Sixty percent of HIV care providers who responded to the MMP Provider Survey reported offering routine HIV screening to all patients 13 to 64 years of age. Nurse practitioners, providers aged <50 years, black providers, and providers with high HIV-infected patient loads were more likely to offer HIV testing to all patients according to the recommendations. Although there is limited information regarding the characteristics of providers who conduct routine HIV testing, there is evidence that physicians who had previously diagnosed an HIV-infected person were more likely to offer HIV testing to their patients 
. Our data extend this finding, by reporting that as the number of HIV-infected patients a provider cares for increases, the likelihood of providing routine testing increases.
No significant differences were found by characteristics of providers' HIV-infected patients. A higher percentage of providers (65% versus 55%) reported offering testing to all patients if their percent of white patients with HIV was less than 25% compared with ≥25%, but this was not significant in the multivariate analysis. However, others have found differences in routine offering of HIV testing by age, race and ethnicity of patients. Myers et al. found that blacks were more likely than whites to be offered testing, and Latinos and persons of other racial/ethnic groups (not Latino, white or black) were less likely to be offered testing 
. The same study found that patients aged less than 55 years, and most notably men aged less than 18 years, were less likely to be offered testing.
Since the release of the CDC recommendations, implementation and acceptance of routine HIV screening programs has been successful in a variety of settings; most reports have been from hospitals or emergency departments. In the first 8 months following the October 2006 implementation of a hospital-wide routine rapid HIV testing program at Howard University Hospital, 57% of 9,810 patients who were offered testing agreed to test 
. In a District of Columbia emergency department (ED) where trained HIV screeners offered rapid testing to 4,187 patients treated in the ED during a 3 month period, nearly 60% accepted testing 
. Routine HIV screening was offered to 954 individuals in three South Carolina community health clinics starting in December 2006, with reported acceptance rates of 62%, 56% and 47%, respectively, in the first 8 months 
. Other attempts at implementing routine testing have not been as successful. Routine testing was offered to 3,467 patients in a District of Columbia Veterans Affairs (VA) hospital from November 2007 through March 2009, but only 25% accepted 
. Similarly, a Denver emergency department offered routine opt-out testing to 28,043 patients from April 2007 through April 2008, with only 24% accepting testing 
. In other facilities and health care systems, routine HIV testing has yet to be implemented. A survey of veterans conducted from mid-October 2008 through mid-February 2009 found that HIV testing is not being routinely offered by VA providers. Of over 31,000 survey respondents, only 9% said they had been offered an HIV test in the past 12 months 
. Further, a 2009 online survey of MSM in the US found that only about half of HIV-negative MSM reported being offered an HIV test by their routine health care provider in the past year 
Although we did not ask providers their reasons for not conducting routine HIV testing, several barriers to implementing the recommendations have been documented 
and are likely similar to barriers experienced by our survey respondents. Barriers identified include: state and federal agency laws 
; providers' concern about lack of prevention counseling 
; stigma and discrimination associated with HIV 
; and the perception that conducting risk-based testing is more cost effective than routine testing 
CDC acknowledged in the recommendations that state statutes and clinic policies might pose barriers to fully implementing the recommendations 
. These barriers were found by Mahajan and colleagues when they examined whether implementing the CDC recommendations was compatible with individual state statutes during the two years following the release of the recommendations 
. They reported that 16 states had statutes that were inconsistent with the key features of the recommendations: 1) opt-out testing; 2) informed consent; and 3) lack of HIV prevention counseling, meaning that implementing one or more of the provisions of the recommendations could not occur without amending existing laws. In the two years following the release of the recommendations, nine of the 16 states passed laws that were consistent 
; in 2010, six states still had laws inconsistent with the recommendations 
. Massachusetts, Michigan, Nebraska, New York (with the exception of rapid testing) and Pennsylvania still required specific written consent for HIV testing, and Michigan, Pennsylvania and West Virginia still required post-test counseling for a negative or positive result. In our survey provider sample sizes were inadequate to assess state as a predictor to determine the impact of state and clinic policies on implementing the recommendations.
Barriers to testing among physicians include pre-test counseling requirements, lack of knowledge of testing recommendations and requirements and lack of training in conducting HIV testing, lack of time, lack of acceptance by patients, burden of the consent process, language barriers, competing priorities and inadequate compensation 
. Lack of training may be less of a barrier for younger providers, who have likely received more training regarding HIV testing, diagnosis and treatment than their older counterparts. This is supported by our finding that younger providers were more likely to offer testing. Lack of knowledge of the recommendations was identified by a survey of internal medicine residents in New York City conducted in early 2007 between five and nine months after the release of the recommendations. Only 32% of those surveyed were aware of the recommendations, and most were not offering routine testing; 36% of the residents used a routine testing approach, while 64% reported utilizing risk-based testing 
. Sixty-eight percent said they would order more HIV tests if consent were oral rather than written and 46% had consent issues (written consent was required by New York State law at the time), 41% reported lack of time, and 20% cited language barriers. Factors associated with ordering 10 or fewer tests included lack of pre-test counseling training, conducting risk-based testing, and taking sexual history never or occasionally. In addition to the barriers above, lack of reimbursement for the test and lack of capacity to provide services or medications for patients with newly diagnosed HIV may also be reasons that testing is not offered.
If recommendations for HIV screening are routinely implemented, high percentages of patients offered testing would accept. Although lack of patient acceptance was noted by physicians 
, the literature shows that when more patients are offered HIV screening, more patients are tested. Six health centers that provide primary care and prevention services to underserved populations in North Carolina, South Carolina and Mississippi conducted almost 3 times the number of HIV tests after implementing routine testing compared with the previous year 
. Of the 9% of VA patients mentioned above who said they had been offered an HIV test in the past 12 months, 91% accepted 
. A survey to assess patients' acceptance of opt-out HIV testing in an urban emergency department asked 529 patients if they would accept opt-out HIV testing; 81% reported they would have accepted 
. In the VA survey mentioned above, 73% of respondents reported they would be “very likely” to accept HIV testing, if recommended by their doctor 
, and focus groups at VA facilities found that both patients and providers agreed that routine testing would be beneficial to public health and to patients 
Although cost may be perceived to be a barrier to offering routine testing, several recent studies have found routine opt-out testing to be cost-effective 
. Holtgrave found that risk-based testing would result in more HIV diagnoses and prevent more HIV infections at a lower cost compared with routine opt-out testing, but in turn would increase the cost burden on the health care system to provide medical care to these newly identified persons 
Overall, HIV testing has increased in the US since the guidelines were introduced, but awareness of the guidelines, the importance of routine HIV testing, and current testing coverage still needs to be conveyed to providers of HIV care and other medical care providers in the US. The baseline for evaluating the effects of CDC's recommendations was developed from the National Health Interview Survey (NHIS). This survey determined that in 2006, an estimated 40.4% (71.5 million) of adults aged 18–64 years in the US reported ever receiving HIV testing 
. Also using NHIS data, CDC reported that among persons aged 18–64 years in 2009, 45.0% (82.9 million) reported ever receiving an HIV test 
. While this increase is encouraging, it still means that by 2009, 55% of persons aged 18–64 years had never received an HIV test. These numbers will need to increase at a faster rate to meet the National HIV/AIDS Strategy goal of increasing the percentage of persons with HIV who are aware of their status to 90% by 2015 
. In addition to the increased resources needed to implement routine HIV testing, the identification of new HIV cases will require linkages to HIV care and treatment resulting in in further demand on our nation's health care system. However, these additional resources required in the short-term will result in savings in costs and lives in the long-term.
There are several limitations to our study. Cost constraints prohibited selecting all 2,550 individual providers participating in the 2007 MMP data collection cycle. Our response rate of 42% was low; however, our sample included mostly physicians, who have lower survey response rates compared to non-physicians 
are based on providers' self-reported responses to survey questions. Given that these are providers of HIV care with the majority caring for patients >10 years, who consider themselves experts in and knowledgeable of HIV treatment, we expected high rates of screening to be offered to their HIV-negative patients. Providers' estimates of the number of HIV-infected patients they provide care to and the racial, ethnic and behavioral characteristics of their HIV-infected patient populations were likely self-determined and not derived using clinic records. We have previously experienced errors in estimates of patient loads during the construction of facility sampling frames in MMP 
. Providers sampled and who responded to the survey may not be representative of HIV care providers in the US, and therefore, results may not be generalizable. We recommend that provider surveys that are conducted in the future as part of MMP include all sampled providers and that analysis weights are incorporated to adjust for selection probabilities and nonresponse.