Search the Bridge

Mailing List

Contact Us


Research Briefs

Search Research Briefs
| Browse Research Briefs

Missing the boat. According to a new study by a research team led by Jeffrey H. Samet, a MED professor of internal medicine, physicians in a variety of health-care settings are missing important opportunities to diagnose HIV infection. These opportunities occur when patients come for treatment during the period after they are infected but before they become symptomatic — a period that can last as long as 11 years.

Of the approximately 900,000 people infected with HIV in the United States, 30 percent, or as many as 275,000, are unaware of their infection. These people are not only at risk themselves from associated diseases that can arise from delayed treatment, but they can be responsible for widely spreading the disease.

In the current study, the researchers examined the medical records of patients who began receiving HIV-related medical care between January 1994 and June 2001 in the HIV Diagnostic Evaluation Unit at Boston Medical Center. All the patients had received medical care at BMC at least once prior to their first positive HIV test.

The researchers noted instances where the record reported patient characteristics or conditions considered to be triggers (diseases or behaviors that indicate some degree of risk for HIV infection) and indications that HIV risk had been discussed with the patient and/or HIV testing had been recommended. They analyzed the data by seriousness of the trigger (unequivocal, suggestible, reasonable, and borderline), by the site where treatment occurred (emergency department, urgent care unit, primary care clinic, or specialty clinic), and other characteristics (such as race/ethnicity, age, and gender).

The analysis revealed that HIV testing was recommended by doctors or other medical personnel in fewer than one in five encounters where clinical triggers were documented. There were striking differences in how often HIV testing was suggested at the different treatment sites, from a high of 78 percent in clinics that treat STDs (sexually transmitted diseases) to a low of 9 percent in obstetrics/gynecology clinics. Emergency departments recommended testing in only 12 percent of the visits where triggers were noted.

Although HIV infection is now considered a treatable chronic illness, it is crucial to identify it early, begin treatment, and prevent its spread through unprotected sexual encounters. Heightened awareness within the medical community of clinical triggers, as well as increased rates of testing, would benefit both the patient and society, according to Samet.

If you build it will they come? There is pressure to control medical costs by increasing regulatory controls on the supply of hospital beds. Since studies have shown that patients in areas that are “over-bedded” generally do not get better quality treatment or have better outcomes overall, it is important to ask if reducing the number of beds in an area can reduce costs without reducing quality.

In making such regulatory decisions, it is essential to know whether having more beds actually leads to more hospitalizations, and conversely, whether fewer available beds result in fewer hospitalizations. According to SMG Associate Professor Erol Peköz, studies that indicate these things use an inherently flawed methodology and could give the false impression of a correlation between bed availability and hospitalization rates when none actually exists. He and his colleagues Michael Shwartz and Joseph Restuccia, both SMG professors, are developing new statistical tools to better analyze hospitalization and outpatient data to more effectively estimate this correlation and help policy-makers reach better decisions.

Estimating the number of beds available in a given geographical area is tricky. In theory, it is possible for anyone to get treatment at any hospital. In reality, health plans and physician preferences limit patients’ access to hospitals. Traditional methodology tries to assign beds as follows: if a certain percentage of the hospital beds at, say, Massachusetts General Hospital (MGH) are occupied by patients from a given geographical area (Lowell, say), then that percentage of MGH’s beds are assigned to Lowell. The number of beds assigned to Lowell in a given year includes actual hospital beds in Lowell, plus beds from hospitals throughout Massachusetts that were occupied by Lowell residents during that year. This assignment methodology by itself creates a misleading statistical correlation between supply (assigned beds) and utilization (number of patients filling these beds) for Lowell, since the number of beds assigned to Lowell will change from year to year to mirror changes in Lowell’s utilization, even though no beds are actually being physically created or eliminated anywhere.

The researchers are developing a method based on a statistical formulation called a Bayesian hierarchical method. It allows them to distinguish between “true” utilization patterns and hospital-induced effects. They will apply the method to two different data sets. The first, inpatient and outpatient data for Medicare recipients in Massachusetts in 1997, will disentangle the effects of three factors: “true” utilization, hospital-induced effect, and an area-specific effect reflecting the amount of diagnosed disease in the area (managed both in-hospital and outpatient). The second data set, of hospital discharge statistics from 1997 to 2000, provides data from a variety of payers and over multiple years.

In preliminary studies the researchers have tested whether high medical admission rates are correlated with high rates of inappropriate admissions and whether high rates for a particular procedure (such as cardiac catheterization) are correlated with inappropriate use of that procedure. They failed to find a connection in either case. A third study, still under way, is attempting to sort out the roles of physicians‚ practice styles (whether physicians tend to admit patients or treat them as outpatients), and the prevalence of disease in a community in determining rates of hospital admissions.

By using new tools to better understand what drives variability in hospitalization rates, the current study, funded by the Agency for Healthcare Research and Quality, will help communities use health-care dollars more efficiently.

"Research Briefs" is written by Joan Schwartz in the Office of the Provost. To read more about BU research, visit


15 May 2003
Boston University
Office of University Relations