Universal Exome Sequencing in Critically Ill Adults: A Diagnostic Yield of 25% and Race-Based Disparities in Access to Genetic Testing

Numerous studies have underscored the diagnostic and therapeutic potential of exome or genome sequencing in critically ill pediatric populations. However, an equivalent investigation in critically ill adults remains conspicuously absent. We retrospectively analyzed whole exome sequencing (WES) data available through the PennMedicine Biobank (PMBB) from all 365 young adult patients, aged 18-40 years, with intensive care unit (ICU) admissions at the University of Pennsylvania Health System who met inclusion criteria for our study. For each participant, two Medical Genetics and Internal Medicine-trained clinicians reviewed WES reports and patient charts for variant classification, result interpretation, and identification of genetic diagnoses related to their critical illness. Of the 365 individuals in our study, 90 (24.7%) were found to have clearly diagnostic results on WES; an additional 40 (11.0%) had a suspicious variant of uncertain significance (VUS) identified; and an additional 16 (4.4%) had a medically actionable incidental finding. The diagnostic rate of exome sequencing did not decrease with increasing patient age. Affected genes were primarily involved in cardiac function (18.8%), vascular health (16.7%), cancer (16.7%), and pulmonary disease (11.5%). Only half of all diagnostic findings were known and documented in the patient chart at the time of ICU admission. Significant disparities emerged in subgroup analysis by EHR-reported race, with genetic diagnoses known/documented for 63.5% of White patients at the time of ICU admission but only for 28.6% of Black or Hispanic patients. There was a trend towards patients with undocumented genetic diagnoses having a 66% increased mortality rate, making these race-based disparities in genetic diagnosis even more concerning. Altogether, universal exome sequencing in ICU-admitted adult patients was found to yield a new definitive diagnosis in 11.2% of patients. Of these diagnoses, 76.6% conferred specific care-altering medical management recommendations. Our study suggests that the diagnostic utility of exome sequencing in critically ill young adults is similar to that observed in neonatal and pediatric populations and is age-independent. The high diagnostic rate and striking race-based disparities we find in genetic diagnoses argue for broad and universal approaches to genetic testing for critically ill adults. The widespread implementation of comprehensive genetic sequencing in the adult population promises to enhance medical care for all individuals and holds the potential to rectify disparities in genetic testing referrals, ultimately promoting more equitable healthcare delivery. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint

T echnologic al innova ti ons ha v e increased the acce ssibi lity and clin ic al utili ty of b r oad, non- t ar g et ed g enetic t e s ting. 1 N o lo ng er c on s tr ained by pr ohib iti v e c o s t o r leng th y turn-ar ound times, g enomic t es ting , including whole e x ome sequencing (WE S) or whole g enome sequencing (W GS), is quick ly b ec oming an integr al t est in the c a r e of crit ic a lly ill p ediatric pa t ients. Numer ous s tudies of ne ona t a l and p edia tr ic i n tensive c a r e uni ts, inc ludin g r andomiz ed c on tr o l trials, show tha t r apid (turnar o und time 1-2 w eek s) or ultr ar apid (tur n ar ound time 2 da y s) WE S/ W GS incr ease diagnos ti c yie ld, shorten the diagnos ti c od y ssey , an d demons tr a te wide- r anging healthc ar e c ost sa vings due to quick er ini tiation of t ar g et ed tr eat men ts, short er hospi t al admissions, and f ew er in v asi v e i n ter v en t ions. 2–8 Expansion o f r apid genomic t esting t o al l criti c al ly ill pe dia t ric patie n ts r ep li c at es these clini c al and f inancial implic a t ions. 9–11 How ev e r , despit e th e w ell-documen ted benefit s in pedia tr ic populations, g enomic tes ting is not r out inel y off er ed du ring the c ar e o f cri ti c ally ill adults. In the f ew r et r ospec tiv e s tud ies o f g enomic t es t ing in int ens iv e c are u nits (ICUs) that ha v e included small number s of adult pa ti en ts (n=7, 36), diagnos tic yields in patients ov er the ag e of 18 y ear s r ang ed fr om 22-57% and ther e w as no st atistic a l di ff er e nce in diagnosis based on pa tie n t ag e. 12,13 The omission of adu lts fr om br oad sequencing s tudi es o f cri ti c ally il l patie n ts par allels the lac k of evidenc e-based guidance on the indica tion s f or br oa d g enomic t es ting in adults mor e g ener a lly , desp ite n umer ous s tudies showing similar r esul ts t o ped ia tr ic popula ti ons. 14–19 In thes e s t udies, t he diagnos tic ra t e of ex ome seque ncing in adults wi th suspect ed g ene tic c ondi tions has b een r epor t ed t o r ang e f r om 14-29 %, demons tr a ting the utili ty o f b r oad g enet ic t esting appr o aches in this popu la ti on. T og ether th ese f acts sug g es t that inc orp or a ting r apid g enomic s equenc ing in t o the c ar e of crit ic a lly ill adu lts ma y yield diagnos ti c r a t es and ben e f its in c o s t and ou t c om e similar to those s een in the pedi a tr i c popula t ions. Her e, w e pr ese n t a r etr o spect iv e c oh ort study o f ou t c omes o f uni v er sal e x ome sequencing in al l 365 adult pa tie n ts, ag ed 18-40 y ear s, admit t ed to an y ICU of the t er tia ry c ar e Uni v er si ty of P enns ylv ania Health S ys t em (UPHS) with g enomic d a t a a v ailable throu gh the P ennMedicine . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint BioBank (PMBB). W e in t egr at ed elec tr onic heal th r e c or d (EHR) inf orma ti on, chart r eview , and g enetic v ar ian t inf orma tion t o asses s r esul ts by se lf-identif ied r ace/ eth nicity and by patie n t knowledg e of the ir g enet ic diagnosis. W e iden tifie d diagnos tic r esul ts in n early 25% of cri ti c all y ill adul ts and c onc erning v ar ian ts of uncert ain signifi c ance (VUSs) in a fur ther 11% of p a ti en ts , with mos t r esults o ccurr ing in medically actionab le g enes. The inciden ce of diagnos tic v ar ian t s w as equiv alen t acr os s all r aces/ ethniciti es and did not decr ease with in cr easing pa tie n t ag e. How ev er , Black and Hispan ic patients w er e signif ic a n tly less lik el y to ha v e t heir g enetic diagnosis documen t ed in thei r medic al charts, a c once rning dispari ty giv en a 66% incr ease in mort ali ty obser v ed f or patie n ts with undocumen t ed g ene tic d iagnoses. Ov er all, ou r s tudy sug g es ts that the bene f its of g enomic t esting in cri tic al i llness is ag e-independe n t, and tha t the univ e r sal use of br oad g enetic t esting modalities in the crit ic a lly i ll adult popu la t ion migh t impr ov e heal thc are out c omes f or a ll patie n ts, r eg ar dless of ag e, r ace, or ethnici ty . . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint

Cohort Demographics and Characteristics Of the 43,612 PMBB participants with exome sequencing data, 365 met inclusion criteria for our study (Figure 1A, Methods). Altogether, 48 patients (13.2%) were age 18-25 years, 92 (25.2%) were age 25-29 years, 116 (31.8%) were age 30-34 years, and 109 (29.9%) were age 35-40 years at the time of first ICU admission ( Figure 2A, Table S1 ). Patients were roughly evenly divided by sex, with 200 patients (54.8%) identifying as female and 165 (45.2%) identifying as male ( Figure 2B). Examining the EHR-recorded race and ethnicity of the 365 patients in our study, 222 (60.8%) identified as White non-Hispanic, 102 (27.9%) identified as Black non- Hispanic, 22 (6%) identified as Hispanic/Latino, 8 (2.2%) identified as Asian, and 11 (3%) identified as an other race/ethnicity ( Figure 2C). The overall mortality rate for the study cohort was relatively low, with 30 patients (8.2%) dying during or after their hospital admission ( Figure 2D). Comparing the broad indications for ICU admission for the study cohort ( Figure 2E), we found that the most common diagnostic categories for ICU admission were: cardiac indications (n=77, 21.1%), cancer-related indications (n=71, 19.5%), vascular indications (n=46, 12.6%), infectious indications (n=25, 6.8%), immunologic indications (n=20, 5.5%), and renal indications (n=19, 5.2%). Broadly dividing participants into three groups by race/ethnicity – White (n=222), Black/Hispanic (individuals identifying as Black and/or Hispanic/Latino, n=124), and Other (individuals identifying as any other race/ethnicity, n=19), we found no significant demographic differences (Table S2). Specifically comparing the two largest groups, White and Black/Hispanic, we found no significant differences in age, sex, or patient status (deceased vs. alive). We did observe two nominally significant differences in ICU admission indications, with Black/Hispanic patients being significantly more likely to be admitted for infectious symptoms (p=0.0172) and less likely to be admitted for cance r-related issues (p=0.009). There were no significant differences in any of the other 13 ICU admission indications examined. V ariant Identification and Interpr etation Aft er r eview of medic al r ec o r ds and the CNV and Exomiser 20 reports generated from exome sequencing data , 18 7 suspicious g en etic v ar ian ts aff ec ting 166 g enes were ide n ti fied acr oss 146 of the 365 individuals (40.0%, Figur e 1B, T able S1 ). Of these v arian ts, 77 (41.2%) w er e missense v arian ts, 36 (19.3%) w er e fr ameshift v arian ts, 24 (12.8%) w er e nonsense v arian ts, 23 (12.3%) . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint w er e v a riants p r edic t ed t o aff ec t mRNA splicing , 18 (9.6%) w er e in -fr ame deletions o r inserti ons, six (3.21%) w er e lar g e del etions spanning mor e than one g ene, and one each (0.5%) w er e s t ar t lo ss, st op loss, or s ynon ymous v aria n ts. O f the 187 susp icious v a rian ts ide n ti fied, 143 (76.5%) acr oss 107 individuals (29.3% of all 365 individua ls studied) wer e f ound t o aff e ct g ene s c ausa t iv e of au t osomal dominan t dis ease, thr ee v ariants (1.6%) acr oss thr ee individuals (0.8%) w er e f ound t o a f f e ct g enes c aus a ti ve of X-link ed d isease, and 41 v aria nts (21.9%) acr oss 20 individuals (5.5%) w er e f ound t o aff ect g enes c ausa t iv e of au t osomal recessive disease. No suspicious v a riants aff ect ing the mit o chondrial g enome w ere ide n ti fied. Assessing v arian t pathog enic ity ( Figure 1B, T able S1 ), w e f ound tha t 111 of the 187 suspic ious v arian ts id en t ified (59.35%) had pr eviously been anno t at ed in ClinV ar 21 as pa thog enic or l ik ely pa thog enic, 31 (16.58%) had pr evio usly been ann ot at ed as v aria n ts o f uncert ain signifi c anc e (VUSs), and 47 (25.1 3%) had not b een pr eviously anno t at ed in ClinV ar a t all. F ollowing the American College of Medical Genetics and Genomics (ACMG) guidelines for clinical sequence interpretation,22 we reassessed all ClinVar-annotated VUSs: we reclassified six of the 31 VUSs (19.4%) as likely pathogenic while the remaining 25 remained classified as VUSs. Of the 47 variants lacking ClinVar annotations, we classified 20 (42.6%) as pathogenic/likely pathogenic while the remaining 27 (57.5) were classified as VUSs. None met criteria to be classified as benign/likely benign. Exome sequencing results identifying pathogenic/likely pathogenic variants were considered “diagnostic” if the result was considered relevant to the individual’s ICU admission; otherwise, pathogenic results were classified as “incidental.” All VUS results were classified as “VUS.” Altogether, we identified 95 diagnostic results, 27 incidental findings, and 44 VUS results across the 365 participants in the study ( Figure 1B, T able S1 ) . While 127 individuals (34.8% of the 365 studied) had a single finding identified (either diagnostic, incidental, or VUS), 19 individuals (5.2%) had more than one finding: four individuals (1.1%) were found to have two diagnostic findings related to their admission; eight individuals (2.2%) were found to have one diagnostic finding and one or more incidental findings; three individuals (0.8%) were found to have one diagnostic result and one suspicious VUS; two individuals (0.6%) were found to have two suspicious VUSs; and one individual each (0.3%) was found to have one suspicious VUS and one incidental finding, and two incidental findings. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint High Diagnostic Rate of Universal Ex ome Sequencing in Adult ICU Patients Acr oss all 365 individua ls included i n our s tudy , 90 (24.7%) w er e f oun d t o hav e one or mor e diagnos tic r esu lts relev ant t o their I CU ad mission ( Figure 3A ). An additional 40 (11.0%) w er e f ound t o ha v e one or mor e suspi cio us VUSs, and an additional 16 (4.4%) w er e f ound t o ha ve one or mor e i ncide n t al findings. O ver all only 219 individuals (60%) had c omplet el y neg a tive e x ome r esults. W e obser v ed no signif ic ant diff erences in diagnos ti c or VUS r a t es acr oss di ff er e nt r acial/ ethnic gr oups, although Whit e individuals had the h ighes t obser v e d diagnos tic r a te and low es t obser v ed VUS rate . The diagnos tic r at e in Whit e individua ls w as f ound t o be 27.0%, 21.0% in Black/Hispanic indi viduals, a nd 21.1% in individuals of an y o ther race/ ethnicity ( Figure 3B). The rate o f VUS iden t ifi c ation in these same gr oups w as 10.4% , 12.1% , and 10.5 % r espect iv e ly . Diagnostic Rate of Ex ome Sequencing Does Not Decrease with Increasing Patient Age Strikingly , and c o n tr a ry t o other pu blished s tudie s, 13–15,23 w e did n ot o bser v e an y s ignifi c an t decr ease in the diagnostic rate o f ex ome sequencing in adu lt pa ti en ts wi th inc r easing patient ag e (p=0.674 by logis tic r egr ess ion, Figure 3C ). The diagnos tic rate i n pati en ts ag ed 18-24 y ear s w as 22.9% , 28.3 % f or pa tie n ts ag ed 2 5-29 y ear s, 25. 9% f or pa tients ag ed 3 0-34 y ear s, and 21.1 % f or pa ti en ts ag ed 35-40 y ear s. Ther e w as a tr end t ow ar ds d ecr eas ing pr ev alence o f inc ident a l findings with incr eas ing pa tient ag e, with incident al f indings iden t ifie d in 8.33% of pa tien ts ag ed 18-24 y ear s, 5.4 3% of pa tients ag ed 2 5-29 y ear s, 2.59 % of pa tie n ts ag ed 30 -34 y ear s, and 3.67% in pa ti en ts ag ed 35-40 y ear s, although this did no t r eac h s ta ti s ti c al s ignifi cance (p = 0.069). Highest Yield of Ex ome Sequencing in Pulmonary , V ascular , and Renal Disease Patients Dividing pa ti en ts by indication for ICU admission, we found that certain indications were more likely to yield diagnostic results than others ( Figure 3D). The five admission indication groups with the highest diagnostic rate were pulmonary disease (n=11, diagnostic rate 81.8%), vascular disease (n=46, diagnostic rate 39.1%), renal disease (n=19, diagnostic rate 36.8%), gastrointestinal disease (n=14, diagnostic rate 35.7%), and cardiac disease (n=77, diagnostic rate 28.6%). The lowest diagnostic rates were observed for infectious diseases (n=25, . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint diagnostic rate 12.0%), endocrine disease (n=14, diagnostic rate 7.1%), and organ donor status, psychiatric disease, and trauma, all of which had a diagnostic rate of 0% (n=2, 2, 3, respectively). Cardiac, V ascular , and Malignancy-Associated Genes Predominate Gr ea ter than 50% of all d iagnos tic r esults ide n tif ied aff ec t ed g en es causing c ar diac (n=18, 18.9%), c ancer (n=16, 16 .8%), or v asc ular (n=16, 16.8%) phenotypes ( Figur e 4A ). The sa me w as true f or VUSs (c ar diac n =17, 38.6 % of VUSs; c ancer n=2, 4.55%; v ascular n=4, 9.09 %). While mos t g enes w er e only r epr ese n t ed once among the diagnos tic r esults (n=48, 51%), 13 w er e diagnosed mor e than once ( Figure 4B ). These include FBN1 (n=8), CFTR (n=7), TTN (n=6), BRCA2 (n=4), VHL (n=4), LMNA (n=3), PKD1 (n=3), and ACVRL1 , CACNA1S , MYH7 , NF1, PLN , and SMAD3 (n=2 each). No g enes appear ed mor e than once in the VUS da t aset. In the da t aset of inc ident a l findings, the majori ty o f g enes w ere identif ied o nly once (n=16, 59%, Figure 4C ), while fi v e g enes appear ed at least tw o times: ALPL (n=3), and BRCA2, MEFV , PALB2 , and TTN (n=2 each). Half of All Diagnostic Results Are Unknown/Undocumented in Patient Charts F or the 95 diagnos t ic e x ome r esu lt s w e identif ied, char t r eview r e v ea led th a t man y w er e unknown or undocumen t ed in the pa tie n t chart a t the time of ICU admission ( Figure 5A ). Ov er all, 44 of the 95 diagnoses w e re absen t fr om p a tie n t charts (46.3%). Ther e w as a t r end t ow ar ds olde r pa t ients being less l ik ely t o ha v e a do cumen t ed d iagnosis; diagnoses w ere documen t ed f or 58.3% of pa t ients ag ed 18-24 y ear s, 59 .3% of pa tien ts ag e d 25-29 y ear s, 6 2.5% of pa tie n ts ag ed 30-34 y ear s, but o nly f or 33.3% of pa tien ts ag ed 35-4 0 y ear s, how ev er thi s r ela t ionship wi th ag e w as not s t atisti cally signifi c ant by logistic regr ession (p = 0.232, Figure 5B ). Black/Hispanic Patients are Significantly Less Lik ely to Have Documented Diagnoses Comparing the r a t es a t which diagno s tic results wer e known and docume n t ed by r ace/ ethnic ity , signific a n t di ff er en ces emer g ed ( Figure 5C ). Whit e pa tients w er e signifi c an t ly mor e lik ely t o ha v e a do cumen t ed diagnosis c o mpar ed t o Black/Hispanic patients (p=0.003 by logis t ic r egr ession). Diagnoses w ere documen t ed f or 63.5% of Whit e pa tie n ts an d 75% of pa tien ts o f other r ace/ ethnic ity , but only f or 28.6% of Black/Hispanic pa t ients. This disparity c anno t b e . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint e xplained by demogr aphic dif f er en c es ( T able S2 ) , or dif f erences in ov erall diagnos tic r at e of e x ome sequencing betw een the se di ff er e n t gr oups ( Figure 3B ). Patients with Unknown/Undocumen ted Diagnoses Hav e Higher Mortality Rates The ov er all mort a lity r a te f or parti ci pan ts in our s tudy w as thank full y rela ti v ely small ( 8.2%, Figure 2D ), but w e obser v ed d iff e rences in mort ali ty r at e by diagnos ti c s ta tus ( Figure 5D ). Ov er all, 6.0% of pa tien ts with chart- documen t ed diagnos ti c r esult s died either during or aft e r their hosp it al iz ation, c ompar ed t o 8. 7% of pa tie n ts with n eg a tiv e e x ome r esults and 10.0% of pa tie n ts with diagnos ti c but undocu men t ed e x ome sequencing r esults. These dif f er ences d id not r each s ta tisti c al signifi c ance, but the 66% incr ease in mort ality rate obser v ed f or pa tie n ts with undocumen t ed diagnos tic res ults c ompar ed t o patients with do cumen t ed diagnos tic r esults is no t able. Patients with Docume nted Di agnoses Have Significantly Longer ICU Admissions W e ne xt e x amined the leng th o f sta y (L OS) in the ICU f or our c ohor t, divided by d iagnos tic stat u s . Pat i e nt s w i t h chart-document ed diagnos t ic r esu lts had a mean ICU L OS of 3.0 days (median 2 da y s), c ompar ed t o 2.2 d a y s (median 1 da y) f or pa tien ts with n eg a tiv e e x ome r esult s and 1.9 da y s (median 1 da y) f or pa t i en ts wi th diagnos t ic but undocument ed diagnostic e x ome sequencing r esu lts. The di ff er en ce in L OS obser v ed f or p a tie n ts with und ocumen t ed d iagnos ti c r esults c ompar ed t o p a tie n ts with documen t ed diagnos t ic r esu lts w as signific a n t by ANO V A (p=0.027). Gener a ting K aplan-Meier cur v es t o c ompar e the L OS betw een gr oups by sur viv a l analy sis the same signifi c an t d iff e r en t w as seen by log-r ank t es t (p=0.045, Figure 5E ). Universal ex ome sequencing r eveals medically actionable results in 8% of patients Ex cluding pa tien ts with known diag noses, 41 of the 36 5 individuals that w e s tudied (11.2%) w ould ha v e g ained a diagnos t ic result fr om univer sal ex ome sequencing , 38 (10.4%) w ould ha ve g ained a diagnosis of a susp icious VUS, and 19 (5.2 %) w ould ha v e g ain ed a diagnosis of an incident al pathog enic finding ( Figure 5F ). Of all of the 166 g enetic findings w e iden t ifie d in our s tudy , 108 (65.1%) occurr ed in g enes with specif ic manag emen t guidelines describ ed in th e . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint NCBI r esour ce “ Gen eR eview s” ( T able S1 ). 24 F or diagnostic r esult s, 72 (76.6%) occurr ed in g enes with medic ally manag emen t r ec ommenda tions; f o r inc ident al f indings 17 occurr ed in medic al ly actionable g en es (60.7%), and f or VUS r esults 19 (43.2%) occurr ed in genes with medic al ly manag e men t r e c ommenda tions. F or diagnos t ic results that w er e spec ifi c all y known/documen t ed in the pa t ient chart 42 (82.4%) occurr ed in g enes with clearly define d medic al manag emen t guidelines, whil e f or results that w ere not known/documen t ed 30 (69.8%) occurr ed in medic al ly actionab le g enes. This su g g es ts tha t f or 30 pa tien t s (8.2% of our en tire c ohor t), e x ome sequencing a t the t ime of admission migh t ha v e sug g es ted specifi c manag emen t chang es. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint

Our s tudy is, t o ou r knowledg e, the fir s t of i ts kind t o in v esti gate the u tili ty of broad e x ome sequencing in the c rit ic al ly ill y o ung adult popul a tion. The r esul ts ser v e as a pi v o t al adv ancemen t in our und er st anding of the impact o f Mende lian d isease in adults, h ighligh ting the f r equency of g eneti c c ontribu tion t o cr iti c al illn ess in th is und er s tud ied population. W e f ind tha t n early a quar t er of all adult U PHS I CU pa tients ag ed 18-40 ha v e a Mendelian g eneti c diagnosis r elat ed t o the ir cri ti c al il l ness. F or ov er 75% of pa tie n ts, the se g enetic diagnoses c onf er spec ific c ar e-alt e ring manag emen t r e c ommenda tions. No t ably , we find c ons ist ently h igh yield of e x ome sequencing acr oss al l ag e br ack ets e x amined, ev en in thi s pa tie n t popul a tion unselec t ed f o r en richmen t of suspec ted g enetic diagnoses. These r esu lts h ighligh t the f act th a t g enetic t es t ing is no t mer ely r e lev an t t o the y oung or t o those wi th a priori susp icions of a g enetic diso r der . The impli c a t ion h er e is prof ound: there e xists an u n t apped po t ential for g enetic t esting tha t could benefit a much br oader adult demogr a phic than pr eviously r ec ogniz ed, alt e ring medic al manag emen t f or a lar g e number of critic a lly ill patie n ts. These findings st and to r eshap e our appr oa ch t o g enetic as sessmen ts in the adu lt patient popu la ti on, which has all t oo of t en b een ov er look ed f or the impleme ntat i o n o f b r oad genetic tes t ing. Our find ings s t and in con tr ast t o th e tr aditiona lly h eld view tha t the l ik el ihood of unc o v ering g enetic diagnoses dec r eases wi th inc r easing pa tie n t ag e bu t is in k eep ing with other s tudi es o f the yie ld o f e x ome/ g enome sequenc ing in adults. In a study o f whole g e nome sequencing in 100 g ener ally health y p a tie n ts ag e d 40–65 y ear s in th e primary c ar e set ting , V ass y et al . disc ov e r ed a new g enetic diagnosis i n 22%, 16 although man y pa tien ts e x hibit ed only minimal s ympt oms of their d iagnosis. In oth er s tud ies o f adult pa t ients with su spicion f o r a g enet ic diagnosis, 14,15,17–19 the diagnos t ic r a te of e x ome sequencing has been r eport ed t o r ang e fr om 14-29 % – low er than tha t obser ved in the pedia t ric popul a ti on, 25 but high enough t o demons t rate t h e u t i l i t y o f b r oad g e netic t e s ting appr oaches in adults. The diagnos tic r at e o f 24% tha t w e report in our s tudy is in k eeping with th ese pr ev ious reports, despite the f a ct th at the population w e e x amined w as not specifi c all y selec t ed f or su spici on of g enetic disease. This . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint sug g es ts tha t the pr es ence of c rit ic al illness in a young adult is as good a pr edict o r of Mende lian disease as “ susp icion for a g enetic dis ease” as det ermined by a g eneti cs-tr ained pr ovide r . The high yield of br oad e x ome seque ncing tha t w e report in cri tic a lly i ll ad ults c an be pu t in th e c ont e xt of a number of r ece n t s imilar studies o f br oad e x ome/ g enom e sequencing in t he criti c al ly ill ped ia tr ic and neonat al pa t ien t popul a ti ons tha t ha ve r epor t ed d iagnos tic r a tes of 21- 38%. 2–4,9,10 The diagnos tic yi eld tha t w e r epo rt h er e is essentiall y the same as th ose r epor t ed in these studies, sug g es ting tha t cri ti c al illn ess is a s tr ong pr ed ic t or o f Men delian diseas e acr os s the li f e span. Import antly , man y of t hese s tud ies f ou nd signif ic a n t c ost s a vings and import ant chang es in manag emen t due t o r api d e x ome/ g enome sequencing. This, along with our r esul ts iden t ify ing specifi c medic al manag emen t r ec ommenda tions f o r ov er 75 % of iden t ifi ed g eneti c diagnoses, sug g es ts tha t r apid e x o me/ g eno me sequencing in cri ti c ally ill adults migh t pr ove similarly beneficia l a t impr oving pa tie n t c are and low ering health c ar e cos ts, although pr ospecti v e studies a r e needed t o v ali da t e this h ypoth esis. In nearly half o f all p a ti en ts wi th diagnos tic e x ome sequenc ing r esult s, w e find that this diagnosis is unknown t o the patient or their tr e a tmen t t eam. W e also f ound tha t patie n ts with documen t ed g enetic d iagnoses had signific a n tly long er ICU stays and tr e nded t ow ar ds ha ving low er mort ali ty r at es than patie n t s with unknown/undocumen t ed g e netic diagnoses. W e c annot be ce rt ain wh y the p r esen ce of a known g enetic diagnosis w as associa ted with signific a n tly long er ICU leng th of s t ay , but it is possib le th a t th ese patie n ts ar e tr e a t ed mor e c autious ly than the ir c ou n t e rparts wi thout known g enetic d isor de r s, and i t is p ossible th a t thi s migh t e xplain the low er mort a lity ra t e th a t w e obser ved in th is gr oup. Futur e pr ospect ive s tudies ar e n eeded to under st and how critic a l c ar e manag emen t chan g es in the f ace of a g enetic diagnosis, and whether making such a diagnosis truly dec r eases pa tie n t mort al ity . This c onsidera tion is not triv ial; the sh ift t ow ar d pr e cision medic ine could he r ald a mor e e ffici ent alloca tion o f r esou r ces and a r educ ti on in health c ar e e xpendi tur es by t a il oring in t er v entions t o individual g ene tic pr o fil es, leading t o impr ov ed outc omes and mor e c ost-e ff ect iv e c ar e. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint One of the mos t disc onc erting r ev e lations of our s tud y is the s t ar k dispari t y in the aw ar eness of g enetic diagnoses along r acial and et hnic lines. W e f ound no di f f e r ence in the diagnos tic ra t e of e x ome sequencing betw een Whit e and Blac k/Hispanic pa t ients; how ev er , our dat a unambiguously show tha t Black and Hispanic pa tie n ts a r e subs t a n tia lly le ss lik el y t o hav e the i r g enetic diagnosis known and documen t ed c ompar ed t o their Whit e c ount erpar ts. This is alarming in the c ont e x t of th e 66% incr eased mort a lity ra t e we obser v ed in patie n ts with undocumen t ed g ene tic d iagnoses. This is mor e than a st a t is t ic; i t is ind ic ativ e of a s ys temic f ailur e in the del iv e ry and implemen ta tion o f clin ic a l g enetic t esting and precision medicin e that pot entiall y c o n tr ibut es t o the s ignifi c an t heal thc a r e dispar ities obs er v ed in non-Whit e pa tie n ts. 26,27 The r eal ity that a p a ti en t's r ace or ethnic ity c ou ld influen ce the ir lik e lihood o f r eceiv ing a t imely and accur at e g en etic diagnosis is a glaring c all t o a c tion. It demands an in t e rr og a tion int o the ac cess, edu c a t ion, and biases that per v ade our he althc are s ys tems, and sug g es ts tha t broad, univ ersal t es t ing appr oaches such as thos e e x amined in our s tudy ma y be needed t o er adica te such inequ iti es. One s trik ing f ea tu r e of our r esu lts w as the high rate at w h i c h we r epea t ed ly identif ied pa thog enic v a riants in the same g e ne in diff e r ent pa ti en ts, with ha lf of all diagnos tic r esul ts af f ec ting g enes app earing mor e tha n once in ou r dat a. Fi v e g enes alon e – FBN1 , CFTR , TTN , BRCA2, and VHL – w er e f ound t o un derly the diagnoses of 29 pa ti en ts i ncluded in our s tu dy , r epr ese n ting 30% of all diagnostic results. This is in c o n tr as t t o what h as been obser v ed i n s tudies o f br oad sequ encing in pedi atric ICU popula t ions, wher e r e curr e nt diagnoses mak e up only a small portion of the diagnos tic r esults. 2,3 Addit ionall y , the r ecur r ent g enes w e iden t ify a r e not ov erlapping with the f ew r ecu rren t g enes identif ied in these ped ia t ric populations. This sug g es ts tha t th e g enetic landsc ap e of Mende lian diseas e in c rit ic al l y ill adu lts ma y be qualit atively di ff eren t, and po t ential ly mor e s ter eo typed, than wha t is obser v ed in pedi a tr ic popula ti on. How ev er , it is also possi ble tha t the spec ific e xper tise o f f er e d a t the t ert iary c ar e hospit al s y st em that w e s tud ied d i r ectl y influ enced this finding , with cert ain diseases and phenotypes be ing specifi c all y enrich ed in the pa t ient population o f UPHS. F or e x a mple, the lar g e and w ell r eg ar ded Aort a Cent e r a t UPHS a t tr acts pa tie n ts from ar ound the c ountr y with . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint aortic aneur y smal disease 28 and ma y ha v e r esult ed in the la r g e number of pa tients with FBN1 pa thog enic v ariants ide n ti fied i n our study . Furthermor e, and similar t o the above poin t, th e v as t majori ty of d iagnoses w e identif ied r esu l t in diseases o f adult onset. W e did not e x clude pa t ients with known g eneti c diagnoses fr om ou r s tudy , and thus the c onspicuou s s c arcity of pediatric-onset g enetic s yndr o mes in our adult ICU c ohor t r aises pr e ssing questions ab out th e i n ter sect ion o f g enet ics wit h pa ti en t sur viv a l and healthc a r e de livery . Ar e indiv iduals w ith ped ia tr ic-onset g enet ic c ondi tions less lik e ly t o sur v ive in to adulthood? Ar e t hey not e xperi encing criti c al i llness i n adulthood? Ar e they s til l seek ing criti c al c are trea tmen t a t pediatr ic hospi t als? Is clini cian bias dis c our aging ICU -lev el tr ea tment? 29 Ar e they not enr ol ling i n the PMBB , a r esear ch-based c ohor t tha t r equi r es sp ecif ic c onsent, and thus not c aptur ed by this s tudy? The answ er s t o these ques tions r emain elusi ve and clearly mor e r esear ch is needed t o under st and this discrepancy . How ev er , it is clear that a r e-ev alua tion of ou r appr oach t o g en etic t esting thr oughout a pa tie n t's li f espan is needed, with the need f or a dynamic model tha t adapts t o the ev olv ing clini c al presen ta tion o f g enetic disease acr o ss the li f espan. G eneti c diseases pr ese n t d iff e r en t ly in a dulthood, of t en with s ympt oms tha t ar e not p r ese n t in ch ildhood and adolesc ence, and the sa me c ar e models an d appr oaches that ha v e been pr o v en eff ect iv e i n pediatric popu la t ions ma y not be appr opri a t e f o r adult pa t ients with g eneti c disease. In c onsidering the br oader impli c a t ions of our s tud y , it is par amoun t t o acknowledg e it s limit a t ions. Fir st and f o r emos t, th is w as a single-cen ter s tudy , which co mes with a number of limit a t ions and pot e n tia l c onf ounders as discussed abov e. Sec ondl y , this w as a r etr ospective s tudy that uti liz ed r es ear ch-gr ade sequencing da t a. The r etr ospec tive na ture of the study mak es it impossibl e f or us t o know how a genetic d iagnosis migh t ha v e aff ec t ed pa ti en t c a r e in r ea l- time, and it is po ssible that some of our sequenc ing r esults, c omplet ed b y a non-CLIA -certifie d sequencing lab, ma y be spur ious. Tha t being sa id, pr evious s tud ies hav e s hown CLIA v alida tio n of r esear ch-gr ade biobank e x ome sequencing t o e x ceed 99%. 30 Las tl y , a major limit a tion of ou r w ork is that it dep ends en t ir el y on p a tie n t enr olment in PMBB. Although ov er all demogr aphics . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint ar e not signif ic a n tly di ff er e n t betw ee n PMBB and UPHS, 31 i t i s poss ible th a t th e popu la t ion we s tudied repr ese n ts a subset of pa t ie n ts that is princ ipally d iff eren t fr om the ov er all UPHS IC U- admit t ed patient popu la ti on. These limit ations necessi t at e c aut ious i n t erp r et ation of th e findings and under s c or e the need f o r pr ospect iv e, multic ent er studies o f clini c al-gr ade e x ome sequencing t o v a lidat e and e x t end ou r r esul ts. In c onclus ion, our find ings pain t a p ic tur e of a medi c al lands c ape on the cu sp of tr ans f o rma tion. The evidence f or the utili ty of uni v ersal e x ome sequencing in criti c ally i ll adults is c ompell ing , with new diagnoses and pot en tial manag emen t chang es iden tif ied in mor e than 10% of all adult pa tie n ts ag ed 18-40 y ear s admit t ed to an y ICU . As w e s t and on the brin k of this new fr o n tie r , i t is incumben t up on us t o f or g e a p a th f o rw ar d th a t no t only r e c ogniz es the impor t ance o f Mendelian g enet ic d isease in adul t morbidity and mort al ity , bu t ac tively in c orp o rate s t h i s knowledg e i n t o r ou tine clin ic a l pr ac tice t o a v o id e x ace rba t ing health c are dispar ities and to impr ov e health c ar e ou t c omes ac r oss the boar d. It is th r ough such e ndea v o r s that w e ma y s tri ve t ow ar d a future wher e pr ec ision medicine is not a lo fty id eal but a s t anda rd c omponen t of adul t criti c al c ar e and a t angible and po t e n t ially l if e-sa ving r eality f o r all p a tie n ts. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint

PMBB patient recruitment and exome sequencing The Penn Medicine BioBank (PMBB) 31 is a University of Pennsylvania academic biobank which recruits patient-participants from the University of Pennsylvania Health System (UPHS) around the greater Philadelphia area in the United States. Appropriate consent was obtained from each participant regarding storage of biological specimens, genetic sequencing, and access to all available EHR data, and permission to recontact for future studies. The study was approved by the Institutional Review Board of the University of Pennsylvania and complied with the principles set out in the Declaration of Helsinki (protocol # 854452). This study included the subset of 43,731 individuals enrolled in PMBB who had previously undergone exome sequencing. Briefly, for each individual, DNA was extracted from stored buffy coats and exome sequences were generated by the Regeneron Genetics Center (Tarrytown, NY) and mapped to GRCh38 as previously described.32 For quality control (QC), sample-level filtering was as follows: individuals with low exome sequencing coverage (less than 75% of targeted bases achieving 20× coverage) or with high missingness (greater than 5% of targeted bases) were removed from analysis, leaving 43,612 samples after sample-level filtering. Variant-level filtering was as follows: in each sample, all single nucleotide variants (SNVs) with a total read depth < 7 were changed to “no-call”, and similarly all insertion/deletion (INDEL) variants with a total read depth < 10 were changed to “no-call.” Subsequently we removed any variant sites where no sample carried an alternate allele balance ≥ 15% (SNVs) or 20% (INDELs). Patent cohort definition Of the 43,612 PMBB participants with exome sequencing data, we excluded individuals with no record of admission to a UPHS intensive care unit, leaving 4,590 individuals. We next removed any individuals with ICU admission diagnosis codes belonging to ICD-10 chapters XIX (Injury, poisoning and certain other consequences of external causes) or XX ( External causes of morbidity and mortality ), resulting in the exclusion of 289 additional individuals. Lastly, we excluded all patient whose age at earliest recorded ICU admission was less than 18 years or greater than or equal to 40 years, leaving a total of 365 individuals for inclusion in the study. Copy Number Variant (CNV) Report Generation Copy number variants (CNVs) were annotated, starting with the same exome sequencing data as described above, using version 1.3 of the CLAMMS pipeline. 33 Standard quality control measures were taken both at the sample level (samples with >40 CNVs or with >40,000 exons . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint called as CNVs were removed) and chromosome level (for samples with >10% of a chromosome covered by >1 CNV call, that chromosome was removed). QC levels ranging from 0-3 were assigned to each CNV call based on Q_non_dip, Q_exact, and allele balance and heterozygosity metrics. Only CNVs meeting the most stringent QC threshold of 3 were included in the analysis. For each individual in the study, a table of the resultant CNV calls, along with affected protein coding genes, was generated for manual review as described below. Patient phenotype definition For the 365 patients included in our study, all International Statistical Classification of Diseases (ICD)-9 and ICD-10 disease diagnosis codes and procedural billing codes ever associated with their care were extracted from the electronic health record (EHR). ICD-9 codes were mapped to ICD-10 using the Center for Medicare and Medicaid Services 2017 General Equivalency Mappings (https://www.cms.gov/Medicare/Coding/ICD10/2017-ICD-10-CM-and-GEMs.html) with unmappable ICD-9 codes dropped from further analysis. For each individual, each unique ICD-10 code was subsequently mapped to a Human Phenotype Ontology (HPO) code using equivalency mappings as published by McArthur et al. 34 and the unique set of HPO terms, per individual, was passed to Exomiser as input. Exomiser Report Generation For each individual included in the study, a list of HPO terms and the QCed VCF file of exome sequencing results, both generated as described above, was passed to the tool Exomiser v.13.2.1, 20 a bioinformatic tool for prioritizing potential disease-associated genetic variants for manual review using random-walk analysis of protein interaction networks, clinical phenotype comparison with known patients based on HPO terms, cross-species phenotype comparisons, as well as a wide range of other computational filters for variant frequency and predicted pathogenicity. Exomiser was run using default/recommended settings considering all possible inheritance patterns, incorporating variant allele frequencies from gnomAD, exAC, and TOPMED, and excluding variants annotated as affecting non-coding regions of the genome. For each individual included in our study Exomiser output in HTML formal was generated for manual review as described below. Chart Review, Variant Pathogenicity Assertions, and Diagnoses For each of the 365 individuals included in our study, two physicians, both trained and board certified in Internal Medicine and Medical Genetics, reviewed the CNV report, Exomiser report, . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint and electronic health record of each patient to identify genetic variants that might be relevant to each patient’s ICU admission. A short clinical summary of each patient’s medical history was prepared and documented. Suspicious genetic variants deemed at last possibly related to an individual’s ICU admission by both physicians (considering both the patient phenotype, variant classification, and the disease inheritance pattern) were selected for further review. Altogether, 187 suspicious variants affecting 166 genes were identified across 146 individuals ( Figure 1B). Using the American College of Medical Genetics and Genomics (ACMG) guidelines for clinical sequence interpretation, 22 we classified/reclassified all variants either absent from the ClinVar database, or with ClinVar entries of VUS only. Subsequently, for all pathogenic/likely pathogenic variants, exome sequencing results were classified as being “Diagnostic” if the pathogenic variant was deemed relevant to the patient’s ICU admission and fit the appropriate inheritance pattern for the disease of interest, or classified as being “Incidental” if the identified pathogenic variant was deemed medically relevant, but not directly related to the patient’s ICU admission. For autosomal recessive diseases, results were considered diagnostic if at least one of the two variants identified was pathogenic/likely pathogenic and if the associated disease phenotype was felt to be a strong match for the patient’s presentation. All results with only VUSs identified were classified into a third group, “VUS.” Patient charts were specifically reviewed to determine if the identified genetic variant had previously been identified clinically and documented in the chart. Patients with non-molecularly confirmed clinical diagnoses were considered to have been previously diagnosed for this purpose. We subsequently classified all genes with identified variants based on the primary organ system affected, and by the presence or absence of specific medical management guidelines in the NCBI GeneReviews resource. 24 Data Visualization and Statistical Analysis All data analysis and visualization was completed in RStudio using R version 4.3.0 and the following packages: ggplot2 v.3.4.3, ggsankey v.0.0.9, dplyr v.1.1.3, tidyverse v.2.0.0, survival v.3.5. Statistical analyses were carried out, as indicated, by constructing linear/logistic regression models or by log-rank test as indicated in the figure legends. For all analyses, two- sided p-values less than 0.05 were considered nominally significant. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint Figures: Figure 1. Cohort Definition and V ariant Interpretation In panel A, the ov er all w o rkflow f or c ohort d e f init ion, r epor t gener a t ion , and chart r eview i s shown. St arting with the 43,731 PM BB parti cipants with e xom e sequenc i ng da t a, w e t ook th e subset of ind ividual s under the age of 40 who had ever be en adm it ted t o an y Uni v er s ity o f P enns ylv ania Health Syste m (UPHS) ICU with an y In t erna tional S t a ti s t ic al Classi fi c a ti on o f Diseases (ICD)- 9/10 ad m ission diag nosis c od e othe r than th ose f all ing under the c at eg ory o f “Injury , poisoning and ce rt ain other c onsequenc es of e xt e rnal c auses” or “Ext ernal c auses o f m orbidity and m ort alit y , ” lea ving 365 individuals in our c ohort. Exom e V arian t Call F orm a t (V CF ) files f or each pa rtic ipan t w ere subjec ted to qual ity c o n tr o l (QC) as shown, a nd all ICD c odes ev e r associat ed with each participa n t ’ s m edic al r ec o r d w er e e xtr ac t ed and m apped t o hu m a n phenotype o n t o log y (HPO) t erm s as described in the m ethods secti o n. These da t a w er e supplied as inputs t o Exom iser v .12. 2.1 and CLAMM S v .1.3 t o gene rate a n E xo m i ser and c op y num ber v arian t (CNV) r eport, which w er e subsequen tly r ev iew ed by tw o Clinic al Geneti cis t s , both boar d cert ified in Int erna l Medi cine and Medic al Genet ics, with c on com it an t r eview of th e pa tie n t cha rt. In pan el B , the w ork flo w and r esul ts o f v ar ian t ide n ti fi c a ti on and int e rpr eta tion i s shown. Of the 365 partic ipan ts, 14 6 had one m or e m or e suspic ious varian ts identif ied o n s e f f f f ) r n e y , e s n . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint either/both the Ex omiser or CNV r eport. W e classified th ese v aria n ts by their associ a t e d inheri t ance p a tt ern and by Clin V ar c la ssifi c a t ion, i f a v ailab le. F or all ClinV ar- annot at ed VUSs, and f or v aria n ts absen t f orm ClinV ar , we classifi ed v ar ian ts using th e Americ a n Colleg e o f Med ic a l Genetics and Genomics (ACMG) guidelines f or c lini c al sequenc e i n t e rpret a tion. The r esu lting v arian ts w er e subsequently classi fied as eithe r diagnos t ic, inc ident al, or VUS, as described in the

section. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint Figure 2. Demographics of the Study Cohort Ov er all dem og r aphics of the 365 par ticipants in cluded in our s tudy are s hown. P anel A show s per cent of the study popul a tion f a lli ng in t o each of f ou r age g r oups a t the tim e of earlie s t IC U adm ission. P anel B show s the per c en t of pa rtic ipan ts by EHR- r epor t ed pa tie n t se x. P anel C show s the percent of pa rtic ipan ts by EHR- r eport ed r ace and ethn icity . P anel D show s th e per cent of part icipants by p a tie n t s ta tus (de ceased or al iv e) a t the tim e of discha r ge. P anel E show s the per c en t of par ticipa n ts d iv ided by r eason f o r ICU adm ission, br oadly divid ed int o 1 5 indica tion g r oups. s U C e E 5 . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint Figure 3. Results of universal ex ome sequencing in critically ill adults age 18-40 year s. (A) Ov er all diag nos ti c rate o f exo m e sequencing in the c om plet e c ohor t of 365 participan ts , along with the r at e of VUS ide n tif ica tion and incide n t al diag nosis ide n ti fica tion. In th is fig ur e each participa n t is c ou n ted only onc e, only f or the hig hes t or der findi ng disc ov e r ed on e xom e sequencing , with diag nos t ic r esult s b eing ev alua ted f ir s t, VUSs se c ond, an d incide n t a l f inding s las t. (B) O v er all d iag nos tic r at e o f e xom e sequencing as in panel A, but s tr a t ifi ed by EHR - r epor t ed r ace/ ethnic ity . No sig nifi can t di ff er en ces in the r a te of dia g nos tic f inding s w a s obser v ed by log istic r eg r ession be tween an y of the th r ee g r oups. (C) Ove r all diag nos ti c r a te o f e xom e sequencing as in panel A, but s tr a t ifi ed by pa ti en t age at y ounge s t ICU adm ission. N o sig nific a n t corr e la t ion w as f ound betw een pa tie n t age and diag nos tic r at e or VUS rate b y l o g i st i c r eg r ession. (D) O v er all d iag nos tic r ate of e xom e sequencing as in panel A , but s tr a ti fied by th e r eason f or ICU adm ission, as in Figure 2E . , e e s - s f o c e . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint Figure 4. Distribution of genes identified by ex ome sequencing acr oss different r esul t categories (A) Pr oportion of e xom e sequencing r esults c o lor ed by the diag nos tic ca teg ory of the in volve d gene. R esult s are str a t ified by d iag nos tic, VUS, and incide n t a l results. (B ) Fr equenc y o f gene s im plic at ed in diag nos t ic f inding s. Genes iden t ifie d only onc e within th is d a t aset ar e c ol lect iv e l y labeled as “ Genes Appearing Only O nce. ” The r em aining genes, appearing tw o or m or e tim e s acr oss all d iag nos tic r esul ts, ar e di spl a y ed. Color coding c or r esponds t o th e c at eg ories d e fined i n panel A. (C) Fr equ ency o f genes im plica t ed in incide n t al find ing s. Gene s identif ied on ly onc e within this da t ase t ar e c ol lect iv el y l abeled as “ Genes Appearing Only O nce. ” The r em ainin g genes, appearing tw o or m or e tim es acr oss all incid ent al r esul ts, ar e di spla y ed. Color codin g c orresponds t o the c at eg ories d e fin e d in panel A. t d s y s n e g g . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint Figure 5. Racial/ ethnic disparities in docume ntation of genetic di agnoses and implications fo r clinical outcomes (A) P er cen t of all 95 diag nos t ic r esul t s ide n tif ied on e xom e sequenc ing which w e r e known an d docum en t ed in p a tie n t charts prio r t o ICU adm ission (y es, in g r een ) or unkn own/undocum en t e d in pa t ien t charts (no , or ange). ( B) P ercen t of diag nostic exom e r esul ts kno wn and docum en t ed , as in pan el A, but s tr atif ied by patient age. No sig nifi c ant c or r elation w as f o und betw een pa t ien t age and docum e ntat i o n o f genetic diag nosis by log is tic r eg r ession. (C) P er cen t of diag nos ti c e xom e r esults known and docum en t ed, as in pan el A, but str a t ifi ed by EHR- r ec or de d r ace/ ethnicity . Bla ck p a tie n ts and Hispanic pa t ients ar e sig nifi c antly l ess lik ely t o hav e a docum en t ed geneti c diag nosis com par ed t o W hite pa t ien t s by log is ti c r eg r ession. (D) Mo rt al it y rate fo r a l l 365 pa tie n ts in cluded in our s tud y , s tr atif ied by r esu lt o f e xom e sequencin g (diag nos tic vs. nega tiv e) and by doc um e ntat i o n o f r esul ts in the pa ti en t chart. P a tien ts wit h docum en t ed diag nos t ic r esul ts w er e l ess lik ely t o die c om par ed t o patie n ts with nega tiv e e xom e r esults, who w er e le ss l ik ely t o d ie c o m par ed t o pa ti en ts with undocum ented diag nos ti c results , althoug h these di f f e r ence w e r e not stat i st i ca l l y s i g n i f i c ant by log is ti c r e g r ession. (E) K aplan - Meier cur v e illustr a t ing the l ik el ihood of r em aining in the ICU by da y of h os pit ali z a t ion, s tr atif ie d r d d , t c d a y g h e , - d . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint in to the same thr ee gr oups as in p anel D . P a tien ts with documen t e d diagnos tic r esul ts had signific a n tly long er leng ths of st a y in the ICU by log-r ank t es t. (F) P er cen t o f pa ti en ts tha t woul d g ain a n ew dia gnosis not alr eady documen t ed in their el ectr on ic medic al r ec or d, s tr atif ied by diagnos tic c at egory , acr os s all 365 pa tie n ts included in the study . Gr e en bar s ind ic at e th e pr oporti on, within each diagnos tic ca t egory , th a t w ou ld speci fi c ally g ain a diagnosis in a g en e with clearl y defined manag emen t r e commenda tions. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint Declaration of Interests The authors declare no competing interests. Acknowledgments We acknowledge the Penn Medicine BioBank (PMBB) for providing data and thank the patient- participants of Penn Medicine who consented to participate in this research program. We would also like to thank the Penn Medicine BioBank team and Regeneron Genetics Center for providing genetic variant data for analysis. We thank Daniel Rader and Michael P. Hart for their

and input on study conception and design. The PMBB is approved under IRB protocol# 813913 and supported by Perelman School of Medicine at University of Pennsylvania, a gift from the Smilow family, and the National Center for Advancing Translational Sciences of the National Institutes of Health under CTSA award number UL1TR001878. T.G.D. is supported in part by the NIH K08 1K08DK127247 and by the Burroughs Wellcome Fund. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint

1. Biesecker, L. G. & Green, R. C. Diagnostic clinical genome and exome sequencing. N. Engl. J. Med. 370, 2418–2425 (2014). 2. NICUSeq Study Group et al. Effect of Whole-Genome Sequencing on the Clinical Management of Acutely Ill Infants With Suspected Genetic Disease: A Randomized Clinical Trial. JAMA Pediatr. 175, 1218–1226 (2021). 3. Dimmock, D. et al. Project Baby Bear: Rapid precision care incorporating rWGS in 5 California children’s hospitals demonstrates improved clinical outcomes and reduced costs of care. Am. J. Hum. Genet. 108, 1231–1238 (2021). 4. Olde Keizer, R. A. C. M. et al. Rapid exome sequencing as a first-tier test in neonates with suspected genetic disorder: results of a prospective multicenter clinical utility study in the Netherlands. Eur. J. Pediatr. 182, 2683–2692 (2023). 5. Australian Genomics Health Alliance Acute Care Flagship et al. Feasibility of Ultra-Rapid Exome Sequencing in Critically Ill Infants and Children With Suspected Monogenic Conditions in the Australian Public Health Care System. JAMA 323, 2503–2511 (2020). 6. Farnaes, L. et al. Rapid whole-genome sequencing decreases infant morbidity and cost of hospitalization. NPJ Genomic Med. 3, 10 (2018). 7. Lumaka, A. et al. Rapid Whole Genome Sequencing Diagnoses and Guides Treatment in Critically Ill Children in Belgium in Less than 40 Hours. Int. J. Mol. Sci. 24, 4003 (2023). 8. Freed, A. S. et al. The Impact of Rapid Exome Sequencing on Medical Management of Critically Ill Children. J. Pediatr. 226, 202-212.e1 (2020). 9. French, C. E. et al. Whole genome sequencing reveals that genetic conditions are frequent in intensively ill children. Intensive Care Med. 45, 627–636 (2019). 10. Bupp, C. P. et al. Breaking Barriers to Rapid Whole Genome Sequencing in Pediatrics: Michigan’s Project Baby Deer. Child. Basel Switz. 10, 106 (2023). . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint 11. Sanford Kobayashi, E. et al. Cost Efficacy of Rapid Whole Genome Sequencing in the Pediatric Intensive Care Unit. Front. Pediatr. 9, 809536 (2021). 12. Kamolvisit, W. et al. Rapid exome sequencing as the first-tier investigation for diagnosis of acutely and severely ill children and adults in Thailand. Clin. Genet. 100, 100–105 (2021). 13. Marouane, A. et al. Lessons learned from rapid exome sequencing for 575 critically ill patients across the broad spectrum of rare disease. Front. Genet. 14, 1304520 (2024). 14. Posey, J. E. et al. Molecular diagnostic experience of whole-exome sequencing in adult patients. Genet. Med. Off. J. Am. Coll. Med. Genet. 18, 678–685 (2016). 15. Mainali, A. et al. Diagnostic yield of clinical exome sequencing in adulthood in medical genetics clinics. Am. J. Med. Genet. A. 191, 510–517 (2023). 16. Vassy, J. L. et al. The Impact of Whole-Genome Sequencing on the Primary Care and Outcomes of Healthy Adult Patients. Ann. Intern. Med. 167, 159–169 (2017). 17. Hou, Y .-C. C. et al. Precision medicine integrating whole-genome sequencing, comprehensive metabolomics, and advanced imaging. Proc. Natl. Acad. Sci. U. S. A. 117, 3053–3062 (2020). 18. Walsh, M. et al. Real world outcomes and implementation pathways of exome sequencing in an adult genetic department. Genet. Med. 24, 1536–1544 (2022). 19. Shickh, S. et al. Exome and Genome Sequencing in Adults with Undiagnosed Disease: A Prospective Cohort Study. J. Med. Genet. 58, 275–283 (2021). 20. Robinson, P. N. et al. Improved exome prioritization of disease genes through cross- species phenotype comparison. Genome Res. 24, 340–348 (2014). 21. Landrum, M. J. et al. ClinVar: public archive of relationships among sequence variation and human phenotype. Nucleic Acids Res. 42, D980-985 (2014). 22. Richards, S. et al. Standards and Guidelines for the Interpretation of Sequence Variants: A Joint Consensus Recommendation of the American College of Medical Genetics and . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint Genomics and the Association for Molecular Pathology. Genet. Med. Off. J. Am. Coll. Med. Genet. 17, 405–424 (2015). 23. Farwell, K. D. et al. Enhanced utility of family-centered diagnostic exome sequencing with inheritance model-based analysis: results from 500 unselected families with undiagnosed genetic conditions. Genet. Med. Off. J. Am. Coll. Med. Genet. 17, 578–586 (2015). 24. GeneReviews®. (University of Washington, Seattle, Seattle (WA), 1993). 25. Chung, C. C. Y . et al. Meta-analysis of the diagnostic and clinical utility of exome and genome sequencing in pediatric and adult patients with rare diseases across diverse populations. Genet. Med. Off. J. Am. Coll. Med. Genet. 25, 100896 (2023). 26. Ong, M.-S. et al. Racial and ethnic disparities in early mortality among patients with inborn errors of immunity. J. Allergy Clin. Immunol. 153, 335-340.e1 (2024). 27. Olufosoye, O., Soler, R. & Babagbemi, K. Disparities in genetic testing for breast cancer among black and Hispanic women in the United States. Clin. Imaging 107, 110066 (2024). 28. Kreibich, M. et al. Influence of Age and the Burden of Ischemic Injury on the Outcome of Type A Aortic Dissection Repair. Ann. Thorac. Surg. 108, 1391–1397 (2019). 29. Jett, J. et al. Retrospective review of the code status of individuals with Down syndrome during the COVID-19 era. Am. J. Med. Genet. C Semin. Med. Genet. e32080 (2023) doi:10.1002/ajmg.c.32080. 30. Blout Zawatsky, C. L. et al. Returning actionable genomic results in a research biobank: Analytic validity, clinical implementation, and resource utilization. Am. J. Hum. Genet. 108, 2224–2237 (2021). 31. Verma, A. et al. The Penn Medicine BioBank: Towards a Genomics-Enabled Learning Healthcare System to Accelerate Precision Medicine in a Diverse Population. J. Pers. Med. 12, 1974 (2022). . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint 32. Park, J. et al. Exome-wide evaluation of rare coding variants using electronic health records identifies new gene-phenotype associations. Nat. Med. 27, 66–72 (2021). 33. Packer, J. S. et al. CLAMMS: a scalable algorithm for calling common and rare copy number variants from exome sequencing data. Bioinforma. Oxf. Engl. 32, 133–135 (2016). 34. McArthur, E., Bastarache, L. & Capra, J. A. Linking rare and common disease vocabularies by mapping between the human phenotype ontology and phecodes. JAMIA Open 6, ooad007 (2023). . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint SUPPLEMENTAL INFORMATION List of Supplemental Tables Table S1. Exome sequencing results and associated participant characteristics and demographics Table S2. Demographic characteristics of the study cohort, divided by EHR-reported race/ethnicity . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint PENN MEDICINE BIOBANK BANNER AUTHOR LIST AND CONTRIBUTION STATEMENTS PMBB Leadership Team Daniel J. Rader, M.D., Marylyn D. Ritchie, Ph.D. Contribution: All authors contributed to securing funding, study design and oversight. All authors reviewed the final version of the manuscript. Patient Recruitment and Regulatory Oversight JoEllen Weaver, Nawar Naseer, Ph.D., M.P.H., Giorgio Sirugo, M.D., P.h.D., Afiya Poindexter, Yi-An Ko, Ph.D., Kyle P. Nerz Contributions: JW manages patient recruitment and regulatory oversight of study. NN manages participant engagement, assists with regulatory oversight, and researcher access. GS assists with researcher access. AP, YK, KPN perform recruitment and enrollment of study participants. Lab Operations JoEllen Weaver, Meghan Livingstone, Fred Vadivieso, Stephanie DerOhannessian, Teo Tran, Julia Stephanowski, Salma Santos, Ned Haubein, P.h.D., Joseph Dunn Contribution: JW, ML, FV, SD conduct oversight of lab operations. ML, FV, AK, SD, TT, JS, SS perform sample processing. NH, JD are responsible for sample tracking and the laboratory information management system. Clinical Informatics Anurag Verma, Ph.D., Colleen Morse Kripke, M.S. DPT, MSA, Marjorie Risman, M.S., Renae Judy, B.S., Colin Wollack, M.S. Contribution: All authors contributed to the development and validation of clinical phenotypes used to identify study subjects and (when applicable) controls. Genome Informatics Anurag Verma Ph.D., Shefali S. Verma, Ph.D., Scott Damrauer, M.D., Yuki Bradford, M.S., Scott Dudek, M.S., Theodore Drivas, M.D., Ph.D., Contribution: AV, SSV, and SD are responsible for the analysis, design, and infrastructure needed to quality control genotype and exome data. YB performs the analysis. TD and AV provides variant and gene annotations and their functional interpretation of variants. . CC-BY-NC-ND 4.0 International licenseIt is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted April 1, 2024. ; https://doi.org/10.1101/2024.03.11.24304088doi: medRxiv preprint