Testing practices in the three clinics evaluated in this study show that physicians test patients with known risk factors to identify HCV infection. The majority of patients with substance abuse (78.1%), alcohol abuse (74.3%), HIV (87.4%), cirrhosis (89.7%), end-stage renal disease (85.1%), ALT elevation (67.2%), or STDs (52.8%) were tested. In addition, a substantial proportion of patients with psychiatric diagnosis (49.7%) were tested. Each of these factors was independently associated with testing in multivariate analysis.
The majority of anti-HCV positive patients identified (73.3%) were born in the high prevalence birth-cohort. Being born in these years was also independently associated with HCV testing and anti-HCV positivity in multivariate analysis. Although testing all patients born in the high prevalence birth cohort may be warranted, evidence suggests that birth cohort-based testing alone would be a less than optimal strategy. First, our data suggest that birth cohort-based testing would fail to identify 26.7% of anti-HCV positive persons, which is similar to the unidentified proportions found when testing only in the birth cohort reported by O'Brien (25.4%),(31
) Armstrong (34.4%),(1
) and Alter (31.3%).(30
) Second, several factors were independently and strongly associated with positivity after adjustment for birth-cohort status including substance abuse, HIV, cirrhosis, and ALT elevation. Lastly, in our study the risk-based screening strategy yielded high rates of anti-HCV positivity in all categories of risk in patients born outside the high-risk birth-cohort. These data suggest that current risk-based screening methods should be continued, and serious consideration should be given to expanding screening recommendations to include birth in the high-risk cohort. Birth cohort testing alone, however, is not recommended.
In this clinic population of an urban academic medical center, the conservative (floor) estimate of the prevalence of hepatitis C antibodies was 4.6%, almost three times the estimated national prevalence.(1
) Our model designed to predict positivity in the untested population estimated a much higher overall prevalence, 7.7%, which is close to the prevalence of 8.3% reported in a similar population by McGinn.(5
) Overall, 39.7% of subjects had been tested. Among those with identified risk (either born in the high prevalence birth-cohort, had a high-risk co-morbidity, or an elevated ALT level), 48.6% had been tested.
It is worth noting that the proportion tested was very high (28.8%) among patients with no identified risk (born outside the high prevalence birth-cohort, no high-risk co-morbidity, and no elevation of ALT) and that the rate of positivity in this group was substantial (3.0%), though less than those with identified risks. Whether a substantial proportion of these tested patients had risk factors not identified through the EMR is not clear. It is also possible that some patients without apparent risk were tested because patients or providers were responding to New York Department of Health efforts, begun in 2004, to raise Bronx community and provider awareness of HCV infection.(39
) Because of the high underlying prevalence of HCV infection (between 4.6% and 7.7%) in this population, universal testing for high-risk urban populations may be more appropriate than the risk-based screening strategy.
This analysis has several important limitations. First, not all patients were tested for anti-HCV so the prevalence we report is an estimate based on risk profile. Second, we utilized an EMR for data collection so we were unable to capture all risks for HCV infection for each patient. Lastly, we did not take into account the temporal relationship between risk factors and HCV tests. It is possible, for example, that a substance abuse diagnosis might have been coded after a HCV test was ordered, and thus we cannot be sure that the diagnosis of substance abuse was present, or in the physician's mind, at the time of testing. Despite these limitations, we were able to uncover a strong relationship between high-risk co-morbidities and physician testing behavior.
In conclusion, we found a very high estimated prevalence of HCV infection in a high-risk urban patient population with a high prevalence of risk factors. We found strong evidence that physicians are using a risk-based screening strategy to identify patients with HCV infection, using known risk factors and other conditions associated with HCV to guide testing. We also found evidence that screening recommendations should be expanded to include the high prevalence birth cohort.