A lethality assessment is a structured screening tool used to evaluate how likely an abusive intimate partner is to kill or seriously harm a victim. It consists of a short set of questions about specific behaviors and circumstances, such as whether the abuser has access to a gun, has threatened to kill, or has choked the victim. Law enforcement officers, social workers, advocates, and healthcare providers use these assessments to identify victims at the highest risk of being murdered and to connect them with safety resources immediately.
How the Assessment Works
The most widely used version is the Danger Assessment (DA), developed by researcher Jacquelyn Campbell. It has two parts. The first asks the victim to look back over the past year using a calendar, marking the approximate dates when physical abuse happened and rating each incident on a 1-to-5 severity scale, from slapping or pushing with no injuries up to the use of a weapon that caused wounds. This gives a picture of how frequent and how severe the violence has been over time.
The second part is a series of 20 yes-or-no questions about specific risk factors tied to intimate partner homicide. These include whether the abuser has threatened to kill the victim, whether the abuser has access to a firearm, whether the abuser uses illegal drugs, whether stalking has occurred, and whether the victim has recently left or separated from the abuser. Other questions ask about threats to harm children, the abuser being unemployed, forced sexual contact, and whether a child in the home is not biologically related to the abuser. Each “yes” answer adds to the overall risk picture.
One question asks the victim directly: “Do you believe he is capable of killing you?” This single question, when asked after the most severe incident of abuse, correctly identifies a meaningful share of eventual homicide cases on its own. But it also misses some, which is why the full set of questions exists. Many victims underestimate their danger, and the structured format helps them see the cumulative weight of multiple risk factors they may have been minimizing individually.
Risk Factors That Carry the Most Weight
Not every warning sign raises the danger equally. A multisite study of intimate partner homicide cases identified several factors that dramatically increase risk. An abuser’s access to a gun increased the odds of the victim being killed by roughly 7.5 times. If the abuser used illegal drugs, the risk was nearly 5 times higher. Prior threats with a weapon raised the risk about 4 times, and direct threats to kill roughly tripled it.
Separation from a controlling partner stands out as one of the most dangerous moments. When a highly controlling abuser and the victim separated after living together, the risk of homicide increased roughly 9-fold. When the worst incident of abuse was triggered by the victim leaving, whether for another partner or for any other reason, the risk jumped 4 to 5 times. This is critical information because it means the period when a victim is trying to get safe is often the period of greatest danger, and it is exactly when a lethality assessment is most useful.
Interestingly, one factor was associated with lower risk: a prior arrest of the abuser for domestic violence cut the odds of homicide by about two-thirds. This suggests that criminal justice intervention, even when it doesn’t stop all abuse, may deter the most extreme violence.
What Happens When Someone Screens as High Risk
In many jurisdictions, lethality assessments trigger a specific protocol when a victim’s answers indicate high danger. Maryland’s Lethality Assessment Program (LAP), one of the most established models, lays out a clear sequence of steps for the professional conducting the screen.
First, the victim is told plainly that their situation matches the profile of cases where intimate partners have killed. This is intentionally direct because many victims do not realize how dangerous their circumstances are. Next, the professional calls a domestic violence hotline and asks the victim to speak with a trained advocate on the phone, right then and there. The goal is to remove barriers: instead of handing someone a phone number and hoping they call later, the connection happens in the moment.
If the victim agrees to talk, the advocate begins working on an immediate safety plan, which can include emergency shelter, transportation, childcare for counseling appointments, help with court appearances, and financial assistance. If the victim declines to speak with the hotline, the professional still calls to get guidance, relays all the safety planning information to the victim, provides the hotline number for future contact, and helps the victim understand which warning signs to watch for going forward. The protocol is designed so that even a victim who refuses help still walks away with more information and resources than they had before.
Where These Assessments Are Used
Lethality assessments were originally developed for use by domestic violence advocates and researchers, but they have expanded into law enforcement and healthcare settings. Police officers in many U.S. states now administer a shortened version of the screening at the scene of a domestic violence call. Emergency departments are often required by hospital accreditation standards to screen patients for intimate partner violence, though how effectively this happens varies. Nurses conducting triage may ask about abuse as part of regulatory requirements, but studies have found that frontline staff sometimes treat these as checkbox questions rather than genuine clinical inquiries. The quality of the screening depends heavily on training and on whether the setting allows for a private, unhurried conversation.
Child welfare workers also use lethality assessments. When a case involves a household with domestic violence, the assessment helps determine not only the risk to the adult victim but also the danger to children in the home. Several of the risk factors on the Danger Assessment, such as the abuser threatening to harm children or having a stepchild in the home, directly address this overlap.
Evidence That It Reduces Homicides
The strongest evidence for lethality assessment programs comes from studying jurisdictions that adopted them and comparing homicide rates before and after. A study examining Maryland’s LAP found that the program reduced female homicides committed by men by approximately 40%, translating to about 2.5 to 3 fewer deaths per million people annually. The effect was strongest in preventing homicides committed by boyfriends rather than spouses.
One important nuance: the program did not reduce nonfatal intimate partner violence. This suggests that lethality assessments work not by stopping abuse overall but by identifying and intervening in the most extreme cases, pulling victims out of the path of lethal violence specifically. The mechanism appears to be the immediate connection to services and safety planning rather than any change in the abuser’s behavior.
Limitations to Keep in Mind
No screening tool predicts the future with certainty. The Danger Assessment correctly flags many high-risk cases, but it also misses some. When victims were asked whether they believed their abuser was capable of killing them, that single question caught about 62% of actual homicide cases while correctly ruling out about 77% of non-lethal cases. The full 20-item tool improves on this, but no combination of questions will identify every case. Risk factors shift over time as circumstances change, so an assessment reflects a snapshot of danger at one moment.
The tool was developed and validated primarily for women in heterosexual relationships with male abusers. While the underlying risk factors, such as gun access, threats to kill, and separation from a controlling partner, are broadly relevant to intimate partner violence across demographics, the specific scoring and validation data come from studies of female victims of male violence. Practitioners working with male victims or in same-sex relationships typically still use the same risk factors as a guide, but the statistical calibration may not apply as precisely.