“You need to be able to stand up for what is right”:MTV Shuga Naija’s transformative impact on youth attitudes towards sexual violence in Nigeria

Background In Nigeria, approxima tely one in t en w omen o f reprod uc tive age repor t e xperiences of sexual violence in the pas t year , wi th po tentiall y end uring conseq uences. The i mpact o f sex ual violence can be particularly devastating for yo unger women and adolescen ts, who may face long-lasting physical, psychological, and social e ffe cts. To a d dress this, MT V Sh uga Naija utilizes ent ertainmen t e duca tion programming, an chored by a TV serial drama, to promot e gen der equalit y and challenge sexual violence norms and behaviors. This research exa mines the impac t o f MT V Sh uga Naija on disclosure of sex ual violence, stig ma an d victi m-bl aming at titu des, and great er dialogu e ab out se xual violence. Metho ds This evaluation uses a panel survey o f Nigerian you th age d 15-24 (574 f emale s; 317 males) with da ta collecte d be fore an d af ter the airing o f MTV Sh uga Naija program ming. The b aseline wave was conduc te d in p erson, while the endline wave was cond uct ed via telephone d ue to th e COVID -19 pandemic . Using th e Theory of Planne d Behavior, survey qu estions foc used on norms, a ttit u des, beliefs, and behaviors related to sexual harassment and sexual violence , we ana lyze chang es in self- reported sexual harassmen t, a tti tu des to wards victi ms, an d discussions with a d ults and fa mily memb ers about sexual violence for tho se in in MTV Sh uga Naija program area s relative to those in two comparator stat es. The analysis uses a fixe d-e ffe cts, do ubly robust di fferen ce-in-dif ferenc es (D ID ) es timation s trateg y to improve t he comparability o f treat men t an d con t rol groups by adjusting for observed dif ferences , ther eby allowing for more pre cise esti matio n of the causal ef fec ts of MTV Sh uga Naija .

At baseline, the majority o f both males and f emales across all stu dy areas repor ted at tit udes bla ming victims rather than perpe trators for exp eriences of se xual violence. Ho wever, r esults from th e doubly-robus t D ID mo dels reveal significant changes in th ese at tit ud es among youth in areas exposed to MTV S hu ga Nija program m ing, as well as grea ter disclosure of inci dents o f se xual violence to int erviewers, possibly d ue t o greater a wareness of wha t consti tu tes sexual violence. For example , agree men t with the s tate men t, “wo men who wear cloth es tha t e xpose their bo dies are . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 3 asking to be rape d,” de creased by 36.6 percentage points [95% CI: -46.8pp, -2 6.3pp] for wo men an d by 31.8 pp [95% C I: -45.6pp, -18.1pp] f or men in areas with MTV Sh uga Naija sexual violence programming relative to those in co mp arison areas. Similarly, acceptanc e of th e at titu de tha t "it is not rape if a woman does no t fight ba c k” decline d by 28.3 pp [95% CI: -40.2p p, -16.4pp] for wo men and by 19.3 pp [95% C I: -35.6pp, -3.0p p] for m en in treat men t versus co mparison areas. Contrary to hypotheses, respon dents in comparison areas showed larger increases in the likelihood of talking with paren ts and fa mily abou t sex ual violence than responden ts in areas targe t ed by MT V Shu ga Naija , although no informa tion was collecte d on the initia tors, circu mstanc es, nor conten t of dialogue with paren ts and fa mily m emb ers, rendering dif ficult the interpre tatio n of this unexp ec ted finding.

This study provides eviden ce that the e ntertainm ent ed uca tion approach of MT V Shu ga Naija can indee d bring abou t significant progress in shifting at titu des an d b ehaviors related to se xual violence. Even with this progress, ho wever, vic tim-blaming nor ms and at titu des, as well as experiences o f sexual violence amon g women -- an d men - - are still wi dely prevalen t. While MTV Sh uga Naija has made hea dway in shifting the landscape surrounding this oft en-taboo subject i n Nigeria, more work is neede d to un derstan d ho w co m muni cation and dialogue wi thin fa milies, sch ools, and comm unities, alongside grea ter e fforts t o support victi ms of se xual violence , ca n be improve d an d enhanced to ensure con tinue d progress . Key words: sexual violence , stig ma, you th, Nigeria . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 4 Intro duction This paper analyzes da ta from a pan el survey of Nigerian a dolescents an d yo un g adul ts in order to estimat e th e impa ct o f M TV Sh uga Naija , an en tertain ment e duca tion program that aire d in Nigeria from 2018 to 2020, on at titu des towar d s sexual violence an d acknowle dge men t of e xperiences o f sexual violence. Th e program was desig ned on the premise that ent ertainm ent- edu cation can leverage the power o f story telling an d me dia to raise a wareness, change at titu des, an d pro mote positive behavior, particularly for taboo topics and d eeply ingrained norms, suc h as acceptan ce o f gender-base d violence ( GBV). Pervasive gender-base d sex ual violence exposes yo ung women to a m yriad o f mental an d ph ysical challenges, warranting the develop men t of e ff ective solu tions tha t a ddress stig ma, vic tim-blaming , and harm ful g ender norms. Th e possible post- trauma tic health conseq uences o f sex ual violence are numerous bo th in the short and long ti me, including unwan te d pregnanc y, se x ually transmit te d infections, gynecological an d reprod uc t ive health problems, physical trau ma, c ardiovascular disease, diabetes, d epression, anxiet y, eating disorders, e motional distress, suicidality , a nd subseq uent engagem ent in po ten tially harmf ul beha viors (e.g., risky sex, s moking, alcohol, drug use) (Ellsberg e t al., 2008; Fle ming K., 2016; Jina & Tho mas, 2013; Me ekers et al., 2013; Ngu ye n et al ., 2021). Worldwide , approxi mately one in three wo men have experience d either ph ysical and/or se xual violence by an inti mate partner or se xu al violence by a non-partn er at some po int in their lives. Among ever- married or partn ered adol escent girls, approxi mately one in four have experience d physical and/or se xual violence from a n intimat e partner , with 16 perc ent exp eriencing it in the prior 12 months (World H ealth Organiza tion, 2021). Sexual violence against wo men, tho ugh und er-reporte d, is also prevalent in Ni geria. According to the 2018 Nigeria Demo graphic and He alth Survey (N DHS ), 9.1% o f women a ged 15 to 49 years reported tha t they ha d ever experienc e d sex ual violence, an d 4.1% reporte d e xperiences of sexual violence in the pas t year . Kad una Sta te , one of the stu dy sites this paper, repor ted even higher rates; 13.8% of women age d 15 to 49 years h ad ever experience d se xual violence, inc luding 7.9% in the last year ( National Population Com mis sion (NP C ) [ Nigeria] & I C F, 2019). Th e WHO es timat es tha t about 13% o f women of reprodu ctive a ge in Nig eria had e xperience d se xual vi olence wi thin the last 12 months (Sardinha e t al. , 2022; World H ealth Organiza tion, 2021). Moreove r, the 2014 Violence Against Children S urvey in Nigeria indi cate d that the lif eti me prevalence of se x ual violence among . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 5 youth 13 to 24 y ears is approximately 2 5% ( National Pop ulation Comission ( NP C) Nigeria, 2016; Nguy en e t al. , 2021). Public discourse about gen der-based se xual violence an d enco uraging th e disclosure of inci dents are crucial steps to wards a d dressing the pre valence and a ccep tance o f sex ual violence, which is frequen tly driven by gend er imbalances , male control, and inequi table gen der a nd an drocentric norms (Antai & An tai, 2008; Gag e & T homas, 2017; P ulerwit z & Barker, 2007; Umana et al ., 2014; World Health Organization, 2021). Ho wever, vic tims’ disclosure to au thorities, fa mily m embers, or others of experiences of g end er-based s exual violence may b e inhibite d beca use of sha me, religious beliefs, gen der po wer imbalanc es, or fe ar of retribu tion (Babalola e t al. , 2015; Fontes & Plum mer, 2010; Nguyen et al., 2021; Ullman et al. , 2008). Nega tive social reactions or disbelief and lack of trust in state ag encies can fur ther discourage disclosure (Ahrens C. E., 2007), resulti ng in low rat es of victims receiving assistance ( Sumn er S. A., 2015). For exa mple, a Nig erian stu dy o f f emal e victims o f sex ual violence age d 13 to 24 revealed tha t only one-third o f the vic tims discussed the incident wi th someon e, with frien ds (38 %) an d parents (32%) being the most co mmon confi dants. Fu rthermore, only one in seven victims k new where to seek help, and only 4% rec eived a ctual assist ance. Many victi ms w ere afraid of the p otential conse quenc es of disclosure (30%), while o thers did n ot believe that reporting the inci dent w as necessary (35%) or think that it was a significant issue (16 %) (Iku teyijo e t al. , 2022; Iku teyijo, 2022; Ng uyen et al., 2021). The latt er reasons likely stem from a tacit acc eptance , an d po tentially even an endorsemen t, of violence against wo men as a social nor m (Gilbert et al., 2022). The M TV Sh uga Naija program consiste d of a TV serial dra ma that is support e d by radio programming an d social me dia an d aim ed to challenge sexual violence norms, promote gen der equality , and enco urage a ttit udinal an d behavior change regarding se xual violence among youn g people. S torylines portrayed th e e xperie nces of young wo men who were victi m s of sex ual violence, including how th ey en coun tered stig ma and sough t jus tice an d support fro m a dults an d pe ers. The program showcase d how great er dialog ue, bo th p ublicly and with parents an d other ad ults, coul d foster lesser tolerance of violence and g reater assistance for vic tims, leading to broader social change. The advantag es of the ent ertainm ent- e duca tion for mat are nu merous. Such m edia initiatives have low access costs , enabling wide reach w ith au diences through television, radio, and film . Comp elling storylines can make conten t relatable a nd eng end er emo tional connec tions wi t h characters an d . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 6 situations, as well as present positive ro le mo dels whose a tti tu des an d behaviors can be e mulat ed by viewers and listeners. Cap tivating narratives can stim ulate conversations about dif ficult topics within comm unities an d social net works, con t ributing to social change (Singhal, 1999, 2004; Singhal & Rogers, 2001; Usdin, 2004; Y ue Z, 201 9). As evidence o f the e ff ec tiveness of MT V Shu g a’s en ter tainmen t e duca tion app roach, an earlier evaluation of prior seasons, which foc u sed on HIV /AIDS and safer sex , used a randomiz ed controlled trial to de monstrate positive causal impacts o f MTV Shuga program s on atti tu des towards dom estic violence, as well as on HIV / AIDS kno wledg e an d at tit ud es, risky sexual behavior, an d reduc tions in STDs a mong fe male yo ut h (Banerjee, 2019). A 2018 stu dy linked exposure to S eason 3 of MT V Shu ga, which containe d a subp lot about a married co uple with a viole nt husband , with improved a ttit ud es among male viewer s, who were 21 p ercent less likely to jus tify do mestic violence (Banerjee, 2019). This research seeks to f urther explore t he impac ts of MTV Sh uga , focusing spe cifically on issues of sexual violence amon g Nigerian you th. To fur ther con tribute to the eviden ce b ase for en tertainm ent edu cation program ming, the stu dy uses a quasi-exp erimen tal approach tha t co mpares out com es for youth in Lagos, wh ere program ming with th em es of se xual violence aire d, with youth in th e sta tes of Kad una and Kano , where su ch programming was absent . Th e principal research objective is to estimat e wh ether MTV S hu ga Naija was ef fec tive in red ucing victi m-blaming a t titu des, increasing reporting of experiences o f se xual hara ssment an d sex ual violence an d increasing interpersonal

about se xual violence a mon g Nigerian yo uth an d their parents . We use an analytical approach, do ubly robust , fi xe d e ffe cts Diff erence-in- Diff erences (D ID ) mo de ls, that adjus ts for possible confounding du e to unobserve d he terogenei ty b et ween trea tmen t an d comparison individuals, which coul d oth erwise po te ntially lead to over- or und er-estima tes of th e true effe ctiveness of MT V Shu ga Naija .

Intervent ion The design an d conten t o f MT V Sh uga Naija is guide d by the Theory o f Plann ed B ehavior (Ajzen, 1991), which posits that changes in the precursors to targe te d behaviors – at tit udes, subjec tive norms, and perceive d behavioral contr ol – lea d even tually to changes in behavioral intention and , in . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 7 turn, to changes in behaviors. Sh ug a pro gramming is designed to includ e con te nt specific to behavioral beliefs, normative belie fs, an d con trol beliefs. The MTV S hu ga Naija t elevision and radio dra mas confron t challenging issues by fea turing diverse characters who express a variety of at tit udes, norms, and b ehaviors. The show presents new , positive attit ud es, norms , and behaviors for you ng people to e mulat e. A t the sam e time , it also depic ts traditional or harm ful a tti tu des, nor ms, and behaviors, d emonstra ting ho w th ese can nega tively affec t youth . On the issue of se xual vio lence, storylines illustrate d ho w b ehaviors such as respecting the au tono my o f sex ual partners, suppo rting victims o f sex ual violence, an d no t engaging in sexual violence led to be tt er ou tco mes, while a lso depicting the possible negative cons equenc es of har mf ul behaviors, both for vic tims an d for p erpetrators of sexual violence . In th e 2018-2020 seasons of MTV Shu g a Naija , issues of se xual violence were portrayed in t wo key storylines – the rape o f a high school girl at a party thro wn by older boys and t he rape of a sex ually experience d “vid eo vixen” and aspiring singer by an influ ential m usic prod ucer . Smaller storylines – such as a t eacher soliciting sex ual favor s from a fe male st ud ent to i mprove her grades or a male stud ent su ggestively touching fe male st uden ts – depict ed forms o f se xual hara ssment. Storylines showed the victims s truggling wi th the after-e ff ects o f se xual assault an d harassment , mo deling proactive behaviors to seek justic e, an d ultima tely prevailing over en trenche d norms and social struct ures, thereby o ff ering viewers a p ath to d eal with th eir own experiences of sex ual violence. The targe t popula tion for M TV Sh uga Naija was you th in the sta tes o f Kad un a, Kano , an d Lagos, although the con ten t in Lagos was broader than in th e north ern t wo sta tes. In Lagos, storylines touche d no t jus t on se xual harassment and violence b ut also on se xual e du cati on and repro duc tive health. From March-April 2018, ten 22 -minut e episod es of the television serial drama were aired on MTV satellite an d NTA2 Lagos ( Nigeri a Television Authority) and again 18 m onths later. The episodes on N TA2 were available only for viewers in the Lagos area while tho se airing on MTV satellite w ere pot entially viewable an yw here in the coun try, although access wa s generally considerably lower o utside of Lagos. E pisodes also aired on the ww w. mtvshug a.com websit e and on the MTV Shuga YouT ube chann el. Th e episodes were comple ment ed by a radio dra ma, social me dia (e.g ., m tvshuga.co m website , Fa cebook , Instagra m, Twit ter ), an d pe er e duca tio n sessions with you th in Lagos. A second 8-episode television drama series that contin ued th e storyli nes in the earlier series were shown in O ctober /Nove m ber 2019. This series aired on MTV sat ellite, MTV Base, . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 8 Startimes TV, Pop Central, NTA , ON TV, and EBON YLIF E TV, alongside a more ex tensive radio drama, peer e du cation, and social me dia. Beginning in 2019, MTV Sh uga com me nced li mite d program activities in Ka du na and Kano, foc used solely on reproduc tive health and absent con tent and storylines abou t sex ual violence. Ac tivities include d a television magazine , a ra dio drama, social me dia, an d pe er e duca tion. The mes of this programming Data To assess the ef fec ts of Sh ug a on the tar get gro ups, surveys were con du ct ed am ong a rando mly sampled cohor t o f urban yo uth . Consistent wi th th e st epwise intro duc tion of t he intervention , you th were int erviewe d at thre e points in time : (1 ) prior to the imple menta tion of any Sh uga Naija activities in any of the s tates (baseline) , (2 ) a fter i mplemen tation of Sh uga Naija a ctivities in Lagos but not elsewhere (mi dline), and (3) af ter the i mplemen tation of Sh uga Naija a ctivities in all three s tates. We focus only on the baseline and en dline waves of data because at mi dline, many of the critical story arcs regarding sexual violence were still emerging , meaning that ke y learnings f or youth were not fully revealed . At baseline, the target sample size for t he stu dy was co mpu te d as 2,000 fe male s and 1,000 males betw een the ag es of 14-24 years. With permission from the Bill and Melinda Gates Instit ute of the Johns Hopkins Univers ity, Baltimore , t he Cen ter for R esearch, Evalua tion Res ources, an d Developmen t ( CRER D ) use d the Perfo rmance Monitoring for Ac coun tability 2020 (PMA2020 2017) Nigeria sample fra me in Kano, L agos, and Ka duna s tates to selec t urban clusters for the Sh ug a survey. Utilization o f PMA2020 enu me ration areas (EAs) , which w ere rando m ly selecte d from each state’s census en umera tion areas, meant that th ere was a previously established and recen t sa mpling frame o f all househol ds locate d in PM A2020 EAs. At baseline, in-person intervie ws were complet ed with 1,900 fe males and 986 males. As wi th most panel surveys, at trition of responden ts occurred across survey waves. Of the b aseline participants, 70% agreed to be re-con tact ed for subs equen t int erviews, an d a tte mp ts were made to re-con tact these individ uals for the en dline survey. Because o f con tact restrictions due to t he COVI D-19 pan demic and to compl y with COV ID-19 regulations, data collection for the en dline survey was und ertaken through pho ne interviews. Phone calls were ma de to all respon dents in th e baseline and midline da tabase who had valid phone . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 9 numbers an d consen ted to callbacks. 1 T he switch to phone surveys co mpoun d ed earlier at trition, resulting in an en dline sample o f 899 re spondents. Of these, only 641 in to tal ( 399 females and 242 males), participate d in all three waves. Survey In stru ments The MTV S hu ga Naija survey instru me nts were d esigned to capture informa ti on on Sh ug a exposur e and indica tors from th e Theory o f Planned Behavior (Ajzen 1991) that guid ed the developm ent o f MTV Sh uga Naija program ming. The q uestionnaire foc used on issues tha t wer e specifically addresse d by the storylines for ea ch of the MT V Sh uga Naija’ s main charac ters. For e xample , questions were asked abou t: 1 ) behavio rs or experiences (e.g ., “Has som eone e ver touche d you in a way that ma de yo u feel uncom fortable? ” “Have yo u ever talked wi th your pare nts or fa mily memb ers about sexual violence?” ); 2 ) a ttit ud es ( e.g ., “ It 's not rape if a girl does n't fight back” ); 3) subjective norms (e. g., “Mos t of m y fri ends think I should use condo ms wh en having sexual intercourse”) ; 4) perceive d behavioral c ontrol (e .g. , “Ho w easy or di ffic ult wou ld it b e for you to use condoms with a sex ual partner?”) ; an d 5) behavioral intentions ( e.g ., “D uring t he nex t 3 months, how likely is it that you will try to p ersuade yo ur partner to use con doms every time you have sex?”) . Because the en dline survey was con duc ted by phon e, a shorter q uestionnaire w as required to minimize responden t fa tigue and drop- off. Q uestions fo cuse d on a s maller set of indica tors from the T PB that were de em ed essential to cover the theory ’s broad do mains and behavioral outcomes (e.g ., contrac eptive use , H IV testing, ex periences and discussions of sex ual violence) . Huma n Subjects Research App roval For the baseline survey, National Healt h Research Com mit tee ( NHR EC ) appr oval was obtained on January 24, 2018 with NHR EC /01/01 /2007-24/01/2018. Institu tional Revie w Board ( IRB ) approval from the Tulane University Bi ome dical IRB was grant ed on February 18, 2018. For the midline survey , I RB approval from the NHR EC was ob tained , which spans August 23, 2019 through Au gust 22, 2020 (N HRE C /01/01/2007-23/08/2019B). Approval from the Tulane University Biomedical I RB was grante d on Sep temb er 20, 2019. For the endlin e survey, CR ER D received approval on July 9 , 2020. Approval from the Tulane University Bio m edical I RB was granted on Aug ust 5, 2020. 1 At baseline, 1,306 females (68.7%) and 712 males (72.2%) consented to be recontacted. . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 10 Statistical A nalysis Our analytical approach is to estima te t wo-period , t wo-group di fferen ce-in- dif ferences mod els, comparing changes over tim e in key o u tcom es in th e trea t ment area of Lagos relative to those in the comparison areas of Kad una and Kano . The primary assump tion un derlying th e D ID mo del is tha t, in the absence of the trea tm ent ( MT V Shu ga Naija ) , average o ut comes in trea t ment and control areas would have follow ed the same tren d from baseline to en dline. Consequ ently, a ny dif ferenc es observed in those tren ds at en dline are attribut ed to the causal influen ce o f M TV Shu ga Naija . In the DI D models, we inclu de a time d um my variable (baseline wave=0, en dline wav e=1) , a binary variable indicating if a p erson resides in the trea tmen t area o f Lagos versus the control areas of Kad una and Kano , an d an in teraction b etw een the two . The e ffec t o f the in tera ction t erms is th e Average Treatm ent Ef fec t on the Treat ed (ATT) . More explicitly , this is a mea sure of the mean impact o f MT V Sh uga Naija on th e pop ulation in areas tha t were e xposed to it. We make use o f th e panel na ture o f the data set to estima te a fix ed ef fec ts regr ession model for each outco me rela ted to exp erience o f sex ua l violence and for at tit udes to wards vict ims of se xual violence. These mod els capt ure differe nces bet ween trea tm ent an d control gr oups through the group fixe d ef fec t an d con trol for any t emporal fac tors tha t af fec t bo th treat m ent an d con trol groups equally through the tim e fi xe d e ffec t. Beca use th ese mo dels also con tain an individual fixe d effe ct , any tim e invariant variable for a n individual drops out of the mo del. T his means tha t th e binary variable for living in the trea tm e nt or co mparison area drops ou t, altho ugh th e interac tion with the binary survey wave variable remains. Because we are concerne d abou t the va lidity of th e parallel tren ds assumption , we use a metho d that combines th e fi xe d-ef fec ts model with inverse probability weighting (I PW ). Th is approach adjusts for poten tial confo unding by weighting observations by the inverse of their pr obability of rec eiving the treat men t based on their surveye d c haracteristics (Jof fe et al. , 2004). The re sulting estima tor, known as the do ubly robust ( DR ) es timator, ( Callaway, 2021; Lunce ford & D avidian, 2004) is robust to misspecification of either the ou tco me (fix ed ef fec ts) regression mod el or t he IPW mo del. The DR mo del in e ffe ct produ ces a weight e d co mbination o f th e es timat es from th e out com e and I PW mod els, wi th th e weights re flec ting th e degree of overlap be t ween the trea tme nt and con trol groups. The D R esti mator consistently estima te s the treat men t e ffe ct parame ters provide d eith er th e regression or the propensity score is co rrectly specifie d. . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 11 While the depen den t variables in each c ase are binary, w e esti mat e th e mod els using a linear fixe d effe cts regression mo del in order to av oid th e inciden tal parame ters problem inherent in non-linear fixe d e ffe cts mod els ( Chamberlain, 198 0). This problem arises because nonline ar fixe d e ffe cts mod els treat individual fixe d e ff ects as parameters to be estima te d, an d, wh en the nu mber o f time periods is small, these mo dels may pro duc e inconsistent estima tes o f the in dividual fix ed ef fec ts, in turn lea ding to inconsistent estima tes o f oth er parame ters in th e model . Cons equen tly, estima tion of the ATT using standar d sta tistical sof t ware like Sta ta is not straight forward , By using th e linear fix ed ef fec ts model , we are able to direc tly id entif y the ATT , th e primary focus o f our stu dy, as the coe fficien t on the inte raction ter m for trea tm ent area and survey wave. Ho wever, we are unable to esti mate average pre di cte d probabilities for observed values o f covariates (e.g ., the predicte d o ut come for those rec eiving t he trea tmen t versus those no t receiving the trea tmen t ) because the in dividual-specific eff ec t, t he intercep t for each individual across a ll time p eriods, is diff erence d ou t o f the mo del thro ugh t he within transforma tion or “ de meaning” of the da ta inherent to fix ed ef fec ts models. Once these individ ual-specific e ff ects are di ff erenced ou t in the fixe d e ffe cts mod els, w e can asc ertain c hanges in the out come variable from c hanges in individual covariates but no t pre dic tions of the o u tcom e variable for diff erent values of c ovariates (Wooldridg e, 2002). Conse quen tly, we present only the average trea tmen t ef fe cts fro m the DI D mod els. All models were es timat ed in Sta ta 18.0 using the drdid com mand d eveloped by Sant ’Anna (2020). This estimator i mplemen ts th e doubly r obust diff erence-in- diff erences es tima t or for the average treat ment e ffec t with the inverse probability of tilting (IPT ) an d weigh ted least squares (WLS ). We also sought to a ddress the possible influence o f chang es in th e sampling pr ocedure associated with moving fro m fa ce- to-fac e intervie ws, as was done for the baseline wave, t o phone interviews , as was necessitat ed by COVI D-19 in th e e ndline wave . Our s trategy involves adju sting our sample weights through an iterative proportional fitting proced ure (Berg mann, 2011). Because the en dline sample was restric te d to responden ts w ho had consen te d to being re-conta cte d by phone, it contained responden ts wi th a dif ferent socioeconomic profile on average than the samples for th e first t wo waves. The unw eighte d endlin e sample, for e xample , was foun d to be more ed ucat ed on average; 27.7% of fe males an d 34.5% o f males in the endline ha d pos t-secon da ry ed ucation as compared wi th 11.8% an d 12.5% respe ctively at mi dline. . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 12 Using the itera tive proportional fit ting procedure (also known as “survey raking”), we adjus te d our sample weigh ts for endline respon dents so that th e demo graphic profile ma tch ed that of the baseline sample in ter ms of gen der, s tate , e du cat ion and marital stat us. Sa mple weights f or the en dline sample were it eratively a djuste d to minimize th e discrepancies with the kno wn popula t ion distribution o f characteristics of the baseline sample u ntil the dif ferenc es bet ween th e t wo sa mples converge d to pre-specified tolerance level. 2 Survey ra king was cond uc ted using the ipfraki ng and ipf weig ht comman ds in S tata 17 ( Kolenikov, 2014, 2019).

De scriptive R e sul ts The characteristics o f the sample chang ed from baseline to en dline, with mos t of th e changes attributable to ma turation an d transitions to ad ulthoo d. All respondents ag ed by three years fro m baseline to en dline; the share o f the fe male sample that was ag ed 13-17 years decrease d fro m 51% a t baseline to 25% at endline in the trea t ment area (Lagos ) an d fro m 42% to 18% in the co mparison area (Ka duna and Kano ). Similar patt erns were observe d in th e male sample. The fe male sample in th e trea t ment are as had higher e du cational at tainmen t a t both baseline an d endline relative to the co mparison sam ple. At bo th baseline an d en dline, fe males in Lagos were approximately 14 p ercentag e points mo re likely to have co mple te d secon dary e duca tion relative to females in the comparison areas (Trea t ment : 85% a t baseline, 81% a t en dline; Comparison: 72% a t baseline, 67% at en dline) . For males, ed ucational at tainmen t was higher in the t reatmen t area a t baseline (75% of males wi th secon dary edu cation in the treat men t area versus 59% in the comparison area) bu t by en dline great er parity was eviden t (79% of males wi th secondary e du cation in the treat men t area versus 82% in the comparison areas). Across all areas, few males, a t bo th baseline and en dline, reporte d that the y w e re currently married , consistent with pa tt erns of consid erable later ages a t marriage for males relativ e to fe males in Nigeria. F emales in the comparison are as were approxi mately three ti mes mor e likely to be marrie d at en dline than fe males in th e treat ment area. R elate dly, rou ghly one in four wo men in the comparison areas had a child by the en dline versus only 8% o f wo men in the t reatmen t area . Less 2 Co nver gence is achieved if the lar gest relative differe nce of t he weig hts in tw o successive iteratio ns (a full cycle over al l raki ng va ri ab les) does n ot exceed a to lera nce o f 0.000001. . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 13 than 3% o f m en in the co mparison areas reported having a child a t en dline, re lative to 75 o f m en in the treat men t areas. At both baseline and endline, th e majority of you th o wne d a c ellphone. Wo me n in the trea tmen t area appear to have got ten a cellphone sooner than wo men in the co mparison areas; 63% of wom en in Lagos owned a cellphone a t baseline versus 51% of wo men in the comparis on areas. By en dline, roughly eight in ten wo men in both are as (trea tmen t: 84%, co mparison: 80%) owned a c ellphone. Approximately equal p ercentag es of ma les in treat men t and co mparison areas owned cellphones at both baseline and at endline. The majority o f youth report exposure to both radio an d television, with t elevision being the more commonly use d than radio. At endline, women in trea tm ent areas were slightly more likely to wa tch television than wo men in comparison a reas (trea tmen t: 98%, co mparison: 90% ) bu t e qually likely to listen to the ra dio ( treat men t: 59%, co mparison: 58%). Males w ere e qually likely to wa tch television at en dline ( treat men t: 96%, co mparison: 95%), bu t males in the trea tm ent area were more likely to listen to the ra dio ( treat men t: 80%, co mparison: 63%). . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 14 Table 1a. Sample Charac teristics, by Su rvey Wave and Sex ( with rake d weights ) F emales Co mp aris on /N orth Treatment/L ag os Baseline % End li ne % Baseline % End li ne % Age a t l ast birt hd ay 13-17 42.2 17.5 51.3 24.7 18-21 36.7 48.0 24.6 44.1 22+ 21.1 34.4 24.1 31.2 Leve l of Ed uca tion None 6.3 5.6 0.0 4.7 Pr ima ry 9.5 18.3 0.7 0.8 Secon da ry 71.9 67.4 85.4 80.7 Hig her 12.3 8.7 13.9 13.8 Marit al s ta tus Curre ntly m ar ried/l ivi ng wit h someo ne 31.4 25.5 8.3 7.5 Not ma rr ied 68.6 74.5 91.7 92.5 Age a t firs t sex Never sex 64.4 62.0 61.6 56.4 Don’t K now /Refused 0.0 3.7 1.9 2.7 You nger t ha n 15 8.2 2.8 3.2 0 15-17 12.5 14.4 9.1 6.3 18-21 13.2 14.8 21.7 30.7 21+ 1.7 2.4 2.4 4.0 N u m b e r o f ch i l dr e n e v e r bor n 0 73.2 75.6 90.1 92.4 1 6.0 6.0 6.9 5.5 2 8.5 6.5 2.1 1.9 3+ 12.3 12.0 0.8 0.3 Own a ph o ne? 51.4 79.8 62.6 84.0 Ever watc h TV? 79.1 90.1 97.0 97.8 Ever l isten to t he r adi o? 67.3 58.4 57.7 59.4 N 386 390 183 184 . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 15 Table 1b. Sample Charac teristics, by Su rvey Wave and Sex ( with rake d weights ), Males Co mp aris on /N orth Treatment/L ag os Baseline % End li ne % Baseline % End li ne % Age a t l ast birt hd ay 13-17 32.5 8.9 46.6 27.0 18-21 36.6 42.7 25.9 37.1 22+ 30.9 48.5 27.5 35.9 Leve l of Ed uca tion None 0.4 0.0 3.4 0.0 Pr ima ry 20.7 5.3 3.1 0.0 Secon da ry 58.9 81.8 75.4 78.8 Hig her 20.0 13.0 18.1 21.2 Marit al s ta tus Curre ntly m ar ried/l ivi ng wit h someo ne 0.8 4.2 4.8 1.7 Not ma rr ied 99.2 95.8 95.2 98.3 Age a t firs t sex Never Sex 86.6 73.2 63.9 47.1 Don’t K now /Refused 1.2 6.7 1.5 3 You nger t ha n 15 1.7 0.5 4.4 3.4 15-17 3.8 2.2 10.6 18.5 18-21 6.1 13.4 17.9 23.6 21+ 0.5 4.0 1.8 4.4 N u m b e r o f ch i l dr e n e v e r bor n 0 99.4 97.5 94.5 93.2 1 0.0 1.0 2.7 1.9 2 0.0 0.9 0.7 1.6 3+ 0.6 0.7 2.1 3.4 Own a ph o ne? 73.6 83.1 70.0 88.1 Ever watc h TV? 89.6 94.5 99.5 96.4 Ever l isten to t he r adi o? 67.0 62.6 71.1 80.3 N 386 390 183 184 . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 16 Se xua l Viol enc e From baseline to en dline, there were significant chang es in exp eriences of sex ual harassment, attit ud es to wards victi ms of sexual viol ence, an d the likelihood of discussing sexual violence with family me mbers for bo th f emales an d males (Tables 2a an d 2b) . Reporting o f e xperiences o f sex ual harassment increase d across all categories f or both f emales an d males, pot entially refle cting a growing a wareness of the issue . Repor ting of un welcom e at te mpts to draw par ticipants into a discussion of sexual mat ters increased from 16.1% to 33.0% for fe males an d 16.3% to 39.7% for males. Unwan ted a tte mp ts to establish a sexual rela tionship also increased fro m 13.7% to 29.4% for f emales an d 17.2% to 43.7% for males. Tou ching in a way that ma de participants feel unco mfortable rose fr om 14.0% to 37.5% for fe males an d 22.7% to 53.7% for males. Attit ud es to wards victi ms of sexual viol ence gen erally improved from baseline to en dline. The belief that a wo man is at least som ewha t resp onsible for being raped while drunk de creased from 80.5% to 60.3% for fe males an d 82.3% to 64.6% for males. The vie w that wo men are as king to be rape d when wearing revealing clothing dropp ed from 86.3% to 64.7% for fe males and decrease d from 86.7% to 81.4% for males. Even with t hese shifts, there remains a significant g ender disparity in some beliefs. For exa mple, the b elief th at wo men wearing revealing clo thing ar e asking to be raped remains high among males. The likelihood of discussing sexual viol ence with family me mbers increased dr amatically. Reporting of conversations wi th fa mily abou t sex ual violence rose from 31.2% to 70.0% for fe males and 9.0% to 38.5% for males. Mo thers were th e most co m mon fa mily memb ers to enga ge in th ese discussions, with the percen tage increas ing from 73.7% to 82.9% for fe males and 42.1% to 56.4% for males. Discussions with fath ers also increased, from 23.6% to 21.3% for fe males and 25.1% to 49.2% for males. These resul ts coincide with a key MTV Sh uga Naija the me : en gaging parents, family memb ers, and societ y at large in open c onversations about sex ual violence ma y contribu te to improved a wareness and und erstanding and ultima tely h elp red uce the occurre nce of sexual harassment and assault. . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 17 Table 2a. Exp eriences of sexual harassment /violence an d a tti tu des to wards se xual violence, Fe males Comparison/North Treatm ent/Lagos Baselin e Endline Baselin e Endline Experie nces Unwelc ome attempts to d raw y ou i nto a discussion of sexu al matters? 12.9 22.2 31.4 54.3 Unw anted attempts to esta blis h a se xual relati o nshi p wit h you despite your eff o rts to disc ourage it? 11.2 17.8 30.9 52.2 Gestured/used body la nguage of a sexual n atur e that o ffended you? 11.9 20.6 35.6 53.3 Some one t ouched y ou i n a w ay th at made y ou f eel uncomf ort ab le? 9.6 21.0 38.8 69.2 Some one imp lied better treatme nt if y ou eng ag ed in sexu al beh avi or wit h them? 2.0 10.3 9.1 25.7 Made y ou feel th reate ned with retal iati o n fo r n ot eng agi ng i n sexual be hav io r? 3.6 7.4 12.5 19.0 At ti tudes to wards Sex ual Vio lence If a w oma n is ra ped w hile s he is dru nk, she is a t least somewh at respo nsi ble. 80.3 71.7 79.3 38.2 Whe n w omen wear c lot hes that e xp ose thei r b odies, they a re asking t o be ra ped. 87.9 77.2 89.1 40.0 If a w oma n i niti ates kissing, she s h ould not be surprised if a guy assumes sex. 92.5 86.3 88.2 47.1 If a w oma n accepts mo ney, gifts, o r fav ors f ro m a ma n w h o is not her husb an d, she sh ould k n ow he e xpects s ex. 81.4 60.7 81.1 51.1 If a guy is d runk, he mig ht r a pe some one u ni ntentio na lly. 78.2 70.9 83.6 65.6 If a gi rl doesn’t p hysically f ight back, y ou ca n’t really say it was ra pe. 69.2 62.4 66.1 32.5 Often, w omen w h o say they we re ra ped agreed to h ave sex & then regretted it. 71.8 53.7 51.1 32.1 N 386 390 183 184 Commun ica tion Have y our pa rents o r f amily mem bers ever talk ed with y ou ab out sexual vi ole nce? 32.5 75.5 36.3 58.7 Wh ich f ami ly membe r? Mot her 75.6 89.6 79.1 65.8 Fat her 28.7 25.7 27.5 10.9 Sister 30.2 33.0 25.4 33.1 Brothe r 13.4 6.1 6.2 5.4 Gra ndm othe r 7.3 9.3 6.5 11.1 Aunt 23.1 21.6 10.6 19.0 N 121 292 68 109 . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 18 Table 2b. Exp eriences of sexual harassment /violence an d a tti tu des to wards se xual violence, Males Comparison/North Treatm ent/Lagos Baselin e Endline Baselin e Endline Experie nces Unwelc ome attempts to d raw y ou i nto a discussion of sexu al matters? 17.6 33.8 21.5 53.8 Unw anted attempts to esta blis h a se xual relati o nshi p wit h you despite your eff o rts to disc ourage it? 20.3 40.2 20.0 51.2 Gestured/used body la nguage of a sexual n atur e that o ffended you? 26.3 35.8 21.6 47.3 Some one t ouched y ou i n a w ay th at made y ou f eel uncomf ort ab le? 23.8 48.8 30.6 65.1 Some one imp lied better treatme nt if y ou eng ag ed in sexu al beh avi or wit h them? 4.6 19.7 10.4 20.3 Made y ou feel th reate ned with retal iati o n fo r n ot eng agi ng i n sexual be hav io r? 7.7 22.9 10.2 14.9 At ti tudes to wards Sex ual Vio lence If a w oma n is ra ped w hile s he is dru nk, she is a t least somewh at respo nsi ble. 83.4 68.4 85.0 56.8 Whe n w omen wear c lot hes that e xp ose thei r b odies, they a re asking t o be ra ped. 89.0 89.8 91.3 61.0 If a w oma n i niti ates kissing, she s h ould not be surprised if a guy assumes sex. 92.4 78.8 95.5 80.9 If a w oma n accepts mo ney, gifts, o r fav ors f ro m a ma n w h o is not her husb an d, she sh ould k n ow he e xpects s ex. 83.0 67.0 87.1 68.2 If a guy is d runk, he mig ht r a pe some one u ni ntentio na lly. 70.6 63.5 67.9 71.0 If a gi rl doesn’t p hysically f ight back, y ou ca n’t really say it was ra pe. 78.6 66.3 75.6 52.9 Often, w omen w h o say they we re ra ped agreed to h ave sex & then regretted it. 60.7 62.2 57.9 60.6 N 222 222 95 95 Commun ica tion Have y our pa rents o r f amily mem bers ever talk ed with y ou ab out sexual vi ole nce? 8.0 33.5 20.8 50.2 Wh ich f ami ly membe r? Mot her 29.6 59.6 59.9 54.0 Fat her 7.4 53.2 66.7 43.0 Sister 31.3 20.8 14.1 9.7 Brothe r 48.7 44.5 26.4 19.9 Gra ndm othe r 0.0 0.2 1.3 0.0 Aunt 0.0 12.4 1.3 5.1 N 17 81 20 49 . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 19 Diff erenc e-in-Dif fer enc es Mod el s Overall, the multivariate DI D resul ts provide support for the e ff ects o f MTV S huga Naija on attit ud es to wards se xual violence an d disclosure of e xperiences of sex ual harass ment and sexual violence, bu t lesser support for increase d co mm unication with fa mily me mbers about se xual violence. The program ha d drama tic and d emon strable eff ects on shifting harmf ul at titu des abou t sex ual violence and victi ms, par ticularly amon g fe males, with no table decreases in vic tim-blaming at titu des in instances of rape un der various scen arios, such as when a wo man is drunk [ ATT=-32.0pp, 95% CI: -43.4pp, -20.7pp], initiates kissing [ATT=-29.3pp, 95% CI : -38.9pp, -19.6pp], or do es not figh t back [ATT=-28.3pp, 95% CI : -40.2pp, -16.4pp], and a tti tu des justi fying rape s uch as “ when wom en where clo thes tha t expose their bo dies, they are asking to be raped” [ATT=-36 .6pp, 95% CI: - 46.8pp, -26.3pp] or “if a guy is drunk, h e can rape so meone uninten tionally” [ATT=-12.0pp, 95% CI: -23.0pp, -1.1pp]. The ef fec ts for ma les were less pronounce d b ut still statist ically significant for justification for rape wh en a woman is wearing revealing clothing [ATT =-31.8 pp, 95% C I: -45.6pp, - 18.1pp] and for characteriza tion of rape as requiring that a girl fight back [AT T=-19.3pp, 95% CI : - 35.6pp, -3.0pp]. These latt er results ma y indicat e that male a tti tu des are hard er to shif t or that diff erent typ es of messages, sourc es of messages, or o ther interven tions may b e more ef fec tive wi th males. In Lagos, bo th males and fe males reported a significant increase in disclosure t o interviewers of experiences o f sex ual harassment, such as unwelco me a tt emp ts to discuss sex a nd of fensive gesturing, rela tive to you th in the comp arison states of Ka duna an d Kano. For females, significant effe cts were observe d for “un wante d a t temp ts at a sex ual relationship” [ATT= 12.3pp, 95% CI : 1.1pp, 23.6pp], an d “ touche d you in an uncom fortable way” [ATT =14.8pp, 95 % C I: 3.2pp , 26.5pp]. For males, statistically significant e ffec t s were observe d for “un wante d a tt emp t s at a sex ual relationship” [ATT=18.5pp, 95% C I: 0 . 5pp, 36.4pp] and “ gestur ed in an of fens ive way” [ATT=24.7pp, 95% C I: 8.1pp, 41.4pp]. These results sug gest i mproved a ware ness and recogni tion of sex ual harassment du e to th e program, lea ding to increase d reporting. A key objec tive of MTV Sh uga Naija w as to remove th e stig ma aroun d victi ms of se xual violence and to promo te op en dialogue so that a ttit ud es coul d shif t, nor ms surrounding appropriate sexual contac t coul d be improved , an d victi ms woul d be more willing to se ek services post-exp erience. While dialogue with family me mbers ab out se xual violence in dee d improve d in the trea tmen t areas, . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 20 the DI D mod els foun d that improvem ents in co mm unication in the comparison areas actually outpace d those in the treat men t areas a mong females [ATT =-19.9pp, 95% CI : -32.1pp, -7.8pp]. Interpre tation o f this finding is challen ging but may lie in th e phrasing of the question (“Have your parents or fa mily me mbers ever talked with yo u abou t se xual violence?) , which provides little information on who initia ted th e conve rsation or what the cont ent of the conversation was. It is possible that such conversations might even have perpe tuat ed or articula ted so me o f the sam e victim-blaming s tereot ypes and at titu de s as admonishmen ts an d warnings towa rd you th. . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 21 Table 3a. Do ubly-Robus t Inverse Prob ability Weighte d (Balanced Panel ) R egr ession Models, Females Outcome ATT [95% con f. interva l] P>t Experie nces Unwelc ome attempts to d iscuss sex 9.2 -2.7 21.1 0.129 Unw anted attempts at relati o nshi p 12.3 1.1 23.6 0.032 Gestured in offensive w ay 2.6 -9.5 14.6 0.677 Touched y ou in a n unc omf orta ble w ay 14.8 3.2 26.5 0.013 Impl ied better treatme nt in exch a nge fo r sex 4.1 -5.2 13.5 0.386 Threate ned wit h reta liati o n fo r no se x 1.5 -7.7 10.6 0.756 At ti tudes Drunk w oma n is resp onsi ble if r aped -32.0 -43.4 -20.7 0.000 Wo men in skim py cl othes are aski ng f o r r ape -36.6 -46.8 -26.3 0.000 If a w oma n i niti ates kissing, guy ca n assume se x -29.3 -38.9 -19.6 0.000 Acceptin g mo ney mea ns sex -5.9 -17.0 5.3 0.301 Drunk guys ca n r ape -12.0 -23.0 -1.1 0.032 It's not ra pe if a gi rl d oesn 't fig ht b ack -28.3 -40.2 -16.4 0.000 Often ra ped w omen agreed t o sex -6.2 -19.0 6.7 0.346 Commun ica tion Sp oke with fami ly a bout sexu al vi ole nce -19.9 -32.1 -7.8 0.001 N Table 3a. Do ubly-Robus t Inverse Prob ability Weighte d (Balanced Panel ) R egr ession Models, Males Outcome ATT [95% con f. interva l] P>t Experie nces Unwelc ome attempts to d iscuss sex 18.4 -2.1 38.9 0.078 Unw anted attempts at relati o nshi p 18.5 0.5 36.4 0.044 Gestured in offensive w ay 24.7 8.1 41.4 0.004 Touched y ou in a n unc omf orta ble w ay 13.3 -4.2 30.7 0.137 Impl ied better treatme nt in exch a nge fo r sex 5.5 -7.3 18.3 0.402 Threate ned wit h reta liati o n fo r no se x -10.9 -26.2 4.3 0.160 At ti tudes Drunk w oma n is resp onsi ble if r aped -25.3 -42.1 -8.6 0.003 Wo men in skim py cl othes are aski ng f o r r ape -31.8 -45.6 -18.1 0.000 If a w oma n i niti ates kissing, guy ca n assume se x -1.3 -12.6 10.1 0.827 Acceptin g mo ney mea ns sex 3.0 -12.9 19.0 0.708 Drunk guys ca n r ape -2.7 -21.7 16.4 0.785 It's not ra pe if a gi rl d oesn 't fig ht b ack -19.3 -35.6 -3.0 0.020 Often ra ped w omen agreed t o sex 4.4 -13.9 22.8 0.637 Commun ica tion Sp oke with fami ly a bout sexu al vi ole nce 2.2 -15.3 19.7 0.806 N . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 22 Discussio n The comp elling and drama tic storylines represente d in the MTV Shu ga Naija se rial drama, co upled with n umero us opportunities for multi- way dialogue through social an d oth er me dia, are target ed to youth in Nigeria, where wo men freq ue ntly have limit ed au tonomy and exp eriences of se xual harassment and sexual violence are co mmonplace . MT V Sh uga Naija ai ms to shift norms that are accepting o f such violence, to foster atti t u de s t h at are suppor tive of vic tims, and to e mpow er you th to speak up an d se ek assistance for inci dents o f violence . MT V Sh uga Naija wa s implemen te d fro m 2018-20 20, which was an ex trem ely op portune time given tha t the Me Too m ovement in Nig eria was also get ting considerable a t tention in the press. Th e results presen te d here contribu te to the scant eviden ce base surroun ding the use of enter tainmen t e du cation to improv e sexual violence attit ud es, norms , and disclosure. This evaluation is unique in tha t it de m onstrates a link bet ween m ultichannel s ocial and behavior change program ming and at tit udes to w ards sexual violence. To dat e, there has been limite d evidenc e of such a link in conte xts similar to th e areas under st u dy here (Bo tt e t al., 2005). Th e WHO an d PAHO have highlighte d the importan t role that mass m edia en ter tainmen t e du cation strat egies can have in influencing g end er norms, com munity responses, and individual a tti tu des towards in tima te partner violence (Worl d Heal th Organi zation & Pan American Health Organiz ation, 2012). The findings in dicating that MT V Shu g a Naija has been succ essful in shifting n orms and a tti tu des are indee d enco uraging, an d y et there appe ars to still be q uite a long way to go . Ev en among those exposed to the campaign , th ere was still a high prevalence of victim blaming a t titu des. Nearly eight of 10 you th e xposed to MTV S hu ga Na ija agreed that , if a woman initia tes kissing, a g uy can assume that se x will occur, and more than half of th e yo uth respon den ts believed tha t it is not rape if a woman does no t fight ba ck. More prog ress is neede d in ord er for vic tims of se xual violence to f eel confiden t o f support and un derstan ding, and ideally for victi mization to b eco me less co mmon . The approaches taken in Shuga’s progr amma tic design and m essage design are important . The overall programmatic design was base d in enter tainmen t-e duca tion chang e ( Singhal, 1999, 2004; Singhal & Rogers, 2001; Usdin, 2004; Y ue Z , 2019). Th e observed beneficial e f fec ts of Sh uga in this stud y can b e th eorize d to f unction thro ugh th e mechanisms that p ermit ent ert ainment-e du cation to grab the a tt ention o f an d cap tivate au di ence memb ers, while also infusing in th em nor ms and . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 23 information per tinent to gend er-based violence. Moreover, the m essage d esign was based in th e Theory of Planned Behavior (Ajzen , 19 91). The television and ra dio dra mas pu rposefully includ ed conten t specific to b ehavioral beliefs, n ormative beliefs, and control b eliefs, w hich are conceive d to be core processes, respe ctively, in the d evelopment of a t titu des, subjective nor ms, an d perceive d behavioral control. Whereas this st udy did no t i mplemen t measures o f each of these s teps in the contex t o f g ender-base d violence , th e e valuation provides evidenc e gen erally in line with the expec te d ou tco mes of this theory an d u nderscores the importan ce o f using pro gramma tic and message-d esign theory in the developm ent an d evalua tion of h ealth ca mpaigns. This research points to areas where a dd itional investigation is warrante d, inclu ding research that focuses specifically on un derstan ding t he fac tors tha t con tribute to the persistence o f victim -blaming attit ud es and misconcep tions about co nsent. Furth er, there re mains a pauci ty of stu dies that se ek to understan d the challenges of repor ting incidences o f sex ual violence to securit y out fits inclu ding police. We sug gest research tha t e xplores the role o f cult ural, societal, and fa m ilial influences in shaping these at tit udes and i den tifying ways to ad dress these fac tors more ef fe ctively thro ugh targete d in terventions. In a ddi tion, fu tu re stu dies of enter tainmen t-e du cation c ould investigat e the impact o f dif ferent m essage framing , c ontent , an d delivery for mats in changin g at titu des towar ds victim-blaming an d consen t an d i denti f ying best prac tices for developing more persuasive campaigns. The Theory o f Planne d B e havior provides one widely use d frame work for message design, b ut fu ture ent ertainmen t-e du cat ion research in this area coul d e mploy others, as well. Addi tionally, stu dies coul d evalua te the effe ctiveness of integra ting oth er strat e gies, such as comm unity-base d initia tives, e duca tional programs, and policy changes, to reinforce the messages conveyed throu gh en tertainm ent- ed uca tion programs like MTV S hu ga Naija . A longer-term perspective is also warranted . The MT V Shu ga Naija ca mpaign achieved som e impressive results in the fe w years that it aired. I mportan t e mpirical issues rem ain, including assessing the long-term e ffec ts o f such campaigns on behavior change an d exp loring whether the impact o f e xposure to these ca mpaigns persists over time or diminishes wi thou t con tinue d reinforcemen t. Li mitat ions and F uture Directi ons It is impor tant to no te tha t this evaluation face d several challenges. In a d dition to selec tive at trition and th e swi tch from face- to- face in terviews to phone in terviews, there may also have been issues . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 24 related to dif ferences in the cul tures an d values be tw een the treat men t an d co mparison areas. The populations of the nor thern sta tes of K aduna an d Kano are pre do minantly Mu slim, reflec te d in much lesser autono my for f emales, pa t t erns of early marriage and early child-b earing, and lesser emphasis on schooling and work ou tsid e the hom e for fe males, in strong contr ast to the more western- focuse d, cosmopolitan pop ulat ions of the south , inclu ding Lagos. Th e original study design articulate d the provision of id entical co ntent across all areas, albeit wi th stag ge red imple men tation betw een the sou th an d the nor th. How ever, local governmen ts in the nor th ob jecte d to th e e xplicit conten t relate d to se x an d sex ual violence in MT V Shu ga Naija program ming a nd prohibite d the airing of any material ex cept tha t focus ed solely on reprodu ctive heal th within the con te xt o f marriage. These dif ferences in the pop ulation charact eristics for th e treat ment and co mparison groups may have implications for on e of th e key assumptions o f this analysis – that treat men t an d control groups wo uld have follow ed parallel trends in sex ual violence at titu des in the absence of the program. While the doubly-robus t DI D esti mation s trategy sough t to am eliorate the possible influence o f these dif ferences, it was simply not f easible to collec t more than one pre-interven tion round o f data that wo uld have permi tte d a more robus t assessments o f trends prior to the implemen tation o f Shu ga Naija ac tivitie s. Hence an e xplicit test o f the parallel t rends assump tion was simply not possible. Second , relate d to a ttrition , th e sample sizes of respond ents, particularly of sex ually active males, were quite small at endline . The stu dy was originally powered un der the assu mption o f low er rates o f attrition, and as a result, th e observed d ifferences re quire d for s tatistical signifi cance for indicators depen den t on se xual ac tivity ten ded to be qui te large. Hen ce, while Sh uga may have shifte d indica tors in the desired direc tion, o ur final sampl e size may have b een too small in som e cases to observe statistically significant shif ts (i. e. , at a p- value of belo w .05 ). Finally, starting in March 2020 – seven months be fore the endline survey – Lagos was placed un der stringent COVID -19 lockdowns tha t w ere not i mplem ente d in the northern s t ates of Ka duna an d Kano. This po tentially a ffe cte d the abili ty o f yo uth in Lagos to in teract wi th ot hers. Hence , chan ges in the rat es of sexual harassmen t an d se xual violence may partially re flect chan ges in mobility circumstanc es.

This evaluation provides evidenc e of M TV Shu ga Naija ’s e ff ects on at titu des an d behaviors relate d to sex ual violence, s tigma tization o f vic tims of sexual violence , an d open dialo gue abou t a d dressing . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 25 sexual violence. During the s tu dy p eriod, more rapid at tit udinal shifts to wards sexual violence an d victims of se xual violence were apparent amon g both male and fe male yo uth in the trea tmen t gro up relative to the comparison group. Thes e results highlight that MT V Shu ga Naij a was able to achieve important ef fec ts in a relatively short p eriod of tim e. It is well es tablished that it takes time for campaign viewers to progress fro m ca mpaign exposure to the d evelopmen t o f revised perc eptions and, subse quen tly, from th e at tit udinal shifts to the develop men t of revised be havioral patterns. By opening and nor malizing dialogu e abou t sex ual violence amon g you th, MT V S huga Naija has set the ground work for subseq uent behavior c hanges that can help re d uce the proble m of sexual violence . . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 26 Declaration of Conflicting In terests The author (s) declare d no po ten tial con flicts of interest wi th respec t to th e rese arch, au thorship, and/or p ublication of this article . Funding This work was f unde d b y the Bill and Melinda Gates Foun dation (B MG F) th rough Grant No . OPP116231. ORCID iDs Paul H utchinson h ttps: / /orcid.org /00 00-0002-5 435-77 99 Christopher E. Beau doin h ttps: / /orcid. org/0000-0003-0 649-33 47 Dominiqu e M eekers ht tps: //orci d.org / 0000-00 02-728 1-3217 Elizabeth O moluabi. h ttps :/ /orcid .org /0000-0003-17 54-538 4 Akanni Akinyemi https: / /orcid.org /00 00-0001-8 652-93 59 . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 27

Ahrens C.E ., C. R ., T ernier-Thames N. K. , Wasco S .M. , Se fl T. . (2007). De cidi ng who m to t ell: Expec tations an d ou tco mes of rape sur vivors’ first disclosures. P sycho logy of Wo men Quar terly , , 31 , 38-49. Ajzen, I . (1991). The th eory of planned behavior. Organizatio nal B ehavior a nd Hu man Decisio n Proc es se s , 50 (2), 179-211. ht tps: // doi.org / . doi:10 .1016/0749-597 8(91)90020-T. Antai, D . E ., & An tai, J. B. (2008). A tti t udes o f wom en to ward intima te par tne r violence: a st ud y of rural women in Nigeria. Rura l R em ot e H ealth , 8 (3), 996. https: / /w ww .ncbi.nlm .nih.gov /pubm e d/18842071 Babalola, S., Kuse miju, B ., Calhoun , L. , Corroon, M. , & Ajao, B. (2015). Fac to rs associated wi th contracep tive idea tion among urban m en in Nigeria. I nt J Gyna ecol Obst et , 130 S uppl 3 , E42- 46. https: / / doi.org/10.1016/j.ijgo.2015.05.006 Banerjee, A. , La Ferrara, E. , an d Orozc o-Olvera, V.H. ,. (2019). Th e En tertaini ng Way to Behavioral Change: Figh ting H IV wi th MTV . . In W. Bank. (E d. ), Po licy R es earch Worki ng Paper . Washington, D C. Bergmann, M. (2011). Ipf w eight : Stata mo dule t o creat e adj us tm en t w eigh ts f or s urvey s . Retrieve d Nove mber 18, 2020 from ht tps: // econpapers.repe c.org/sof t ware/bocbo code /s457353.htm . Bott , S. , Morrison, A., & World Bank. ( 2005). Prev e ntin g and re spondi ng to g end er-based vi ol enc e in middle and l ow-inco me cou ntrie s a g lobal revie w a nd analysis . World Bank,. http :/ /econb eta. worldbank.org /e xtern al/de fault / main?pagePK =64165259&p iPK =6416542 1&menu P K=64166093&theSiteP K =4 69372&entityI D=000012009_2005 060 116110 1 Callaway, B . a . S. A., P edro H. C. ,. (2021). Diff erence-in- Diff erences with mul t iple time periods. Jour nal of Eco nom etrics , 225, 200-230. Chamberlain, G. (1980). Analysis of Co variance with Qualita tive Da ta. In N. B . o. E. Research (E d. ). Cambrid ge, MA. Ellsberg, M. , Jansen, H. A ., Heise, L ., Watts, C . H ., Garcia-Moreno , C., Heal t h, W. H. O. M.- c. S . o . W. s., & Do mestic Violence against W omen S tu dy, T. (2008). Inti mat e partne r violence and women 's physical and men tal health in the WHO m ulti-coun try s tu dy on wo m en's health and domes tic violence: an observational stud y. Lanc et , 371(9619), 1165-1172. https: / /doi.org /10.1016/S0140-6736(0 8)60522-X Fleming K. , K. L. (2016). “Sh e keeps hi s secrets”: A gen dere d analysis of th e i mpact of sha me on the non- disclosure of sexual violence in one low-inco me South African co m m unity. African Saf ety Prom otio n: A Jour nal of I njury a nd V iole nce Pr eve ntio n , 11 (2), 107-124. Fontes, L. A ., & Plum mer, C . (2010). C ultural issues in disclosures of child se x ual abuse. J C hild Se x Abus , 19(5) , 491-518. ht tps: // doi.org /10.1080/10538 712.201 0.51252 0 Gage, A. J. , & Tho mas, N. J. (2017). W omen's Work, Gen der Roles , and In tim ate Partner Violence in Nigeria. Arch S e x B ehav , 46(7) , 1923-1938. https: // doi.org /10.1007/s10508- 017-102 3-4 Gilbert, L . K. , Annor, F. B ., & Kress, H. (2022). Associations Bet ween En do rsement o f Ineq uitable Gen der Nor ms and In tima te Partner V iolence and S ex ual Risk Behaviors Among You th in Nigeria: Violence Against Children Sur vey, 2014. J I nterp ers Vio le nce , 37(11-12), NP8507- NP8533. ht tps: // doi.org /10.1177/088626052 0978 196 Ikute yijo, O. O ., Akinyemi, A . I ., & Me rten, S . (2022). E xposure to job-relate d violence among young female sex workers in urban slu ms of South west Nig eria. BM C Pub lic Hea lth , 22(1) , 1021. https: // doi.org /10.1186/s12889- 022-134 40-1 Ikute yijo, O. O ., Akinyemi, A .I ., Mert e n, S. , F et ters, M. D. ,. (2022). Removing barriers to utilisation of support services for abuse d f emale adolescents in Nigeria slums. Eur opea n J o urnal of P ublic Hea lth , 32. h ttps: / /doi.org / doi.org/10. 1093/eurpub/ckac131.455 . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 28 Jina, R., & Tho mas, L . S. (2013). Heal t h consequen ces of sexual violence against wo men. B es t Pract Res C lin Ob ste t Gyna ecol , 27(1 ), 15-26. h tt ps:// doi.org /10.1016/j.bpobgyn.2012. 08.012 Joffe, M . M ., Ten Have , T . R ., Fel dman , H. I ., & Kim mel, S. E. (2004). Mod el selection, con foun der control, an d marginal struct ural mo dels: review an d n ew applica tions. Th e A me rican Stati stician , 58(4), 272-279. Kolenikov, S. (2014). Calibrating surve y da ta using i terative proportional fi ttin g (raking). S tata Jour nal , 14, 22-59. Kolenikov, S. (2019). Up dat esd to the i pfraking ecosyste m. Sta ta J our nal , 19, 14 3-184. Luncefor d, J . K., & Davidian, M. (2004). S tratifica tion an d w eighting via the pr opensity score in estimation o f ca usal treat men t e ff ects : a comparative s tu dy. S tatis tics in m edicine , 23(19), 2937- 2960. Meekers, D. , Pallin, S . C. , & Hu tchinso n, P . (2013). Prevalenc e and correlates of physical, psychological, and se xual inti mate part ner violence in Bolivia. Glob Pub lic H ealt h , 8 (5 ), 588- 606. https: / /doi.org /10.1080/1744169 2.2013.7760 93 National Pop ulation Comission ( NP C) Nigeria, U. N. , an d the U. S. Cen ters for Disease Con trol and Prevention, . (2016). Vi ol enc e A gains t C hildren in Nig eria: Findin g s fr om a Na tiona l S urvey, 2014 . National Pop ulation Com mission (N P C) [Nigeria], & I C F. (2019). Nig eria D e mographic a nd H eal th Surv ey 2018 . N. a . IC F. Nguy en, K. H ., Kress, H ., Atu chukw u, V., Ono tu, D ., S wa minathan, M. , Ogba nufe, O. , Msun gama, W., & Su mner, S. A. (2021). Disclosure of Se xual Violence Among Girls and Young Wo men Aged 13 to 24 Years: Results From the Violence Against Children Surveys in Nigeria and Malawi. J I nt erper s Vi ol enc e , 36(3-4 ), N P2188-220 4NP. https: / /doi.org /10.1177/08862605 187 57225 Pulerwit z, J., & Barker, G . (2007). M easuring atti tu des toward gen der nor ms a mong yo ung men in Brazil: Develop ment and psychom etric evalua- tion o f the GE M scale. Me n a n d Mascu linitie s , 10 (3), 322-338. ht tps: // doi.org /ht tps:/ /doi . org/10.1177/1097184X06298 778 Sant’Anna, P. H. C ., Zhao, Jun, . (2020). Do ubly robust di fferen ce-in- differen c es estima tors. J our nal of Ec on om etrics , 219, 101-122. Sardinha, L., Mahe u- Giroux, M. , S tockl , H. , Meyer , S. R ., & Garcia-Moreno, C . (2022). Global, regional, and na tional prevalence esti m ates of physical or sexual, or both , inti mate partner violence against wo men in 2018. Lanc et, 399(10327), 803-813. https: / /doi.org /10.1016/S0140-6736(2 1)02664-7 Singhal, A., & Ro gers E. M .,. (1999). E nter tainm en t-educati on: A c om mu nication str ategy for social chan ge . Routle dge . Singhal, A., Co dy, M . J., Rog ers, E. M ., & Sabido , M . (E ds.) . (2004). E nt ertainm e nt-ed ucation and social chang e. Rou tledg e. Singhal, A., & Ro gers, E . M. (2001). Th e Ent ertainmen t-E du cation S trate gy in Campaigns. In R . E . Rice & C . Atkins (Eds .) , P ublic Com mu ni cation Campaign s (3rd edi tion e d. ). Sag e Publications. Sumner S . A. , M. , A.A. , Saul J., Mo tsa-Nzu za N. , Kw esigabo G. , B uluma R. , Kilbane T. (2015). Prevalence of sexual violence against c hildren and use of social services-seven countries, 2007-20 13. Morbidity and Mor tality W ee kl y Repor t , 64(21) , 565-569. Ullman, S. E., Star zynski, L. L ., Long , S . M ., Mason, G. E. , & Long, L . M . (2008). E xploring the relationships of wom en's sex ual assault disclosure, social reactions, an d proble m drinking. J Int erper s Viol enc e , 23 (9), 1235-1257. http s:// doi.org /10.1177/08862605 0831 42 98 Umana, J. E. , Fa wole, O. I. , & Adeo ye, I. A. (2014). Prevalence and correlates of intima te partner violence towar ds fe male s tu dents o f th e University of Ibadan , Nig eria. BM C Wome ns H eal th , 14 , 131. https :/ / doi.org/10.1186/1472-6874-14-13 1 . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint 29 Usdin, S. , Singhal, A. , Shong we , T. , Go ldstein, S ., & Shabalala, A ., . (2004). No short cu ts in entertain ment -ed uca tion: Designing So ul Cit y step -by-st ep. In M . J. C . A. Sing hal, E. M. Rogers & M. Sabido, (E d .), E nter tainm e nt-ed ucation and social chan ge : Hist ory, r es ear ch, and practice (pp. 153-175). La wrence Erlbaum Asso ciates. Wooldridge, J. M. (2002). Econ om etric an alysis of cr os s sec tion and pa ne l data . M IT Press. World Health Organization. (2021). Vi ole nce again st w ome n pr eval enc e es timate s, 2018. . World Health Organization, & Pan A m erican Health Organiza tion. (2012). U n dersta nding and addressi ng vi ol enc e agai nst w om en : I ntima te partner vi ol enc e . https: / /apps.who .int /iris/bitstream /ha ndle/10665/77432/WHO_R HR _12.36_eng.p df Yue Z , W . H ., Singhal A., . (2019). Using television drama as ent ertainm ent-e duca tion to tackle dom estic violence in China. Jo urna l o f D eve lopm en t C omm u nication , 30, 30-44. . CC-BY-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 August 24, 2023. ; https://doi.org/10.1101/2023.08.23.23293994doi: medRxiv preprint