Ap il 2017 | Volume 8 | A icle 3841
O iginal esea ch
published: 05 Ap il 2017
doi: 10.3389/ immu.2017.00384
F on ie s in Immunology | www. on ie sin.o g
Edi ed by:
Isabelle Mey s,
KU Leu en, Belgium
Re iewed by:
Es he De V ies,
Tilbu g Uni e si y, Ne he lands
Filomeen Hae ynck,
Ghen Uni e si y, Belgium
Isabella Quin i,
Sapienza Uni e si y o Rome, I aly
*Co espondence:
Lo enz G igull
g igull.lo enz@mh-hanno e .de
Special y sec ion:
This a icle was submi ed o
P ima y Immunode iciencies,
a sec ion o he jou nal
F on ie s in Immunology
Recei ed: 16No embe 2016
Accep ed: 17Ma ch2017
Published: 05Ap il2017
Ci a ion:
MückeU, KlemannC, BaumannU,
Meye -Bahlbu gA, Ko umX,
KlawonnF, Lechne WM and G igullL
(2017) Pa ien ’s Expe ience in
Pedia ic P ima y Immunode iciency
Diso de s: Compu e ized
Classi ica ion o Ques ionnai es.
F on . Immunol. 8:384.
doi: 10.3389/ immu.2017.00384
Pa ien ’s expe ience in Pedia ic
P ima y immunode iciency
Diso de s: compu e ized
classi ica ion o Ques ionnai es
U s Mücke1, Ch is ian Klemann2, Ul ich Baumann3, Almu Meye -Bahlbu g4,
Xiaowei Ko um5, F ank Klawonn5,6, We ne M. Lechne 7 and Lo enz G igull1*
1 Depa men o Pedia ic Hema ology and Oncology, Hanno e Medical School, Hanno e , Ge many, 2 Depa men o
Pedia ic Su ge y, Hanno e Medical School, Hanno e , Ge many, 3 Depa men o Pedia ic Pulmonology, Hanno e Medical
School, Hanno e , Ge many, 4 Depa men o Pedia ics, Uni e si y Medicine G ei swald, G ei swald, Ge many, 5 Helmhol z
Cen e o In ec ion Resea ch, B aunschweig, Ge many, 6 Os alia Uni e si y o Applied Sciences, Wol enbue el, Ge many,
7 Imp o ed Medical Diagnos ics, IMD GmbH, Hanno e , Ge many
in oduc ion: P ima y immunode iciency diso de s (PIDs) a e a he e ogeneous g oup
o mo e han 200 a e diseases. Timely diagnosis is o u e mos impo ance. The e o e,
we aimed o de elop a diagnos ic ques ionnai e wi h compu e ized pa e n- ecogni ion
in o de o suppo physicians o iden i y suspicious pa ien his o ies.
Ma e ials and me hods: S anda dized in e iews we e conduc ed wi h gua dians o
child en wi h PID. The ques ionnai e based on pa en al obse a ions was de eloped
using Colaizzis’ amewo k o con en analysis. Answe s om 64 PID pa ien s and 62
con ols we e analyzed by da a mining me hods in o de o make a diagnos ic p edic ion.
Pe o mance was e alua ed by k- old s a i ied c oss- alida ion.
esul s: The diagnos ic suppo ool achie ed a diagnos ic sensi i i y o up o 98%.
The analysis o 12 in e iews e ealed 26 main phenomena obse ed by pa en s in he
p e-diagnos ic pe iod. The ques ions we e sys ema ically ph ased and selec ed esul ing
in a 36-i em ques ionnai e. This was answe ed by 126 pa ien s wi h o wi hou PID o
e alua e p edic ion. I em analysis e ealed signi ican ques ions.
Discussion: Ou app oach p o ed sui able o ecognizing pa e ns and hus di e en-
ia es be ween obse a ions o PID pa ien s and con ol g oups. These indings p o ide
he basis o de eloping a ool suppo ing physicians o conside a PID wi h a ques ion-
nai e. These da a suppo he no ion ha pa ien ’s expe ience is a co ne s one in he
diagnos ic p ocess.
Keywo ds: p ima y immunode iciency disease, da a mining, diagnos ic suppo , ques ionnai e, colaizzi
Abb e ia ions: PID, p ima y immunode iciency diso de ; SVMs, suppo ec o machines, RFs, andom o es s; LR, logis ic
eg ession; NB, nai e Bayes classi ie s; LD, linea disc iminan ; NNs, analysis and nea es neighbo classi ie s.
Table 1 | spec um o diseases in he in e iews: 12 in e iews wi h
pa en s o child en su e ing om di e en diseases we e conduc ed.
in e iew ca ego y Disease
A P edominan ly an ibody de iciencies CVID
B P edominan ly an ibody de iciencies CVID
C Well-de ined synd omes wi h
immunode iciency
A axia eleangiec asia
D P edominan ly an ibody de iciencies CVID
E Complemen de iciencies C2 de iciency
F Combined immunode iciencies Ce nunnos
G Unde ined immunode iciency Combined immunode iciency
wi h di e en cy openia
H Well-de ined synd omes wi h
immunode iciency
Nijmegen b eakage synd ome
I Congeni al de ec s o phagocy e
numbe , unc ion, o bo h
X-linked ch onic
g anuloma ous disease
J Unde ined immunode iciency Unde ined se e e
immunode iciency
K P edominan ly an ibody de iciencies CVID
L Well-de ined synd omes wi h
immunode iciency
Nijmegen b eakage synd ome
Mos common diagnosis was common a iable immunode iciency (CVID, n=4).
2
Mücke e al. Compu e ized Classi ica ion o Ques ionnai es
F on ie s in Immunology | www. on ie sin.o g Ap il 2017 | Volume 8 | A icle 384
inT ODUcTiOn
P ima y immunode iciency diso de s (PIDs) in child en a e a
g oup o mo e han 200 a e diseases p esen ing a wide spec um
o symp oms (1, 2). Al hough p og ess in gene ic de ini ion has
been made, clinical diagnos ics emains a challenging ask o
he gene al p ac i ione (GP) o pedia ician. Ea ly diagnosis is
o pa amoun impo ance because he delay leads o inc eased
mo ali y, mo bidi y, and educed quali y o li e (3, 4). Al hough
he ime o diagnosis a ies, diagnos ic delay is common (5–8).
Hence, se e al e o s ha e been made o suppo pa ien –
p o ide communica ion and ea ly e e al o specialis s.
Diagnos ic wa ning signs, educa ion campaigns, and guidelines
ha e been in oduced in o de o aise awa eness and o sho en
diagnos ic delay (9, 10). Howe e , Subba ayan e al. (11) s a e
ha exis ing ools do no wo k su icien ly and new app oaches
a e eques ed (12). In his ega d, pa ien ’s medical his o ies o e
clues o conside ing a PID. P e ious s udies ha e emphasized
he impo ance o medical his o y aking as app oxima ely 80%
o he diagnoses could be es ablished by ca e ul his o y aking
only (13, 14). Howe e , hese s udies do no ocus on PID. Due o
he mul i ude o di e en immune de ec s and he highly a iable
clinical p esen a ion, es ablishing he diagnosis o PID is pa icu-
la ly challenging. In some cases, he unde lying a e condi ion
mimics common diseases. In all cases, physician’s expe ience
and backg ound de e mine whe he a e e al and u he es ing
a e o de ed (15). Be o e es ablishing a diagnosis, pa en s o en
ecognize peculia i ies. Ye , wi hou an immunological expe
a hand, i is di icul o pu hese obse a ions in o he con ex
o a pa icula disease. The e o e, he ques ion o how o decide
which pa ien should ecei e u he in es iga ion gains sup eme
signi icance.
P io o es ablishing a co ec diagnosis, diagnos ic e o s
occu in all medical ields (16, 17) and migh be caused by a i-
ous ac o s such as unusual o silen p esen a ion o he disease,
una ailabili y o expe ise, o inadequa e knowledge (18). Key
indings sugges ha pa ien s should be enabled o ell hei s o y
app op ia ely, and doc o s ha e o compensa e an una oidable
lack o expe ience conce ning a e diseases (19, 20).
The e o e, we in oduce compu e ized analysis o pa en al
obse a ions collec ed in a no el ques ionnai e o p o ide
addi ional suppo . Ou aim was no o de ine a co ec and
speci ic diagnosis bu o iden i y he need o imely e e al o
a PID specialis . Some s udies al eady indica e he po en ial use
o pa ien -cen e ed ques ionnai es and da a mining in diseases
wi h a a he na ow spec um o symp oms (21–23). To aid he
disco e y o unknown co ela ions o o de i e ecommenda-
ions o u he ac ion, compu e -assis ed analysis o huge
amoun s o da a is use ul (24). This a ge ed pa e n analysis
has al eady been es ablished in in e ne sea ch engines, bank-
ing, insu ance, and ma ke ing. I is used o make a p edic ion
o o se e as an immedia e ale unc ion. In medicine such
algo i hms ha e been applied success ully in di e en con ex s
(25, 26).
We hypo hesized ha p e-diagnos ic expe iences o PID
pa ien s could be used o de elop a ques ionnai e. Such a ques-
ionnai e should dis inguish di e en pa ien coho s using da a
mining classi ie s. This in es iga ion could es ablish a basis o
de elop a compu e ized diagnos ic suppo ool.
MaTe ials anD MeThODs
A i s , pa ien -cen e ed semi-s uc u ed in e iews wi h gua d-
ians o child en wi h a con i med PID we e conduc ed. The s udy
p o ocol was app o ed by he e hics commi ee o Hanno e
Medical School and w i en in o med consen was ob ained
om each gua dian. An o e iew o he diseases included in he
in e iew p ocess is p o ided in Table1.
The selec ion o in e iew pa ne s ollowed p ede ined
p inciples: su icien speech comp ehension, w i en consen ,
con i med PID, child en’s age 0–18yea s, majo i y o he eigh
IUIS-ca ego ies (27) should be ep esen ed by a leas one in e -
iew, inclusion o u he in e iewees un il heo e ical sa u a ion
(28). In e iewe s used a guideline o s anda dized p ocedu e
and a uni o m beginning (wha did you obse e ega ding you
child’s heal h and gene al de elopmen be o e he doc o s could
inally ell you ha you child su e s om a PID?). Each in e -
iew was analyzed using Colaizzi’s amewo k wi h espec o
phenomena expe ienced by pa en s in he p e-diagnos ic pe iod
(see Figu e 1). This s anda dized p ocedu e con ains se en
de ined s eps and is well es ablished in social sciences o con en
analysis (29). The esul s o he analysis we e documen ed in a
able con aining a sepa a e column o each s ep. The phenomena
we e so ed depending on occu ence in he in e iews in o de
o in eg a e ele an obse a ions in he ques ionnai e. Fo s udy
pu pose, we added a s ep o ques ion gene a ion (see Figu e1).
Membe s o he s udy g oup used he esul s o d a ques-
ions. The pa en al poin o iew and he choice o wo ds we e
in eg a ed in he de elopmen o he ques ionnai e. All ques ions
we e summa ized in a ques ion pool o u he selec ion. Each
ques ion ecei ed an iden i ica ion numbe o e aceabili y
o i s o igin. The o al numbe o ques ions was sys ema ically
FigU e 1 | colaizzi’s amewo k con ains de ined s eps o
s anda dized con en analysis. Modi ica ion: “ph asing o ques ions” was
inse ed o s udy pu pose.
3
Mücke e al. Compu e ized Classi ica ion o Ques ionnai es
F on ie s in Immunology | www. on ie sin.o g Ap il 2017 | Volume 8 | A icle 384
educed ollowing p ede ined equi emen s: each phenomenon
ele an o he p e-diagnos ic pe iod was ep esen ed by a leas
one i em, and he mos ele an phenomena we e inco po a ed by
addi ional ques ions. Duplica es we e canceled, a p e es o com-
p ehensibili y, consensus in he s udy g oup, and expe opinion
we e in eg a ed. The ull esul ing ques ionnai e consequen ly
e lec s he p e-diagnos ic expe ience o pa en s wi h a child
a ec ed by PID (Table S1 in Supplemen a y Ma e ial).
Dis ibu ion o he Ques ionnai e
To gene a e a da a se , he ques ionnai e was dis ibu ed o
pa ien s who isi ed he Immunological Ou pa ien Clinic o he
Hanno e Medical School in 2013 and 2014 o egula appoin -
men . All pa ien s had an es ablished diagnosis o PID. An addi-
ional online e sion was accessible o membe s o he pa ien
g oup o PID in Ge many [Deu sche Selbs hil e Angebo ene
Immunde ek e (DSAI)] wi h a special access code. As a con ol
g oup, we andomly collec ed ques ionnai es om gua dians o
heal hy child en and child en who we e hospi alized due o a
disease o he han PID.
Da a Mining Me hods
S a is ical so wa e lib a ies o e di e en compu e -based
me hods o analyze and classi y da a. Mos o hese me hods a e
a ia ions o main s a is ical concep s like ec o space me hods
o a i icial neu al ne wo ks. The ool unde discussion uses sup-
po ec o machines, andom o es s, logis ic eg ession (LR),
nai e Bayes classi ie s, linea disc iminan analysis and nea es
neighbo classi ie s (30). Classi ie s we e ained and es ed wi h
ques ionnai es p ocessed in nume ic able o ma . A usion algo-
i hm combined he di e en p edic ions made by each single
classi ie o one inal decision (31). Valida ion o he sys em was
pe o med in wo s eps. Fi s , he diagnos ic accu acy was chal-
lenged using wo se s o ques ionnai es: child en wi h PID and
heal hy child en. In a second s ep, he sys em had o dis inguish
be ween ques ionnai es o PID child en and a combina ion
o andomly chosen heal hy and sick child en wi h di e en
diseases, e.g., se e e b onchi is, b ain umo , cys ic ib osis, and
ulce a i e coli is ( ull spec um, see Table S2 in Supplemen a y
Ma e ial). On bo h le els, alida ion was pe o med by k- old
s a i ied c oss- alida ion. In addi ion, we analyzed ques ions
conce ning hei con ibu ion o he co ec classi ica ion based
on he p-Value o he coe icien s in LR, i.e., wi h he null hypo h-
esis ha he coe icien o he ques ion is 0, meaning ha he
co esponding ques ion does no con ibu e signi ican ly o he
p edic ion based on LR.
esUlTs
co e Phenomena We e e ealed h ough
F equen Pa en al Obse a ions in
s anda dized, semi-s uc u ed in e iews
The analysis o he in e iews using Colaizzi’s amewo k e ealed
majo hemes such as ch onological cha ac e is ics o in ec ions,
pa en al pe cep ion o in ec ions, suscep ibili y o in ec ions in
e e yday li e, in ec ions o he espi a o y ac , and e ec i eness
o an ibio ic ea men (Table S3 in Supplemen a y Ma e ial).
These we e exp essed in wo ds and concep s by he a ge g oup.
gene a ion o a 36-i em Ques ionnai e by
s anda dized con en analysis
Based on majo hemes and quo a ions, 186 p elimina y ques-
ions we e ph ased o ep esen he pa en al pe spec i e (no
shown). Fo he inal e sion, he ques ions we e sys ema ically
educed (Table S1 in Supplemen a y Ma e ial). Reasons o exclu-
sion we e: duplica ions in o m and con en , i ele ance o he
p e-diagnos ic pe iod, and incomp ehensibili y. Modi ica ions
we e discussed wi h expe ienced immunologis s and app o ed
by consensus o he s udy g oup. The inal ques ionnai e was sub-
sequen ly comple ed by 126 pa en s. A low cha o he p ocess
is shown in Figu e2.
He e, he inal ques ionnai e was comple ed by pa en s om
di e en pa ien collec i es. The esul ing aw da a se consis ed o
64 ques ionnai es om child en wi h PID and 62 ques ionnai es
wi hou PID (35 heal hy+27 pa ien s wi h illness o he han PID,
see Figu e3). The mos common diagnoses in he PID g oup
we e common a iable immunode iciency diso de s (CVID)
(n=11) and agammaglobulinemia (n=10) (see Table2).
Diagnoses in he illness-o he - han-PID g oup we e e.g., acu e
lympha ic leukemia, coli is ulce osa, cys ic ib osis, and ch onic
enal ailu e (Table S2 in Supplemen a y Ma e ial).
iden i ica ion o suspicious answe
Pa e ns by Da a Mining shows a
sensi i i y o up o 98%
99 indi iduals answe ed he ques ionnai e o s ep 1. This g oup
consis ed o 64 indi iduals wi h PID and 35 heal hy child en
se ing as a con ol g oup (see Figu e 2). A 11- old s a i ied
FigU e 2 | Flow cha o he s udy p ocedu e. A e conduc ion o 12
in e iews wi h pa en s o p ima y immunode iciency diso de (PID) pa ien s,
con en analysis was used o de elop a 36-i em ques ionnai e. The no el
ques ionnai e was u ilized o collec da a o aining and es ing o da a
mining echniques conce ning classi ica ion app oaches esul ing in sensi i i y
up o 98%.
4
Mücke e al. Compu e ized Classi ica ion o Ques ionnai es
F on ie s in Immunology | www. on ie sin.o g Ap il 2017 | Volume 8 | A icle 384
c oss alida ion showed an o e all sensi i i y o 98%. Rega ding
he g oup o child en wi h PID, 63 o 64 (98%) ecei ed he co -
ec diagnosis, and 34/35 o he con ols we e classi ied co ec ly.
In o al, 97/99 indi iduals ecei ed he co ec diagnosis in he
c oss- alida ion. Subsequen ly, 27 ques ionnai es o andomly
chosen child en wi h di e en diseases we e added o he da a
se (see Figu e3). A 21- old s a i ied c oss- alida ion was pe -
o med o examine diagnos ic sensi i i y.
In his g oup (PID pa ien s compa ed o heal hy + illness
o he han PID), 113/126 (90%) o he ques ionnai es we e clas-
si ied co ec ly. In de ail, 61/64 indi iduals wi h PID ecei ed
he co ec diagnosis, bu only 52/62 indi iduals we e co ec ly
classi ied as “no-PID.”
analysis Disco e ed Ques ions signi ican
o Di e en ia ion
The analysis o he 36 ques ions, using p-Value compu a ion,
e ealed 32 ques ions which con ibu ed o he co ec classi ica-
ion signi ican ly (p<0.05). Ques ions wi h highes signi icance
in s ep 2 (PID s. heal hy+illness o he han PID) we e
Q1 (Did you child su e om ill heal h cons an ly?)
Q31 (Is i ue ha you child’s in ec ions las ed longe han he
ones o o he child en?)
Q32 (Is i ue ha you child was ea ed wi h an ibio ics
egula ly?)
Q35 (Is i ue ha he doc o s could no ell wha you child was
su e ing om?)
Th ee-dimensional Visualiza ion o answe
Pa e ns Di e en ia es heal hy om sick
child en
In e ms o a classi ica ion p oblem he answe pa e n classi i-
ca ion wi h 36 i ems consis s o 36 dimensions as 36 ques ions
we e answe ed. Based on Sammon mapping, a h ee-dimensional
ep esen a ion o he da a was gene a ed (see Figu e4) (32). I
isualizes he ques ionnai es conce ning he co ec classi ica-
ion, and illus a es he gene al p inciple o dis inguishing wo
g oups o pa ien s. Compa ing s ep 1 and s ep 2 o he s udy
(Figu es 4A,B) mo e o e lapping o he coho s is shown in
Figu e4B. O e lapping ques ionnai es can ep esen misdiag-
nosed cases.
DiscUssiOn
The diagnosis o PID p esen s a challenge o pedia icians due
o a e incidence and unspeci ic symp oms, esul ing in delayed
e e al and nega i e ou come o pa ien s and pa en s (3, 4). The
p e-diagnos ic expe ience o pa ien s and hei amilies is no
su icien ly in eg a ed in o he diagnos ic p ocess. The s udy a
hand indica es ha a no el combina ion o a ques ionnai e and
da a mining echniques p o ides he means o iden i y suspicious
answe pa e ns.
Ou da a show ha pa ien in e iews a e a easible ool
o gene a e ques ionnai es. In o de o imp o e he quali y
o he ques ions, pa en al in e iews we e ob ained o collec
obse a ions om he p e-diagnos ic ime. Colaizzi’s me hod
is widely applied in nu sing science and quali a i e esea ch
and has p o en use ul o de i e knowledge om in e iews
(33, 34). I appea s o be use ul o collec pe sonal expe i-
ence and o mine i o he gene a ion o a ques ionnai e
(35). The da a gene a ed in his s udy suppo s he impac o
quali a i e esea ch in he diagnos ic p ocess (36). Ou da a
gi e impo an insigh s in o he pa en s’ pe cep ions p io o
diagnosis as opposed o clinical ocus, and likewise easons
o he diagnos ic delay could be exposed. An ad an age o
Colaizzi’s amewo k is he s anda dized p ocedu e, which
makes he connec ion be ween ci a ion and gene a ed ques ion
e aceable (37).
In ou app oach, we combined six classi ie s and added
a usion algo i hm o inc ease sensi i i y. Ou combina ion
showed a sensi i i y o 84–98%. In compa ison o s ep 1 and s ep
2, sensi i i y dec eased when adding sick child en o heal hy
con ols. The o e all sensi i i y o 90% in s ep 2 s ill unde sco es
he p omising pe o mance o he echniques pilo ed he e.
The applica ion o a usion algo i hm imp o ed he diagnos ic
quali y o he ool. Recen ly, he concep o ques ionnai es and
da a mining was success ully pilo ed by Ro he e al. (23). They
applied ela ed echniques on pulmona y diseases and achie ed
encou aging esul s.
The combina ion o a ques ionnai e and da a mining has
al eady been es ed in diagnosing gas oesophageal e lux dis-
ease (GERD) (21). The ool di e en ia es be ween complain s
caused by GERD and o he dyspep ic diso de s. In con as
o ou s udy, Ho owi z e al. used a sho e ques ionnai e (15
s. 36 i ems), did no use ex ensi e in e iews o accumula e
obse a ions, and applied less da a mining echniques.
I is impo an o emphasize ha he p ocess o p o essional
medical his o y aking is nei he in ended no sui able o be
eplaced by ques ionnai es. In ac , his o y aking migh be
imp o ed because he diagnos ic ool highligh s hose ques ions,
Table 2 | The ou mos equen diagnoses o child en wi h p ima y
immunode iciency diso de s (PiDs): he ou mos equen diagnoses o
a o al o 64 ques ionnai es ecei ed om PiD pa ien s.
Diagnosis n
Common a iable immune de iciency (CVID) 11
Agammaglobulinemia 10
Synd omes wi h ecu en e e (TRAPS+PFAPA) 6
Ch onic g anuloma ous disease 5
TRAPS, TNF ecep o -associa ed pe iodic synd ome; PFAPA, pe iodic e e wi h
aph hous s oma i is, pha yngi is, and adeni is.
FigU e 3 | Da a sou ce o ques ionnai es o de ed by ca ego y showing s epwise analysis app oach. Ques ionnai es we e dis ibu ed o h ee g oups and
subsequen ly used o da a mining aining and es . Tes o classi ica ion skills was pe o med in s ep 1 [p ima y immunode iciency diso de s (PIDs) s. heal hy] and
s ep 2 (PID s. no PID). No-PID con ains ques ionnai es o heal hy child en and child en wi h illness o he han PID.
5
Mücke e al. Compu e ized Classi ica ion o Ques ionnai es
F on ie s in Immunology | www. on ie sin.o g Ap il 2017 | Volume 8 | A icle 384
which da a mining sugges ed o be o high diagnos ic ele ance.
Physicians can cla i y he pas medical his o y and ecei e addi-
ional hin s (38–40). The accep ance o diagnos ic ques ionnai es
has al eady been p o en (41).
Some expe s s ess he no ion ha mo e awa eness o
PID is needed among physicians in e ia y hospi als (11). Ye ,
Lankisch e al. (42) sugges ocusing on p ima y ca e physicians
and pedia icians wi hou p o ound expe ience in PID. Lankisch
e al. p o ided a modi ied ca alog o wa ning signs and achie ed
imp o ed esul s o de ec ing child en wi h PID compa ed o
classic wa ning signs (12, 42). Howe e , hey unde lined he need
o imp o emen s in he diagnos ic pa hway in o de o de ec
p e e ably 9 ou o 10 child en su e ing om PID. The ool
pilo ed he e migh help o ill his gap despi e he p elimina y
cha ac e o he s udy as well as i s limi a ions. Di e en pape -
based ques ionnai es a e nowadays used on occasion in ou pa-
ien clinics o collec da a abou pa ien s’ medical his o y. They
a e based on expe opinion, mos a e no empi ically e i ied,
and hey a e in e p e ed manually de i ing hin s depending on
examine s expe ise. In con as , ou ool ocuses on he pa ien s
o gua dians pe spec i e, can be s a is ically e iewed, is ex end-
able, imp o ed by adding mo e da a, and i wo ks independen ly
om he doc o ’s expe ience.
Pa s o ou ques ionnai e e lec cu en guidelines (43).
Addi ionally, he ques ions wi h high impac o he diagnosis
PID sha e simila i ies o he classical “wa ning signs” bu also
di e ences. The usage o an ibio ics is included in bo h ools.
Simple ques ions like Q1 (did you child su e om ill heal h
cons an ly?) and Q36 (is i ue ha you child was absen om
school/p e-school/kinde ga en due o sickness mo e o en han
o he child en?) a e solely de i ed om pa en al expe ience and
e ealed an impo an con ibu ion o co ec classi ica ion. In
con as o an analysis by Subba ayan e al. (11), a posi i e amily
his o y was no signi ican o inding he co ec diagnosis in
ou coho . This esul suppo s a p esump ion by B odszki e al.
(44) who s a es ha he deg ee o consanguini y in he collec i e
is essen ial. Kallus (45) unde lines ha a speci ic a ibu e has
o be ep esen ed in a sample o en enough o each su icien
disc imina o y powe . Fo his eason, a ques ionnai e con aining
mo e han 10 i ems inc eases he p obabili y o ca ch su icien
obse a ions. Ne e heless, i is no a single ques ion ha makes
he di e ence. I is he combina ion o se e al ques ions wi h
di e en con ibu ions o he co ec classi ica ion ha makes
diagnos ic hin s possible.
Ou s udy has se e al limi a ions: i s o all, as i is a s udy
o p oo o concep , only a na ow spec um o diagnoses was
included in he in e iew phase. Likewise, only a small pa o all
known PID was ep esen ed. Second, he numbe o pa ien s and
con ols is limi ed. Ne e heless, c oss- alida ion suppo s he
esul s which need e-e alua ion in a p ospec i e ial. Besides,
diagnos ic suppo was also possible o hose diagnoses no
included in he in e iews indica ing he plas ici y o he sys em.
Thi d, using a w i en ques ionnai e limi s he usage o a ool. Fo
example, i equi es enough language comp ehension. The ques-
ionnai e was gene a ed and e alua ed in Ge man. Today, he
ans e in o di e en languages and cul u es and ep oduc ion
o he excellen esul s is unp o en. In he e a o mul icul u al
socie ies, a uni e sal comp ehensibili y independen o language
abili y would be desi able (46). Table compu e s, which ha e
al eady been posi i ely e alua ed in o he a eas could be used
o educe he e o o comple ion o he ques ionnai e and i s
in e p e a ion (47, 48). Fu he mo e, possible online adap ions
o pape e sions al eady p oo ed usabili y in o he a eas (49).
A u he limi a ion is ela ed o he selec ion o con ols. The
coho s do no e lec he egula spec um o pa ien s seen by
GP o pedia icians. Ideally, his s udy should ha e inco po a ed
a con ol g oup ep esen a i e o he day- o-day popula ion
o GP and/o gene al pedia icians. Con ols p esen ing wi h
symp oms suspicious o a PID would ha e answe ed a ques ion-
nai e p io o he e e al o an immunologis . Due o limi ed
esou ces, his e alua ion will be pa o a u u e ial based on
his pilo p oo o concep . Thus, u he p ospec i e s udies wi h
pa ien s p esen ing PID-like symp oms a e needed o e i y ou
app oach.

FigU e 4 | Visualiza ion o answe s indica es di e en answe pa e n. Answe pa e n can be isualized by he dimension educ ion echnique Sammon
mapping. X=p ima y immunode iciency diso de (PID), O=con ol. (a) Classes PIDs s. heal hy a e easy o dis inguish. (b) Due o inc eased he e ogenei y i is
mo e di icul o disc imina e classes PID s. no-PID. No-PID con ains ques ionnai es om heal hy child en and child en su e ing om illnesses o he han PID.
6
Mücke e al. Compu e ized Classi ica ion o Ques ionnai es
F on ie s in Immunology | www. on ie sin.o g Ap il 2017 | Volume 8 | A icle 384
Taken oge he , we success ully pilo ed a ques ionnai e-based
ool in ended o classi y di e en pa ien coho s.
A diagnos ic ool o de ec ion o child en wi h PID could be
used in di e en se ings: he GP o he pedia ician could hand
i o pa en s wi h child en showing ecu en in ec ions o eas-
su e he need o u he diagnos ics. The ques ionnai e could be
answe ed on a able compu e while si ing in he wai ing a ea.
A diagnos ic sugges ion would only be p esen ed o he physi-
cian who could include he compu e ized diagnos ic sugges ion
in o he diagnos ic wo k low and use eye-ca ching answe s o
cla i ying.
A se ies o i e clinical cases as an example o he p ospec-
i e classi ica ion p ocess is gi en in Table S4 in Supplemen a y
Ma e ial. Ex ensi e in es iga ion o use iendliness is pa o a
p ospec i e e alua ion. The p inciples applied in his s udy migh
also be ex ended o o he g oups o a e diseases beyond PID.
aUThO cOnT ibUTiOns
UM and LG concep ualized he s udy, conduc ed he in e iew,
analyzed ansc ip s, and w o e he manusc ip . WL, FK, and XK
pe o med all da a mining. CK, UB, and AM-B supe ised he
clinical p ocess. All au ho s discussed he esul s and commen ed
on he manusc ip .
acKnOWleDgMenTs
This s udy would no ha e been possible wi hou he help o
pa en s and pa ien s who pa icipa ed in he in e iew p ocess
and comple ed he ques ionnai es. We also acknowledge he sup-
po o he nu ses in he Immunological Ou pa ien Clinic o he
Medical Uni e si y Hanno e and he DSAI (Gab iele G ündel).
We a e hank ul o Da id Dieckmann o in-dep h p oo eading.
FUnDing
This wo k ecei ed, in pa s, inancial suppo by he Robe -
Bosch S i ung, S u ga (Ge many).
sUPPleMenTa Y MaTe ial
The Supplemen a y Ma e ial o his a icle can be ound
online a h p://jou nal. on ie sin.o g/a icle/10.3389/ immu.
2017.00384/ ull#supplemen a y-ma e ial.
eFe ences
1. Pica d C, Al-He z W, Bous iha A, Casano a JL, Cha ila T, Conley ME, e al.
P ima y immunode iciency diseases: an upda e on he classi ica ion om he
in e na ional union o immunological socie ies expe commi ee o p ima y
immunode iciency 2015. J Clin Immunol (2015) 35:696–726. doi:10.1007/
s10875-015-0201-1
2. Boyle JM, Buckley RH. Popula ion p e alence o diagnosed p ima y immu-
node iciency diseases in he Uni ed S a es. J Clin Immunol (2007) 27:497–502.
doi:10.1007/s10875-007-9103-1
3. Joshi AY, Iye VN, Hagan JB, S Sau e JL, Boyce TG. Incidence and empo al
ends o p ima y immunode iciency: a popula ion-based coho s udy. Mayo
Clin P oc (2009) 84:16–22. doi:10.1016/S0025-6196(11)60802-1
4. Tabolli S, Giannan oni P, Pul i en i F, La Ma a F, G ana a G, Mili o C, e al.
Longi udinal s udy on heal h- ela ed quali y o li e in a coho o 96 pa ien s
wi h common a iable immune de iciencies. F on Immunol (2014) 5:1–9.
doi:10.3389/ immu.2014.00605
5. Michos A, Rap aki M, Tan ou S, Tzanoudaki M, Spanou K, Lia sis
M, e  al. P ima y immunode iciency diseases: a 30-yea pa ien
egis y om he e e al cen e o p ima y immunode iciencies in
G eece. J Clin Immunol (2014) 34(7):836–43. doi:10.1007/s10875-014-
0066-8
6. Aghamohammadi A, Bah ami A, Mamishi S, Mohammadi B, Abolhassani H,
Pa aneh N, e al. Impac o delayed diagnosis in child en wi h p ima y an i-
body de iciencies. J Mic obiol Immunol In ec (2011) 44:229–34. doi:10.1016/
j.jmii.2011.01.026
7
Mücke e al. Compu e ized Classi ica ion o Ques ionnai es
F on ie s in Immunology | www. on ie sin.o g Ap il 2017 | Volume 8 | A icle 384
7. Maa schalk-Elle b oek LJ, Hoepelman AIM, Van Mon ans JM, Elle b oek
PM. The spec um o disease mani es a ions in pa ien s wi h common a iable
immunode iciency diso de s and pa ial an ibody de iciency in a uni e si y
hospi al. J Clin Immunol (2012) 32:907–21. doi:10.1007/s10875-012-
9671-6
8. Quin i I, So esina A, Spada o G, Ma ino S, Donnanno S, Agos ini C, e al.
Long- e m ollow-up and ou come o a la ge coho o pa ien s wi h common
a iable immunode iciency. J Clin Immunol (2007) 27:308–16. doi:10.1007/
s10875-007-9075-1
9. Modell V. The impac o physician educa ion and public awa eness on
ea ly diagnosis o p ima y immunode iciencies: Robe A. Good immu-
nology symposium. Immunol Res (2007) 38:43–7. doi:10.1007/s12026-007-
0048-5
10. Fa mand S, Baumann U, Bo e M, F anke K, Habe mehl P, Kapaun P,
e  al. In e disziplinä e AWMF-Lei linie zu Diagnos ik on p imä en
Immunde ek en (S2k) in e disciplina y AWMF guideline o he diagnos ics
o p ima y immunode iciency. Klin Padia (2011) 223:378–85. doi:10.1055/
s-0031-1287835
11. Subba ayan A, Cola usso G, Hughes MS, Genne y RA, Sla e M, Can JA,
e al. Clinical ea u es ha iden i y child en wi h p ima y immunode iciency
diseases. Pedia ics (2011) 127:810–6. doi:10.1542/peds.2010-3680
12. A kw igh PD, Genne y AR. Ten wa ning signs o p ima y immunode i-
ciency: a new pa adigm is needed o he 21s cen u y. Ann N Y Acad Sci
(2011) 1238:7–14. doi:10.1111/j.1749-6632.2011.06206.x
13. Hamp on JR, Ha ison MJ, Mi chell JR, P icha d JS, Seymou C. Rela i e
con ibu ions o his o y- aking, physical examina ion, and labo a o y in es i-
ga ion o diagnosis and managemen o medical ou pa ien s. B Med J (1975)
2:486–9. doi:10.1136/bmj.2.5969.486
14. Pe e son MC, Holb ook JH, Von Hales D, Smi h NL, S ake LV.
Con ibu ions o he his o y, physical examina ion, and labo a o y
in es iga ion in making medical diagnoses. Wes J Med (1992) 156:163–5.
doi:10.1097/00006254-199210000-00013
15. Budych K, Helms TM, Schul z C. How do pa ien s wi h a e diseases expe-
ience he medical encoun e ? Explo ing ole beha io and i s impac on
pa ien -physician in e ac ion. Heal h Policy (2012) 105:154–64. doi:10.1016/
j.heal hpol.2012.02.018
16. Singh H, Thomas EJ, Wilson L, Kelly PA, Pie z K, Elkeeb D, e al. E o s o
diagnosis in pedia ic p ac ice: a mul isi e su ey. Pedia ics (2010) 126:70–9.
doi:10.1542/peds.2009-3218
17. Singh H, Pe e sen LA, Thomas EJ. Unde s anding diagnos ic e o s in
medicine: a lesson om a ia ion. Qual Sa Heal h Ca e (2006) 15:159–64.
doi:10.1136/qshc.2005.016444
18. G abe M, Go don R, F anklin N. Reducing diagnos ic e o s in medicine:
wha ’s he goal? Acad Med (2002) 77:981–92. doi:10.1097/00001888-
200210000-00009
19. G abe ML. Taking s eps owa ds a sa e u u e: measu es o p omo e imely
and accu a e medical diagnosis. Am J Med (2008) 121:S43–6. doi:10.1016/
j.amjmed.2008.02.006
20. Sa ka U, Bonacum D, S ull W, Spi zmuelle C, Jin N, López A, e  al.
Challenges o making a diagnosis in he ou pa ien se ing: a mul i-si e su ey
o p ima y ca e physicians. BMJ Qual Sa (2012) 21:641–8. doi:10.1136/
bmjqs-2011-000541
21. Ho owi z N, Moshkowi z M, Halpe n Z, Leshno M. Applying da a mining
echniques in he de elopmen o a diagnos ics ques ionnai e o GERD. Dig
Dis Sci (2007) 52:1871–8. doi:10.1007/s10620-006-9202-5
22. G igull L, Lechne WM. Suppo ing diagnos ic decisions using hyb id
and complemen a y da a mining applica ions: a pilo s udy in he pedi-
a ic eme gency depa men . Pedia Res (2012) 71:725–31. doi:10.1038/
p .2012.34
23. Ro he AK, Schwe k N, B inkmann F, Klawonn F, Lechne W, G igull
L. Diagnos ic suppo o selec ed paedia ic pulmona y diseases using
answe -pa e n ecogni ion in ques ionnai es based on combined da a min-
ing applica ions-a monocen ic obse a ional pilo s udy. PLoS One (2015)
10:1–17. doi:10.1371/jou nal.pone.0135180
24. Yoo I, Ala ai ee P, Ma ino M, Pena-He nandez K, Gopidi R, Chang J-F, e al.
Da a mining in heal hca e and biomedicine: a su ey o he li e a u e. J Med
Sys (2012) 36:2431–48. doi:10.1007/s10916-011-9710-5
25. Yang ST, Lee JD, Chang TC, Huang CH, Wang JJ, Hsu WC, e  al.
Disc imina ion be ween Alzheime ’s disease and mild cogni i e impai men
using SOM and PSO-SVM. Compu Ma h Me hods Med (2013) 2013:253670.
doi:10.1155/2013/253670
26. Bax WG, Sho e FS, Si es FD, Hollande JE. A neu al ne wo k aid o he
ea ly diagnosis o ca diac ischemia in pa ien s p esen ing o he eme gency
depa men wi h ches pain. Ann Eme g Med (2002) 40:575–83. doi:10.1067/
mem.2002.129171
27. Al-He z W, Bous iha A, Casano a JL, Chapel H, Conley ME, Cunningham-
Rundles C, e  al. P ima y immunode iciency diseases: an upda e on he
classi ica ion om he In e na ional Union o Immunological Socie ies Expe
Commi ee o P ima y Immunode iciency. F on Immunol (2011) 5:162.
doi:10.3389/ immu.2011.00054
28. Flick U. An In oduc ion o Quali a i e Resea ch. 4 h ed. SAGE (2009). 529 p.
A ailable om: h p://books.google.cz/books?id=sF 1oWX2DoEC
29. Colaizzi PF. Psychological Resea ch as he Phenomenologis ’s View I . In: Valle
RS, King M, edi o s. Exis en ial-Phenomenological Al e na i es o Psychology.
New Yo k: Ox o d Uni e si y P ess (1978). p. 48–71.
30. Fe nández-Delgado M, Ce nadas E, Ba o S, Amo im D. Do we need hun-
d eds o classi ie s o sol e eal wo ld classi ica ion p oblems? J Mach Lea n
Res (2014) 15:3133–81.
31. G igull L, Lechne W, Pe i S, Kollewe K, Dengle R, Mehmecke S, e  al.
Diagnos ic suppo o selec ed neu omuscula diseases using answe -pa e n
ecogni ion and da a mining echniques: a p oo o concep mul icen e
p ospec i e ial. BMC Med In o m Decis Mak (2016) 16:31. doi:10.1186/
s12911-016-0268-5
32. Sammon JW. A nonlinea mapping o da a s uc u e analysis. IEEE T ans
Compu (1969) C-18:401–9. doi:10.1109/T-C.1969.222678
33. Kea on KA, Pie ce LL. Ca diac he apy o men wi h co ona y a e y
disease: he li ed expe ience. J Holis Nu s (2000) 18:63–85. doi:10.1177/
089801010001800107
34. Hillia d C, O’Neill M. Nu ses’ emo ional expe ience o ca ing o child en
wi h bu ns. J Clin Nu s (2010) 19:2907–15. doi:10.1111/j.1365-2702.2009.
03177.x
35. Nada ajah SR, Buchholz SW, Wiegand DL, Be ge A. The li ed expe ience
o indi iduals in ca diac ehabili a ion who ha e a posi i e ou look on hei
ca diac eco e y: a phenomenological inqui y. Eu J Ca dio asc Nu s (2016)
16(3):230–9. doi:10.1177/1474515116651977
36. Hin on D, Ki k S. Paedia ic mul iple scle osis: a quali a i e s udy o am-
ilies’ diagnosis expe iences. A ch Dis Child (2015) 100:623–9. doi:10.1136/
a chdischild-2014-306523
37. Sande s C. Applica ion o Colaizzi’s me hod: in e p e a ion o an audi able
decision ail by a no ice esea che . Con emp Nu se (2014) 14:292–302.
doi:10.5172/conu.14.3.292
38. A o a S, Goldbe g AD, Menchine M. Pa ien imp ession and sa is ac ion
o a sel -adminis e ed, au oma ed medical his o y- aking de ice in he
eme gency depa men . Wes J Eme g Med (2014) 15:35–40. doi:10.5811/
wes jem.2013.2.11498
39. Koeneman MM, K use A-J, Sep SJS, Gubbels CS, Slangen BFM, an Go p T,
e al. A amily his o y ques ionnai e imp o es de ec ion o women a isk o
he edi a y gynecologic cance : a pilo s udy. Fam Cance (2014) 13:469–75.
doi:10.1007/s10689-014-9711-x
40. DeAngelis AF, Chambe s IG, Hall GM. The accu acy o medical
his o y in o ma ion in e e al le e s. Aus Den J (2010) 55:188–92.
doi:10.1111/j.1834-7819.2010.01221.x
41. He shey CO, G an BJB. Con olled ial o a pa ien -comple ed his o y
ques ionnai e: e ec s on quali y o documen a ion and pa ien and physician
sa is ac ion. Am J Med Qual (2016) 17:126–35. doi:10.1177/106286060
201700402
42. Lankisch P, Schi ne J, Ghosh S, Babo F, Bo kha d A, Laws HJ. The
Duesseldo wa ning signs o p ima y immunode iciency: is i ime o change
he ules? J Clin Immunol (2015) 35:273–9. doi:10.1007/s10875-015-0149-1
43. Fa mand S, Baumann U, on Be nu h H, Bo e M, Foe s e -Waldl E, F anke
K, e al. [In e disciplina y AWMF guideline o he diagnos ics o p ima y
immunode iciency]. Klin Padia (2011) 223:378–85. doi:10.1055/s-0031-
1287835
44. B odszki N, Jönsson G, Ska um L, T uedsson L. P ima y immunode iciency
in in ec ion-p one child en in sou he n Sweden: occu ence, clinical
cha ac e is ics and immunological indings. BMC Immunol (2014) 15:31.
doi:10.1186/s12865-014-0031-6
45. Kallus KW. E s ellung on F abeogen. 1. Au l. Wien: Facul as.wu (2010).
8
Mücke e al. Compu e ized Classi ica ion o Ques ionnai es
F on ie s in Immunology | www. on ie sin.o g Ap il 2017 | Volume 8 | A icle 384
46. Gog ay NJ, Tha e UM, Dasgup a B, Deshpande S. Use o he WOMAC
ques ionnai e in Mumbai and he challenges o ansla ion and c oss cul u al
adap a ion. Indian J Med E hics (2016) 10:33–5. A ailable a : h p://www.ncbi.
nlm.nih.go /pubmed/23439195
47. Mickan S, A he on H, Robe s NW, Heneghan C, Tilson JK. Use o handheld
compu e s in clinical p ac ice: a sys ema ic e iew. BMC Med In o m Decis
Mak (2014) 14:56. doi:10.1186/1472-6947-14-56
48. Fleischmann R, Duhm J, Huppe s H, B and SA. Table compu e s wi h
mobile elec onic medical eco ds enhance clinical ou ine and p omo e
bedside ime: a con olled p ospec i e c osso e s udy. J Neu ol (2015)
262:532–40. doi:10.1007/s00415-014-7581-7
49. Bland JDP, Rudol e S, Welle P. P ospec i e analysis o he accu acy o diag-
nosis o ca pal unnel synd ome using a web-based ques ionnai e. BMJ Open
(2014) 4:e005141. doi:10.1136/bmjopen-2014-005141
Con lic o In e es S a emen : WL, FK, and LG a e co- ounde s o IMD (IMD
GmbH, Deu schland). The o he au ho s ha e no con lic s o in e es o disclose.
The handling edi o decla ed a pas co-au ho ship wi h one o he au ho s
UB and s a es ha he p ocess ne e heless me he s anda ds o a ai and
objec i e e iew.
Copy igh © 2017 Mücke, Klemann, Baumann, Meye -Bahlbu g, Ko um,
Klawonn, Lechne and G igull. This is an open-access a icle dis ibu ed unde he
e ms o he C ea i e Commons A ibu ion License (CC BY). The use, dis ibu ion
o ep oduc ion in o he o ums is pe mi ed, p o ided he o iginal au ho (s) o
licenso a e c edi ed and ha he o iginal publica ion in his jou nal is ci ed, in
acco dance wi h accep ed academic p ac ice. No use, dis ibu ion o ep oduc ion is
pe mi ed which does no comply wi h hese e ms.