Provider-initiated HIV testing and counseling among adult general outpatients in two high-volume primary care clinics in Gauteng Province, South Africa resulted in a 2.85 fold increase in odds of HIV test acceptance as compared to provider referral to onsite VCT services in the same clinics. Patients' reported experiences of the two models were similar and positive, though significantly more patients reported that their providers answered their questions about HIV in the PITC model. The median age of study participants was 33 years, 63% were women, and 66% had never been married; thus clinic patients were representative of a population with high HIV prevalence in South Africa. Among study participants in both the PITC and VCT referral model, more than one in five among those tested was HIV positive. In both models, documented linkage to HIV care among those who tested positive was extremely low. Providers expressed appreciation of the value of PITC in answers to written questions and in discussions, indicating that it assisted them with patient care; however they tested only a small percentage of their patients.
Among those who declined testing with either model, almost one-third (31%) refused because they were uncomfortable or afraid of an HIV test and 19% reported not feeling the need to be tested. These reasons are similar to the published literature, and indicate that continuing widespread fear of HIV testing must be addressed 
. In both the VCT referral and PITC models, over half the participants reported that they did not feel they could decline the test. These findings are unclear as many of these same participants did in fact refuse testing. Nonetheless, guidance and ongoing supervision must be provided to health care providers implementing PITC to ensure patients can opt out of testing 
. Evidence from our study suggests that test uptake was associated with having had a previous HIV test and ever being forced or coerced into sex, but not with any other reported risk behaviors for HIV acquisition.
One of the strengths of our study is that during PITC model implementation, the providers themselves offered and performed PITC as part of the general outpatient visit. Increased acceptance of HIV testing by general outpatients offered PITC has been previously reported in Zambia and South Africa, but in both those studies lay counselors rather than clinicians offered and provided the HIV testing and counseling in the outpatient department 
. Similar rates of increased testing were seen in those studies as the 1.8 fold increase in ours: in Zambia, the addition of lay counselor-conducted PITC to referral to VCT doubled the number tested for HIV in 9 primary care clinics compared with referral to VCT alone. In Durban, South Africa, acceptance of testing was 1.5 times higher with PITC conducted by lay counselors compared with referral to VCT by clinicians. Clinicians themselves performing routine HIV testing, and associated increased testing uptake, has been reported in antenatal, TB, and STI clinics in southern Africa 
, but to our knowledge this is the first report of clinicians implementing PITC in general outpatient clinics with very high daily patient volumes.
Another strength is that we assessed provider and patient attitudes and perceptions of PITC during its implementation, and compared and contrasted these with HIV test acceptance results. Contradictory findings included that providers expressed appreciation for the value of PITC for improving patient care, but tested very few patients. Confirmatory findings included that patients who reported that it was possible to get a confidential HIV test in their community were more likely to accept testing. These findings from provider and patient surveys can inform program improvements.
Furthermore, we followed participants beyond uptake of HIV testing to determine the linkage of those who tested positive to HIV care and treatment services. Many studies have reported an increase in HIV test acceptance with PITC; few have documented whether the HIV-infected persons identified benefited from their known status by accessing HIV clinical services 
. We found only four percent of HIV-positive patients had a registered visit to the onsite HIV treatment clinic three months after their test result. A priority area for further research is to investigate the reasons for this lack of follow-up.
There were several limitations to this study. First, the study design, a pre-intervention/post-intervention evaluation, lacks the rigor of a randomized controlled trial. The two health centers were typical of health centers in South Africa, but may not be representative of other types of health facilities. There was a decline in the rate of participation and follow up interviews during the PITC data collection period, which was observed in one of the two clinics. This difference was likely due to a drop in staffing at that health center during PITC implementation, so that many patients left without being seen by a provider, including enrolled participants who had completed baseline questionnaires. It is unlikely that this affected HIV test acceptance at the clinic. The use of self-reported data from participants carries the inherent possibility of social desirability bias. However, it seems unlikely there would be differential reporting between the participants in the two models of testing. Furthermore, self-reported HIV status has been shown to have similar validity to other self-reported variables 
. Lastly, the follow-up of patients at the HIV treatment clinic at the CHC where they were tested may not be an indication of an individual accessing care. Patients may have chosen to go to another HIV treatment clinic for reasons of convenience or perceived confidentiality.
Several programmatic recommendations follow from our study results. First, regarding the low rate of testing by providers during the PITC model implementation. Provider performance even in high-volume clinics can be influenced by strong leadership from all administrative levels of the health system to create a sense of professional responsibility for improving patients' knowledge of their HIV status. Furthermore, in settings such as South Africa where overall one in every five adults is HIV-infected, determining HIV status should be considered a necessary part of a differential diagnosis for any acute medical conditions. Since this study was completed, the South African Minister of Health has endorsed PITC, which should lead to changed expectations of providers' performance 
Under current staffing conditions, it will be very difficult to achieve universal HIV testing through PITC in South African community health centers. To do so, for a CHC serving 400 patients a day with 12 providers offering PITC (the averages from our study), each provider would need to test 33 patients per day on average (results not shown). If however, HIV testing was recommended once per year for those first testing negative, that number could fall to 11, as patients reported visiting the same health center a median of three times per year. Under these conditions, encouraging providers to test 6 patients per day on average would ensure that roughly 50% of outpatients would be offered an HIV test in a given year.
A second essential area for long-term prevention programming in addition to increasing testing rates, is determining the barriers to successfully linking patients who test HIV positive to treatment services, and implementing interventions to overcome these barriers at the structural and individual level. For example, Gauteng Province is instituting a patient locater system, which will include all government HIV care programs, so that patient access to care can be tracked across facilities. Determining the effectiveness of this system in improving retention will be key.
Finally, using a parallel rather than a serial HIV rapid testing algorithm would reduce the time necessary for processing HIV tests, and improve the efficiency of both models of HIV testing. Recent legislation in South Africa has for the first time allowed lay counselors to conduct HIV rapid testing, which will streamline the VCT referral model.
In conclusion, PITC increased the uptake of HIV testing compared with referral to onsite VCT in two government-operated, free of charge, community health centers in South Africa, and patients reported a positive response to PITC. The proportion of patients who were tested was low in both models of HIV testing, a concern in a country with high prevalence of HIV infection; among those tested, the proportion of patients who tested HIV positive was high. PITC allowed health care providers to identify many HIV infected general outpatients, but some key challenges should be addressed as it is scaled up to complement existing VCT services. Health facilities implementing PITC in the future will benefit from regional and facility-level PITC implementation plans including the development of training schedules, optimization of clinic flow and floor plans to ensure patient confidentiality, and administrative support to supervise and motivate health care providers. Finally, strengthening referral systems within and between health facilities to ensure that patients are effectively linked to treatment and prevention services will be vital to ensuring successful patient and programmatic outcomes.