]\gt ; Study of a priori Next , is zero or negative , and the higher the infection rate the more injurious is the happening .
is zero when .
It generally increases at first with and afterwards decreases ; it is negative when proyided negative .
In other words , in these cases an increase of is beneficial at the beginning of the outbreak , but injurious later on .
Now and diminishes if the case-mortality increases ; in fact Thus an increase in the case-mortality is injurious at first , but may be ultimately beneficial , because an affected individual who dies ceases to be infective .
Application of the Theory of : to the Study of a priori -Part III .
By Lieut.-Colonel Sir TALD P , K.C.B. , , and HILDA P. HUDSON , M.A. , Sc. D. ( Received October 19 , 1916 .
) CONTENTS .
PAGE SECTION XI .
( vi ) Assigned 232 ( vii ) More General Equations 234 SECTION XII .
( i ) Mortality .
First Modification of the Equ tions of Part I 236 ( ii ) Second Modification of the Equations 237 SECTION XIII .
( i ) Discussion 238 XI .
( i ) Variable Happening.\mdash ; In Section IX , we commenced by touching upon the necessity of studying the effect of changes in the happeningelement , but there dealt only with the case of hypometric happening .
We now proceed to examine other in this element .
226 Lieut.-Colonel Sir R. Ross and Dr. Hilda P. Hudson .
It is unlikely that in any infectious disease the infectivity-element will vays r constant .
It may remain practically such for some length of time ; and hypothetical results of this condition have therefore been analysed above in detail ; but nature abhors a straight line , and we may infer from general experience that changes are sure to occur from time to time .
As gested in Section I , changes in infectivity may be due\mdash ; ( a ) to the action , so to speak , of the organisms themselves\mdash ; which is difficult to believe in because it involves the conception that organisms living in large numbers of different hosts should , as it were , make a kind of concerted and simultaneous effort ; and ( b ) to cres in season , climate , or other environment which may ( 1 ) increase or diminish the infective strength of the infecting organisms ; or ( 2 ) facilitate or hinder their transference from one host to another .
At present we can proceed only on the supposition that the change has occurred , without attempting to enquire how or why it has occurred ; and only some brief lotes on the subject can be given at present .
( ii ) Discontinuous Chcmge.\mdash ; We have hitherto supposed that when the number of affected individuals is , which may be small or large .
If the happening is hypermetric KL and is less than , the number of affected individuals , will increase towards the latter figure ; but if the happening is isometric KL , or is hypometric KL will diminish to the limits discussed in Section IX .
What will happen if the infectivity is suddenly changed during one unit of time from to , and thereafter remains at the latter figure ?
If , the -curve suddenly becomes steeper ; itself is continuous , but its gradient is discontinuous .
The -curve is discontinuous , and has a sudden rise at the moment when changes ; thereafter both curves proceed according to the equations with substituted for .
If at the moment of discontinuity more than half the population is affected , so that is descending , there is an epidemic manifestation , a sudden rise followed by a decline .
Both and are increased .
These are probably just the conditions which apply to many zymotic diseases , in which most of the population appears to remain affected , that is immune , for years or permanently .
In these cases , at the of discontinuity is nearly , and is nearly unity , and the augmented infectivity acts merely upon the small residue of non-affected individuals , let us say , newly-born infants , or children , or others who hitherto escaped infection by accident or by residence in remote villages .
It is well known that at the beginning of the present war recruits from many highland villages were attacked by very virulent measles on joining camps priori in the lowlands .
On the other hand , the same sort of epidsmic may occur in numbers of villages and small in which the considered infection has died out from want of material , and so allowed of a gradual accumulation of fresh material , which is then suddenly ( so to speak ) by the entry of one or two infective cases from outside .
It is extremely improbable , of course , that the infectivity actually makeb a saltum ; but we may still apply the equations already obtained if we give to a series of sive discontinuous augnnentations for successive units of time , by similar discontinuous diminutions until returlls to its value .
In his case the -curve will epidemic nanifestations if nd will then decline agaiu a value .
This is one way of to a continuous in ( iii ) from.\mdash ; We vevcr , ) new equations ) the hypothesi that iufectivity a function of the time .
Let us suppose that , where .
Then in the fundamental eqnation 4 namely , we have and KL Put . ?
that ; then KLdt KLdt)constant the solution .
As ever , KLdt ) : hence KLdt ) If is ible , as may often hapl ) in short and sharp epidemics immunit we vrite , lnany diseases at least , KLdt have at present no knowledge of the nature of the function or even if there is such a function .
if there is , we may suppose that , for positive values of is small when [ is small , increases at first as increases , reaches a maximum for a certain value of , and then declines to a small value again when is larger\mdash ; as is obseryed in mauy phenomena .
In other words should expect a symmetrical or non-symmetrical bell-shaped curve such as those with which we are familiar in statistical work .
of x.\mdash ; We may also suppose that is given in terms of the time , not by an analytic function , but by a curve whose equation VOL. XCIII.\mdash ; A. Lieut.-Colonel R. and Dr. Hilda .
Hudson .
then ( leter1nine the nature of ?
; and by raphical Since , there only a value of ' whPn or when K , then KL does ; when the ) ceases only me of , : is when Now since by definition lies ) ) yeen 0 and 1 , we call only have if between and 1 .
: the iagralu of IX ( i ) shows that this only occur when .
Hence unless ( -curve rises ) the level the no flnctuations , either rises steadily to 1 its limit , sinks to wherever -curve has a portion rising above can plot the values of corresponding to this rtion , which a cap , part the -curve , lying entirely between the and 1 .
If the -curye fluctuates , there are a series of these -caps , are only of the -curve that the -curve can possibly cross .
Now , whenevel the -curve meets an -cap , , and has a turning valne .
the -cnrve to meet the -cap from outside , it to ascend on till it meets of a priori the opposite side of the same cap , yhen it passes outside and nnlediatel begins to descend , until it meets another -cap .
The -curve descends eadil except where it is inside a cap ; and soon as the -caps are plotted , the eneral character of the curve is fairly apparent .
If has a definite limit when and this limit is less than , then the last } ) is closed , is ultimately descending and tends to .
But if , then has a limit and ' tends to the ) limit as , and ) approach it ' from above or from elow .
, and is positiv-e , ises a with the top of an -cnp ponds exactly to a of ' But a of 1 occurs where it crosses a descending ) lnaima of occur somewhat later the of alnple , that is periodic , rising every , and .
the level a year .
has a series of with equal lna of , say .
Start from a of ' etween two -caps , suppose the value of to iven .
Then .
to descend .
If is may pass the top of one more -caps , descending ] the while , but , sooner or later , it lneets an -cap , passes it , and begins to rise .
It thereafCer meets } ) , for it leaves each cap at a height less than , and , since it descends } ) the caps , it cannot pass over the top of the next .
year } rises and fnlls , the maxirnum occurring later than that or , and we yearly epidemic .
) thele no reason to make these epidemics equal ; , if two are equal , they all must be .
For if equal values for ints of the periods , the whole coursc these the FIG. 2 .
same ; in other words , if two periods are superposed , and if the two parts of the -curve have one point in common , they coincide , but , if meets any L-cap at a point lower than the preceding cap , its whole course lside that cap is lower , and lower still in the next cap , and so on , tending to a form which is periodic , and which may be the axis itself , the happening 230 Lieut.-Colonel Sir R. Ross and Dr. Hilda P. Hudson .
dying out .
But , if meets any -cap higher than the last , it meets the next higher still , and so on , and tends to a periodic limiting form of finite fluctuations .
In order to plot more closely , we have to replace the fundamental differential equation by the approximate difference equation , where are calculated from the value of measured ofl the given curve , and is given ; or by some moditication of the method .
( v ) rnphical Teatment of is and has been obtained , ws have next to deal with , which can be directly calculated from the equation Hence always increases with : as increases , increases if , and decreases if .
If we wish to know the general behaviour of , the following Table is useful ; it allows for a single bell of each of the and , -curves , the maximum of occurring later than that of In using this , we follow down a column until ) passes through the value , when we cross to the next -column .
Since in each colnmn is increasing in the third row and decreasing iu the seventh , there is bound to be a maximum of between them , which occurs before the maximum of if and after if is seen that if and , : are both decreasing and , we cannot say what is doing , as the two causes of its change act in opposite a priori thometry .
ways .
It is possible for to have a minimum followed by a short rise and a maximum , and , if then becomes equal to 3 , falls again to another 1uinimum .
This would happen , for example , if suddenly became constant after decreasing , before drops .
Then we have a slight recrudescence of the epidemic near its close .
Since and in the case considered and are both ative , and is positiye , so that ises again , provided is small enough , but , decreases , the factor decreases and changes sign ; the second ter in becomes negative , and sinks .
In rises and falls in a symmetrical bell and then remains constant : Lieut.-Colonel Sir R. Ross and .
Hilda P. Hudson .
does the same ; , has an unsymmetrical bell , rising more steeply th.an it falls , and sinking to the same limit as ; and shows an epidemic with slight double recrudescence before sinking to its limit .
The graphs of , and are on the same scale , and that of on 10 times the scale .
A \mdash ; The last sub-section started by assuming that is given in so experimental way .
But is not given except conjecture , and the nlethod is useful to whether any particular con , ecture leads to numerical results which gree with the records of actual epidemics .
Now the most and accul.ate records are the notifications of fresh cases and deaths made by the medical officers of in particular , the records of measles in Aberdeen for the 20 1883-1902 are very suitable material though even there the numbers are not enough to be quite satisfactory .
The death-rates do not correspond to ally of the symbols we have duced so far ; for is the number of deatbs of affected persons per unit time , a different from the number of deaths due the affection .
We return to this in a later section .
We start therefore with the fresh .
assumption } notiiications , that is , the values of , are known .
also iven , either by interpolation census returns , 01 because in a } and llild epidemic is nearly constant .
Then is known a function of .
The fundamental equations are best taken in the forms and can respectively be determined , if reasonable esCimates be made of the constants , R. When , are found and plotted , we can construct raphically .
is the point .
the vertical axis mark off OA , OC ; then AB , AC Let PA meet the axis NO in ; let meet .
PN in and in and let meet in R. Stndy of a priori and so is the point on the c-curve ) '-curve .
If and ) ) , hen In this case , ( the the -curve always In a reat is htly 1 1uell , lore f than it alls . .
( is ) the Ller 1 ) auct for the of the ternl is tive if , and positive if il is , ( the nlit negative .
1 this , the ttll ) souner than of is .
But its , and tho nlllit of .
These results an tenbion of those in .
) where l'onud that in the , , are { is But we do not to , we assunle Then is the ) that rises .
at and a T. Then for asynnnet .
an )( facto1 ; ich still vanishes at and , but a maximum before ; Lieut. and Dr. If , is , let pt , where all these , the ative exponentials decrease the last , ' oscillates about nlea , the phase ) ehind that of by Then the same ) eriod .
( vii ) More \mdash ; In nlany diseases , example those in which tfiere is period of incubation , the infectivity is variable , and depends not only on the absolute but also on the of the , which is the time } ) since the case was first infected .
In order to take this into account , , must be considered as a fnnction of two independent variables , the , and the .
fluctuations of the infectivity of the disease as a whole cause to vary as a function of ; episodes in the course of a .
case cause to vary as a function of , very possibly a discontinuous function .
For example , an tion period of days is represented by taking when , suppose that , all depend on and ; but on only .
Let the nnnlbor of fresh cases between times and Let be 1lumbel of the of cases which survive , having neither died , elnigrated , nor reyerted , at time then of dnration Thus is the of cases time whose xtion .
The total numl ) .
of cases xisting at is .
( 86 ) As , ( the nunnber of is rate of rted at time is ritte ( 87 ) Study of a priori The ndanlental equations now take the forms .
188 These equations not independent .
If 8is integrated with respect to from to , since we have which is equivalent to 88 by virtue of 86 .
These equations they quite unwieldy , ) they are being applied a large number of simpler cases .
If , are independent of and constant for a particular of to , then for any intermediate value of , from 87 and if have other constant values to } ' value of in th latter range , : and so on .
The simplest case is when is constant from to , zero all values of , and independent of , while all the other elements are neglected .
Since there are then no deaths nor recoveries , and is constant .
Then , the suffixes showing the value to be substituted for in , then this difference equation takes the simpler for1n .
The best way of dealing with this equation is to assume a Dirichlet ries for , substitute and equate coefficients .
lient.-Colonel Sir and Dr. Hilda P. Hudson .
then 1\ldquo ; rbitrary , , and the general reduction formula for he is lnce is itive , the bigns the ients alternate ; essentially itive , therefore is posive : and fore ] is negative and is : First ]of .
of 1.\mdash ; It has been thed t death-rate of affected popularion is a constant of this , say , is due to other causes , and is due to the .
If is constant , so is , and the deaths from disease are simply and their ratio to the total population is proportional to therefore increases steadily ; this does not with the records .
It mole leasonable to assume that the deaths from the disease all 0CCU1 altel infeclion and their number is ) to ; let it be ] Then its atio to the total laCio is , and and falls in propol.tion .
This is nnuch ' nearly true .
But now the fundamental equations .
We that the deaths are in addition to deatbs to the ] ' causes , an that they appen beforo cases have had infect an appreciable nunlber othel people .
Then efore W result 01 elimination is , say , where Study priori of which integral is The forDl of this curye depends on the signs of Let \mdash ; If the last term is positiye , , then same ; and if is a positive quantity , les than or is tive , ( begins to decrease , cannot ( 'hangesign and eases steadily to zero as its \mdash ; If the last in , then is positive is tive .
and is positive ; therefol L ositiye , crins to ease , bnt it does not till henc cannot change ' incl c to as '(1 ) , from to in a shal ) with a at In case ( i ) steadily ) of the cnrve , In if to , bn if a when then to its .
If are both igible , a maxintulJl and sinks to as its limit .
( ii ) lShUlll)tion a infectious is that the case ceases to infective a taiu 1 , when also thele ceases to be of death ihe ) remains anently .
To lepresent this , } ) the population to consist of sses : , heal A are unaffected , X infective and immuncz , immune .
For } ) licity , neglect itll " " natural ' ' variations , but supposc there are MX deaths from the disease per unit fresh cases .
The ndalnental equations arc Hence decreases and increases steadily .
Suppose .
Then ; 238 Lieut.-Colonel Sir R. and Dr. Hibda P. Hudson .
which is integra ) in finite form if is , and leads to ]gebraic functions and logarithms .
In the final state , is by a real root ) , etween P and .
A definite number of persons escape the disease ether .
Now , is of the same sign as is positive at first , ovided ( ; in this case X , the total number of infective , rises to a maximum before to is found to be of the salnc s as , which is positive at first if , and steadily decreases .
of the sanJesign as , and , if , this is less than , the of ( if it one ) is therefore earlier than that of or of , that is to , the number of eaths occurq thal ] the height of th epid which is the time of eatest .
of nolifications .
For , let Then .
Let X is stationary whelt or , leadin to approxinlately , the time which before the lnaximum of the mortality .
XIII .
( i ) work , very little attention has been paid to the actual of the constants .
We have rather been which can be use in a reat viety of special cases .
will be seen that the nunnber of dependent variables is , and that ( fluctuation in any one can be ) uGed for in a great many different ways by supposing there is a suitably adjusted ariation in almost any one of the } ) arameters .
And the variation of any one parameter when it is over a range small compared with the mean value ban be represented approximately by of several analytic functions , which would lead to the most diverse if the variation were to continue over a wider .
It is therefore Study ) riori P extremely rash to out priori any oue possible cause of fluctuation any one phenomenon .
The cases already considered have led exactly to the selies of required by the facts : 1 .
The steadily rising curve of a that pernlea s whole population 2 .
The bell-shaped an } idelnic that dies entirely .
The uymetrical bell of ) ) with idemic , and settles down to a steady ) level 4 .
The periodic with ' rise ] XI .
The ) is cencc the of an epidemic , or where diffel j. in ) occur intervals [ XI ( v ) ] .
This that , ( in of , the of epidemics as as we see at ( in ] be explained by the laws of , as studied in this ) The methods which have been used call , , for the rellarks:\mdash ; A. In the tions of Section , the -mortality and infectivity vere both arded as constants .
It is very unlikely that this is strictly true ; and it is much more probable that their values are highest at ) definite short period after infection then gradually die away .
B. In Section XI some attempt is to lneet this objection by intl ducing the case-duration as a second indepcndent variable , in addition the time , but there are innumerable other considerations that may also be taken into account , such as the , sex , and social position of the patients , climate , temperature , and , in the case of insect-borne diseases , the habits the rier .
C. Except in ( ii ) , the whole population has been divided into two classes only , the unaffected , and the affected who considered to be both infective and also immune .
But it is again unlikely that this is strictly , and , in order to represent the facts accurately , a much larger nunlbe ] of classes would have to be considered .
1 ) .
it would be easy to write down equations any given hypothesis as to the number and behaviour of the variables , these equations , like those of XI ( vii ) , would probably be quite unsuited ) numerical work , and we should have to fall back on laborious ) methods such as expansion in series or mechanical quadratures .
Prof T. H. elock . .
The qnantities } have treated continuous functions of the time , , as .
of persons , they can only take values , and diqcontinuous unless they are constant ; in the bame Wfly , and vary nultjples of .
This is a selio } , hen the nnnnbers are large .
In a small village , ge nun ber of cases in the period ] ) two epidemics be only or thYec , and , if these are re1lloyed by any chance , ) hole events is there is no one to keep aliye the infection , a new tion may oorow up entirely non-imlnune .
Then of source infection same effect as if rose to its ) .
Tn , in the case of small populathe 11o ) ofte1l obtain , so that the hole of paper is ) ) than to small ones .
The ace By T. H. ELOCK , , Professor of plied Mathematics in the University of Received January 1 .
The study of the water waves produced by the motion of an signed pressure distribution over the surface has hitherto been limited to the steady state attained when the systenl has been moving with uniform velocity for a very long time .
In his latest series of papers on water waves , Lord Kelvi 1nade an elaborate phical and numerical study problems , and expressed the hope of apply i ' his lnethods to calculate the initiation and continued growth 0 canal vaves due to the sudden commencement and continued application of a moving , steady rface pressure .
Ill the ] ) , I have not attempted any of the surface olevation itself , but I )oceeded directly to the calculation of the corresponding wave .
At the wave resistance is known only for the steady state tain localised pressure systems in mifornl motion , and it se to attempt some estimate of ) time taken to attain his state when we into account the of the motion .
One examine eHecC of initial acceleration , .but I have limited the blem by only the case of a system which is suddenly shed , and is the same instant set in motion with uniform velocity .
Kelvin , ' Math. and Phvs .
Papers , ' , p. 456 ( 1906 ) .

