Stalking Risk Assessment: Evidence, Tools, and Safety Planning
Author: Ukrainian Psychological Hub · Published: October 11, 2026 · Editorial Policy
Stalking risk assessment is the structured evaluation of persistent unwanted pursuit, the possibility of further stalking or associated harms, and the measures that may interrupt those patterns. Its most useful questions are specific: What behavior has occurred, what might happen next, under which circumstances, to whom, and what can be done to reduce exposure? A well-founded assessment integrates chronology, corroborated facts, behavior over time, contextual circumstances, and the preferences of the person being targeted. Its conclusions remain revisable as new information arrives.
This article explains the evidence behind stalking risk assessment, the SAM, SRP, and SHARP tools, what predictive studies actually establish, the consequences of false positives and false negatives, and how risk formulation supports individualized safety planning. It separates psychology from legal determinations, documented conduct from speculation, and a professional judgment from a guaranteed forecast.
What is stalking risk assessment?
Stalking involves a pattern of unwanted conduct directed at a person. It may include surveillance, repeated messages, unwanted approaches, following, interference with daily life, threats, online monitoring, location tracking, and tactics carried out through other people. The Centers for Disease Control and Prevention's stalking definition describes patterned harassing or threatening tactics that cause fear or safety concerns. Context and repetition matter: apparently minor incidents can acquire significance when connected through time.
Stalking risk assessment uses information about this pattern to evaluate separately the likelihood of continued pursuit, renewed contact, escalation in intrusiveness, stalking of new targets, physical or sexual violence, threats to others, and serious disruption of housing, employment, or other life activities. These outcomes have different frequencies and cannot be combined into a single universal prediction. Every assessment should identify its outcome, time frame, target population, and reason for use.
Depending on the case, professionals may assess an identified suspect with access to records and interviews, or conduct an incident-based assessment with the targeted person and available documentation. Police, courts, forensic mental health professionals, victim advocates, schools, workplaces, and multidisciplinary threat-management teams have different responsibilities and legal powers. Their tasks overlap, but their conclusions are not interchangeable.
Stalking behavior, criminal liability, and clinical diagnosis
Stalking is a behavioral phenomenon; particular conduct may also satisfy a criminal statute. The legal elements depend on jurisdiction, including requirements concerning patterns, fear, distress, intent, and relevant territorial or interstate facts. The US Department of Justice offers a general federal description. The text of 18 U.S.C. § 2261A sets out specific federal offenses. State, tribal, and international provisions may differ.
An evaluator may identify behaviors, analyze risk markers, and formulate possible future scenarios. Only the appropriate legal decision-maker determines whether the conduct satisfies an applicable legal standard. Likewise, a psychiatric diagnosis is a separate clinical judgment. It neither establishes that stalking occurred nor proves future dangerousness; many stalking cases do not require a mental-disorder explanation at all.
Why stalking requires its own assessment framework
Stalking is typically a course of conduct unfolding over time. Repeated appearances, calls, digital account access, unwanted gifts, surveillance, indirect communication, and interference with a person's relationships may appear unrelated unless the investigator reconstructs the pattern. A chronology can show persistence after explicit boundaries, a change of contact channels after blocking, or a new attempt to gain physical access.
The Stalking Prevention, Awareness, and Resource Center organizes identification around surveillance, life invasion, intimidation, and interference through sabotage or attack. Its framework helps practitioners ask about behavior rather than relying on the victim's choice of terminology. It is an identification aid rather than an actuarial predictor.
General violence assessment may add useful background, but stalking persistence, confrontation, new-victim stalking, and serious violence remain different outcomes. A person may persistently harass without physical contact, or pose a meaningful confrontation concern despite a shorter observed history. The assessment should not collapse all scenarios into one unexplained low, medium, or high label.
What information should be collected?
A competent review constructs a dated behavioral record using the victim's account, messages, call logs, incident reports, witnesses, digital records, previous legal restrictions, workplace or campus documentation, and evidence from the person assessed when available. Each source has limitations. Practitioners distinguish direct observation, self-report, corroborated information, allegation, and reasonable inference instead of presenting them as equal proof.
Frequency, duration, persistence, and escalation
Important questions include how often contact occurred, how long the pattern has continued, which boundaries were communicated, whether efforts to avoid contact were circumvented, and whether tactics changed after an intervention. Escalation can concern frequency, intimacy of intrusion, surveillance, access to private locations, intimidation, or force. It need not follow a single predictable sequence.
Prior aggression, threats, and access to harm
Documented prior aggression, forced entry, explicit or implicit threats, weapon-related behavior, and breaches of restrictions warrant careful contextual scrutiny. A review and preliminary meta-analysis by Rosenfeld identified correlations between stalking-related violence and threats, prior violence, substance misuse, and former intimate relationships in the studies available at that time. These are group associations, and the source populations and definitions limit transfer to individual cases.
An absence of verbal threats is not evidence that physical harm is impossible. Conversely, ambiguous language and unverified allegations should not automatically become proof of imminent assault. Assessors examine the content of threats, access, preparation, timing, and alternatives supported by evidence.
Relationship context, separation, and opportunities
A former intimate partner may have access to shared finances, passwords, addresses, children, pets, or regular routines. A workplace, academic, or service relationship can provide recurrent access without prior intimacy. Strangers and online contacts may use other information pathways. The relevance of a relationship category depends on concrete opportunities and conduct, not stereotypes about a class of people.
Dynamic changes and protective conditions
Events such as separation, job loss, new legal proceedings, relocation, custody transitions, a change in access to the target, or new digital monitoring may alter risk. Their relevance depends on context and the individual's observable response. A restraining order, intervention, arrest, or blocked account can change access and accountability, but none guarantees that stalking has ended.
Protective factors can include genuine adherence to boundaries, effective monitoring and intervention, reduced access, stable support, engagement with relevant care, and practical safety resources. Their protective value is conditional and must be verified. Risk management requires reassessment after material new incidents.
The targeted person's perspective
The person experiencing stalking often knows details of repeated contact, contextual meaning, and possible access that case files omit. Their safety concerns deserve serious examination without treating emotional presentation as a truth test. Calmness does not disprove harm; distress does not by itself prove an allegation. Effective planning also considers disabilities, caregiving responsibilities, housing, employment, financial constraints, language, and access to services.
Stalking risk assessment tools compared
Risk instruments serve different purposes and users. Some are structured professional judgment frameworks designed for trained assessors. Others support situational identification and victim-centered planning. Reliability, discrimination, calibration, populations studied, and real-world usefulness should be evaluated separately. Neither a familiar brand name nor a numerical score establishes validity for every setting.
SAM: Guidelines for Stalking Assessment and Management
The SAM is a structured professional judgment approach for assessing stalking-related risks and guiding management. It helps organize risk factors, the relevance of those factors to a particular case, and potential interventions. Its initial development and validation research reported promising preliminary reliability and concurrent validity, while acknowledging file-based evidence limitations.
In a 2018 study of 146 adult referrals, consistency between raters improved when coding procedures were clarified. Reported AUCs were approximately .69 for case prioritization and .76 for continued-stalking judgments. These figures measure group discrimination in that sample; they are not a 69% or 76% probability that a named individual will stalk again.
A long-term study of 100 offenders found the combined presence of SAM factors associated with subsequent stalking-related investigations, charges, or convictions over a lengthy period, yet overall professional summary risk ratings did not significantly predict recidivism in that sample. The distinction between useful item-level information and the performance of an overall judgment is central to rigorous interpretation.
SRP: Stalking Risk Profile
The SRP is a structured approach to formulating stalking-specific risks and corresponding management needs. In a 2018 validation study of 241 specialist referrals, risk judgments showed performance that varied by outcome and observation period. Certain full-follow-up AUC values were around .66–.68, while short-term results were less consistent.
These studies do not establish that SAM and SRP are identical or that a professional may use either without specialized training. The setting, assessment date, coding standards, and available collateral information influence findings. Tools must be selected on the basis of the case's purpose and the evidence available for a comparable population.
SHARP: Stalking and Harassment Assessment & Risk Profile
SHARP is a web-based resource designed to help examine a stalking situation and inform safety discussion. SPARC's risk and safety page describes a 48-item assessment, a situational profile using 14 factors associated with several harms, and accompanying safety information. SPARC encourages victims and responders to collaborate when using it.
SHARP's accessibility and usefulness for organizing a narrative should be distinguished from independent evidence of predictive accuracy for severe harm in a particular population. A SHARP profile is not a diagnosis, a legally binding finding, or a validated percentage of imminent violence. A comprehensive forensic opinion may require methods and data that SHARP is not designed to replace.
General violence instruments and overlapping domains
Some forensic teams also consider broader violence risk tools or intimate-partner assessment frameworks. Their use depends on the specific referral question. Predicting violent recidivism, repeated stalking, domestic abuse, or lethal violence involves partly overlapping but distinct evidence bases. No general risk scale becomes a specialized stalking prediction rule merely because the same case involves both violence and pursuit.
A 2023 study comparing SAM, SRP, and HCR-20V3 followed 86 forensic psychiatric patients and found AUCs around .72–.73 for further stalking, with dynamic measures providing useful information over time. Results apply to the studied sample and outcome. They should not be generalized automatically to campuses, workplace incidents, online-only cases, or community complaints.
How accurate are stalking risk assessments?
Accuracy must be described with respect to a particular outcome and a relevant evaluation sample. Empirical studies provide meaningful, though bounded, support for structured stalking assessment. Their findings vary by tool, follow-up time, observer agreement, and definition of repeat conduct. The evidence does not support treating an assessment as a guaranteed prediction of a person's actions.
Predictive validity and AUC
The area under the receiver operating characteristic curve (AUC) describes how often a model assigns a higher risk rating to a case with the measured outcome than a case without it. An AUC of .50 is equivalent to chance discrimination and 1.00 is perfect discrimination for that dataset. An AUC of .76 therefore does not mean there is a 76% chance an individual will reoffend. Converting AUC into an individual probability is a serious interpretive error.
Calibration and base rates
Calibration concerns agreement between forecast probabilities and observed event frequencies. A tool may rank people reasonably well while producing inaccurate probabilities in a new setting. Many structured professional judgment frameworks do not provide empirically calibrated individual percentages. A person should not attach a seemingly precise percentage to a qualitative risk judgment without appropriate calibration evidence.
The base rate of the outcome changes the practical significance of positive findings. Serious violence is rarer than renewed unwanted messages in many settings. A screening strategy that flags many people can create numerous false positives when the severe outcome is infrequent, even when discrimination is better than chance. Positive and negative predictive values therefore require knowledge of outcome frequency and threshold selection.
False positives and false negatives
A false positive means a defined outcome is predicted or flagged but not observed during follow-up. It may lead to disproportionate restrictions, stigma, anxiety, or misplaced resources. A false negative means a person categorized as lower risk later engages in the measured behavior. It may leave someone exposed to avoidable harm. Each error has implications for different decision-makers, so reporting sensitivity, specificity, predictive values, and decision thresholds is preferable when those values are available.
Police reports are also imperfect measures of continued stalking. Nonreporting, lack of access to services, enforcement differences, and incomplete follow-up can conceal incidents. Conversely, an investigation, a charge, and a conviction are distinct outcomes. Studies combining these events should not be paraphrased as proven violent convictions.
Interrater reliability and reproducibility
Two assessors may disagree about whether a behavior occurred, whether stalking is still current, or how relevant a factor is. Shea and colleagues found that operational definitions affected SAM interrater consistency. Assessors should document disputed evidence, use explicit coding procedures, and recognize limitations when another trained practitioner might reasonably reach a different judgment.
Time-sensitive reassessment
A rating applies to circumstances at a defined time. New threats, contacts, access routes, separations, changes in custody, or digital intrusions can alter the assessment. Penney and colleagues' longitudinal research emphasizes changing risk indicators. A historic label must not replace investigation of current behavior or predefined triggers for reassessment.
Population limits, cultural fairness, and bias
Stalking tool validation has often used specialist or forensic psychiatric samples, sometimes relatively small, highly selected, and predominantly male. Performance in one setting is not automatically evidence for adolescents, all-gender populations, community complainants, minority groups, people with disabilities, or cases involving technology alone. Tools can also behave differently across languages and jurisdictions.
Measurement bias arises when opportunities to report, access to legal systems, police recording practices, or interpretation of communication differ across communities. Fairness demands attention to both the person assessed and the person seeking protection. Overall AUC values do not demonstrate equality of error rates, calibration, or access to support across subgroups.
The American Psychological Association's Specialty Guidelines for Forensic Psychology call for competent methods, corroboration, impartiality, and attention to individual differences. Where subgroup studies or local calibration are unavailable, that evidence gap belongs in the report. A label must never be elevated into certainty by concealing data limitations.
How professionals formulate scenarios and management plans
A useful formulation explains how observable behaviors interact with opportunities, vulnerabilities, and protective measures. It identifies plausible scenarios: continued online contact, unwanted appearance at a workplace, disruption of custody exchanges, physical confrontation, or renewed pursuit after intervention. For each scenario, it asks what might enable or interrupt the behavior and what evidence supports the inference.
Professional reports should identify the referral question, exact observed and alleged facts, source limitations, instruments and their relevant validation evidence, risk factors and counterevidence, scenarios, time frame, proportionate management options, responsible actors, and reassessment triggers. A high-risk label without such explanation offers little practical protection.
An assessor must also test reasonable competing explanations. An encounter in a shared workplace may be coincidental in one context and evidence of pursuit in another. File-only evaluations may lack the alleged stalker's account or independent corroboration. A transparent report acknowledges these limits instead of converting uncertainty into conviction or dismissal.
Safety planning for people experiencing stalking
A stalking safety plan is developed around the targeted person's circumstances, preferences, resources, and changing exposure. Qualified victim advocates can help evaluate alternatives and coordinate with police, schools, employers, legal professionals, or family. SPARC provides safety planning and documentation materials. Planning is an ongoing practical process, not a test that someone must pass to deserve help.
If an individual faces immediate danger, forced entry, an active assault, or a credible imminent threat, they should seek safe shelter and contact local emergency services. In the United States, call 911. The US Department of Justice's stalking resource also describes reporting and victim-assistance options. Other countries have different emergency numbers, laws, and support systems.
Identify contact points and immediate vulnerabilities
Review where the unwanted behavior has occurred and how access was obtained: home, work, school, transit, social networks, shared devices, childcare arrangements, or accounts controlled by another person. Consider likely upcoming events, including a court date, relocation, scheduled exchanges, or release from custody. Prioritize realistically changeable exposures without assuming the person can simply abandon employment, education, or family responsibilities.
Coordinate trusted people and urgent response
With the person's consent, identify trusted contacts and institutions that can help if an incident occurs. Agree on a simple code phrase or other signal, a safe place to go, who will contact emergency services, and how important details will be relayed. Share only the necessary personal information; broad public disclosures may create unintended privacy risks.
Preserve a reliable incident record
A dated log can record actions, locations, channels of contact, witnesses, and responses. Preserve screenshots with surrounding context, messages, voicemails, call histories, and relevant reports when it is safe and lawful. A secure backup can help prevent lost documentation. A targeted person should not be asked to confront, follow, communicate with, or investigate the alleged stalker in order to collect stronger evidence.
Review physical safety without one-size-fits-all demands
Possible measures include visitor controls, safer travel arrangements, secure doors, escorts, trusted pickup procedures, and coordination with workplace or campus security. Their value depends on actual circumstances, accessibility, and available resources. An advocate can help weigh whether routine changes improve safety or create additional hardship and exposure.
Take technology-facilitated stalking seriously
Location tracking, shared cloud accounts, password-reset access, monitoring applications, smart-home devices, and social media can create routes for surveillance. If monitoring is suspected, use a safer device and seek qualified technology-safety advice before changing settings. Disabling access, deleting spyware, blocking accounts, or replacing a phone can alert someone who is monitoring the device or destroy useful evidence. The National Domestic Violence Hotline's safety planning resource discusses these risks.
Technological safety can involve auditing shared calendars, app permissions, recovery contacts, vehicle or object trackers, connected cameras, workplace accounts, and shared family services. This review should be careful, lawful, and proportional. Technical certainty should not be claimed from unexplained battery drain or other nonspecific symptoms. When practical, trained digital safety personnel can document findings.
Consider legal and institutional options
Reporting, no-contact arrangements, protective orders, legal aid, security measures, and employer or school responses may help. Eligibility and procedures vary by place. Protective orders can create enforceable restrictions but do not guarantee compliance or substitute for a safety plan. Decisions about applying, communicating, documenting breaches, and pursuing enforcement should be made with local legal advice where needed.
Update the plan when circumstances change
Predefine triggers for review, such as repeated violations, new threats, movement from remote to in-person access, attempted entry, weapon-related behavior, threats toward children or pets, or sudden changes in location knowledge. Agree who receives updated information and how urgent assistance will be obtained. Plans should preserve autonomy and be workable for a person's day-to-day life.
Forensic ethics and legal application
Police, courts, correctional agencies, mental health services, and victim advocates each have a distinct role. Police investigate alleged offenses; courts apply legal standards; forensic evaluators address defined psycholegal questions; victim advocates prioritize safety and access to support. Information-sharing and confidentiality duties vary. A professional should state the scope of any opinion clearly.
The APA forensic guidelines emphasize scientific foundations, impartiality, corroboration, and disclosure of limits. If a report uses SAM or SRP, it should identify the version, training, population, outcome, assessment date, coding limitations, and basis for conclusions. A psychometric score does not prove guilt or substitute for statutory elements.
Criminal responsibility, competency to stand trial, civil protection, and prediction of future stalking answer distinct questions. A competent evaluator may provide evidence relevant to one, but legal findings belong to the authorized decision-maker under jurisdiction-specific standards. A risk score is not a criminal conviction, and the presence of mental illness does not make a person a criminal.
Common myths about stalking and risk
The claim that stalking always includes an explicit threat ignores patterned surveillance and repeated intrusive behavior. The claim that everyone who sends unwanted messages is a legally proven stalker ignores required statutory elements and context. The claim that a high tool score proves an impending assault confuses group discrimination with individual certainty.
Other myths include believing that mental illness explains stalking in general, that victims can simply block all contact, that a protective order always ends harassment, or that a single incident tells the entire story. Evidence-oriented practice reconstructs chronology, examines actual access and behavior, and adapts safety measures to the case.
Practical takeaways for different readers
For a forensic professional, the essential questions are the predicted outcome, relevant time horizon, accuracy evidence, population fit, information quality, and proportionate management options. A defensible assessment explains both what is known and what could change. The professional should avoid inventing individualized probabilities when only qualitative or group-level data exist.
For a person experiencing stalking, a risk category is a starting point for a concrete safety conversation. Ask what behavior was considered, what new incidents should trigger action, who can help, how digital and physical access can be reduced, and what support is feasible. A low-risk label never removes the right to seek assistance and report unwanted conduct.
For legal decision-makers, psychological research and assessment support informed judgment only within their tested scope. Allegations, observations, clinical findings, psychometric results, risk opinions, statutory elements, and judicial conclusions must remain distinguishable. Transparency about uncertainty increases the practical utility of the evidence.
Frequently asked questions
What is the best stalking risk assessment tool?
There is no universally best tool. SAM and SRP are structured professional judgment frameworks with validation studies in selected specialist populations. SHARP supports victim and responder understanding and safety discussion. The choice depends on purpose, training, population, and available evidence. None establishes guilt.
How accurate is the SAM?
Published studies report moderate discrimination for some stalking outcomes, with AUCs varying by rating, sample, and follow-up. One long-term study found that combined SAM factors were predictive while overall summary risk judgments were not significantly predictive. Reported performance cannot be converted into a precise probability for an individual.
Can stalking assessment predict homicide?
Evidence supporting reliable individual homicide prediction from stalking-specific instruments is limited. Rare severe outcomes pose particular base-rate and false-positive problems. Credible threats, access, and urgent safety needs still require attention even when no validated individual probability can be calculated.
Is SHARP a clinical or forensic diagnosis?
No. SHARP helps structure information about a situation and safety concerns. It is not a clinical diagnosis or legal decision, and its risk profile should not be read as proof of future conduct.
Can stalking occur entirely online?
Yes. Persistent unwanted surveillance, tracking, monitoring, repeated contact, and other technology-facilitated tactics may form a stalking pattern. Legal classification depends on specific facts and jurisdiction. The CDC includes digital stalking behaviors in its survey definition.
Does a psychiatric diagnosis make someone more likely to stalk?
A diagnosis alone does not establish stalking behavior or individual dangerousness. Where mental health information is relevant, professionals must examine the actual behavior, context, and needs rather than assigning a criminal motive from a label.
Will a protective order stop stalking?
An order may establish enforceable boundaries, but its existence cannot guarantee compliance. Its role depends on legal rules, enforcement, and an individualized plan. A victim advocate or attorney can help consider local options.
When should risk be reassessed?
Reassessment is appropriate after meaningful changes such as new contact, intensified intrusions, threats, legal developments, a change in access, or a failed intervention. Scheduled reviews should also reflect the urgency of the case and the responsible agency.
What should someone do if they suspect stalking?
If there is immediate danger, contact local emergency services. Otherwise, speak with a qualified local victim-service provider, preserve relevant evidence safely, review physical and digital access, and consider legal or police guidance. An individualized safety plan is preferable to an inflexible checklist.
Related articles
For the broader evidence behind group-level discrimination, prediction error, and forensic risk reports, see Violence Risk Assessment: What It Can Predict and Where It Fails.
For professional scope, ethics, and distinctions among psychological findings and legal determinations, see Forensic Psychology: What It Is and What Forensic Psychologists Do.
For a comparison with numerical decision-support methods used in criminal justice, see Algorithmic Risk Assessment in Criminal Justice: Accuracy, Bias, and Fairness and COMPAS Risk Scores: What They Are, Evidence, and Controversy. These separate articles examine algorithmic calibration, subgroup fairness, and legal decision contexts; stalking-specific tools are designed around different outcomes and evidence.
Related Articles
References
American Psychological Association. (2013). Specialty Guidelines for Forensic Psychology. American Psychologist, 68(1), 7–19. APA source.
Centers for Disease Control and Prevention. (2025). National Intimate Partner and Sexual Violence Survey: Frequently Asked Questions. CDC source.
Coupland, S. H., Storey, J. E., Kropp, P. R., & Hart, S. D. (2023). Forecasting stalking recidivism using the Guidelines for Stalking Assessment and Management (SAM). Assessment, 30(4), 1168–1181. https://doi.org/10.1177/10731911221086050.
Kropp, P. R., Hart, S. D., Lyon, D. R., & Storey, J. E. (2011). The development and validation of the guidelines for stalking assessment and management. Behavioral Sciences & the Law, 29(2), 302–316. https://doi.org/10.1002/bsl.978.
McEwan, T. E., Shea, D. E., Daffern, M., MacKenzie, R. D., Ogloff, J. R. P., & Mullen, P. E. (2018). The reliability and predictive validity of the Stalking Risk Profile. Assessment, 25(2), 259–276. https://doi.org/10.1177/1073191116653470.
National Domestic Violence Hotline. (n.d.). Stalking Safety Planning. Hotline source.
Penney, S. R., Ulrich, R., & Maheandiran, M. (2023). An examination of predictive validity and change in risk factors for stalking over time. Journal of the American Academy of Psychiatry and the Law, 51(3), 377–389. https://doi.org/10.29158/JAAPL.220110-22.
Rosenfeld, B. (2004). Violence risk factors in stalking and obsessional harassment: A review and preliminary meta-analysis. Criminal Justice and Behavior, 31(1), 9–36. https://doi.org/10.1177/0093854803259241.
Shea, D. E., McEwan, T. E., Strand, S. J. M., & Ogloff, J. R. P. (2018). The reliability and predictive validity of the Guidelines for Stalking Assessment and Management (SAM). Psychological Assessment, 30(11), 1409–1420. https://doi.org/10.1037/pas0000589.
Stalking Prevention, Awareness, and Resource Center. (n.d.). Identifying Stalking Behaviors. SPARC source.
Stalking Prevention, Awareness, and Resource Center. (n.d.). Risk & Safety. SPARC source.
United States Department of Justice, Office on Violence Against Women. (n.d.). Stalking. DOJ source.
United States Government Publishing Office. (2024). 18 U.S.C. § 2261A: Stalking. US Code source.