In re the Detention of Ritter

*495[As amended by order of the Court of Appeals April 12, 2016.]

Korsmo, J.

¶1 — After remanding for a hearing following our initial consideration of this appeal, we now consider Steven Ritter’s challenges to the jury’s decision to commit him as a sexually violent predator. In the published portion of this opinion, we address his challenge to the dynamic risk assessment tool used at trial. We affirm.

FACTS

¶2 The salient facts in this appeal largely concern procedural matters. Additional facts related to the issues considered in the unpublished portion of this opinion will be addressed in conjunction with those arguments.

¶3 Mr. Ritter, at age 15, sexually assaulted his 46-year-old aunt. He spent about 30 months in juvenile sex offender treatment in Oklahoma and was released at age 18. Within the year, he molested a 9-year-old girl at a public library in Yakima. He was convicted of that offense and served his sentence at the Twin Rivers facility in Monroe. There were additional uncharged incidents of sexual misconduct as a juvenile that were admitted into evidence at trial.

¶4 When his sentence was drawing to a close, the State had Mr. Ritter evaluated by Dr. Dale Arnold. Dr. Arnold applied three actuarial instruments to Mr. Ritter’s static risk factors and his own clinical judgment to Mr. Ritter’s dynamic risk factors. Dr. Arnold concluded in written reports in 2006 and 2009 that Mr. Ritter met the criteria of a sexually violent predator (SVP). In late 2011, after the State had filed SVP proceedings against Mr. Ritter, Dr. Arnold revised his reports to apply the forensic version of *496the Structured Risk Assessment—Forensic Version (SRA-FV) to Mr. Ritter’s dynamic factors.

¶5 Mr. Ritter unsuccessfully tried to exclude use of the SRA-FV and two of the static instruments at trial. After he was committed by the jury, Mr. Ritter timely appealed to this court. His appeal raised four issues, including a challenge to the use of the SRA-FV. We exercised our authority to remand for a Frye1 hearing on that issue. In re Det. of Ritter, 177 Wn. App. 519, 520-21, 312 P.3d 723 (2013).

¶6 Both sides presented expert testimony at the remand hearing. The State presented the testimony of Dr. Amy Phenix to establish the inception and validity of the SRA-FV. The defense presented two experts: a statistician, Dr. Dale Glaser, and a psychologist, Dr. Brian Abbott. The basics of forensic testing were not in dispute. The first step in analyzing a sexual offender’s risk of future reconviction is to score that person on one or more of several actuarial instruments. These are widely used, validated, and well established since at least 1998. They look at the presence or absence of various static factors that affect the risk of sexual reoffense. These static factors are immutable, and consist primarily of facts about the offender and the offense committed, such as the number of offenses and the sex of the victim(s).

¶7 The static factors were established individually by various studies2 looking at populations of sex offenders that were released from prison, and then correlating reoffense with the presence or absence of the various factors. In 1998, Dr. Karl Hanson published a meta-analytic study, compiling all the existing studies into a cohesive, single framework. This gave rise to the Static-99 actuarial instrument. Subsequent studies and analysis have further refined the factors and given rise to several newer instruments that *497may incorporate additional factors or structure the analysis differently. All of these instruments have moderate predictive accuracy; employing additional instruments incrementally increases that accuracy.

¶8 Because an analysis based only on static risk factors will never change, the psychological community began looking for dynamic factors that could be used both to refine the risk analysis and help guide treatment. In 2002-2003, Drs. Thornton and Beecham published a series of analytical papers that served as a methodological foundation for the SRA-FV. They looked at each dynamic factor as falling into one of four constructs: sexual interest, relational style, self-management, and attitudes.3 They posited that in order to have any degree of accuracy, a comprehensive analysis would need to examine at least three of those constructs. They then developed the SRA-FV to examine the first three constructs.4

¶9 In 2010, a meta-analytic study was published on the research into dynamic risk factors comparable to the 1998 study and provided the statistical basis for developing an instrument based on those dynamic factors. The SRA-FV was released to the psychological community for use that same year, essentially providing a structured application of the meta-analysis. Subsequently, in 2013, Dr. Thornton published a peer-reviewed article establishing the development and validity of the SRA-FV.

¶10 A professional administering the SRA-FV looks to their diagnostic interactions with the individual and to facts available in that person’s record, and then scores each dynamic risk factor against an operational guideline, from 0 to 2: 0—the factor is absent; 1—the factor is present; 2—the factor is strongly present. Those factors are then weighted *498and summed to arrive at three domain scores, corresponding to those three constructs the instrument is assessing. Higher overall scores on each domain correspond to a higher absolute probability of reoffense. However, the SRA-FV does not return any actual probability of reoffense, but is instead used in conjunction with the Static-99R.

¶11 Because the statistical data underpinning the Static-99 was derived from many different studies, those studies were amalgamated in order to create a large population base. However, different data sets involve different types of people. Consequently, as the Static-99 was refined, the instrument was adjusted to account for the varying inherent recidivism rates in the studied populations by separating the studies into several normative groups. Under the revised Static-99R, the examiner must score the static risk factors, then compare that score against one of the normative groups to arrive at a probability that the offender will be convicted of a future sex crime.5 The SRA-FV is used to sort the individual into one of those normative groups.

¶12 The SRA-FV was constructed from a sample obtained from the Massachusetts Hospital in Bridgewater6 and then cross validated on a separate sample from that same hospital. Trial testimony showed there is some criticism in the psychological community that the dated sample might not correlate with a modern sample. However, contemporary samples employed in comparable instruments, the Stable-2007 and the VRS-SO (Violence Risk Scale-Sexual Offender), suggest that the sample should be accurate. Of note, the SRA-FV sample is the only sample set that includes long-term, incarcerated offenders rather than people in the community. Employing the SRA-FV in con*499junction with the Static-99R leads to an incremental increase in predictive accuracy from 0.68 to 0.74.

¶13 In addition to the Bridgewater sample issue, the SRA-FV was criticized for its lack of construct validity and low inter-rater reliability. All of these were stated limitations in the peer-reviewed article. First, construct validity has not been established for any of the particular dynamic risk factor ratings employed by the SRA-FV. Construct validity refers to a measure of whether a psychometric test measures what it claims to measure. In the context of the SRA-FV, the question is whether the assessment of the particular risk factors and composite constructs actually measures what they purport to measure. The concern is that the mechanisms for measuring the dynamic factors are not identical between the SRA-FV and the studies used to establish correlations between the factors and reoffense.

¶14 The final limitation to the SRA-FV is that it has shown a relatively low inter-rater reliability. Essentially, this is a measure of how frequently different people administering the instrument reach the same result. Although low, it is not low enough to be considered invalid.

¶15 After hearing the testimony and reviewing the exhibits, the trial court determined that opinions based on the SRA-FV are admissible under Frye. The trial court entered extensive findings of fact and conclusions of law. The parties filed supplemental briefs concerning the Frye hearing; Mr. Ritter challenged many of the court’s findings. A panel subsequently considered the case without oral argument.

ANALYSIS

¶16 In light of the previous remand, the primary issue presented by this appeal is whether the SRA-FV satisfies the Frye standard for admissibility. We conclude, as did Division Two of this court while this matter was on remand, that the SRA-FV does satisfy Frye.

¶17 Whether novel scientific evidence is admissible presents a mixed question of law and fact that this court *500reviews de novo. In re Det. of Pettis, 188 Wn. App. 198, 204-05, 352 P.3d 841 (finding that the SRA-FV satisfies Frye), review denied, 184 Wn.2d 1025 (2015). Pettis involved the same two primary psychological experts who testified in this case—Dr. Amy Phenix and Dr. Brian Abbott. Id. at 208-10. Dr. Abbott did not testify in Pettis, but his critical article concerning the test was discussed in the opinion. Id. at 209.

¶18 Washington applies the Frye test to gauge whether expert testimony premised on scientific evidence may be admissible. State v. Copeland, 130 Wn.2d 244, 261, 922 P.2d 1304 (1996). Frye requires that expert testimony be based on principles generally accepted in the scientific community. State v. Canaday, 90 Wn.2d 808, 812, 585 P.2d 1185 (1978). The test is two pronged: (1) whether the underlying theory is generally accepted in the scientific community and (2) whether there are techniques utilizing the theory which are capable of producing reliable results. State v. Riker, 123 Wn.2d 351, 359, 869 P.2d 43 (1994). The court does not assess the reliability of the evidence, but if there is significant dispute between qualified experts as to its validity, it may not be admitted. Copeland, 130 Wn.2d at 255. If the scientific principle satisfies Frye, the trial court applies ER 702 in determining whether to admit testimony. Pettis, 188 Wn. App. at 205. This court reviews the trial court’s ER 702 ruling for abuse of discretion. Id. Discretion is abused if it is exercised on untenable grounds or for untenable reasons. State ex rel. Carroll v. Junker, 79 Wn.2d 12, 26, 482 P.2d 775 (1971).

¶19 Here, there is no dispute that the principles underlying the SRA-FV are generally accepted in the scientific community. It is based on research linking dynamic risk factors with the probability that a sex offender will reoffend in the future. There also is general agreement that a structured analysis of those factors leads to a more reliable prediction than a haphazard, individualized inquiry. Accord Pettis, 188 Wn. App. at 207-10. This is essen*501tially the same process used in applying static risk factors. The first prong of the Frye test is satisfied.

¶20 The real dispute is whether the SRA-FV is capable of producing reliable results, thereby satisfying the second prong of the Frye test. The defense challenged the test in the trial court by arguing several weaknesses in the current model. First, the defense experts challenged the efficacy of the test by pointing out the lack of additional7 validation studies. The statistician, Dr. Glaser, was dissatisfied with the data presented in support of the SRA-FV, but he agreed that what was available did establish that the instrument showed a significant incremental improvement in predictive accuracy. More critically, neither Dr. Glaser nor any other witness suggested that the SRA-FV was inaccurate or produced invalid results.

¶21 The defense also challenged the reliability of the test, stressing that the inter-rater reliability was somewhat low. This challenge is significant because inter-rater reliability, the ability of different evaluators to obtain similar results, represents the instrument’s precision. Subsequent studies, however, have indicated higher rates of inter-rater reliability that are well within the range accepted by the psychological community. This evidence establishes that there are generally accepted methods of applying the SRA-FV. Pettis, 188 Wn. App. at 210.

¶22 Finally, at trial and on appeal, the defense placed great weight on the lack of construct validity. In psychometric testing, construct validity is of paramount importance because a test purporting to establish a construct is useless if it does not actually establish that construct. However, the SRA-FV is not primarily a psychometric test; it is a predictive test. Dr. Phenix pointed out that construct validity might be useful in refining the test in the future, but if any of the metric components of the instrument measured something other than what they were supposed to measure, *502it did not affect the predictive accuracy of the SRA-FV. As with the previous arguments, this challenge is unavailing.

¶23 The trial court correctly determined that the arguments presented against the SRA-FV went to the weight of the assessment, not its admissibility. Pettis, 188 Wn. App. at 211. Accordingly, we reach the same conclusion that the Pettis court did:

We hold that there are generally accepted methods of applying the SRA-FV in a manner capable of producing reliable results, and thus it passes the second prong of the Frye test. Thus, we hold that the SRA-FV passes the Frye test.

Id.

¶24 The trial court properly admitted the SRA-FV assessment in Mr. Ritter’s trial. Thus, we affirm the commitment order.

¶25 A majority of the panel having determined that only the foregoing portion of this opinion will be printed in the Washington Appellate Reports and that the remainder having no precedential value shall be filed for public record pursuant to RCW 2.06.040, it is so ordered.

Siddoway, C.J., and Fearing, J., concur.

Review denied at 185 Wn.2d 1039 (2016).

Frye v. United States, 54 App. D.C. 46, 293 F. 1013 (1923).

The impetus for these studies arose out of other studies that showed that treatment did no better than random in predicting reoffense.

For example, sexual interest in children or sexual violence falls into the sexual interest construct, while impulsivity or response to authority falls into the self-management category.

Attitudes were omitted because there is no valid way of determining their presence or absence in an individual.

After arriving at that number, practitioners will also look at individual case factors that may affect their determinations but were not included in the instruments.

The hospital treated high-risk sex offenders who had been civilly committed from the 1960s through the 1980s.

The SRA-FV has been cross validated with the Static-99R.