Violence against women is a global health problem of epidemic proportions (World Health Organization, 2013, 2021). It is estimated that around 35% of women suffer violence in their lifetime, being the most common violence type perpetrated by intimate partners, which affects approximately 30% of females worldwide (Devries et al., 2013). Thus, a woman is more likely to be injured, raped, or killed by her intimate partner than other people (World Health Organization, 2005; United Nations, 2006). There are different terms to refer to this violence, such as domestic violence, violence against women, intimate partner violence, marital violence, or wife assault. The denomination “partner violence against women” (PVW) is used in the current study. It refers to a physical, sexual, and psychological assault against women from their current or former men partners (Cunha and Goalves, 2016; Spencer and Stith, 2020).

Victimization has negative consequences for women's health, causing injuries and anxiety-stress responses that affect the gastrointestinal, cardiac, reproductive, and neurological systems (Wisner et al., 1999; Ruiz-Pérez et al., 2007; Ellsberg et al., 2008). Irreversible and permanent cessation of vital functions of the organism causing death is the most severe outcome of PVW (Snider et al., 2009). It affects 30 thousand women's lives annually, constituting approximately 40% of all homicides (Stöckl et al., 2013; United Nations Office on Drugs and Crime, 2018). The term “intimate partner femicide” (IPF) is used in this study when referring to severe violence resulting in women's deaths that are committed by their present or former intimate partners in heterosexual relationships (Campbell et al., 2009; Storey and Hart, 2014; Messing et al., 2015a; Messing and Campbell, 2016).

The high rates and severity of PVW highlight the importance of predicting future violence, based on scientific evidence, to manage this risk (Messing et al., 2020b). This prediction is possible through the development of risk assessment instruments that evaluate the danger in the violent relationship and take the information for aggressors' intervention and victim's safety planning (Campbell et al., 2009; Echeburúa et al., 2009). Numerous risk assessment instruments exist to assess different outcomes related to intimate partner violence as physical and sexual violence, recidivism or reassault, and homicide (Dutton and Kropp, 2000; Hanson et al., 2007; Kropp, 2008; Nicholls et al., 2013; Graham et al., 2019). Tools with the power to predict homicide are required to discern between potential lethal and non-lethal violence cases. Scientific evidence reveals that there are notable differences in victim, offender and situational characteristics that contribute to the probability of PPW escalating to IPF (Jung and Stewart, 2019; Overstreet et al., 2021). The increase in frequency and severity of violence, separation/divorce and kill threats are some of the main factors on which research in the field agree (Nicolaidis et al., 2003; Belfrage and Rying, 2004; Campbell et al., 2009; Dobash and Dobash, 2011; Kivivuori and Lehti, 2012; Vatnar and Bjørkly, 2013; Cunha and Goncalves, 2016; Johnson et al., 2020; Monckton Smith, 2020; Abrunhosa et al., 2021). It is the occurrence of certain elements that lead to IPF, not a simple progression of violence (Dobash et al., 2007). This information assists professionals to effectively manage the limited resources available by focusing their efforts on quickly, comprehensively, and effectively protecting those victims who are at high risk of being killed by their intimate partners (Storey and Hart, 2014).

Risk assessment instruments refer to tools that assist the “decision-making process through which we have to determine the best course of action by estimating, identifying, qualifying, or quantifying risk” (Nicholls, 2006). In this context, the risk is understood as the probability that an individual will engage in a certain kind of behavior in the future (Otto and Douglas, 2011; Fedock and Covington, 2019). Hence, specialized risk assessment tools designed for intimate partner homicide assess the risk of a lethal assault perpetrated by one partner against the other.

Most of the intimate partner homicide risk assessment instruments are applied regardless of the sex of both victim and aggressor (men to women, women to men, men to men, and women to women relationships) (Nicholls et al., 2013; Graham et al., 2019). PVW is an entirely different category of violence from other forms of intimate partner violence because it is the manifestation of a historical gender asymmetry and unequal power in relationships between men and women that led to domination and subordination (Russell and Harmes, 2001; De Jesus and da Silva, 2018). Men use violence as a demonstration to women that they have the authority in the relationship, having women controlled and subjected to their criteria (Anderson, 2005; González and Rodríguez-Planas, 2020). As violence, in this case, is different from those happening in other types of relationships, this phenomenon needs to be considered when using sensitive risk assessment instruments.

The inexistence of a gold standard in homicide risk assessment instruments for the diverse groups of intimate partner relationships could lead to errors in predictions (Nicholls et al., 2013). Deaths are predictable and preventable if adequate tools are used to target the population-based on factors known associated with it in each case (Johnson et al., 2019). Thus, for the prediction of IPF, it is essential to use tools that include risk factors items specific to female victimization by male aggressors in relationships. These might not coincide with the most convenient tools for intimate partner homicide in general. Summarizing sensible risk assessment instruments for IPF is a priority because of its high prevalence compared with homicide in other intimate partner relationships groups (Garcia et al., 2007; Messing et al., 2013). However, no studies focusing on this aspect have been identified to date. Graham et al. (2019) identified the need for future research to assess the reliability, validity, and feasibility of intimate partner violence and homicide risk assessment instruments in the diverse intimate relationship population.

Reliability and validity are common psychometric properties used to evaluate intimate partner homicide and reassault risk assessment instruments (Graham et al., 2019). Reliability refers to the reproducibility or consistency of measurement tools in obtaining the same results on repeated application to a person or group under similar circumstances (Cook and Beckman, 2006). The procedures most used to determine reliability are internal consistency reliability and interrater reliability (Heale and Twycross, 2015; Graham et al., 2019). The first refers to the degree to which the different items of an instrument perform together to measure a construct consistently, using the Cronbach's α commonly. In the area of violence risk assessment, Semahegn et al. (2019) indicated that the minimum acceptable value of the Cronbach's α is 0.7 (Nunnally, 1994; DeVellis, 2003; Kimberlin and Winterstein, 2008; Taber, 2018). Interrater reliability analyses consistency under agreement responses among multiple raters on different items of an instrument, using standard statistics of percentage agreement, interclass and intraclass correlation, Pearson's r, Spearman's p, and Cohen's k (Nunnally, 1994; DeVellis, 2003). Intraclass correlation coefficient is widely used and, in the field of violence risk assessment, Telles et al. (2009) mentioned that values from 0.4 to 0.6 are acceptable.

Validity refers to the degree of accuracy of an instrument in measuring the theoretical construct that it is intended to measure, revealing whether it can be used for its intended purpose (Kimberlin and Winterstein, 2008; Sampieri, 2018). There are different types of validity, including content, construct, and criterion validity. The content validity is concerned with the extent to which the substance of the instrument's elements is adequate to assess the specific domains that encompass the construct measured (Carmines and Zeller, 1979). The construct validity refers to how well an instrument reflects and measures a theoretical concept by determining the strength of correlation between the components of the instrument to know they are parts of the exact theoretical concept measurement and differ from other measures (Cronbach and Meehl, 1955). The criterion validity is based on a comparison of the instrument with another external criterion that measures the same construct through correlations' analysis of the results obtained in them (DeVellis, 2003). If all criteria types apply at the same time, the validity is said to be concurrent (Kaplan and Saccuzzo, 2013).

Predictive validity is a specific form of criterion validity, and it refers to the accuracy of an instrument for predicting a future criterion measure such as homicide (Messing et al., 2013, 2015a). This point is typically assessed in terms of sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), Receiver Operating Characteristic (ROC), and Area under the Curve (AUC). Sensitivity refers to the correct identification of cases expected to meet the predictive criterion, whereas specificity targets cases that are not expected to meet it. For instance, the sensitivity in risk assessment instruments for intimate partner homicide refers to the correct classification of lethal cases and the specificity to the correct classification of non-lethal cases (Parikh et al., 2008; Loinaz, 2017; Graham et al., 2019). PPV is the probability of cases that are expected to meet the criterion, and it occurs. NPV is the probability of cases not expected to meet the criterion, and it does not occur. Following the previous example, PPV corresponds to the probability of victims that are expected to be killed and indeed die, and NPV to the probability of victims that are not expected to be killed and do not die (Faller, 2005; Akobeng, 2007). The ROC is a graph that plots sensitivity as a function of 1-specificity obtaining the AUC. These provide information of predictive accuracy on a scale of 0 to 1. An AUC of 0.50 indicates an inability to predict. The closer to 1.0, the better the prediction accuracy (Messing and Thaller, 2013; Messing et al., 2013; Loinaz, 2017). For the field of violence risk assessment, Rice and Harris (2005) indicate that values between 0.6 and 0.7 are considered acceptable.

For the mentioned, the purpose of the current systematic review is to synthesize the scientific knowledge of risk assessment instruments used specifically for IPF, which aid in predicting cases in danger and, subsequently, the prevention of lethal results. Hence, the research questions are the following:

The reporting of the current systematic review was guided by the standards of the Preferred Reporting Items for Systematic Review and Meta-Analysis (PRISMA) Statement (Page et al., 2021). It details a process of identification, selection, appraisal, and synthetization of the studies to ensure a quality scientific review. Even though this guide was primarily used in the health framework, it has been applied to other areas of research related to intimate partner violence (Gerino et al., 2018; Velotti et al., 2018). A meta-analysis has not been developed because of the heterogeneity of statistical information available on the publications included in the current study, as the findings were not comparable.

The present systematic review established three research questions to synthesize the scientific knowledge of risk assessment instruments for IPF: What are the specific risk assessments instruments for IPF?, What are the risk factors for IPF included in the instruments? And what are the reliability and validity of the instruments? The findings respond to them and are discussed in the following paragraphs.

Addressing the first research question, the instruments for IPF are identified in different studies included in the systematic review. The quality search strategy allowed us to obtain fundamental studies to comprehensively understand these instruments. It is an innovative aspect in this review since no quality assessment of the strategy has been identified in other studies in the area. In the study selection process, a high number of them didn't meet the inclusion criteria and were excluded. In the field of risk assessment, it is not common to use specific keywords to refer to violence against women by current or former male partners, so we used general terms related to intimate partner violence and homicide across the different databases, proceeding from general partner violence to particular partner violence against a specific gender. In some studies included, different terms were detected to refer to PVW and IPF, such as “intimate partner violence against women” and “intimate partner femicide.” The greater integration of these terms in the studies on the matter is essential to facilitate the identification of these studies and promote more visibility of violence against women. In addition, the quality assessment of the included studies allows the analysis of the risk of bias, not being an judgment on how good or bad they are.

Regarding study characteristics results, research on risk assessment instruments for IPF over time with updates and cultural and contextual adaptations is observed. However, most of the studies have been carried out in the United States, making it difficult to extrapolate the findings equally to all parts of the world, hence the need for further research in other countries. In addition, the analysis of the studies' characteristics has also allowed identifying samples based mainly on victims' self-reports which are not officially corroborated. This information is crucial, but focusing exclusively on the victim is insufficient. A completion with other sources is essential as evidence reveals that there are victims with distorted perceptions of victimization (Patró Hernández et al., 2011; Storey and Hart, 2014). Analysis of official data is also required to include lethal cases to contribute to a greater number of risk assessment instruments for IPF and not just for attempted IPF, which are the majority. Regarding the sample, some of the included studies use a recurrent one. These studies used sample data collected for the National Institute of Justice-funded Oklahoma Lethality Assessment Study (OKLA) and Risk Assessment Validation study (RAVE) (Messing et al., 2013, 2015a, 2017, 2020a,b; Messing and Campbell, 2016). This data was used to develop and test different risk assessment instruments for IPF, which poses limitations regarding the obtained solid results, highlighting the need to carry out future studies to validate or refute them using diverse samples.

The findings also respond to the second research question "what are the risk factors for IPF included in the instruments?". The most frequent factors identified in the instruments are validated by several studies that demonstrated an association of them with the IPF. It reveals that they are essential to continue to be part of the instrument to predict the phenomenon. These are escalation of frequency and severity of violence (Nicolaidis et al., 2003; Kivivuori and Lehti, 2012; Vatnar and Bjørkly, 2013; Cunha and Goncalves, 2016; Johnson et al., 2020; Monckton Smith, 2020), sexual abuse (Bagwell-Gray, 2016; Dobash and Dobash, 2016), victim's perception that aggressor is capable of killing her (Nicolaidis et al., 2003; Vatnar and Bjørkly, 2013; Johnson et al., 2020), drug and alcohol problems (Belfrage and Rying, 2004; Kivivuori and Lehti, 2012; Cunha and Goalves, 2016; Dobash and Dobash, 2016; Johnson et al., 2020), battering during pregnancy (Decker et al., 2004), suicide threats or attempts (Belfrage and Rying, 2004), separation (Belfrage and Rying, 2004; Dobash and Dobash, 2011; Cunha and Goncalves, 2016; Abrunhosa et al., 2021), kill threats (Nicolaidis et al., 2003; Belfrage and Rying, 2004; Cunha and Goncalves, 2016), stalking (Nicolaidis et al., 2003; Johnson et al., 2020), aggressor control daily victim's activities (Decker et al., 2004; Dobash and Dobash, 2011; Bagwell-Gray, 2016; Monckton Smith, 2020), aggressor is violently and constantly jealous of victim (Nicolaidis et al., 2003; Dobash and Dobash, 2011; Bagwell-Gray, 2016; Johnson et al., 2020), and access, possession and use of weapons (Cunha and Goncalves, 2016; Reckdenwald et al., 2019; Johnson et al., 2020; Monckton Smith, 2020; Abrunhosa et al., 2021).

There are more factors supported by numerous studies that show their association with IPF, but they are included in a small number of instruments, contrary to the previous case. These are low economic income, unemployment, young women, stepchildren, no children, marital status, history of violence, criminal records, violence with injuries, history of mental health problems, lack of remorse, justifications, minority membership, immigration, non-foreign born aggressor, and non-lethal threats. Concerning the factor of low economic income, several studies reveal that it is an element associated with the IPF. This could explain why other elements already included in the instruments are risk factors as unemployment without economic benefits nor pension (Fernández-Montalvo and Echeburúa, 2005; Cunha and Goncalves, 2016). Furthermore, the evidence specifies that the lack of work by itself also has an impact on IPF perpetration due to being retired or responsible for the household chores are also associated with the deaths that should be considered for inclusion in the instruments (Kivivuori and Lehti, 2012; Sebire, 2017; Ward-Lasher et al., 2018). The factor women age between 15 and 34 years the evidence clarifies the range itself is not relevant, but the fact that the victim is younger than the aggressor (Cunha and Goalves, 2016; Sebire, 2017). Thus, modifying this item to this condition should be considered. The presence of children who are not biological offspring of the aggressor is also corroborated by scientific studies but has not yet been identified. This association with childlessness requires more research and needs to be explored (Sebire, 2017; Soria Verde et al., 2019). Regarding the marital status, it is a factor associated with the IPF. Also, serious partnerships without this status that are living together is another element to be considered for inclusion on the instruments (Dobash and Dobash, 2016). The evidence points at the history of interpersonal violence as a strong predictor of IPF. It could be evidenced in criminal records, and both are integrated into only a few instruments (Kivivuori and Lehti, 2012; Dobash and Dobash, 2016; Sebire, 2017; Soria Verde et al., 2019; Monckton Smith, 2020).

The use of violence resulting in injuries is a significant risk indicator to IPF when it is on the victim's face, head, or neck, which is not specified in the instruments (Reckdenwald et al., 2019). The factor history of mental health problems could be complete with the evidence, indicating the studies that the presence of any mental disorder in the aggressor is not associated with IPF. Specifically, anxiety, affective, psychotic, and personality disorders are related to it (Belfrage and Rying, 2004; Cunha and Goalves, 2016; Caman et al., 2022). Moreover, the suicide threats or attempts risk factor items commonly included in the tools are a manifestation of the affective disorder and personality disorder such as depressive disorder and borderline personality disorder (Belfrage and Rying, 2004; Caman et al., 2022). The above elements of suicide are related to the victim's danger of death, but also that of the aggressor as evidence indicates that a significant proportion of IPF cases are followed by the suicide of the offender (Vatnar et al., 2022). Simultaneous lack of empathy and remorse are factors associated with IPF (Dobash and Dobash, 2011, 2016). Although not included in the instruments, this evidence suggests the need for consideration in the future. Cognitive justifications to violence by aggressors are an important factor associated with IPF because they neutralize these acts (Fernández-Montalvo and Echeburúa, 2005; Dobash and Dobash, 2011). Immigration is an evidenced factor associated with IPF, and it could be interconnected with the other non-foreign born aggressor factor due to the link to lethal violence is greater when victim and aggressor are immigrants and come from the same ethnic background (Belfrage and Rying, 2004). Acculturation stress, welfare deficiencies and ethnic discrimination could be related to the involvement of immigrants with IPF (Vatnar et al., 2017). Ethnic minority membership is another factor related to immigration validated by scientific studies (Belfrage and Rying, 2004; Fernández-Montalvo and Echeburúa, 2005; Cunha and Goncalves, 2016; Sebire, 2017; Ward-Lasher et al., 2018). Death threats is another factor commonly repeated in the instruments, but it is not for threats against the victim to harm children and report to Child Protective Services, immigration, and other authorities. This is in the face of the victim's desire to denounce the victimization, seeks professional help, or separation/divorce, which are also included in the instruments with low frequency. More research is needed to validate or refute this inclusion to discriminate between lethal and non-lethal violence.

Evidenced factors associated with IPF and not included in the instrument have been identified. Specifically, regarding the aggressor these are protection order and prison sentences (Kivivuori and Lehti, 2012; Dobash and Dobash, 2016; Monckton Smith, 2020), childhood problems (Dobash and Dobash, 2016), distorted beliefs and rigid cognition of aggressor about possessiveness and control over women and fear of abandonment (Nicolaidis et al., 2003; Dobash and Dobash, 2016; Monckton Smith, 2020). Relative to victims there are isolation and submissive behaviors factors (Sebire, 2017; Monckton Smith, 2020). It would be convenient to consider incorporating these factors in the instruments and examining the improvement of psychometric properties with this inclusion. Nevertheless, not only the content and number of factors should be considered, but also the combination of these together to predict IPF. The study of Dawson and Piscitelli (2021) identified that the relationship between the number of factors and the risk of death is not necessarily linear. Different clusters of homicide have been defined and each of them has unique combinations of risk factors. The study of Gnisci and Pace (2016) reveals that IPF is a dynamic process and the sequence in which risk factors appear is important as well as the number and grouping of them. Some elements may be predictive of IPF by themselves, but when combined sequentially with others that occur, their predictive strength may be greater. Therefore, this heterogeneity must be taken into account for the prediction and prevention. Future studies could analyze in representative groups of PPW and IPF a wide variety of factors (related to victim, aggressor, relationship and environment) and consider how the combination of these and their occurrence over time increase or decrease the likelihood of homicide.In other words, not to focus attention on identifying specific factors are risk or protective but whether groups of factors are risk or protective. This would address the diversity and complexity of the phenomenon, leading to new risk assessment instruments and updates of current ones.

The knowledge of factors associated with IPF is essential for inclusion in risk assessment instruments. Their weights need to be pondered since not all have the same influence to predict homicides. Only one study of all included in the review has analyzed risk factor weights which is relative to the DA-I as mentioned in the results (Messing et al., 2013). There are certain studies that, although they do not directly display weights, could be extracted from the Odds Ratio (OR) and Relative Risk Ratio (RRR) scores reported statistically significantly. These are common measures of effect sizes that provide information on the probability that the homicide will occur, given particular factors. Factors with values of 1 indicate no effect on the homicide result, below one decreased risk and above one increased risk for this outcome in this (Schechtman, 2002; Andrade, 2015). The factors considered as low weight are partner control of women's life with OR ranged from 1.73 and 1.90 (Snider et al., 2009; López-Ossorio et al., 2021) and suicide ideas and attempts with RRR of 0.90 (Messing et al., 2020b) and OR of 1.994 (López-Ossorio et al., 2021). The last one could reduce the risk of IPF by a score lower than 1, being necessary to analyze it in future studies. There are two factors considered as medium weight, which is woman believe that man is capable of killing her by OR scores between 2.5 (Messing et al., 2017) and 5 (Snider et al., 2009) and battering during pregnancy by OR between 2.1 (Messing et al., 2017) and 3.4 (Snider et al., 2009). No high weight factors have been identified as there are controversies. In particular, escalation of violence has low weigh by an OR of 1.9 (Messing et al., 2017) but also high weight by an OR of 4.7 (Snider et al., 2009), violent and jealousy partner has a low weight by an OR ranged between 1.9 and 2.1 (Messing et al., 2017; López-Ossorio et al., 2021) and high weight by an OR of 5.5 (Snider et al., 2009), and strangulation has low weights by OR of 1.74 (Messing and Campbell, 2016) and 2.6 (Messing et al., 2017) and high weights by OR 4.1 (Snider et al., 2009). In this regard, it would be convenient to carry out studies on the weights of the mentioned factors and on those where this is not yet known. Moreover, it is crucial to also know the factors with RRR or OR scores lower than 1 to obtain information on elements that dampen the risk factors know up to know. This could improve homicide risk management by providing greater protection to cases with many risk factors and few protective factors.

The results also respond to the third research question, “what are the reliability and validity of the instruments?.” Some studies reported reliability, indicating that the instruments are 70% reliable and, consequently, 30% unreliable based on the consistency of the items for the outcome measured. This affirmation is extracted by a mean Cronbach's α score of 0.70 that is close to 1 (Campbell, 1986; Brown, 2002; Wang, 2015; Messing and Campbell, 2016; Messing et al., 2020b). The interrater reliability is only reported in one study, and it indicates good consistency of the instrument based on the degree of agreement on the items by an intraclass correlation coefficient of 0.83, which is close to 1 (Storey and Hart, 2014; Koo and Li, 2016). There is also one study that examined the construct validity, which indicates medium agreement with other similar measures due to a moderate positive correlation of 0.50 (Campbell, 1986; Adams et al., 2014; Chiu et al., 2015). The mentioned psychometric properties are not commonly revealed in studies included in the review, and future research should focus on these elements. The predictive validity is an exception due to these specific forms of criterion validity reported in most studies.

The balanced sensitivity and specificity scores of the instruments presented in the results do not always correspond to the prediction of IPF or attempted IPF but to specific cut-off scores and levels of danger. For instance, the updated version of the DA (Campbell et al., 2009; Storey and Hart, 2014) has a sensitivity of 0.55 and specificity of 0.97 for the extreme danger, while the balance is at the severe level with a sensitivity of 0.79 and specificity of 0.86. The consideration of extreme level is essential for predicting IPF or attempted IPF since it is closer to lethal outcome measured, although unbalanced. The same is applicable to the updated version of the DA-I (Messing et al., 2020b) which has a sensitivity of 0.29 and specificity of 0.969 for extreme levels compared to a sensitivity of 0.86 and specificity of 0.63 for severe levels. This data indicates underprediction contrary to the overprediction of the lethality screen (Messing et al., 2015a; Messing and Campbell, 2016) which has a high sensitivity of 0.93 and low specificity of 0.21. What is more, the DA-LE (Messing and Campbell, 2016; Messing et al., 2020b) has a balanced sensitivity of 0.60 and specificity of 0.70 approximately at the cut-off point of 7 over 11. It indicates low risk under that score and high risk above it, but as the cut-off increases, the sensitivity decreases and specificity increases significantly. For example, at a cut-off score of 9, the sensitivity is 0.24 and specificity is 0.90, and at 11, sensitivity is 0.04, and specificity is 0.98. The same is observed for the DA-5 (Snider et al., 2009), the updated version of the DA-5 (Messing et al., 2017), and the SIVIPAS (Echeburúa et al., 2009). These instruments could also lead to an underprediction of high-risk cases. In this regard, studies that assessed the PVW cases and then followed up on them to see whether the lethal or non-lethal result occurred report tool overprediction problems, whereas studies that analyzed know lethal and non-lethal PVW display underprediction problems (Campbell et al., 2009; Storey and Hart, 2014). It is necessary for future research to apply both types of studies to each of the instruments for IPF for stronger results.

The equilibrium of the mentioned elements is essential to discriminate high-risk cases from those that are not preventing a high number of false positives and false negatives. The first case makes it difficult to assist victims at risk because of the confusion generated by non-high-risk cases and the existence of limited resources, and the second case generates problems in the detection of high-risk cases and, therefore, in their prevention. The recently updated version of the DA (Messing et al., 2020a) overcomes this unbalanced limitation of the updated version of the DA (Campbell et al., 2009; Storey and Hart, 2014) due to an acceptable equilibrium in sensitivity and specificity for the extreme danger level. However, these measures should be tested again for the other mentioned studies.

In this section limitations of the included studies and future lines of research to address them have been presented. The process of systematic review also has several limitations that must be contemplated. First, it is possible to have additional studies related to risk assessment instruments for IPF, beyond those analyzed in the review. They have not been included because they are not in the databases used or didn't meet inclusion criteria such as not being in English or not being scientific articles. Second, the synthesis of the information from the studies has been carried out as neutrally and objectively possible, but there is the possibility that the interpretation of some results may not correspond with the original authors initially intended to transmit. The systematic review has also strengths. First, the systematic review has performed under the quality standards. The PRISMA guidelines (Page et al., 2021) was used allowing the course of the rigorous review process. Additionally, a quality assessment of the search strategy was conducted using an innovative text mining methodology LSA, and a quality assessment of the included studies through an adaptation of CASP, EPHPP, and CRD tools (Geraghty and Woodhams, 2015). Second, the review not only evidence the specific instruments for IPF but also detects their weaknesses and proposes future research to address them.

The systematic review provides an overview of evidence-based risk assessment instruments for IPF, which are the Danger Assessment, the Danger Assessment for Immigrants, the Danger Assessment for Law Enforcement, the Danger Assessment-5, the Taiwan Intimate Partner Violence Danger Assessment, the Severe Intimate Partner Risk Prediction Scale, The Lethality Screen, and the H-Scale. Content of risk factors items of IPF, validity, and reliability of these instruments are synthesized. Information on country, year, sample, and data sources in which the studies that support these instruments were conducted is also summarized. This comprehensible knowledge could assist professionals involved in PVW use tools within an evidence-practice framework to identify women victims at risk of near-lethal or lethal violence by their partners and respond with risk mitigation strategies at a time to prevent it. In addition, this review highlights research gaps to be considered in future studies on this field.

The original contributions presented in the study are included in the article/supplementary material, further inquiries can be directed to the corresponding author.

EG-V, NA, FF-N, and DB-A have designed the systematic study. They have involved to determine the aims, the method and results configuration. EG-V and NA have searched for studies on databases and collaborated for the peer review. EG-V has drafted the manuscript and prepared the figures and tables with the NA, FF-N, and DB-A assistance. NA, FF-N, and DB-A have been responsible for supervising the paper. All authors contributed to results interpretation and critical discussion.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.