In this 43-month double-blind, randomized trial, we found that an automated EHR-based alert increases CHB testing in API individuals by more than twofold. Of those who completed HBsAg testing, 4.9% were positive. Thus as a proof of concept, we have shown that EHR-based alerts provide incremental benefit towards increasing CHB screening. We have previously shown that electronic messages to health care providers placed within 24 h prior to a clinic visit were effective at increasing CHB testing for at-risk individuals. However, this approach was resource intensive, could not be easily automated, and interrupted workflows14. In this regard, the automated alert described herein is an improvement on our previous method.
Automated electronic medical alerts have been validated as clinical decision support tools in a variety of disease states15,16,17,18. However, unlike chronic hepatitis C or other common preventive health care measures where screening is based on age and is accurately captured in the EHR, mass screening for CHB requires more personalization as it is based upon country of nativity, which is typically not recorded in the EHR. Thus, we used a novel approach to identify patients who were at risk for CHB using imputed API race or ethnicity based upon surname, language preference, or country of origin. Additional strengths of this study include its randomized design and length of blinding. Many similar studies measure the impact of EHR interventions before and after implementation, which can be confounded by other changes in practice and knowledge over time. The nearly 4 years of implementation of the alert in Cohort 1 also demonstrates that there is a potential for saturation of an alert and that uptake may not continue in a linear fashion. Interestingly, saturation appeared to have been reached much more rapidly in Cohort 2, which may be due to alert “burn out” among healthcare providers.
Results of multivariable logistic regression analysis showed that patients with Medicaid insurance or those who self-pay for medical care had lower odds of completing HBV testing compared to privately insured patients and those with Medicare. This healthcare access disparity may explain the lower than expected detection of HBsAg positives in the alert group. In theory, patients with Medicare and Medicaid should be at higher risk for CHB. Medicare patients are older and less likely to have been vaccinated against hepatitis B (HBV) since universal infant vaccination in the United States began in 198219. Medicaid patients likewise may have decreased access to medical care, perinatal HBV screening, and HBV vaccination. Despite inclusion of these higher risk groups, we did not find an increased number of HBsAg positives in the alert group.
Of the 4141 total patients in the alert group (Cohort 1 and 2 combined), only 389 (9.4%) completed HBsAg testing so while the alert did increase the chance of HBsAg completion, the overall effect was small.
There were limitations to the CHB alert and the current study. The first limitation is that the alert is passive. That is, it lays dormant in the medical chart of the at-risk person until he or she visits their physician. Only after the physician opens the medical chart and discovers the alert, can the effect of the alert be manifest. However, even among patients that attended an office visit during the study period, a similar twofold increase in HBsAg testing without an increase in HBsAg positive detection was found. Multivariate analysis did show that both the alert and attendance of office visits were independently associated with completion of HBsAg testing. More active measures to engage and educate at-risk patients are needed to bring them to their physicians for screening. The second limitation to the CHB alert is that some API may not have been identified in the population. Our algorithm utilizes surnames that are associated with API ethnic groups but excludes surnames that are ambiguous for API ethnicity, for example “Lee”. Adopted APIs, APIs that have taken on a non-API married surname, or non-APIs that have taken on API surnames may also lead to misclassification. Third, we did not screen other risk groups such as African born, persons who inject drugs, and men who have sex with men. Fourth, since CMS did not initially cover HBsAg testing, we needed to use 2 Cohorts. This may have decreased the effectiveness of the alert in Cohort 2, due to contamination. That is, PCP awareness of hepatitis B testing may have increased during Cohort 1 and more tests could have been ordered regardless of alert status during Cohort 2.
In conclusion, EHR alerts increase completion of HBsAg testing, but do not increase detection of HBsAg positive cases even over years of follow up, but this may be explained by differences in insurance status. Screening for CHB requires a personalized approach since the decision to screen is based upon country of birth. While electronic alerts may have a role in increasing CHB screening, they will likely be ineffective alone. A multifaceted approach involving patient outreach and engagement to draw at-risk patients to CHB screening opportunities is needed.