OPINION
ANDERSON, Justice.Appellant Cedrick Scott Ince was civilly committed as a sexually dangerous person pursuant to Minn.Stat. § 25BB.185 (2012).1 The petition for civil commitment was filed 1 day before Ince’s release from the Minnesota correctional facility where he had served a sentence imposed following a guilty plea to third-degree criminal sexual conduct. Ince appealed his commitment, arguing that respondent Sibley County failed to prove by clear and convincing evidence that he is “likely to engage in acts of harmful sexual conduct” within the meaning of Minn.Stat. § 253B.02, subd. 18e(a)(3) (2012), and that the district court failed to adequately address the less restrictive alternative he presented, consistent with Minn.Stat. § 253B.185, subd. 1(d). In a divided, unpublished opinion, the court of appeals affirmed.
We granted Ince’s petition for review to consider whether the factors relevant to the determination that a person is “likely to engage in acts of harmful sexual conduct,” as set forth in In re Linehan (Linehan I), 518 N.W.2d 609 (Minn.1994), require clarification or modification, and whether Ince met his burden of proving the existence of a less restrictive alternative to commitment. After a thorough review of the record and our precedent, we now reverse the decision of the court of appeals, vacate the district court’s order for commitment, and remand to the district court for reconsideration.
I.
The appellant, Cedrick Ince, was 22 years old at the time of the commitment hearing and had twice been charged with criminal sexual conduct. The first occasion arose from events on February 11, 2007, while Ince, then 17 years old, was at a party. The victim, a 17-year-old known to Ince, passed out at the party after drinking. Ince, who was also drinking, removed the victim’s underwear and sexually penetrated her while she remained unconscious. Ince later gave several versions of the events, all of which attempted to diminish his role and culpability in the assault. Eventually, however, Ince admitted that the sexual contact was not consensual, that he “took things way too far,” and that the victim “was heavily intoxicated.” *18He pleaded guilty to fifth-degree criminal sexual conduct, was adjudicated delinquent, and was placed on probation.
On October 5, 2008, 3 weeks after he was placed on probation for the first offense, Ince broke into the house of a second acquaintance and raped her. The victim, who was 19 years old and had been in a brief relationship with Ince, testified that she was sleeping on the couch in her house and woke up to find Ince on top of her. When she struggled, Ince began choking her, making it difficult for her to breathe. After raping her, Ince told the victim he had a gun and would shoot her if she called the police. Ince again sought to minimize the circumstances of this violent assault, telling police that the sex was consensual and that the victim liked “weird stuff.” In later reports, however, Ince took responsibility, saying, “I was drinking and I broke into her house and raped her.... I forced it upon her ... she said no, but I didn’t listen.” Ince pleaded guilty to third-degree criminal sexual conduct and was sentenced to 48 months in prison with a 10-year conditional release term. While in prison, Ince, who has consistently been diagnosed with alcohol dependence, completed chemical dependency treatment, though not without some problems and setbacks. He was placed on a probation contract for rule violations, and he was terminated from aftercare. Ince did not enter or complete any sex offender programs in prison.
The petition for civil commitment was filed by Sibley County the day before Ince was scheduled to be released from prison. Ince was allowed to leave the facility subject to intensive supervised release from September 2011 until his commitment trial concluded in May 2012. During the 8 months that Ince was on supervised release, he obtained employment on a dairy farm, began renting a nearby house from his employer, and purchased a truck. Ince’s employer reported that Ince has “done an excellent job” and has an “ongoing, continuous” offer of employment. In early 2012, Ince began court-ordered sex offender treatment at an outpatient, community treatment program, although the record suggests that Ince’s disclosures to program staff were less than complete. Finally, the evidence was undisputed that Ince was considered in “sustained full remission” based on his alcohol and drug abstinence since 2009; had participated consistently in Alcoholics Anonymous (AA); and had supportive family relationships.
At the commitment hearing, the district court heard testimony from two court-appointed examiners, Dr. Penny Zwecker and Dr. Peter Marston, and from the County’s expert witness, Dr. Rosemary Linderman. The district court recognized that expert testimony carries “much weight” in civil commitment proceedings and found Dr. Marston’s testimony to be “particularly persuasive and convincing.” The court announced that it accepted Dr. Marston’s testimony and, unless otherwise noted, rejected the testimony of the other experts who did not agree with Dr. Mar-ston’s opinions.
Based on an interview with Ince, Dr. Marston concluded that after his release from prison, Ince expressed remorse and regret for his behavior, and concern for the victims of his offenses. Based on evaluations and risk assessments, Dr. Marston diagnosed Ince with antisocial personality disorder, a psychopathic personality disorder, alcohol dependence, and attention deficit hyperactivity disorder (ADHD), and concluded that as a result of those disorders, Ince had serious difficulty controlling his sexually harmful behavior.
Dr. Marston and the other experts attempted to determine, using risk assess*19ments based on actuarial tools, the likelihood that Ince will sexually reoffend in the future. Those assessments produced varied results, placing Ince at a medium risk for reoffending (using the RRASOR assessment); a high risk of reoffending (using the Static~99R assessment); and a high risk for reoffending compared to other sex offenders (using the MnSOST-3.1 assessment). Numeric probabilities also varied, showing Ince’s 5-year risk of reof-fending as 31.2 percent (Static-99R assessment); or, a 4-year probability of reof-fending as 7.92 percent (which is higher than 89.9 percent of sex offenders in Minnesota) (MnSOST-3.1 assessment); or, a “rule of thumb” suggested by one of the test’s developers, in which the 5-year esti- . mated rate of reoffending (31.2 percent) is doubled to establish a lifetime risk of reof-fending (62.4 percent). Dr. Marston also testified regarding the Linehan factors, concluding that Ince’s youth placed Ince at an increased risk of reoffending; that his offenses were serious, recent, and extremely severe; that base rate statistics indicate Ince is at a high risk of reoffend-ing compared to a typical sex offender; that Ince’s recent successes (employment, family support, and AA commitment) could be threatened by the stress associated with the commitment proceedings and his supervised release; and that, although Ince had begun sex offender treatment after his release, he did not have a relapse plan in place.
Dr. Marston acknowledged that controlling Ince’s alcoholism and ADHD would generally reduce the risk of reoffending. Further, he testified that Ince’s successes while supervised in the months following his release from prison represented a decrease in the risk of future harm to others. All three experts recommended, however, that Ince be committed to inpatient treatment with the Minnesota Sex Offender Program (MSOP).
Based on the evidence, the district court concluded that clear and convincing evidence demonstrated that Ince met the statutory requirements for commitment as a sexually dangerous person. The court, concluding that no less restrictive alternative was available, then ordered Ince committed to MSOP. The court of appeals affirmed, and we granted Ince’s petition for review.
II.
We begin with the standard for commitment as a sexually dangerous person, specifically the element of “likely” future harmful sexual conduct as a result of a personality disorder. Minn.Stat. § 253B.02, subd. 18c(a) (2012).2 Ince urges us to refine our previous construction of this provision, which requires the petitioner to show by clear and convincing evidence that future harmful conduct is “highly likely.” In re Linehan (Linehan III), 557 N.W.2d 171, 180 (Minn.1996), vacated sub nom. Linehan v. Minnesota, 522 U.S. 1011, 118 S.Ct. 596, 139 L.Ed.2d 486 (1997) (remanding for reconsideration in light of Kansas v. Hendricks, 521 U.S. 346, 117 S.Ct. 2072, 138 L.Ed.2d 501 (1997)). Ince argues that in showing that future harmful conduct is “highly likely,” the petitioner should be required to demonstrate it is “substantially certain” that *20future harmful sexual conduct will occur. The County opposes Ince’s proposed construction, noting that the current standard comports with constitutional mandates, and there is no compelling reason to overrule our precedent.
In order to commit someone as a “sexually dangerous person” (SDP), the district court must find by clear and convincing evidence that the person: (1) has engaged in a course of harmful sexual conduct; (2) has manifested a sexual, personality, or other mental disorder or dysfunction; and (3) as a result, is likely to engage in acts of harmful sexual conduct. Minn.Stat. §§ 253B.02, subd. 18c(a), 253B.18, subd. 1, 253B.185, subd. 1. The question of how to interpret the phrase “likely to engage in acts of harmful sexual conduct” as used in Minn.Stat. § 253B.02, subd. 18c(a)(3), is a matter of statutory interpretation, and thus a legal question. Hince v. O’Keefe, 632 N.W.2d 577, 582 (Minn.2001). Our review of this issue, therefore, is de novo. Id.
We have previously interpreted the term “likely” in the definition of “sexually dangerous person” to mean “highly likely.” Linehan III, 557 N.W.2d at 180. In Line-han III, we rejected the argument that “likely” means “more likely than not,” or more simply, at least a 50.1 percent probability. Id. We relied on the need for “accurate factual findings” and a “degree of overall certainty” in concluding that the Legislature would not have intended “to weaken the standard of likelihood in the SDP Act by combination with a relatively high burden of persuasion — the clear and convincing evidence standard.” Id. This holding was further developed by our decision in Linehan IV, in which we held that to commit a person as an SDP, the district court must find that it is “ ‘difficult, if not impossible, for the person to control his dangerous behavior.’ ” In re Linehan (Linehan IV), 594 N.W.2d 867, 875 (Minn.1999) (quoting Kansas v. Hendricks, 521 U.S. 346, 358, 117 S.Ct. 2072, 138 L.Ed.2d 501 (1997)).
In interpreting “likely” to mean “highly likely,” we also considered the due process concerns that constrain legislative discretion when imposing restraints on liberty. Linehan III, 557 N.W.2d at 180 (“[D]ue process concerns under the state and federal constitutions constrain legislative discretion to set standards of likelihood when liberty is at stake.”). Noting that due process requires evaluation of the risk of a “false prediction of future harmful conduct,” we said “ ‘[t]he individual should not be asked to share equally with society the risk of error when the possible injury to the individual is significantly greater than any possible harm to the state.’ ” Id. (quoting Addington v. Texas, 441 U.S. 418, 427, 99 S.Ct. 1804, 60 L.Ed.2d 323 (1979)). Therefore, we rejected an interpretation of “likely” that diminished the level of certainty and imposed too much of a risk of error on the individual. Id. “Highly likely” has thus operated as the statutory standard to show a person is “likely to engage in acts of harmful sexual conduct” since Linehan III. See Karl v. Uptown Drink, LLC, 835 N.W.2d 14, 17 (Minn.2013) (“Once we interpret a statute, our interpretation becomes part of the statute as though written therein.” (citation omitted) (internal quotation marks omitted)).
With this background, we turn to Inee’s proposed construction, which equates “likely” in Minn.Stat. § 253B.02, subd. 18c(a)(3), with “substantially certain.” Ince argues that constitutional considerations support a “substantially certain” standard to avoid inappropriately shifting the risk of error from society to the individual. While we agree that constitutional considerations support the construction of “likely” as “highly likely,” Li*21nehan III, 557 N.W.2d at 180, construing “likely” to mean “substantially certain” would be a departure from the ordinary usage of the term that is not required by the Constitution. Cf. Addington, 441 U.S. at 428, 99 S.Ct. 1804 (observing that because civil commitment does not implicate the same risks as in criminal cases, in which “the risk of error to the individual must be minimized even at the risk that some who are guilty might go free[,] ... [t]he full force of [avoiding error] does not apply to a civil commitment”).
Further, Ince’s construction cannot be accepted given statutory construction principles. See Minn.Stat. § 645.08(1) (2012) (instructing that “words and phrases are construed according to rules of grammar and according to their common and approved usage”). We acknowledge that one acceptable meaning of “likely” is “more probable ... than not.” See Black’s Law Dictionary 925 (6th ed.1990) (defining likely as being “of such nature or so circumstantial as to make something probable and having better chance of existing or occurring than not”). But the doctrine of stare decisis requires that we not overturn our previous construction, now “part of the statute as though written therein,” Caldas v. Affordable Granite & Stone, Inc., 820 N.W.2d 826, 836 (Minn.2012), without a compelling reason to do so. Oanes v. Allstate Ins. Co., 617 N.W.2d 401, 406 (Minn.2000). Furthermore, the distinct nature of commitment proceedings, with a focus on assessments and evaluations that require interpretation by experts, does not fit well with a demand for substantial certainty. See Addington, 441 U.S. at 429, 99 S.Ct. 1804 (“Given the lack of certainty and the fallibility of psychiatric diagnosis, there is a serious question as to whether a state could ever prove beyond a reasonable doubt that an individual is both mentally ill and likely to be dangerous.”).
We therefore reaffirm that the “highly likely” standard as enunciated in Linehan III ensures that the demands of due process are met. See generally Linehan III, 557 N.W.2d at 180 (“If the state were to require only a 10% probability of dangerousness (the fact to be demonstrated) and a clear and convincing evidence standard (say, a 75% degree of certainty), then the demand of due process that the citizen not share equally the risk of error would be undermined.”). The “highly likely” standard outlined in Linehan III also maintains the risk of error at a level that is permissible under due process for a civil proceeding. Id. (“[T]he error that due process seeks to avoid is a false prediction of future harmful conduct, and not only a prediction that is less accurate than the statutory standard.”).
We also conclude that “highly likely” cannot be defined by a numeric value. We recognize that the risk of error in commitment proceedings stems from the challenge of attempting to predict the future conduct of an individual. See Jones v. United States, 463 U.S. 354, 378, 103 S.Ct. 3043, 77 L.Ed.2d 694 (1983) (Brennan, J., dissenting) (“Although a substantial body of research suggests that a consistent pattern of violent behavior may, from a purely statistical standpoint, indicate a certain likelihood of further violence in the future, mere statistical validity is far from perfect for purposes of predicting which individuals will be dangerous.”). A highly likely standard, particularly when coupled with a clear-and-convincing-evidence standard, provides a necessary degree of confidence in that prediction. See Addington, 441 U.S. at 423, 99 S.Ct. 1804 (“The function of a standard of proof, as that concept is embodied in the Due Process Clause and in the realm of factfinding, is to ‘instruct *22the factfinder concerning the degree of confidence our society thinks he should have in the correctness of factual conclusions for a particular type of adjudication.’” (quoting In re Winship, 397 U.S. 358, 370, 90 S.Ct. 1068, 25 L.Ed.2d 368 (1970) (Harlan, J., concurring))); see also Heller v. Doe, 509 U.S. 312, 322, 113 S.Ct. 2637, 125 L.Ed.2d 257 (1993) (discussing how the burden of proof relates to the “risk of error faced by the subject of the proceeding”). Adopting a numeric value will not change the difficulties in predicting the likelihood of future behavior. Linehan III, 557 N.W.2d at 180 (noting that “committing courts cannot combine a factual element that requires only 50.1% probability -with an evidentiary standard of less-than-certainty” while still retaining a degree of overall certainty that is consistent with the term “likely”).
Thus, we reaffirm our previous construction of “likely,” as used in Minn.Stat. § 253B.02, subd. 18c(a)(3), to require clear and convincing evidence that the person is “highly likely” to engage in acts of harmful sexual conduct. The constitutionality of the highly likely standard is long settled, see Linehan IV, 594 N.W.2d at 872-76, and we see no compelling reason to undo this longstanding precedent.
III.
We now turn to the evidence relevant to the district court’s determination of whether Ince is highly likely to engage in acts of harmful sexual conduct. We review the district court’s factual findings under a clear error standard to determine whether they are supported by the record as a whole. See In re Joelson, 385 N.W.2d 810, 811 (Minn.1986).
The factors used to evaluate the likelihood of future harmful sexual conduct were first enumerated in Linehan I, as follows:
(a) the person’s relevant demographic characteristics (e.g., age, education, etc.); (b) the person’s history of violent behavior (paying particular attention to recen-cy, severity, and frequency of violent acts); (c) the base rate statistics for violent behavior among individuals of this person’s background (e.g., data showing the rate at which rapists recidi-vate, the correlation between age and criminal sexual activity, etc.); (d) the sources of stress in the environment (cognitive and affective factors which indicate that the person may be predisposed to cope with stress in a violent or nonviolent manner); (e) the similarity of the present or future context to those contexts in which the person has used violence in the past; and (f) the person’s record with respect to sex therapy programs.
Linehan I, 518 N.W.2d at 614.
Ince argues that the Linehan factors remain valid but have been displaced by more recent emphasis on actuarial assessments. This shift in focus, he concludes, diminishes the role and significance of dynamic factors; improperly permits “double counting” of some factors, such as age and criminal history; and ultimately deprives the commitment process of an individualized determination. The County agrees that a “multi-factor” analysis best addresses the complex issues that surround a commitment determination. But the County disagrees that an overemphasis on actuarial assessment tools has led to the problems Ince identifies.3
*23We acknowledge that the commit ment determination is made neither lightly nor easily. It is a “difficult task” often requiring consideration of a “voluminous and complex” record followed by a “careful balancing of all the relevant facts.” Linehan III, 557 N.W.2d at 191. We have also recognized that “dangerousness prediction methodology is complex and contested.” Id. at 189. Given these challenges, we have declined to confine the inquiry for predicting whether an individual is highly likely to engage in harmful sexual conduct, and we have not foreclosed the district courts from considering all evidence relevant to that determination. See, e.g., Minn. R. Civ. Commit. P. 15 (“The court may admit all relevant, reliable evidence ... without requiring foundation witnesses.”); Linehan III, 557 N.W.2d at 178, 189 (stating the district court used a “mul-ti-factor analysis” including statistics and “other factors it believed were sound indicators of future conduct,” and concluding the court did not err in “considering] evidence not specifically listed in Linehan I ”); In re Blodgett, 510 N.W.2d 910, 915 (Minn.1994) (noting that district court could consider, under predecessor test, “other factors that bear on the predatory sex impulse and the lack of power to control it”).
Indeed, the need for a multi-factor analysis lies in the very purpose for civil commitment. We have acknowledged that “mere dangerousness [is] not sufficient to justify civil commitment,” but “dangerousness coupled with proof of an additional statutory factor such as mental illness or personality disorder” may support that commitment. Linehan IV, 594 N.W.2d at 872-78; see also Linehan III, 557 N.W.2d at 184 (noting that “[m]ere social malad-justments might not satisfy” due process, and “the government’s ability to commit those whom it fears” is limited). The distinction between “dangerousness” and “dangerousness coupled with” serves “to limit involuntary civil confinement to those who suffer from a volitional impairment rendering them dangerous beyond their control.” Linehan IV, 594 N.W.2d at 873 (citation omitted) (internal quotation marks omitted). The State’s legitimate interests in treatment and public protection, Linehan III, 557 N.W.2d at 186-87, are furthered by consideration of all evidence relevant to the civil commitment decision.
But we remind the district courts that the “dangerousness prediction” is neither “a purely ‘clinical’ prediction” nor “simply a matter for statisticians.” Linehan III, 557 N.W.2d at 189, 191. Rather, with the benefit of all the relevant and rehable evidence, the district court must make a “good faith attempt[ ] ... to isolate the most important factors in predicting harmful sexual conduct.” Id. at 189. As the trier of fact, the district court will be in the best position to determine *24the weight to be attributed to each factor, as well as to evaluate the credibility of witnesses-a critical function in these cases that rely so heavily on the opinions of experts. In re Knops, 536 N.W.2d 616, 620 (Minn.1995); see also State v. Al-Naseer, 788 N.W.2d 469, 478 (Minn.2010) (“We recognize that the trier of fact is in the best position to determine credibility and weigh the evidence.”).
We agree that actuarial assessment evidence is relevant to the determination of whether a person is highly likely to engage in future1 harmful sexual conduct. We long ago endorsed a “multi-factor analysis for dangerousness prediction,” Linehan III, 557 N.W.2d at 189, and we see no logic to confining the scope of the inquiry on this issue. The necessity for accuracy — or put differently, for avoiding mechanical and arbitrary results — demands a thorough consideration of all relevant evidence. We caution the district courts, however, to be wary of the potential factor repetition that can result from considering the Linehan factors in addition to multiple actuarial assessments that use different approaches based on factors that are the same as or similar to the Linehan factors. We do not believe it is necessary to delineate here examples in which that repetition can occur. Rather, recognizing that the potential for duplication exists, we rely on the ability of district courts to weigh the evidence in each case, drawing the appropriate conclusions based on consideration of all the evidence. See Al-Naseer, 788 N.W.2d at 473.
We also do not consider it necessary to limit the time frame for assessing the likelihood of future harmful sexual conduct, as Ince suggests. Ince argues that the time frame should be limited due to the unreliability of long-term predictions. We have previously stated that “the [SDP] Act does not limit the prediction by time period.” Linehan III, 557 N.W.2d at 190. We will not add words to a statute that the Legislature has “purposely omitted or inadvertently overlooked.” Premier Bank v. Becker Dev., LLC, 785 N.W.2d 753, 760 (Minn.2010). The district court is free to determine the weight to be attributed to any particular piece of evidence, including predictions of future short- or long-term recidivism rates, based on the record in an individual case.
The only remaining issue is whether the district court’s assessment of the evidence in this case was consistent with the framework we have set out above. Recognizing that the district court is in the best position to weigh the evidence and assess credibility, we nonetheless conclude that a remand is necessary here. While we are sympathetic to the challenges the district court faced, we cannot discern from the court’s findings and conclusions the extent to which the court considered itself constrained by the Linehan factors once it addressed the actuarial assessment evidence. The court simply reviewed the Li-nehan factors after largely accepting Dr. Marston’s opinions on the actuarial evidence, but without indicating the significance of any of those factors within the context of a multi-factor analysis.
Similarly, the experts’ opinions reach mixed, if not contradictory, results based on the actuarial evidence as compared to the Linehan factors. For example, Dr. Marston concluded based on the Linehan factors that Ince’s young age placed him at an increased risk of reoffending, but based on other assessment tools including the Static-99 assessment, agreed that “the first year might be regarded as the most likely year to reoffend” and, significantly, that Ince’s sustained compliance after his release from prison reduced the risk of future harm. Dr. Marston also agreed that Ince’s risk of reoffending would de*25crease as he aged, because his antisocial personality disorder would abate. Yet applying a “rule of thumb,” Dr. Marston calculated a lifetime recidivism rate that exceeded 60 percent. Likewise, Dr. Mar-ston concluded that Ince’s personality disorder, ADHD, and alcoholism made it difficult for him to control his sexual impulses, but also admitted that controlling Ince’s alcoholism and ADHD would reduce Ince’s risk of reoffending.
We do not intend these observations to suggest that the district court’s findings and conclusions lack any support in the record. Rather, these observations illustrate the difficult task the court faced in this unique case, in which the experts agreed that the picture was complicated by, among other things, Ince’s sustained period of community living, compliance with supervised release conditions, or work and family successes, as well as his alcoholism and mental health conditions. Because we cannot determine whether the district court adhered to the Linehan factors after considering the other evidence, we conclude that a remand is necessary so the court can comprehensively consider all relevant, reliable evidence and make findings in light of our directions above.
IV.
The final issue we consider is whether the record supports the district court’s conclusion that Ince did not prove that a less restrictive alternative to commitment in MSOP would suffice. Ince presented evidence that his treatment needs could be met with ongoing therapy, intensive supervision through the Department of Corrections, and treatment in the outpatient community program. Ince also argued that, based on his successful transition from prison to community living under intensive supervision, he can remain in the community without posing a threat to public safety. Both the district court and the court of appeals disagreed.
If the conditions for civil commitment as a sexually dangerous person are met, the district court “shall commit the patient to a secure treatment facility unless the patient establishes by clear and convincing evidence that a less restrictive treatment program is available that is consistent with the patient’s treatment needs and the requirements of public safety.” MinmStat. § 25BB.185, subd. 1(d). Thus, by statute, the burden of proving that a less restrictive alternative exists rests on Ince.
We have not previously addressed the necessity or nature of findings of fact on the availability of a less restrictive alternative.4 A constellation of competing concerns are posed by the choice between commitment “to a secure treatment facility” or to a “less restrictive treatment program,” Minn.Stat. § 25BB.185, subd. 1(d), most obviously an individual’s liberty interests and the State’s interest in public safety. Linehan IV, 594 N.W.2d at 872 (explaining that “freedom from physical restraint has always been at the core of the liberty protected by the Due Process Clause” and the State has “a compelling interest in ... *26protecting the public from sexual violence” (citation omitted) (internal quotation marks omitted)). Although there is no statutory definition for a “less restrictive treatment program,” the district court must consider a proposed less restrictive alternative in light of the objectives of commitment: “the patient’s treatment needs and the requirements of public safety.” Minn.Stat. § 253B.185, subd. 1(d). Particularly given the difficulty of the commitment determination, see Linehan III, 557 N.W.2d at 191 (noting that “[d]angerousness prediction is a difficult task that the legislature has delegated to the district courts”), findings of fact on a proposed less restrictive alternative will no doubt contribute to meaningful appellate review and confidence in the decision. See e.g., Minn. R. Civ. P. 52.01 (stating that in actions tried upon facts without a jury, “the court shall find the facts specially and state separately its conclusions of law”); Rosenfeld v. Rosenfeld, 311 Minn. 76, 82, 249 N.W.2d 168, 171 (1976) (stating that findings of fact will “assure consideration of the statutory factors ... facilitate appellate review ... [and] satisfy the parties that this important decision was carefully and fairly considered”).
The district court concluded that “no less restrictive treatment ■ program” existed that could meet Ince’s need for “inpatient treatment,” and that MSOP was capable of meeting that need. The evidence in the record may be sufficient to support this conclusion, but in the absence of findings of fact about that evidence, we cannot be certain. Particularly because the unusual nature of the facts and circumstances here require a remand for reconsideration in light of what we have said about the factors for determining whether Ince is highly likely to reoffend, the determination of a possible less restrictive alternative must be revisited as well.5 Cf. Hurt v. Johnston, 242 Minn. 329, 337, 65 N.W.2d 193, 198 (1954) (holding that a “new trial on all the issues here [would] be in the interests of justice” because “[b]oth sides are entitled to a fair, impartial, conscientious consideration of all the evidence”).
V.
In summary, we reaffirm the principles first announced in the Linehan decisions: the risk of harmful sexual conduct occurring must be “highly likely” based on consideration of the Linehan factors and all relevant evidence. Given our decision today, reconsideration of the unique and specific facts of Ince’s case is warranted. Therefore, we reverse the court of appeals, vacate the district court’s order for commitment, and remand to the district court for further proceedings consistent with this opinion.
Reversed and remanded.
. In 2013, the Legislature amended portions of chapter 253 B (2012) to recodify in chapter 253D (Supp.2013) certain provisions related to civil commitment as a sexually dangerous person or sexual psychopathic personality. Act of May 9, 2013, ch. 49, § 22, 2013 Minn. Laws 210, 229-31. Because Ince's commitment proceedings were conducted under chapter 253B, we refer only to the relevant pre-2013 sections of chapter 253B in this opinion.
. Commitment also requires clear and convincing evidence that the person has engaged in a "course of harmful sexual conduct” and has a personality or other disorder or dysfunction. Minn.Stat. § 253B.02, subd. 18c(a)(l)-(2); see Minn.Stat. §§ 253B.18, subd. 1 (2012) (using "clear and convincing” standard for commitment), 253B.185, subd. 1 (adopting standard used in section 253B.18 (2012)). Ince does not challenge the district court’s findings on these requirements, and we therefore do not consider the legal standards or the evidence related to these requirements.
. The third Linehan factor, "base rate statistics,” does not mention the more individualized predictions that can be drawn from actuarial tools. Linehan I, 518 N.W.2d at 614. The term "base rate statistics” refers to recidivism rates for a particular class, such as the *23overall percentage of sex offenders that reof-fend. Id. (describing base rate statistics as the statistics "for violent behavior among individuals of this person’s background (e.g., data showing the rate at which rapists recidivate, the correlation between age and criminal sexual activity, etc.)”). In contrast, actuarial statistics are a relatively new tool developed from risk assessment models used in the insurance context. See Eric S. Janus & Robert A. Prentky, Forensic Use of Actuarial Risk Assessment with Sex Offenders: Accuracy, Admissibility and Accountability, 40 Am.Crim. L.Rev. 1443, 1454 (2003). In actuarial evidence, risk factors are identified and the overall risk for an individual, rather than the risk for a class, is computed based on the number of risk factors the individual presents. Linehan III, 557 N.W.2d at 189 n. 14 ("Actuarial predictions are based on statistics that can be determined mathematically (e.g., age and the number of previous offenses), and on a formula for evaluating the significance of such variables.”).
. In In re Senty-Haugen, 583 N.W.2d 266 (Minn.1998), we held that, in the absence of a statutory requirement for commitment to the least restrictive alternative, there was no error in failing to make findings of fact on the existence of such an alternative. Id. at 269. Following this decision, the Legislature amended Minn.Stat. § 253B.18, subd. 1, to require commitment to a secure treatment facility unless the evidence establishes "that a less restrictive treatment program is available that is consistent with the patient’s treatment needs and the requirements of public safety.” Act of May 4, 1999, ch. 118, § 3, 1999 Minn. Laws 481, 482.
. Because we remand to the district court for reconsideration, we do not address Ince’s argument that he met his burden of proof to show the existence of a less restrictive alternative. We leave to the district court the issue of whether to reopen the record to permit additional testimony on any of the issues presented here. But we note that, given the liberty interest at stake, the conclusory nature of the district court’s finding that no less restrictive alternative existed, and, most importantly, the highly unusual fact pattern presented here, additional evidence and testimony, expert and otherwise, may well be warranted.