K iego ae al. J T ansl Med (2021) 19:68
h ps://doi.o g/10.1186/s12967-021-02714-8
RESEARCH
A heo e ical model o heal h managemen
using da a-d i en decision-making: he u u e
o p ecision medicine andheal h
E a K iego a1†, Milos Kudelka2†, Ma in Rad ansky2 and Ji i Gallo3,4*
Abs ac
Backg ound: The bu den o ch onic and socie al diseases is a ec ed by many isk ac o s ha can change o e ime.
The minimalisa ion o disease-associa ed isk ac o s may con ibu e o long- e m heal h. The e o e, new da a-d i en
heal h managemen should be used in clinical decision-making in o de o minimise u u e indi idual isks o disease
and ad e se heal h e ec s.
Me hods: We aimed o de elop a heal h ajec o ies (HT) managemen me hodology based on elec onic heal h
eco ds (EHR) and analysing o e lapping g oups o pa ien s who sha e a simila isk o de eloping a pa icula disease
o expe iencing speci ic ad e se heal h e ec s. Fo mal concep analysis (FCA) was applied o iden i y and isualise
o e lapping pa ien g oups, as well as o decision-making. To demons a e i s capabili ies, he heo e ical model
p esen ed uses genuine da a om a local o al knee a h oplas y (TKA) egis e (a o al o 1885 pa ien s) and shows
he in luence o s ep by s ep changes in i e li es yle ac o s (BMI, smoking, ac i i y, spo s and long-dis ance walking)
on he isk o ea ly eope a ion a e TKA.
Resul s: The heo e ical model o HT managemen demons a es he po en ial o using EHR da a o make da a-
d i en ecommenda ions o suppo bo h pa ien s’ and physicians’ decision-making. The model example de eloped
om he TKA egis e ac s as a clinical decision-making ool, buil o show su geons and pa ien s he likelihood o
ea ly eope a ion a e TKA and how he likelihood changes when ac o s a e modi ied. The p esen ed da a-d i en
ool sui s an indi idualised app oach o heal h managemen because i quan i ies he impac o a ious combina ions
o ac o s on he ea ly eope a ion a e a e TKA and shows al e na i e combina ions o ac o s ha may change he
eope a ion isk.
Conclusion: This heo e ical model in oduces u u e HT managemen as an unde s andable way o concei ing
pa ien s’ u u es wi h a iew o posi i ely (o nega i ely) changing hei beha iou . The model’s abili y o in luence
bene icial heal h ca e decision-making o imp o e pa ien ou comes should be p o ed using a ious eal-wo ld da a
om EHR da ase s.
Keywo ds: P ecision medicine, P ecision heal h, Elec onic heal h eco d, Clinical decision-making ool, Heal h
ajec o y, Ea ly eope a ion, Re ision a e, To al knee a h oplas y, Li es yle ac o s, Fo mal concep analysis
© The Au ho (s) 2021. This a icle is licensed unde a C ea i e Commons A ibu ion 4.0 In e na ional License, which pe mi s use, sha ing,
adap a ion, dis ibu ion and ep oduc ion in any medium o o ma , as long as you gi e app op ia e c edi o he o iginal au ho (s) and
he sou ce, p o ide a link o he C ea i e Commons licence, and indica e i changes we e made. The images o o he hi d pa y ma e ial
in his a icle a e included in he a icle’s C ea i e Commons licence, unless indica ed o he wise in a c edi line o he ma e ial. I ma e ial
is no included in he a icle’s C ea i e Commons licence and you in ended use is no pe mi ed by s a u o y egula ion o exceeds he
pe mi ed use, you will need o ob ain pe mission di ec ly om he copy igh holde . To iew a copy o his licence, isi h p://c ea i eco
mmons .o g/licen ses/by/4.0/. The C ea i e Commons Public Domain Dedica ion wai e (h p://c ea i eco mmons .o g/publi cdoma in/
ze o/1.0/) applies o he da a made a ailable in his a icle, unless o he wise s a ed in a c edi line o he da a.
Backg ound
Long- e m heal h is a delica e combina ion o nu i ion,
li es yle, en i onmen and gene ics, dedica ion o main-
aining and imp o ing one’s heal h, and he assiduous
a oidance o heal h-damaging beha iou s. A heal h a-
jec o y (HT) is a use ul way o po aying he dynamic
Open Access
Jou nal o
T ansla ional Medicine
*Co espondence: ji i.gallo@ olny.cz
†E a K iego a, Milos Kudelka con ibu ed equally
3 Depa men o O hopedics, Facul y o Medicine and Den is y, Palacky
Uni e si y Olomouc, Hne o inska 3, 775 15 Olomouc, Czech Republic
Full lis o au ho in o ma ion is a ailable a he end o he a icle
Page 2 o 12
K iego ae al. J T ansl Med (2021) 19:68
cou se o heal h and disease and p esen s an indi idual’s
heal h as a ac o dependen on ime. The isks o a pa -
icula disease a e in luenced by many ac o s, which may
change in speci ic si ua ions o e ime [1, 2]. Nowadays,
many isk ac o s, as well as o he heal h and/o disease-
ela ed da a, a e de ailed in a hospi al o ou pa ien elec-
onic heal h eco ds (EHR) [3–6]. Inc easingly, pa ien s
a e willing o sha e mo e and be e da a wi h he heal h
ca e sys em. The e o e, he e is an u gen need o de elop
a p ocess o au oma ed analysis o his da a, which could
esul in es ablishing a clinical decision-making ool
(CDMT) as a componen o a clinical decision suppo
sys em (CDSS), which in u n, ul ima ely leads o he
educ ion o he indi idual isks associa ed wi h ce ain
diseases o ad e se heal h e ec s [7–9].
The quali y o decision-making in he e a o p ecision
heal h and p ecision medicine (see desc ip ion o hese
e ms below) is in luenced by h ee g oups o da a ha
a e ela ed o ime. The i s g oup is da a desc ibing
he pa ien ’s cu en condi ion epo ed as a se o ac-
o s in hei EHR. The second g oup is da a ep esen ing
he pa ien ’s his o y, which is (o should be) included in
he EHR, such as he pa ien ’s ini ial condi ion and i s
changes o e he ime p eceding hei cu en condi-
ion. The hi d g oup o da a is ela ed o a desc ip ion
o he pa ien ’s speci ic li ing condi ions and hei u u e
changes, which a e no included in he EHR.
Based on he huge amoun o da a a ailable in EHRs,
including hund eds o demog aphic, labo a o y and
clinical ac o s, he e is an u gen need o de elop com-
pu a ional app oaches and CDMTs based on combina-
ions o pa ien ac o s o suppo decision-making abou
e ec i e heal h and disease managemen [10, 11]. These
app oaches should allow he clinician(s) and pa ien (s)
o e alua e oge he he quali a i e and quan i a i e con-
ibu ions o nume ous ac o s o he medical isk, such
as he disease, ea men esponse, ailu e, complica ion
and/o p ognosis in indi idual pa ien s, as al eady shown
in eal-wo ld coho s [12–14]. Addi ionally, pa ien s
could be in o med abou he impac o pa icula ac-
o s on he likely ou come. The CDMT would allow deci-
sion-making, sha ed be ween pa ien s and clinicians, o
be based on in elligible ecommenda ions. Finally, his
app oach migh modi y pa ien s’ expec a ions, which is a
ac o s ongly a ec ing no only u u e ca e o in e en-
ions [11], bu also hei ou comes. Ne e heless, he e is
a lack o compu a ional app oaches ha could quan i y
he con ibu ion o isk ac o s on heal h [15, 16].
We aimed o de elop an HT managemen s a egy ha
could iden i y and u ilise ac o s ha can a ec , indi idu-
ally o in combina ion, an indi idual’s u u e heal h. The
in oduced heo e ical model was p esen ed using he
clinical egis y da ase p esen ed in ou p e ious s udy
[17], e ealing he posi i e e ec o i e li es yle ac o s
(no mal BMI, non-smoking, ac i i y, spo s and long-dis-
ance walking) on educing he isk o ea ly eope a ion
a e o al knee a h oplas y (TKA). HT managemen
based on con inuous da a-d i en decision-making is a
long- e m s a egy o manage heal h, i espec i e o he
b anch o medicine.
Ma e ial andme hods
HT managemen
Wo king wi h a la ge amoun o pa ien da a in he o m
o EHRs (all he in o ma ion collec ed and a chi ed in
hospi al o ou pa ien elec onic da abases, including
egis ies o pa icula ea men s) and using au oma ed
p ocessing and analysis me hods based on machine
lea ning o , gene ally, on a i icial in elligence should
esul in CDMTs ha can in elligen ly suppo clinicians’
and pa ien s’ decision-making [11]. As men ioned in he
in oduc ion, ou app oach ocuses on he pa ien ’s in lu-
enceable u u e, s a ing wi h hei ini ial condi ion and
his o y sa ed in he EHR. The pa ien ’s u u e is unde -
s ood as a isk (o se o isks) o disease and ad e se
heal h e ec s. Thei op ions o educing he isk a e hen
analysed based on he ac o s ha p obably in luence
hei isk.
Fou assump ions ha e been made as ollows:
1. The pa ien can in luence hei indi idual ac o s (o
a leas some o hem).
2. Each combina ion o selec ed ac o s de ines a g oup
o pa ien s as simila in e ms o hese ac o s, and
hey a e exposed o a simila isk le el.
3. The deg ee o isk o de eloping a disease o medical
condi ion can be quan i ied o each combina ion o
alues o he selec ed ac o s. Combina ions o ac-
o s may o e lap, o one combina ion may be pa o
o he la ge combina ions.
4. The mo e speci ic he combina ion o ac o s con-
side ed, he smalle he g oup o pa ien s. Indi idual
g oups hen di e in hei deg ee o isk. Because
combina ions o ac o s may o e lap, g oups o
pa ien s may also o e lap, o an e en mo e speci ic
g oup o pa ien s may be included in a less speci ic
la ge g oup o pa ien s. This is he mos impo an
assump ion; i is a consequence o he p e ious h ee
assump ions and is ela ed o he ac o s ha will be
examined.
These assump ions unde pin he logic o which
g oups he pa ien belongs o. A he same ime, hanks
o a change in he ac o s examined and in luenced
by he pa ien s, hey can mo e o ano he g oup wi h
a lowe (o highe ) le el o isk. Due o he complex
Page 3 o 12
K iego ae al. J T ansl Med (2021) 19:68
ela ionships be ween g oups, he e a e always mo e
op ions o mo ing o a g oup wi h less (o highe ) isk.
Mo eo e , he mo e owa ds a lowe isk may o may
no be a one-o . On he con a y, i is assumed ha i
will be epea ed o e ime wi h a clinician’s o physi-
o he apis ’s possible supe ision. This c ea es a p ocess
called ‘heal h ajec o y managemen ’. The indi idual
s eps o his p ocess pe o med o e ime c ea e a a-
jec o y leading, ideally, o a con inuous educ ion o
isk and/o an imp o emen in heal h cha ac e is ics.
Implemen ing hese ac o s on he pa icula pa ien (s)
can a ec he ou comes o he apeu ic in e en ions, a
leas in pa , ia he educ ion o ha m associa ed wi h
an in e en ion. The e o e, all s akeholde s (pa ien s,
hei physicians, clinical se ings and insu ance sys-
ems) may p o i om such app oaches.
Al hough his is a ask ha is gene ally e y complex in
scope and con en , he essence o HT managemen can
be shown in a simple and unde s andable example, which
will be gi en below. This example is based on he esul s
o p e iously published esea ch [17], which showed he
posi i e e ec o i e ac o s (no mal BMI, non-smoking,
ac i i y, spo s and long-dis ance walking) on educing
he isk o ea ly TKA eope a ion. Fo his s udy on HT
managemen , a da ase wi h amilia pa ien s was used
whe e hei condi ions we e known be o e unde going
TKA su ge y, and whe he hey unde wen ea ly eop-
e a ion and he i e ac o s we e bina ised ( o example,
non-smoke s/ex-smoke s = ze o, smoke s = one).
As men ioned abo e, an essen ial equi emen o
he analysis is ha he e can be o e lapping g oups o
pa ien s cha ac e ised by ag eemen in he ac o s s ud-
ied. A adi ional me hod ha can de ec o e lapping
combina ions o bina y ac o s and o e lapping g oups is
o mal concep analysis (FCA), which esul s in a isual
s uc u e desc ibing o e lapping g oups (clus e s), he
so-called ‘concep la ice’ [18–20]. The use o bina y ac-
o s and a concep la ice may seem o be limi ing ac o s
in his app oach, bu his is no he case. The concep la -
ice o e s a simple in oduc ion o HT managemen . Fo
non-bina y ac o s, i is possible o use one o he o e -
lapping clus e ing me hods o he same pu pose [21, 22].
HT managemen is no ocused on a one- ime p edic-
ion: i s usual goal is o place he pa ien in o a g oup
wi h common cha ac e is ics and subsequen ea men .
The pu pose o HT managemen is o in e ac wi h he
explo a i e in e ace o CDMT epea edly and mo e
pa ien s who, in e ms o hei condi ion and his o y,
belong o one g oup in o one o he mo e speci ic g oups
wi h less isk. HT managemen a ge s ac o s ha can be
a ec ed ei he by he pa ien o hei physicians and ha ,
indi idually o in combina ion, can posi i ely o nega-
i ely in luence he pa ien ’s u u e condi ion.
Analy ical model: g ouping pa ien s in oo e lapping
clus e s
FCA and he concep la ice a e used o illus a e ou
app oach (see Addi ional ile1: Tables S1 and S2 and
Figu e S1). Fi s , pa ien g oups a e o med in o a con-
cep la ice using FCA [23–25] applied o a selec ed se
o bina y ac o s. In his eal-wo ld example, he e we e
i e p eope a i e li es yle ac o s. The o mal concep s
a e clus e s ha indica e ela ionships hidden in he
da a among pa ien s wi h a common subse o li es yle
ac o s. Concep s a e de i ed om he able con aining
pa ien ac o s called ‘con ex ’ [26]. By o de ing con-
cep s, a ma hema ical s uc u e called a concep la ice
is ob ained ha desc ibes ela ionships be ween indi-
idual g oups (concep s) o pa ien s wi h a sha ed se o
ac o s. Such a s uc u e enables he isualisa ion o he
concep s in hie a chical o m. Fo example, he concep
can be a g oup o pa ien s who a e non-smoke s and
ha e a BMI < 30, ega dless o o he ac o s. When physi-
cal ac i i y is added o hese wo ac o s, a new, smalle
(and mo e speci ic) concep con aining a g oup o non-
smoking pa ien s wi h a BMI < 30 who also pa icipa e in
physical ac i i y will be o med.
Thus, we can c ea e a sequence in which he i s con-
cep con ains he mos pa ien s, and he las con ains he
leas pa ien s. The smalle he concep , he mo e speci ic
i is and he mo e simila he pa ien s a e. Each o hese
concep s (clus e s) ca ies a di e en likelihood o isk.
Quan i ying con ibu ions o changes inmodi iable clinical
ac o s
I he in es iga ed ac o s a e modi iable, hen a sequence
o concep s could be iden i ied in which (i) he ac o s in
he p eceding concep a e also con ained in he succeed-
ing concep o he sequence, (ii) pa ien s in he succeed-
ing concep a e con ained in he p eceding concep o he
sequence and (iii) he likelihood o isk in he succeed-
ing concep is lowe han in he p eceding concep . All
such sequences can hen be unde s ood as possible HTs
because hey con inually educe he isk’s likelihood. In
each o hese ajec o ies, he i s clus e desc ibes he
pa ien ’s cu en condi ion, and in subsequen clus e s o
he ajec o y, he numbe o pa ien s dec eases and he
modi iable ac o s ha con ibu e o he expec ed ou -
comes in he u u e inc ease. On he o he hand, i he
pa ien ’s condi ion changes o he p e ious clus e o he
ajec o y, hei isk’s likelihood inc eases.
Tha said, some concep s can con ain e y ew pa ien s
and ze o isk e en s. The empi ical p obabili y o eop-
e a ion is ze o in hese cases. To e lec ha eope a ion
may also occu in hese small g oups, some unce ain y
was in oduced in o he da ase be o e u he analysis
Page 4 o 12
K iego ae al. J T ansl Med (2021) 19:68
was pe o med ( o mo e de ails, see he Addi ional
ile1). This sligh ly changed he empi ical p obabili y. The
modi ica ion did no change he eope a ion p obabili y
o he en i e da ase . Fo a highe empi ical p obabili y
han o al, eope a ion p obabili y dec eases, and o a
lowe empi ical p obabili y han o al, eope a ion p ob-
abili y inc eases o he indi idual concep s. This is also
ue o he ze o alue ha also inc eased he p obabili y.
Visualisa ion o HT ajec o ies
Concep s o each pa ien subg oup and he ela ionships
be ween hem can be isualised as a weigh ed-di ec ed
ne wo k. The e ices (ci cles) o he ne wo k ep e-
sen indi idual concep s. The di ec ed edges (a ows)
ep esen whe he he isk o eope a ion inc eased
o dec eased a e adding a ac o . The g oups o con-
cep s connec ed in a sequence by a ows ep esen HT
wi h g adually added ac o s ha dec ease he isk o
eope a ion.
The size o he e ices (concep s) co esponds o he
isk o eope a ion. The same holds o e ex labels wi h
ac o s. The edge (a ow) s eng h and i s label co e-
spond o he educ ion o eope a ion isk a e adding a
posi i e ac o (indica ing how much he isk o eope a-
ion would be educed). On he o he hand, emo ing a
posi i e ac o may be unde s ood as adding a nega i e
ac o , leading o a highe isk o eope a ion. The col-
ou s o he e ices and edges indica e he eliabili y o
he es ima ion. Concep s ( e ices) con aining a leas
en pe cen o pa ien s a e g een, concep s ha ha e an
o iginal empi ical p obabili y equal o ze o a e ed, and
o he concep s a e yellow. The en pe cen h eshold was
selec ed based on he size o he da ase o d aw a en ion
o he lowe eliabili y o he ecommenda ions when
examining isualised HT and in e ac ing wi h a CDMT.
Heal h ajec o y example
Ou app oach was applied o a eal-wo ld coho o
pa ien s wi h TKA, and he con ibu ion o modi iable
li es yle ac o s o he isk o eope a ion was e alua ed.
Ou me hodology o HT managemen consis s o h ee
componen s: (i) con ex , leading o he de ini ion o a
medical p oblem and isk e en , acquisi ion and e alua-
ion o pa ien da a, (ii) an analy ical da a model iden i y-
ing isk ac o s and possible HTs, and (iii) implemen ing
he model in a CDMT and p o iding a use in e ace o
suppo clinical decision-making (see Fig.1).
Da ase (pa ien coho )
To p esen ou model, an unselec ed eal-wo ld coho
o 1885 pa ien s (695 men and 1190 women) who unde -
wen TKA su ge y be ween Sep embe 2010 and Ap il
Fig. 1 Scheme o gene al heal h ajec o y (HT) managemen . HT managemen consis s o h ee s eps: (1) con ex leading o he de ini ion
o a medical p oblem and isk e en , acquisi ion and e alua ion o pa ien da a; (2) an analy ical da a model based on da a analysis, analysis
o ac o s associa ed wi h isk e en s, iden i ica ion o isk ac o s associa ed wi h isk e en s and a da a model o a CDMT; and (3) CDMT o
pa ien managemen based on a pa ien ’s pe sonal cha ac e is ics. Newly gene a ed pa ien da a can en e he da a modelling s ep, e ining he
assessmen o he likelihood o a medical e en
Page 5 o 12
K iego ae al. J T ansl Med (2021) 19:68
2017 a a single e ia y o hopaedic cen e was ana-
lysed. Fo all pa ien s, he li es yle and clinical ac o s
be o e TKA su ge y, as well as in o ma ion ega ding
ea ly eope a ion (de ined as less han wo yea s a e
p ima y su ge y), we e a ailable in he clinical egis e .
Based on di e en eope a ion a es in younge and olde
pa ien s, subg oups we e c ea ed based on he median
numbe o eope a ions in he male and emale g oups
(younge emales ≤ 71 yea s, olde emales > 71 yea s;
younge males ≤ 66yea s, olde males > 66yea s), espec-
i ely [17]. Fo clinical and li es yle ac o s in he en olled
pa ien s and gende and age subg oups, see Table1 and
Addi ional ile1: TableS3.
In es iga ed li es yle ac o s
To demons a e he capabili ies o ou model, he ollow-
ing p eope a i e ac o s we e included: physical ac i -
i y, spo s ac i i y, smoking, body mass index (BMI)
and he abili y o walk long dis ance (1000m). Physical
ac i i y was e alua ed using he Uni e si y o Cali o -
nia Los Angeles (UCLA) ac i i y scale [27]. In e ms o
UCLA, an inac i e pa ien was one who epo ed no o
low physical ac i i y (ca ego ies one o h ee). An ac i e
pa ien (ca ego ies ou o six) epo ed egula pa ici-
pa ion in mild (walking) o mode a e ac i i ies, such as
swimming, unlimi ed housewo k o shopping. A high
deg ee o ac i i y was de ined as ca ego ies se en o en,
acco ding o UCLA. Spo s ac i i y was e alua ed based
on he pa ien s’ subjec i e es ima ions o hei pa icipa-
ion in spo , dis inguishing be ween none, ec ea ional,
compe i i e and p o essional pe o mance le els. A BMI
(calcula ed as weigh in kilog ams di ided by heigh in
squa e me es) o 30 o o e was conside ed obese (obe-
si y I: BMI 30–35; obesi y II: BMI > 35).
The indi idual ac o s we e bina ised as (i) no physi-
cal ac i i y (UCLA ca ego ies ≤ ou ) e sus physical
ac i i y (UCLA ca ego ies > ou , pe o ming unlimi ed
housewo k and shopping), (ii) no spo s ac i i y e sus
spo s ac i i y ( ec ea ional, compe i i e and p o essional
pe o mance le els), (iii) smoking e sus non-smoking
(including ex-smoke s) and (i ) no mal/o e weigh
(BMI < 30) e sus obese (BMI ≥ 30).
Resul s
Obse ed concep s inmales and emales
Table1 shows he demog aphic and li es yle ac o s o
a eal-wo ld pa ien coho wi h TKA om he clini-
cal egis e o join eplacemen s used o es ing ou
app oach. The concep s we e calcula ed o younge and
olde emales (see Addi ional ile1: Tables S4 and S5),
and younge and olde males (see Addi ional ile1: Tables
S6 and S7) sepa a ely as o he ac o s in luence he a e
o eope a ions in each pa ien subg oup. The sequences
o concep s associa ed wi h educing he likelihood o
eope a ion in TKA pa ien subg oups a e shown in
Addi ional ile1: Figu e S2. Fo each concep , he numbe
o pa ien s in he concep , he numbe o pa ien s who
unde wen ea ly eope a ion, he pe cen age o p obabili-
ies, including he empi ical p obabili y o eope a ion in
he concep , and he gi en unce ain y a e p esen ed.
Table 1 Demog aphic andli es yle pa ame e s in heTKA pa ien coho
TKA: o al knee a h oplas y; BMI: body mass index; UCLA: Uni e si y o Cali o nia Los Angeles; NA: no a ailable.
a one pa ien wi h UCLA high (7–10) included
Pa ame e s Value Younge emales
(≤ 71yea s)
N = 670
Olde emales
(> 71yea s)
N = 520
Younge males
(≤ 66yea s)
N = 275
Olde males
(> 66yea s)
N = 420
N % N % N % N %
BMI [kg/m2] < 30 235 35.1 261 50.2 116 42.1 237 56.4
31–35 226 33.7 178 34.2 89 32.4 147 35.0
> 35 209 31.2 81 15.6 70 25.5 36 8.6
Smoking No 538 80.3 479 92.1 166 60.4 298 71
S op 56 8.4 23 4.4 56 20.4 90 21.4
Yes 76 11.3 18 3.5 53 19.3 32 7.6
UCLA ac i i y No/low (1–3) 545 81.3 460 88.5 181 65.8 325 77.4
Middle (4–6) 125 18.7 60 11.5 94a34.2 95 22.6
Spo ac i i y No 607 90.6 483 92.9 216 78.5 342 81.4
Ac i e 43 6.4 24 4.7 49 17.8 56 13.3
NA 20 3.0 13 2.5 10 3.6 22 5.2
Reope a ion No 643 96.0 495 95.2 245 89.1 393 93.6
Yes 27 4.0 25 4.8 30 10.9 27 6.4

Page 6 o 12
K iego ae al. J T ansl Med (2021) 19:68
As an example, obse ed concep s in olde women
will be discussed (Addi ional ile1: TableS5). The i s
ow o Addi ional ile1: TableS5 is a concep con aining
olde women wi h no common ac o s. The pe cen age o
eope a ions in his concep (and, hus, he o al p opo -
ion o eope a ions among olde women) is 4.98%. The
nex ows show he pe cen age o eope a ions in each
subg oup de ined by combina ions o ac o s. Fo exam-
ple, adding he ac i i y ac o , ega dless o BMI, smok-
ing, spo and long-dis ance walking, p oduces a smalle
g oup wi h 11.75% o olde women and mo e p ecise
in o ma ion on he eope a ion a e (3.39%).
Da a analysis andou pu s o  heCDMT
To ob ain in o ma ion abou he isk o ea ly eope a-
ion in a pa ien be o e he p ima y TKA, a CDMT was
de eloped using an app op ia ely s uc u ed and ali-
da ed da ase . In s ep one, he pa ien ypes hei gende
and age in o he CDMT. The o e all eope a ion a e in
pa ien s wi hin ha gende and age g oup will be p e-
sen ed based on eal-wo ld da a om a egis y o o al
join a h oplas y. In s ep wo, he pa ien selec s hei
p eope a i e ac o s: non-smoking s a us (Y/N), ac i i y
(Y/N), abili y o walk 1000m (Y/N) and spo s ac i i y
(Y/N). The CDMT shows he pe cen age o pa ien s wi h
he same p eope a i e ac o s o TKA, and he eope a-
ion a e in his pa ien g oup based on he egis y da a.
In a u he s ep, he pa ien could add indi idual ac o s
ha hey wish o change p io o he p ima y su ge y,
and he CDMT calcula es he size o he g oup and he
eope a ion a es based on he co esponding g oup o
pa ien s wi h hose ac o s. The CDMT can p o ide he
pa ien wi h in o ma ion abou how o posi i ely change
hei le el o isk and p omo e con idence in aking ha
s ep. A e es ing he impac o indi idual ac o s, com-
bina ions could be es ed. The pa ien may choose o
modi y he ac o s: o example, hose ha esul in he
lowes eope a ion a es and/o hose ha hey can in lu-
ence hemsel es.
To show he p ac ical ou pu o ou CDMT, examples
o wo TKA pa ien s will be p esen ed: a non-smoking
olde woman and an olde man who smokes. The CDMT
shows he bes combina ions o posi i e ac o s, as well as
he o de o changes needed o achie e he bes ou come
( he lowes isk o ea ly eope a ion). To gain ull insigh
in o he calcula ions, all he da a is shown, e en when he
di e ence in he likelihood o eope a ion by changing a
pa icula ac o o hei combina ions is ela i ely small.
Ne e heless, e en small changes may mo e he pa icu-
la pa ien in o he g oup wi h be e o wo se ou comes.
Case s udy A: Woman, 78yea s old, non-smoke , no
ac i i y (limi ed housewo k, no shopping), no long-dis-
ance walking, a BMI o 36, no spo s ac i i y.
The e ision a e in he whole g oup o olde women
is 4.98% (see Fig.2 and Addi ional ile1: TableS5). In
his g oup, only 11% o women we e physically ac i e
Fig. 2 Sequence o concep s associa ed wi h educing he likelihood
o eope a ion in a pa icula woman (shown in colou : 78 yea s
old, non-smoke , no ac i e, no long-dis ance walking, BMI o 36,
no spo s ac i i y). A ep esen a i e example o a CDMT based on
eal-wo ld da a. The edge (a ow) s eng h and i s label co espond
o he educ ion o he isk o eope a ion a e adding a ac o
(pe cen age o how much he isk o eope a ion would be educed).
The same holds o he e ex labels wi h ac o s and he numbe s
o pa ien s. Me hods o educing he likelihood o eope a ion in his
speci ic case a e colou ed ligh g een, and he mos e ec i e me hod
is shown in da k g een. Posi i e ac o s we e ac i i y (Ac i i y),
long-dis ance walking (LongDis Walk), no smoking (NoSmoking),
a BMI < 30 (lowBMI) and no posi i e ac o s p esen (NO COMMON
FACTORS). The colou o he p esen ed case changes ( om ed o
o ange hen g een) as he p obabili y o eope a ion dec eases
Page 7 o 12
K iego ae al. J T ansl Med (2021) 19:68
using UCLA’s classi ica ion, 7% epo ed spo s ac i i y,
14% could walk 1000m, abou 50% had a BMI o below
o equal 30, and 92% we e non-smoke s. The sequences
o concep s associa ed wi h educing he likelihood o
eope a ion in olde women a e shown in Fig.2.
A e adding non-smoking, which is he only p eop-
e a i e ac o educing he likelihood o eope a ion
o his pa icula woman, he CDMT calcula es he
p obabili y o a e ision a e o 4.91%. A e including
ano he indi idual posi i e ac o o a combina ion
o ac o s o his woman, he CDMT calcula es he
likelihood o eope a ion and co esponding imp o e-
men when hose ac o s a e modi ied (Fig.3).
Fo his woman, he e a e h ee sugges ed ways o
educe he likelihood o eope a ion. Fi s , when add-
ing spo s ac i i y, he likelihood o eope a ion lowe s
by 87% o a e ision a e o 0.62%. Howe e , obese peo-
ple wi h no o low physical ac i i y canno suddenly be
expec ed o s a spo s ac i i y p io o TKA su ge y.
This would no be easible o his pa icula woman. The
second is o add long-dis ance walking (1000m), which
may lowe he p obabili y o eope a ion by 37% ( o a
e ision a e o 3.10%). Howe e , i may be di icul o
Fig. 3 The ou pu o he clinical decision-making ool (CDMT) o he olde woman (78 yea s old, a BMI o 36, no ac i i y, no spo , non-smoking)–a
ep esen a i e example. The sc eens show a he e ision a e in he whole g oup o olde women; b he likelihood o e ision a e in a pa icula
olde woman, based on he li es yle pa ame e s; c he likelihood o e ision a e and imp o emen s a e adding physical ac i i y o his
pa icula woman ( educ ion o he likelihood o eope a ion by 29%); d he likelihood o e ision a e and imp o emen s a e adding physical
ac i i y + BMI < 30 o his pa icula woman (likelihood o eope a ion educed by 45%)
Page 8 o 12
K iego ae al. J T ansl Med (2021) 19:68
s a long-dis ance walking in he case o a woman wi h
a se e e os eoa h i ic knee and no o low ac i i y (see
Fig.2). The hi d way seems o be he mos easible o
his pa icula women: she may s a wi h physical ac i -
i y in he o m o unlimi ed housewo k and shopping,
which will lowe he p obabili y o eope a ion by 29% ( o
a e ision a e o 2.90%). I his is ollowed by lowe ing
he BMI, he combina ion o hese ac o s may u he
dec ease he p obabili y o eope a ion by 16% ( o a e i-
sion a e o 2.04%). Fo a emale pa ien wi h a BMI < 30,
who ollows hese ecommenda ions and becomes ac i e,
he e ision a e educes o 0.15% by including long-dis-
ance walking.
Case s udy B: Man, 75yea s old, smoke , no ac i i y, a
BMI o 33, no spo s ac i i y.
In he g oup o olde men, only 23% we e physically
ac i e in e ms o UCLA’s classi ica ion, 19% epo ed
spo s ac i i y, 22% could walk 1000m, abou 50% had a
BMI o below 30, and 71% we e non-smoke s. This olde ,
obese man (smoke , no physical ac i i y, no spo s ac i -
i y) has no p eope a i e ac o s educing he likelihood
o eope a ion, meaning he p obabili y o eope a ion is
6.72% (see Fig.4). Fo his man, he e a e h ee sugges ed
ways o p eope a i ely educe he likelihood o eope a-
ion: adding no-smoking, long-dis ance walking (1000m)
and lowe ing his BMI ia a die o ope a i ely (see Fig.4).
S ep by s ep, he bes me hod o his man o imp o e
his chances o a oiding eope a ion a e o s op smoking
(imp o emen o 6%), hen s a ing o walk longe dis-
ances (imp o emen o 16%) ollowed by lowe ing his
BMI (imp o emen o 20%). By ollowing hese s eps, he
man’s likelihood o eope a ion is educed o 4.31% (see
Fig.4).
Addi ionally, he model can also isualise wha happens
i a nega i e ac o is added. Take, o example, a 75-yea -
old man indica ed o TKA. He is an ex-smoke , who
does no ac i i y o spo s ac i i y wi h a BMI o 33 who
s a s smoking. By s a ing smoking, his pa ien ’s likeli-
hood o eope a ion inc eases om 6.29 o 6.72 (a de e-
io a ion o 6%).
Discussion
We in oduced he concep o HT managemen based on
analysing he ela ionships be ween modi iable ac o s
and he deg ee o medical isk. We ha e shown how he
con ibu ions o indi idual ac o s o hei combina ions
ha educe u u e medical isk can be iewed in de ail
based on he pa ien ’s condi ion. A signi ican ad an-
age o his app oach is he au oma ed suppo h ough a
CDMT, which o e s al e na i e decisions and aces how
he choice o an al e na i e c ea es an HT o educe isks
g adually.
This esea ch ocuses p ima ily on he possibili ies
o in luencing he pa ien ’s u u e conce ning ac o s
ha p o ably a ec hei heal h and isk o disease [28].
Acco dingly, he pa ien s can in luence, a leas in pa ,
Fig. 4 Concep s associa ed wi h educing he likelihood o
eope a ion in a pa icula man (shown in colou : 75 yea s old,
smoke , no ac i e, no long-dis ance walking, BMI o 33). A
ep esen a i e example o a CDMT based on eal-wo ld da a. Men
and women a e expec ed o unde ake di e en physical ac i i ies.
The edge (a ow) s eng h and i s label co espond o he educ ion
o he isk o eope a ion a e adding a ac o (pe cen age o how
much he isk o eope a ion would be educed). The same holds
o e ex labels wi h ac o s and he numbe s o pa ien s. Me hods
o educing he likelihood o eope a ion in his speci ic case a e
colou ed ligh g een, and he mos e ec i e me hod is shown in da k
g een. Posi i e ac o s we e ac i i y (Ac i i y), long-dis ance walking
(LongDis Walk), no smoking (NoSmoking), a BMI < 30 (lowBMI) and
no posi i e ac o s p esen (NO COMMON FACTORS). The colou o
he p esen ed case changes ( om ed o o ange hen g een) as he
p obabili y o eope a ion dec eases
Page 9 o 12
K iego ae al. J T ansl Med (2021) 19:68
some o he ac o s on hei HT by deciding o change
hei beha iou and li es yle. The e o e, he managemen
o u u e HT should be dependen on a pa icula disease
wi h g owing pa icipa ion om he pa ien . The CDMT
se es o suppo he clinician’s and pa ien ’s decisions.
In ou app oach o HT managemen using FCA, we wo k
only wi h speci ic isk and a se o bina y ac o s, a leas
some o which can be in luenced by a change in pa ien
beha iou a pa icula ime poin s in hei li e. Al hough
he bina isa ion o ac o s may appea o be limi ing,
EHR ac o s a e o en inhe en ly bina y (posi i e e sus
nega i e) o can be easily and na u ally bina ised (such
as non-smoke /ex-smoke e sus smoke , BMI ≤ 30 e -
sus BMI > 30). Fo mo e complex asks, nume ical ac-
o s and o e lapping clus e ing me hods can be used
[21, 22]. Thanks o he isual hie a chical o m, FCA
p o ides well-explained and in e p e able ou comes and
enables he nume ical calcula ion o an e en ’s p obabil-
i y o occu ence wi hin a clus e [29, 30]. This allows he
deg ee o isk o be easily quan i ied o di e en combi-
na ions o ac o s and p oposes selec i e ajec o ies o
educe isk. A signi ican ad an age o his app oach is
he use o o e lapping clus e s, hus p o iding pa ien (s)
wi h mo e op ions o educe hei quan i iable isk and
conside how o educe hei isk in he longe e m.
These ea u es a e no a ailable using adi ional me h-
ods, such as p edic ion in he meaning o classi ica ion o
non-o e lapping clus e ing.
This me hodology was applied o a eal-wo ld da a-
se om o hopaedics, showing he in luence o li es yle
ac o s on he isk o TKA eope a ion in a coho o
1885 pa ien s om a egis y o TKAs [17]. In his case,
he pa ien ’s condi ion is unde s ood as a se o ac o s
eco ded in he egis e o TKAs ( he EHR). A leas some
o he ac o s a e assumed o be modi iable by he pa ien .
Mo eo e , any combina ion o ac o s de ines a g oup
o pa ien s as simila in e ms o hese ac o s. Fo each
combina ion o selec ed ac o s, he deg ee o medical
isk can be quan i ied, and combina ions o hese ac o s
may o e lap. As a a ge ool based on his app oach, a
use - iendly CDMT was c ea ed ha implemen s he
HT managemen model. Using he CDMT, he pa ien
can make decisions abou hei sho - e m o long- e m
u u e, ei he by hemsel es o unde he supe ision o
hei clinician o physio he apis . Thanks o he isuali-
sa ion o one o mo e HT, decisions can be made wi h
a longe - e m expec a ion. The clinical ele ance o he
obse ed esul s and cu en o hopaedic opinion a e
discussed in de ail in he Addi ional ile1. Impo an ly,
his model may be adap ed o local da a de i ed om
he hospi al in which he pa ien will be ope a ed on, hus
es ablishing pa ien expec a ions based on local eal-
wo ld pa ien da a. We a e awa e ha a CDMT based on
da a om o he TKA/hospi al egis e s may o e o he
esul s, as he pa ame e s may be in luenced by o he ac-
o s, such as gene ic backg ound, li es yle, en i onmen
and he local heal h ca e sys em, con ibu ing o he
pa ien ou come.
The p esen ed model o HT managemen can be
b oadly applied. In he e a o p ecision medicine and
heal h, i is c ucial o iden i y c i ical ac o s ha signi i-
can ly inc ease o educe heal h isk(s) in all b anches o
medicine [10, 31]. The essen ial ask, no jus in o ho-
paedics, is o iden i y he bes -sui ed he apy o an indi-
idual pa ien , as well as o minimise he ha m associa ed
wi h a pa icula in e en ion, because e en well-es ab-
lished he apeu ic in e en ions ha e been ques ioned
in he las ew decades [14, 32]. Se e al o he examples
in he li e a u e iden i y isk ac o s o a ious diseases,
such as diabe es [33, 34], ca dio ascula disease [35, 36],
b eas [37, 38] and lung [39, 40] cance and many o he s
ha a e s aigh o wa dly applicable o HT managemen ,
as shown in Table2.
The ad an age o his and o he da a-d i en app oaches
is ha i , o example, he impac o di e en ac o s on
heal h a ies in di e en egions, hen HT managemen
Table 2 Examples o possible uses o HT managemen
Disease Modi iable nega i e ac o s
Diabe es [33, 34] Being o e weigh o obese, physical inac i i y, high blood p essu e, high choles e ol, obacco smoking, unheal hy
ea ing, hea y alcohol consump ion
Ca dio ascula disease [35, 36] Being o e weigh o obese, physical inac i i y, unheal hy ea ing, alcohol consump ion, smoking, high blood p essu e,
diabe es, sodium in ake
B eas cance [37, 38] Long- e m use o combina ion ho mone eplacemen he apy (oes ogen–p oges in), obesi y, alcohol consump ion,
la e p egnancy o ne e being p egnan , nigh -shi wo k, physical inac i i y
Lung cance [39, 40] Smoking (ciga e e, ciga and pipe), second-hand smoking, be a ca o ene supplemen s in hea y smoke s, alcohol
consump ion, exposu e o chemicals, ai pollu ion
TKA eope a ion [17] BMI, smoking, low ac i i y, no spo s, no long-dis ance walking
Au oimmune diseases [47] Being obese, smoking, unheal hy ea ing, physical inac i i y, exposu e o ce ain in ec ions, ce ain medica ions, expo-
su e o oxic agen s