This assessment shows that the strategies employed during the STRETCH trial resulted in significant increases in total integration scores at intervention clinics. The specific areas in which the integration score increased were in providing HIV care in primary care clinics not previously involved in the ART programme (mainstreaming HIV score) and in the provision of elements of ART care, namely, the taking of baseline blood tests, drug readiness training and monthly supply of ART for patients eligible for ART (ART score). These findings have been independently confirmed by a qualitative process evaluation of the STRETCH trial which found that patients and nurses appreciated the convenience of patients being able to access HIV care including ARVs at their local clinic instead of having to travel to an ART clinic 
. There was no increase in mean pre-ART integration scores during the trial because these elements of HIV care, namely VCT, initial CD4 count and routine care for those not yet eligible for ART which had been identified by staff as critical elements of pre-ART needing integration, had already been substantially integrated into primary care services by local managers in the months leading up to the trial.
In contrast it appears that the strategies used in the STRETCH trial had no effect on internal integration scores at intervention clinics, with no significant shift towards patients being able to access HIV care from all nurses within the clinic. There may be other more effective strategies to achieve integration of HIV care into the work of all nurses within primary care clinics, or there may be factors that mitigate against internal integration. Topp et al described some strategies to integrate the provision of HIV care into the work of all nurses within two primary care clinics in Zambia 
. These strategies included training of all staff in HIV care, as in the STRETCH trial, but also the use of other strategies not used in the STRETCH trial – combined medical records and waiting areas and the inclusion of HIV testing into triage of all patients. They did document increased uptake of HIV testing and good standards of HIV care. However, they also reported resistance on the part of nurses and patients to completely integrated ART services because of issues such as increased waiting times and the loss of informal support for patients on ART with the loss of ART waiting areas 
. A synthesis of the findings of three qualitative studies on internal integration in Free State clinics conducted at the same time as the STRETCH trial found that administrative issues and patient and nurse preferences tended to mitigate against internal integration of HIV care (manuscript submitted for publication).
The increase in mean total integration, ARV and mainstreaming HIV scores by the fourth assessment late in the trial at control clinics, resulted from provincial implementation of a new national AIDS policy including nurse prescription of ART and the provision of ART in all primary care clinics – the two main interventions of the trial 
. The STRETCH trial was a pragmatic trial conducted under real conditions which include such policy changes. The research team was able to negotiate with the province that nurse initiation of ART would not be implemented in control clinics till after the trial, but was not able to delay integration of HIV care into primary care services in control clinics in the last few months of the trial.
One of the strengths of this study is that it is a prospective assessment using a new semi- quantitative tool to document integration of HIV care. The contents of this questionnaire were likely to be valid as the elements of HIV care and the need to integrate them at both levels were identified in consultation with staff at ART clinics. Internal consistency as shown by Cronbach's alpha was good. Real validity of the questionnaire was demonstrated in that it captured an increase in integration scores at control clinics as a result of the implementation of a new policy to integrate HIV care into primary care in the last months of the trial.
There were some potential limitations to this study. The first two interviews were conducted during clinic visits while the last two were conducted telephonically, interviews were not always conducted with the same staff member and data on services at referring primary care clinics were based on reports from staff at the ART site and not at the primary care clinic. However, all interviews were conducted by the trial coordinator who was well known to the clinic staff and involved in local management teams implementing integration and thus was able to independently confirm progress of integration as described by the interviewee at each clinic. Though there is a possibility that the coordinator may have influenced answers, the results of inter-observer reliability tests suggest that this was negligible. The lack of progress in internal integration compared with the progress in mainstreaming HIV, captured by the questionnaire, suggests that the interviewees were not unduly influenced to report integration where there was none. The integration questionnaire was developed to assess the integration of HIV care into primary care as it affected service delivery for patients and was therefore not able to assess the effects of integration of other areas of health system functioning. The questionnaire was not able to document the impact of integration of HIV care on the provision of other primary care services. This is an important area of research, and is the subject of a project currently being conducted in all primary care clinics in the Free State.
This questionnaire was validated in the specific context of the Free State and may need some further development, but it could be a valuable tool for assessing integration of HIV care into primary care clinics in other settings. The main results of the STRETCH trial showed that patient survival was not significantly different in intervention clinics compared with control clinics 
.The integration scores obtained in this study will be correlated with survival of patients with CD4 below 350 and not yet on ART, from the STRETCH trial to determine if integration of HIV care may have had an independent effect on patient survival. These results, together with the process evaluation and results of the STRETCH study, should be useful in identifying whether integration is an effective strategy to improve survival of HIV-positive patients in need of ART.