After Commander Robert McBride, a retired naval officer, suffered a fatal heart attack, his widow and minor children commenced this wrongful death action against the United States under 28 U.S.C. § 1346 (b). They claim the death was proximately caused by negligent failure of the duty doctor at Tripler Army Hospital, Hawaii, to admit McBride to a coronary care unit.

The parties agreed to bifurcate the trial and to litigate first the issue of liability, with the court as trier of fact. At the close of the plaintiff’s evidence, the trial judge granted the government’s motion to dismiss under Rule 41(b), F. R.Civ.P., on the alternate grounds that the government doctor had not been negligent and that the plaintiffs had not established the requisite causal proximity between lack of hospital treatment and McBride's death.

We reverse and remand because we believe the trial judge may have applied an incorrect standard of care on the negligence issue, and because the finding on proximate causation is clearly erroneous.

I.

In January 1968 McBride spent five days in the Tripler Hospital heart ward, undergoing testing to diagnose the source of pain in his lower chest. The tests revealed no evidence of heart disease, but the staff did not rule it out. McBride was released and requested to return in a few weeks for further testing.

Three nights later McBride again experienced severe chest pains. He went to the Tripler emergency room in the early morning hours of January 28. The physician on duty, a young resident, examined him, read the report of the earlier testing, and took an electrocardiogram, a tracing showing the changes in electric potential produced by heart contractions. McBride’s pain had subsided quickly, and throughout the examination his vital signs appeared normal.

The physician told McBride his pain probably resulted from a gastrointestinal disturbance, but that heart disease could not be eliminated as a possibility. He advised admission to the coronary care unit. McBride expressed a preference to return home, saying he felt fine and the previous hospitalization had disclosed no problems with his heart. The doctor allowed McBride to leave, on condition that he return at once should the pain recur. McBride died shortly after reaching home.

The plaintiffs sought to prove that the duty doctor had been negligent in his diagnosis and should have insisted that McBride be hospitalized. The trial judge correctly stated the appropriate standard of conduct against which the doctor’s acts must be measured. See Restatement (Second) of Torts § 299A:

We are unable to determine from the record whether the court actually applied this standard in making its find ing. The judge’s comments and questions suggest strongly that he was judging the doctor on the basis of what one could reasonably expect from a young resident, instead of measuring his acts against a community standard.

The duty doctor acknowledged at trial that he had erroneously interpreted McBride’s electrocardiogram as normal, although in truth it revealed an abnormal pattern. The plaintiffs produced experts who said the general practitioner with ordinary skill would have read the electrocardiogram accurately. The Chief of Cardiology at Tripler testified that many interns and residents would not have recognized the abnormal tracings. The court relied heavily upon this last bit of testimony, saying that the duty doctor’s misinterpretation did not demonstrate negligence if one viewed it against the backdrop of his lack of special training and experience.

The appellants contend that the court erred in holding the duty doctor to a standard of conduct founded upon his personal experience. We agree. Under the American Law Institute formula, the duty of care owed to the patient does not vary according to the doctor’s individual knowledge or education. The wording of the Restatement section indicates that the normal standard will be altered only if the doctor represents to his patients that he possesses special skill. The change depends upon the nature of the representation to the patient, not upon an after-the-fact assessment of what one could expect from a doctor with comparable training and practice.

Commander McBride had the right to expect the quality of care usually found in the medical community and the hospital was obliged to provide physicians who could meet that standard. If the hospital staff had reason to believe that its interns and residents could not reasonably be expected to discern subtle abnormalities in electrocardiogram tracings, then it should not have permitted them to make unaided electrocardiogram analyses.

We do not mean to imply that we believe the duty doctor acted negligently. Indeed, had the trial court made its finding after applying the correct standard, we could not say on this record that the finding would be clearly erroneous. But we do not know whether the judge would have reached the same conclusion if he had applied the standard as we have outlined it. So we must remand to allow him to reconsider the evidence in light of the duty of care imposed by the Restatement section.

II.

The trial court found that the plaintiffs’ proof did not show the essential causal connection between nonadmit-tance and McBride’s death. The expert testimony presented statistical data on two phenomena: the effect of hospital treatment upon a patient’s chance of surviving a heart attack, and the longevity of patients who have survived initial attacks. All medical witnesses agreed that treatment in a coronary care unit significantly enhances the likelihood of successfully surviving a heart attack. Although hospitalization cannot prevent the attack or reduce its severity, treatment does lessen the impact of the heart failure upon the patient’s vital functions. Only 15 percent of those admitted to coronary care die from their first heart attacks. The mortality rate outside hospitals is 30 to 35 percent.

One doctor testified that heart patients’ longevity statistics, which measure reduction in life expectancy after one has lived through a heart attack, are an unreliable indicator of any given patient’s likely life span. These figures lump together all heart patients who have had one heart attack, without regard to the type or severity of their heart disease.

The trial court applied this criticism of the life span figures to the data showing the efficacy of hospital treatment for the initial attack. Nothing in the record justified treating the two sets of statistics identically for purposes of this criticism. No expert testimony supported the view that the coronary care survival rate data are unreliable because they do not differentiate among types of heart disease.

One doctor did testify that the figures given would be inaccurate if applied to patients who experienced a massive congestive heart failure. He said failures of that type account for most of the deaths in coronary care units. The same witness, however, also testified that the pathologist’s report showed that McBride had not died from heart failure of this moi’e severe kind. This doctor, on the basis of specific information about McBride’s condition as well as his knowledge of coronary care unit survival rates, concluded that McBride’s chances of living would have been improved at least 50 percent by admission to the hospital. No evidence contradicted this opinion or undermined its factual basis.

When a plaintiff’s cause of action rests upon an allegedly negligent failure to give necessary treatment, he must show, with reasonable medical probability, that the treatment would have successfully prevented the patient’s injury. He need not prove with certainty that the injury would not have occurred after proper treatment. In most situations the. best medical treatment in the world cannot provide an absolute guarantee of success; medicine is not an exact science in that sense. Yet the absence of positive certainty should not bar recovery if negligent failure to provide treatment deprives a patient of a significant improvement in his chances for recovery. We think the plaintiff demonstrated the requisite reasonable medical probability in this case.

Reversed and remanded for further proceedings consistent with this opinion.

. The law of Hawaii governs this ease. The few cases reported from Hawaii employ a standard of care quite similar to that proposed by the Restatement section, and we feel confident the Supreme Court of Hawaii would follow the American Law Institute formulation. See, e. g., Burrows v. Hawaiian Trust Co., 49 Haw. 351 , 417 P.2d 816 (1966) ; Lyu v. Shinn, 40 Haw. 198 (1953). . “THE COURT: Now, when you say you have to tie everything in with the patient’s history, you’re tying it in with the knowledge of how many years as a GP? “THE WITNESS: I’ve been practicing for about eight years now. “THE COURT: Now, presupposing we had a GP who was in his second year of residency, would you expect him to have the same grasp of these subtle but definite changes on the electrocardiogram?” R.T. 170 “THE COURT: Now, at the time that [the duty doctor] read the X-ray, he had had three months training in the cardiac section of Tripler and had some weekly conferences during the year and a half about after that training. He had been a resident at Tripler for but one year and a half after one year’s internship. Thus, he would not be expected to have the same degree of expertise as either Doctor Jay or Doctor Dodge.” [both general practitioners], R.T. 247 . See, e. g., Johnson v. Vaughn, 370 S.W. 2d 591 (Ky.1963); Agnew v. Los Angeles, 134 Cal.App.2d 433 , 286 P.2d 556 (1955). . See, e. g., Hicks v. United States, 368 F.2d 626 (4th Cir. 1966); Johnson v. Vaughn, 370 S.W.2d 591 (Ky.1963); Agnew v. Los Angeles, 134 Cal.App.2d 433 , 286 P.2d 556 (1955).