Are interventions with batterers effective? A meta-analytical review

Psychosocial In e en ion Psychosocial In e en ion (2020) 29(3) 153-164 ISSN:1132-0559/© 2020 Colegio O icial de la Psicología de Mad id. This is an open access a icle unde he CC BY-NC-ND license (h p://c ea i ecommons.o g/licenses/by-nc-nd/4.0/). A e In e en ions wi h Ba e e s E ec i e? A Me a-analy ical Re iew Ramón A ce, Es he A ias, Me cedes No o, and F ancisca Fa iña Uni e sidad de San iago de Compos ela, Spain h ps://jou nals.copmad id.o g/pi Funding: This esea ch has been sponso ed by a g an o he Spanish Minis y o Economy, Indus y and Compe i i eness (PSI2017-87278-R). Co espondence: [email p o ec ed] (R. A ce). In e en ion p og ams o ba e e s ha e been he subjec o con- o e sy e e since hei concep ion. These in e en ions ha e been open o c i icism om bo h a es o a i e pe spec i e and a eminis pe spec i e demanding esou ces should be alloca ed o ic ims, no o ba e e s. Such c i icism, howe e , comes in o di ec con lic wi h he legal and judicial manda e o p ison ins i u ions ha a e obliged ARTICLE INFO Recei ed 5 Decembe 2019 Accep ed 13 Ap il 2020 A ailable online 26 May 2020 Keywo ds: Ba e e In e en ion assessmen Dulu h model Cogni i e-beha iou al ea men p og ams O icial eco ds Couple epo s ABSTRACT The inconsis ency in he esul s bo h in e nally and be ween o p e ious me a-analyses on ba e e in e en ion p og am e icacy, and he publica ion o new ba e e in e en ions unde sco ed he need o an up- o-da e me a-analy ical e iew. A o al o 25 p ima y s udies we e ound om li e a u e sea ch, ob aining 62 e ec sizes, and a o al sample o 20,860 in e ened ba e e s. The esul s o a global me a-analysis showed a posi i e, signi ican , and o a medium magni ude e ec size o ba e e in e en ions, bu no gene alizable. Ne e heless, he esul s exhibi ed a signi ican ly highe a e o ecidi ism measu ed in couple epo s (CRs) han in o icial eco ds (ORs). As a consequence, in e en ion e icacy measu ing in CRs was null, whils in ORs was posi i e and signi ican . As o he in e en ion model, posi i e and signi ican e ec s we e obse ed unde he Dulu h Model and cogni i e-beha iou al ea men p og ams (CBTPs), bu a highe e ec size was ob ained wi h CBTPs in compa ison o he Dulu h Model (unde his model, in e en ions may ha e nega i e e ec s, i.e., an inc ease in ecidi ism a e). In ela ion o in e en ion leng h, sho in e en ions ailed o educe ecidi ism in ORs and may ha e nega i e e ec s, while long in e en ions we e e ec i e in educing ecidi ism a e in ORs wi hou nega i e e ec s. E icacy e alua ions in sho ollow-ups we e in alid as a i icially boos ed ecidi ism educ ion a e. Limi a ions o ORs and sho ollow-ups as measu es o he in e en ion e icacy and implica ions o esul s o ba e e in e en ion a e discussed. ¿Son e ec i as las in e enciones con los mal a ado es? Una e isión me aanalí ica RESUMEN La inconsis encia in e na y en e las e isiones me aanalí icas en los esul ados sob e la e icacia de los p og amas de in e ención con mal a ado es, así como la publicación de nue os es udios, pone de mani ies o la necesidad de lle a a cabo una e isión me aanalí ica ac ualizada. Se encon ó un o al de 25 es udios p ima ios, de los que se ob u ie on 62 amaños del e ec o pa a una mues a o al de 20,860 mal a ado es in e enidos. Los esul ados del me aanálisis global mos a on un amaño del e ec o p omedio posi i o, signi ica i o y de una magni ud mode ada pa a la in e ención con mal a ado es, pe o no gene alizable. Sin emba go, los esul ados e ela on una asa de eincidencia mayo medida en los in o mes de las pa ejas (IPs) que en los egis os o iciales (ROs). Como consecuencia, la e icacia de la in e ención medida en los IPs esul ó nula, mien as que en los ROs ue posi i a y signi ica i a. En elación al modelo de in e ención, se encon a on amaños del e ec o posi i os y signi ica i os con el Modelo Dulu h y los p og amas de a amien o cogni i o-conduc uales (PTC-Cs), pe o el amaño del e ec o ob enido con los PTC-Cs e a signi ica i amen e mayo que con el Modelo Dulu h (con es e modelo las in e enciones pueden ene e ec os nega i os, es deci , un inc emen o en la asa de eincidencia). En elación a la longi ud de la in e ención, las in e enciones b e es alla on en la educción de la eincidencia en los ROs y pueden ene e ec os nega i os, en an o que las in e enciones la gas ue on e icaces en la educción de la asa de eincidencia en los ROs y no dan luga a e ec os nega i os. Las e aluaciones de la e icacia de la in e ención en pe íodos co os de seguimien o esul a on no álidas al inc emen a a i icialmen e la asa de educción de la eincidencia. Se discu en las limi aciones de la medida de la e icacia de la in e ención en los ROs y en pe íodos co os de seguimien o, así como las implicaciones pa a la in e ención con mal a ado es. Palab as cla e: Mal a ado E aluación de la in e ención Modelo Dulu h P og amas de a amien o cogni i o-conduc uales Regis os o iciales In o mes de las pa ejas Ci e his a icle as: A ce, R., A ias, E., No o, M., & Fa iña, F. (2020). A e in e en ions wi h ba e e s e ec i e? A me a-analy ical e iew. Psychosocial In e en ion, 29 (3), 153-164. h ps://doi.o g/10.5093/pi2020a11 154 R. A ce e al. / Psychosocial In e en ion (2020) 29(3) 153-164 o emba k on he esocializa ion and ehabili a ion o inma es. A u - he bone o con en ion conce ns he e icacy o in e en ions, ha is, i he in e en ion lacks e icacy, i is un ounded. Ini ially, in e en- ions we e designed o iolen physical o sexual o ende s, bu in e- cen yea s hey ha e been gene alized o less se ious o ences, mainly mino o ences and misdemeanou s wi h non-cus odial communi y sen encing. Claims ha in e en ions wi h ba e e s we e no e ec- i e (Rosen eld, 1992) ha e been e u ed by me a-analy ical e iews ha ha e ound in e en ions o be e ec i e. Ne e heless, in se e al e iews he mean e ec size was signi ican , bu modes (Babcock e al., 2004), whe eas in o he s he mean e ec size was impo an , bu no gene alizable o all in e en ions (A ias e al., 2013). No wi h- s anding, es ima es on he bene i s o ba e e in e en ions ange om 5% (Babcock e al., 2004) o 20% (A ias e al., 2013). Mo eo e , he compa ison o he mean e ec size o he e icacy o in e en ions (BESD; Rosnow & Rosen hal, 1988) in he mos ecen me a-analysis (A ias e al., 2013), = .20, wi h he esul s o he ea men o delin- quency in gene al (Redondo e al., 2002), and he esul s o sexual o ende s in pa icula , = .13 (Schmucke & Lösel, 2015), showed ba - e e in e en ions we e simila o mo e e ec i e. Mode a o s o he in e en ion we e analysed as he e ec s we e subjec o he e ogene- i y. The a iable measu ing e ec s has been inconsis en , in compa - ison o he sys ema ic posi i e and signi ican , bu no gene alizable e ec epo ed in o icial eco ds (ORs) (A ias e al., 2013; Babcock e al., 2004; Fede & Wilson, 2005; Rosen eld, 1992), he esul s o cou- ple epo s (CRs) we e inconsis en . In ac , Fede and Wilson (2005) and A ias e al. (2013) ound a null e ec in CRs, whils Rosen eld (1992) encoun e ed a no able decline in he ecidi ism a e (33% in ea ed ba e e s s. 47.3% in d opou s), and Babcock e al. (2004) he same e ec as in ORs, d = 0.18. Likewise, he disc epancies obse ed in he esul s o he ype o in e en ion. Thus, in e ms o in e en ion p og ams based on he Dulu h Model, a eminis psychoeduca ional app oach, Babcock e al. obse ed a signi ican e ec on he educ ion o ecidi ism in ORs, d = 0.25 (12.4%) and CRs, d = 0.24 (11.9%), bu no signi ican e ec was obse ed in Cogni i e-Beha iou al T ea men P og ams (CBTPs). Mo eo e , A ias e al. ound a mode a e e ec size, bu no gene alizable o all s udies (i.e., wi h ad e se-nega i e-ou - comes) in ORs bo h in he Dulu h Model, d = 0.41, and in CBTPs, d = 0.47, which was insigni ican in CRs wi h ei he ype o ea men . As o he leng h o he in e en ion, Babcock e al. (2004) de e - mined sho and long in e en ions we e as equally e ec i e in ORs, ( d = 0.16 and d = 0.20, espec i ely), and A ias e al. (2013) no iced long in e en ions had a signi ican and gene alizable e ec , d = 0.49, bu sho ones did no . In CRs, Babcock e al. ob ained small signi i- can e ec s in bo h sho and long in e en ions. In con as , A ias e al. a ained bo h sho and long in e en ions could inc ease ecidi- ism. Du ing long- e m ollow-up, Babcock e al. obse ed a signi i- can e ec in he decline o ecidi ism in ROs, d = 0.25, which A ias e al. claimed was no gene alizable, and could po en ially inc ease he ecidi ism a e by 17.7%. In sho - e m ollow-ups, he me a-analysis by Babcock e al. e ealed (unexpec edly gi en ha he de ec ed a e o ecidi ism is ob iously smalle in sho e han in longe pe iods) a smalle e ec o he in e en ion, d = 0.13, bu signi ican , han in he wo k by A ias e al., who ob ained a null e ec , d = 0.04, which could e en ually ha e an ad e se e ec in he ange o 22%. In CRs, Babcock e al. ob ained a simila pa e n o esul s: signi ican e ec s in bo h sho and long ollow-ups, and pa adoxically la ge in long ollow-ups ha p o ide a b oade measu e o ecidi ism han sho pe iods. In compa ison, A ias e al. ound a null e ec in sho ollow-ups, d = 0.03, wi h a nega i e e ec o up o 28.3%, and negligible, d = 0.12, wi h ad e se e ec s o up o 18.2% in long ollow-ups. The inconsis ency in he esul s bo h in e nally and be ween me a-analysis, and he ac ha ini ially hey measu ed physical iolence, which was ecen ly ex ended o encompass psychologi- cal iolence, as well as he p oli e a ion o new ba e e in e en- ion p og ams unde sco e he need o a me a-analy ical e iew o es ablish he ac ual s a e-o - he-a o ba e e in e en ions, and o elucida e he in e - and in a-analysis inconsis encies. Me hod S udy Sea ch The sea ch o s udies was designed o upda e he 2012 me a- analysis by A ias e al. (2013). Thus, he ollowing sea ch s a egies we e employed: a) sea ch in b oad da abases PsycIn o, ERIC, EBSCO, and Google Schola ; b) sea ch in gende iolence obse a o ies ( .g ., www.wo k-wi h-pe pe a o s.eu; www.VAWne .o g; www.minca a. umn.edu; www.cou inno a ion.o g; www.cienciaspenales.ne ; www.i esweb.o g); c) con ac ing p ominen esea che s in he ield ( ha is, esea che s who we e he co esponding au ho s o all he pape s ound, bo h he included and excluded pape s); and d) e iew o all he bibliog aphy in he e e ences lis s o all he pape s ound, and p e ious me a-analy ical e iews. The mos p oduc i e keywo ds we e: “ba e e ”, “in e en ion p og am”, “e alua ion”, “assessmen ”, “e ec i eness”, “in ima e pa ne iolence”, “pa ne - iolen men”, “ ecidi ism”, “ eo ending”, “a i ion”, “domes ic iolence”, “cou manda es ba e e in e en ion”, and “p ison in e en ion”. The ela ion o keywo ds was gene a ed by a sys em o succes- si e app oxima ions whe eby he ini ial keywo ds o p e ious me- a-analyses de e mined he inclusion o ele an keywo ds om all he pape s ound. Inclusion and Exclusion C i e ia Bea ing in mind he objec i es o he me a-analysis, he ollowing inclusion c i e ia we e applied o he pape s eligible in he sea ch: a) hey epo ed he numbe o pa icipan s; b) hey p o ided he ecidi ism a e o he sample o subjec s who comple ed he in e en ion; c) hey applied an expe imen al o quasi-expe imen al design (wi h o wi hou con ol g oup); d) hey indica ed heo e ical app oach, con en s, and du a ion o he in e en ion p og am; and e) hey s ipula ed he ollow-up pe iod o measu ing ecidi ism. S udies in ol ing a 6-mon h ollow-up pe iod o sho e we e excluded owing o he lack o alidi y in he measu es. A o al o 25 pape s we e included unde hese c i e ia, wi h 62 e ec sizes, and a sample o al o 20,860 in e ened ba e e s. The s udy sea ch lowcha is shown in Figu e 1. Da a Analysis A me a-analysis o expe imen s was pe o med by co ec ing he e ec sizes acco ding o he dis ibu ion o a i ac s (Schmid & Hun e , 2015). As he measu e o ecidi ism was no mally exp essed in pe cen ages, o when his was no he case hey could be compu ed, in his analysis he measu e o ecidi ism was de ined as he pe cen age o ba e e s who had eo ended in gende iolence (da a on ecidi ism in o he o ences we e excluded) du ing he ollow-up pe iod. Two indices ha e been p oposed o calcula ing he e ec size in p opo ions: Cohen’s h and Hedges and Olkin’s (1985) δ, leading o sligh ly highe , bu quali a i ely simila esul s o h (A ias e al., 2013). In his me a-analysis, Hedges and Olkin’s δ was employed. None heless, all o he analyses wi h h we e eplica ed and ob ained quali a i ely simila esul s. In e ms o in e p e a ion, δ and h a e in e changeable wi h Cohen’s d , and an e ec size o 0.20, 0.50, 0.80 is conside ed o be small, medium, and la ge, espec i ely. Howe e , his classi ica ion has gene a ed disc epan in e p e a ions, whe e small e ec sizes we e classi ied as negligible by ce ain au ho s bu ele an by o he s. Hence, he e ec size was ans o med in o pe cen iles (Mon ei o e al., 2018) and he magni ude was in e p e ed in e ms o pe cen age supe io i y o e all possibili ies (Vila iño e al., 2018). 155 A e In e en ions wi h Ba e e s E ec i e? A Me a-analy ical Re iew As o he e ec s o he calculus o δ, in p e ious e iews i has been calcula ed wi h con ol g oups bo h in expe imen al and quasi-expe imen al designs. The esul s showed he ype o design media ed di e ences in e ec sizes. Howe e , he ype o analysis design is no an e ec o he in e en ion. Tha is, he di e ences in esul s we e due o he es alue, no he in e en ion. Fo his eason, and in o de o homogenize he in e -s udies con as alue, o con ol he bias o con ol g oups in p ima y s udies and o include in e en ions wi hou con ol g oups, he con as alue o ecidi ism was calcula ed as eo ending wi hou in e en ion and he ecidi ism a e weigh ed by he sample e o o he con ol g oup o al (Schmid & Hun e , 2015), which was .22 in ORs and .28 in CRs. Compu ed e ec sizes om p ima y s udies (see Appendix), he ollowing s a is ics we e calcula ed: sample size weigh ed mean e ec size ( d ), s anda d de ia ion o d ( SD d ), s anda d de ia ion p edic ed o sampling e o alone ( SD p e ), s anda d de ia ion o d a e emo ing sampling e o a iance ( SD es ), mean ue e ec size (δ), s anda d de ia ion o δ ( SD δ ), pe cen age o a iance a ibu able o s a is ical a i ac s (%Va ), 95% con idence in e al o d (95% CI d ), and 80% c edibili y in e al o δ (80% CI δ ). Though esul s opened he doo o u he me a-analyses ( he a iance explained by a i ac ual e o s was below 75%, i.e., 75% ule; Schmid & Hun e , 1981), u he me a-analyses we e impossible as numbe o sizes ( k ≤ 3) o sample sizes ( N < 400) we e insu icien , o no o he mode a o s had been de ined in he p ima y s udies. When me a-analyses a e calcula ed o he s udy o mode a- o s o he g ouping o a iables acco ding o le els, he esul s do no include he compa ison be ween me a-analyses. Thus, he esul s we e insu icien , as is he case o he analysis o in e en- ion models whe e signi ican e ec s we e obse ed in ORs, bu i hey we e he same o highe in one condi ion o ano he was no epo ed. The same occu ed wi h he du a ion o he in e - en ion, and he ollow-up pe iod. To o e come his con ingency, he solu ion p oposed by Amado e al. (2015) was adop ed, whe e he s a is ic q s was calcula ed o compa e wo e ec sizes by con- e ing , and hen compa ing he s. Fu he mo e, he esul s o he me a-analysis we e also qui e limi ed in disce ning he impli- ca ions o p ac ice, in his case, he in e en ion. Rega ding he implica ions o he me a-analysis on p o essional p ac ice, Fa iña e al. (2017) ecommend, acco ding o he speci ic objec i es a hand, sensi i i y es ima ion (s a is ical U ), e ec quan i ying (in p esen me a-analy ical analysis, he e icacy o he in e en ion, BESD), and he p obabili y o supe io i y. C i e ion Reliabili y As CRs we e ga he ed wi h di e en measu emen ins umen s, he co ec ion o he c i e ion un eliabili y was compu ed wi h Mosie ’s (1943) composi e eliabili y coe icien , = .87. Coding Fo he analysis o mode a o s, he ollowing ecidi ism a iables we e encoded (ORs, N = 19,429 and k = 46, and CRs, N = 1,351 and k = 16); ollow-up ime (≤ 12 mon hs, N = 3,509 and k = 21, and > 12 Reco ds iden i ied h ough da abase sea ching ( n = 1,910) Addi ional eco ds iden i ied h ough o he sou ces ( n = 23) Reco ds a e duplica es emo ed ( n = 1,547) Reco ds sc eened ( n = 374) Full- ex a icles assessed o eligibili y ( n = 76) S udies included in quan i a i e syn hesis (me a-analysis) ( n = 25) Full- ex a icles assessed wi h easons ( n = 51) Reco ds excluded ( n = 1,173) Iden i ica ion Sc eening Eligibili yIncluded Figu e 1. Flow Diag am o he Me a-analysis. 156 R. A ce e al. / Psychosocial In e en ion (2020) 29(3) 153-164 mon hs, N = 16,050 and k = 26); ea men du a ion (< 16 sessions/ weeks, N = 3,631 and k = 14; and >16 sessions/weeks, N = 15,878 and k = 32); in e en ion le el (indi idual s. mul i-le el); ype o sessions (indi idual sessions, g oup sessions o combined sessions); and ype o ea men (Dulu h Model: N = 15,027 and k = 25; CBTPs: N = 1,629 and k = 9, and OTIs: N = 2,853 and k = 12). The encoding was ca ied ou by an encode who no ed he le els o each o he ca ego ies c ea ed and desc ibed by esea che s, and ma ked exac ly whe e hey we e e e ed o in he ex . In he encoding o he ype o in e en ion, he c i e ion s a ed by he au ho s o he p ima y s udies was applied, bu i should be no ed ha inc easingly in e en ions based on he Dulu h Model also include cogni i e-beha iou al aining echniques o ange managemen , and CBTPs a gende pe spec i e (e.g., pa ia chal a i udes and alues). No wi hs anding, he app oach asc ibed o he in e en ion was in line wi h he models p oposed. A second encode e iewed all o he s udies using he same encoding sys em. The ea e , bo h he egis e ed ca ego ies, and he exac co espondence in he encodings we e checked o es ima e he ue conco dance ( ) wi h he kappa s a is ic (A ce e al., 2000). Kappa co ec s he andom e ec in conco dance, bu no he co espondence in encodings. Thus, he egis a ion o a ca ego y by wo encode s in wo di e en places a e wo encoding e o s in he ue kappa (lack o co espondence), whe eas i is encoded as co ec wi h Cohen’s kappa. Mo eo e , consis ency be ween wo encode s is no su icien o es ima ing he ideli y o encodings (i.e., coding co ec ness in ela ion o he con en o he ca ego y), which would equi e o he encoding o emain s able h ough ime, in a- encode , and in o he e alua ion con ex s (code s in o he s udies). The ue conco dance in e - and in a-encode s was e y good ( > .81). Mo eo e , he encode s we e consis en in o he con ex s i.e., e alua ions (A ias e al., 2013; Gallego e al., 2019). Bea ing in mind he in e - and in a-e alua o and in e -con ex consis ency o code s, encodings we e eliable, ha is, e lec ed he o iginal da a. Resul s Con as o In e -c i e ia Consis ency o he Recidi ism Measu emen The esul s showed a signi ican ly highe a e (+.1540), Z = 12.32, p < .001, o ecidi ism measu ed in CRs, han in ORs. Thus, ORs masked a signi ican amoun o ecidi ism (no su p isingly, be- cause many couples e use o collabo a e, a p edic able beha iou associa ed o e ic imiza ion; B ame e al., 2015; and concealmen o inju ies; A ce, Fa iña, Seijo, e al., 2015). Mo eo e , he con idence ma gin o he mean a e o ecidi ism, highe in CRs han in ORs, CI 95% [.1491, .1589], e ealed i was s able a ound 15%. Analysis o Ou lie s Taking in o accoun ha he e icacy o a ea men a ies acco ding o he a iable measu ed, an analysis o ou lie s was ca ied ou o each measu e, ORs and CRs. Da a showed ha one o he s udies, S i h e al. (2004) wi h CRs, was an ex eme alue (± 3 * IQR). As o he e ec sizes o s udies measu ing ecidi ism in ORs, no ex eme cases we e obse ed no ou lie s wi h he c i e ion ± 1.5 * IQR, bu ex eme cases we e obse ed wi h a much mo e conse a i e c i e ion ± 2 SD (Chau ene ’s c i e ion). This c i e ion equi es ha each me a-analysis should e i y he co ec classi ica ion o ou lie s and incon enien s udies (con a y Table 1. Me a-Analyses o he E ec i eness o Ba e e In e en ion P og ams k N d SD d SD p e SD es δ SD δ% Va 95% CI d 80% CIδ Gene al Me a-Analysis 162 20,860 0.43 0.4485 0.1114 0.4345 0.44 0.4300 6.17 0.40, 0.46 -0.13, 1.00 259 20,215 0.48 0.3278 0.1108 0.3275 0.49 0.3332 10.28 0.45, 0.51 0.06, 0.91 Me a-Analysis o In e en ion E ec i eness Measu ed in CRs 116 1,351 0.004 0.2218 0.2194 0.0321 0.005 0.0344 97.90 -0.10, 0.11 -0.03, 0.05 215 1,340 -0.008 0.0000 0.2131 0.0000 -0.009 0.0000 100.00 -0.11, 0.09 -0.009 Me a-Analysis o In e en ion E ec i eness Measu ed in ORs 146 19,509 0.45 0.4490 0.0988 0.4338 0.45 0.4338 4.93 0.42, 0.48 -0.10, 1.01 244 18,875 0.51 0.3279 0.0984 0.3129 0.51 0.31298 99.00 0.48, 0.54 0.11, 0.91 Me a-Analysis o In e en ion E ec i eness Measu ed in ORs on Sho Follow-up Pe iods 121 3,509 0.35 0.8191 0.1563 0.8041 0.35 0.8041 3.64 0.28, 0.42 -0.68, 1.37 219 2,875 0.69 0.4004 0.1679 0.3635 0.69 0.3635 17.58 0.61, 0.77 0.23, 1.15 Me a-Analysis o In e en ion E ec i eness Measu ed in ORs on Long Follow-up Pe iods 26 16,050 0.48 0.3026 0.0818 0.2914 0.48 0.2914 7.31 0.45, 0.51 0.11, 0.85 Me a-Analysis o In e en ion E ec i eness o Dulu h Model Measu ed in ORs 125 15,027 0.37 0.4627 0.0824 0.4553 0.37 0.4553 3.17 0.34, 0.40 -0.20, 0.96 223 14,393 0.44 0.3168 0.0811 0.3168 0.44 0.3168 6.15 0.41, 0.47 0.04, 0.85 Me a-Analysis o E ec i eness o CBTPs Measu ed in ORs 91,629 0.88 0.1382 0.1562 0.0000 0.88 0.0000 100.00 0.78, 0.98 0.88 Me a-Analysis o E ec i eness o OTIs Measu ed in ORs 12 2,853 0.63 0.2155 0.1333 0.1693 0.63 0.1693 38.26 0.55, 0.71 0.41, 0.85 Me a-Analysis o E ec i eness o B ie In e en ions Measu ed in ORs 114 3,631 0.03 0.7064 0.1244 0.6953 0.03 0.6953 3.10 -0.03, 0.09 -0.86, 0.92 212 2,997 0.29 0.4379 0.1274 0.4379 0.29 0.4379 7.81 0.22, 0.36 -0.27, 0.85 Me a-Analysis o E ec i eness o Long In e en ions Measu ed in ORs 32 15,878 0.55 0.2785 0.0917 0.2630 0.55 0.2630 10.84 0.52, 0.58 0.21, 0.89 No e. k = numbe o e ec sizes; N = o al sample size; d = sample size weigh ed mean e ec size; SD d = s anda d de ia ion o d ; SD p e = s anda d de ia ion p edic ed o sampling e o alone; SD es = s anda d de ia ion o d a e emo ing sampling e o a iance; δ = mean ue e ec size; SD δ = he s anda d de ia ion o δ; % a = pe cen o obse ed a iance accoun ed by a i ac ual e o s; 95% CId = 95% con idence in e al o d; 80% CIδ = 80% c edibili y in e al o δ; 1me a-analysis including all e ec sizes om p ima y s udies; 2me a-analysis emo ing ex eme and/o ou lie s e ec sizes. 157 A e In e en ions wi h Ba e e s E ec i e? A Me a-analy ical Re iew o he hypo hesis o analysis), o mode a o s (Tukey, 1960). Thus, a me a-analysis was pe o med wi h and wi hou hese s udies o de e mine hei impac and hei e ec s on es ima o s. Mo eo e , a me a-analysis wi h and wi hou ex eme alues was pe o med. Analysis o he Gene al E icacy o Ba e e In e en ion P og ams The me a-analysis o he o al e ec size o he s udies on he e icacy o ba e e in e en ions (see Table 1) showed, o 62 e ec sizes and o a sample o 20,860 ba e e s, a posi i e, signi ican (con idence in e al o d does no include ze o), and medium-mode a e mean ue e ec size, δ = 0.44, o ba e e in e en ions. Howe e , hese esul s we e no gene alizable (c edibili y in e al o δ includes ze o) o all s udies on ba e e in e en ions, since in e en ions may ha e nega i e e ec s. Likewise, he esul s wi hou ou lie s ( N = 20,215 and k = 59) showed a posi i e, signi ican (con idence in e al o d does no include ze o), medium magni ude, and gene alizable (c edibili y in e al o δ does no include ze o), i.e., ba e e in e en ions had no nega i e e ec s (on ecidi ism) mean ue e ec size, δ ≈ 0.50. Compa a i ely, he e ec sizes wi h and wi hou a ypical alues we e equi alen , q s = .024, ns . Thus, he elimina ion o ou lie s excluded incon enien esul s ha had no e ec s on global esul s. In e ms o ne in e en ion, e icacy ( educ ion in ecidi ism a e) would be app oxima ely 21.49% ( = .2149), bu wi h a nega i e lowe limi , i.e., an inc ease in ecidi ism o 6.49% (80% LCV [lowe c edibili y alue] con e ed o = -.0649). The magni ude o he e ec size was g ea e han 62.17% o all possibili ies, and 24.34% o all he posi i e e ec sizes o he in e en ions. Ne e heless, he pe cen age o a iance explained by a i ac ual e o s was below 75% in bo h me a-analyses, so hese esul s we e media ed by mode a o s o e ec . The mos ex ensi ely esea ched mode a o is he a iable me- asu ing e ec s (c i e ion), i.e., ecidi ism, which has been assessed in p ima y s udies in bo h ORs (i.e., police, cou s, co ec ional ins- i u ions), and CRs. Analysis o he E ec s o Ba e e In e en ions on he Va iable Measu ing Recidi ism The me a-analysis on he s udies measu ing in e en ion e icacy on he ecidi ism a e in CRs (see Table 1), wi h a sample o 1,351 ba e e s and 16 e ec sizes, e ealed ha in e en ions had no e ec on ecidi ism, wi h a null (δ = 0.005) mean ue e ec size ( U 1 = 0.007, i.e., he independence o he dis ibu ions o ea ed and non- ea ed ba e e s was only 0.7%), and could be nega i e by up o -0.10 o , in o he wo ds, he in e en ion could ha e a nega i e e ec inc easing ecidi ism a e by up o 4.99% ( = -.0499). The lack o e ec s o he in e en ion measu ed in he ecidi ism a e in CRs was no media ed by mode a o s (% VAR > 75). Thus, he esul s we e conclusi e. The esul s we e eplica ed (see Table 1) including he ex eme case ( N = 1,340 and k = 15), which unde pins hei s abili y. The me a-analysis on s udies es ima ing in e en ion e icacy on ecidi ism in ORs, wi h a sample o 19,509 ba e e s and 46 e ec sizes, showed (see Table 1) a posi i e, signi ican (con idence in e al o d does no include ze o), small-medium (δ = 0.45) and non-gene alizable (c edibili y in e al o δ includes ze o) mean ue e ec size in he in e en ion, wi h possible nega i e e ec s o up o 4.99% (80% LCV con e ed o = -.0499), which we e media ed by mode a o s o he ela ionship be ween ea men and ecidi ism (% VAR > 75%). Wi hou ou lie s ( N = 18,875 and k = 44), he esul s we e eplica ed: a posi i e, signi ican , medium, and media ed by mode a o s mean ue e ec size, bu gene alizable (c edibili y in e al o δ does no include ze o) o all he s udies. Compa a i ely, he e ec size o he me a-analysis wi h and wi hou ou lie s we e simila , q s = .029, ns . In sho , he e ec s o ou lie s we e cons ained o he lack o gene aliza ion o he esul s o all o he s udies, so by elimina ing hem we would be disca ding incon enien esul s, no eal ou lie s. Thus, he mean ecidi ism educ ion a e in ORs due o he in e en ion ( e sus non-in e ened o ende s) was 21.95%, bu wi h a lowe nega i e limi , ha is, an inc ease in ecidi ism o up o 4.99%. Succinc ly, on a e age, in e en ions educed ecidi ism in ORs, bu hey could also ha e ad e se e ec s by inc easing ecidi ism a e by mo e han 50% (Taylo & Maxwell, 2009). The magni ude o he e ec was highe han 62.55% o all possibili ies, and highe han 25.1% o all he posi i e e ec sizes o ba e e in e en ions. Thus, he posi i e in e en ion e ec is explained by he measu emen me hod, ROs, a he han he measu ed cons uc , ecidi ism (Podsako e al., 2003). Howe e , he pe cen age o a iance explained by a i ac ual e o s was lowe han 75%. So, he esul s we e media ed by mode a o s. Analysis o he E ec s o Follow-up Time on Recidi ism in ORs The ecidi ism ollow-up pe iod is a c i ical ac o o c i e ion alidi y. In ac , sho ollow-up pe iods can a i icially inc ease he e icacy a e, gi en ha app oxima ely 2/3 o eo ending occu s in he i s wo yea s (Redondo e al., 2001). Acco ding o Gondol , 2000 and Jones and Gondol (2002), eo ending occu s in ba e e s in hal he ime, and he e is widesp ead ag eemen ha mos ba e y goes un epo ed, and no documen ed in ORs (Eu opean Union Agency o Fundamen al Righ s, 2014). Fo his eason, he ollow-up pe iod was aken as a mode a o o he esul s o he in e en ion and we e subdi ided in o sho ollow-up pe iods o 6 o 12 mon hs, and long ollow-up pe iods o mo e han 12 mon hs, a classi ica ion a es ed o be alid (A ias e al., 2013; Gondol , 2000). The me a-analysis on he e icacy o ba e e in e en ions in sho ollow-up pe iods, i.e., 12-mon h o less in ORs (see Table 1), wi h a sample o 3,509 ba e e s and a o al o 21 e ec sizes, ound a posi i e and signi ican (con idence in e al o d does no include ze o) mean ue e ec size (δ = 0.35), bu no gene alizable o o he s udies (c edibili y in e al o δ includes ze o), and o a small magni ude. These esul s we e eplica ed in he me a-analysis wi hou ou lie s ( N = 2,875, k = 19), wi h a posi i e, and signi ican (con idence in e al o d does no include ze o) mean ue e ec size o mode a e magni ude (δ = 0.69), which was gene alizable o o he s udies (c edibili y in e al o δ does no include ze o). Compa a i ely, he e ec size o he in e en ion in he me a-analysis wi hou ou lie s was signi ican ly highe han wi h ou lie s, q s = .164, p < .01, which was an unexpec ed and incong uous esul (ou lie s diminished he e icacy o he in e en ion). The e o e, he esul s should be analysed wi hou ou lie s. In p ac ical e ms, ba e e in e en ions educed he ecidi ism a e in ORs by an a e age 32.61%, wi hou in e en ions wi h ad e se e ec s, (80% LCV = .23), and he e ec size was highe han 68.79% o all possibili ies and 37.58% o all posi i e in e en ion e ec sizes. Howe e , hese esul s we e media ed by mode a o s, bo h in he me a-analysis wi h all o he e ec sizes, and wi hou ou lie s (% VAR < 75). The me a-analysis on he e icacy o ba e e in e en ions on 12-mon h ollow-up pe iods longe in ORs (see Table 1), wi h 26 e - ec sizes and a sample o 16,050 ba e e s, e ealed a posi i e and signi ican (con idence in e al o d does no include ze o) mean ue e ec size o mode a e magni ude (δ ≈ 0.50). Mo eo e , his e- sul was gene alizable o o he s udies (c edibili y in e al o δ does no include ze o), bu was media ed by mode a o s (% VAR < 75). As o he educ ion in he ecidi ism a e in ORs, ba e e in e en ions educed he ecidi ism a e by 23.34% ( = .2334) and, once again, wi hou de ec ing nega i e in e en ion e ec s (80% LCV = .11). The 158 R. A ce e al. / Psychosocial In e en ion (2020) 29(3) 153-164 magni ude o he e ec size was abo e he 63.31% o all possibili ies and 26.62% o all he posi i e in e en ion e ec sizes. The mean ue e ec size o he in e en ion in ORs wi h a sho ollow-up pe iod o up o 12 mon hs (δ = 0.69) o less was signi ican ly la ge , q s = .101, p < .05, han ha ob ained o he long ollow-up pe iod o mo e han 12 mon hs (δ = 0.48). Mode a o s could no be analysed as N s (< 400) o e ec sizes we e insu icien ( k ≤ 3). Analysis o he E ec s o he Modali y o he In e en ion on Recidi ism in ORs Ano he mode a o epo ed in he li e a u e as ha ing e ec s on he in e en ion is he modali y o he in e en ion, ca ego ized as in e - en ions based on he Dulu h Model, Cogni i e-Beha iou al T ea men in e en ions, and O he Type o In e en ions (A ias e al., 2013; Bab- cock e al. 2004; Fede & Wilson, 2005; Le esque & Gelles, 1998). Me a-analysis o he e ec s on ecidi ism in ORs o in e en- ions based on he Dulu h Model wi h a sample o 15,027 ba e e s, and 25 e ec sizes e ealed (see Table 1) a posi i e and signi ican (con idence in e al o d does no include ze o) mean ue e ec size o small magni ude ( d = 0.37). Mo eo e , hese esul s we e no gene alizable (c edibili y in e al o δ includes ze o) o all s udies wi h in e en ions based on he Dulu h Model. These esul s we e eplica ed wi hou ou lie s ( N = 14,393, k = 23), wi h he excep ion ha hey we e gene alizable o all he s udies (c edibili y in e al o δ does no include ze o). Compa a i ely, he e ec sizes o all he s udies, and hose wi hou ou lie s we e simila , q s = .034, ns . Thus, he ou lie s we e no such, bu incon enien s udies wi h nega i e e ec s. In p ac ical e ms, in e en ions based on he Dulu h Model educed on a e age ecidi ism by 18.19% ( = .1819), wi h po en ial nega i e e ec s o up o 9.95% (80% LCV con e ed o = -.0995), and an e ec size abo e 60.25% o all possibili ies, and 20.50% o all he posi i e in e en ion e ec sizes. Howe e , he esul s we e media ed by mode a o s (% VAR < 75). The me a-analysis o he e ec s o ba e e in e en ions wi h CBTPs on ecidi ism in ORs (see Table 1), wi h a o al o 9 e ec sizes and a sample o 1,629 ba e e s, exhibi ed a posi i e, signi ican o he in e en ion (con idence in e al o d does no include ze o), and o a la ge magni ude (δ = 0.88) mean ue e ec size, implying an a e age educ ion in he ecidi ism a e o 40.27% ( = .4027). The gene aliza ion o he ue e ec size could no be s udied as he a iance was ze o (i.e., he s udies o his me a-analysis we e no andomly dis ibu ed); hus, i was analysed wi h he e ec size weigh ed by sampling e o (conse a i e es ima e e sus he ue size), and was gene alizable (c edibili y in e al o d does no include ze o) o o he s udies, and wi hou nega i e e ec s, 80% CI [0.70, 1.06]. Addi ionally, he lowe limi o he e ec (90% o he in e en ions su passed his limi ) was 33.40% (80% LCV con e ed o = .3340). The magni ude o he e ec size was highe han 73.24% o all possibili ies, and 46.48% o all he posi i e in e en ion e ec sizes. The me a-analysis o he e ec s o O he Types o In e en ion (OTIs) on ecidi ism in ORs (see Table 1), wi h a o al o 12 e ec sizes, and a sample o 2,853 ba e e s, ound a posi i e, mode a e, and signi ican (con idence in e al o d does no include ze o) mean ue e ec size (δ = 0.63), gene alizable o all he s udies (c edibili y in e al o δ does no include ze o). The educ ion in he ecidi ism a e wi h OTIs ( e sus he a e o non-in e ened ba e e s) was on a e age 30.04% ( = .3004), and wi hou in e en ions wi h nega i e e ec s, and a minimum h eshold in he educ ion o he ecidi ism a e o 20.08% (80% LCV con e ed o = .2008). The educ ion in he ecidi ism a e by OTIs ( e sus he a e o non-in e ened ba e e s) was on a e age 30.48%, wi hou nega i e e ec s, and wi h a minimum h eshold in he educ ion o ecidi ism o 20.55%. The magni ude o he e ec size was highe han 67.36% o all possibili ies, and 34.72% o all he posi i e in e en ion e ec sizes. Ne e heless, he e ec o hese in e en ion modali ies on ecidi ism was media ed by mode a o s (% VAR < 75). Once again, he analysis o mode a o s in he di e en ypes o in e en ions was no possible due o insu icien N s and k s. Compa a i ely, he e ec o he in e en ion on he educ ion o ecidi ism was signi ican ly highe in he CBTPs han in he Dulu h Model, q s = .243, p < .01 and in he OTIs, q s = .117, p < .01, and in he OTIs han in he Dulu h Model, q s = .126, p < .01. Analysis o E ec s o Du a ion o he In e en ion on Recidi ism in ORs A u he mode a o o he e ec s ha is sys ema ically e iewed conce ns he du a ion o he in e en ion (A ias e al., 2013; Babcock e al., 2004). In acco dance wi h Babcock e al. (2004), wo ca ego ies we e c ea ed: sho du a ion o p og ams unde 16 sessions/weeks (4 mon hs) and long du a ion o p og ams o mo e han 16 weeks/ sessions. The esul s o he me a-analysis o sho in e en ions (< 16 sessions), wi h a o al o 14 e ec sizes and 3,631 ba e e s (see Table 1), showed a non-signi ican (con idence in e al o d includes ze o) mean ue e ec size and, consequen ly, i was no gene alizable o o he s udies. The me a-analysis eplica ed wi hou ou lie s ( N = 2,997, k = 12) ound a posi i e and signi ican (c edibili y in e al o δ does no include ze o) mean ue e ec size o small magni ude (δ = 0.29), no gene alizable (c edibili y in e al o δ includes ze o), and media ed by mode a o s (% VAR < 75) ha could no be analysed owing o he insu icien N s and k s. The compa ison o e ec sizes wi h all o he s udies wi h he exclusion o ou lie s e ealed signi ican di e ences, q s = .129, p < .01, anging om a non-signi ican (all o he s udies) o a signi ican (wi hou ou lie s) mean ue e ec size. Thus, he ou lie s we e no such, bu incon enien s udies wi h e y nega i e e ec s. In ac , b ie p og ams can inc ease he ecidi ism a e by 39.5% (80% LCV con e ed o = .3950). Rega ding long in e en ion p og ams (> 16 sessions), he esul s o he me a-analysis wi h a o al o 32 e ec sizes and a sample o 15,878 ba e e s (see Table 1) displayed a posi i e and signi ican (con idence in e al o d does no include ze o) mean ue e ec size o mode a e magni ude (δ = 0.55), gene alizable o o he s udies (c edibili y in e al o δ does no include ze o). As o he e icacy o he in e en ion, long in e en ions educed he ecidi ism a e in ORs by 26.52% ( = .2652), and wi hou in e en ions wi h nega i e e ec s, and a minimum h eshold in he educ ion o ecidi ism o 10.44% (80% LCV con e ed o = .1044). The magni ude o he e ec size was highe han 65.17% o all possibili ies and 30.34% o all he posi i e in e en ion e ec sizes. Howe e , he a e age e ec size was media ed by mode a o s (% VAR < 75) ha we e no s udied owing o insu icien N s and k s. Compa a i ely, long p og ams we e mo e e ec i e in educing ecidi ism in compa ison o sho in e en ions, q s= .257, p < .01. Discussion The p esen me a-analyses ha e se e al limi a ions ha should be bo ne in mind in he gene aliza ion o esul s. Fi s , he classi ica ion o s udies aken om p ima y s udies ou lined he cha ac e is ics o de ining hem (e.g., an in e en ion was ca ego ized unde a pa icula model, bu he desc ip ion also encompassed o he models; he leng h o he ollow-up was di e en ly measu ed); hus, i was ca ego ically assumed in he me a-analysis, when he e may be subs an ial a iabili y in ca ego ies o o e lapping ca ego ies. Second, no all s udies consis en ly epo ed mode a o s (e.g., ea men ype), wi h he subsequen exclusion o hem. Thi d, he alidi y o he measu e o 159 A e In e en ions wi h Ba e e s E ec i e? A Me a-analy ical Re iew e icacy o in e en ions on ecidi ism ailed o be a good es ima o , as i may no e en be he eal objec i e o in e en ions unde Spanish law ea men in co ec ional ins i u ions is olun a y, and he manda e o co ec ional ins i u ions is o de elop ap i udes o o e come de iciencies and o modi y un a ou able and nega i e a i udes; Ley O gánica 1/1979, de 26 de sep iemb e, Gene al Peni encia ia). Fou h, he e alua ion o he e icacy o he in e en ion on ecidi ism based on o icial eco ds was a c i e ion o limi ed alidi y (a s ong posi i e biased app oxima ion o ecidi ism; in o he wo ds, in compa ison o CRs, i ailed o cap u e app oxima ely hal o he ecidi ism). Fi h, ollow-up ime, as a measu e o he e icacy o in e en ions on ecidi ism, had a di ec and big e ec on he alidi y o he measu e. In gene al, an es ima ed 50 o 60% o ecidi ism occu ed du ing he i s yea (Redondo e al., 2001), wi h a o al con i med ecidi ism a e anging om 40% o mo e han 80%, which a ied acco ding o he measu e and he ollow-up ime (S o e , 2005). Six h, he measu e o he e icacy o in e en ions in CRs was also p obably posi i ely biased (o e es ima ion o he esul s o he in e en ions) bo h in he measu e i sel (e.g., no all o he ba e e s who had a pa ne a e ea men epo ed i ; he exposu e imes o couples wi h di ec e ec s on ecidi ism we e no epo ed; a signi ican numbe o women ailed o espond o encode s) and due o esponse bias (e.g., bias o leniency, i.e., endency o minimize o conceal assaul s). In ac , one o he dis inc i e cha ac e is ics o ic ims o gende iolence (inju y) is he ailu e o epo o ences and he concealmen o assaul s in epo s (Hol e al., 2003), concealing o ences (A ce, 2017), minimizing inju ies (A ce, Fa iña, & Vila iño, 2015; Vila iño e al., 2009), and ailing o epo he ecidi ism o in e ened ba e e s (B ame e al., 2015), in pa icula du ing cohabi a ion wi h he ba e e . Se en h, e alua ions o e icacy we e aken om he au ho s o he in e en ions, who ely on e icacy o p og am con inui y. Thus, he numbe o unpublished s udies wi h no e ec s o nega i e e ec s may be high. Eigh h, es alues we e also subjec o simila measu emen e o s. Nin h, he d opou s linked o ecidi ism we e no gene ally encoded as such, which would a i icially inc ease he e icacy a e (Ta e al., 2001). Ten h, o he mode a o s (no s udied as N s o k s we e insu icien ) may modi y he magni ude o di ec ion o he ea men e icacy (Ma ín & Mo al, 2019; Ma ín e al., 2019). Thus, u he esea ch wi h addi ional mode a o s should be designed. In sho , he esul s had a ce ain deg ee o alidi y o measu ing he e icacy o in e en ions e sus non-in e en ion, bu no su icien o de e mine he eal a es o ecidi ism, which we e p esumably signi ican ly highe han egis e ed. These limi a ions, in e ms o alidi y o conclusions, can be classi- ied as sys ema ic measu emen e o s (when hey gene a e an al e - na i e explana ion o he esul s), o a biased measu emen me hod (i.e., a iance a ibu ed o me hod and no o measu ed cons uc ) (Podsako e al., 2003). Bea ing in mind he implica ions o hese limi a ions on he alidi y o he conclusions o his me a-analysis, hey a e discussed u he below: a) The con ol and analysis o he e ec s in he esul s o ex eme alues and ou lie s was open o wo di e en in e p e a ions. The ex eme alue e e ed o a ze o ecidi ism a e, an impossible esul ha also dis o s esul s. Mo eo e , ou lie s ( he classi ica ion c i e ion was he s a is ical signi icance, wi h a p obabili y o occu ence o 5%, e sus he common c i e ion ± 1.5 * IQR, much mo e s ingen and ha would no classi y any as an ou lie ) we e gene ally (wi h he excep ion o sho ollow-up pe iods) incon enien s udies (con a y o he hypo hesis o he e icacy o in e en ions pos ula ed in he analysis i sel ), which we e no p ecisely ou lie s, bu mode a o s (Tukey, 1960). The e o e, hey a e alid cases whe e in e en ions a e obse ed o ha e no esul s o ad e se e ec s, a possibili y ha has been alida ed in p e ious e iews (A ias e al., 2013; Babcock e al., 2004; Fede & Wilson, 2005; Le esque y Gelles, 1998). b) The ecidi ism a e was almos 14% in in e ened ba e e s, nea ly 13% (.1295) in ORs, and 26.13% (.2613) in CRs. Thus, hal o ecidi ism wen un epo ed in ORs. Ne e heless, his was s ill a om he ue a e, gi en ha , as p e i- ously s a ed, sou ces o measu emen e o we e inhe en o he me hod i sel (measu e o in e en ion e icacy), no he cons uc unde assessmen ( ecidi ism), and ended o maximize he e icacy o in e en ions ( a iance a ibu ed o he me hod, no o he cons uc ). Hence, e alua ion was a sou ce o sys ema ic e o : maximizing he e ec s o in e - en ions. c) The esul s o he gene al me a-analysis e ealed ha in e ened ba e e s bene i ed om in e en ions compa ed o non-in e ened ba e e s. Acco ding o his me a-analysis, he gene al e icacy in he educ ion o he ecidi ism a e was app oxima ely 21%. Though he e icacy o in e en ions was es ima ed o be as negligible in e iews on he issue (Babcock e al., 2004), his would no be so in ela ion o he o al po en ial in e en ion e ec s, whe e e icacy ob ained would be abo e 62.17%, P62.17, and 24.34% o posi i es ( educ ion in ecidi ism a e). Fu he mo e, e icacy was highe han ha ob ained o delinquency in gene al (Redondo e al., 2002), and highe han ha ob ained o sexual ba e e s (Schmucke & Lösel, 2015). Ne e heless, in e en ions can ha e no e ec on ecidi ism, o e en nega i e e ec s (inc ease ecidi ism). In sho , in gene al in e en ions, educed measu ed ecidi ism a es, bu no all in e en ions, we e e icacious. d) Mos o he e ec s o in e en ions (93.5%) we e measu ed in ecidi ism in ORs. Me a-analysis on he e icacy o in e en ions wi h his measu e showed a signi ican educ ion in ecidi ism a e (a smalle p obabili y o being accused o sen enced o in ima e pa ne iolence han non-in e ened ba e e s), es ima ed o be a ound 22%, an e icacy a e abo e 62.55%, P62.55, o all possibili ies, and 25.1% o posi i e e ec s. Once again, no all in e en ions had he e ec o educing ecidi ism, and some may e en ha e nega i e e ec s (inc ease ecidi ism), implying no all in e en ions we e alid. e) The ollow-up pe iod on e icacy o in e en ions media ed di e ences in ecidi ism in ORs. Whils ecidi ism a e in sho ollow-ups ell o a ound 33%, educ ion in long ollow- ups was 23%, which was a s a is ically signi ican educ ion. Thus, sho ollow-up pe iods a i icially boos ed ecidi ism educ ion a e by 10%. Mo eo e , hese esul s co obo a ed ha 2/3 o all ecidi ism, in acco dance wi h esul s by Redondo e al. (2001), occu ed du ing he i s wo yea s, o in hal ha ime o ba e e s acco ding o Gondol , 2000 and Jones and Gondol , 2002. Thus, e alua ions in sho ollow- up pe iods we e in alid. ) In e en ion con en s had a di ec impac on he educ ion o ba e e ecidi ism in ORs. As o in e en ion model, he highes e ec s we e ob ained wi h CBTPs, a esul ha has been sys ema ically subs an ia ed in li e a u e e iews (A ias e al., 2013; Babcock e al., 2004; Fede & Wilson, 2005; Le esque & Gelles, 1998), and suppo s he iew ha in e en ions should be mul imodal: cogni ion and beha iou (Lila e al., 2019). A small, and signi ican ly lowe e ec , was ob ained wi h in e en ions based on he Dulu h Model, which could e en ha e ad e se e ec s (mo e ecidi ism). The small e ec size, in addi ion o he high numbe o d opou s du ing hese in e en ions, has led Ta e al. (2001) o he conclusion ha hese ypes o in e en ion we e possibly no e ec i e. This claim was no co obo a ed by he esul s, as e ec size was signi ican and associa ed o a mean ecidi ism educ ion a e o app oxima ely o 18%. 160 R. A ce e al. / Psychosocial In e en ion (2020) 29(3) 153-164 No wi hs anding, he high a iabili y obse ed could ha e nega i e e ec s. I we ake in o accoun ha p og ams based on he Dulu h Model we e ini ially aimed a educa ing in a gende pe spec i e, bu we e subsequen ly ans o med in o cogni i e-beha iou al models wi h a gende pe spec i e, addi ion o gende pe spec i e may be he cause o he decline in e icacy e sus exclusi ely cogni i e-beha iou al in e en ions. Tha is, gende pe spec i e o in e en ion may ail o e-educa e ba e e s, o e en ha e ad e se e ec s ( eac ance o ea men o a eigning adhe ence o ea men could explain his con ingency). g) OTIs (gene ally ocused on cogni ion and beha iou , bu supplemen ed and guided by o he in e en ion models, such as ecological, he apeu ic, o mul ile el models) had he e ec o signi ican ly educing ecidi ism in ORs, and in e media e be ween Dulu h Model and CBTPs. Ne e heless, hese esul s we e limi ed o he me a- analysis s udies and we e no gene alizable o models o he han hose examined ( his ca ego y includes any model ha was no s ic ly cogni i e-beha iou al o Dulu h). A close inspec ion o esul s indica es cogni i e-beha iou al in e en ions in addi ion o an ecological o mul ile el in e en ion models imp o ed he e icacy o in e en ions, he eby he need o u he s udies o e alua e bo h o hese combined models. h) Sho in e en ions ailed o educe ecidi ism in ORs and could ha e nega i e e ec s (inc ease ecidi ism as compa ed o no in e en ion) by almos 40%. In compa ison, long in e - en ions we e e ec i e in educing ecidi ism a e in ORs (≈ 26%), wi hou nega i e e ec s. This unde sco es ba e e in e - en ions should be long- e m o be e ec i e. Ba e e s ha bou oxic cogni ions (Ma una, 2004), which a e highly esis an o in e en ions (No o e al., 2012), in pa icula sho in e - en ions ha may be ine ec i e o e en coun e p oduc i e. i) Howe e , in e ms o ecidi ism in CRs, he e icacy o in e en ions e sus non-in e en ion was o ally null, wi h a de ini e esul (i.e., gene alizable and no media ed by mode a o s). Hence, in e en ions we e e ec i e in educing he a e o accusa ions and guil y e dic s, bu no he a e o eal, di ec in ima e pa ne iolence. This esul unde sco ed he limi a ions o ORs as a measu e o he e icacy o in e en ions, ha is, ORs we e a poo alid measu e as a subs an ial amoun o ecidi ism wen unde ec ed. Se e al causes may be esponsible o his limi a ion in alidi y: un epo ed ecidi ism, ejec ed complain s by he police, no all gende iolence eco ded in CRs cons i u es gende iolence ac oss all coun ies, and i may be eco ded as sexis iolence which alls ou o he scope o ba e e in e en ions. In conclusion, he e is a co pus o li e a u e on he e icacy o in- e en ions, showing signi ican e ec s in educing ecidi ism in o - icial eco ds. In o he wo ds, in e ened ba e e s we e less likely o be accused/sen enced again in (ORs) o he same o ence. No wi h- s anding, no all o he in e en ions we e e ec i e in ORs. Thus, sho in e en ions we e comple ely ine ec i e and could ha e nega i e e ec s o up o 40%, and ce ain in e en ions based on he Dulu h Model may ha e nega i e e ec s o up o 10%. In con as , long in e - en ions based on CBTPs o OTIs ( he esul s may no be gene alised o o he echniques han hose e ised) we e on a e age e ec i e and wi hou nega i e e ec s on ecidi ism in ORs. The e alua ion o he e icacy o he in e en ion in ORs was no a alid measu e as i was subjec o a sys ema ic measu emen e o (i minimized he a e), which could each o e 80% (S o e , 2005). Howe e , ega dless he ac ha ORs we e no a alid measu e o ecidi ism, we may d aw om he me a-analysis wi h ORs he conclusion ha in e en ion p og ams should be based on a long cogni i e-beha iou al app oach. Howe e , his does no imply ha in e en ions had a signi ican e ec on educing ecidi ism in all measu es o iolence. The e was a null e ec in couple epo s. Fu he esea ch should be unde aken o e alua e he e icacy o in e en ions in bo h ypes o measu e o iden i y he ype o ecidi ism los in ORs and o iden i y mode a o s o he e icacy o in e en ions on his measu e. Mo eo e , s udies a e equi ed o assess he e icacy o in e en ions on o he c i e ia such as he in e nal mechanisms unde lying gende iolence (A ce e al., 2014), and acquisi ion o skills and abili ies ha enable ba e e s o success ully manage isk si ua ions o in ima e pa ne iolence (A ce & Fa iña, 2010; Lila e al., 2018; Ma ín-Fe nández, G acia, Ma co, e al., 2018; Ma ín- Fe nández, G acia, & Lila, 2018). Con lic o In e es The au ho s o his a icle decla e no con lic o in e es . Re e ences Re e ences ma ked wi h an as e isk indica e s udies included in he me a- analysis Amado, B. G., A ce, R., & He aiz, A. (2015). Psychological inju y in ic ims o child sexual abuse: A me a-analy ic e iew. Psychosocial In e en ion, 24 (1), 49-62. h ps://doi.o g/10.1016/j.psi.2015.03.002 A ce, R. (2017). Análisis de con enido de las decla aciones de es igos: e aluación de la alidez cien í ica y judicial de la hipó esis y la p ueba o ense [Con en analysis o he wi ness s a emen s: E alua ion o he scien i ic and judicial alidi y o he hypo hesis and he o ensic p oo ]. Acción Psicológica, 14 (2), 171-190. h ps://doi.o g/10.5944/ ap.14.2.21347 A ce, R., & Fa iña, F. (2010). Diseño e implemen ación del p og ama Galicia de eeducación de mal a ado es: una espues a psicosocial a una necesidad social y peni encia ia [Design and implemen a ion o he Galicia p og am o ba e e e-educa ion: A psychosocial answe o a social and peni en ia y need] Psychosocial In e en ion, 19 (2), 153- 166. h ps://doi.o g/10.5093/in2010 19n2a7 A ce, R., Fa iña, F., & F aga, A. I. (2000). Géne o y o mación de juicios en un caso de iolación [Gende and ju o judgmen making in a case o ape]. Psico hema, 12 (4), 623-628. h p://www.psico hema.com/pd /381.pd A ce, R., Fa iña, F., & No o, M. (2014). Compe encia cogni i a en penados p ima ios y einciden es: implicaciones pa a la eeducación [Cogni i e compe ence among ecidi is and non- ecidi is p isone s: Implica ions o he ehabili a ion]. Anales de Psicología, 30 (1). 259- 266. h ps://doi.o g/10.6018/analesps.30.1.158201 A ce, R., Fa iña, F., Seijo, D., & No o, M. (2015). Assessing imp ession managemen wi h he MMPI-2 in child cus ody li iga ion. Assessmen , 22 (6), 769-777. h ps://doi.o g/10.1177/1073191114558111 A ce, R., Fa iña, F., & Vila iño, M. (2015). Daño psicológico en casos de íc imas de iolencia de géne o: un es udio compa a i o de las e aluaciones o enses [Psychological inju y in in ima e pa ne iolence cases: A con as i e analysis o o ensic measu es]. Re is a Ibe oame icana de Psicología y Salud, 6 (2), 72–80. h ps://doi. o g/10.1016/j. ips.2015.04.002 A ias, E., A ce, R., & Vila iño, M. (2013). Ba e e in e en ion p og ammes: A me a-analy ic e iew o e ec i eness. Psychosocial In e en ion, 22 (2), 153-160. h ps://doi.o g/10.5093/in2013a18 Babcock, J. C., G een, C. E., & Robie, C. (2004). Does ba e e s’ ea men wo k? A me a-analy ic e iew o domes ic iolence ea men . Clinical Psychology Re iew, 23 (8), 1023-1053. h ps://doi.o g/10.1016/j. cp .2002.07.001 *Babcock, J. C., & S eine , R. (1999). The ela ionship be ween ea men , inca ce a ion and ecidi ism o ba e ing: A p og am e alua ion o Sea le’s coo dina ed communi y esponse o domes ic iolence. Jou nal o Family Psychology, 13 (1), 46-59. h ps://doi. o g/10.1037//0893-3200.13.1.46 *Benne , L., Call, C., Fle , H., & S oops, C. (2005). P og am comple ion, beha io al change and e-a es o he ba e e in e en ion sys em o Cook Coun y, Illinois. Illinois C iminal Jus ice In o ma ion Au ho i y. Chicago, IL. h p://ci esee x.is .psu.edu/ iewdoc/ download?doi=10.1.1.589.9366& ep= ep1& ype=pd *Bowen, E., Gilch is , E. A., & Beech, A. R. (2005). An examina ion o he impac o communi y-based ehabili a ion on he o ending beha iou o male domes ic iolence o ende s and he cha ac e is ics associa ed wi h ecidi ism. Legal and C iminological Psychology, 10 (2), 189-209. h ps://doi.o g/10.1348/135532505x36778 B ame, R., Kaukinen, C., Go e , A. R., & La imo e, P. K. (2015). No-con ac o de s, ic im sa e y, and o ende ecidi ism in cases o misdemeano c iminal domes ic iolence: A andomized expe imen . Ame ican 161 A e In e en ions wi h Ba e e s E ec i e? A Me a-analy ical Re iew Jou nal o C iminal Jus ice, 40 (2), 225-249. h ps://doi.o g/10.1007/ s12103-014-9242-x *Coul e , M., & VandeWee d, C. (2009). Reducing domes ic iolence and o he c iminal ecidi ism: E ec i eness o a mul ile el ba e e s in e en ion p og am. Violence and Vic ims, 24 (2), 139-152. h ps:// doi.o g/10.1891/0886-6708.24.2.139 *Da is, R. C., Taylo , B. G., & Maxwell, C. D. (1998). Does ba e e ea men educe iolence? A andomized expe imen in B ooklyn. Jus ice Qua e ly, 18 (1), 171-201. h ps://doi.o g/10.1080/07418820100094861 *Dobash, R., Dobash, R. E., Ca anagh, K., & Lewis, R. (1996). Reeduca ion p og ams o iolen men: An e alua ion (Resea ch Findings, No. 46). h p://weba chi e.na ionala chi es.go .uk/20110314171826/h p:// ds.homeo ice.go .uk/ ds/pd s/ 46.pd *Dun o d, F. W. (2000). The San Diego Na y expe imen : An assessmen o in e en ions o men who assaul hei wi es. Jou nal o Consul ing and Clinical Psychology, 68 (3), 468-476. h ps://doi.o g/10.1037//0022- 006x.68.3.468 Eu opean Union Agency o Fundamen al Righ s. (2014). Violence agains women: An EU-wide su ey. Resul s a a glance. h ps:// a.eu opa.eu/ si es/de aul / iles/ a-2014- aw-su ey-a -a-glance-oc 14_en.pd Fa iña, F., Redondo, L., Seijo, D., No o, M., & A ce, R. (2017). A me a-analy ic e iew o he MMPI alidi y scales and indexes o de ec de ensi eness in cus ody e alua ions. In e na ional Jou nal o Clinical and Heal h Psychology, 17 , 128-138. h ps://doi.o g/10.1016/j.ijchp.2017.02.002 *Fede , L. R., & Dugan, L. (2004). Tes ing a cou -manda ed ea men p og am o domes ic iolence o ende s: The B owa d expe imen . Na ional Ins i u e o Jus ice. Washing on, DC. h ps://www.ncj s.go / pd iles1/nij/199729.pd Fede , L., & Wilson, D. B. (2005). A me a-analy ic e iew o cou -manda ed ba e e in e en ion p og ams: Can cou s a ec abuse s’ beha io ? Jou nal o Expe imen al C iminology, 1 (2), 239-262. h ps://doi. o g/10.1007/s11292-005-1179-0 Gallego, R., No o, M., Fa iña, F., & A ce, R. (2019). Child- o-pa en iolence and pa en - o-child- iolence: A me a-analy ic e iew. Eu opean Jou nal o Psychology Applied o legal Con ex . Ad ance online publica ion. h ps://doi.o g/10.5093/ejpalc2019a4 Gondol , E. W. (2000). Reassaul a 30-mon hs a e ba e e p og am in ake. In e na ional Jou nal o O ende The apy and Compa a i e C iminology, 44 (1), 111-128. h ps://doi.o g/10.1177/0306624x00441010 *G aña, J. L., Redondo, N., Muñoz-Ri as, M. J., & Cuenca, M. L. (2017). Sho - e m ea men esponse, a i ion and ecidi ism in a pa ne iolen men ypology compa ed wi h a con ol g oup. Psicología Conduc ual, 25 (3), 465-482. *Haggå d, U., F eij, I., Danielsson, M., Wenande , D., & Långs öm, N. (2017). E ec i eness o he IDAP ea men p og am o male pe pe a o s o in ima e pa ne iolence: A con olled s udy o c iminal ecidi ism. Jou nal o In e pe sonal Violence, 32 (7), 1027-1043. h ps://doi. o g/10.1177/0886260515586377 *Hasisi, B., Shoham, E., Weisbu d, D., Ha i , N., & Zelig, A. (2016). The “ca e package” p ison domes ic iolence p og ams and ecidi ism: A quasi- expe imen al s udy. Jou nal o Expe imen al C iminology, 12 (4), 563- 586. h ps://doi.o g/10.1007/s11292-016-9266-y Hedges, L. V., & Olkin, I. (1985). S a is ical me hods o me a-analysis. Academic P ess. *He man, K., Ro unda, R., Williamson, G., & Vodano ich, S. (2014). Ou comes om a Dulu h model ba e e in e en ion p og am a comple ion and long e m ollow-up. Jou nal o O ende Rehabili a ion, 53 (1), 1-18. h ps://doi.o g/10.1080/10509674.2013.861316 Hol , V. L., Ke nic, M. A., Wol , M. E., & Ri a a, F. P. (2003). Do p o ec ion o de s a ec he likelihood o u u e pa ne iolence and inju y? Ame ican Jou nal o P e en i e Medicine, 24 , 16-21. h ps://doi. o g/10.1016/S0749-3797(02)00576-7 *Jenkins, J. A., & Men on, C. (2003). The ela ionship be ween inca ce a ed ba e e s’ cogni i e cha ac e is ics and he e ec i eness o beha io al ea men . Co ec ions Compendium, 28 (3), 1-27. h p://wheels o jus ice. wu.edu/publica ions/Co ec ionsCompendium-Ma 2003.pd *Jones, A. S., & Gondol , E. W. (2002). Assessing he e ec o ba e e p og am comple ion on eassaul : An ins umen al a iables analysis. Jou nal o Quan i a i e C iminology, 18 (1), 71-98. h ps://doi. o g/10.1023/a:1013244929733 *Lab iola, M., Rempel, M., & Da is, R. C. (2005). Tes ing he e ec i eness o ba e e p og ams and judicial moni o ing. Resul s om a andomized ial a he B onx Misdemeano Domes ic Violence Cou . Cen e o Cou Inno a ion. Nue a Yo k, NY. h p://www.cou inno a ion.o g/ si es/de aul / iles/ba e e p og amse ec i eness.pd Le esque, D. A., & Gelles, R. J. (1998, July). Does ea men educe ecidi ism in men who ba e ? A me a-analy ic e alua ion o ea men ou come. Pape p esen ed a he P og am E alua ion and Family Violence Resea ch: An In e na ional Con e ence, Du ham, NH. Ley O gánica 1/1979, de 26 de sep iemb e, Gene al Peni encia ia. (1979). Bole ín O icial del Es ado, 281 , 33987-34058. h ps://www.boe.es/ busca /pd /1995/BOE-A-1995-25444-consolidado.pd *Lila, M., G acia, E., & Ca alá-Miñana, A. (2018). Indi idualized mo i a ional plans in ba e e in e en ion p og ams: A andomized clinical ial. Jou nal o Consul ing and Clinical Psychology, 86 (4), 309-320. h ps:// doi.o g/10.1037/ccp0000291 Lila, M., Ma ín-Fe nández, M., G acia, E., López-Osso io, J. J., & González, J. L. (2019). Iden i ying key p edic o s o ecidi ism among o ende s a ending a ba e e in e en ion p og am: A su i al analysis. Psychosocial In e en ion, 28 (3), 157-167. h ps://doi.o g/10.5093/ pi2019a19 *Lin, S., Su, C., Chou, F. H., Chen, S., Huang, J., Wu, G. T., Chen, W. J., Chao, S. S., & Chen, C. (2009). Domes ic iolence ecidi ism in high- isk Taiwanese o ende s a e he comple ion o iolence ea men p og ams. Jou nal o Fo ensic Psychia y & Psychology, 20 (3), 458-472. h ps://doi.o g/10.1080/14789940802638341 Ma ín, A. M., Pad ón, F., & Redondo, S. (2019). Ea ly na a i es o desis ance om c ime in di e en p ison egimes. Eu opean Jou nal o Psychology Applied o Legal Con ex , 11 (2), 71-79. h ps://doi. o g/10.5093/ejpalc2019a2 Ma ín, B., & Mo al, M. V. (2019). Relación en e dependencia emocional y mal a o psicológico en o ma de ic imización y ag esión en jó enes [Rela ionship be ween emo ional dependence and psychological abuse in he o m o ic imiza ion and agg ession in young people]. Re is a Ibe oame icana de Psicología y Salud, 10 (2), 75-89. h ps://doi. o g/10.23923/j. ips.2019.02.027 Ma ín-Fe nández, M., G acia, E., & Lila, M. (2018). Assessing ic im- blaming a i udes in cases o in ima e pa ne iolence agains women: De elopmen and alida ion o he VB-IPVAW scale. Psychosocial In e en ion, 27 (3), 133-143. h ps://doi.o g/10.5093/pi2018a18 Ma ín-Fe nández, M., G acia, E., Ma co, M., Va gas, V., San i so, F. A., & Lila, M. (2018). Measu ing accep abili y o in ima e pa ne iolence agains women: De elopmen and alida ion o he A-IPVAW scale. Eu opean Jou nal o Psychology Applied o Legal Con ex , 10 (1), 26-34. h ps://doi.o g/10.5093/ejpalc2018a3 Ma una, S. (2004). Desis ance and explana o y s yle: A new di ec ion in he psychology o e o m. Jou nal o Con empo a y C iminal Jus ice, 20 , 184-200. h ps://doi.o g/10.1177/1043986204263778 Mon ei o, A., Vázquez, M. J., Seijo, D., & A ce, R. (2018). ¿Son los c i e ios de ealidad álidos pa a clasi ica y disce ni en e memo ias de hechos au o-expe imen ados y de e en os is os en ídeo? [A e he eali y c i e ia alid o classi y and o disc imina e be ween memo ies o sel -expe ienced e en s and memo ies o ideo-obse ed e en s?]. Re is a Ibe oame icana de Psicología y Salud, 9 (2), 149-160. h ps:// doi.o g/10.23923/j. ips.2018.02.020 Mosie , C. I. (1943). On he eliabili y o a weigh ed composi e. Psychome ika, 8 (3), 161-168. h ps://doi.o g/10.1007/BF02288700 *Mu phy, C. M., Musse , P. H., & Ma on, K. I. (1998). Coo dina ed communi y in e en ion o domes ic abuse s: In e en ion sys em in ol emen and c iminal ecidi ism. Jou nal o Family Violence, 13 (3), 263-284. h ps://doi.o g/10.1023/a:1022841022524 No o, M., Fa iña, F., Seijo, D., & A ce, R. (2012). Assessmen o a communi y ehabili a ion p og amme in con ic ed male in ima e-pa ne iolen o ende s. In e na ional Jou nal o Clinical and Heal h Psychology, 12 (2), 219-234. h p://www.aepc.es/ijchp/a iculos_pd /ijchp-408.pd *Pé ez, M. Giménez-Salinas A., & Juan, M. (2012). E aluación del p og ama “Violencia de géne o: p og ama de in e ención pa a ag eso es”, en medidas al e na i as. Ins i u o de Ciencias Fo enses y de la Segu idad y Sec e a ía Gene al de Ins i uciones Peni encia ias. Mad id, Spain. h p://www.ins i ucionpeni encia ia.es/web/expo /si es/de aul / da os/desca gables/desca gas/VDG_EVALUACION_AUTONOMA_NIPO. pd Podsako , P. M., MacKenzie, S. B., Lee, J.-Y., & Podsako , N. P. (2003). Common me hod bias in beha io al esea ch: A c i ical e iew o he li e a u e and ecommended emedies. Jou nal o Applied Psychology, 88 (5), 879-903. h ps://doi.o g/10.1037/0021-9010.88.5.879 Redondo, S., Sánchez-Meca, J., & Ga ido, V. (2001). T ea men o o ende s and ecidi ism: Assessmen o he e ec i eness o p og ammes applied in Eu ope. Psychology in Spain, 5 (1), 47-62. h p://www. psychologyinspain.com/con en / ep in s/2001/6.pd Redondo, S., Sánchez-Meca, J., & Ga ido, V. (2002). Los p og amas psicológicos con delincuen es y su e ec i idad: la si uación eu opea [Psychological p og ammes wi h o ende s and hei e ec i eness: The Eu opean si ua ion]. Psico hema, 14 (Supl.), 164-173. h p://www. psico hema.com/pd /3487.pd Rosen eld, B. D. (1992). Cou o de ed ea men o spouse abuse. Clinical Psychology Re iew, 12 (2), 205-226. h ps://doi.o g/10.1016/0272- 7358(92)90115-o Rosnow, R. L., & Rosen hal, R. (1988). Focused es s o signi icance and e ec size es ima ion in counseling psychology. Jou nal o Counseling Psychology, 35 , 203-208. h ps://doi.o g/10.1037/0022-0167.35.2.203 *Saunde s, D. G. (1996). Feminis -cogni i e-beha io al and p ocess- psychodynamic ea men s o men who ba e : In e ac ion o abuse ai s and ea men models. Violence and Vic ims, 11 (4), 393-414. h ps://doi.o g/10.1891/0886-6708.11.4.393 Schmid F. L., & Hun e , J. E. (1981). Employmen es ing: Old heo ies and new esea ch indings. Ame ican Psychologis , 36 (10), 1128-1137. h ps://doi.o g/10.1037/0003-066X.36.10.1128 Schmid , F. L., & Hun e , J. E. (2015). Me hods o me a-analysis: Co ec ing e o s and bias in esea ch indings (3 h ed.). Sage. Schmucke , M., & Lösel, F. (2015). The e ec s o sexual o ende ea men on ecidi ism: An in e na ional me a-analysis o sound quali y