted in an ordinary index of causes, as in that of the Census manual, without increasing its length by a single page, and that the decisions between any two out of, say, 5,000 terms could be instantly found. Many millions of separate deci- sions would be thus condensed, which, expressed in any other way, would require many pages to record, and would be a vast, perhaps insuperable, labor to frame. The principle of each individual decision, as made between two independent terms coming before the registrar, would be precisely the same as that derived from his previously recorded numerical judgment as expressed in a rating. It is not my intention to urge that the classification of causes of death shall be an entirely mechanical process. To accomplish the best results it must necessarily be conducted under medical direction, and first- hand examination of the original certificates will often disclose valuable suggestions as to the true place of the return in classification. Never- theless, the most accomplished medical statistician will find cases where his judgment is uncertain, or where, at different times, he may make contrary decisions. A series of rating numbers would then be most useful, and would tend to unify practice. For many offices, however, where the compilation of data is left more or less entirely to clerks who do not possess medical qualifications, the plain and explicit directions available by means of the ratings would be invaluable. The advantages of such a system for the criticism and revision of the classification are obvious. With each title and each constituent term assigned a definite value in the opinions of the chief users of statistics, it would be merely necessary to lop off titles of little consequence, and to exclude terms of very uncertain validity under the titles, to greatly improve the precision and usefulness of the system. In some cases titles of slight precision would have to be retained on account of the large number of returns received under them, but their low value would be forcibly called to attention, and in time the physicians might be brought to avoid them. Indefinite terms that now vitiate the statistics of certain causes would be thrown out by this process of sifting, and a thoroughly revised and renovated classification would be the result. It is physically impossible for a commission to consider all of the points that require attention in an ordinary session of a few days, but with the concrete opinion of registrars, clinicians, pathologists, sanitarians, and other 18 classes of users of mortality statistics, definitely expressed in advance and capable of numerical analysis, the work of revision should be most complete and satisfactory. At present all classifications are lumbered to some extent with useless titles. These are titles of little precision, of rare occurrence, or divided more minutely than the accuracy of returns will warrant. On the other hand, some diseases of great importance are lumped together without reference to the very important clinical, pathologic, and etiologic distinc- tions that should be observed. It might be possible, by eliminating titles of small consequence, to obtain statistics of the causes of death of first importance, in all their necessary details, and also to add a selected list of the most important combinations of causes, without exceeding the usual limit of registration reports. What is needed at the present time is a full expression of opinion on the practical matters relating to the statistical presentation of causes of death, from the entire medical profession. The public health workers have begun the task, because the collection and publication of vital sta- tistics is usually a part of the duties devolving upon sanitary adminis- trations, and the preliminary announcement from the committee of the American Public Health Association, indorsed by the committee of the conference of state and provincial boards of health of North America, shows how the matter has been gone about. Sanitarians and vital-stat- isticians alone can not make this movement a complete success. The interest and help of the entire profession are necessary adjuncts. The clinician, whether practicing in city or country, whether a teacher in a medical school or a private observer, should render aid. The patholo- gist should consider the questions involved from his point of view. Men devoted to special branches of medicine, as pediatrics, gynecology, surgery, obstetrics, etc., should give the benefit of their experience. The aid of the various special societies should be obtained, and also, through the cooperation of the state societies, with their constituent county organizations, the friendly feeling and interest of the entire medical profession of the country should be secured for this movement to improve the accuracy and usefulness of mortality statistics. Natu- rally, with this object in view, we come to the American Medical Asso- ciation, and one of the purposes of the presentation of this paper is to ask the practical cooperation of the association, by means of a special committee on vital statistics, with the corresponding committees already organized by the American Public Health Association and by the con- ference of state and provincial boards of health. Many years ago a very able committee was appointed by this associa- tion, charged with the duty of "preparing a uniform plan for registration reports of births, marriages, and deaths, including a nomenclature and classification of causes of death. ' ' Members of this committee were Dr. Edward Jarvis, Dr. W. L. Sutton, Dr. Wilson Jewell, Dr. Edwin M. Snow, and Dr. R. W. Gibbes, and very able reports were presented to the 19 association at the sessions of 1858 and 1859. The intervention of the Civil War prevented the continuance of their work, and we are just beginning to be ready to-day, after over forty years, to carry out the principles of uniform and effective registration laws and uniform and comparable presentation of statistical data, especially relative to causes of death, that these pioneers of registration work so zealously advocated at that time. We now have the immense advantage of a permanently organized Census Office, representing the Federal Government, and pre- pared to cooperate and now effectively cooperating with the state and municipal registration offices and with the committees of those organi- zations that have undertaken to aid in this work. We now have the great additional advantage of a more thoroughly organized and united medical profession, and by its aid, under the auspices of the American Medical Association, these desirable reforms can be rapidly brought to accomplishment. It is time that a committee was again organized by the association for this purpose, and I trust that your recommendation will be heartily given for this purpose.