This is a class action brought pursuant to 42 U.S.C.A. § 1983 (West 1981) to challenge the Georgia procedures governing the involuntary commitment and release of persons who have been acquitted of criminal charges by reason of insanity (insanity ae-quittees). The plaintiffs’ class was certified as consisting of “all persons who are, or will be, confined in mental hospitals pursuant to Ga.Code § 27-1503 (Ga.Laws 1977, pp. 1293, 1295-96, Sec. 2) following findings of not guilty by reason of insanity.” Defendant Edwards is the Chief Executive Officer and Commissioner of the Department of Human Resources. Defendant John Branning is the Superintendent of Northwest Georgia Regional Hospital in Rome, Georgia. The defendants will be referred to as the “State” or as “Georgia.” The procedures are challenged as violating both the Due Process Clause and the Equal Protection Clause of the Fourteenth Amendment.
The plaintiffs attack Ga.Code Ann. § 27-1503 (1978),1 which sets forth the pro*513cedures that govern the disposition of defendants found not guilty by reason of insanity. These provisions, however, must be read in conjunction with the interpretive gloss which the appellate courts of Georgia have applied to the statute. See Clark v. State, 245 Ga. 629, 266 S.E.2d 466 (1980); Skelton v. Slaton, 243 Ga. 426, 254 S.E.2d 704 (1979); Pitts v. State, 151 Ga.App. 691, 261 S.E.2d 435 (1979); Pennewell v. State, 148 Ga.App. 611, 251 S.E.2d 832 (1979); Du-bose v. State, 148 Ga.App. 9, 251 S.E.2d 15 (1978). The statutory scheme, as construed by the Georgia courts, can be summarized as follows.2
Following a finding of not guilty by reason of insanity,3 the trial court retains jurisdiction over the insanity acquittee and inquires into the present mental state of the person, and “upon a showing of good cause by the prosecutor”4 may order such person to be confined to a mental hospital for not less than 30 days. In fact, the commitment is for an indefinite period of time since the State does not initiate a hearing to determine the current mental state of the insanity acquittee.
Once so committed to a state mental hospital the acquittee or the hospital must petition the committing court to secure the acquittee’s release. A petition cannot be entertained within the initial thirty day period nor within twelve months of any prior petition. A valid petition will set in motion a hearing at which the sole issue is whether the insanity acquittee meets the criteria for civil commitment5 under the Georgia Men-
*514tal Health Code, Georgia Code Chapter 88-5 or 88-25.6
In contrast, an M.H.C. committee must automatically receive a hearing within 17 days of his involuntary hospitalization, unless he in writing waives the right to this hearing. Although insanity acquittees and M.H.C. committees are guaranteed many of the same rights at their respective hearings (e.g., the right to counsel, the right to confront and cross-examine witnesses), see 501 F.Supp. at 1056, the release hearing for insanity acquittees differs substantially in several ways from the commitment and release hearings for M.H.C. committees. Not only are insanity acquittees denied a state-initiated hearing at the start of their commitment, but they may apply for release only once every twelve months; after the initial commitment of the M.H.C. committee, the state must “recommit” the committee within six months, and at twelve month intervals thereafter. Insanity acquittees are presumed to be mentally ill; M.H.C. committees are not presumed to be mentally ill at either their commitment or recom-mitment hearings. Insanity acquittees bear the burden of proving their fitness for release and the State, therefore, need not prove by clear and convincing evidence that an acquittee meets the Chapter 88-5 criteria for continued confinement; the State must prove by clear and convincing evidence that the M.H.C. committee meets the Chapter 88-5 criteria. The release of an insanity acquittee must be ordered by the committing court; the hospital may release the M.H.C. committee at any time without any judicial approval. See Clark v. State, supra; Pitts v. State, supra; Pennewell v. State, supra; Ga.Code Ann. §§ 27-1503, 38-118, 88-501(u), 88-506.2, 88-506.5, and 88-506.6 (1978, 1979, 1981 and Supp. 1981).
In a comprehensive and scholarly opinion, Judge Murphy granted plaintiffs motion for preliminary injunction,7 holding:8
1. That both the Equal Protection Clause and the Due Process Clause require *515Georgia to provide insanity acquittees with a state-initiated initial commitment hearing;
2. That equal protection and due process considerations require that the State bear the burden of proving, by clear and convincing evidence, the commitment criteria at the initial commitment hearing;
3. With respect to Georgia’s presumption that an insanity acquittee’s mental state continues, reliance on this presumption to justify procedural differences in the commitment process violates both equal protection and due process; and
4. That some of Georgia’s procedures for the release of insanity acquittees violate equal protection and due process, but that Georgia’s requirement of court approval for the release of insanity acquittees is constitutional when limited to acquittees who have committed violent crimes.
The State challenges each of the foregoing rulings. We will consider each in turn. In dealing with the release issue, we will address separately its three aspects: the requirement of court approval for the release of insanity acquittees, the burden of proof at the release hearing, and the frequency of applications for release.9
*516I. PRESUMPTION OF CONTINUING MENTAL STATE
Departing from the order in which the district court discussed the issues, we address first the constitutionality of Georgia’s presumption that the mental state of an acquittee continues — i.e., once a defendant has been found not guilty by reason of insanity, in all subsequent legal proceedings Georgia law presumes, subject to rebuttal, that the acquittee is still legally insane. We address this issue first because the state asserts that this presumption is a partial justification for denying insanity acquittees a state-initiated commitment hearing and for the placement of the burden of proof on insanity acquittees.
The presumption derives from Ga.Code § 38-118, which provides: “Other presumptions of law, such as of ... continuance of ... a mental state once proved to exist ... may be rebutted by proof.” Ga.Code § 38-118 (1981). Applying this statutory presumption, the Georgia scheme permits the presumption of present mental illness from the fact of the finding of insanity at the criminal trial. Clark v. State, 245 Ga. 629, 266 S.E.2d 466, 470, 476-77 (1980). Al*517though the statute provides support only for a presumption that mental illness continues, Clark arguably creates a judicial presumption that the dangerousness10 evidenced by the crime also continues.11 Thus, the Georgia scheme combines a statutory presumption and an implicit judicial presumption of both present mental illness and present dangerousness from the fact of the insanity acquittal.
The district court held that both due process and equal protection prohibit utilization of the presumption either to justify denying insanity acquittees a state-initiated commitment hearing or to justify imposing on them a higher burden of proof. We need not decide whether the presumption violates due process, because it is, beyond peradventure, a violation of the Equal Protection Clause.
At the initial commitment hearing for an M.H.C. committee, the State must demonstrate by clear and convincing evidence that the M.H.C. committee is both mentally ill and dangerous. Six months after that initial determination, and annually thereafter, the State must “recommit” the M.H.C. committee, sustaining again a burden of proof, by clear and convincing evidence, that the M.H.C. committee is both mentally ill and dangerous. No presumption of continued mental illness or continued dangerousness is operative, see Clark v. State, 266 S.E.2d at 476, even though the standard to be proved at the subsequent hearing — i.e., mental illness and dangerousness — is identical to the standard at the initial hearing.
By contrast, Georgia applies the presumption against insanity acquittees, notwithstanding the fact that the insanity acquittal findings are somewhat different from the findings necessary for commitment, and notwithstanding the fact that the criminal trial, unlike the civil commitment hearing, is not focused on the two crucial findings, i.e., mental illness and dangerousness.
Several reasons convince us that there is no rational basis for applying the presumption against insanity acquittees and not against M.H.C. committees. First, although we acknowledge that the finding of insanity, whether it be pursuant to the M’Naghten standard or the delusional compulsion standard,12 probably encompasses a finding of mental illness as that term is defined by Georgia Code § 88-501(a), we nevertheless doubt that the reliability of the insanity finding equals that of the finding of mental illness at a commitment hearing. The normal statutory procedures for commitment hearings are meticulous and insure a significant measure of reliability. Two preliminary hearings precede the commitment hearing itself, as well as a five day evaluation period at a mental hospital. Certifications from several physicians are required. Ga.Code §§ 88-505.2, 88-505.3, 88-505.5, 88-506.2 (1979 & Supp. 1981).13 Moreover, *518in a criminal trial the finding of insanity is only by a preponderance of the evidence; in a civil commitment hearing the finding of mental illness is supported by the higher standard of clear and convincing evidence. Even if we cannot say with certainty that the civil commitment finding of mental illness is more reliable than the insanity finding, we can say with confidence that there is no rational basis at all to assume that the insanity finding is more reliable.14
Second, we believe that the finding of dangerousness at a commitment hearing is more reliable than that at an insanity acquittal both because of the meticulous procedures above mentioned, and because the commitment hearing is focused exclusively on only two crucial issues, i.e., mental illness and dangerousness, whereas the criminal trial is focused on conviction and the issue of insanity, but the issue of dangerousness is involved, if at all, only incidentally. Even crimes such as murder, which implicate the dangerousness criterion, do so by revealing past violent behavior; the Mental Health Code criterion of dangerousness, however, speaks to the probability of future dangerous conduct. A person whose mental illness has manifested itself in violence in the past may in some cases pose no threat to himself or others in the future. The application of the dangerousness criterion in a civil commitment hearing entails a prediction of future conduct which cannot be equated with a finding of a single past criminal act, even if that act constituted a substantial risk of imminent harm to the perpetrator or others. While Georgia does require that this past act must provide reasonable cause to believe that such a prediction of future dangerousness may be made, this reasonable cause finding also cannot be equated with the actual prediction of dangerousness. Simply stated, the criminal trial looks toward the past, the civil commitment hearing and criteria look toward the future.
Third, and most important, the dangerousness finding at a commitment hearing is a finding that the person involved is sufficiently dangerous to require involuntary commitment. On the other hand, the finding of dangerousness at an insanity acquittal is considerably more anemic. The finding is merely that the insanity acquittee has committed a crime that implicates the dangerousness criterion.15 The fact of the crime is merely relevant to the dangerousness criterion; it is not a finding that the dangerousness criterion has been met. It is true that the Georgia procedure provides for a summary finding by the criminal trial judge, immediately after the insanity acquittal, that there is “reasonable cause to believe that the defendant presently meets the criteria for civil commitment,” Clark v. State, 266 S.E.2d at 475. However, such a probable cause finding is clearly less reliable than the final judicial finding of dangerousness at a commitment hearing, a full and fair hearing following the meticulous statutory procedures mentioned above. A finding that there is probable cause to believe a person is dangerous is considerably weaker than a finding, supported by clear and convincing evidence, that a person is dangerous. The contrast between the two findings of dangerousness might be more starkly stated as follows; with respect to M.H.C. committees, fact A (the dangerousness criterion) has been finally proved at the commitment hearing following meticulous procedures, but at subsequent hearings the identical fact A must be proved again without the benefit of any presumption. On the other hand, with respect to insanity acquittees fact X (the commission of a *519crime implicating danger in some manner) has been proved at the insanity acquittal, from which fact the trial judge has found that there is probable cause to believe fact A (the dangerousness criterion), all of which the State argues, by application of the presumption, make it unnecessary to prove at all, or justifies a lesser burden of proof with respect to, the existence of fact A.
We conclude that there is no conceivable rational basis for applying the presumption against insanity acquittees and not against M.H.C. committees, when the factual predicate for the presumption would be stronger in the context of M.H.C. committee hearings. Accordingly, we hold that application of the presumption in the instant case violates the Equal Protection Clause.15.1
II. STATE-INITIATED HEARING
Under the Georgia scheme, insanity acquittees may be summarily committed immediately following their acquittal by reason of insanity, without a hearing to determine their present status with respect to mental illness and dangerousness. The summary commitment is made on the basis of the criminal trial itself. See 501 F.Supp. at 1055, n.4; Clark v. State, 245 Ga. 629,266 S.E.2d 466, 475 (1980). Because a state-initiated commitment hearing is provided to determine whether M.H.C. committees meet the commitment criteria at the time of commitment — i.e., present mental illness and dangerousness — the district court held that Georgia’s failure to provide such a hearing for insanity acquittees violates the Equal Protection Clause. The district court also held that such a hearing is required by the Due Process Clause. We agree.15.2
This precise issue was addressed in Powell v. Florida, 579 F.2d 324 (5th Cir. 1978). In Powell, the Florida statute empowered the trial court to commit an insanity acquit-tee if the court considers that the discharging of the insanity acquittee would be manifestly dangerous to the safety of the people. 579 F.2d at 326, n. 2. Powell was thus committed on the basis of evidence presented at his trial in chief and without any evidence of present mental condition. 579 F.2d at 329. The panel held that this original commitment without a hearing violated Powell’s due process rights, saying:
An acquittal by reason of insanity is based upon the jury’s determination that there is a reasonable doubt as to the defendant’s sanity at the time of the offense. See Farrell v. State, 101 So.2d 130 (Fla.1958). It does not speak to the defendant’s present mental condition. Bolton v. Harris, 130 U.S.App.D.C. 1, 5-9, 395 F.2d 642, 646-650 (1968). Before a trial judge can refuse to discharge a person acquitted by reason of insanity, Fla. R.Crim.P. 3.460 requires a finding that “the discharge or going at large of such insane person shall be . . dangerous to the peace and safety of the people.” The section thus requires the trial judge, before he can order the defendant committed, to make a determination that the defendant is presently insane and that he *520now would be dangerous if discharged. In re Connors, 332 So.2d 336, 338 (Fla. 1976). Commitment, therefore, depends upon findings of fact different from those established by the jury’s acquittal by reason of insanity. Due process requires that such findings be made following a hearing at which the defendant is present with counsel, has an opportunity to be heard and to present evidence, and has an opportunity to confront and cross examine witnesses. Specht v. Patterson, 386 U.S. 605, 87 S.Ct. 1209,18 L.Ed.2d 326 (1967); Bolton v. Harris, 130 U.S.App. D.C. 1, 9, 395 F.2d 642, 650 (1968). The trial judge’s commitment of Powell based only on “the evidence adduced at trial, the plea of not guilty by reason of insanity and the verdict entered in this cause” and his failure to hold a separate hearing to take evidence on Powell’s then present mental condition violated Powell’s due process rights.
579 F.2d at 330. Accord Jackson v. Foti, 670 F.2d 516 (5th Cir. 1982).
The State attempts to distinguish Powell and Jackson by pointing to the Georgia presumption of continuing state of mind. The State argues that the presumption fills in the time gap which concerned the Powell panel. Since the law presumes that an insanity acquittee’s mental deficiency at the time of the crime continues to the time of commitment, the State argues that there is no need to hear evidence concerning present mental state and thus no need for a separate commitment hearing. We need not address other possible deficiencies in this argument,16 because we have held in Part I of this opinion that application of this presumption in the manner urged is constitutionally forbidden. Thus, we reject the State’s attempted distinction of Powell and Jackson on this ground.
The State also argues that these plaintiffs’ due process rights are not violated because, while they are not entitled to an automatic, state-initiated hearing, they do have a right after the expiration of the 30 day evaluation period to notice of their right to a hearing, which they waive if they decline to file an application for release.17 Finding Powell v. Florida, supra, and Jackson v. Foti, supra, controlling, we reject the state’s argument that the mere availability of a hearing is sufficient. In both cases, this court held that due process requires a commitment hearing to determine an insanity acquittee’s current mental condition and dangerousness.18 In both cases the insanity acquittee had the right, under the state procedures, to file a state habeas corpus petition to seek release;19 yet in both cases the state procedures were found to violate due process and a state-initiated commitment hearing was required. We find no significant difference between the state ha-beas corpus petition which was available in Powell and Jackson to obtain a commitment hearing, and the elective application for release which was available to the plaintiffs here. We agree with the implicit holdings of Powell and Jackson that the mere availability of a commitment hearing is not sufficient due process in the case of an insanity acquittee who is confined in a mental institution under circumstances that may be adverse to his intelligent exercise of his rights,20 and for whom at this stage of the proceedings counsel is not appointed.21 Ac*521cord United States ex rel. Schuster v. Herold, 410 F.2d 1071, 1085 (2d Cir. 1969). We also find support for our position in a recent Ninth Circuit decision, Doe v. Gallinot, 657 F.2d 1017 (9th Cir. 1981), which expressly rejected, on grounds similar to those we find persuasive, the adequacy of state habeas corpus relief as a substitute for a mandatory, state-initiated hearing before the involuntary civil commitment of mentally ill persons. Because the Ninth Circuit decision involved commitment for only a limited time, i.e., 14 days, the instant case involving commitment for an indeterminate time presents a much more compelling case.
We conclude, following Powell v. Florida, supra, and Jackson v. Foti, supra, that Georgia’s procedure affording only the mere availability of a commitment hearing violates the Due Process Clause.
Both Powell v. Florida, 579 F.2d at 334, and Jackson v. Foti, 670 F.2d at 520-21, also hold that a state-initiated commitment hearing is mandated for insanity acquittees by the Equal Protection Clause, when civil committees are provided such a hearing before involuntary commitment. We see no way to distinguish the instant case from Powell and Jackson, and accordingly we also hold that the Equal Protection Clause requires that the plaintiffs here receive the same state-initiated commitment hearing which Georgia provides for M.H.C. committees.
Accordingly, we affirm the district court’s holding that Georgia, at the expiration of the 30 day evaluation period,22 must provide for insanity acquittees the same state-initiated commitment hearing that it provides for M.H.C. committees. Accord Bolton v. Harris, 395 F.2d 642 (D.C.Cir. 1968). See Specht v. Patterson, 386 U.S. 605, 87 S.Ct. 1209, 18 L.Ed.2d 326 (1967).
III. BURDEN OF PROOF AT COMMITMENT HEARING
Having determined that there must be a state-initiated commitment hearing, we next address the issue of what burden of proof is constitutionally required. The State defends the Georgia scheme which places the burden on the insanity acquittee to prove by a preponderance of the evidence that he is not mentally ill and dangerous. The district court held on equal protection and due process grounds that, at the initial commitment hearing, the State must bear the burden of proving mental illness and dangerousness by clear and convincing evidence. We agree.
Our analysis begins with Addington v. Texas, 441 U.S. 418, 99 S.Ct. 1804, 60 L.Ed.2d 323 (1979). There the Supreme Court of Texas held that the appropriate standard of proof for the involuntary commitment of civil committees was the preponderance of the evidence standard. The Supreme Court of the United States disagreed, holding that the preponderance standard falls short of the demands of the Due Process Clause. The court noted:
This Court repeatedly has recognized that civil commitment for any purpose constitutes a significant deprivation of liberty that requires due process protection. See e.g., Jackson v. Indiana, 406 US 715 [92 S.Ct. 1845, 32 L.Ed.2d 435] ...; Humphrey v. Cady, 405 US 504 [92 S.Ct. 1048, 31 L.Ed.2d 394] ....
. . . [T]he state has no interest in confining individuals involuntarily if they are not mentally ill or if they do not pose some danger to themselves or others. Since the preponderance standard creates the risk of increasing the number of individuals erroneously committed, it is at least unclear to what extent, if any, the state’s interests are furthered by using a preponderance standard in such commitment proceedings.
... We conclude that the individual’s interest in the outcome of a civil commit*522ment proceeding is of such weight and gravity that due process requires the state to justify confinement by proof more substantial than a mere preponderance of the evidence.
441 U.S. at 425-27, 99 S.Ct. at 1809-10. Having rejected the preponderance standard as being insufficient to meet due process requirements, the Court proceeded to reject also the very strict criminal law standard, j.e., beyond a reasonable doubt. The Court found the strict criminal law standard inappropriate for several reasons. First, the Court noted that the continuous professional review of a patient’s condition will afford opportunities for correction of any initially erroneous commitments. Second, the Court was concerned that the “subtleties and nuances of psychiatric diagnosis render certainties virtually beyond reach in most situations.” 441 U.S. at 430, 99 S.Ct. at 1811. In contrast to the criminal prosecution context where specific, knowable facts make the reasonable doubt standard practicable, “psychiatric diagnosis . .. is to a large extent based on medical ‘impressions’ drawn from subjective analysis and filtered through the experience of the diagnostician. This process often makes it very difficult for the expert physician to offer definite conclusions about any particular patient.” 441 U.S. at 430, 99 S.Ct. at 1811. The Court thus concluded that the reasonable doubt standard is inappropriate “because, given the uncertainties of psychiatric diagnosis, it may impose a burden the state cannot meet.” 441 U.S. at 432, 99 S.Ct. at 1812.
Having held that the preponderance standard falls short of the due process demands, but that the strict reasonable doubt standard is not required, the Court adopted the intermediate “clear and convincing” evidence standard. 441 U.S. at 431, 99 S.Ct. at 1812.
We cannot conclude that there is sufficient difference between the instant class of insanity acquittees and the class of civil committees in Addington to justify a departure from the Addington “clear and convincing” evidence standard. Insanity ac-quittees, like civil committees, have at stake a significant deprivation of their liberty.
The State argues that the insanity acquittal, with its determination of insanity, justifies a different burden of proof, i.e., placement of the burden of proof on the insanity acquittee. In Powell v. Florida, supra, we rejected in another context the State’s attempt to attach significance to this previous finding of insanity. There, we pointed out that the previous finding of insanity related to the time of the offense, and “does not speak to the defendant’s present mental condition.” 579 F.2d at 330. This time gap cannot be bridged by utilizing the Georgia presumption of a continued mental state; as discussed above, the presumption is not constitutionally permissible.
In the only significant post -Addington decision by a federal circuit court of appeals,22.1 Warren v. Harvey, 632 F.2d 925 *523(2d Cir. 1980), the Second Circuit considered a constitutional challenge to the Connecticut statute providing for the commitment of insanity acquittees. The statute required that the state prove that the insanity acquittee was mentally ill and dangerous by a preponderance of the evidence. It was challenged as falling short of the clear and convincing evidence standard required in Addington v. Texas, supra, for the commitment of civil committees. The Second Circuit held that the fact that the insanity acquittee had committed the criminal act justified the lesser standard of proof, implicitly holding that insanity acquittees were more dangerous than civil committees.
After careful analysis of Warren v. Harvey, several considerations persuade us to decline to follow it.
First, we cannot square the Second Circuit case with the controlling Supreme Court precedent, primarily Addington v. Texas, supra. The factor relied upon by the Second Circuit to justify the lesser burden of proof was the insanity acquittee’s commission of the crime. However, Addington also had committed a crime. His civil commitment was triggered by an “assault by threat” against his mother, for which he was arrested and taken into custody. Thus, in balancing the interest of Addington (protection against the risk of erroneous commitment, a significant deprivation of his liberty) against the interest of the State (protection for the community against the dangerous tendencies of such mentally ill persons), the Supreme Court was considering a person whose commitment was triggered by circumstances involving the commission of a crime, just as was the commitment in Warren v. Harvey, and as was the commitment of each class member before us. This fact significantly undermines the Second Circuit’s distinction of Addington. We conclude that the State of Georgia’s interest in protection against the erroneous release of insanity acquittees generally is not significantly different from Texas’ interest in protection against the erroneous release of Addington. And, as noted above, the interest of insanity acquittees generally is identical to Addington’s interest — i.e., protection against erroneous commitment, a significant deprivation of liberty.
Second, we also think that the implicit assumption in Warren v. Harvey that insanity acquittees are more dangerous than civil committees is inconsistent with the second prong of Baxstrom v. Herold, 383 U.S. 107, 113, 115, 86 S.Ct. 760, 763, 764, 15 L.Ed.2d 620 (1966). There the Supreme Court held that the fact of Baxstrom’s past crime did not establish that he was “such a danger to others that the strict security of a Department of Correction hospital is warranted,” 383 U.S. at 115, 86 S.Ct. at 764. The Court held that Baxstrom’s past crime did not justify denying Baxstrom judicial review of the administrative decision assigning him to the maximum security institution when all civil committees could be assigned to that institution only after a judicial determination that the “person is so dangerously mentally ill that his presence in a civil hospital is dangerous to the safety of other patients or employees, or to the community.” 383 U.S. at 113, 86 S.Ct. at 763. The Supreme Court so held despite the fact that Bax-strom’s crime involved an assault on a police officer with an ice pick at night, injuring the officer’s face, forehead, and collar bone. Brief for Respondent at App. 4r-5, Baxstrom v. Herold, supra. All civil committees are by definition dangerous; the implicit assumption of Warren v, Harvey is that insanity acquittees are more dangerous. This is the precise assumption that
*524was found insufficient in Baxstrom’s second holding. See dicta to the same effect in Jackson v. Indiana, 406 U.S. at 728, 92 S.Ct. at 1853.23
Third, we find unpersuasive the three explicit considerations relied upon by Warren v. Harvey in distinguishing Addington. The Second Circuit minimized the interest of insanity acquittees in protection against erroneous commitment, assuming that any such error would probably have been offset by a similar error in the determination of his insanity in connection with his insanity acquittal. We harbor considerable doubt that there is any logical necessity or even probability that one such error would lead to the other. In any event, the assumption runs counter to Powell v. Florida, 579 F.2d at 330, which emphasizes that the finding of insanity as of the time of the crime is discrete from the finding of mental illness and dangerousness at the time of the commitment hearing. An acquittal, even on grounds of insanity, generally absolves the defendant of criminal responsibility for his act. Notwithstanding this well settled principle, the Second Circuit’s justification for the different burden of proof is in fact an attempt to punish an “erroneously” acquitted insanity acquittee. Even if an “error” were made in finding the defendant not guilty by reason of insanity, this “error” cannot be “corrected” by committing the defendant under the State’s mental health power. See infra, n. 24. The Warren v. Harvey court also minimized the stigma which would be suffered by the erroneous commitment of an insanity acquittee, observing that insanity acquittees had already incurred stigma as a result of the judicial determination, by virtue of their insanity acquittal, that they committed a crime. We find this observation inconsistent with Vi-tek v. Jones, 445 U.S. 480, 492, 494, 100 S.Ct. 1254,1263,1264, 63 L.Ed.2d 552 (1980), which held that commitment of those already incarcerated as prisoners constitutes a significant stigma. Moreover, the stigma element is certainly of lesser significance than the deprivation of liberty. Finally the Second Circuit relied upon the danger of calculated abuse of the insanity defense, and the deterrent effect that might be afforded by the utilization of the lesser burden of proof. This consideration also contains overtones of punitive consequences lingering after an insanity acquittal, which is inconsistent with Georgia’s rejection of the notion that any punitive purpose remains after as insanity acquittal. Bailey v. State, 210 Ga. 52, 77 S.E.2d 511 (1953) (A person confined to ... a hospital for the insane after his acquittal on the ground of insanity, is not to be considered as a criminal undergoing punishment; he is there solely to be cared for, protected, and guarded so that he may not injure another or himself).24 Even if these considerations are proper, and we do not believe that they are, we note also that, as compared to the Georgia scheme, there is a greater risk in the Connecticut scheme under consideration in Warren that an accused could receive an insanity acquittal, but that the state could not satisfy its burden of proving the criteria for involuntary commitment. In Connecticut, an accused need only establish a reasonable doubt as to his sanity in order to obtain an insanity acquittal. By contrast, in Georgia, the accused must prove his insanity by a preponderance of the evidence. *525We conclude that little weight should be accorded to the suggested deterrent effect to guard against calculated abuse of the insanity defense,25 and we agree with the Warren court’s own acknowledgment that such deterrent effect is “slight.”
For the foregoing reasons, we find Warren v. Harvey, supra, unpersuasive.26 We can find no satisfactory ground to distinguish Addington v. Texas, supra, and we therefore affirm the district court’s conclusion that due process requires the State to bear the burden of proof at the initial commitment hearing by clear and convincing evidence.
We hold alternatively on equal protection grounds that Georgia cannot place the burden of proof on insanity acquittees while simultaneously providing that the State must prove the initial commitment criteria of M.H.C. committees by clear and convincing evidence. The foregoing discussion indicates that there is no rational basis to justify treating the two classes differently with respect to the burden of proof. The strongest justification offered by the State is the intuitive feeling that the insanity acquittees are more dangerous than M.H.C. committees. However, as noted above, the Supreme Court in Baxstrom v. Herold, supra, rejected this precise justification in the second prong of its equal protection analysis. 383 U.S. at 115, 86 S.Ct. at 764. Accord, Bolton v. Harris, 395 F.2d 642, 647, 649 (D.C.Cir.1968) (referring to the “principle derived from the Supreme Court’s decision in Baxstrom v. Herold that the commission of criminal acts does not give rise to a presumption of dangerousness which, standing alone, justifies substantial difference in commitment procedures and confinement conditions for the mentally ill”); Waite v. Jacobs, 475 F.2d 392, 397 (D.C.Cir. 1973); Cameron v. Mullen, 387 F.2d 193, 201-02 (D.C.Cir.1967); United States ex rel. Schuster v. Herold, 410 F.2d 1071, 1081, 1085-86 (2d Cir. 1969)27 With respect to *526the purported state interest of isolating the super-dangerous28 for different treatment, the classification of insanity acquittees is overinclusive and underinclusive in precisely the same way and to precisely the same degree as was the classification in Bax-strom’s second prong. The disadvantaged class of insanity acquittees is overinclusive to the extent that they might have committed crimes which evidenced a degree of danger less than or comparable to, but not greater than, the dangerousness inherent in M.H.C. committees generally. This is identical to the overinclusiveness in Baxstrom’s second prong where the Supreme Court observed that Baxstrom “may or may not be . . . such a danger to others that the strict security of a Department of Correction hospital is warranted.” 383 U.S. at 114-15, 86 S.Ct. at 764. Similarly, the disadvantaged class of insanity acquittees is underinclusive to the extent that M.H.C. committees also might have committed past dangerous crimes and to the extent that they might for other reasons be more dangerous than the average M.H.C. committee. The classification which failed the equal protection scrutiny in Baxstrom’s second prong was underinclusive in the same way and to the same degree. 383 U.S. at 115, 86 S.Ct. at 764. We conclude that Baxstrom’s second prong forecloses the State’s reliance on the proffered justification that insanity acquit-tees generally are more dangerous than M.H.C. committees generally.
Moreover, the instant record does not support a conclusion that insanity acquit-tees are in fact more dangerous.29 Com*527mentators have questioned the accuracy of the intuitive assumption that insanity acquittees are more dangerous. See e.g. Note: Commitment and Release of Persons Found Not Guilty by Reason of Insanity, A Georgia Perspective, 15 Ga.L.Rev. 1065, 1079 (1981) (“Scientific studies,[30] while not conclusive, do indicate that .. . [insanity acquittees] as a group are not substantially more dangerous than their civilly committed counterparts.”) The Second Circuit in United States ex rel. Schuster v. Herold, 410 F.2d 1071 (2d Cir. 1969), reporting on the subsequent history after Baxstrom v. Herold, raised similar doubts about the assumption:
Most judicial reform is accompanied by cries of horror and dismay that the action by the court has surely carried society over the brink and into the abyss of administrative chaos. Certainly this was true of Baxstrom. When § 384 was invalidated, the New York authorities decided to transfer all patients confined in Dannemora or Matteawan pursuant to that provision — a total of 992 inmates — to civil hospitals. This transfer was designated by them as “Operation Baxstrom.” When this decision was announced, some segments of the community were anxious and outraged. Involved labor unions demanded special pay and training for working with the Operation Baxstrom transferees because they were presumptively “dangerous.” Residents in the neighborhoods of civil mental hospitals protested the presence of “criminal lunatics” in their midst. Yet in only one year the protests evaporated. The transfer has been an astounding success. Of these nearly 1000 exprisoners whom the appropriate state authorities had prior to Bax-strom determined to be too dangerous to be placed in a civil hospital, 176 have been fully discharged — 147 to their home communities and the rest to other hospitals; 454 others have elected to remain in civil mental hospitals on voluntary or informal status; and only seven have had to be returned to Matteawan after a judicial determination that they were dangerous. As one author has noted, the Bax-strom patients were almost as pure as Ivory Snow; they were 9928/ioo per cent free from dangerous mental illness.
410 F.2d at 1085-86 (footnote omitted).
The argument that insanity acquittees generally are more dangerous than M.H.C. committees generally is especially weak in light of the Georgia statutory scheme. Before the state-initiated hearing — which is the relevant time for purposes of determining the appropriate burden of proof at that hearing — the statutory scheme contemplates that the following procedures normally will have been completed and the following findings will have been made with respect to M.H.C. committees: (1) either a preliminary investigation by the county board of health has found that there is probable cause to believe that such person is mentally ill and dangerous, or a physician has certified that he has examined such person within the preceding five days and found that such person may be mentally ill and dangerous, Ga.Code Ann. § 88-505.2 (1979); and (2) the court, after reviewing the aforementioned finding or certificate, has found that there is reasonable cause to believe that such person may be mentally ill and dangerous, and has therefore held a full and fair hearing thereon, and after such hearing has satisfied itself that an immediate evaluation is necessary, whereupon the court orders such person to be delivered to the hospital, for a five day evaluation period, id. § 88-505.3; and (3) the Chief Medical Officer of the evaluating hospital, supported by the opinions of two physicians who have examined such person within the five day evaluation period, has recommended that such person is mentally ill and dangerous and petitioned the court for the state-initiated hearing at issue. Id. §§ 88-505.5 & 88-506.2.
*528Thus, the M.H.C. committee has been certified to be mentally ill and dangerous by a physician, followed by a probable cause determination by a court, followed by a finding by the court after a full and fair hearing, followed by a five day evaluation at the hospital, followed by a recommendation by the hospital that such person is mentally ill and dangerous. It is also significant that the foregoing findings of dangerousness are made pursuant to the Georgia definition which requires that the danger be “manifested by either recent overt acts or recent expressed threats of violence, which present a probability of physical injury to himself or to other persons.” Id. § 88-501(v).
By contrast, the insanity acquittee has been judicially determined to be legally insane at some previous time, i.e., the time of the crime, and to have committed the crime at that past time. In addition, the court has made the “good cause” finding that there is reasonable cause to believe that the insanity acquittee is presently mentally ill and dangerous, Clark v. State, 266 S.E.2d at 475, but that finding is a summary one made on the basis of evidence at criminal trial and without a further hearing. See infra, Part IV.A.2.
In this setting, we cannot conclude that there is any rational basis for the intuitive assumption — that insanity acquittees generally are more dangerous than M.H.C. committees generally — sufficient to justify discriminating against insanity acquittees with respect to the burden of proof. Of course, the egregious nature of any particular crime might justify an inference that such particular person is more dangerous than M.H.C. committees generally, e.g., the brutal rape and murder in United States v. Ecker, 543 F.2d 178 (D.C.Cir.1976), cert. denied, 429 U.S. 1063, 97 S.Ct. 788, 50 L.Ed.2d 779 (1977), or the murder in Warren v. Harvey, supra. However, such extreme cases should not skew the proper location of the burden of proof; in such cases the State will easily carry its burden of proof by clear and convincing evidence. Rather, it is only in the closer, more ambiguous cases that the burden of proof standard will be significant.
For the foregoing reasons, we hold that there is no rational basis to justify treating insanity acquittees differently from M.H.C. committees with respect to burden of proof at the initial commitment hearing.
We thus hold, on both due process and equal protection grounds, that the State must bear the burden of proving the commitment criteria by clear and convincing evidence.
IV. RELEASE PROCEDURES
A. Requirement of Court Approval for Release
As noted above, Georgia law provides that insanity acquittees committed pursuant to Georgia Code § 27-1503, must secure judicial approval in order to be released from the State hospital. Persons committed under Georgia Code Chapter 88-5 (i.e., civil commitment) may be discharged from the hospital without court approval if the chief medical officer of the hospital finds that the patient is either no longer mentally ill or no longer dangerous. Ga.Code Ann. § 88-506.6(b)(l) (1979). The plaintiffs contend that this difference in release procedures violates the Equal Protection Clause.
1. The District Court’s Opinion.
Although the district court astutely noted that Georgia Code § 27-1503 “must be read in conjunction with the interpretive gloss which appellate courts of Georgia have read into the statute,” 501 F.Supp. at 1055, the district court did not refer to these state court interpretations when it deemed the statute to require “all insanity-acquittees to obtain an order from the committing court before they could be released.” Id. at 1071 (emphasis added). Having so construed § 27-1503, the district court found “no rational basis for withholding from all insanity-acquittees the release procedure available to all M.H.C. committees in Georgia.” Id. (emphasis added).
The district court’s constitutional analysis paralleled the second prong of the Baxstrom case. Compare 501 F.Supp. at 1071-*52972 with 383 U.S. at 114-15, 86 S.Ct. at 764.31 The district court noted that justifying release procedures for all insanity acquittees different from those available for all M.H.C. committees on the ground that the former group is more dangerous than the latter exposes the legislative classification to problems of overinclusiveness and under-inclusiveness. Some M.H.C. committees have committed acts evidencing a much greater degree of dangerousness than have some acquittees (i.e., underinclusiveness), and many insanity acquittees have shown no greater likelihood of harming others than have M.H.C. committees (i.e., overin-clusiveness). In light of these considerations, the district court held that requiring all acquittees, but not all M.H.C. committees, to obtain judicial approval prior to release violated equal protection. Acknowledging, however, that some acquittees have committed acts which constitute the elements of “violent crimes” (i.e., “murder, rape, armed robbery, arson, aggravated assault, and other offenses of like nature,” 501 F.Supp. at 1072), the district court concluded that the State constitutionally could require acquittees who were charged with serious violent crimes to secure court approval for release.
Analysis of the State Law. 2.
Before reaching the difficult constitutional question of whether a state could validly impose more stringent release procedures for a 11 acquittees than those required for all M.H.C. committees, we should ascertain whether Georgia has indeed attempted to do so. A well-settled doctrine of constitutional law is that a court should not reach out and decide difficult questions of federal constitutional law if these questions can be avoided by a decision based on state law. Siler v. Louisville & Nashville Railroad Co., 213 U.S. 175, 29 S.Ct. 451, 53 L.Ed. 753 (1909).32 Moreover, in this case, if Georgia has not established one set of release procedures for all acquittees and another only for all M.H.C. committees, then a pronouncement on the constitutionality of such a plan would be merely an advisory opinion.
Our examination of the statute in question, as authoritatively interpreted by the Georgia Supreme Court, leads us to conclude that, as a matter of state law, § 27-1503 does not require judicial approval for the release of all acquittees involuntarily committed in Georgia. Although the statute does provide that in “all criminal trials” where the jury returns a verdict of *530not guilty by reason of insanity the trial court “shall immediately inquire into the sanity of the person at the time of acquittal,” the statute further requires that only “upon a showing of good cause by the prosecutor, may [the court] defer ruling on the same and order such person to be confined in a state mental hospital ... for a period of not less than 30 days.” Ga.Code Ann. § 27-1503(a) (1978) (emphasis added). The Georgia Supreme Court’s most recent interpretation of this statutory language reveals that not all insanity acquittees can be committed under this section; therefore, we must consider the release provisions to apply only to a more limited class than all acquittees.
In Clark v. State, 245 Ga. 629, 266 S.E.2d 466 (1980), the court construed the “good cause” language “to mean a showing that there is reasonable cause to believe that the defendant presently meets the criteria for civil commitment.” Id. 266 S.E.2d at 475, that is, mental illness and dangerousness.33 Although perhaps not compelled by the statutory language, the Clark construction was strongly suggested by the statute itself, which incorporates the civil commitment criteria as the standard for determining the release of insanity acquittees. Ga. Code Ann. § 27-1503 (1978).
Clark, however, clearly does not view Georgia law as requiring or providing for a hearing on the issue of dangerousness at this point in the proceedings. Not only does the court refer to the “good cause” showing as “summary commitment,” 266 S.E.2d at 475, but also the Georgia court’s discussion of the due process rights of ac-quittees reveals that no “hearing” (i.e., in the sense of taking new or additional evidence) takes place when the court “immediately” inquires into the sanity of the acquit-tee. Clark held that the acquittee’s due process right to a hearing would be waived if the acquittee did not file an application for release after the 30-day evaluation commitment period ends. Clark also held that the acquittee was entitled to receive notice of his right to a hearing “after expiration of the 30-day period.” Id. If Georgia law contemplated a hearing at the time of the “good cause” showing on the acquittee’s present sanity, then there would have been no need for Clark to find a waiver of the due process right to a hearing in failure to apply for release after the 30-day period. Similarly, if a hearing were held at the time of the good cause showing, then receiving notice of a right to a hearing after the expiration of the 30-day period would be superfluous. The record in the instant case confirms that in practice the Georgia courts do not hold a hearing at this stage in the proceedings. 501 F.Supp. at 1055 n.4.
Thus, at the time of summary commitment, the trial court, under Georgia law, does not hold a hearing or take new evidence, but nonetheless is required to determine if there is reasonable cause to believe that the defendant presently meets the criteria for civil commitment (i.e., mental illness and dangerousness). Since no new evidence is taken for this showing of “good cause,” the crime for which the insanity acquittee was acquitted constitutes the only basis for the “good cause” finding of the dangerousness criterion. Unless facts, introduced in the criminal trial of the acquit-tee, relating to the acquittee’s crime itself, reveal reasonable cause for believing that the acquittee is dangerous, as that term is precisely defined by the civil commitment statutes, Georgia law does not authorize the commitment of that acquittee under § 27-1503.34 Clark v. State, supra. Accord Whitfield v. State, 158 Ga.App. 660, 281 S.E.2d 643 (1981).
*531Our conclusion that Clark has so construed and limited § 27-1503 to apply only to acquittees acquitted of crimes evidencing dangerousness is reinforced by the reasons Clark advanced for upholding the different release procedures against equal protection attack. Clark justified the difference between release procedures for acquittees committed under § 27-1503 as compared to the procedures provided for other committees on the ground that “insanity acquittees have demonstrated their dangerous propensities by committing prior criminal acts.” 266 S.E.2d at 472, 477. The appellant in Clark, in fact, had been charged with and found not guilty by reason of insanity for murder. This justification for the difference in release procedures is true only if the class of acquittees required to obtain judicial approval for release is confined to those acquittees who have been tried for crimes evidencing dangerousness. See 501 F.Supp. at 1071-72.
The cases cited by Clark in support of the different release procedures for acquittees also involve acquittals for crimes evidencing dangerousness. See 266 S.E.2d at 472, citing Powell v. Florida, 579 F.2d 324 (5th Cir. 1978) (murder); United States v. Ecker, 543 F.2d 178 (D.C.Cir.1976), cert. denied, 429 U.S. 1063, 97 S.Ct. 788, 50 L.Ed.2d 779 (1977) (murder and rape). These cases, moreover, in justifying the different release provisions applied to the acquittees there seeking relief, emphasized the proven dangerousness of the acquittees in question as established by their insanity acquittals for crimes clearly showing dangerousness. Powell v. Florida, 579 F.2d at 331-33; United States v. Ecker, 543 F.2d at 195-98 & n.80.
Thus, we conclude that § 27-1503, as a matter of state law, only authorizes the commitment of insanity acquittees who have been tried for crimes that evidence dangerousness. Not all insanity acquittees can properly be committed pursuant to § 27-1503, and thus not all acquittees who are or have been committed are subject to the release provisions of that statute.
Since § 27-1503, as interpreted by Clark, cannot and does not govern the commitment and release of persons acquitted by reason of insanity of crimes not evidencing dangerousness, the release of these acquit-tees must be controlled by Chapter 88-5, the civil commitment standards. When the chief medical officer of a state hospital determines that an acquittee, who had been charged with a crime not evidencing dangerousness, should be discharged (i.e., the acquittee is no longer mentally ill or dangerous), that acquittee must be released pursuant to Georgia Code § 88-506.6; there is no statutory authority for his continued detention or for requiring court approval for his release. The continued detention of such an acquittee, in the absence of statutory authorization for such restraint, would violate due process of law. To the extent that the commitment, ostensibly under § 27-1503, of any of the class members here is based on an insanity acquittal for a crime not relevant to the dangerousness criterion, those class members will be entitled to appropriate relief on remand. See, infra.
What we have said thus far about the release procedures for acquittees flows directly from state, not constitutional, law. Yet, it would be unrealistic to overlook the fact that Georgia courts inevitably have interpreted § 27-1503 with an eye toward constitutional principles. Clark interpreted § 27 — 1503 in the context of a constitutional challenge, inter alia, to the validity of requiring court approval for the release of acquittees. Clark’s narrowing construction of § 27-1503 accords with sound principles of statutory interpretation — i.e., courts should strive to construe a statute so as to avoid constitutional defects. See, e.g., Cro-well v. Benson, 285 U.S. 22, 62, 52 S.Ct. 285, 296, 76 L.Ed. 598 (1932); Bolton v. Harris, 395 F.2d 642 (D.C.Cir.1968); People v. Lally, 19 N.Y.2d 27, 277 N.Y.S. 654, 224 N.E.2d 87 (1966). Indeed, serious constitutional questions would be raised if the statute had been read broadly to require judicial approval for the release of all acquittees. Although we need not, and do not, decide these questions, we note that the constitu*532tional issues are of significant dimension. We have already noted that the second prong of Baxstrom v. Herold, supra, rejected on equal protection grounds the asserted justification that prisoners were super-dangerous. Part III, supra, Part IV.A.l., n. 31. In addition, the first prong of Baxstrom’s equal protection analysis held that the fact of Baxstrom’s crime did not justify denying him a jury trial at the commitment hearing when a jury was afforded to all civil committees. In Jackson v. Indiana, 406 U.S. 715, 92 S.Ct. 1845, 32 L.Ed.2d 435 (1972), the Supreme Court applied the Baxstrom rationale in striking down differences in release standards applicable to those found incompetent to stand trial, as compared to the standards which civil committees had to satisfy to gain release from mental hospitals. The Court reasoned:
If criminal conviction and imposition of sentence are insufficient to justify less procedural and substantive protection against indefinite commitment than that generally available to all others, the mere filing of criminal charges surely cannot suffice.
406 U.S. at 724,92 S.Ct. at 1851. The Court also quoted with apparent approval from a New York Bar Association Report:
‘The basic and unifying thread which runs through our recommendations is a rejection of the notion that the mere fact of a criminal charge or conviction is a proper basis upon which to build other unnecessary, unprofitable, and essentially unfair distinctions among the mentally ill.’
Id. at 724, n. 4, 92 S.Ct. at 1851, n. 4. Finally, in Humphrey v. Cady, 405 U.S. 504, 92 S.Ct. 1048, 31 L.Ed.2d 394 (1972), the Court noted the constitutional difficulties inherent in using past criminal conduct as a basis for distinguishing between classes of mentally ill persons with respect to the procedures available to such persons, when those procedural differences “are in no way limited by the nature of the defendant’s crime.” Id. at 511, 92 S.Ct. at 1053. The Court suggested further that the state might seek to justify the discrimination against mentally ill offenders “in some special characteristics” of those offenders that relate to the particular difference in procedures. This sort of justification, however, would be undermined and “the equal protection claim would seem especially persuasive if . . . [the procedural protections a committee receives were determined] by the arbitrary decision of the state to seek his commitment under one statute rather than the other.” Id. at 512, 92 S.Ct. at 1053.
Grave constitutional problems would be raised had it been necessary to subject to the Baxstrom-Jackson-Humphrey scrutiny a classification composed of all insanity ac-quittees. The underinclusiveness and over-inclusiveness which concerned the court below would have to survive an uncomfortably close comparison to the sort of ill fit of legislative means to ends which Baxstrom’s second prong condemned. The challenge to a statute’s underinclusiveness and overin-clusiveness takes on a particular poignancy when raised in a context, such as in Bax-strom and here, where everyone in any event will receive a hearing prior to commitment. Since we have already held in Part II that the State is obligated to provide a hearing in the first place, there would be very little burden on the State and no interference with its legitimate interest if the State were required to identify at that hearing those for whom the State’s particular interest was applicable, e.g., the super-dangerous or the criminally dangerous. It is hard to conceive of a strong State interest in classifying all insanity acquittees as criminally dangerous or all prisoner-committees as super-dangerous when, for every member of these groups the State must in any event hold a hearing on the very issue of whether the particular acquittee or prisoner is dangerous at all.35
*533We realize that none of the above Supreme Court cases concern release procedures for insanity acquittees. Relying on qualifying language in Baxstrom that equal protection does not require identical treatment but only that the difference in procedures be relevant to the purpose of the classification, this circuit and the District of Columbia circuit have held that some differences in release provisions for some acquittees will survive equal protection scrutiny. Powell v. Florida, 579 F.2d 324 (5th Cir. 1978); United States v. Ecker, 543 F.2d 178 (D.C.Cir.1976), cert. denied, 429 U.S. 1063, 97 S.Ct. 788, 50 L.Ed.2d 779 (1977). Nevertheless, we are aware of no decision which has upheld such differences in release procedures in a context where the court clearly viewed the statute as intending to treat all acquittees (including those acquitted for crimes not evidencing dangerousness) differently from all M.H.C. committees. Cf., e.g., Powell v. Florida, supra (murder)36; United States v. Ecker, supra (rape and murder). See also Lee v. Kolb, 449 F.Supp. 1368 (W.D.N.Y.1978) (murder) (dictum); Allen v. Radack, 426 F.Supp. 1052 (D.S.D.1977) (murder) (dictum).37 The fore*534going cases have upheld such differences in release procedures on the grounds that the acquittees challenging the release provisions, unlike M.H.C. committees, manifested their dangerousness in conduct which had been found beyond a reasonable doubt to constitute the elements of a crime (i.e., the criminal conduct of such acquittees in itself evidenced dangerousness); thus focusing only on insanity acquittees whose crimes evidence dangerousness, as did the Clark v. State court. At least one federal court has held that requiring judicial approval for the release of all insanity acquittees but not for the release of any M.H.C. committees violates equal protection. Reynolds v. Neill, 381 F.Supp. 1374, 1384-85 (N.D.Tex.1974) (three-judge court), vacated on other grounds sub nom. Sheldon v. Reynolds, 422 U.S. 1050, 95 S.Ct. 2671, 45 L.Ed.2d 703, reinstated and enforced, 404 F.Supp. 1004 (1975). See State v. Clemons, 110 Ariz. 79, 515 P.2d 324 (1973); Wilson v. State, 259 Ind. 375, 287 N.E.2d 875 (1972); People v. McQuillan, 392 Mich. 511, 221 N.W.2d 569 (1974).
The Supreme Court has not yet directly addressed this issue, but the foregoing discussion suggests the substantial constitutional problems involved in a requirement of court approval for the release of all acquittees. The Clark court’s interpretation of § 27-1503 must therefore be understood in light of the significant equal protection problems that would arise from a broader reading of the statute.
Moreover, in addition to the foregoing equal protection concerns, an expansive construction of the Georgia statute to include all insanity acquittees (including non-dangerous persons acquitted of crimes such as criminal tax fraud, for example) would raise serious due process concerns, and the Clark opinion must be viewed against this background also. In assessing the constitutionality of the statutory scheme in Humphrey v. Cady, supra, the Court observed that “Wisconsin conditions such confinement not solely on the medical judgment that the defendant is mentally ill and treatable, but also on the social and legal judgment that his potential for doing harm, to himself or others, is great enough to justify such a massive curtailment of liberty.” 405 U.S. at 509, 92 S.Ct. at 1052 (emphasis added) (footnote omitted). Several federal courts have interpreted this language to imply that due process requires a finding of dangerousness, “great enough to justify such a massive curtailment of liberty,” as a prerequisite for involuntary confinement for the precautionary purpose of protecting the safety of the committee or others. E.g., Lynch v. Baxley, 386 F.Supp. 378, 390 (M.D. Ala.1974) (three-judge court); Bell v. Wayne County General Hospital, 384 F.Supp. 1085, 1096 (E.D.Mich.1974) (three-judge court); Lessard v. Schmidt, 349 F.Supp. 1078, 1093 (E.D.Wis.1972) (three-judge court), vacated on other grounds, 414 U.S. 473, 94 S.Ct. 713, 38 L.Ed.2d 661 (1974), reinstated and enforced, 379 F.Supp. 1376 (E.D.Wis.), vacated on other grounds, 421 *535U.S. 957, 95 S.Ct. 1943, 44 L.Ed.2d 445 (1975), reinstated, 413 F.Supp. 1318 (E.D. Wis.1976). This view proved prescient in light of the Supreme Court’s landmark decision in O’Connor v. Donaldson, 422 U.S. 563, 95 S.Ct. 2486, 45 L.Ed.2d 396 (1975), which held that the custodial detention of one who is neither dangerous nor incapable of caring for himself violated that person’s constitutional rights. Dangerousness (or in appropriate cases inability to care for oneself) thus was held to be a sine qua non of non-treatment-based confinement.38
In the face of this recent trend of federal decisions, the Georgia legislature and supreme court have carefully restricted the scope of § 27-1503. The Georgia civil commitment standards mandate that a person may not be confined unless he presents “a substantial risk of imminent harm to himself or others as manifested by either recent overt acts or recent express threats of violence which present a probability of physical injury to himself or other persons . ... ” Ga.Code Ann. §§ 88-501(v), 504.2, 506.2 (Supp.1981). Clark has imported these standards into the determination of whether an insanity acquitee should be detained for the 30-day evaluation period. Thus, no person may properly be committed in Georgia without a finding of dangerousness made according to a standard that would appear to be as stringent as the most exacting statement of the due process requirement in the above-cited federal cases. This statutory standard of dangerousness and the judicial construction of the procedures for commitment of acquittees have thereby avoided the constitutionally suspect practice of automatically committing all insanity ac-quittees, for at least observational confine*536ment,39 based solely on their commission of the elements of a crime, regardless of the nature or dangerousness vel non of that crime.
We have outlined these constitutional questions, not for the purpose of resolving them, but to illustrate the context in which, and background against which, the Clark case interpreted § 27-1503. We fortunately need not, and do not, decide these constitutional questions because we understand the Georgia Supreme Court to have construed § 27-1503 narrowly so that such substantial constitutional challenges are avoided. The district court considered the statute to have a substantially broader application than we perceive it to have and thus addressed a much different constitutional question than we now have before us. We therefore vacate that portion of the district court’s order concerning the requirement of judicial approval for release, and we now turn to a more limited issue of the constitutionality of the narrowed Georgia statute.
3. Equal Protection Analysis of the Statute as Construed by Clark.
Given Clark’s narrowing construction of § 27-1503, the classifications created by Georgia law are, on one hand, acquit-tees charged with crimes indicating dangerousness, who must obtain judicial approval for release, and, on the other hand, all other acquittees and M.H.C. committees, who can be released by the chief medical officer of the state hospital. We think that when viewed in this fashion the Georgia statutory scheme survives equal protection scrutiny.
The courts and the commentators have recognized that commitment and release decisions comprise “mixed questions,” involving not only medical considerations but also a legal and social judgment, i.e., a balancing of the interests of the patient against the risk of harm, including criminal harm, to the public. Humphrey v. Cady, 405 U.S. at 509, 92 S.Ct. at 1052; Powell v. Florida, 579 F.2d at 333; Goldstein and Katz, Dangerousness and Mental Illness: Some Observations on the Decision to Release Persons Acquitted by Reason of Insanity, 70 Yale L.J. 225, 230-31 (1960); Reid, Disposition of the Criminally Insane, 16 Rutgers L.Rev. 75,122-23 (1961). See Diamon, The Psychiatric Prediction of Dangerousness, 123 U.Pa.L.Rev. 439 (1974). There are, therefore, grounds in general for judicial participation in decisions regarding the release of committed persons; these reasons assume a heightened significance where the release decision concerns an insanity acquittee whose dangerousness has been evidenced by criminal conduct. Insanity acquittees who are subject to the release provisions of § 27-1503 are persons for whom there has been a judicial determination that the mental illness and dangerousness which triggered their commitment manifested itself in the commission, as established by proof beyond a reasonable doubt, of the elements of a crime. The criminal overtones inherent in such an insanity acquittee’s circumstance provides a rational basis for requiring judicial approval before release, and justifies the State’s unwillingness to abdicate these release decisions to the psychiatric community. Matters involving crime are peculiarly within the province of the judiciary and therefore make it eminently reasonable to delegate these release decisions to a judge, rather than leaving them in medical hands. The special experience and expertise of the courts in dealing with crime and persons who have engaged in criminal conduct justify vesting in the judiciary release decisions concerning insanity acquittees whose potential for harm has been shown by the commission of the elements of a dangerous crime. This choice of decision makers by the State is no less rational than would be a system whereby the release of paranoid schizophrenic com*537mittees might be assigned to a medical official specializing in that disorder while the release decisions concerning manic depressive committees might be delegated to a specialist in that psychosis. “Equal protection does not require that all persons be dealt with identically, but it does require that a distinction made have some relevance to the purpose for which the classification is made.” Baxstrom v. Herold, 383 U.S. at 111, 86 S.Ct. at 762. The distinction Georgia has drawn, as circumscribed in Clark, is substantially related to the purpose that the classification is intended to serve.
When the Georgia statute is viewed in light of the judicial gloss provided by Clark, the constitutional problems under Baxstrom become manageable. Here, unlike Bax-strom, the classification made by the State, as refined by Clark, is not overinclusive. The only persons required to obtain court approval for release are those acquittees who, because their dangerousness and mental illness have been revealed through and manifested in criminal conduct, implicate the State’s interest in invoking the judiciary’s expertise in criminal matters. There are no members of this class of acquittees for whom requiring court approval for release does not serve the State’s interest.
Similarly, Jackson and Humphrey are distinguishable from the present case because here, in contrast to those cases, the procedural difference at issue is precisely limited by the nature of the defendant’s crime. Indeed, whether the judicial release procedure is invoked depends solely on the nature of the acquittee’s crime (i.e., whether or not that crime evidences dangerousness). The mere criminal charge by itself is not the basis for the procedural difference; rather the dangerousness evidenced by the criminal conduct, i.e., a “special characteristic of . . . [the class in question] which ... render [s] a ... [judicial] determination uniquely [ Appropriate ...,” Humphrey v. Cady, 405 U.S. at 512, 92 S.Ct. at 1053, justifies the difference here. Here, unlike Jackson and Humphrey, there is a finding of dangerousness, and of dangerousness of the particular type, i.e., criminal, which implicates the State’s legitimate interest. In sum, the means Georgia has chosen are substantially related to the end the State legitimately seeks to achieve.
Although the statute remains somewhat underinclusive even after being focused by Clark, this underinclusiveness is not substantial enough to undermine the statute’s rationality. Admittedly, the state’s interest would also seem to require court approval for the release of M.H.C. committees whose commitment was triggered by dangerous conduct which constituted the elements of a crime. Yet there are differences between such committees and the acquittees subject to § 27-1503 that did not exist for the present and former prisoners who were treated differently in the Baxstrom case. In Baxstrom, a former prisoner released from prison the day before his commitment was entitled to have a judge decide whether he should be sent to a maximum security hospital; the hospital assignment of a prisoner committed on the last day of his sentence was left entirely to the discretion of administrative officials. This difference, which amounted to merely when the State instituted commitment proceedings, cast serious doubt on the state’s avowed purpose of segregating from ordinary committees those with “proven criminal tendencies” or those whose extreme dangerousness warranted maximum security.
Here, acquittees covered by the statute differ in a substantial way from those M.H.C. committees who, despite seeming to implicate the State’s interest, are not subject to the special release provisions. Ac-quittees have had the benefit of a judicial hearing (i.e., the criminal trial itself) where the fact that they committed the elements of a crime has been shown beyond a reasonable doubt. In the summary commitment proceeding immediately following the verdict of not guilty by reason of insanity, a finding by the court is made linking this criminal conduct to the civil commitment criteria of mental illness and dangerousness. See Clark v. State, 266 S.E.2d at 475. M.H.C. committees whose commitment is based on conduct arguably constituting the elements of a crime, have not received the *538benefit of a judicial hearing, replete with the full panoply of criminal procedural safeguards (including the beyond a reasonable doubt burden of proof), to establish the fact of their criminal conduct. The civil commitment hearing does not address whether the committee has engaged in conduct that constitutes the elements of a crime; rather, that hearing focuses on whether the committee is mentally ill and dangerous, a completely different inquiry. M.H.C. committees might indeed have a strong due process objection if, without the opportunity for a hearing to establish or to refute charges of past criminal conduct, they were nonetheless subjected to consequences flowing from such alleged conduct. See Specht v. Patterson, 386 U.S. 605, 87 S.Ct. 1209, 18 L.Ed.2d 326 (1967).
Even though Georgia’s legislative scheme may suffer from some underinclusiveness, it can still survive equal protection scrutiny. The legislature is not required to address every facet of a problem to satisfy the demands of equal protection. The legislature may single out what it believes to be the most serious aspects of an evil and focus its attention in those directions. Buckley v. Valeo, 424 U.S. 1, 105-06, 96 S.Ct. 612, 675-676, 46 L.Ed.2d 659 (1976); Katzenbach v. Morgan, 384 U.S. 641, 657, 86 S.Ct. 1717, 1727, 16 L.Ed.2d 828 (1966). The legislature may also approach a problem of considerable depth and complexity one step at a time. Michael M. v. Superior Court of Sonoma County, 450 U.S. 464, 481 n.13, 101 S.Ct. 1200, 1210 n.13, 67 L.Ed.2d 437 (1981) (Stewart, J., concurring); Fullilove v. Klutznick, 448 U.S. 448, 485-86, 100 S.Ct. 2758, 2778, 65 L.Ed.2d 902 (1980). Where a statute is challenged solely on the basis of its underinclusiveness, that statute does not violate the Equal Protection Clause unless its underinclusiveness is substantial enough to create doubts about the statute’s basic rationality. We are not prepared to say that the omissions of the Georgia statute so significantly undermine its rationality. Georgia could well have concluded that insanity acquittees whose criminal conduct evidences dangerousness constitute the core of the state’s concern over mentally ill criminal offenders.
Of course, Georgia might have eliminated any problems of underinclusiveness (as well as overinclusiveness) by requiring the court at the initial commitment hearing for both M.H.C. committees and insanity acquittees to take evidence and expressly to make findings on whether the potential committee has exhibited dangerousness in the form of criminal conduct.40 These findings could automatically invoke the continuing jurisdiction of the court to oversee the release of such committees, be they M.H.C. committees or insanity acquittees. Since the State must hold these commitment hearings anyway,41 this would appear to be a practical, efficacious, and fair means of achieving the State’s legitimate purposes. We are not, however, empowered to strike down the instant statute because, with the benefit of hindsight, we can think of a wiser, fairer, or more precise method than the State has chosen. Legislating in the mental health area is fraught with complexities and uncertainties, and the State must be allowed some latitude in addressing these problems. So long as the State’s classification bears a fair and substantial rational relationship to its legitimate interest, and the State has not encroached upon protected interests, we must uphold its laws against constitutional attack.
We therefore hold that the release provisions of § 27-1503, as that statute has been circumscribed by Clark, do not violate the Equal Protection Clause. Our discussion of the scope of the State law at issue here, however, may make some relief appropriate *539for members of the appellee’s class. To the extent that a class member may have been summarily committed following an insanity acquittal for a crime that does not evidence dangerousness, § 27-1503 does not and cannot control the continued confinement of that person. The only authority for confining such a person would appear to be the civil commitment statutes, Georgia Code Chapter 88-5, and the continued commitment of such acquittees must conform with the procedures specified in that Chapter, as well as other parts of this opinion. The district court on remand may fashion such relief as it deems appropriate for those class members whose commitment is not properly governed by § 27-1503 and who meet the civil commitment criteria for release, Ga. Code Ann. § 88-506.6 (1979}.
We should note, however, that the contours of the relief contemplated under our view of the Georgia law differ from the initial order entered by the district court. The district court held that Georgia may require court approval for the release of “individuals who have committed acts which are considered violent crimes: murder, rape, armed robbery, arson, aggravated assault, and other offenses of like nature.” 501 F.Supp. at 1072. We find that Georgia has extended its judicial release provisions to the insanity acquittee who “presents a substantial risk of imminent harm to himself or others as manifested by” his criminal conduct “which present[s] a probability of physical injury to himself or to other persons.” To be subject to the strictures of § 27 — 1503, the insanity acquittee’s criminal conduct must (1) reveal that the acquittee presents a substantial risk of imminent harm to himself or others, and (2) present a probability of physical injury to himself or others. We note that while violent crimes may evidence the sort of dangerousness contemplated by this standard, crimes not ordinarily considered “violent” may also do so (e.g., statutory rape). We leave to the district court to decide what sort of criminal conduct evidences this dangerousness requirement and the district court should be guided in its determination by the application of Georgia’s civil commitment criteria and the constitutional considerations detailed in this opinion.
B. Burden of Proof at the Release Hearing
Having sustained Georgia’s procedure for judicial approval of the release decision with respect to the limited class described in Part IV.A. of this opinion, we now consider the burden of proof standard at the release hearing. The State defends the Georgia scheme which places the burden on the insanity acquittee to prove by a preponderance of the evidence that he is no longer mentally ill and dangerous. The district court held, on equal protection grounds, that Georgia could not place on insanity acquittees the burden of proving their fitness for release, while simultaneously requiring the State to bear the burden of proof in the comparable release (i.e., the “recommitment”) hearing for M.H.C. committees. The district court held that the Equal Protection Clause did not require identical treatment, concluding that there should be a shared burden of proof at the release hearing for certain insanity acquit-tees, i.e., those who had committed violent crimes.
We address first the State’s argument that the district court erred in refusing to honor the Georgia procedure requiring that insanity acquittees bear the burden of proof. We reject the State’s argument on due process grounds following Addington v. Texas, 441 U.S. 418, 99 S.Ct. 1804, 60 L.Ed.2d 323 (1979), and on equal protection grounds.
Except for any possible differences between the initial commitment hearing and the release hearing, the reasons described in Part III of this opinion, dealing with the proper burden of proof at the initial hearing, would foreclose the State’s argument. The State might have argued 42 that once it *540has proved, at the initial commitment hearing, that the insanity acquittee is mentally ill and dangerous, due process would no longer require the State to marshall its evidence and bear the burden of proof at subsequent release hearings.43 Although the argument has superficial appeal, closer analysis reveals its weakness. The thrust of the argument is the administrative burden on the State. However, the statutory scheme makes it clear that Georgia contemplates periodic review of the insanity ac-quittee’s need for continued institutionalization. Georgia Code § 27-1503(b) creates the opportunity for annual release hearings, at the instance of either the person confined or the superintendent of the hospital. The statutory scheme also contemplates continuous treatment.44 Because such continuous treatment and review are being conducted in any event, and because evidence at any release hearing would come largely from the psychiatrists and other hospital staff already participating in the insanity acquit-tee’s program, the administrative burden of producing evidence cannot be considered a significant problem. When a hearing will be held in any event, whether the State must respond to a preponderance of the evidence showing by the acquittee or itself produce clear and convincing evidence would appear to have no impact on the administrative burden faced by the State. In either case the State no doubt will come forward with virtually the same proof. This is especially true given the uncertainties of litigation in this area, the subtlety of the difference in degree between the two standards for the burden of proof, and the significance of the interests at stake. In this situation the allocation of the burden of proof may affect the outcome of the hearing, but it will not affect the State’s task at that hearing. Moreover, the fact that Georgia’s scheme with respect to M.H.C. committees provides not only continuous treatment and review, but also provides that the State will bear the burden of proof at release hearings suggests that the administrative burden is no obstacle.
A variation on the preceding argument might also have been made by the State. Dictum in Waite v. Jacobs, 475 F.2d 392, 400 (D.C.Cir.1973), suggests that it would be reasonable to place the burden of proof on an insanity acquittee at the release hearing, because his mental illness and dangerousness had been convincingly proved at the initial commitment hearing, and because it is rational to presume the continuation of this status. We also reject this variation. The argument runs counter to Powell v. Florida, 579 F.2d at 330, which emphasized the due process problems implicated in relying upon findings of a previous mental state as a substitute for findings of a present mental state. The problems would be exaggerated in this context because continuation of a previous mental state is even less rational here where the patient has been undergoing continuous medical treatment since the time of the initial commitment hearing.
The Waite v. Jacobs argument also runs counter to Jackson v. Indiana, 406 U.S. 715, 738, 92 S.Ct. 1845, 1858, 32 L.Ed.2d 435 (1972) (“At the least, due process requires that the nature and duration of commitment bear some reasonable relation to the purpose for which the individual is commit*541ted”), and O’Connor v. Donaldson, 422 U.S. 563, 575, 95 S.Ct. 2486, 2493, 45 L.Ed.2d 396 (1975). (“[E]ven if his voluntary confinement was initially permissible, it could not constitutionally continue after that basis no longer existed”). See also Addington v. Texas, 441 U.S. at 426, 99 S.Ct. at 1809. If the State has no interest in confining a person who is no longer mentally ill and dangerous, and if the administrative burden on the State is no obstacle, we cannot discern a sufficient difference between the initial hearing and the release hearing to justify a placement of the burden of proof different from that required by Addington. The significant considerations — on the one hand, the interest of the patient in protection against the significant deprivation of his liberty, and on the other hand the interest of the community in protection against the dangerous tendencies of mentally ill persons — remain identical.
Finally, the argument based upon the Waite v. Jacobs dictum must be rejected because of equal protection considerations. As noted above, the argument relies upon a presumption that the status of an insanity acquittee — i.e., mentally ill and dangerous — continues from the time of the initial commitment hearing to the time of the release hearing. Even if evenhanded use of such a presumption were permissible, which we doubt for the above-discussed reasons, it would violate equal protection principles because the presumption is not applied in release hearings for M.H.C. committees notwithstanding the identical positions of the two classes. See supra, Part I.
The preceding discussion, together with the discussion in Part III of this opinion, reveals that there is no rational basis to justify placing the burden of proving the release criteria on insanity acquittees while simultaneously requiring the State to bear the burden of proof at release hearings for M.H.C. committees. We thus conclude that equal protection principles preclude the State from requiring that insanity acquit-tees bear the burden of proof.
Accordingly, we hold on due process grounds and on equal protection grounds that Georgia cannot place the burden of proving the release criteria on insanity ac-quittees while simultaneously providing that the State must bear the burden of proof with respect to the release of M.H.C. committees.
In rejecting the State’s argument and concluding that there is no justification for Georgia’s attempt to place the burden of proof at the release hearing on the insanity acquittee, our analysis of both the due process considerations as expounded in Adding-ton and the equal protection considerations lead inevitably to the conclusion that the State must bear the burden of proof by clear and convincing evidence. Our discussion above and our discussion in Part III demonstrate our inability to find any satisfactory way to distinguish the due process analysis of Addington v. Texas, supra, and demonstrate the absence of any rational basis to justify a different burden of proof in determining the release of insanity acquittees as opposed to M.H.C. committees. Accordingly, we modify the district court’s shared burden ruling 45 and conclude, on both due process and equal protection grounds, that the State must bear the burden of proving the release criteria by clear and convincing evidence.
C. Frequency of Applications for Release
As discussed in Part II of this opinion, due process and equal protection con*542siderations require Georgia to afford insanity acquittees a state-initiated commitment hearing at the expiration of the 30 day evaluation period. Georgia Code § 27-1503(b) provides that insanity acquittees can apply for release only at one year intervals. By contrast, M.H.C. committees can be released at any time when and if the hospital determines that the patient no longer meets the commitment criteria, and can apply for a judicial release hearing notwithstanding the hospital’s adverse recommendation six months after the initial commitment and annually thereafter. Ga.Code Ann. §§ 88-506.2 and 88-506.5 (Supp.1981). The district court held that limiting the frequency of release applications by insanity acquittees lacks any rational basis, and ordered that insanity acquittees be afforded the opportunity to apply for release with the same frequency as permitted M.H.C. committees. We agree.45.1
The State’s brief devotes only three sentences to this issue, Brief of Appellant at 34, and offers no justification at all for the different treatment of the two classes. Nor can we conceive of any justification. The district court observed that there could be no conceivable reason to deny a prompt hearing when the hospital has determined that an insanity acquittee no longer meets the commitment criteria, in light of the fact that the hospital’s recommendation will constitute the primary, though not conclusive,46 evidence at the hearing, and in light of the fact that a M.H.C. committee would be automatically released in such a situation. We agree. Nor can we conceive of any reason to deny insanity acquittees the opportunity enjoyed by M.H.C. committees to apply for release six months after the initial commitment. See Bolton v. Harris, 395 F.2d 642, 652 (D.C.Cir.1968). Accordingly, we hold on equal protection grounds that insanity acquittees must be afforded the opportunity to apply for release with the same frequency as M.H.C. committees.
V. CONCLUSION
In sum, we hold that the Georgia presumption of continuing insanity as applied only to insanity acquittees and not to M.H.C. committees denies insanity acquit-tees the equal protection of the laws. We hold also that the Due Process and Equal Protection Clauses require Georgia, after the expiration of the 30-day evaluation period, to hold a state-initiated commitment hearing for all insanity acquittees committed under Georgia Code § 27-1503, and to bear the burden of proving by clear and convincing evidence at that hearing that the acquittee meets the Mental Health Code commitment criteria. With 'respect to the judicial release provision of § 27-1503, we conclude that, as that provision has been narrowed by the Georgia courts to apply only to acquittees who were charged with crimes evidencing the civil commitment criterion of dangerousness, it passes constitutional muster. Lastly, we hold with respect to release hearings that due process and equal protection compel the State to bear *543the burden of proving by clear and convincing evidence that the acquittee still meets the civil commitment criteria, and that equal protection requires that all acquittees be afforded the opportunity to apply for release with the same frequency as permitted M.H.C. committees.
AFFIRMED IN PART, VACATED AND MODIFIED IN PART, AND REMANDED.
. Ga.Code § 27-1503 provides in pertinent part:
(a) . .. . If such verdict of acquittal [by reason of insanity] is returned by the jury, the court shall retain jurisdiction over the person so acquitted and shall immediately inquire into the sanity of the person at the time of acquittal and, upon a showing of good cause by the prosecutor, may defer ruling upon the same and order such person to be confined in a State mental hospital, to be selected by the Department of Human Resources for a period of not less than 30 days. A person committed to the Department of Human Resources pursuant to *513this section shall not be released from confinement unless and until the court which committed him, after notice and hearing, shall find and determine that such person does not meet the criteria for civil commitment under Chapter 88-5 or 88-25, as now or hereafter amended. Nothing in this section contained shall prevent the transfer of such person from one State hospital to any other State hospital by the Department of Human Resources or the transfer of such patient to a hospital in another state in the manner provided by law, upon order of the superior court in the county from which he was committed, or in which he is detained.
(b) An application for the release of a person who has been committed to the Department of Human Resources under subsection (a) upon the ground he does not meet the civil commitment criteria under Chapter 88-5 or 88-25, as now or hereafter amended, may be made to the superior court of the county from which he was committed, either by such person or by the superintendent of the State hospital in which the said person is confined. No hearing upon such application shall be allowed until the person committed shall have been confined for a period of not less than 30 days from the date of the order of commitment. If the finding of the court is adverse to releasing such person on the ground that such person meets the civil commitment criteria under Chapter 88-5 or 88-25, as now or hereafter amended, a further application shall not be heard by the court until one year has elapsed from the date of hearing upon his last preceding application.
. This summary is taken in substantial measure from the district court’s opinion, including liberal verbatim quotation. A more detailed statement of the facts is contained in that opinion. Benham v. Edwards, 501 F.Supp. 1050, 1054 56 (N.D.Ga.1980).
. Georgia recognizes the M’Naghten and the delusional compulsional standards as a complete defense to criminal culpability. Ga.Code § 26-702 provides:
A person shall not be found guilty of a crime, if at the time of the act, omission, or negligence constituting the crime, such person did not have mental capacity to distinguish between right and wrong in relation to such act, omission, or negligence.
Ga.Code § 26-703 provides:
A person shall not be found guilty of a crime when at the time of the act, omission, or negligence constituting the crime, such person, because of mental disease, injury, or congenital deficiency, acted as he did because of a delusional compulsion as to such act which overmastered his will to resist committing the crime.
. “Good cause” has been defined as “a reasonable cause to believe that the defendant presently meets the criteria for civil commitment.” Clark v. State, 266 S.E.2d at 475. Clark does not appear to contemplate that this “good cause” showing amounts to a hearing on this issue, see infra, Part IV.A.2., and the court below found that in practice no hearing is held. 501 F.Supp. at 1055 n. 4.
. The district court refused to refer to the classifications at issue in this case as “criminal committees” (i.e., persons committed following insanity acquittals) and “civil committees” (i.e., persons committed pursuant to the Georgia Mental Health Code). The court considered these labels to be somewhat misleading, to lack constitutional significance, and to be likely to obscure the similarities of circumstance and legal status of these two groups. Since the *514plaintiffs challenge the validity of treating these two groups differently, reliance on these labels might tend to beg the legal questions at hand. 501 F.Supp. at 1051 n. 5. We agree. The district court adopted the terminology of “insanity acquittees” to designate those persons committed under Ga.Code § 27-1503, and “M.H.C. committees” to designate those committed pursuant to the Georgia Mental Health Code. We will follow this usage.
. Georgia Code Chapter 88-5 pertains to the hospitalization and treatment procedures for the mentally ill. Georgia Code Chapter 88-25 pertains to the mentally retarded. An individual is subject to involuntary commitment under Chapter 88-5 if he is mentally ill and (1) presents a substantial risk of imminent harm to himself or others as manifested by either recent overt acts or recent expressed threats of violence which present a probability of physical injury to himself or other persons, or (2) is so unable to care for his own physical health and safety as to create an imminently life-endangering crisis. Ga.Code Ann. § 88-50 l(v) (Supp. 1981). But see infra, Part IV.A.2. at n. 33.
The criteria for the involuntary commitment of the mentally retarded are scattered throughout numerous provisions of the Georgia Code. See Ga.Code Ann. §§ 88-2502(d), 88-2504(f), 99-3303(c) (1981 and Supp.1981).
The mentally retarded are afforded rights and procedures similar to those granted to M.H.C. committees. See 501 F.Supp. at 1056 n. 8. Compare Ga.Code Ann. Ch. 88-25 with Ga. Code Ann. Ch. 88-5. The district court found that at the time of trial only one insanity ac-quittee was in a hospital for the mentally retarded. 501 F.Supp. at 1056 n. 8. In the interest of simplicity, this court, as did the district court, will refer solely to the mentally ill, and not the mentally retarded, provisions throughout this opinion.
. On appeal the State has challenged the district court’s ruling only on the merits, and has not suggested any error with respect to the district court’s findings of irreparable injury, harm to the defendants and public interest, the other factors which must be demonstrated to obtain injunctive relief. Moreover, at oral argument the parties agreed that no further evidence will be offered and that it is appropriate to treat the instant preliminary injunction as a permanent injunction.
. The State has not appealed the rulings in Part III.D. of the district court’s opinion, 501 F.Supp. at 1070-71, holding that insanity ac-quittees can be committed only if they meet the criteria for civil commitment, i.e., mental illness and dangerousness, and holding that the evidentiary rules and standard which apply in hearings for M.H.C. committees also apply in hearings for insanity acquittees.
. Since we must review several equal protection points decided by the district court, we ordinarily might have to address the threshold question of what level of judicial scrutiny is appropriate for evaluating the constitutionality of the challenged provisions. San Antonio Independent School District v. Rodriguez, 411 U.S. 1, 93 S.Ct. 1278, 36 L.Ed.2d 16 (1973). The court below expressly declined to determine what level of scrutiny the facts of this case would invoke, finding that the Georgia statute could not survive even the most minimal and deferential equal protection scrutiny. 501 F.Supp. at 1054 n.l. We also conclude, but for reasons different from those expressed by the district court, that we need not identify the precise form of scrutiny mandated by the Equal Protection Clause.
The Supreme Court’s traditional approach in analyzing an equal protection challenge to a statute might be perceived as invoking one of two distinct modes or levels of judicial scrutiny. According to this view, where a statute involved racial or other “suspect" classifications, or, as subsequently expanded, fundamental interests, the Court has applied “strict scrutiny” and has required that the classification in question be necessary to achieve a compelling governmental interest, that the means chosen to achieve this interest be carefully tailored, and that no less drastic means be available. See San Antonio Independent School District v. Rodriguez, 411 U.S. at 16-17, 93 S.Ct. at 1287-88. Where suspect classes or fundamental interests were not present, and especially in cases concerning economic regulation, the Court has employed “minimum rational basis analysis” and has upheld the classification if it was in any way rationally related to the purpose of the legislation. See Railway Express Agency v. New York, 336 U.S. 106, 110, 69 S.Ct. 463, 465, 93 L.Ed. 533 (1949). This perception of the Court’s equal protection jurisprudence has given rise to considerable criticism of and dissatisfaction with the rigidity and inflexibility inherent in the so-called “two-tier” approach to equal protection analysis. Harris v. McRae, 448 U.S. 297, 341-44, 100 S.Ct. 2671, 2707, 65 L.Ed.2d 784 (1980) (Marshall, J., dissenting); Craig v. Boren, 429 U.S. 190, 210-11 & n.*, 97 S.Ct. 451, 464, 50 L.Ed.2d 397 (1976) (Powell, J., concurring); Id. at 211-12, 97 S.Ct. at 464-465 (Stevens, J., concurring); Vlandis v. Kline, 412 U.S. 441, 458, 93 S.Ct. 2230, 2239, 37 L.Ed.2d 63 (1973) (White, J., concurring); San Antonio Independent School District v. Rodriguez, 411 U.S. at 97-110, 93 S.Ct. at 1329-1336 (Marshall, J., dissenting). One commentator has observed that the scrutiny of the upper tier of this dichotomy is “ ‘strict’ in theory and fatal in fact,” while the lower tier consists of “minimal scrutiny in theory and virtually none in fact.” Gunther, The Supreme Court, 1971 Term — Forward: In Search of Evolving Doctrine on a Changing Court: A Model for a Newer Equal Protection, 86 Harv.L.Rev. 1, 8 (1972).
Arguably, some of the Court’s most recent decisions embody an emerging middle level or intermediate form of equal protection scrutiny. See, e.g., Cabell v. Chavez-Salido, - U.S. -,-, 102 S.Ct. 735, 739, 70 L.Ed.2d 677, 684 (1982); Fullilove v. Klutznick, 448 U.S. 448, 517-19, 100 S.Ct. 2758, 2794, 65 L.Ed.2d 902 (Marshall, J., concurring); Lalli v. Lalli, 439 U.S. 259, 265, 99 S.Ct. 518, 523, 58 L.Ed.2d 503 (1978); University of California Regents v. Bakke, 438 U.S. 265, 356-62 & n.30, 98 S.Ct. 2733, 2781-85 & n.30, 57 L.Ed.2d 750 (1978) (Brennan, J., concurring and dissenting); Trimble v. Gordon, 430 U.S. 762, 767, 97 S.Ct. 1459, 1463, 52 L.Ed.2d 31 (1977); Craig v. Boren, 429 U.S. at 197-99, 97 S.Ct. at 456-57; Id. at 210-11 & n.*, 97 S.Ct. at 463-64 & n.* (Powell, J., concurring); Gunther, supra; Wilkinson, The Supreme Court, the Equal Protection Clause, and the Three Faces of Constitutional Equality, 61 Va.L.Rev. 945 (1975). Some have suggested that in cases involving distinctions among classes of mentally ill persons, the Court has brought to bear some form of this heightened scrutiny; Professor Gunther has termed this analysis “vigorous rational basis scrutiny” or *516the traditional standard applied “with new bite.” Gunther, supra, at 22, 31. See Jackson v. Indiana, 406 U.S. 715, 92 S.Ct. 1845, 32 L.Ed.2d 435 (1972); Humphrey v. Cady, 405 U.S. 504, 92 S.Ct. 1048, 31 L.Ed.2d 394 (1972); Baxstrom v. Herold, 383 U.S. 107, 86 S.Ct. 760, 15 L.Ed.2d 620 (1965); Gunther, supra, at 20-32. The Supreme Court, however, has not yet expressly declared what standard of review should apply in this area, and most courts confronting equal protection claims analogous to those raised in this case have not paused to define the appropriate level of scrutiny. See, e.g., Jackson v. Foti, 670 F.2d 516 (5th Cir. 1982); Powell v. Florida, 579 F.2d 324 (5th Cir. 1978); United States v. Ecker, 543 F.2d 178 (D.C.Cir. 1976), cert. denied, 429 U.S. 1063, 97 S.Ct. 788, 50 L.Ed.2d 779 (1977); Bolton v. Harris, 395 F.2d 642 (D.C.Cir. 1968). See generally Schweiker v. Wilson, 450 U.S. 221, 231 & n.13, 101 S.Ct. 1074, 1081 & n.13, 67 L.Ed.2d 186, 195 & n.13 (1981).
We conclude that we need not decide what level of scrutiny the Constitution mandates, except that we do hold that traditional strict scrutiny is not required. In the areas where we find that the Georgia law violates the Equal Protection Clause, we base this decision on either binding Fifth Circuit precedent, see, infra, Part II, or direct application of specific principles clearly established in Supreme Court cases (e.g., Baxstrom's holding that the past crime of a mentally ill offender does not justify different treatment from all other committed persons based on the notion that the crime reveals a heightened degree of dangerousness), see, infra, Parts III and IV.B., or an a fortiori conclusion that the statute here is less rational than were comparable provisions that were struck down by binding precedent, see infra, Parts I and IV.C. In Part IV.A., where we uphold the requirement of judicial approval for the release of acquittees, we find that the Georgia law withstands all but the strictest scrutiny. Although several commentators have urged that strict scrutiny be applied to classifications concerning involuntary commitment, e.g., Kirschner, Constitutional Standards for Release of the Civilly Committed and Not Guilty by Reason of Insanity: A Strict Scrutiny Analysis, 20 Ariz.L.Rev. 233 (1978); Note, Commitment Following Acquittal by Reason of Insanity and the Equal Protection of the Laws, 116 U.Pa.L.Rev. 924, 933 (1968); Note, Mental Illness: A Suspect Classiñcation?, 83 Yale L.J. 1237 (1974), we decline to do so. The Supreme Court in evaluating classifications concerning involuntary commitment in the Baxstrom, Humphrey, and Jackson cases has employed the language and basic analysis, albeit probably with “a sharper focus,” Craig v. Boren, 429 U.S. at 210 n.*, 97 S.Ct. at 464 n.* of the traditional rational basis test. See German and Singer, Punishing the Not Guilty: Hospitalization of Persons Acquitted by Reason of Insanity, 29 Rutgers L.Rev. 1011, 1012 n. 4 (1976). Although these cases do hold in favor of the persons challenging the classifications, as a court of appeals we are reluctant to extend strict scrutiny to an area where the Supreme Court, despite three opportunities to do so, has not invoked this standard of review. The deprivation of liberty alone has not been deemed sufficient to necessitate strict scrutiny for classifications pertaining to involuntary confinement. Marshall v. United States, 414 U.S. 417, 94 S.Ct. 700, 38 L.Ed.2d 618 (1974). Although insanity acquit-tees may be somewhat isolated from the political process, they would appear to be no more of a suspect class or a “discreet and insular minoritfy],” United States v. Carotene Products Co., 304 U.S. 144, 153 n.4, 58 S.Ct. 778, 783 n.4, 82 L.Ed. 1234 (1938), than women, aliens, or illegitimate children. We therefore hold that our standard of review is not strict scrutiny. Since where we find the Georgia statute to be unconstitutional our decision is based on controlling precedent and where we sustain the Georgia procedure we do so assuming arguen-do that heightened, but not strict, scrutiny applies, we conclude that no further articulation of the applicable scrutiny is necessary under the circumstances of this case.
. Throughout this opinion we use the term “mental illness and dangerousness” as shorthand for the Georgia commitment criteria. The criteria for commitment of M.H.C. committees is set out in Ga.Code Ann. § 88-50 l(v) (Supp. 1981), which in pertinent part authorizes the commitment of “[a] person who is mentally ill and ... who presents a substantial risk of imminent harm to himself or others as manifested by either recent overt acts or recent expressed threats of violence which present a probability of physical injury to himself or to other persons.” The same substantive criteria apply in the commitment of insanity acquittees. Clark v. State, 266 S.E.2d at 475; Ga.Code Ann. § 27-1503 (1978).
. We evaluate the Georgia presumption assuming arguendo that the Georgia Supreme Court did create a presumption of continuing dangerousness. Otherwise, with respect to the dangerousness criterion, there would be no basis at all for Georgia’s scheme which denies a mandatory initial commitment hearing and shifts the burden of proof to insanity acquit-tees. Indeed, grave due process questions would be raised if Georgia confined persons for whom there is no basis to believe that they are dangerous. See infra, Part IV.A.2.
. The insanity acquittal finding is either: (1) that, the defendant did not have the mental capacity to distinguish right from wrong (the M’Naghten standard codified in Georgia Code § 26-702); or (2) that the defendant, because of mental disease, acted as he did because of a delusional compulsion which overmastered his will (the delusional compulsion standard codified in Georgia Code § 26-703).
. See infra, Part III.
. There is further reason to doubt the validity of a presumption of continuing insanity or mental illness for acquittees. Even though the acquittee was legally insane at the time he committed the elements of a crime, since that time, and prior to commitment, he has become sufficiently mentally competent to stand trial or enter a plea. See Drope v. Missouri, 420 U.S. 162, 95 S.Ct. 896, 43 L.Ed.2d 103 (1975); Pate v. Robinson, 383 U.S. 375, 86 S.Ct. 836, 15 L.Ed.2d 815 (1966). The strength of the presumption of continuing mental illness or insanity is thereby further weakened by this intervening period of sufficient mental competence to stand trial. See Bolton v. Harris, 395 F.2d 642, 646-52 (D.C.Cir.1968).
. See infra, Part IV.A.
. But see Jones v. United States, 432 A.2d 364, 371, 373-74 (D.C.App.1981) (en banc), cert, granted,-U.S.-, 102 S.Ct. 999, 71 L.Ed.2d 292 (1982) (ipse dixit assertion that presumption of continuing dangerousness and mental illness “is both reasonable and valid.”).
. The appellees and the dissent have suggested that Georgia Act of April 16, 1982, Act No. 1439, House Bill 1240, may have mooted the issue of whether the State is constitutionally required to initiate a commitment hearing for insanity acquittees. We disagree, and find that notwithstanding Act No. 1439, this case presents a live controversy which must be decided by this court. Act No. 1739 amends Georgia Code § 27-1503 by adding a new subsection (e) that requires a mandatory hearing prior to the commitment of an insanity acquit-tee. This hearing will be initiated by the State after the expiration of the 30 day evaluation period and within 30 days of receipt by the prosecuting attorney of an evaluation report. Act No. 1439, however, does not become effective until July 1, 1982, and the present version of § 27-1503 does not contemplate either the court-ordered evaluation report which triggers the state-initiated hearing or the state-initiated hearing itself. Thus, all persons who have been or will be acquitted by reason of insanity prior to July 1, 1982, have been and will be subject to the present version of § 27-1503 and will not receive the benefit of either the court-ordered evaluation report or the state-initiated hearing triggered by this report.
. Another possible deficiency in the State’s argument is that the Georgia presumption is rebuttable and would seem in any event to require a hearing and opportunity for rebuttal.
. Clark v. State, 266 S.E.2d at 475.
. Powell v. Florida, 579 F.2d at 327; Jackson v. Foti, 670 F.2d at 521.
. Powell v. Florida, 579 F.2d at 327; Jackson v. Foti, 670 F.2d at 518.
. See e.g., Fasulo v. Arafeh, 173 Conn. 473, 378 A.2d 553 (1977).
. Counsel is automatically appointed for an M.H.C. committee, unless the patient indicates in writing that he does not desire to be represented by counsel. Ga.Code Ann. § 88-506.-2(a)(2) (Supp.1981). The patient’s right to refuse the appointment of counsel is also subject to the discretion of the court. Id. § 88-501(u). Therefore, when an M.H.C. committee files a written waiver of the commitment hearing, which is permissible under § 88-506.-2(a)(5), he has the benefit of representation by counsel. See also id. § 88-505.3(a). By contrast, the Georgia procedure does not provide for appointment of counsel to assist an insanity *521acquittee in deciding whether to file an application for a commitment hearing; the acquittee only receives notice of his right to counsel. Clark v. State, 266 S.E.2d at 475.
. Plaintiffs did not challenge the 30 day evaluation period in the court below, 501 F.Supp. at 1057 n. 10, or on appeal.
. The District of Columbia Court of Appeals has also confronted this issue. Jones v. United States, 432 A.2d 364 (D.C.1981) (en banc), cert, granted, - U.S. -, 102 S.Ct. 999, 71 L. Ed.2d 292 (1982). We find that court’s discussion of this issue unsatisfactory and its reasoning unpersuasive. The court approved placing the burden of proof at release proceedings on the insanity acquittee because the acquittee was “the one to advocate the fact of his past insanity” and because “the presumption that a mental state continues is a reasonable one.” Id. at 374. We have already rejected the use of such a presumption to justify differences in the burden of proof for insanity acquittees and for M. H.C. committees. See supra, Part I. The court also concluded that the difference in the allocation of the risk of erroneous commitment for M.H.C. committees and insanity acquittees was “minimized by the fact that when the acquittee bears the burden, he need prove entitlement to release by only a preponderance of the evidence, the lowest standard of proof, whereas when the government bears the burden [i.e., in civil commitments], it must prove insanity and dangerousness by a higher standard, namely that of clear and convincing evidence.” 432 A.2d at 375. The fallacy in this reasoning becomes apparent when, for example, you compare the results under the different burdens of proof when the evidence is in equilibrium. Under the standard applicable to M.H.C. committees, the committee would be released because the government failed to muster clear and convincing evidence; the standard applicable to insanity acquittees would require that the acquittee remain confined because the *523acquittee failed to establish his fitness for release by a preponderance of the evidence.
Ironically, the court in Jones in another part of its opinion cited Addington and observed that while Addington “did not deal with commitment of persons found not guilty by reason of insanity, a § 301(d) commitment [i.e., the District of Columbia’s procedure for committing insanity acquittees] is actually a civil, not a criminal proceeding and Addington’s reasoning should apply with equal force to commitment of acquittees.” Id. at 370, n. 10. Despite this acknowledgment of Addington’s relevance, the Jones court did not address the argument that Addington requires the government to bear the burden of proving by clear and convincing evidence that the acquittee meets the criteria for commitment.
. We note also that this implicit assumption by the Warren v. Harvey panel would seem to conflict with the conclusions of a previous Second Circuit case, United States ex. rel Schuster v. Herold, 410 F.2d 1071, 1085-86 (2d Cir. 1969). See also United States ex. rel Schuster v. Vincent, 524 F.2d 153 (2d Cir. 1975).
. Even if the factual premises underlying these fears of abuse of the insanity defense and “error” in insanity acquittals were true, these considerations still could not justify incarcerating an acquittee in a mental hospital. What interest does the State have in committing to a mental hospital a non-mentally ill person who erroneously was found not guilty by reason of insanity? If a person is not mentally ill, the State has no basis for committing that person to a mental hospital. See O'Connor v. Donaldson, 422 U.S. 563, 95 S.Ct. 2486, 45 L.Ed.2d 396 (1975); Ga.Code Ann. §§ 88-501(v), 88-506.-2(c). Commitment under these circumstances can only be punishment for the original criminal act or for escaping conviction and this would raise grave double jeopardy problems.
. Dicta in Lynch v. Overholser, 369 U.S. 705, 715, 82 S.Ct. 1063, 1069, 8 L.Ed.2d 211 (1962), also suggests, possible legislative purposes relating to minimizing calculated abuses of the insanity defense and other considerations indicative of punitive purpose. The Supreme Court dicta is inapposite because Georgia has rejected the notion that any punitive purpose remains after an insanity acquittal. We note also that the Supreme Court dicta was made without consideration of any constitutional implications.
. United States v. Brown, 478 F.2d 606 (D.C. Cir.1973), is a pre-Addington case based on equal protection grounds which employs an analysis very similar to that in Warren v. Harvey. Brown did hold, however, that equal protection requires the same burden of proof for insanity acquittees and civil committees, once insanity acquittees have served a reasonable time, not to exceed the maximum sentence for the offense of which they were acquitted. 478 F.2d at 612. The Brown holding that there is no justification for a different burden of proof after a reasonable penal term, but that there is justification initially, necessarily assumes a lingering punitive purpose after an insanity acquittal. The fact that the District of Columbia statutory scheme uniquely contemplates some continuing criminal responsibility after an insanity acquittal, was expressly recognized in Brown, 478 F.2d at 611, and has been a factor, either explicitly or implicitly, in other cases in the D.C. Circuit. United States v. Ecker, 543 F.2d 178, 199 (D.C.Cir. 1976), cert. denied, 429 U.S. 1063, 97 S.Ct. 788, 50 L.Ed.2d 779 (1977); Waite v. Jacobs, 475 F.2d 392, 396 (D.C.Cir. 1973); Dixon v. Jacobs, 427 F.2d 589, 601-3 (D.C.Cir. 1970) (concurring opinion). But see Jones v. United States, 432 A.2d 364, 376 (D.C. 1981) (en banc), cert. granted, -U.S. -, 102 S.Ct. 999, 71 L.Ed.2d 292 (1982) (rejecting this reasoning in Brown and Waite and holding that confinement of an insanity acquittee under the D.C.Code is not based on any punitive considerations). The district court in the instant case noted that this D.C. Circuit view runs counter to the overwhelming authority that an insanity acquittee is not responsible and subject to punishment, 501 F.Supp. at 1065, and suggested that the D.C. Circuit’s view might have resulted from their pioneering decisions expanding the scope of the insanity defense beyond the traditional M'Naghten standard. 501 F.Supp. at 1066. In any event, Georgia has rejected the notion that any continuing punitive purpose remains after an insanity acquittal, and therefore United States v. Brown is inapposite in assessing the Georgia scheme.
. We recognize that there is language in United States v. Ecker, supra, supporting the intuitive assumption that insanity acquittees are more dangerous than civil committees. For several reasons we decline to accord significant weight to this language in Ecker. First, the Ecker court v/as expressly influenced by the violence of Ecker’s particular crime, a brutal rape and murder of a twenty-four year old *526senatorial aide in her own apartment. 543 F.2d at 181, 197, 198 n. 80, 199. That is, Ecker himself probably was more dangerous than civil committees generally. Second, the Ecker court simply assumed that insanity acquittees generally are more dangerous than civil committees generally, without offering any supporting reasoning and without citing any supporting evidence. As indicated in the text below we remain unconvinced that insanity ac-quittees generally are in fact more dangerous than civil committees generally. Third, the actual holding of Ecker was expressly premised on the fact that Ecker’s period of confinement was within the maximum sentence he could have received, i.e., life, thus invoking the D.C. Circuit’s rationale that criminal responsibility remains after an insanity acquittal which justifies different treatment. 543 F.2d at 199. Therefore, the Ecker language supporting the assumption that insanity acquittees are super-dangerous is dictum. Moreover, the Ecker holding — that equal protection principles permit the requirement of judicial approval before the release of insanity acquittees notwithstanding the fact that civil committees may be released by decision of the hospital alone — need not be understood as presupposing that insanity acquittees are more dangerous than M.H.C. committees. In Part IV.A. of this opinion, we also sustain a similar differential in the Georgia scheme, but we do so on grounds unrelated to the assumption that insanity acquittees are more dangerous and unrelated to the punitive purpose rationale. Fourth, we note that the Ecker court focused only on the first prong of Baxstrom, which did not squarely address the validity of the assumption of super-dangerousness, and did not discuss the more relevant second prong of Baxstrom v. Herold, which rejected on equal protection grounds an analogous assumption that Baxstrom’s past crime established that he was too dangerous for the civil hospitals in which civil committees generally were confined. Finally, we note that the Ecker court did not consider the potential over-inclusiveness and underinclusiveness of the disadvantaged class of insanity acquittees which is identical to that which failed to survive the equal protection analysis of Baxstrom’s second prong.
. We will sometimes use the term “super-dangerous” as a shorthand way to refer to the assumption that insanity acquittees are more dangerous than the average civil committee.
. We note that the district court in the instant case limited the class of insanity acquittees to whom the release provisions would be applied to those who had committed “violent crimes: murder, rape, armed robbery, arson, aggravated assault, and other offenses of like nature.” 501 F.Supp. at 1072. We need not consider whether such a class might in fact be more dangerous than M.H.C. committees generally, because we have no such statutory classification before us. In Part IV.A. of this opinion we modify the district court’s ruling in that regard, and hold that Georgia Code § 27-1503 applies only to acquittees whose crimes implicate the dangerousness criterion for civil commitment. See Ga.Code Ann. § 88-501(v) (Supp.1981). However, our definition of the statutory classification says nothing with respect to whether these acquittees are more dangerous than M.H.C. committees generally. All M.H.C. committees are by definition dangerous, just as are all of the persons who are properly committa-ble under § 27-1503. Thus, persons subject to commitment under the statute before us have *527committed crimes which implicate danger, just as Baxstrom had committed the crime of assault, 383 U.S. at 108, 86 S.Ct. at 761.
. Id. at n. 78. The author of the Note cites authority, including several studies, to support the quoted proposition.
. The second prong of Baxstrom struck down as violating equal protection a statutory scheme whereby administration officials would determine whether prisoners such as Baxstrom would be assigned to a maximum security mental institution while all others civilly committed were entitled to have a judge decide whether they should be sent to a maximum security facility. With regard to considering prisoners (such as Baxstrom) more dangerous than other persons committed, the Court observed, “It may or may not be that Baxstrom is presently mentally ill and such a danger to others that the strict security of a Department of Correction hospital is warranted” (i.e., overinclusiveness — some prisoners were not so exceptionally dangerous as to require maximum security confinement), 383 U.S. at 114-15, 86 S.Ct. at 764. The Court also noted that when former prisoners were civilly committed, even if committed very soon after being released from prison, they, like other civil committees, could not be sent to a maximum security hospital without court approval, despite the fact that such former prisoners might be as dangerous as, or more so than, Baxstrom (i.e., underinclusiveness). Id. at 115, 86 S.Ct. at 764. The Court did not examine how substantial in practice was this underinclusiveness and overinclusiveness. Since all persons (including prisoners) were committed after a judicial hearing, requiring the same treatment for prisoners as accorded all others (i.e., a court decision on whether confinement in a maximum security institution was necessary) would protect both the State and the individual interests while placing only a minimal burden on the state. See Stanley v. Illinois, 405 U.S. 645, 652-57 & n.9, 92 S.Ct. 1208, 1213-1215 & n.9, 31 L.Ed.2d 551 (1972). See also Trimble v. Gordon, 430 U.S. 762, 770-73, 97 S.Ct. 1459, 1465-1466, 52 L.Ed.2d 31 (1977); Carrington v. Rash, 380 U.S. 89, 95-96, 85 S.Ct. 775, 779-780, 13 L.Ed.2d 675 (1965).
. In this case, our conclusion that the Georgia Supreme Court has construed the statute more narrowly than did the district court will not obviate the need to determine the constitutionality of the statute. This limiting interpretation, however, does substantially simplify the constitutional inquiry. See infra.
. While civil commitment may also be predicated upon a finding that the person “is so unable to care for his own physical health and safety as to create an imminently life-endangering crisis,” Ga.Code Ann. § 88-501(v) (Supp. 1981), that such a finding could or would be made following a criminal trial is highly unlikely. Moreover, as the reasoning of Clark makes clear, a special type of dangerousness, and not incapacity, is the basis for the different commitment and release procedures for acquittees. See, infra.
. Of course, acquittees whose criminal conduct does not evidence dangerousness may still be committed, based on other indicia of dangerousness, under Georgia’s civil commitment statutes, Georgia Code Chapter 88-5.
. In the instant case, Georgia might have better achieved its interest in judicial supervision of the release of criminally or especially dangerous mentally ill persons if it simply designated at the initial commitment hearing all committees, whether insanity acquittees or M.H.C. committees, who are criminally or especially dangerous. Those designated persons *533could then reasonably be required to seek judicial approval before release, thus fully serving the legitimate state interests.
. The state court proceedings underlying the constitutional challenge in Powell v. Florida, supra, are remarkably similar to the Georgia commitment standards and procedures as expounded by Clark. Following a jury verdict of not guilty by reason of insanity, the state criminal trial judge committed Powell to a state mental hospital. The then applicable Florida Rules of Criminal Procedure required the trial court to discharge an insanity acquittee unless the court found that the acquittee was “manifestly dangerous.” Fla.R.Crim.P. 3.460 (West 1965) (repealed). An acquittee committed under this rule could only obtain release by an order of the committing court. Id. The trial judge in Powell, without holding a hearing, made the requisite finding of dangerousness expressly based only on “the evidence adduced at trial, the plea of not guilty by reason of insanity, and the verdict entered in this case....” See Powell v. Florida, 579 F.2d at 327. We have no reason to believe that such a summary procedure was not in accordance with prevailing Florida practice — indeed, the Florida Supreme Court denied Powell’s habeas corpus petition, finding no error in his commitment and continued confinement. Powell v. Genung, 306 So.2d 113 (Fla.1975). Thus, in Powell, the special requirement of judicial approval for release, as is true for the comparable Georgia provision as construed by Clark, was triggered only because the petitioner had been acquitted of a crime evidencing dangerousness. Both the language and reasoning used by the Powell panel indicate that the court contemplated a class comprised only of insanity ac-quittees who had been acquitted of crimes implicating dangerousness. 579 F.2d at 331, 333 & n. 10 (“the purpose of this requirement [dangerousness] is to assure that only those whose potential for doing harm is great enough to justify the deprivation of their liberty are committed. See Humphrey v. Cady, 405 U.S. at 509 [92 S.Ct. at 1052] . . . Powell’s dangerousness to society because of his mental condition has been established by anti-social behavior, the killing of another human being”; “[t]he basis for commitment of an individual acquitted by reason of insanity is his ‘dangerousness’ to society”; “[s]ince the defendant in Jackson had never been tried for the crimes of which he was accused, there was no finding and little suggestion that he was dangerous”). Thus the Powell panel did not address the constitutionality of a procedure requiring judicial approval for the release of all insanity acquittees, regardless of dangerousness.
Moreover, the Powell court did not consider the due process problems, described in the text below, inherent in a practice which permits the involuntary confinement of insanity acquittees who are not dangerous. Indeed, there is some hint of such due process concerns in the language quoted above from Powell.
. Bolton v. Harris, 395 F.2d 642 (D.C.Cir. 1968), is not to the contrary. First, the crime for which Bolton was found not guilty by reason of insanity, auto theft, may, depending on the particular facts of that crime, indicate dangerousness. See DuBose v. State, 148 Ga.App. 9, 251 S.E.2d 15 (1978); Brief of Appellant at 34. Second, Bolton’s two-sentence, conclusory sanction of requiring court approval for Bolton’s release is probably dictum. In Bolton’s case, the hospital superintendent did not recommend his release and in fact opposed it. Bolton was thus in the same position in terms of the relief and release procedures available to him as were all other M.H.C. committees for whom the hospital opposed release (i.e., release could only be sought by the patient in a habeas corpus action). The constitutionality of requiring only acquittees to secure court approval for release after the hospital has recommended release was not properly before the court (Bolton was not a class action) nor probably ripe for decision. Third, although Bolton in other respects (e.g., the right of acquittees to a hearing on their present sanity and dangerousness) is a *534landmark case that in part anticipated Jackson, Humphrey, and O’Connor v. Donaldson, 422 U.S. 563, 95 S.Ct. 2486, 45 L.Ed.2d 396 (1975), Bolton’s dictum on release procedure differences must be viewed in light of those subsequent Supreme Court decisions.
Similarly, notwithstanding broad language in Overholser v. Leach, 257 F.2d 667 (D.C.Cir. 1958), Ragsdale v. Overholser, 281 F.2d 943 (D.C.Cir. 1960), and Overholser v. O’Beirne, 302 F.2d 852 (D.C.Cir. 1962), that different procedures for acquittees than for M.H.C. committees are necessary and appropriate because acquittees are “an exceptional class of people,” those cases are readily distinguishable. None of those cases involve a challenge to requiring court approval for the release of all insanity acquittees but not for the release of any M.H.C. committees. Moreover, as in Bolton, the release of the petitioner in each of those cases was opposed, not recommended, by the hospital (i.e., an attack on release procedure differences, even if it had been raised, might not have been ripe for decision). Two of these cases, Leach and Ragsdale, involved crimes (respectively robbery and armed robbery) which almost surely evidence dangerousness and the third, O’Beirne, involved a crime (larceny) which the court expressly viewed as indicating dangerousness. 302 F.2d at 861 and n.18. Finally, the broad language and distinctions advanced in support of different treatment for acquittees and M.H.C. committees probably do not survive Baxstrom, Jackson, Humphrey, and Donaldson, and were called into question and “modified” by Bolton. 395 F.2d at 649, 653.
. Donaldson did not attempt to define the parameters of the constitutional requirement of dangerousness, nor need we in the instant case. Although dicta in the early Supreme Court case of Greenwood v. United States, 350 U.S. 366, 76 S.Ct. 410, 100 L.Ed. 412 (1956), and in the three-judge district court case of Reynolds v. Sheldon, 404 F.Supp. 1004 (N.D.Tex.1975), suggest that danger to property, as opposed to danger to persons, might be sufficient, most federal courts which have addressed the issue since Humphrey and Donaldson have suggested that the danger must relate to persons, and that the dangerousness must be of a substantial degree and imminence to justify involuntary commitment. E.g., Suzuki v. Alba, 438 F.Supp. 1106 (D.Haw.1977), aff’d. in relevant part sub nom, Suzuki v. Yuen, 617 F.2d 173 (9th Cir. 1980); Johnson v. Solomon, 484 F.Supp. 278, 287 (D.Md.1979); Bension v. Meredith, 455 F.Supp. 662, 669-72 (D.D.C.1978); Stamus v. Leonhardt, 414 F.Supp. 439, 451 (S.D.Iowa 1976); Doremus v. Farrell, 407 F.Supp. 509, 514-15 (D.Neb.1975) (three-judge court); Lynch v. Baxley, 386 F.Supp. at 389-91; Bell v. Wayne County General Hospital, 384 F.Supp. at 1095-96, 1102; Lessard v. Schmidt, 349 F.Supp. at 1093; Dixon v. Attorney General of Pennsylvania, 325 F.Supp. 966, 972, 974 (M.D. Pa. 1971) (three-judge court). See also Cross v. Harris, 418 F.2d 1095, 1101-03 (D.C.Cir.1969); Suzuki v. Quisenberry, 411 F.Supp. 1113, 1124 (D. Haw. 1976).
The district court in the Suzuki case, supra, held that dangerousness to property was per se an insufficient constitutional basis for commitment in an emergency or nonemergency situation. The Ninth Circuit declined to decide whether dangerousness to property could ever justify involuntary commitment, but instead held that the Hawaii statute was too broad and unconstitutional “[b]ecause it would permit the state to deprive one of his liberty when he threatens harm to any property,” “regardless of [its] value or significance.” 617 F.2d at 176 (emphasis in original). Citing Humphrey v. Cady, supra, and Addington v. Texas, 441 U.S. 418, 99 S.Ct. 1804, 60 L.Ed.2d 323 (1979), the court reasoned; “Under the current Hawaii definition of ‘danger to property,’ a person could be committed if he threatened to shoot a trespassing dog. The state’s interest in protecting animals must be outweighed by the individual’s interest in personal liberty.” 617 F.2d at 176. The Hawaii statute had defined “dangerousness to property” as “inflicting, attempting or threatening imminently to inflict damage to any property in a manner which constitutes a crime, as evidenced by a recent act, attempt or threat.” Id. (emphasis added).
Even if the state is conceded to have an interest in treating mentally ill offenders whose conduct does not meet these stringent standards of dangerousness, this still may not justify involuntary commitment. The federal courts have evidenced a growing concern that involuntary confinement should be an option of last resort and that the states should seek to achieve their interests through the least drastic or restrictive alternative available. See Lake v. Cameron, 364 F.2d 657 (D.C.Cir.1966) (en banc); Johnson v. Solomon, 484 F.Supp. at 300-10; Stamus v. Leonhardt, 414 F.Supp. at 452-53; Lynch v. Baxley, 386 F.Supp. at 392; Lessard v. Schmidt, 349 F.Supp. at 1095-97. See also Davis v. Hubbard, 506 F.Supp. 915, 934-38 (N.D.Ohio 1980).
. Several courts have held that even temporary, emergency, or observational commitment must also be justified by a showing of the constitutional requirement of dangerousness. Bension v. Meredith, 455 F.Supp. at 668-69; Suzuki v. Alba, 438 F.Supp. at 1109-10; Stamus v. Leonhardt, 414 F.Supp. at 449-51; Bell v. Wayne County General Hospital, 384 F.Supp. at 1091, 1095-96, 1102.
. Similarly, if in future legislation the State asserts an interest in isolating those persons, whether they be M.H.C. committees or insanity acquittees, who are especially dangerous, it could require the court at the initial commitment hearing for both classes to take evidence and make express findings of super-dangerousness, and thus rationally justify invoking a requirement of judicial approval before release.
. See, e.g., Baxstrom v. Herold, supra; Powell v. Florida, supra; United States ex rel. Schus-ter v. Herold, 410 F.2d 1071 (2d Cir. 1969); Lynch v. Baxley, supra; Part II, supra.
. The State’s brief on appeal, Brief of Appellant at 20-25, treated as a single mold the burden of proof issues at the initial commitment hearing and at the release hearing, offer*540ing no arguments to justify a different standard at the two hearings.
. Dicta in Dorsey v. Solomon, 604 F.2d 271, 274, 276 n. 4 (4th Cir. 1979), would seem to support such an argument. However, the statements in the Fourth Circuit case — to the effect that due process permits the placement of the burden of proof on insanity acquittees at the release hearing when the State has shouldered a proper burden of proof at the initial commitment hearing — are dicta because the plaintiffs did not appeal the district court’s ruling on the burden of proof issue. 604 F.2d at 274, 277. The statements are also casual and without any supporting reasoning. Moreover, the district court had justified the difference in procedures available to insanity acquittees by observing that there were “other important counterbalancing rights accorded an insanity acquittee but not a civil committee.” Dorsey v. Solomon, 435 F.Supp. 725, 737, 740 (D.Md. 1977).
. Georgia Code § 88-502.4 provides that each person receiving services for mental illness shall receive care and treatment suited to his needs.
. We noted in Part III, at footnote 29, that the district court had limited the class of insanity acquittees, to whom the reléase procedures would apply, to those who had committed violent crimes. We need not consider how that more limited, and possibly more dangerous class, would affect our due process and equal protection analyses, or whether that class would justify the shared burden of proof found appropriate by . the district court. We need not, and do not, address such issues because we do not view the statute as constitutionally limited in application to only those insanity acquittees who have committed the violent crimes listed by the district court, 501 F.Supp. at 1072. Instead we consider the statute to apply to those insanity acquittees who have committed any crime that implicates the dangerousness criterion for civil commitment. See Part IV.A., supra.
. The appellees have suggested that Georgia Act No. 1439 may have mooted the appellants’ claim that the district court erred in striking down the twelve month limitation on applications for release under Georgia Code § 27-1503(b). See supra, n. 15.2. Act No. 1439 adds a new subsection (f) to § 27-1503, which the appellees read as providing for a release hearing when requested by the acquittee or the hospital superintendent at any time following commitment and, if the initial application is denied, at twelve month intervals thereafter. Even if the appellees have correctly interpreted the new law, and we express no opinion on their interpretation, we find for the same reasons as expressed in note 15.2 that Act No. 1439 does not moot this issue. Moreover, § 27-1503 as amended by Act No. 1439 still does not afford acquittees the opportunity to seek release with the same frequency as can M.H.C. committees. M.H.C. committees can be released at any time by the hospital’s chief medical officer; acquittees, even under the new law, must wait twelve months after an unsuccessful release application to reapply for release, even when the hospital superintendent subsequently recommends the acquittee’s release.
. Powell v. Florida, 579 F.2d 324, 332 (5th Cir. 1978), held that “a judge cannot reject unanimous medical opinion that the insanity acquit-tee is not mentally ill when the judge does not have such power in all other civil commitment cases.”