Correspondence
Physicians' Experience and Survival in Patients with AIDS
N Engl J Med 1996; 335:349-351August 1, 1996
- Article
To the Editor:
In the report by Kitahata et al. (March 14 issue),1 we were somewhat surprised by the conclusion that the experience of physicians is associated with the length of survival after the development of AIDS in patients seropositive for the human immunodeficiency virus (HIV). We recently showed that patients who had previously received medical care lived a significantly shorter time after the development of AIDS than patients who had an AIDS-defining illness when they first presented, probably because effective intervention delayed the onset of AIDS until immunosuppression was more advanced (as shown by the presence of diseases associated with severe immunosuppression).2
A major omission in the paper weakens the conclusion. The AIDS-defining illnesses are not broken down according to the category of physician's experience. For example, if fewer AIDS-defining illnesses associated with less severe degrees of immunosuppression (Kaposi's sarcoma or esophageal candidosis) were found in one category of experience than in another, the interpretation of survival after the AIDS diagnosis would be altered. In addition, since the authors showed that the most experienced physicians made more use of primary prophylaxis against Pneumocystis carinii pneumonia, one would expect this to be reflected in the AIDS-defining diagnoses presenting to the various groups.3
Effective medical intervention should both improve life and prolong it. As evidence grows that medical intervention delays the onset of AIDS, shorter survival after a diagnosis of AIDS does not necessarily mean that care has been less effective.
3 ReferencesRichard J. Coker, M.R.C.P.
St. Mary's Hospital, London W2 1NY, United KingdomMark C. Poznansky, Ph.D.
St. Mary's Hospital Medical School, London W2 1PG, United Kingdom1
Kitahata MM ,Koepsell TD ,Deyo RA ,Maxwell CL ,Dodge WT ,Wagner EH . Physicians' experience with the acquired immunodeficiency syndrome as a factor in patients' survival. N Engl J Med 1996;334:701-706
Full Text | Web of Science | Medline2
Poznansky MC ,Coker R ,Skinner C , et al. HIV positive patients first presenting with an AIDS defining illness: characteristics and survival. BMJ 1995;311:156-158[Erratum, BMJ 1995;311:356.]
CrossRef | Web of Science | Medline3
Graham NM ,Zeger SL ,Park LP , et al. Effect of zidovudine and Pneumocystis carinii pneumonia prophylaxis on progression of HIV-1 infection to AIDS. Lancet 1991;338:265-269
CrossRef | Web of Science | Medline
To the Editor:
The direction of the effect revealed in the fine study by Dr. Kitahata and her colleagues is probably correct. However, the actual magnitude of the effect of physicians' experience on patients' survival is less than that reported because of what I will call “recursive identification bias.”
Suppose that a cardiologist and a medical student enroll patients with diastolic murmurs in a study evaluating the quality of care. One would find that the medical student's patients do not survive as long. Why not? The medical student tends to identify patients with more obvious murmurs who are sicker and then tends to take less good care of them. The problem with the study design is that the clinicians whose care is being evaluated identify the cases by which they are judged.
In the study by Kitahata et al., survival was measured from the time of the first AIDS-defining diagnosis until death. Consequently, a systematic bias is introduced, because there is likely to be a positive correlation between the physician's clinical experience (with patients with AIDS) and his or her diagnostic aptitude in identifying AIDS-related conditions. To assume that the strategy of enrolling the patients does not introduce this recursive identification bias is to assume that diagnostic skills are distributed either randomly or uniformly.
Control for diagnostic category does not constitute adjustment for the variation in clinicians' abilities to make each particular diagnosis. Similarly, CD4+ cell counts are highly correlated with particular diagnostic categories,1,2 so correction for these counts does not discriminate well between patients in the same category.
The effect of recursive identification bias on the study findings is hard to estimate. However, it seems likely that the large decrease in the relative risk of death among the first three patients cared for by a “naive” physician (from 1.00 for the first patient to 0.73 for the second patient and 0.54 for the third) results largely from having appreciated the diagnosis of AIDS sooner, rather than from having learned so quickly to treat the many complex conditions affecting these patients.
2 ReferencesPeter B. Bach, M.D.
University of Chicago, Chicago, IL 606371
Hanson DL ,Chu SY ,Farizo KM ,Ward JW . Distribution of CD4+ T lymphocytes at diagnosis of acquired immunodeficiency syndrome-defining and other human immunodeficiency virus-related illnesses. Arch Intern Med 1995;155:1537-1542
CrossRef | Web of Science | Medline2
Moore RD ,Chaisson RE . Natural history of opportunistic disease inan HIV-infected urban clinical cohort. Ann Intern Med 1996;124:633-642
Web of Science | Medline
To the Editor:
In his editorial1 on the study by Kitahata et al., Dr. Volberding's comment on monitoring viral load in routine practice is misleading. It is more and more evident that serum RNA titers are good prognostic markers and valuable surrogate end points. However, there is a substantial difference between knowing this information and recommending its use in daily practice. Tests of viral load have analytic aspects that must be validated, and to date no therapeutic intervention has been delineated that is based on RNA titers. We do not yet know any specific RNA threshold at which antiretroviral therapy should be initiated or changed, in the way we know that patients with fewer than 200 CD4+ cells per cubic millimeter need prophylaxis against P. carinii. As soon as we acquire this critical piece of information, the test will have its proper place in the standard of care.
1 ReferencesLuis Guerra-Romero, M.D., Ph.D.
Francisco Parras Vazquez, M.D., Ph.D.
National Plan on AIDS, Ministry of Health, 28071 Madrid, Spain1
Volberding PA . Improving the outcomes of care for patients withhuman immunodeficiency virus infection. N Engl J Med 1996;334:729-731
Full Text | Web of Science | Medline
To the Editor:
Dr. Volberding mentions that there is little coordination in the provision of quality care to patients with HIV. In some of our nation's correctional facilities, the situation is even more disheartening.
On April 7, 1995, an inmate presented with a CD4+ count of 146 cells per cubic millimeter and progressive lower-extremity weakness. Urinary and fecal incontinence followed. He was hospitalized from May 22 to May 27 and then discharged to a jail cell. In the cell he could not walk to the toilet and urinated and defecated in his bed. No member of the medical staff examined him, and he received no medications. He died on June 6.
In June 1995 a second inmate presented with a persistent respiratory syndrome and bilirubinuria. By November, although he reported that “coughing was waking me up at night,” no chest film had been ordered. In December he was hospitalized, and a chest film demonstrated dense opacification of the right upper lobe. His CD4+ count was 150 cells per cubic millimeter, sputum smears revealed many acid-fast bacilli, and sputum cultures grew Mycobacterium tuberculosis. Progressive hepatic and renal insufficiency ensued, and he died on January 24, 1996.
Local jurisdictions should provide for routine peer review and expert consultation in the clinical care of HIV-infected inmates.
Abe M. Macher, M.D.
Eric P. Goosby, M.D.
Office of HIV/AIDS Policy, Washington, DC 20201Author/Editor Response
The authors reply:
To the Editor: Drs. Coker and Poznansky report that patients who presented for medical care at the time of their AIDS-defining illness had less severe diagnoses and therefore survived longer than patients who had received prior care.1 As they point out, effective medical interventions, such as chemoprophylaxis against P. carinii pneumonia, may delay the diagnosis of AIDS in some patients until they are more severely immunosuppressed.
All the patients we studied received care for at least a year before their AIDS diagnoses. Because of space limitations, we did not present data on the severity of the AIDS-defining disease or on CD4+ counts as stratified according to the category of the physician's experience. We reported that the patients cared for by the physicians with more experience were more likely to receive prophylaxis against P. carinii pneumonia. We also found that patients of more experienced physicians were less likely to have P. carinii pneumonia as their AIDS-defining illness. However, there was no systematic increase in the severity of illness with regard to the other AIDS-defining diagnoses among the patients of the more experienced physicians. Nor was there a systematic association between the physician's experience and the CD4+ cell count at diagnosis.
We took into account the degree of underlying immunosuppression at the time of a patient's AIDS diagnosis by adjusting for the CD4+ count. At a given CD4+ count, patients are susceptible to a spectrum of HIV-related conditions of several levels of severity. As a reflection of this, CD4+ count and severity of AIDS-defining illness were independent predictors of survival in our cohort. After we controlled for both factors, the physician's increased level of experience continued to have an observable effect in lowering the patient's risk of death.
Dr. Bach is concerned that physicians with less experience may have been slower to identify AIDS-defining illnesses. Because the final determination of AIDS-defining events at the Group Health Cooperative of Puget Sound is made through a case review involving the primary care physician and the medical record, the date of the AIDS-defining illness is a definitive starting point for follow-up. Our study supports the hypothesis that the physician's experience with AIDS management can lower the risk of death among patients with AIDS. We hope this information will be useful elsewhere in optimizing outcomes for such patients.
1 ReferencesMari M. Kitahata, M.D., M.P.H.
Clare L. Maxwell, M.S.
University of Washington, Seattle, WA 98122-43041
Poznansky MC ,Coker R ,Skinner C , et al. HIV positive patients first presenting with an AIDS defining illness: characteristics and survival. BMJ 1995;311:156-158[Erratum, BMJ 1995;311:356.]
CrossRef | Web of Science | Medline
Author/Editor Response
Drs. Coker and Poznansky note that variables other than the physician's experience may affect outcomes in HIV care. Certainly, imbalances between groups of physicians in the severity of their patients' illnesses could be important. There is little reason to believe, however, that sicker patients in the health maintenance organization studied by Kitahata et al. were preferentially assigned to the most inexperienced group of physicians. If anything, the greater use of prophylaxis against P. carinii pneumonia by the more experienced physicians would argue that they were either treating a population with more advanced disease or that they were seeing more of the difficult infections not prevented by antibiotic prophylaxis.
The comments of Guerra-Romero and Parras Vazquez about RNA testing are true to a point. Assays of HIV RNA titers have yet to be shown prospectively to produce optimal treatment. This may take years, and meanwhile the test is certainly becoming a “standard of care,” at least in the staging of HIV disease. The prognosis is poor for those with higher RNA titers,1 even if the CD4+ cell count is higher than 500 cells per cubic millimeter, a fact that clearly supports considering HIV RNA titers in deciding when antiretroviral therapy should be started.
Macher and Goosby raise a very important point about the practical difficulty of delivering the complex care required to treat HIV disease. If the care system is limited in dealing with even basic management (as is true with many patients, not just those in prison), how realistic is it to expect competence in the use of RNA markers, therapy with protease inhibitors, and the like? We have to recognize these current limitations, and we must improve the training of those engaged in HIV care.
1 ReferencesPaul A. Volberding, M.D.
University of California, San Francisco, San Francisco, CA 94110
