On January 4, 1989, the Board of Registration in Medicine (board) revoked Dr. Paul E. Bettencourt’s license to practice medicine in the Commonwealth. Bettencourt appealed the board’s decision pursuant to G. L. *222c. 112, § 64 (1988 ed.), and a single justice of this court reported the case to the full court without decision.
The board ruled that, because of inappropriate sexual conduct and other activities with a male patient, Bettencourt (1) committed gross misconduct in the practice of medicine; (2) engaged in gross incompetence in practicing medicine and in conduct that placed into question his competence to practice medicine, in violation of G. L. c. 112, §§ 5(c) and 61 (1988 ed.), and 243 Code Mass. Regs. § 1:03 (5)(a)(3) (1987); (3) was guilty of misconduct in the practice of medicine in violation of 243 Code Mass. Regs. § 1:03 (5)(a)(18) (1987); (4) evidenced a lack of good moral character to practice medicine as required by G. L. c. 112, § 2; and (5) undermined public confidence in the integrity of the medical profession. Bettencourt does not argue that conduct of the type that the board found that he had engaged in is not conduct warranting the various rulings of the board. Instead, he challenges the board’s factual findings that he engaged in the alleged conduct. The fundamental issue for the board, and before that for the hearing officer appointed to hear the allegations, was whether to believe the patient or the doctor.
We conclude that the board’s decision was based in part on findings that are not supported in the record. Furthermore, in deciding whether to believe the patient or Bettencourt, the board improperly disregarded relevant evidence tending to show that Bettencourt would not have engaged in the alleged misconduct. The proceeding must be remanded to the board for further consideration of the credibility issue.
We summarize the board’s findings of fact to the extent necessary for a discussion of the legal questions on which we decide this case. Bettencourt specializes in pulmonary disease and internal medicine. The patient needed a preemployment physical examination and selected Bettencourt in the belief that the doctor was a homosexual with whom the patient would feel comfortable in discussing a “gay related” medical *223problem.1 The patient had his first appointment with Bettencourt on September 23, 1986. The patient already knew that he had tested positive for hepatitis B. The patient returned for examinations on October 9 and October 23. Bettencourt concluded that the patient was a chronic carrier of the hepatitis B virus, would not develop acute hepatitis, and could safely undergo surgery for a condition that had troubled the patient. The surgery was performed by another physician on November 6.
Around Thanksgiving, the patient saw Bettencourt at his office. During that visit, according to the patient and as found by the board, the doctor stroked the patient’s penis, kissed him, and continued until the patient ejaculated. Prior to the next office visit, Bettencourt arranged to have lunch with the patient. The doctor, according to the patient and as found by the board, told the patient that his male lover would be going away for the holidays and invited the patient to his house for dinner on Christmas Eve. After lunch, they went to the patient’s apartment where, as the board found, they had “sexual relations.” The patient spent the night at Bettencourt’s house on December 24, 1986. They kissed, masturbated each other, slept, and then engaged in the same sexual activities in the morning. During the patient’s January 15 and March 17, 1987, office visits, the doctor again kissed and masturbated the patient. During the patient’s May 15 and August 12, 1987, office visits, Bettencourt started to masturbate the patient, but the patient told the doctor that he could not “do this,” and the doctor stopped. Bettencourt did not charge the patient, whose insurance coverage had expired, for the March, May, and August visits.
In September, 1987, the patient told his therapist, who had been counseling him since June, 1987, and also a friend, that his doctor had engaged in sexual conduct with him.
We first discuss the board’s handling of evidence bearing on the risk of infection a person would run in engaging in *224sexual activities, such as those alleged here, with a person who is a carrier of the hepatitis B virus. We then discuss the board’s reliance on the patient’s disclosures to his friend and to his therapist as an aid to resolving the credibility question.
1. Bettencourt argued to the board that the patient’s testimony concerning sexual contacts between them should be disbelieved because he knew the patient was a carrier of the hepatitis B virus. Bettencourt reasoned that, because sexual contact with a carrier of the virus would have exposed him to a significant risk of infection, it was highly improbable that he would have engaged in such contact. In support of this argument, Bettencourt presented evidence, based on a blood test performed in June, 1988, that he had not been exposed to the hepatitis B virus. An expert for Bettencourt testified that he was highly susceptible to infection with that virus. She also testified concerning the degree of risk of infection that would exist in particular circumstances, including kissing and exposure to semen. The board submitted evidence on the degree of risk in the form of an affidavit from a physician whose clinical and investigative specialty was viral hepatitis. That expert’s opinion of the degree of risk, however, was based on irrelevant data concerning the risk of infection from one who had acute hepatitis B, and not from one who, like the patient, was a carrier of the virus (as to whom apparently there were no data concerning the risk).2
*225The board found that hepatitis B could be transmitted by exposure to semen and saliva. It also concluded that there was a twenty-five percent chance that Bettencourt would have contracted the hepatitis B virus. While the statistical evidence considered by the board was weak over-all, the twenty-five percent figure is consistent with testimony by Bettencourt’s expert that the probability of getting hepatitis B through kissing is about twenty-five percent.
The important point, however, was not the actual risk of infection but, rather, Bettencourt’s perception of that risk. The findings on this point were not well presented. The board recited that Bettencourt and his expert witness presented evidence that Bettencourt was aware of the risks of contracting hepatitis B. Unfortunately, there was no explicit finding whether Bettencourt was or was not aware of that risk, but we may fairly assume from the board’s discussion that it found that Bettencourt knew of the risk to some unquantified degree.3
In response to Bettencourt’s argument that his knowledge of the risk of infection would deter him from sexual contact with the patient, the board found that “on those occasions when the [doctor] did come in contact with [the patient’s] semen, he appeared to follow the usual response to avoid infection which involves the rapid and immediate washing of the skin with a good antibacterial or anti-viral soap.” That finding was based on the testimony of Bettencourt’s expert as to the usual response to such an exposure. There was no evidence, however, from any witness (including the patient) that Bettencourt cleaned up in this manner after the alleged incidents. There was a finding that Bettencourt washed up after the incidents in his office, but, even if we were to permit the inference that the appropriate soap was available and that Bettencourt used it in his office, there was no evidence of his washing up after the incidents that allegedly occurred away *226from the office. Moreover, the board’s conclusion that Bettencourt faced a twenty-five percent chance of acquiring hepatitis B through his conduct with the patient was apparently based on testimony concerning the risk of infection through kissing. Any cleanup procedure followed by Bettencourt after exposure to the patient’s semen would detract neither from Bettencourt’s risk of infection from kissing (which, according to the expert testimony, may have been substantially greater than from exposure to semen) nor from his perception of that risk. Thus, the board’s conclusion that Bettencourt always appeared to follow the usual response to avoid infection is not supported by substantial evidence when the record is viewed as a whole. See Fox Ridge Assocs. & Co. v. Assessors of Marshfield, 392 Mass. 652, 653 (1984); New Boston Garden Corp. v. Assessors of Boston, 383 Mass. 456, 466 (1981).
The board also discounted Bettencourt’s argument by concluding that a physician might not avoid unnecessary exposure to the virus because “the sexual impulse might dominate and control a more reasoned approach,” Bettencourt argues that the board was not warranted in reaching such a conclusion in the absence of expert testimony bearing on sexual impulses in the face of known risks of infection. This was a subject on which the trier of fact did not have to have assistance from an expert. It is also a subject, however, that the board should have considered in light of the specific facts of the case, and in terms of what physicians reasonably would do to avoid infection. Relevant facts in this case were Bettencourt’s perception of the risk of infection, and the precautions he undertook to minimize that risk.4 The board, however, made no finding as to Bettencourt’s perception of the risk, and its conclusions concerning precautions undertaken by Bettencourt are unsupported by substantial evidence. Consequently, the board’s conclusion that “the sexual impulse” dominated Bettencourt’s perception of the risk was *227not based on substantial evidence, was arbitrary, and was an abuse of discretion.
At the end of its discussion discrediting Bettencourt’s argument concerning the risk of infection, the board stated that “[f]or resolution of this case, we look to the credibility of the witnesses and the consistency and inherent probability of their testimony.” The board thereafter never discussed the risk of infection issue in its decision, and thus ignored it as a factor in passing on credibility. While the task of assessing the credibility of witnesses is one uniquely within an agency’s discretion (School Comm. of Wellesley v. Labor Relations Comm’n, 376 Mass. 112, 120 [1978]), this court may modify or set aside findings and conclusions that are arbitrary or unsupported by substantial evidence. See G. L. c. 30A, § 14 (7) (1988 ed.); Embers of Salisbury, Inc. v. Alcoholic Beverages Control Comm’n, 401 Mass. 526, 529 (1988). In passing on the credibility issue, the board could not properly ignore (a) Bettencourt’s perception of the risk, (b) the absence of evidence that, if he did what he is alleged to have done, he took appropriate precautions to avoid infection, and (c) the fact that there was no evidence that there was any precaution that could be taken against infection from kissing.
2. The board found substantial corroboration of the patient’s testimony in the testimony of a friend of the patient and of a licensed social worker who had counseled the patient. In September, 1987, the patient told his friend that he had had sexual encounters with his physician. The friend’s testimony concerning the incidents was generally consistent with, but less detailed than, the patient’s testimony. The former therapist’s testimony was more detailed and was also substantially consistent with the patient’s testimony. It was the therapist, and not the patient, who filed a complaint against Bettencourt with the board.5
*228The hearing officer had admitted the evidence only for the purpose of corroborating the patient’s testimony, not to prove the truth of the facts presented, under an analogy to the fresh complaint doctrine that is applicable to out-of-court statements of victims of sexual assault. See, e.g., Commonwealth v. LeFave, 407 Mass. 927, 941 (1990); Glover v. Callahan, 299 Mass. 55, 56-58 (1937). He treated the allegedly corroborative statements as being reliable and admissible as fresh complaints because, in his view, they were timely. The board, however, neither explicitly acknowledged the limited purpose for which the testimony of the friend and the therapist was admitted, nor referred to the fresh complaint principle. It did, however, acknowledge that the timeliness of a complaint might be a factor in determining whether it corroborates a witness’s claim. In this case, the first alleged sexual encounter occurred in late November, 1986, and the last in August, 1987. The first disclosure by the patient was in September, 1987.
The board took a different approach from that of the hearing officer on the matter of timeliness, and it did so on the basis of an opinion that has no support in the record. It stated that the case before it was more closely analogous to one involving a battered or sexually abused child in the care of an abusing adult than to a victim of rape. The board thus *229concluded “that there are many strong forces which can produce, through no fault of the victim, a long delay in revealing events comprising a physician’s sexual abuse of a patient.”
It is not apparent on the record that the patient was in a condition analogous to a child sexually abused by an adult who had that child in his care. The patient was thirty years old when he first saw Bettencourt. Bettencourt was not providing mental health services to the patient, nor was there evidence that the patient was helplessly under Bettencourt’s care as a child might be under an adult’s care. The patient’s supposed dependency on Bettencourt is not supported by substantial evidence. The patient was receiving therapy from someone else and had been doing so for approximately four months before he revealed his alleged sexual encounters with Bettencourt. The board’s conclusion that it could properly consider the allegedly corroborating evidence, because the patient’s relationship to Bettencourt explains the delay in the making of any complaint, lacks evidentiary support.6
3. A judgment shall be entered remanding the proceeding to the Board of Registration in Medicine for further consideration in light of this opinion.
So ordered.
There was no direct evidence that the patient’s perception of Bettencourt as a homosexual was correct.
The discrepancy might have been brought out if this opinion evidence had been presented at the hearing, where objections by counsel for Bettencourt and cross-examination of the expert could have clarified the point. After the board’s decision was issued, the board’s expert gave a second affidavit stating that the board had “misunderstood, misrepresented, and misused” his affidavit as a supporting basis for its decision. Bettencourt sought reconsideration by the board and moved to reopen the record to consider the second affidavit (and other matters). The board denied the motion to reopen, citing a portion of the testimony of Bettencourt’s expert to the effect that it is extremely rare (a 1% chance or so) to contract hepatitis B through contact between semen and intact skin, i.e., where there are no abrasions. The doctor had testified, however, that in the winter his hands are dry because he washes them frequently, and that at times he has hangnails.
There was evidence that before the patient’s operation, Bettencourt took special steps to warn the surgeon and operating room personnel of the hepatitis risk.
Also worthy of consideration would have been evidence of Bettencourt’s concern for avoiding risks of infection generally, his personal habits, and his concern that others might become infected during this patient’s operation.
The counseling service at which the licensed social worker worked had ceased operations by the time of the hearing. The files of the service had been destroyed, and thus the therapist had no notes or records to support his testimony or to provide a basis for cross-examination. The only document the therapist had was his written report to the board. The board may *228have been warranted in declining to infer an improper motive in the destruction of the records, but it did so in the belief that the destruction of the files was a business decision made in the ordinary course of business. There was no direct evidence to support that finding, and the basis for any inference to that effect would need to be explained in any agency decision.
Contemporaneous documentation can have a major influence on credibility determinations. For example, the patient’s testimony that all his appointments but one were late in the day is inconsistent with (a) contemporaneous written records made in the doctor’s appointment book and (b) certain hospital laboratory report summaries showing when particular samples were collected. The board concluded that these “minor discrepancies” did not detract from the patient’s over-all credibility, even though the record would support a finding that six of the nine appointments were before 10 a.m.
We see no error in the unavailability of the patient’s therapist for questioning by Bettencourt’s counsel before the day the therapist testified.
We see no harm or error in the admission of the therapist’s and the friend’s testimony. The rules of evidence generally do not apply to administrative proceedings. See G. L. c. 30A, § 11 (2). A legal issue may arise later, if evidence is relied on in support of the result reached by the agency. The question then is whether that evidence properly contributes to the presence of substantial evidence on the whole record supporting the agency’s findings and rulings.